changelog

pandoc (2.1)

  * Filter changes (#4196).  Previously we ran all lua filters before
    JSON filters.  Now we run filters in the order they are presented on
    the command line, whether lua or JSON.  There are two incompatible API
    changes: The type of `applyFilters` has changed, and `applyLuaFilters`
    has been removed.

  * Text.Pandoc.Class: make `FileTree` opaque (don't export
    `FileTree` constructor).  This forces users to interact with it using
    `insertInFileTree` and `getFileInfo`, which normalize file names.

pandoc (2.0.6)

  * Added `jats` as an input format.

    + Add Text.Pandoc.Readers.JATS, exporting `readJATS` (API
      change) (Hamish Mackenzie).
    + Improved citation handling in JATS reader.  JATS citations
      are now converted to pandoc citations, and JATS ref-lists
      are converted into a `references` field in metadata, suitable
      for use with pandoc-citeproc. Thus a JATS article with embedded
      bibliographic information can be processed with pandoc and
      pandoc-citeproc to produce a formatted bibliography.

  * Allow `--list-extensions` to take an optional FORMAT argument.
    This lists the extensions set by default for the selected FORMAT.
    The extensions are now alphabetized, and the `+` or `-`
    indicating the default setting comes before, rather than after,
    the extension.

  * Markdown reader:

    + Preserve original whitespace between blocks.
    + Recognize `\placeformula` as context.
    + Be pickier about table captions.  A caption starts with a `:` which
      can't be followed by punctuation.  Otherwise we can falsely interpret
      the start of a fenced div, or even a table header line like
      `:--:|:--:`, as a caption.
    + Always use four space rule for example lists.  It would be awkward
      to indent example list contents to the first non-space character after
      the label, since example list labels are often long.  Thanks to
      Bernhard Fisseni for the suggestion.
    + Improve raw tex parsing.  Note that the Markdown reader is also
      affected by the `latex_macros` extension changes described below
      under the LaTeX reader.

  * LaTeX reader:

    + `latex_macros` extension changes (#4179).  Don't pass through macro
      definitions themselves when `latex_macros` is set.  The macros
      have already been applied.  If `latex_macros` is enabled, then
      `rawLaTeXBlock` in Text.Pandoc.Readers.LaTeX will succeed in parsing
      a macro definition, and will update pandoc's internal macro map
      accordingly, but the empty string will be returned.
    + Export `tokenize`, `untokenize` (API change).
    + Use `applyMacros` in `rawLaTeXBlock`, `rawLaTeXInline`.
    + Refactored `inlineCommand`.
    + Fix bug in tokenizer.  Material following `^^` was
      dropped if it wasn't a character escape.  This only affected
      invalid LaTeX, so we didn't see it in the wild, but it appeared
      in a QuickCheck test failure.
    + Fix regression in LateX tokenization (#4159).  This mainly affects the
      Markdown reader when parsing raw LaTeX with escaped spaces.
    + Add tests of LaTeX tokenizer.
    + Support `\foreignlanguage` from babel.
    + Be more tolerant of `&` character (#4208).  This allows us to parse
      unknown tabular environments as raw LaTeX.

  * Muse reader (Alexander Krotov):

    + Parse anchors immediately after headings as IDs.
    + Require that note references does not start with 0.
    + Parse empty comments correctly.

  * Org reader (Albert Krewinkel):

    + Fix asterisks-related parsing error (#4180).
    + Support minlevel option for includes (#4154).  The level of headers
      in included files can be shifted to a higher level by specifying a
      minimum header level via the `:minlevel` parameter. E.g.
      `#+include: "tour.org" :minlevel 1` will shift the headers in
      tour.org such that the topmost headers become level 1 headers.
    + Break-up org reader test file into multiple modules.

  * OPML reader:

    + Enable raw HTML and other extensions by default for notes
      (#4164).  This fixes a regression in 2.0.  Note that extensions can
      now be individually disabled, e.g.  `-f opml-smart-raw_html`.

  * RST reader:

    + Allow empty list items (#4193).
    + More accurate parsing of references (#4156).  Previously we erroneously
      included the enclosing backticks in a reference ID (#4156).  This
      change also disables interpretation of syntax inside references, as
      in docutils.  So, there is no emphasis in `` `my *link*`_ ``.

  * Docx reader:

    + Continue lists after interruption (#4025, Jesse Rosenthal).
      Docx expects that lists will continue where they left off after an
      interruption and introduces a new id if a list is starting again. So
      we keep track of the state of lists and use them to define a "start"
      attribute, if necessary.
    + Add tests for structured document tags unwrapping (Jesse Rosenthal).
    + Preprocess Document body to unwrap `w:sdt` elements (Jesse Rosenthal,
      #4190).

  * Plain writer:

    + Don't linkify table of contents.

  * RST writer:

    + Fix anchors for headers (#4188).  We were missing an `_`.

  * PowerPoint writer (Jesse Rosenthal):

    + Treat lists inside BlockQuotes as lists.  We don't yet produce
      incremental lists in PowerPoint, but we should at least treat lists
      inside BlockQuotes as lists, for compatibility with other slide formats.
    + Add ability to force size.  This replaces the more specific
      `blockQuote runProp`, which only affected the size of blockquotes. We
      can use this for notes, etc.
    + Implement notes.  This currently prints all notes on a final slide.
      Note that at the moment, there is a danger of text overflowing the
      note slide, since there is no logic for adding further slides.
    + Implement basic definition list functionality to PowerPoint writer.
    + Don't look for default template file for Powerpoint (#4181).
    + Add pptx to isTextFormat list.  This is used to check standalone
      and not writing to the terminal.
    + Obey slide level option (Jesse Rosenthal).
    + Introduce tests.

  * Docx writer:

    + Ensure that `distArchive` is the one that comes with pandoc
      (#4182).  Previously a `reference.docx` in `~/.pandoc` (or the user data
      dir) would be used instead, and this could cause problems because a
      user-modified docx sometimes lacks vital sections that we count
      on the `distArchive` to supply.

  * Org writer:

    + Do not wrap "-" to avoid accidental bullet lists (Alexander Krotov).
    + Don't allow fn refs to wrap to beginning of line (#4171, with help from
      Alexander Krotov).  Otherwise they can be interpreted as footnote
      definitions.

  * Muse writer (Alexander Krotov):

    + Don't wrap note references to the next line (#4172).

  * HTML writer:

    + Use br elements in line blocks instead of relying on CSS
      (#4162).  HTML-based templates have had the custom CSS for
      `div.line-block` removed.  Those maintaining custom templates will want
      to remove this too.  We still enclose line blocks in a div with class
      `line-block`.

  * LaTeX writer:

    + Use `\renewcommand` for `\textlatin` with babel (#4161).
      This avoids a clash with a deprecated `\textlatin` command defined
      in Babel.
    + Allow fragile=singleslide attribute in beamer slides (#4169).
    + Use `\endhead` after `\toprule` in headerless tables (#4207).

  * FB2 writer:

    + Add cover image specified by `cover-image` meta (Alexander Krotov,
      #4195).

  * JATS writer (Hamish Mackenzie):

    + Support writing `<fig>` and `<table-wrap>` elements
      with `<title>` and `<caption>` inside them by using Divs with class set
      to one of `fig`, `table-wrap` or `caption` (Hamish Mackenzie).  The
      title is included as a Heading so the constraint on where Heading can
      occur is also relaxed.
    + Leave out empty alt attributes on links.
    + Deduplicate image mime type code.
    + Make `<p>` optional in `<td>` and `<th>` (#4178).
    + Self closing tags for empty xref (#4187).
    + Improve support for code language.

  * Custom writer:

    + Use init file to setup Lua interpreter (Albert Krewinkel).
      The same init file (`data/init`) that is used to setup the Lua
      interpreter for Lua filters is also used to setup the interpreter of
      custom writers.lua.
    + Define instances for newtype wrapper (Albert Krewinkel).  The custom
      writer used its own `ToLuaStack` instance definitions, which made
      it difficult to share code with Lua filters, as this could result
      in conflicting instances.  A `Stringify` wrapper is introduced to
      avoid this problem.
    + Added tests for custom writer.
    + Fixed definition lists and tables in `data/sample.lua`.

  * Fixed regression: when target is PDF, writer extensions were being
    ignored.  So, for example, `pandoc -t latex-smart -o file.pdf`
    did not work properly.

  * Lua modules (Albert Krewinkel):

    + Add `pandoc.utils` module, to hold utility functions.
    + Create a Haskell module Text.Pandoc.Lua.Module.Pandoc to
      define the `pandoc` lua module.
    + Make a Haskell module for each Lua module. Move definitions for the
      `pandoc.mediabag` modules to a separate Haskell module.
    + Move `sha1` from the main `pandoc` module to `pandoc.utils`.
    + Add function `pandoc.utils.hierarchicalize` (convert list of
      Pandoc blocks into (hierarchical) list of Elements).
    + Add function `pandoc.utils.normalize_date` (parses a date and
      converts it (if possible) to "YYYY-MM-DD" format).
    + Add function `pandoc.utils.to_roman_numeral` (allows conversion
      of numbers below 4000 into roman numerals).
    + Add function `pandoc.utils.stringify` (converts any AST element
      to a string with formatting removed).
    + `data/init.lua`: load `pandoc.utils` by default
    + Turn pipe, read into full Haskell functions.  The `pipe` and `read`
      utility functions are converted from hybrid lua/haskell functions
      into full Haskell functions. This avoids the need for intermediate
      `_pipe`/`_read` helper functions, which have dropped.
    + pandoc.lua: re-add missing MetaMap function.  This was a bug
      introduced in version 2.0.4.

  * Text.Pandoc.Class: Add `insertInFileTree` [API change].  This gives
    a pure way to insert an ersatz file into a `FileTree`.  In addition, we
    normalize paths both on insertion and on lookup.

  * Text.Pandoc.Shared: export `blocksToInlines'` (API change, Maura Bieg).

  * Text.Pandoc.MIME: Add opus to MIME type table as audio/ogg (#4198).

  * Text.Pandoc.Extensions:   Alphabetical order constructors for
    `Extension`.  This makes them appear in order in `--list-extensions`.

  * Allow lenient decoding of latex error logs, which are not always
    properly UTF8-encoded (#4200).

  * Update latex template to work with recent versions of beamer.
    The old template produced numbered sections with some recent
    versions of beamer.  Thanks to Thomas Hodgson.

  * Updated reference.docx (#4175).  Instead of just "Hello, world", the
    document now contains exemplars of most of the styles that have an
    effect on pandoc documents.  This makes it easier to see the effect
    of style changes.

  * Removed `default.theme` data file (#4096).  It is no longer needed now
    that we have `--print-highlight-style`.

  * Added `stack.lts9.yaml` for building with lts 9 and ghc 8.0.2.
    We still need this for the alpine static linux build, since
    we don't have ghc 8.2.2 for that yet.

  * Removed `stack.pkg.yaml`.  We only really need `stack.yaml`; we
    can put flag settings for pandoc-citeproc there.

  * Makefile: Add 'trypandoc' and 'pandoc-templates' targets to
    make releases easier.

  * MANUAL.txt:

    + Add note on what formats have `+smart` by default.
    + Use native syntax for custom-style (#4174, Mauro Bieg).
    + Introduce dedicated Extensions section, since some extensions
      affect formats other than markdown (Mauro Bieg, #4204).
    + Clarify default html output for `--section-divs` (Richard Edwards).

  * filters.md: say that Text.Pandoc.JSON comes form pandoc-types.
    Closes jgm/pandoc-website#16.

  * epub.md: Delete removed `-S` option from command (#4151, Georger Araújo).

pandoc (2.0.5)

  * Fix a bug in 2.0.4, whereby pandoc could not read the theme files
    generated with `--print-highlight-style` (#4133).  Improve JSON
    serialization of styles.

  * Fix CSS issues involving line numbers (#4128).
    Highlighted code blocks are now enclosed in a div with class `sourceCode`.
    Highlighting CSS no longer sets a generic color for pre and code; we only
    set these for class `sourceCode`.

  * `--pdf-engine-opt`: fix bug where option order was reversed (#4137).

  * Add PowerPoint (pptx) writer (Jesse Rosenthal).
    It works following the standard Pandoc conventions for making other
    sorts of slides. Caveats:

    + Syntax highlighting is not yet implemented. (This is difficult
      because there are no character classes in Powerpoint.)
    + Footnotes and Definition lists are not yet implemented. (Notes will
      usually take the form of a final slide.
    + Image placement and auto-resizing has a few glitches.
    + Reference powerpoint files don't work dependably from the command
      line. This will be implemented, but at the moment users are advised
      to change themes from within Powerpoint.

  * Create shared Text.Pandoc.Writers.OOXML module (Jesse Rosenthal).
    This is for functions used by both Powerpoint and Docx writers.

  * Add default pptx data for Powerpoint writer (Jesse Rosenthal).

  * Add `empty_paragraphs` extension.

    + Deprecate `--strip-empty-paragraphs` option.  Instead we now
      use an `empty_paragraphs` extension that can be enabled on
      the reader or writer.  By default, disabled.
    + Add `Ext_empty_paragraphs` constructor to `Extension`.
    + Revert "Docx reader: don't strip out empty paragraphs."
      This reverts commit d6c58eb836f033a48955796de4d9ffb3b30e297b.
    + Implement `empty_paragraphs` extension in docx reader and writer,
      opendocument writer, HTML reader and writer.
    + Add tests for `empty_paragraphs` extension.

  * Markdown reader:

    + Don't parse native div as table caption (#4119).
    + Improved computation of column widths in pipe tables.
      Pipe tables with lines longer than the text width (as set
      by `--columns`) are now scaled to text width, with the relative
      widths of columns determined by the ratios between the
      header lines.  Previously we computed column widths using
      the ratio of header line lengths to column width, so that
      tables with narrow header lines were extremely thin, which
      was very rarely the desired result.

  * LaTeX reader: fix `\` before newline (#4134).  This should be a space,
    as long as it's not followed by a blank line. This has been fixed at the
    tokenizer level.

  * Muse reader (Alexander Krotov):

    + Add test for `#disable-tables` directive in Emacs mode.
    + Don't allow emphasis to be preceded by letter.
    + Add underline support in Emacs Muse mode..
    + Support multiline directives in Amusewiki mode

  * Man writer: omit internal links (#4136).  That is, just print the link
    text without the URL.

  * Markdown reader: accept processing instructions as raw HTML (#4125).

  * Lua filters (Albert Krewinkel):

    + Use script to initialize the interpreter.  The file `init.lua` is
      used to initialize the Lua interpreter which is used in Lua filters.
      This gives users the option to require libraries which they want to
      use in all of their filters, and to extend default modules.
    + Fix package loading for Lua 5.1.  The list of package searchers is
      named `package.loaders` in Lua 5.1 and LuaJIT, and `package.searchers`
      in Lua 5.2 and later.
    + Refactor lua module handling.  The integration with Lua's package/module
      system is improved: A pandoc-specific package searcher is prepended to
      the searchers in `package.searchers`. The modules `pandoc` and
      `pandoc.mediabag` can now be loaded via `require`.
    + Bump lower bound of hslua.  The release hslua 0.9.3 contains a new
      function which makes using Haskell functions as package loaders much
      easier.

  * reveal.js template:  add title-slide identifier to title slide (#4120).
    This allows it to be styled more easily.

  * LaTeX template: Added support for `pagestyle` variable (#4135,
    Thomas Hodgson)

  * Add `-threaded` to ghc-options for executable (#4130, fixes a build
    error on linux).


pandoc (2.0.4)

  * Add `--print-highlight-style` option.  This generates a JSON version
    of a highlighting style, which can be saved as a `.theme` file, modified,
    and used with `--highlight-style` (#4106, #4096).

  * Add `--strip-empty-paragraphs` option.  This works for any input format.
    It is primarily intended for use with docx and odt documents where
    empty paragraphs have been used for inter-paragraph spaces.

  * Support `--webtex` for `gfm` output.

  * Recognize `.muse` file extension.

  * Support beamer `\alert` in LaTeX reader. Closes #4091.

  * Docx reader: don't strip out empty paragraphs (#2252).
    Users who have a conversion pipeline from docx may want to consider adding
    `--strip-empty-paragraphs` to the command line.

  * Org reader (Albert Krewinkel): Allow empty list items (#4090).

  * Muse reader (Alexander Krotov):

    + Parse markup in definition list terms.
    + Allow definition to end with EOF.
    + Make code blocks round trip.
    + Drop common space prefix from list items.
    + Add partial round trip test.
    + Don't interpret XML entities.
    + Remove `nested`.
    + Parse `~~` as non-breaking space in Emacs mode.
    + Correctly remove indentation from notes.  Exactly one space is
      required and considered to be part of the marker.
    + Allow list items to be empty.
    + Add ordered list test.
    + Add more multiline definition tests.
    + Don't allow blockquotes within lists.
    + Fix reading of multiline definitions.
    + Add inline `<literal>` support.
    + Concatenate inlines of the same type

  * Docx writer: allow empty paragraphs (#2252).

  * CommonMark/gfm writer:

    + Use raw html for native divs/spans (#4113).  This allows a pandoc
      markdown native div or span to be rendered in gfm using raw html tags.
    + Implement `raw_html` and `raw_tex` extensions.  Note that `raw_html`
      is enabled by default for `gfm`, while `raw_tex` is disabled by default.

  * Muse writer (Alexander Krotov):

    + Test that inline math conversion result is normalized.
      Without normalization this test produced
      `<em>a</em><em>b</em><em>c</em>`.
    + Improve inline list normalization and move to writer.
    + Escape hash symbol.
    + Escape `----` to avoid accidental horizontal rules.
    + Escape only `</code>` inside code tag.
    + Additional `<verbatim>` is not needed as `<code>` is verbatim already.

  * LaTeX writer:

    + Allow specifying just width or height for image size.
      Previously both needed to be specified (unless the image was
      being resized to be smaller than its original size).
      If height but not width is specified, we now set width to
      textwidth. If width but not height is specified, we now set
      height to textheight.  Since we have `keepaspectratio`, this
      yields the desired result.
    + Escape `~` and `_` in code with `--listings` (#4111).

  * HTML writer: export `tagWithAttributes`.  This is a helper allowing
    other writers to create single HTML tags.

  * Let papersizes `a0`, `a1`, `a2`, ... be case-insensitive by
    converting the case as needed in LaTeX and ConTeXt writers.

  * Change `fixDisplayMath` from `Text.Pandoc.Writers.Shared`
    so that it no longer produces empty `Para`'s as an artifact.

  * `Text.Pandoc.Shared.blocksToInlines`:  rewrote using builder.
    This gives us automatic normalization, so we don't get
    for example two consecutive Spaces.

  * Include default CSS for 'underline' class in HTML-based templates.

  * revealjs template:  add `tex2jax` configuration for the
    math plugin.  With the next release of reveal.js, this will
    fix the problem of `$`s outside of math contexts being
    interpreted as math delimiters (#4027).

  * `pandoc.lua` module for use in lua filters (Albert Krewinkel):

    + Add basic lua List module (#4099, #4081).  The List module is
      automatically loaded, but not assigned to a global variable. It can be
      included in filters by calling `List = require 'List'`.  Lists of blocks,
      lists of inlines, and lists of classes are now given `List` as a metatable,
      making working with them more convenient.  E.g., it is now possible to
      concatenate lists of inlines using Lua's concatenation operator `..`
      (requires at least one of the operants to have `List` as a metatable):

          function Emph (emph)
            local s = {pandoc.Space(), pandoc.Str 'emphasized'}
            return pandoc.Span(emph.content .. s)
          end

      The `List` metatable is assigned to the tables which get passed to
      the constructors `MetaBlocks`, `MetaInline`, and `MetaList`. This
      enables the use of the resulting objects as lists.
    + `Lua/StackInstances`: push Pandoc and Meta via constructor.
      Pandoc and Meta elements are now pushed by calling the respective
      constructor functions of the pandoc Lua module. This makes serialization
      consistent with the way blocks and inlines are pushed to lua and allows
      to use List methods with the `blocks` value.
    + Add documentation for pandoc.List in `lua-filters.md`.

  * Use latest tagsoup.  This fixes a bug in parsing HTML tags with
    `&` (but not a valid entity) following them (#4094, #4088).

  * Use skylighting 0.4.4.1, fixing the color of unmarked code text
    when `numberLines` is used (#4103).

  * Make `normalizeDate` more forgiving (Mauro Bieg, #4101), not
    requiring a leading 0 on single-digit days.

  * Fix `--help` output for `--highlight-style` to include `FILE` (Mauro
    Bieg, #4095).

  * Clearer deprecation warning for `--latexmathml, --asciimathml, -m`.
    Previously we only mentioned `--latexmathml`, even if `-m` was
    used.

  * Changelog: fix description of lua filters in 2.0 release
    (Albert Krewinkel).  Lua filters were initially run *after* conventional
    (JSON) filters.  However, this was changed later to make it easier to deal
    with files in the mediabag. The changelog is updated to describe that
    feature of the 2.0 release correctly.

  * Change Generic JSON instances to TemplateHaskell (Jasper Van der Jeugt,
    #4085).  This reduces compile time and memory usage significantly.

  * `lua-filters.md`: Added tikz filter example.

  * Create alternative zip file for macOS binaries.

  * Create alternative zip file for Windows binaries.

  * Update INSTALL.md since we now provide zips for binaries.

  * Relax `http-types` dependency (Justus Sagemüller, #4084).

  * Add `epub.md`, `getting-started.md` to docs.  These used to live in
    the website repo.

  * Add `packages` target to Makefile.

  * Bump bounds for binary, http-types, tasty-hunit

pandoc (2.0.3)

  * Lua filters: preload text module (Albert Krewinkel, #4077).
    The `text` module is preloaded in lua. The module contains some UTF-8
    aware string functions, implemented in Haskell.  The module is loaded on
    request only, e.g.:

        text = require 'text'
        function Str (s)
          s.text = text.upper(s.text)
          return s
        end

  * Allow table-like access to attributes in lua filters (Albert Krewinkel,
    #4071).  Attribute lists are represented as associative lists in Lua. Pure
    associative lists are awkward to work with. A metatable is attached to
    attribute lists, allowing to access and use the associative list as if
    the attributes were stored in as normal key-value pair in table.
    Note that this changes the way `pairs` works on attribute lists. Instead
    of producing integer keys and two-element tables, the resulting iterator
    function now returns the key and value of those pairs.  Use `ipairs` to
    get the old behavior.  Warning: the new iteration mechanism only works if
    pandoc has been compiled with Lua 5.2 or later (current default: 5.3).

  * Text.Pandoc.Parsing.uri:  allow `&` and `=` as word characters (#4068).
    This fixes a bug where pandoc would stop parsing a URI with an
    empty attribute:  for example, `&a=&b=` wolud stop at `a`.
    (The uri parser tries to guess which punctuation characters
    are part of the URI and which might be punctuation after it.)

  * Introduce `HasSyntaxExtensions` typeclass (Alexander Krotov, #4074).

    + Added new `HasSyntaxExtensions` typeclass for `ReaderOptions` and
      `WriterOptions`.
    + Reimplemented `isEnabled` function from `Options.hs` to accept both
      `ReaderOptions` and `WriterOptions`.
    + Replaced `enabled` from `CommonMark.hs` with new `isEnabled`.

  * Add `amuse` extension (Alexander Krotov) to enable Amuse wiki
    behavior for `muse`.  New `Ext_amuse` constructor for
    `Extension`. Note: this is switched on by default; for
    Emacs behavior, use `muse-amuse`.

  * Muse reader (Alexander Krotov):

    + Count only one space as part of list item marker.
    + Produce SoftBreaks on newlines. Now wrapping can be preserved
      with `--wrap=preserve`.
    + Add Text::Amuse footnote extensions.  Footnote end is indicated by
      indentation, so footnotes can be placed anywhere in the text,
      not just at the end of it.
    + Accept Emacs Muse definition lists when `-amuse`.
      Emacs Muse does not require indentation.

  * HTML reader:

    + Ensure we don't produce level 0 headers (#4076), even for chapter
      sections in epubs.  This causes problems because writers aren't set
      up to expect these.
    + Allow spaces after `\(` and before `\)` with `tex_math_single_backslash`.
      Previously `\( \frac{1}{a} < \frac{1}{b} \)` was not parsed as math in
      `markdown` or `html` `+tex_math_single_backslash`.
    + Parse div with class `line-block` as LineBlock.
    + Don't fail with block-level content in figcaption (Mauro Bieg, #4183).

  * MANUAL: clarify that math extensions work with HTML.
    Clarify that `tex_math_dollars` and `tex_math_single_backslash`
    will work with HTML as well as Markdown.

  * Creole reader: Fix performance issue for longer lists (Sascha Wilde,
    #4067).

  * RST reader: better support for 'container' directive (#4066).
    Create a div, incorporate name attribute and classes.

  * LaTeX reader:

    + Support column specs like `*{2}{r}` (#4056).  This is equivalent to
      `rr`.  We now expand it like a macro.
    + Allow optional args for parbox (#4056).
    + Allow optional arguments on `\footnote` (#4062).

  * EPUB writer: Fixed path for cover image (#4069).  It was previously
    `media/media/imagename`, and should have been `media/imagename`.

  * Markdown writer: fix bug with doubled footnotes in grid tables
    (#4061).

  * LaTeX template:  include natbib/biblatex after polyglossia (#4073).
    Otherwise we seem to get an error; biblatex wants polyglossia
    language to be defined.

  * Added examples to lua filters documentation.


pandoc (2.0.2)

  * Deprecated ancient HTML math methods: `--latexmathml`, `--gladtex`,
    `--mimetex`, `--jsmath`.

  * Fixed URIs in `data/jats.csl`.  They were being rendered twice,
    leading to invalid XML in default JATS output with pandoc-citeproc.

  * `lua-filters.md`: use real-world man page filter as example.

  * Add lua filter functions `walk_inline` and `walk_block`
    in the pandoc module, to apply filters inside particular
    inline and block elements.

  * Refactored some code from `Text.Pandoc.Lua.PandocModule`
    into new internal module `Text.Pandoc.Lua.Filter`.

  * Markdown reader:

    + Allow fenced code blocks to be indented 1-3 spaces (#4011).
      This brings our handling of them into alignment with CommonMark's.
    + Fix YAML metadata with "chomp" (`|-`).  Previously if a
      YAML block under `|-` contained a blank line, pandoc would
      not parse it as metadata.

  * Removed `etc.` from abbreviations file.  Often `etc.` ends a
    sentence, and we want the period to be treated as a
    sentence-ending period.

  * Fix regression with `--metadata` (#4054).  Values specified with
    `--metadata` should replace a metadata value set in the document
    itself, rather than creating a list including a new value.

  * EPUB writer:

    + Fix EPUB OCF structure.  #3720 had been improperly implemented.
    + Fix modified paths for raw HTML tags (src, poster, etc.)
      (#4050, #4055). This had not been updated for the new EPUB
      container layout, with a separate text/ subdirectory.
    + Fix image paths with empty `--epub-subdirectory`.

  * Miscellaneous code cleanup (Alexander Krotov).

  * Use pandoc-types 1.17.3, which adds `Walkable` instances
    for `[Block] Block` and `[Inline] Inline`.

  * Remove obsolete `stack.full.yaml` (#4052).

  * Change to using pandoc-citeproc 0.12.1 in binary packages.

  * Consolidate math output method documentation (#4049, Mauro Bieg).

  * `MANUAL.txt`: fix header level of "Extension: emoji" (Albert Krewinkel).

  * Use lua filter to generate man page from `MANUAL.txt`, replacing old
    Haskell filters.  This is easier and faster.

  * Improved `INSTALL.md`.

  * Update commands to extract deb archive on Linux (#4043, Salim B).


pandoc (2.0.1.1)

  * Improved fix to #3989 (parsing of HTML tags containing
    `>` in an attribute or comment). The previous fix (in 2.0.1) only
    worked in certain cases.

  * FB2 writer (Alexander Krotov):

    + Add `unrecognised` genre to `<title-info>`
      (Alexander Krotov).  XML schema requires at least one genre.
    + Remove `<annotation>` from `<body>`.

  * CommonMark writer: fix strikethrough for `gfm` (#4038).

  * Use texmath 0.10, which adds support for a wider range of
    symbols and fixes default column alignments in MathML
    and OMML.

  * Highlighting fixes, using skylighting 0.4.3.2:

    + Fix invalid CSS.
    + Support `lineAnchors` (or `line-anchors`) in HTML code blocks.
    + Ensure that code lines don't get duplicate identifiers (#4031).
      The line identifiers are built using the code block's identifier
      as a prefix. If the code block has null identifier, we use
      `cb1`, `cb2`, etc.

  * Added a few abbreviations to `data/abbreviations`,
    and sorted the list (#3984, Wandmalfarbe).

  * Improved support for columns in HTML writer (#4028).

    + Remove `width` attribute from the `div`.
    + Remove space between `<div class="column">` elements,
      since this prevents columns whose widths sum to 100%
      (the space takes up space).
    + Move as much as possible of the CSS to the template.
    + Ensure that all the HTML-based templates (including epub)
      contain the CSS for columns.
    + Columns default to 50% width unless they are given a width
      attribute.  So if you want two equal-width columns, you
      can use a div with class `column` and no `width` attribute.

  * SelfContained: use `base64` for css links with media attribute (#4026).
    This fixes `--self-contained` with S5.

  * Improve `pandoc-template-mode.el` (Vaclav Haisman).

  * Issue INFO, not WARNING, when a .sty file cannot be
    read in LaTeX reader.  It is normally not an issue requiring
    a fix from the user if .sty files are not found.

  * INSTALL.md: MacOS instructions needed xar -f (adam234).

  * MANUAL.txt:

    + Clarify that --setext-headers doesn't affect gfm output (#4035).
    + Clarify what is needed to open and close a div in `fenced_divs`
      (#4039, Tristano Ajmone).
    + Removed reference to `default.beamer` in docs (#4024).
      Also added mention of other templates affecting PDF output
      with different settings.

pandoc (2.0.1)

  * Fixed regression in parsing of HTML comments in markdown and other
    non-HTML formats (`Text.Pandoc.Readers.HTML.htmlTag`) (#4019).
    The parser stopped at the first `>` character, even if it wasn't
    the end of the comment.

  * Creole reader (Sascha Wilde):

    + Fix some minor typos and formatting.
    + Add additional test on nowiki-block after para.
    + Fix lists with trailing white space.

  * LaTeX reader: handle `%` comment right after command.
    For example, `\emph%`.

  * Markdown reader:  make sure fenced div closers work in lists.
    Previously the following failed:

        ::: {.class}
        1. one
        2. two
        :::

    and you needed a blank line before the closing `:::`.

  * Make `fenced_divs` affect the Markdown writer.  If `fenced_divs` is
    enabled, Divs will be rendered as fenced divs.

  * LaTeX/Beamer writer: support "blocks" inside columns and other Divs
    (#4016).

  * HTML Writer: consistently use dashed class-names (Mauro Bieg, #3556).
    Note: this change may require some changes in CSS rules.
    `footnoteRef` has become `footnote-ref`, `titleslide` has
    become `title-slide`, and `footnoteBack` has become `footnote-back`.

  * JATS writer: Properly pass through author metadata (#4020).

  * FB2 writer (Alexander Krotov):

    + Write blocks outside of `<p>` in definitions.
    + Make bullet lists consistent with ordered lists, repeating
      the marker for the outer list rather than indenting sublists,
      since indentation does not work in readers.
    + Add new style FB2 tests.

  * `Text.Pandoc.ImageSize`: Add `Millimeter` constructor to `Dimension`
    (#4012) [API change]. Now sizes given in 'mm' are no longer converted
    to 'cm'.

  * Revise documentation of small caps syntax (Andrew Dunning, #4013).

  * Fix broken reference links in manual (Andrew Dunning, #4014)

  * Fixed example of slide columns structure in changelog (#4015).
    Also documented this feature in MANUAL.txt.


pandoc (2.0.0.1)

  * EPUB writer:

    + Fixed filepaths for nonstandard epub-subdirectory values.
    + Ensure that epub2 is recognized as a non-text format,
      so that a template is used.
    + Don't include "prefix" attribute for ibooks for epub2.
      It doesn't validate.
    + Fix stylesheet paths; previously we had an incorrect
      stylesheet path for the cover page and nav page.

  * LaTeX reader:

    + Insert space when needed in macro expansion (#4007).
      Sometimes we need to insert a space after a control sequence
      to prevent it merging with a following letter.
    + Allow unbraced arguments for macros (#4007).
    + Allow body of macro definition to be unbraced (#4007).

  * Linux package build: ensure that pandoc-citeproc is statically linked.

  * trypandoc: add native, ms.

pandoc (2.0)

  [new features]

  * New output format `ms` (groff ms). Complete support, including
    tables, math, syntax highlighting, and PDF bookmarks. The writer uses
    texmath's new eqn writer to convert math to eqn format, so a ms file
    produced with this writer should be processed with `groff -ms -e` if
    it contains math.

  * New output format `jats` (Journal Article Tag Suite).  This is an XML
    format used in archiving and publishing articles.  Note that a
    URI-encoded CSL stylesheet (`data/jats.csl`) is added automatically
    unless a stylesheet is specified using `--css`.

  * New output format `gfm` (GitHub-flavored CommonMark) (#3841).
    This uses bindings to GitHub's fork of cmark, so it should parse
    gfm exactly as GitHub does (excepting certain postprocessing
    steps, involving notifications, emojis, etc.).  `markdown_github`
    has been deprecated in favor of `gfm`.

  * New output format `muse` (Emacs Muse) (Alexander Krotov, #3489).

  * New input format `gfm` (GitHub-flavored CommonMark) (#3841).
    This uses bindings to GitHub's fork of cmark.  `markdown_github`
    has been deprecated in favor of `gfm`.

  * New input format `muse` (Emacs Muse) reader (Alexander Krotov, #3620).

  * New input format `tikiwiki` (TikiWiki markup) (rlpowell, #3800).

  * New input format `vimwiki` (Vimwiki markup) (Yuchen Pei, #3705).
    Note that there is a new data file, `data/vimwiki.css`, which can
    be used to display the HTML produced by this reader and
    pandoc's HTML writer in the style of vimwiki's own HTML
    export.

  * New input format `creole` (Creole 1.0) (#3994, Sascha Wilde).

  * New syntax for Divs, with `fenced_divs` extension enabled by
    default (#168).  This gives an attractive, plain-text way to create
    containers for block-level content.

  * Added new syntax for including raw content in any output format,
    enabled by the `raw_attribute` extension (which is on by default
    for `markdown` and `multimarkdown`).  The syntax is the same as
    for fenced code blocks or code inlines, only with `{=FORMAT}` for
    attributes, where `FORMAT` is the name of the output format
    (e.g., `ms`, `html`).

  * Implement multicolumn support for slide formats (#1710).
    The structure expected is:

        :::::::::::::: {.columns}
        ::: {.column width="40%"}
        contents...
        :::
        ::: {.column width="60%"}
        contents...
        :::
        ::::::::::::::

    Support has been added for beamer and all HTML slide formats.

  * Allows line comments in templates, beginning with `$--` (#3806).
    (Requires doctemplates 0.2.1.)

  * Add `--eol=crlf|lf|native` flag and writer option to control line endings
    (Stefan Dresselhaus, #3663, #2097).

  * Add `--log` option to save log messages in JSON format to a file (#3392).

  * Add `--request-header` option, to set request headers when pandoc
    makes HTTP requests to fetch external resources. For example:
    `--request-header User-Agent:blah`.

  * Added lua filters (Albert Krewinkel, #3514).  The new `--lua-filter`
    option works like `--filter` but takes pathnames of special lua filters
    and uses the lua interpreter baked into pandoc, so that no external
    interpreter is needed.  Note that lua filters are all applied before
    regular filters, regardless of their position on the command line.
    For documentation of lua filters, see `doc/lua-filters.md`.

  * Set `PANDOC_READER_OPTIONS` in environment where filters are run.
    This contains a JSON representation of `ReaderOptions`, so filters
    can access it.

  * Support creation of pdf via groff `ms` and pdfroff.
    `pandoc -t ms -o output.pdf input.txt`.

  * Support for PDF generation via HTML and `weasyprint` or `prince`
    (Mauro Bieg, #3909).  `pandoc -t html5 -o output.pdf --pdf-engine=prince`.

  * Added `--epub-subdirectory` option (#3720).  This specifies the
    subdirectory in the OCF container that holds the EPUB specific content.
    We now put all EPUB related content in an `EPUB/` subdirectory by default
    (later this will be configurable).

    ```
      mimetype
      META-INF/
        com.apple.ibooks.display-options.xml
        container.xml
      EPUB/ <<--configurable-->>
        fonts/ <<--static-->>
        font.otf
      media/ <<--static-->>
        cover.jpg
        fig1.jpg
      styles/ <<--static-->>
        stylesheet.css
      content.opf
      toc.ncx
      text/ <<--static-->>
        ch001.xhtml
    ```

  * Added `--resource-path=SEARCHPATH` command line option (#852).
    SEARCHPATH is separated by the usual character, depending on OS
    (: on unix, ; on windows).  Default resource path is just working
    directory.  However, the working directory must be explicitly
    specified if the `--resource-path` option is used.

  * Added --abbreviations=FILE option for custom abbreviations file
    (#256).  Default abbreviations file (`data/abbreviations`) contains
    a list of strings that will be recognized by pandoc's
    Markdown parser as abbreviations.  (A nonbreaking space will
    be inserted after the period, preventing a sentence space in
    formats like LaTeX.) Users can override the default by putting a file
    abbreviations in their user data directory (`~/.pandoc` on *nix).

  * Allow a theme file as argument to `--highlight-style`.
    Also include a sample, `default.theme`, in `data/`.

  * Allow `--syntax-definition` option for dynamic loading of syntax
    highlighting definitions (#3334).

  * Lists in `markdown` by default now use the CommonMark variable
    nesting rules (#3511). The indentation required for a block-level
    item to be included in a list item is no longer fixed, but is
    determined by the first line of the list item.  To be included in
    the list item, a block must be indented to the level of the first
    non-space content after the list marker. Exception: if are 5 or more
    spaces after the list marker, then the content is interpreted as an
    indented code block, and continuation paragraphs must be indented
    two spaces beyond the end of the list marker.  See the CommonMark
    spec for more details and examples.

    Documents that adhere to the four-space rule should, in most cases,
    be parsed the same way by the new rules.  Here are some examples
    of texts that will be parsed differently:

        - a
          - b

    will be parsed as a list item with a sublist; under the four-space
    rule, it would be a list with two items.

        - a

              code

    Here we have an indented code block under the list item, even though it
    is only indented six spaces from the margin, because it is four spaces
    past the point where a continuation paragraph could begin.  With the
    four-space rule, this would be a regular paragraph rather than a code
    block.

        - a

                code

    Here the code block will start with two spaces, whereas under
    the four-space rule, it would start with `code`.  With the four-space
    rule, indented code under a list item always must be indented eight
    spaces from the margin, while the new rules require only that it
    be indented four spaces from the beginning of the first non-space
    text after the list marker (here, `a`).

    This change was motivated by a slew of bug reports from people
    who expected lists to work differently (#3125, #2367, #2575, #2210,
     #1990, #1137, #744, #172, #137, #128) and by the growing prevalance
    of CommonMark (now used by GitHub, for example).  Those who
    prefer the old behavior can use `-f markdown+four_space_rule`.

  * Added `four_space_rule` extension.  This triggers the old pandoc
    parsing rule for content nested under list items (the "four space
    rule").

  * Added `spaced_reference_links` extension (#2602).  It allows whitespace
    between the two parts of a reference link:  e.g.

        [a] [b]

        [b]: url

    This was previously enabled by default; now it is now forbidden by default.

  * Add `space_in_atx_header` extension (#3512).  This is enabled by default
    in pandoc and GitHub markdown but not the other flavors.
    This requires a space between the opening #'s and the header
    text in ATX headers (as CommonMark does but many other implementations
    do not).  This is desirable to avoid falsely capturing things like

        #hashtag

    or

        #5

  * Add `sourcefile` and `outputfile` template variables (Roland Hieber,
    #3431).

  * Allow ibooks-specific metadata in epubs (#2693).  You can now have
    the following fields in your YAML metadata, and it will be treated
    appropriately in the generated EPUB:

    ```
      ibooks:
        version: 1.3.4
        specified-fonts: false
        ipad-orientation-lock: portrait-only
        iphone-orientation-lock: landscape-only
        binding: true
        scroll-axis: vertical
    ```


  [behavior changes]

  * Reader functions no longer presuppose that CRs have been
    stripped from the input. (They strip CRs themselves, before
    parsing, to simplify the parsers.)

  * Added support for translations (localization) (#3559).
    Currently this only affects the LaTeX reader, for things
    like `\figurename`.  Translation data files for 46 languages
    can be found in `data/translations`.

  * Make `--ascii` work with DocBook output too.

  * Rename `--latex-engine` to `--pdf-engine`,
    and `--latex-engine-opt` to `--pdf-engine-opt`.

  * Removed `--parse-raw` and `readerParseRaw`.  These were confusing.
    Now we rely on the `+raw_tex` or `+raw_html` extension with latex or html
    input.  Thus, instead of `--parse-raw -f latex` we use `-f latex+raw_tex`,
    and instead of `--parse-raw -f html` we use `-f html+raw_html`.

  * With `--filter` R filters are now recognized, even if they are
    not executable (#3940, #3941, Andrie de Vries).

  * Support SVG in PDF output, converting with `rsvg2pdf` (#1793).

  * Make epub an alias for epub3, not epub2.

  * Removed `--epub-stylesheet`; use `--css` instead (#3472, #847).
    Multiple stylesheets may be used.  Stylesheets will be taken both from
    `--css` and from the `stylesheet` metadata field (which can contain
    either a file path or a list of them).

  * `--mathml` and MathML in HTMLMathMethod no longer take an argument.
    The argument was for a bridge JavaScript that used to be necessary
    in 2004.  We have removed the script already.

  * `--katex` improvements.  The latest version is used, and the
    autoload script is loaded by default.

  * Change MathJax CDN default since old one is shutting down (#3544).
    Note:  The new URL requires a version number, which we'll have
    to update manually in subsequent pandoc releases in order to
    take advantage of mathjax improvements.

  * `--self-contained`: don't incorporate elements with `data-external="1"`
    (#2656).  You can leave an external link as it is by adding the attribute
    data-external="1" to the element.  Pandoc will then not try to
    incorporate its content when `--self-contained` is used.  This is
    similar to a feature already supported by the EPUB writer.

  * Allow `--extract-media` to work with non-binary input formats
    (#1583, #2289).  If `--extract-media` is supplied with a non-binary
    input format, pandoc will attempt to extract the contents of all
    linked images, whether in local files, data: uris, or external uris.
    They will be named based on the sha1 hash of the contents.

  * Make `papersize: a4` work regardless of the case of `a4`.
    It is converted to `a4` in LaTeX and `A4` in ConTeXt.

  * Make `east_asian_line_breaks` affect all readers/writers (#3703).

  * Underlined elements are now treated consistently by readers
    (#2270, hftf); they are always put in a Span with class `underline`.
    This allows the user to treat them differently from other emphasis,
    using a filter.  Docx, Org, Textile, Txt2Tags, and HTML readers
    have been changed.

  * Improved behavior of `auto_identifiers` when there are explicit ids
    (#1745).  Previously only autogenerated ids were added to the list
    of header identifiers in state, so explicit ids weren't taken
    into account when generating unique identifiers.  Duplicated
    identifiers could result.  This simple fix ensures that explicitly given
    identifiers are also taken into account.

  * Use `table-of-contents` for contents of toc, make `toc` a boolean
    (#2872).  Changed markdown, rtf, and HTML-based templates accordingly.
    This allows you to set `toc: true` in the metadata; this
    previously produced strange results in some output formats.
    For backwards compatibility, `toc` is still set to the
    toc contents.  But it is recommended that you update templates
    to use `table-of-contents` for the toc contents and `toc`
    for a boolean flag.

  * Change behavior with binary format output to stdout.
    Previously, for binary formats, output to stdout was disabled
    unless we could detect that the output was being piped (and not
    sent to the terminal).  Unfortunately, such detection is not
    possible on Windows, leaving windows users no way to pipe binary
    output.  So we have changed the behavior in the following way:

    + Output to stdout is allowed when it can be determined that
      the output is being piped (on non-Windows platforms).
    + If the `-o` option is not used, binary output is never sent
      to stdout by default; instead, an error is raised.
    + If `-o -` is used, binary output is sent to stdout, regardless
      of whether it is being piped. This works on Windows too.

  * Better error behavior:  uses of `error` have been replaced by
    raising of `PandocError`, which can be trapped and handled by the
    calling program.

  * Removed `hard_line_breaks` extension from `markdown_github` (#3594).
    GitHub has two Markdown modes, one for long-form documents like READMEs
    and one for short things like issue coments. In issue comments, a line
    break is treated as a hard line break. In README, wikis, etc., it is
    treated as a space as in regular Markdown.  Since pandoc is more likely to
    be used to convert long-form documents from GitHub Markdown,
    `-hard_line_breaks` is a better default.

  * Include `backtick_code_blocks` extension in `mardkown_mmd` (#3637).

  * Escape `MetaString` values (as added with `-M/--metadata` flag) (#3792).
    Previously they would be transmitted to the template without any
    escaping.  Note that `--M title='*foo*'` yields a different result from

        ---
        title: *foo*
        ---

    In the latter case, we have emphasis; in the former case, just
    a string with literal asterisks (which will be escaped
    in formats, like Markdown, that require it).

  * Allow `em`, `cm`, `in` for image height/width in HTML, LaTeX (#3450).

  * HTML writer: Insert `data-` in front of unsupported attributes.  Thus,
    a span with attribute `foo` gets written to HTML5 with `data-foo`, so
    it is valid HTML5.  HTML4 is not affected.  This will allow us to use
    custom attributes in pandoc without producing invalid HTML. (With help
    from Wandmalfarbe, #3817.)

  * Plain writer:  improved super/subscript rendering.  We now
    handle more non-digit characters for which there are
    sub/superscripted unicode characters.  When unicode
    sub/superscripted characters are not available, we use
    `_(..)` or `^(..)` (#3518).

  * Docbook, JATS, TEI writers: print INFO message when omitting interior
    header (#3750).  This only applies to section headers inside list items,
    e.g., which were otherwise silently omitted.

  * Change to `--reference-links` in Markdown writer (#3701).  With
    `--reference-location` of `section` or `block`, pandoc will now repeat
    references that have been used in earlier sections.  The Markdown
    reader has also been modified, so that *exactly* repeated references
    do not generate a warning, only references with the same label but
    different targets.  The idea is that, with references after every block,
    one might want to repeat references sometimes.

  * ODT/OpenDocument writer:

    + Support `lang` attribute (#1667).
    + Added support for `--toc` (#2836).  Thanks to @anayrat.

  * Docx writer:

    + `lang` meta, see #1667 (Mauro Bieg, #3515).
    + Change `FigureWithCaption` to `CaptionedFigure` (iandol, #3658).
    + Use `Table` rather than `Table Normal` for table style (#3275).
      `Table Normal` is the default table style and can't be modified.
    + Pass through comments (#2994).  We assume that comments are defined as
      parsed by the docx reader:

        I want <span class="comment-start" id="0" author="Jesse Rosenthal"
        date="2016-05-09T16:13:00Z">I left a comment.</span>some text to
        have a comment <span class="comment-end" id="0"></span>on it.

      We assume also that the id attributes are unique and properly
      matched between comment-start and comment-end.
    + Bookmark improvements.  Bookmark start/end now surrounds content rather
      than preceding it.  Bookmarks generated for Div with id
      (jgm/pandoc-citeproc#205).
    + Add `keywords` metadata to docx document properties (Ian).

  * RST writer: support unknown interpreted text roles by
    parsing them as `Span` with `role` attributes (#3407).  This
    way they can be manipulated in the AST.

  * HTML writer:

    + Line block: Use class instead of style attribute (#1623).  We now
      issue `<div class="line-block">` and include a default definition
      for `line-block` in the default templates, instead of hard-coding a
      `style` on the div.
    + Add class `footnoteBack` to footnote back references (Timm Albers).
      This allows for easier CSS styling.
    + Render SmallCaps as span with smallcaps class (#1592), rather than
      using a style attribute directly.  This gives the user more flexibility
      in styling small caps in CSS.
    + With reveal.js we use `data-src` instead of `src` for images for
      lazy loading.
    + Special-case `.stretch` class for images in reveal.js (#1291).
      Now in reveal.js, an image with class `stretch` in a paragraph
      by itself will stretch to fill the whole screen, with no
      caption or figure environment.

  * Added warnings for non-rendered blocks to writers.

  * Writers now raise an error on template failure.

  * When creating a PDF via LaTeX, warn if the font is missing some
    characters (#3742).

  * Remove initial check for PDF-creating program (#3819).
    Instead, just try running it and raise the exception if it
    isn't found at that point.  This improves things for users of Cygwin
    on Windows, where the executable won't be found by `findExecutable`
    unless `.exe` is added.  The same exception is raised as before, but
    at a later point.

  * Readers issue warning for duplicate header identifiers (#1745).
    Autogenerated header identifiers are given suffixes so as not to clash
    with previously used header identifiers. But they may still coincide with
    an explicit identifier that is given for a header later in the document,
    or with an identifier on a div, span, link, or image. We now issue
    a warning in this case, so users can supply an explicit identifier.

  * CommonMark reader now supports `emoji`, `hard_line_breaks`, `smart`,
    and `raw_html` extensions.

  * Markdown reader:

    + Don't allow backslash + newline to affect block structure (#3730).
      Note that as a result of this change, the following, which formerly
      produced a header with two lines separated by a line break, will
      now produce a header followed by a paragraph:

        # Hi\
        there

      This may affect some existing documents that relied on
      this undocumented and unintended behavior.  This change makes pandoc
      more consistent with other Markdown implementations, and with itself
      (since the two-space version of a line break doesn't work inside ATX
      headers, and neither version works inside Setext headers).


  * Org reader (Albert Krewinkel, unless noted):

    + Support `table.el` tables (#3314).
    + Support macros (#3401).
    + Support the `#+INCLUDE:` file inclusion mechanism (#3510).
      Recognized include types are `example`, `export`, `src`, and
      normal org file inclusion.  Advanced features like line numbers
      and level selection are not implemented yet.
    + Interpret more meta value as inlines.  The values of the following
      meta variables are now interpreted using org-markup instead of
      treating them as pure strings: `keywords` (comma-separated list of
      inlines), `subtitle` (inline values), `nocite` (inline values, can
      be repeated).
    + Support `\n` export option (#3940).  This turns all newlines in the
      text into hard linebreaks.

  * RST reader:

    + Improved admonition support (#223).  We no longer add an
      `admonition` class, we just use the class for the type of admonition,
      `note` for example.  We put the word corresponding to the label in
      a paragraph inside a `Div` at the beginning of the admonition with
      class `admonition-title`.  This is about as close as we can get to
      RST's own output.
    + Initial support of `.. table` directive.  This allows adding captions
      to tables.
    + Support `.. line-block` directive.  This is deprecated but may still
      be in older documents.
    + Support scale and align attributes of images (#2662).
    + Implemented implicit internal header links (#3475).
    + Support RST-style citations (#853).  The citations appear at the end
      of the document as a definition list in a special div with id
      `citations`. Citations link to the definitions.
    + Recurse into bodies of unknown directives (#3432).
      In most cases it's better to preserve the content than
      to emit it.  This isn't guaranteed to have good results;
      it will fail spectacularly for unknown raw or verbatim directives.
    + Handle chained link definitions (#262).  For example,

          .. _hello:
          .. _goodbye: example.com

      Here both `hello` and `goodbye` should link to `example.com`.
    + Support anchors (#262).  E.g.

          `hello`

          .. _hello:

          paragraph

      This is supported by putting "paragraph" in a `Div` with id `hello`.
    + Support `:widths:` attribute for table directive.
    + Implement csv-table directive (#3533).  Most attributes are supported,
      including `:file:` and `:url:`.
    + Support unknown interpreted text roles by parsing them as Span
      with "role" attributes (#3407).  This way they can be manipulated in
      the AST.

  * HTML reader: parse a span with class `smallcaps` as `SmallCaps`.

  * LaTeX reader:
 
    + Implemented `\graphicspath` (#736).
    + Properly handle column prefixes/suffixes.  For example, in
      `\begin{tabular}{>{$}l<{$}>{$}l<{$} >{$}l<{$}}`
      each cell will be interpreted as if it has a `$`
      before its content and a `$` after (math mode).
    + Handle komascript `\dedication` (#1845).  It now adds a
      `dedication` field to metadata.  It is up to the user to supply
      a template that uses this variable.
    + Support all `\textXX` commands, where XX = `rm`, `tt`, `up`, `md`,
      `sf`, `bf` (#3488).  Spans with a class are used when there is
      nothing better.
    + Expand `\newenvironment` macros (#987).
    + Add support for LaTeX subfiles package (Marc Schreiber, #3530).
    + Better support for subfigure package (#3577).
      A figure with two subfigures turns into two pandoc
      figures; the subcaptions are used and the main caption
      ignored, unless there are no subcaptions.
    + Add support for \vdots (Marc Schreiber, #3607).
    + Add basic support for hyphenat package (Marc Schreiber, #3603).
    + Add basic `\textcolor` support (Marc Schreiber).
    + Add support for `tabularx` environment (Marc Schreiber, #3632).
    + Better handling of comments inside math environments (#3113).
      This solves a problem with commented out `\end{eqnarray}` inside
      an eqnarray (among other things).
    + Parse tikzpicture as raw verbatim environment if `raw_tex` extension
      is selected (#3692).  Otherwise skip with a warning.  This is better
      than trying to parse it as text!
    + Add `\colorbox` support (Marc Schreiber).
    + Set identifiers on Spans used for `\label`.
    + Have `\setmainlanguage` set `lang` in metadata.
    + Support etoolbox's `\ifstrequal`.
    + Support `plainbreak`, `fancybreak` et al from the memoir class
      (bucklereed, #3833).
    + Support `\let`.  Also, fix regular macros so they're expanded at the
      point of use, and NOT also the point of definition.  `\let` macros,
      by contrast, are expanded at the point of definition.  Added an
      `ExpansionPoint` field to `Macro` to track this difference.
    + Support simple `\def` macros.  Note that we still don't support
      macros with fancy parameter delimiters, like `\def\foo#1..#2{...}`.
    + Support \chaptername, \partname, \abstractname, etc.  (#3559,
      obsoletes #3560).
    + Put content of `\ref`, `\label`, `\eqref` commands into `Span` with
      attributes, so they can be handled in filters (Marc Schreiber, #3639)
    + Add Support for `glossaries` and `acronym` package (Marc Schreiber,
      #3589). Acronyms are not resolved by the reader, but acronym and
      glossary information is put into attributes on Spans so that they
      can be processed in filters.
    + Use `Link` instead of `Span` for `\ref`.  This makes more sense
      semantically and avoids unnecessary `Span [Link]` nestings when
      references are resolved.
    + Rudimentary support for `\hyperlink`.
    + Support `\textquoteleft|right`, `\textquotedblleft|right` (#3849).
    + Support `\lq`, `\rq`.
    + Implement `\newtoggle`, `\iftoggle`, `\toggletrue|false` from etoolbox
      (#3853).
    + Support `\RN` and `\Rn`, from biblatex (bucklereed, #3854).
    + Improved support for `\hyperlink`, `\hypertarget` (#2549).
    + Support `\k` ogonek accent.
    + Improve handling of accents.  Handle ogonek, and fall back correctly
      with forms like `\"{}`.
    + Better support for ogonek accents.
    + Support for `\faCheck` and `\faClose` (Marc Schreiber, #3727).
    + Support for `xspace` (Marc Schreiber, #3797).
    + Support `\setmainlanguage` or `\setdefaultlanguage` (polyglossia)
      and `\figurename`.
    + Better handling of `\part` in LaTeX (#1905).  Now we parse chapters as
      level 0 headers, and parts as level -1 headers.  After parsing, we
      check for the lowest header level, and if it's less than 1 we bump
      everything up so that 1 is the lowest header level.  So `\part` will
      always produce a header; no command-line options are needed.
    + Add block version of `\textcolor` (Marc Schreiber).
    + `\textcolor` works as inline and block command (Marc Schreiber).
    + `\textcolor` will be parse as span at the beginning of a paragraph
      (Marc Schreiber).
    + Read polyglossia/babel `\text(LANG){...}` (bucklereed)
    + Improved handling of include files in LaTeX reader (#3971).
      Previously `\include` wouldn't work if the included file
      contained, e.g., a begin without a matching end.
    + Support `\expandafter` (#3983).
    + Handle `\DeclareRobustCommand` (#3983).  Currently it's just treated
      as a synonym for `\newcommand`.
    + Handle `\lettrine` (Mauro Bieg).

  * Math improvements due to updates in texmath:

    + Improved handling of accents and upper/lower delimiters.
    + Support for output in GNU eqn format (used with *roff).
    + Allow `\boldsymbol` + a token without braces, and similarly
      with other styling commands.
    + Improve parsing of `\mathop` to allow multi-character operator names.
    + Add thin space after math operators when "faking it with
      unicode."

  * `walk` is now used instead of `bottomUp` in the `ToJSONFilter`
    instance for `a -> [a]` (pandoc-types).  Note that behavior
    will be slightly different, since `bottomUp`'s treatment of
    a function `[a] -> [a]` is to apply it to each sublist of a
    list, while walk applies it only to maximal sublists.
    Usually the latter behavior is what is wanted, and the
    former can be simulated when needed.  But there may be
    existing filters that need to be rewritten in light of the
    new behavior.  Performance should be improved.

  * There are some changes to syntax highlighting due to revisions
    in the `skylighting` library:

    + Support for `powershell` has been added, and many syntax
      definitions have been updated.
    + Background colors have been added to the `kate` style.
    + The way highlighted code blocks are formatted in HTML has
      been changed (David Baynard), in ways that may require
      changes in hard-coded CSS affecting highlighting.
      (If you haven't included hard-coded highlighting CSS in
      your template, you needn't change anything.)


  [API changes]

  * New module `Text.Pandoc.Class` (Jesse Rosenthal, John MacFarlane).
    This contains definitions of the `PandocMonad` typeclass, the
    `PandocIO` and `PandocPure` monads, and associated functions.

  * Changed types of all writers and readers.

    + We now use `Text` instead of `String` in the interface (#3731).
      (We have not yet changed the internals of most readers to work
      with `Text`, but making this change in the API now opens up a
      path to doing that.)
    + The result is now of form `m a` with constraint `PandocMonad m`.
      Readers and writers can be combined to form monadic values which
      can be run using either `runIO` or `runPure`.  If `runIO` is used,
      then both readers and writers will be able to do IO when needed
      (for include files, for example); if `runPure` is used,
      then the functions are pure and will not touch IO.
    + Where previously you used
      `writeRST def (readMarkdown def "[foo](url)")`, now you
      would use
      `runPure $ readMarkdown def (pack "[foo](url)") >>= writeRST def`.

  * New module `Text.Pandoc.Readers` (Albert Krewinkel).  This
    contains reader helper functions formerly defined in the
    top-level `Text.Pandoc` module.

    + Changed `StringReader` -> `TextReader`.
    + `getReader` now returns a pair of a reader and
      `Extensions`, instead of building the extensions into the
      reader (#3659).  The calling code must explicitly set
      `readerExtensions` using the `Extensions` returned.  The
      point of the change is to make it possible for the calling
      code to determine what extensions are being used.

  * New module `Text.Pandoc.Writers` (Albert Krewinkel).
    This contains writer helper functions formerly defined in the
    top-level `Text.Pandoc` module.

    + Changed `StringWriter` -> `TextWriter`.
    + `getWriter` now retuns a pair of a reader and
      `Extensions`, instead of building the extensions into the
      reader (#3659).  The calling code must explicitly set
      `readerExtensions` using the `Extensions` returned.  The
      point of the change is to make it possible for the calling
      code to determine what extensions are being used.

  * New module `Text.Pandoc.Lua`, exporting `runLuaFilter` (Albert Krewinkel,
    #3514).

  * New module `Text.Pandoc.App`.  This abstracts out the functionality
    of the command line program (`convertWithOpts`), so it can be reproduced
    e.g. in a desktop or web application.  Instead of exiting, we throw errors
    (#3548), which are caught (leading to exit) in pandoc.hs, but allow other
    users of `Text.Pandoc.App` to recover.  `pandoc.hs` is now a 2-liner.
    The module also exports some utility functions for parsing options
    and running filters.

  * New module `Text.Pandoc.Logging` (exported module) (#3392).
    This now contains the `Verbosity` definition previously in
    `Text.Pandoc.Options`, as well as a new `LogMessage` datatype that will
    eventually be used instead of raw strings for warnings.  This will enable
    us, among other things, to provide machine-readable warnings if desired.
    Include ToJSON instance and showLogMessage.  This gives us the possibility
    of both machine-readable and human-readable output for log messages.

  * New module `Text.Pandoc.BCP47`, with `getLang`, `Lang(..)`, `parseBCP47`.

  * New module `Text.Pandoc.Translations`, exporting `Term`,
    `Translations`, `readTranslations`.

  * New module `Text.Pandoc.Readers.LaTeX.Types', exporting `Macro`, `Tok`,
    `TokType`, `Line`, `Column`.

  * `Text.Pandoc.Error`: added many new constructors for `PandocError`.

  * Expose some previously private modules (#3260).  These are often
    helpful to people writing their own reader or writer modules:

    + `Text.Pandoc.Writers.Shared`
    + `Text.Pandoc.Parsing`
    + `Text.Pandoc.Asciify`
    + `Text.Pandoc.Emoji`
    + `Text.Pandoc.ImageSize`
    + `Text.Pandoc.Highlighting`
`
  * New module `Text.Pandoc.Extensions` (Albert Krewinkel):
    Extension parsing and processing functions were defined in the top-level
    `Text.Pandoc` module.  These functions are moved to the Extensions
    submodule as to enable reuse in other submodules.

  * Add `Ext_raw_attribute` constructor for `Extension`.

  * Add `Ext_fenced_divs` constructor for `Extension'.

  * Add `Ext_four_space_rule` constructor in `Extension`.

  * Add `Ext_gfm_auto_identifiers` constructor for `Extension`.

  * Add `Monoid` instance for `Extensions`.

  * Add `Text.Pandoc.Writers.Ms`, exporting `writeMs`.

  * Add `Text.Pandoc.Writers.JATS`, exporting `writeJATS`.

  * Add `Text.Pandoc.Writers.Muse`, exporting `writeMuse`.

  * Add `Text.Pandoc.Readers.Muse`, exporting `readMuse`.

  * Add `Text.Pandoc.Readers.TikiWiki`, exporting `readTikiWiki`.

  * Add `Text.Pandoc.Readers.Vimwiki`, exporting `readVimwiki`.

  * Add `Text.Pandoc.Readers.Creole`, exporting `readCreole`.

  * Export `setVerbosity` from `Text.Pandoc`.

  * `Text.Pandoc.Pretty`: Add `Eq` instance for `Doc`.

  * `Text.Pandoc.XML`: `toEntities`: changed type to `Text -> Text`.

  * `Text.Pandoc.UTF8`:

    + Export `fromText`, `fromTextLazy`, `toText`, `toTextLazy`.
      Define `toString`, `toStringLazy` in terms of them.
    + Add new functions parameterized on `Newline`: `writeFileWith`,
      `putStrWith`, `putStrLnWith`, `hPutStrWith`, `hPutStrLnWith`.

  * `Text.Pandoc.MediaBag`: removed `extractMediaBag`.

  * `Text.Pandoc.Highlighting`:

    + `highlighting` now returns an Either rather than Maybe.
      This allows us to display error information returned by the skylighting
      library.  Display a warning if the highlighting library throws an error.
    + Add parameter for `SyntaxMap` to `highlight`.

  * `Text.Pandoc.Writers.Math`:

    + Export `defaultMathJaxURL`, `defaultKaTeXURL`.  This will ensure that
      we only need to update these in one place.

  * `Text.Pandoc.SelfContained`:

    + Removed `WriterOptions` parameter from `makeSelfContained`.
    + Put `makeSelfContained` in PandocMonad instead of IO.  This removes
      the need to pass MediaBag around and improves exceptions.  It also
      opens up the possibility of using makeSelfContained purely.
    + Export `makeDataURI`.

  * `Text.Pandoc.ImageSize`:

    + Export `lengthToDim`, new function `scaleDimension`.
    + Export `inEm` from ImageSize (#3450).
    + Change `showFl` and `show` instance for `Dimension` so
      extra decimal places are omitted.
    + Added `Em` as a constructor of `Dimension`.
    + Add `WriterOptions` parameter to `imageSize` signature (Mauro Bieg).

  * `Text.Pandoc.Templates`:

    + Change type of `renderTemplate'`.  Now it runs in `PandocMonad`
      and raises a proper `PandocTemplateError` if there are problems, rather
      than failing with uncatchable `error`.
    + Change signature of `getDefaultTemplate`.  Now it runs in any instance
      of `PandocMonad`, and returns a `String` rather than an `Either` value.
      And it no longer takes a `datadir` parameter, since this can be
      retrieved from `CommonState`.

  * `Text.Pandoc.Options`:

    + Added `writerEpubSubdirectory` to `WriterOptions` (#3720).
      The EPUB writer now takes its EPUB subdirectory from this option.
    + In `WriterOptions`, rename `writerLaTeXEngine` to `writerPdfEngine`
      and `writerLaTeXArgs` to `writerPdfArgs` (Mauro Bieg, #3909).
    + Add `writerSyntaxMap` to `WriterOptions`.
    + Removed `writerEpubStylesheet` from `WriterOptions`.
    + Remove `writerUserDataDir` from `WriterOptions`.  It is now carried
      in `CommonState` in `PandocMonad` instances.  (And thus it can be used
      by readers too.)
    + Changed `writerEpubMetadata` to a `Maybe String`.
    + Removed `readerApplyMacros` from `ReaderOptions`.  Now we just check
      the `latex_macros` reader extension.
    + FromJSON/ToJSON instances for `ReaderOptions`.
    + In `HTMLMathMethod`, the `KaTeX` contsructor now takes only
      one string (for the KaTeX base URL), rather than two.
    + Removed `writerSourceURL` from `WriterOptions`.  We now use
      `stSourceURL` in `CommonState`, which is set by `setInputFiles`.

  * `Text.Pandoc.Shared`:

    + `tabFilter` now takes a `Text`, not `String`.
    + `openURL`: Changed type from an Either.  Now it will just raise
      an exception to be trapped later.
    + Remove `normalizeSpaces` (#1530).
    + Remove `warn`.  (Use `report` from `Text.Pandoc.Class` instead.)
    + Export a new function `crFilter`.
    + Add `eastAsianLineBreakFilter` (previously in Markdown reader).
    + Provide custom `isURI` that rejects unknown schemes.
      (Albert Krewinkel, #2713).  We also export the set of known
      `schemes`.  The new function replaces the function of the same name
      from `Network.URI`, as the latter did not check whether a scheme is
      well-known.  All official IANA schemes (as of 2017-05-22) are
      included in the set of known schemes.  The four non-official schemes
      `doi`, `isbn`, `javascript`, and `pmid` are kept.
    + Remove `err`.
    + Remove `readDataFile`, `readDefaultDataFile`, `getReferenceDocx`,
      `getReferenceODT`. These now live in `Text.Pandoc.Class`,
      where they are defined in terms of `PandocMonad`
      primitives and have different signatures.
    + Remove `openURL`.  Use `openURL` from `Text.Pandoc.Class` instead.
    + Add `underlineSpan`.

  * `Text.Pandoc.Readers.HTML`: export new `NamedTag` class.

  * `Text.Pandoc.Readers.Markdown`: remove `readDocxWithWarnings`.
     With the new API one can simply use `getLog` after running
     the reader.

  * `Text.Pandoc.Readers.LaTeX`: Changed types for `rawLaTeXInline`
     and `rawLaTeXBlock`.  (Both now return a `String`, and they are
     polymorphic in state.)


  [bug fixes and under-the-hood improvements]

  * TEI writer: Added identifiers on `<div>` elements.

  * DokuWiki reader: Better handling for code block in list item (#3824).

  * Custom writer: Remove old preprocesesor conditionals (Albert Krewinkel).

  * ZimWiki writer: Removed internal formatting from note and table cells,
    because ZimWiki does not support it (Alex Ivkin, #3446).

  * MediaWiki writer:

    + Updated list of syntax highlighting languages (#3461).
      Now `r` gets you `<source>` rather than `<code>` (among others).
    + Add display attribute on `<math>` tags (#3452).  This allows display
      math to be rendered properly.
    + Remove newline before `</ref>` (#2652).
    + Don't softbreak lines inside list items (#3531).

  * Org writer:

    + Reduce to two spaces after bullets (#3417, Albert Krewinkel).
    + Add unit tests (Alexander Krotov).
    + Stop using raw HTML to wrap divs (Albert Krewinkel, #3771).
    + Do not strip `#` from Org anchor links (Alexander Krotov).

  * CommonMark writer:

    + Avoid excess blank lines at end of output.
    + Prefer pipe tables to HTML tables even if it means losing relative
      column width information (#3734).
    + Support table, strikethrough extensions, when enabled (as with gfm).
      Note that we bypass the commonmark writer from cmark and construct our
      own pipe tables, with better results.
    + Properly support `--wrap=none`.
    + Use smallcaps class for `SmallCaps` (#1592).
    + Omit "fig:" prefix in image titles.  This is used internally to
      indicate internal figures.

  * RST writer:

    + Properly handle table captions.
    + Don't wrap lines in in definition list terms.  Wrapping is not allowed.
    + Implemented `+/-smart` and improved escaping with `+smart`.
    + Add empty comments when needed to avoid including a blockquote
      in the indented content of a preceding block (#3675).
    + Improve grid table output, fix bug with empty rows (#3516).
      Uses the new `gridTable` in Writers.Shared, which is here
      improved to better handle 0-width cells.
    + Remove space at beginning/end of RST code span (#3496).  Otherwise
      we get invalid RST.  There seems to be no way to escape the space.
    + Add header anchors when header has non-standard id (#3937).
    + Correctly handle inline code containing backticks, using a `:literal:`
      role (#3974).
    + Don't backslash-escape word-internal punctuation (#3978).

  * Markdown writer:

    + Don't include variables in metadata blocks.  Previously variables set
      on the command line were included in e.g. YAML metadata, contrary to
      documentation and intentions.
    + Improved escaping with `+smart`.
    + Fixed grid tables embedded in grid tables (#2834).
    + Use span with class 'smallcaps' for SmallCaps, instead of a style
      attribute as before (#1592).
    + Escape initial `%` in a paragraph if the `pandoc_title_blocks`
      extension is enabled (#3454).  Otherwise in a document starting with
      a literal `%` the first line is wrongly interpreted as a title.
    + Fixed false ordered lists in YAML metadata (#3492, #1685).  Now we
      properly escape things that would otherwise start ordered lists,
      such as

          ---
          title: 1. inline
          ...
    + Better handling of tables with empty columns (#3337).  We now
      calculate the number of columns based on the longest row (or the
      length of aligns or widths).
    + Escape unordered list markers at beginning of paragraph (#3497), to
      avoid false interpretation as a list.
    + Escape `|` appropriately.
    + Ensure space before list at top level (#3487).
    + Avoid spurious blanklines at end of document after tables and list,
      for example.
    + Fixed bugs in simple/multiline list output (#3384).
      Previously we got overlong lists with `--wrap=none`.  This is fixed.
      Previously a multiline list could become a simple list (and would
      always become one with `--wrap=none`).
    + Don't emit a simple table if `simple_tables` disabled (#3529).
    + Case-insensitive reference links (David A Roberts, #3616).
      Ensure that we do not generate reference links whose labels differ only
      by case.  Also allow implicit reference links when the link
      text and label are identical up to case.
    + Put space before reference link definitions (Mauro Bieg, #3630).
    + Better escaping for links (David A. Roberts, #3619).  Previously the
      Markdown writer would sometimes create links where there were none
      in the source.  This is now avoided by selectively escaping bracket
      characters when they occur in a place where a link might be created.
    + Added missing `\n` (David A. Roberts, #3647).
    + Fixed duplicated reference links with `--reference-links`
      and `--reference-location=section` (#3674).  Also ensure that there
      are no empty link references `[]`.
    + Avoid inline surround-marking with empty content (#3715).
      E.g. we don't want `<strong></strong>` to become `****`.
      Similarly for emphasis, super/subscript, strikeout.
    + Don't allow soft break in header (#3736).
    + Make sure `plain`, `markdown_github`, etc. work for raw.
      Previously only `markdown` worked.  Note: currently a raw block labeled
      `markdown_github` will be printed for any `markdown` format.
    + Ensure that `+` and `-` are escaped properly so they don't cause
      spurious lists (#3773).  Previously they were only
      if succeeded by a space, not if they were at end of line.
    + Use pipe tables if `raw_html` disabled and `pipe_tables` enabled,
      even if the table has relative width information (#3734).
    + Markdown writer: don't crash on `Str ""`.
    + Make `Span` with null attribute transparent.  That is, we don't use
      brackets or `<span>` tags to mark spans when there are no attributes;
      we simply output the contents.
    + Escape pipe characters when `pipe_tables` enabled (#3887).
    + Better escaping of `<` and `>`.  If `all_symbols_escapable` is set,
      we backslash escape these.  Otherwise we use entities as before.
    + When writing plain, don't use `&nbsp;` to separate list and indented
      code.  There's no need for it in this context, since this isn't to be
      interpreted using Markdown rules.
    + Preserve classes in JS obfuscated links (Timm Albers, #2989).
      HTML links containing classes originally now preserve them when using
      JavaScript email obfuscation.
    + Render `SmallCaps` as a native span when `native_spans` are enabled.
    + Always write attributes with `bracketed_spans` (d-dorazio).

  * Man writer:

    + Fix handling of nested font commands (#3568).  Previously pandoc emitted
      incorrect markup for bold + italic, for example, or bold + code.
    + Avoid error for definition lists with no definitions (#3832).

  * DocBook writer:

    + Fix internal links with `writerIdentifierPrefix opt`
      (#3397, Mauro Bieg).

  * Docx writer:

    + Don't include bookmarks on headers unless non-null id (#3476).
    + Support 9 levels of headers (#1642).
    + Allow 9 list levels (#3519).
    + Don't take `distArchive` from datadir (#3322).  The docx writer takes
      components from the distribution's version of `reference.docx` when it
      can't find them in a user's custom `reference.docx`.  Previously, we
      allowed a `reference.docx` in the data directory (e.g. `~/.pandoc`)
      to be used as the distribution's reference.docx.  This led to a
      bizarre situation where pandoc would produce a good docx using
      `--template ~/.pandoc/ref.docx`, but if `ref.docx` were moved to
      `~/.pandoc/reference.docx`, it would then produce a corrupted docx.
    + Fixed handling of soft hyphen (0173) (#3691).
    + Better handling of keywords (#3719).
    + Cleaner code for handling dir and style attributes for `Div`.
    + Use `Set` for dynamic styles to avoid duplicates.
    + Removed redundant element from data/docx/word/numbering.xml.
      The elements we need are generated when the document is
      compiled; this didn't do anything.
    + Activate `evenAndOddHeaders` from reference docx (#3901,
      Augustín Martín Barbero).

  * ODT/OpenDocument writer:

    + Calculate aspect ratio for percentage-sized images (Mauro Bieg, #3239).
    + Use more widely available bullet characters (#1400).  The old
      characters weren't available in some font sets.  These seem to work
      well on Windows and Linux versions of LibreOffice.
    + Wider labels for lists (#2421).  This avoids overly narrow labels for
      ordered lists with `()` delimiters.  However, arguably it creates
      overly wide labels for bullets.  Also, lists now start flush with
      the margin, rather than indented.
    + Fixed dropped elements in some ordered lists (#2434).

  * FB2 writer:

    + Don't render `RawBlock` as code.
    + Don't fail with an error on interior headers (e.g. in list) (#3750).
      Instead, omit them with an INFO message.
    + Add support for "lang" metadata (Alexander Krotov, #3625).
    + Format `LineBlock` as poem (Alexander Krotov).  Previously writer
      produced one paragraph with `<empty-line/>` elements, which are not
      allowed inside `<p>` according to FB2 schema.
    + Replace `concatMap` with `cMap` (Alexander Krotov).
    + Write FB2 lists without nesting blocks inside `<p>` (Alexander
      Krotov, #4004)

  * HTML writer:

    + Make sure `html4`, `html5` formats work for raw blocks/inlines.
    + Render raw inline environments when `--mathjax` used (#3816).
      We previously did this only with raw blocks, on the assumption
      that math environments would always be raw blocks. This has changed
      since we now parse them as inline environments.
    + Ensure we don't get two style attributes for width and height.
    + Report when not rendering raw inline/block.
    + Issue warning if no title specified and template used (#3473).
    + Info message if `lang` is unspecified (#3486).
    + Removed unused parameter in `dimensionsToAttributeList`.
    + Avoid two class attributes when adding `uri` class (#3716).
    + Fix internal links with `writerIdentifierPrefix opt` (#3397, Mauro
      Bieg).
    + Use revealjs's math plugin for mathjax (#3743).  This is a thin
      wrapper around mathjax that makes math look better on revealjs.
    + Slidy:  use h1 for all slides, even if they were originally
      level 2 headers (#3566).  Otherwise the built-in table of contents
      in Slidy breaks.

  * LaTeX writer:

    + Don't render LaTeX images with data: URIs (#3636).  Note that
      `--extract-media` can be used when the input contains data: URIs.
    + Make highlighted code blocks work in footnotes (Timm Albers).
    + Don't use figure inside table cell (#3836).
    + Use proper code for list enumerators (#3891).  This should fix problems
      with lists that don't use arabic numerals.
    + Always add hypertarget when there's a non-empty identifier (#2719).
      Previously the hypertargets were only added when there was actually
      a link to that identifier.
    + Use `%` after hypertarget before code block.
    + Add `\leavevmode` before hypertarget at start of paragraph (#2704,
      fixes formatting problems in beamer citations).
    + Don't use `lstinline` in \item[..] (#645).  If you do, the contents
      of item disappear or are misplaced.  Use `\texttt` instead.
    + Fix problem with escaping in `lstinline` (#1629).  Previously the
      LaTeX writer created invalid LaTeX when `--listings` was specified and
      a code span occured inside emphasis or another construction.
    + Fix error with line breaks after empty content (#2874).  LaTeX
      requires something before a line break, so we insert a `~` if no
      printable content has yet been emitted.
    + Use BCP47 parser.
    + Fixed detection of otherlangs (#3770).  We weren't recursing into
      inline contexts.
    + Handle language in inline code with `--listings` (#3422).
    + Write euro symbol directly in LaTeX (Andrew Dunning, #3801).
      The textcomp package allows pdfLaTeX to parse `€` directly, making the
      `\euro` command unneeded.
    + Fixed footnotes in table captions (#2378).  Note that if the table has
      a first page header and a continuation page header, the notes will
      appear only on the first occurrence of the header.
    + In `writeBeamer` output, allow hyperlinks to frames (#3220).
      Previously you could link to a header above or below slide level but
      not *to* slide level.  This commit changes that.  Hypertargets are
      inserted inside frame titles; technically the reference is to just
      after the title, but in normal use (where slides are viewed full
      screen in a slide show), this does not matter.
    + Remove `\strut` at beginning of table cells (#3436).  This fixes a
      problem with alignment of lists in table cells.  The `\strut` at the
      end seems to be enough to avoid the too-close spacing that motivated
      addition of the strut  in #1573.
    + Add partial siunitx Support (Marc Schreiber, #3588).

  * ConTeXt writer:

    + Refactored to use BCP47 module.
    + Remove unnecessary `$` (Alexander Krotov, #3482).
    + Use header identifiers for chapters (#3968).

  * EPUB writer:

    + `title_page.xhtml` is now put in `text/`.
    + Don't strip formatting in TOC (#1611).

  * Textile reader:

    + Fix bug for certain links in table cells (#3667).
    + Allow 'pre' code in list item (#3916).

  * HTML reader:

    + Added warnings for ignored material (#3392).
    + Better sanity checks to avoid parsing unintended things as
      raw HTML in the Markdown reader (#3257).
    + Revise treatment of `li` with `id` attribute (#3596).  Previously we
      always added an empty div before the list item, but this created
      problems with spacing in tight lists.  Now we do this: If the list
      item contents begin with a `Plain` block, we modify the `Plain`
      block by adding a `Span` around its contents.  Otherwise, we add a
      `Div` around the contents of the list item (instead of adding an
      empty `Div` to the beginning, as before).
    + Add `details` tag to list of block tags (#3694).
    + Removed `button` from block tag list (#3717).  It is already in the
      `eitherBlockOrInlineTag` list, and should be both places.
    + Use `Set`s instead of lists for block tag lookup.
    + Rewrote to use `Text` throughout.  Effect on memory usage is modest
      (< 10%).
    + Use the lang value of `<html>` to set the lang meta value (bucklereed,
      #3765).
    + Ensure that paragraphs are closed properly when the parent block
      element closes, even without `</p>` (#3794).
    + Parse `<figure>` and `<figcaption>` (Mauro Bieg, #3813).
    + Parse `<main>` like `<div role=main>` (bucklereed, #3791).
      `<main>` closes `<p>` and behaves like a block element generally
    + Support column alignments (#1881).  These can be set either
      with a `width` attribute or with `text-width` in a `style` attribute.
    + Modified state type to be an instance of `HasLogMessages`, so
      `registerHeader` can issue warnings.
    + `</td>` or `</th>` should close any open block tag (#3991).
    + `<td>` should close an open `<th>` or `<td>`.
    + `htmlTag` improvements (#3989).  We previously failed on cases
      where an attribute contained a `>` character. This patch fixes the
      bug, which especially affects raw HTML in Markdown.

  * Txt2Tags reader:

    + Newline is not indentation (Alexander Krotov).

  * MediaWiki reader:

    + Allow extra hyphens after `|-` in tables (#2649).
    + Allow blank line after table start (#2649).
    + Fixed more table issues (#2649).
    + Ensure that list starts begin at left margin (#2606).  Including when
      they're in tables or other list items.
    + Make smart double quotes depend on `smart` extension (#3585).
    + Don't do curly quotes inside `<tt>` contexts (#3585).  Even if `+smart`.
    + Modified state type to be an instance of `HasLogMessages`, so
      `registerHeader` can issue warnings.

  * TWiki reader (Alexander Krotov):

    + Remove unnecessary `$` (#3597).
    + Simplify `linkText` (#3605).

  * EPUB reader:

    + Minor refactoring, avoiding explicit MediaBag handling.
      This all works behind the scenes in CommonState plumbing.

  * Docx reader:

    + Don't drop smartTag contents (#2242).
    + Handle local namespace declarations (#3365).  Previously we didn't
      recognize math, for example, when the xmlns declaration occured on
      the element and not the root.
    + More efficient trimSps (#1530).  Replacing `trimLineBreaks`.  This
      does the work of `normalizeSpaces` as well, so we avoid the need for
      that function here.
    + Avoid 0-level headers (Jesse Rosenthal, #3830).  We used to parse
      paragraphs styled with "HeadingN" as "nth-level header." But if a
      document has a custom style named "Heading0", this will produce a
      0-level header, which shouldn't exist. We only parse this style
      if N>0. Otherwise we treat it as a normal style name, and
      follow its dependencies, if any.
    + Add tests for avoiding zero-level header (Jesse Rosenthal).

  * ODT reader:

    + Replaced `collectRights` with Rights from `Data.Either`.
    + Remove dead code (Albert Krewinkel).

  * Org reader (Albert Krewinkel, unless noted).

    + Don't allow tables inside list items (John MacFarlane, #3499).
    + Disallow tables on list marker lines (#3499).
    + Convert markup at beginning of footnotes (John MacFarlane, #3576).
    + Allow emphasized text to be followed by `[` (#3577).
    + Handle line numbering switch for src blocks.
      The line-numbering switch that can be given to source blocks (`-n` with
      an start number as an optional parameter) is parsed and translated to a
      class/key-value combination used by highlighting and other readers and
      writers.
    + Stop adding rundoc prefix to src params.  Source block parameter names
      are no longer prefixed with `rundoc`. This was intended to simplify
      working with the rundoc project, a babel runner. However, the rundoc
      project is unmaintained, and adding those markers is not the reader's
      job anyway.  The original language that is specified for a source
      element is now retained as the `data-org-language` attribute and only
      added if it differs from the translated language.
    + Allow multi-word arguments to src block params (#3477).  The reader now
      correctly parses src block parameter list even if parameter arguments
      contain multiple words.
    + Avoid creating `nullMeta` by applying `setMeta` directly
      (Alexander Krotov).
    + Replace `sequence . map` with `mapM`.
    + Fix smart parsing behavior.  Parsing of smart quotes and special
      characters can either be enabled via the `smart` language extension or
      the `'` and `-` export options. Smart parsing is active if either the
      extension or export option is enabled.  Only smart parsing of special
      characters (like ellipses and en and em dashes) is enabled by default,
      while smart quotes are disabled.  Previously, all smart parsing was
      disabled unless the language extension was enabled.
    + Subject full doc tree to headline transformations (Albert Krewinkel,
      #3695).  Emacs parses org documents into a tree structure, which is
      then post-processed during exporting. The reader is changed to do the
      same, turning the document into a single tree of headlines starting
      at level 0.
    + Fix cite parsing behaviour (Herwig Stuetz).  Until now, `org-ref`
      cite keys included special characters also at the end. This caused
      problems when citations occur right before colons or at the end of
      a sentence.  With this change, all non alphanumeric characters at
      the end of a cite key are ignored.  This also adds `,` to the list
      of special characters that are legal in cite keys to better mirror
      the behaviour of org-export.
    + Fix module names in haddock comments.  Copy-pasting had lead to
      haddock module descriptions containing the wrong module names.
    + Recognize babel result blocks with attributes (#3706).  Babel
      result blocks can have block attributes like captions and names.
      Result blocks with attributes were not recognized and were parsed
      as normal blocks without attributes.
    + Include tags in headlines.  The Emacs default is to include tags in the
      headline when exporting.  Instead of just empty spans, which contain the
      tag name as attribute, tags are rendered as small caps and wrapped in
      those spans.  Non-breaking spaces serve as separators for multiple tags.
    + Respect export option for tags (#3713).  Tags are appended to
      headlines by default, but will be omitted when the `tags` export option
      is set to nil.
    + Use `tag-name` attribute instead of `data-tag-name`.
    + Use `org-language` attribute rather than `data-org-language`.
    + Modified state type to be an instance of `HasLogMessages`, so
      `registerHeader` can issue warnings.
    + End footnotes after two blank lines.  Footnotes can not only be
      terminated by the start of a new footnote or a header, but also by two
      consecutive blank lines.
    + Update emphasis border chars (#3933).  The org reader was updated to
      match current org-mode behavior: the set of characters which are
      acceptable to occur as the first or last character in an org emphasis
      have been changed and now allows all non-whitespace chars at the
      inner border of emphasized text (see `org-emphasis-regexp-components`).

  * RST reader:

    + Fixed small bug in list parsing (#3432).  Previously the parser didn't
      handle properly this case:

          * - a
            - b
          * - c
            - d
    + Handle multiline cells in simple tables (#1166).
    + Parse list table directive (Keiichiro Shikano, #3432).
    + Make use of `anyLineNewline` (Alexander Krotov, #3686).
    + Use `anyLineNewline` in `rawListItem` (Alexander Krotov, #3702).
    + Reorganize block parsers for ~20% faster parsing.
    + Fixed `..include::` directive (#3880).
    + Handle blank lines correctly in line blocks (Alexander Krotov, #3881).
      Previously pandoc would sometimes combine two line blocks separated
      by blanks, and ignore trailing blank lines within the line block.
    + Fix indirect hyperlink targets (#512).

  * Markdown reader:

    + Allow attributes in reference links to start on next line (#3674).
    + Parse YAML metadata in a context that sees footnotes defined in
      the body of the document (#1279).
    + When splitting pipe table cells, skip tex math (#3481).
      You might have a `|` character inside math.  (Or for that matter
      something that the parser might mistake for raw HTML.)
    + Treat span with class `smallcaps` as SmallCaps.
      This allows users to specify small caps in Markdown this way:
      `[my text]{.smallcaps}` (#1592).
    + Fixed internal header links (#2397).
      This patch also adds `shortcut_reference_links` to the list
      of mmd extensions.
    + Treat certain environments as inline
      when they occur without space surrounding them (#3309, #2171).
      E.g. equation, math.  This avoids incorrect vertical space
      around equations.
    + Optimized `nonindentSpaces`.  Makes the benchmark go from 40 to 36 ms.
    + Allow latex macro definitions indented 1-3 spaces.
      Previously they only worked if nonindented.
    + Improved parsing of indented raw HTML blocks (#1841).
      Previously we inadvertently interpreted indented HTML as
      code blocks.  This was a regression.  We now seek to determine the
      indentation level of the contents of an HTML block, and (optionally)
      skip that much indentation.  As a side effect, indentation may be
      stripped off of raw HTML blocks, if `markdown_in_html_blocks` is
      used. This is better than having things interpreted as indented
      code blocks.
    + Fixed smart quotes after emphasis (#2228).  E.g. in `*foo*'s 'foo'`.
    + Warn for notes defined but not used (#1718).
    + Use `anyLineNewline` (Alexander Krotov).
    + Interpret YAML metadata as Inlines when possible (#3755).  If
      the metadata field is all on one line, we try to interpret it as
      Inlines, and only try parsing as Blocks if that fails.  If it
      extends over one line (including possibly the `|` or `>` character
      signaling an indented block), then we parse as Blocks.  This was
      motivated by some German users finding that `date: '22. Juin 2017'`
      got parsed as an ordered list.
    + Fixed spurious parsing as citation as reference def (#3840).
      We now disallow reference keys starting with `@` if the
      `citations` extension is enabled.
    + Parse `-@roe` as suppress-author citation (pandoc-citeproc#237).
      Previously only `[-@roe]` (with brackets) was recognized as
      suppress-author, and `-@roe` was treated the same as `@roe`.
    + Fixed parsing of fenced code after list when there is no intervening
      blank line (#3733).
    + Allow raw latex commands starting with `\start` (#3558).  Previously
      these weren't allowed because they were interpreted as starting
      ConTeXt environments, even without a corresponding `\stop`...
    + Added `inlines`, `inlines1`.
    + Require nonempty alt text for `implicit_figures` (#2844).
      A figure with an empty caption doesn't make sense.
    + Removed texmath macro material; now all this is handled
      in the LaTeX reader functions.
    + Fixed bug with indented code following raw LaTeX (#3947).

  * LaTeX reader:

    + Rewrote LaTeX reader with proper tokenization (#1390,
      #2118, #3236, #3779, #934, #982).  This rewrite is primarily
      motivated by the need to get macros working properly.  A side benefit
      is that the reader is significantly faster.  We now tokenize the
      input text, then parse the token stream.  Macros modify the token
      stream, so they should now be effective in any context, including
      math. Thus, we no longer need the clunky macro processing
      capacities of texmath.
    + Parse `\,` to `\8198` (six-per-em space) (Henri Werth).
    + Allow `\newcommand\foo{blah}` without braces.
    + Support `\lstinputlisting` (#2116).
    + Issue warnings when skipping unknown latex commands (#3392).
    + Include contents of `\parbox`.
    + Allow `\hspace` and `\vspace` to count as raw block or inline.
      Previously we would refuse to parse anything as raw inline if
      it was in the `blockCommands` list.  Now we allow exceptions
      if they're listed under ignoreInlines in inlineCommands.
      This should make it easier e.g. to include an `\hspace`
      between two side-by-side raw LaTeX tables.
    + Don't drop contents of `\hypertarget`.
    + Handle spaces before `\cite` arguments.
    + Allow newpage, clearpage, pagebreak in inline contexts as well as
      block contexts (#3494).
    + Treat `{{xxx}}` the same as `{xxx}` (#2115).
    + Use `pMacroDefinition` in macro (for more direct parsing).
      Note that this means that `macro` will now parse one
      macro at a time, rather than parsing a whole group together.
    + Fixed failures on \ref{}, \label{} with `+raw_tex`.  Now these
      commands are parsed as raw if `+raw_tex`; otherwise, their argument
      is parsed as a bracketed string.
    + Don't crash on empty `enumerate` environment (#3707).
    + Handle escaped `&` inside table cell (#3708).
    + Handle block structure inside table cells (#3709).  `minipage` is no
      longer required.
    + Handle some width specifiers on table columns (#3709).  Currently
      we only handle the form `0.9\linewidth`.  Anything else would have
      to be converted to a percentage, using some kind arbitrary assumptions
      about line widths.
    + Make sure `\write18` is parsed as raw LaTeX.  The change is in the
      LaTeX reader's treatment of raw commands, but it also affects the
      Markdown reader.
    + Fixed regression with starred environment names (#3803).
    + Handle optional args in raw `\titleformat` (#3804).
    + Improved heuristic for raw block/inline.  An unknown command at the
      beginning of the line that could be either block or inline is
      treated as block if we have a sequence of block commands followed by
      a newline or a `\startXXX` command (which might start a raw ConTeXt
      environment).
    + Don't remove macro definitions from the output, even if
      `Ext_latex_macros` is set, so that macros will be applied.
      Since they're only applied to math in Markdown, removing the macros
      can have bad effects.  Even for math macros, keeping them should be
      harmless.
    + Removed `macro`.  It is no longer necessary, since the
      `rawLaTeXBlock` parser will parse macro definitions.  This also avoids
      the need for a separate `latexMacro` parser in the Markdown reader.
    + Use `label` instead of `data-label` for label in caption (#3639).
    + Fixed space after \figurename etc.
    + Resolve references to section numbers.
    + Fix `\let\a=0` case, with single character token.
    + Allow `@` as a letter in control sequences.  `@` is commonly used
      in macros using `\makeatletter`.  Ideally we'd make the tokenizer
      sensitive to `\makeatletter` and `\makeatother`, but until then this
      seems a good change.
    + Track header numbers and correlate with labels.
    + Allow `]` inside group in option brackets (#3857).
    + lstinline with braces can be used (verb cannot be used with braces)
      (Marc Schreiber, #3535).
    + Fix keyval funtion: pandoc did not parse options in braces correctly
      (Marc Schreiber, #3642).
    + When parsing raw LaTeX commands, include trailing space (#1773).
      Otherwise things like `\noindent foo` break and turn into
      `\noindentfoo`.  Affects `-f latex+raw_tex` and `-f markdown` (and other
      formats that allow `raw_tex`).
    + Don't treat "..." as Quoted (#3958).  This caused quotes to be omitted in
      `\texttt` contexts.
    + Add tests for existing `\includegraphics` behaviour (Ben Firshman).
    + Allow space before `=` in bracketd options (Ben Firshman).
    + Be more forgiving in parsing command options.  This was needed, for
      example, to make some minted options work.
    + Strip off quotes in `\include` filenames.

  * Added `Text.Pandoc.CSV`, simple (unexported) CSV parser.

  * `Text.Pandoc.PDF`:

    + Got `--resource-path` working with PDF output (#852).
    + Fetch images when generating PDF via context (#3380).
      To do this, we create the temp directory as a subdirectory
      of the working directory. Since context mk IV by default looks
      for images in the parent directory, this works.
    + Use `report` instead of `warn`, make it sensitive to verbosity settings.
    + Use `fillMediaBag` and `extractMedia` to extract media to temp dir.
      This reduces code duplication.
    + `html2pdf`: use stdin instead of intermediate HTML file
    + Removed useless `TEXINPUTS` stuff for `context2pdf`.  mkiv context
      doesn't use `TEXINPUTS`.

  * `Text.Pandoc.Pretty`:

    + Simplified definition of `realLength`.
    + Don't error for blocks of size < 1.  Instead, resize to 1 (see #1785).

  * `Text.Pandoc.MIME`:

    + Use `application/javascript` (not `application/x-javascript`).
    + Added `emf` to mimeTypes with type `application/x-msmetafile` (#1713).

  * `Text.Pandoc.ImageSize`:

    + Improve SVG image size code (Marc Schreiber, #3580).
    + Make `imageSize` recognize basic SVG dimensions (Mauro Bieg, #3462).

  * Use `Control.Monad.State.Strict` throughout.  This gives 20-30% speedup
    and reduction of memory usage in most of the writers.

  * Use `foldrWithKey` instead of deprecated `foldWithKey`.

  * `Text.Pandoc.SelfContained`:

    + Fixed problem with embedded fonts (#3629).
    + Refactored getData from `getDataURI` in `SelfContained`.
    + Don't use data URIs for script or style (#3423).  Instead, just use
      script or style tags with the content inside.  The old method with
      data URIs prevents certain optimizations outside pandoc.  Exception:
      data URIs are still used when a script contains `</script>` or a
      style contains `</`.
    + SelfContained: Handle URL inside material retrieved from a URL
      (#3629).  This can happen e.g. with an @import of a google web font.
      (What is imported is some CSS which contains an url reference
      to the font itself.) Also, allow unescaped pipe (|) in URL.
    + Load resources from `data-src` (needed for lazy loading in
      reveal.js slide shows).
    + Handle `data-background-image` attribute on section (#3979).

  * `Text.Pandoc.Parsing`:

    + Added `indentWith` (Alexander Krotov, #3687).
    + Added `stateCitations` to `ParserState`.
    + Removed `stateChapters` from `ParserState`.
    + In `ParserState`, make `stateNotes'` a Map, add `stateNoteRefs`.
    + Added `gobbleSpaces` and `gobbleAtMostSpaces`.
    + Adjusted type of `insertIncludedFile` so it can be used with token
      parser.
    + Replace old texmath macro stuff from Parsing.  Use Macro from
      Text.Pandoc.Readers.LaTeX.Types instead.
    + Export `insertIncludedFile`.
    + Added `HasLogMessages`, `logMessage`, `reportLogMessages` (#3447).
    + Replace partial with total function (Albert Krewinkel).
    + Introduce `HasIncludeFiles` type class (Albert Krewinkel).  The
      `insertIncludeFile` function is generalized to work with all parser
      states which are instances of that class.
    + Add `insertIncludedFilesF` which returns F blocks (Albert Krewinkel).
      The `insertIncludeFiles` function was generalized and renamed
      to `insertIncludedFiles'`; the specialized versions are based on that.
    + `many1Till`: Check for the end condition before parsing (Herwig
      Stuetz).  By not checking for the end condition before the first
      parse, the parser was applied too often, consuming too much of the
      input. This only affects `many1Till p end` where `p` matches on a
      prefix of `end`.
    + Provide `parseFromString` (#3690).  This is a verison of
      `parseFromString` specialied to ParserState, which resets
      `stateLastStrPos` at the end.  This is almost always what we want.
      This fixes a bug where `_hi_` wasn't treated as emphasis in the
      following, because pandoc got confused about the position of the
      last word: `- [o] _hi_`.
    + Added `takeP`, `takeWhileP` for efficient parsing of `[Char]`.
    + Fix `blanklines` documentation (Alexander Krotov, #3843).
    + Give less misleading line information with `parseWithString`.
      Previously positions would be reported past the end of the chunk.
      We now reset the source position within the chunk and report
      positions "in chunk."
    + Add `anyLineNewline` (Alexander Krotov).
    + Provide shared F monad functions for Markdown and Org readers
      (Albert Krewinkel).  The `F` monads used for delayed evaluation
      of certain values in the Markdown and Org readers are based on a
      shared data type capturing the common pattern of both `F` types.
    + Add `returnF` (Alexander Krotov).
    + Avoid parsing `Notes:**` as a bare URI (#3570).  This avoids parsing
      bare URIs that start with a scheme + colon + `*`, `_`, or `]`.
    + Added `readerAbbreviations` to `ParserState`.  Markdown reader
      now consults this to determine what is an abbreviation.
    + Combine grid table parsers (Albert Krewinkel, #3638).  The grid table
      parsers for markdown and rst was combined into one single
      parser `gridTable`, slightly changing parsing behavior of both
      parsers: (1) The markdown parser now compactifies block content
      cell-wise: pure text blocks in cells are now treated as paragraphs
      only if the cell contains multiple paragraphs, and as plain blocks
      otherwise. Before, this was true only for single-column tables. (2)
      The rst parser now accepts newlines and multiple blocks in header
      cells.
    + Generalize tableWith, gridTableWith (Albert Krewinkel).
      The parsing functions `tableWith` and `gridTableWith` are generalized
      to work with more parsers. The parser state only has to be an
      instance of the `HasOptions` class instead of requiring a concrete
      type. Block parsers are required to return blocks wrapped into a
      monad, as this makes it possible to use parsers returning results
      wrapped in `Future`s.

  * `Text.Pandoc.Shared`:

    + Simplify `toRomanNumeral` using guards (Alexander Krotov, #3445)
    + `stringify`: handle Quoted better (#3958).  Previously we were losing
      the quotation marks in Quoted elements.

  * `Text.Pandoc.Writers.Shared`:

    + Export `metaToJSON'`, `addVariablesToJSON` (#3439).
      This allows us to add the variables AFTER using the metadata
      to generate a YAML header (in the Markdown writer).
    + Added `unsmartify` (previously in RST writer).
      Undo literal double curly quotes.  Previously we left these.
    + Generalize type of `metaToJSON` so it can take a Text.  Previously a
      String was needed as argument; now any ToJSON instance will do.
    + Added `gridTable` (previously in Markdown writer).
    + `gridTable`: Refactored to use widths in chars.
    + `gridTable`:  remove unnecessary extra space in cells.
    + Fixed `addVariablesToJSON`.  It was previously not allowing multiple
      values to become lists.
    + Pipe tables: impose minimum cell size (see #3526).


  [default template changes]

  * HTML templates (including EPUB and HTML slide show templates):

    + Make default.html5 polyglot markup conformant (John Luke Bentley,
      #3473).  Polyglot markup is HTML5 that is also valid XHTML. See
      <https://www.w3.org/TR/html-polyglot>.  With this change, pandoc's
      html5 writer creates HTML that is both valid HTML5 and valid XHTML.
    + Regularized CSS in html/epub/html slide templates (#3485).
      All templates now include `code{white-space: pre-wrap}`
      and CSS for `q` if `--html-q-tags` is used.  Previously some templates
      had `pre` and others `pre-wrap`; the `q` styles were only sometimes
      included.
    + CSS for `.smallcaps`, (Mauro Bieg, #1592)
    + `default.revealjs`: make `history` default to true.
    + `default.revealjs`: use lazy loading (#2283).
    + `default.revealjs`: add `mathjax` variable and some conditional code
      to use the MathJaX plugin.
    + `default.slidy` uses `https` instead of `http` (ickc, #3848).
    + `default.dzslides`: Load Google Font using HTTPS by default
      (Yoan Blanc).

  * DocBook5 template: Use `lang` and `subtitle` variables (Jens Getreu,
    #3855).

  * LaTeX/Beamer template:

    + Combine LaTeX/Beamer templates (Andrew Dunning, #3878).
      `default.beamer` has been removed; beamer now uses the
      `default.latex` template.  Beamer-specific parts are conditional
      on the `beamer` variable set by the writer. Note that
      `pandoc -D beamer` will return this (combined) template.
    + Use `xcolor` for `colorlinks` option (Andrew Dunning, #3877).
      Beamer loads `xcolor` rather than `color`, and thus the
      `dvipsnames` option doesn't take effect. This also provides a wider
      range of colour selections with the `svgnames` option.
    + Use starred versions of `xcolor` names (Andrew Dunning).
      Prevents changes to documents defined using the `dvipsnames` list (e.g.
      `Blue` gives a different result with svgnames enabled).
    + Load `polyglossia` after header-includes (#3898).  It needs to be
      loaded as late as possible.
    + Use `unicode-math` (Vaclav Haisman).  Use `mathspec` with only
      XeLaTeX on request.
    + Don't load `fontspec` before `unicode-math` (over there).
      The `unicode-math` package loads `fontspec` so explict loading of
      `fontspec` before `unicode-math` is not necessary.
    + Use `unicode-math` by default in default.latex template.  mathspec will
      be used in xelatex if the `mathspec` variable is set; otherwise
      unicode-math will be used (Václav Haisman).
    + Use `dvipsnames` options when `colorlinks` specified (otherwise
      we get an error for `maroon`) (Thomas Hodgson).
    + Added beamer `titlegraphic` and `logo` variables (Thomas Hodgson).
    + Fix typo in fix for notes in tables (#2378, zeeMonkeez).
    + Fix `hyperref` options clash (Andrew Dunning, #3847) Avoids an options
      clash when loading a package (e.g. `tufte-latex`) that uses
      `hyperref` settings different from those in the template.
    + Add `natbiboptions` variable (#3768).
    + Fix links inside captions in LaTeX output with links-as-notes
      (Václav Haisman, #3651).  Declare our redefined `\href` robust.
    + Load `parskip` before `hyperref` (Václav Haisman, #3654).
    + Allow setting Japanese fonts when using LuaLaTeX (Václav Haisman,
      #3873).  by using the `luatexja-fontspec` and `luatexja-preset`
      packages. Use existing `CJKmainfont` and `CJKoptions` template
      variables. Add `luatexjafontspecoptions` for `luatexja-fontspec`
      and `luatexjapresetoptions` for `luatexja-preset`.
    + Added `aspectratio` variable to beamer template (Václav Haisman,
      #3723).
    + Modified template.latex to fix XeLaTex being used with tables
      (lwolfsonkin, #3661).  Reordered `lang` variable handling to
      immediately before `bidi`.

  * ConTeXt template: Improved font handling: `simplefonts` is now
    obsolete in ConTeXt (Pablo Rodríguez).


  [documentation improvements]

  * MANUAL.txt:

    + Add URL for Prince HTML > PDF engine (Ian, #3919).
    + Document that content above slide-level will be omitted in
      slide shows.  See #3460, #2265.
    + Explain `--webtex` SVG url (Mauro Bieg, #3471)
    + Small clarification in YAML metadata section.
    + Document that html4 is technically XHTML 1.0 transitional.
    + Remove refs to highlighting-kate (#3672).
    + Document ibooks specific epub metadata.
    + Clarify that mathml is used for ODT math.
    + Mention limitations of Literate Haskell Support (#3410,
      Joachim Breitner).
    + Add documentation of limitations of grid tables (Stephen
      McDowell, #3864).
    + Clarify that meta-json contains transformed values (Jakob Voß,
      #3491) Make clear that template variable `meta-json` does not
      contain plain text values or JSON output format but field values
      transformed to the selected output format.

  * COPYRIGHT:

    + Clarify that templates are dual-licensed.
    + Clarify that pandoc-types is BSD3 licensed.
    + List new files not written by jgm (Albert Krewinkel).
    + Update dates in copyright notices (Albert Krewinkel).  This follows
      the suggestions given by the FSF for GPL licensed software.
      <https://www.gnu.org/prep/maintain/html_node/Copyright-Notices.html>

  * INSTALL.md:

    + Improved instructions for tests with patterns.
    + Put RPM-based distros on separate point (Mauro Bieg, #3449)

  * CONTRIBUTING.md:

    + Fixed typos (Wandmalfarbe, #3479).
    + Add "ask on pandoc-discuss" (Mauro Bieg).

  * Add lua filter documentation in `doc/lua-filters.md`.  Note that the
    end of this document is autogenerated from `data/pandoc.lua`
    using `make doc/lua-filters.md`, which uses `tools/ldoc.ltp`
    (Albert Krewinkel).

  * Add `doc/filters.md`. This is the old scripting tutorial from
    the website.

  * Add `doc/using-the-pandoc-api.md` (#3289).  This gives an introduction
    to using pandoc as a Haskell library.


  [build infrastructure improvements]

  * Removed `data/templates` submodule.  Templates are now a subtree
    in `data/templates`.  This removes the need to do `git submodule
    update`.

  * Renamed `tests` -> `test`.

  * Remove `https` flag.  Always build with HTTPS support.

  * Use `file-embed` instead of `hsb2hs` to embed data files when
    `embed_data_files` flag is set.  `file-embed` gives us better dependency
    tracking:  if a data file changes, ghc/stack/cabal know to recompile
    the Data module.  This also removes `hsb2hs` as a build dependency.

  * Add `custom-setup` stanza to pandoc, lowercase field names.

  * Add `static` Cabal flag.

  * Name change OSX -> MacOS.  Add a -MacOS suffix to mac package rather
    than -OSX.  Changed local names from osx to macos.

  * make_macos_package.sh - Use strip to reduce executable size.

  * Revised binary linux package.  Now a completely static executable
    is created, using Docker and alpine.  We create both a deb and a
    tarball.  The old `deb` directory has been replaced with a `linux`
    directory.  Running `make` in the `linux` directory should
    perform the build, putting the binary packages in `artifacts/`.

  * `linux/control.in`: add `Replaces:`, so existing pandoc-citeproc and
    pandoc-data packages will be uninstalled; this package provides
    both (#3822).  Add latex packages as 'suggested', update
    description.

  * Remove cpphs build requirement -- it is no longer needed.

  * Replaced `{deb,macos,windows}/stack.yaml` with `stack.pkg.yaml`.

  * Name change OSX -> macOS (ickc, #3869).

  * Fix casing of Linux, UNIX, and Windows (ickc).

  * `.travis.yml`:  create a source dist and do cabal build and test there.
    That way we catch errors due to files missing from the data
    section of pandoc.cabal.

  * Makefile:

    + Split `make haddock` from `make full`.
    + Add BRANCH variable for winpkg.
    + Add `lint` target.
    + Improve `make full`. Disable optimizations.
      Build everything, inc. trypandoc and benchmarks.  Use parallel build.
    + Allow `make test` to take `TESTARGS`.

  * Added new command tests (`Tests.Command`), using small text files
    in `test/command/`.  Any files added in this directory will be treated
    as shell tests (see smart.md for an example).  This makes it very easy
    to add regression tests etc.

  * Test fixes so we can find data files.  In old tests & command tests,
    we now set the environment variable `pandoc_datadir`.  In lua tests,
    we set the datadir explicitly.

  * Refactored `compareOutput` in docx writer test.

  * Consolidated some common functions in `Tests.Helper`.

  * Small change to unbalanced bracket test to speed up test suite.

  * Speed up Native writer quickcheck tests.

  * Use tasty for tests rather than test-framework.

  * Add simple Emacs mode to help with Pandoc templates editing.
    (Václav Haisman, #3889). `tools/pandoc-template-mode.el`


pandoc (1.19.2)

  * Use skylighting library instead of highlighting-kate for syntax
    highlighting. Skylighting is faster and more accurate (#3363).
    Later we'll be able to add features like warning messages, dynamic
    loading of xml syntax definitions, and dynamic loading of themes.

  * Added a new highlight style, `breezeDark`.

  * Text.Pandoc.Highlighting: Update list of `listings` languages (#3374).
    This allows more languages to be used when using the `--listings`
    option.

  * OpenDocument writer:

    + Small refactoring.  Removed separate 'parent' parameter in paraStyle.
    + Don't profilerate text styles unnecessarily (#3371).
      This change makes the writer create only as many temporary
      text styles as are absolutely necessary. It also consolidates
      adjacent nodes with the same style.

  * Org reader (Albert Krewinkel):

    + Allow short hand for single-line raw blocks (Albert Krewinkel,
      #3366).  Single-line raw blocks can be given via `#+FORMAT: raw line`,
      where `FORMAT` must be one of `latex`, `beamer`, `html`, or `texinfo`.
    + Accept org-ref citations followed by commas (Albert Krewinkel).
      Bugfix for an issue which, whenever the citation was immediately
      followed by a comma, prevented correct parsing of org-ref citations.
    + Ensure emphasis markup can be nested.  Nested emphasis markup (e.g.
      `/*strong and emphasized*/`) was interpreted incorrectly in that the
      inner markup was not recognized.
    + Remove pipe char irking the haddock coverage tool (Albert Krewinkel).

  * Docx reader: Empty header should be list of lists (Jesse Rosenthal).
    In the past, the docx reader wrote an empty header as an empty list. It
    should have the same width as a row (and be filled with empty cells).

  * MediaWiki reader:

    + Improved handling of display math (#3362).  Sometimes display math is
      indented with more than one colon.  Previously we handled these cases
      badly, generating definition lists and missing the math.
    + Fix quotation mark parsing (#3336, tgkokk).  Change MediaWiki reader's
      behavior when the smart option is parsed to match other readers'
      behavior.

  * Markdown reader:

    + Fixed `-f markdown_github-hard_line_breaks+escaped_line_breaks`
      (#3341).  Previously this did not properly enable escaped line breaks.
    + Disallow space between inline code and attributes (#3326, #3323,
      Mauro Bieg).

  * DocBook5 writer: make id attribute xml:id, fixes #3329 (#3330, Mauro Bieg).

  * Added some test cases for ODT reader (#3306, #3308, Hubert Plociniczak).

  * LaTeX writer: allow tables with empty cells to count as "plain."
    This addresses a problem of too-wide tables when empty cells
    are used.  Thanks to Joost Kremers for reporting the issue.

  * Org writer: prefix footnote numbers with `fn:` (Albert Krewinkel).
    Unprefixed numbers where used by older org-mode versions, but are no
    longer supported.

  * HTML writer: don't process pars with empty RawInline, (#1040, #3327,
    Mauro Bieg).

  * Markdown writer: Fix display math with `--webtex` (#3298).

  * Fix sample.lua so it properly handles raw blocks/inlines (#3358,
    bumper314).

  * Templates:

    + default.latex: Moved geometry after hyperref (Václav Haisman).
      Otherwise PDF sizes can be wrong in some circumstances.
    + Copied a few changes from default.latex to default.beamer
      (Wandmalfarbe).
    + default.latex, default.beamer: Changed position of `\VerbatimNotes`
      and `fancyvrb`.  This fixes hyperlinks on footnotes in documents
      that contain verbatim in notes (#3361).  (Note: the beamer template
      was updated to match the LaTeX template, but at this point verbatim
      in notes seems not to work in beamer.)
    + default.latex: Allow passing `microtypeoptions` to microtype
      (Václav Haisman).
    + default.latex: Add hyphen option to url package.
    + default.docbook5: Fix namespace declarations (Mauro Bieg).

  * Moved `make_osx_package.sh` to `osx/` directory.

  * Travis continuous integration:

    + Fix false positives with dist build.
    + Speed improvements (Kolen Cheung, #3304, #3357).

  * MANUAL.txt:

    + Clarify that blank space is needed around footnotes (#3352).
    + Fixed typo (#3351, Alexey Rogechev).
    + Note that `--wrap=auto` does not work in HTML output.
    + Default `--columns` width is 72, not 80.
    + Fixed broken links (#3316, Kolen Cheung).
    + Document usage of `@*` in nocite section (#3333, John Muccigrosso).

  * INSTALL.md:

    + Indent code so it's properly formatted (#3335, Bheesham Persaud).
    + Added instructions for extracting binary from OSX, Windows packages.

  * CONTRIBUTING.md: Describe labels currently used in issue tracker
    (Albert Krewinkel).  The labels have changed over time, the list of
    labels is updated to reflect the current set of labels used in the
    issue tracker.

  * Rearrange and extend badges in README (Albert Krewinkel, #3354)

  * Bumped version bounds for dependencies.


pandoc (1.19.1)

  * Set `PANDOC_VERSION` environment variable for filters (#2640).
    This allows filters to check the pandoc version that produced
    the JSON they are receiving.

  * Docx reader: Ensure one-row tables don't have header (#3285,
    Jesse Rosenthal).  Tables in MS Word are set by default to have
    special first-row formatting, which pandoc uses to determine whether
    or not they have a header. This means that one-row tables will, by
    default, have only a header -- which we imagine is not what people
    want. This change ensures that a one-row table is not understood to
    be a header only.  Note that this means that it is impossible to
    produce a header-only table from docx, even though it is legal
    pandoc. But we believe that in nearly all cases, it will be an
    accidental (and unwelcome) result

  * HTML reader:

    + Fixed some bad regressions in HTML table parser (#3280).
      This regression leads to the introduction of empty rows
      in some circumstances.
    + Understand `style=width:` as well as `width` in `col` (#3286).

  * RST reader:

    + Print warnings when keys, substitition, notes not found.
      Previously the parsers failed and we got raw text.  Now we get a
      link with an empty URL, or empty inlines in the case of a note or
      substitution.

    + Fix hyperlink aliases (#3283).

  * Man writer: Ensure that periods are escaped at beginning of line
    (#3270).

  * LaTeX writer: Fix unnumbered headers when used with `--top-level`
    (#3272, Albert Krewinkel).  Fix interaction of top-level divisions
    `part` or `chapter` with unnumbered headers when emitting LaTeX.  Headers
    are ensured to be written using stared commands (like `\subsection*{}`).

  * LaTeX template: use comma not semicolon to separate keywords for
    `pdfkeywords`.  Thanks to Wandmalfarbe.

  * Markdown writer: Fixed incorrect word wrapping (#3277).  Previously pandoc
    would sometimes wrap lines too early due to this bug.

  * Text.Pandoc.Pretty:  Added `afterBreak` [API change].  This makes it
    possible to insert escape codes for content that needs escaping at the
    beginning of a line.

  * Removed old MathMLInHTML.js from 2004, which should no longer
    be needed for MathML with modern browsers.

  * Fixed tests with dynamic linking (#2709).

  * Makefile: Use stack instead of cabal for targets.  This is just
    a convenience for developers.

  * Fixed bash completion of filenames with space (#2749).

  * MANUAL: improved documentation on how to create a custom `reference.docx`.

  * Fix minor spelling typos in the manual (#3273, Anthony Geoghegan)

pandoc (1.19)

  * Changed resolution of filter paths.

    + We now first treat the argument of `--filter` as a full (absolute
      or relative) path, looking for a program there. If it's found, we
      run it.
    + If not, and if it is a simple program name or a relative path, we
      try resolving it relative to `$DATADIR/filters`.
    + If this fails, then we treat it as a program name and look in the
      user's PATH.
    + Removed a hardcoded '/' that may have caused problems with
      Windows paths.

    Previously if you did `--filter foo` and you had `foo` in your path and
    also an executable `foo` in your working directory, the one in the path
    would be used. Now the one in the working directory is used.

    In addition, when you do `--filter foo/bar.hs`, pandoc will now find a
    filter `$DATADIR/filters/foo/bar.hs` -- assuming there isn't a
    `foo/bar.hs` relative to the working directory.

  * Allow `file://` URIs as arguments (#3196). Also improved default reader
    format detection. Previously with a URI ending in .md or .markdown,
    pandoc would assume HTML input. Now it treats these as markdown.

  * Allow to overwrite top-level division type heuristics (#3258,
    Albert Krewinkel). Pandoc uses heuristics to determine the most
    reasonable top-level division type when emitting LaTeX or
    Docbook markup. It is now possible to overwrite this implicitly set
    top-level division via the `top-level-division` command line parameter.

  * Text.Pandoc.Options \[API changes\]:

    + Removed `writerStandalone` field in `WriterOptions`, made
      `writerTemplate` a `Maybe` value. Previously setting
      `writerStandalone = True` did nothing unless a template was provided
      in writerTemplate. Now a fragment will be generated if
      `writerTemplate` is `Nothing`; otherwise, the specified template
      will be used and standalone output generated.
    + `Division` has been renamed `TopLevelDivision` (#3197). The
      `Section`, `Chapter`, and `Part` constructors were renamed to
      `TopLevelSection`, `TopLevelChapter`, and
      `TopLevelPart`, respectively. An additional `TopLevelDefault`
      constructor was added, which is now also the new default value of
      the `writerTopLevelDivision` field in `WriterOptions`.

  * Improved error if they give wrong arg to `--top-level-division`.

  * Use new module from texmath to lookup MS font codepoints in Docx reader.
    Removed unexported module Text.Pandoc.Readers.Docx.Fonts. Its code now
    lives in texmath (0.9).

  * DocBook reader: Fixed xref lookup (#3243). It previously only worked
    when the qnames lacked the docbook namespace URI.

  * HTML reader:

    + Improved table parsing (#3027). We now check explicitly for non-1
      rowspan or colspan attributes, and fail when we encounter them.
      Previously we checked that each row had the same number of cells,
      but that could be true even with rowspans/colspans. And there are
      cases where it isn't true in tables that we can handle fine -- e.g.
      when a tr element is empty. So now we just pad rows with empty cells
      when needed.
    + Treat `<math>` as MathML by default unless something else is
      explicitly specified in xmlns. Provided it parses as MathML,
      of course. Also fixed default which should be to inline math if no
      display attribute is used.
    + Only treat "a" element as link if it has href (#3226). Otherwise
      treat as span.

  * Docx reader (Jesse Rosenthal):

    + Add a placeholder value for CHART. We wrap `[CHART]` in a
      `<span class="chart">`. Note that it maps to inlines because, in
      docx, anything in a drawing tag can be part of a larger paragraph.
    + Be more specific in parsing images We not only want `w:drawing`,
      because that could also include charts. Now we specify
      `w:drawing/pic:pic`. This shouldn't change behavior at all, but it's
      a first step toward allowing other sorts of drawing data as well.
    + Abstract out function to avoid code repetition.
    + Update tests for img title and alt (#3204).
    + Handle Alt text and titles in images. We use the "description" field
      as alt text and the "title" field as title. These can be accessed
      through the "Format Picture" dialog in Word.
    + Docx reader utils: handle empty namespace in `elemName`. Previously,
      if given an empty namespace `(elemName ns "" "foo")` `elemName`
      would output a QName with a `Just ""` namespace. This is never what
      we want. Now we output a `Nothing`. If someone *does* want a
      `Just ""` in the namespace, they can enter the QName
      value explicitly.

  * ODT reader/writer:

    + Inline code when text has a special style (Hubert Plociniczak). When
      a piece of text has a text `Source_Text` then we assume that this is
      a piece of the document that represents a code that needs to
      be inlined. Adapted the writer to also reflect that change.
      Previously it was just writing a 'preformatted' text using a
      non-distinguishable font style. Code blocks are still not recognized
      by the ODT reader. That's a separate issue.
    + Infer table's caption from the paragraph (#3224,
      Hubert Plociniczak). ODT's reader always put empty captions for the
      parsed tables. This commit

        1.  checks paragraphs that follow the table definition
        2.  treats specially a paragraph with a style named 'Table'
        3.  does some postprocessing of the paragraphs that combines tables
            followed immediately by captions

      The ODT writer used the `TableCaption` style for the caption
      paragraph.  This commit follows the OpenOffice approach which allows
      for appending captions to table but uses a built-in style named
      `Table` instead of `TableCaption`. Users of a custom `reference.odt`
      should change the style's name from `TableCaption` to `Table`.

  * ODT reader: Infer tables' header props from rows (#3199,
    Hubert Plociniczak). ODT reader simply provided an empty header list
    which meant that the contents of the whole table, even if not empty, was
    simply ignored. While we still do not infer headers we at least have to
    provide default properties of columns.

  * Markdown reader:

    +   Allow reference link labels starting with `@...` if `citations`
        extension disabled (#3209). Example: in

            \[link text\]\[@a\]

        `link text` isn't hyperlinked because `[@a]` is parsed as
        a citation. Previously this happened whether or not the `citations`
        extension was enabled. Now it happens only if the `citations`
        extension is enabled.
    +   Allow alignments to be specified in Markdown grid tables. For
        example,

            +-------+---------------+--------------------+
            | Right    | Left                 | Centered |
            +=========:+:=================+:=============:+
            | Bananas | $1.34         | built-in wrapper |
            +-------+---------------+--------------------+

    +   Allow Small Caps elements to be created using bracketed spans (as
        they already can be using HTML-syntax spans) (#3191, Kolen Cheung).

  * LaTeX reader:

    + Don't treat `\vspace` and `\hspace` as block commands (#3256).
      Fixed an error which came up, for example, with `\vspace` inside
      a caption. (Captions expect inlines.)
    + Improved table handling. We can now parse all of the tables emitted
      by pandoc in our tests. The only thing we don't get yet are
      alignments and column widths in more complex tables. See #2669.
    + Limited support for minipage.
    + Allow for `[]`s inside LaTeX optional args. Fixes cases like:
    + Handle BVerbatim from fancyvrb (#3203).
    + Handle hungarumlaut (#3201).
    + Allow beamer-style `<...>` options in raw LaTeX (also in Markdown)
      (#3184). This allows use of things like `\only<2,3>{my content}` in
      Markdown that is going to be converted to beamer.

  * Use pre-wrap for code in dzslides template (Nicolas Porcel). Otherwise
    overly long code will appear on every slide.

  * Org reader (Albert Krewinkel):

    +   Respect column width settings (#3246). Table column properties can
        optionally specify a column's width with which it is displayed in
        the buffer. Some exporters, notably the ODT exporter in org-mode
        v9.0, use these values to calculate relative column widths. The org
        reader now implements the same behavior. Note that the org-mode
        LaTeX and HTML exporters in Emacs don't support this feature yet,
        which should be kept in mind by users who use the column
        widths parameters.
    +   Allow HTML attribs on non-figure images (#3222). Images which are
        the only element in a paragraph can still be given HTML attributes,
        even if the image does not have a caption and is hence not a figure.
        The following will add set the `width` attribute of the image to
        `50%`:

            +ATTR\_HTML: :width 50%
            =======================

            \[\[file:image.jpg\]\]

    +   Support `ATTR_HTML` for special blocks (#3182). Special
        blocks (i.e. blocks with unrecognized names) can be prefixed with an
        `ATTR_HTML` block attribute. The attributes defined in that
        meta-directive are added to the `Div` which is used to represent the
        special block.
    +   Support the `todo` export option. The `todo` export option allows to
        toggle the inclusion of TODO keywords in the output. Setting this to
        `nil` causes TODO keywords to be dropped from headlines. The default
        is to include the keywords.
    +   Add support for todo-markers. Headlines can have optional
        todo-markers which can be controlled via the `#+TODO`, `#+SEQ_TODO`,
        or `#+TYP_TODO` meta directive. Multiple such directives can be
        given, each adding a new set of recognized todo-markers. If no
        custom todo-markers are defined, the default `TODO` and `DONE`
        markers are used. Todo-markers are conceptually separate from
        headline text and are hence excluded when autogenerating
        headline IDs. The markers are rendered as spans and labelled with
        two classes: One class is the markers name, the other signals the
        todo-state of the marker (either `todo` or `done`).

  * LaTeX writer:

    + Use `\autocites*` when "suppress-author" citation used.
    + Ensure that simple tables have simple cells (#2666). If cells
      contain more than a single Plain or Para, then we need to set
      nonzero widths and put contents into minipages.
    + Remove invalid inlines in sections (#3218, Hubert Plociniczak).

  * Markdown writer:

    + Fix calculation of column widths for aligned multiline tables
      (#1911, Björn Peemöller). This also fixes excessive CPU and memory
      usage for tables when `--columns` is set in such a way that cells
      must be very tiny. Now cells are guaranteed to be big enough so that
      single words don't need to line break, even if this pushes the line
      length above the column width.
    + Use bracketed form for native spans when `bracketed_spans`
      enabled (#3229).
    + Fixed inconsistent spacing issue (#3232). Previously a tight bullet
      sublist got rendered with a blank line after, while a tight ordered
      sublist did not. Now we don't get the blank line in either case.
    + Fix escaping of spaces in super/subscript (#3225). Previously two
      backslashes were inserted, which gave a literal backslash.
    + Adjust widths in Markdown grid tables so that they match
      on round-trip.

  * Docx writer:

    + Give full detail when there are errors converting tex math.
    + Handle title text in images (Jesse Rosenthal). We already handled
      alt text. This just puts the image "title" into the docx
      "title" attr.
    + Fixed XML markup for empty cells (#3238). Previously the Compact
      style wasn't being applied properly to empty cells.

  * HTML writer:

    + Updated `renderHtml` import from blaze-html.

  * Text.Pandoc.Pretty:

    + Fixed some bugs that caused blank lines in tables (#3251). The bugs
      caused spurious blank lines in grid tables when we had things like
      `blankline $$ blankline`.
    + Add exported function `minOffet` \[API change\] (Björn Peemöller).
    + Added error message for illegal call to `block` (Björn Peemöller).

  * Text.Pandoc.Shared:

    + Put `warn` in MonadIO.
    + `fetchItem`: Better handling of protocol-relative URL (#2635). If
      URL starts with `//` and there is no "base URL" (as there would be
      if a URL were used on the command line), then default to http:.

  * Export Text.Pandoc.getDefaultExtensions \[API change\] (#3178).

  * In --version, trap error in `getAppUserDataDirectory` (#3241). This
    fixes a crash with `pandoc --version` on unusual systems with no real
    user (e.g. SQL Server 2016).

  * Added weigh-pandoc for memory usage diagnostics (#3169).

  * Use correct mime types for woff and woff2 (#3228).

  * Remove make\_travis\_yml.hs (#3235, Kolen Cheung).

  * changelog: Moved an item that was misplaced in the 1.17.2 section to the
    1.18 section where it belongs.

  * CONTRIBUTING.md: minor change in wording and punctuation (#3252,
    Kolen Cheung).

  * Further revisions to manual for `--version` changes (#3244).


pandoc (1.18)

  * Added `--list-input-formats`, `--list-output-formats`,
    `--list-extensions`, `--list-highlight-languages`, and
    `--list-highlight-styles` (#3173).  Removed list of highlighting
    languages from `--version` output.  Removed list of input and output
    formats from default `--help` output.

  * Added `--reference-location=block|section|document` option
    (Jesse Rosenthal).  This determines whether Markdown link references
    and footnotes are placed at the end of the document, the end of the
    section, or the end of the top-level block.

  * Added `--top-level-division=section|chapter|part` (Albert Krewinkel).
    This determines what a level-1 header corresponds to in LaTeX,
    ConTeXt, DocBook, and TEI output.  The default is `section`.
    The `--chapters` option has been deprecated in favor of
    `--top-level-division=chapter`.

  * Added `LineBlock` constructor for `Block` (Albert Krewinkel).  This
    is now used in parsing RST and Markdown line blocks, DocBook
    `linegroup`/`line` combinations, and Org-mode `VERSE` blocks.
    Previously `Para` blocks with hard linebreaks were used.  `LineBlock`s
    are handled specially in the following ouput formats: AsciiDoc
    (as `[verse]` blocks), ConTeXt (`\startlines`/`\endlines`),
    HTML (`div` with a style), Markdown (line blocks if `line_blocks`
    is enabled), Org-mode (`VERSE` blocks), RST (line blocks). In
    other output formats, a paragraph with hard linebreaks is emitted.

  * Allow binary formats to be written to stdout (but not to tty) (#2677).
    Only works on posix, since we use the unix library to check whether
    output is to tty.  On Windows, pandoc works as before and always requires
    an output file parameter for binary formats.

  * Changed JSON output format (Jesse Rosenthal).  Previously we used
    generically generated JSON, but this was subject to change depending
    on the version of aeson pandoc was compiled with.  To ensure stability,
    we switched to using manually written ToJSON and FromJSON
    instances, and encoding the API version.  **Note:**  pandoc filter
    libraries will need to be revised to handle the format change.
    Here is a summary of the essential changes:

    + The toplevel JSON format is now `{"pandoc-api-version" :
      [MAJ, MIN, REV], "meta" : META, "blocks": BLOCKS}`
      instead of `[{"unMeta": META}, [BLOCKS]]`.
      Decoding fails if the major and minor version numbers don't
      match.
    + Leaf nodes no longer have an empty array for their "c" value.
      Thus, for example, a `Space` is encoded as `{"t":"Space"}`
      rather than `{"t":"Space","c":[]}` as before.

  * Removed `tests/Tests/Arbitrary.hs` and added a `Text.Pandoc.Arbitrary`
    module to pandoc-types (Jesse Rosenthal).  This makes it easier
    to use QuickCheck with pandoc types outside of pandoc itself.

  * Add `bracketed_spans` Markdown extension, enabled by default
    in pandoc `markdown`.  This allows you to create a native span
    using this syntax:  `[Here is my span]{#id .class key="val"}`.

  * Added `angle_brackets_escapable` Markdown extension (#2846).
    This is needed because github flavored Markdown has a slightly
    different set of escapable symbols than original Markdown;
    it includes angle brackets.

  * Export `Text.Pandoc.Error` in `Text.Pandoc` [API change].

  * Print highlighting-kate version in `--version`.

  * `Text.Pandoc.Options`:

    + `Extension` has new constructors `Ext_brackted_spans` and
      `Ext_angle_brackets_escapable` [API change].
    + Added `ReferenceLocation` type [API change] (Jesse Rosenthal).
    + Added `writerReferenceLocation` field to `WriterOptions` (Jesse
      Rosenthal).

  * `--filter`:  we now check `$DATADIR/filters` for filters before
    looking in the path (#3127, Jesse Rosenthal, thanks to Jakob
    Voß for the idea).  Filters placed in this directory need not
    be executable; if the extension is `.hs`, `.php`, `.pl`, `.js`,
    or `.rb`, pandoc will run the right interpreter.

  * For `--webtex`, replace deprecated Google Chart API by CodeCogs as
    default (Kolen Cheung).

  * Removed `raw_tex` extension from `markdown_mmd` defaults (Kolen Cheung).

  * Execute .js filters with node (Jakob Voß).

  * Textile reader:

    + Support `bc..` extended code blocks (#3037).  Also, remove trailing
      newline in code blocks (consistently with Markdown reader).
    + Improve table parsing.  We now handle cell and row attributes, mostly
      by skipping them.  However, alignments are now handled properly.
      Since in pandoc alignment is per-column, not per-cell, we
      try to devine column alignments from cell alignments.
      Table captions are also now parsed, and textile indicators
      for thead and tfoot no longer cause parse failure.  (However,
      a row designated as tfoot will just be a regular row in pandoc.)
    + Improve definition list parsing.  We now allow multiple terms
      (which we concatenate with linebreaks).  An exponential parsing
      bug (#3020) is also fixed.
    + Disallow empty URL in explicit link (#3036).

  * RST reader:

    + Use Div instead of BlockQuote for admonitions (#3031).
      The Div has class `admonition` and (if relevant) one of the
      following:  `attention`, `caution`, `danger`, `error`, `hint`,
      `important`, `note`, `tip`, `warning`.  **Note:** This will change
      the rendering of some RST documents!  The word ("Warning", "Attention",
      etc.) is no longer added; that must be done with CSS or a filter.
    + A Div is now used for `sidebar` as well.
    + Skip whitespace before note (Jesse Rosenthal, #3163).  RST requires a
      space before a footnote marker. We discard those spaces so that footnotes
      will be adjacent to the text that comes before it. This is in line with
      what rst2latex does.
    + Allow empty lines when parsing line blocks (Albert Krewinkel).

  * Markdown reader:

    + Allow empty lines when parsing line blocks (Albert Krewinkel).
    + Allow attributes on autolinks (#3183, Daniele D'Orazio).

  * LaTeX reader:

    + More robust parsing of unknown environments (#3026).
      We no longer fail on things like `^` inside options for tikz.
    + Be more forgiving of non-standard characters, e.g. `^` outside of math.
      Some custom environments give these a meaning, so we should try not to
      fall over when we encounter them.
    + Drop duplicate `*` in bibtexKeyChars (Albert Krewinkel)

  * MediaWiki reader:

    + Fix for unquoted attribute values in mediawiki tables (#3053).
      Previously an unquoted attribute value in a table row
      could cause parsing problems.
    + Improved treatment of verbatim constructions (#3055).
      Previously these yielded strings of alternating Code and Space
      elements; we now incorporate the spaces into the Code.  Emphasis
      etc. is still possible inside these.
    + Properly interpret XML tags in pre environments (#3042).  They are meant
      to be interpreted as literal text.

  * EPUB reader:  don't add root path to data: URIs (#3150).
    Thanks to @lep for the bug report and patch.

  * Org reader (Albert Krewinkel):

    + Preserve indentation of verse lines (#3064).  Leading spaces in verse
      lines are converted to non-breaking spaces, so indentation is preserved.
    + Ensure image sources are proper links.  Image sources as those in plain
      images, image links, or figures, must be proper URIs or relative file
      paths to be recognized as images.  This restriction is now enforced
      for all image sources.  This also fixes the reader's usage of uncleaned
      image sources, leading to `file:` prefixes not being deleted from
      figure images.  Thanks to @bsag for noticing this bug.
    + Trim verse lines properly (Albert Krewinkel).
    + Extract meta parsing code to module.  Parsing of meta-data is well
      separable from other block parsing tasks.  Moving into new module to
      get small files and clearly arranged code.
    + Read markup only for special meta keys.  Most meta-keys should be read
      as normal string values, only a few are interpreted as marked-up text.
    + Allow multiple, comma-separated authors.  Multiple authors can be
      specified in the `#+AUTHOR` meta line if they are given as a
      comma-separated list.
    + Give precedence to later meta lines.  The last meta-line of any given
      type is the significant line.  Previously the value of the first line
      was kept, even if more lines of the same type were encounterd.
    + Read LaTeX_header as header-includes.  LaTeX-specific header commands
      can be defined in `#+LaTeX_header` lines.  They are parsed as
      format-specific inlines to ensure that they will only show up in LaTeX
      output.
    + Set documentclass meta from LaTeX_class.
    + Set classoption meta from LaTeX_class_options.
    + Read HTML_head as header-includes.  HTML-specific head content can be
      defined in `#+HTML_head` lines.  They are parsed as format-specific
      inlines to ensure that they will only show up in HTML output.
    + Respect `author` export option.  The `author` option controls whether
      the author should be included in the final markup.  Setting
      `#+OPTIONS: author:nil` will drop the author from the final meta-data
      output.
    + Respect `email` export option.  The `email` option controls whether the
      email meta-field should be included in the final markup. Setting
      `#+OPTIONS: email:nil` will drop the email field from the final
      meta-data output.
    + Respect `creator` export option.  The `creator` option controls whether
      the creator meta-field should be included in the final markup.  Setting
      `#+OPTIONS: creator:nil` will drop the creator field from the final
      meta-data output.  Org-mode recognizes the special value `comment` for
      this field, causing the creator to be included in a comment.  This is
      difficult to translate to Pandoc internals and is hence interpreted the
      same as other truish values (i.e. the meta field is kept if it's
      present).
    + Respect unnumbered header property (#3095).  Sections the `unnumbered`
      property should, as the name implies, be excluded from the automatic
      numbering of section provided by some output formats.  The Pandoc
      convention for this is to add an "unnumbered" class to the header.  The
      reader treats properties as key-value pairs per default, so a special
      case is added to translate the above property to a class instead.
    + Allow figure with empty caption (Albert Krewinkel, #3161).
      A `#+CAPTION` attribute before an image is enough to turn an image into
      a figure. This wasn't the case because the `parseFromString` function,
      which processes the caption value, would fail on empty values. Adding
      a newline character to the caption value fixes this.

  * Docx reader:

    + Use XML convenience functions (Jesse Rosenthal).
      The functions `isElem` and `elemName` (defined in Docx/Util.hs) make
      the code a lot cleaner than the original XML.Light functions, but they
      had been used inconsistently. This puts them in wherever applicable.
    + Handle anchor spans with content in headers.  Previously, we would only
      be able to figure out internal links to a header in a docx if the
      anchor span was empty. We change that to read the inlines out of the
      first anchor span in a header.
    + Let headers use exisiting id.  Previously we always generated an id for
      headers (since they wouldn't bring one from Docx). Now we let it use an
      existing one if possible. This should allow us to recurs through anchor
      spans.
    + Use all anchor spans for header ids.  Previously we only used the first
      anchor span to affect header ids. This allows us to use all the anchor
      spans in a header, whether they're nested or not (#3088).
    + Test for nested anchor spans in header.  This ensures that anchor spans
      in header with content (or with other anchor spans inside) will resolve
      to links to a header id properly.

  * ODT reader (Hubert Plociniczak)

    + Include list's starting value.  Previously the starting value of
      the lists' items has been hardcoded to 1. In reality ODT's list
      style definition can provide a new starting value in one of its
      attributes.
    + Infer caption from the text following the image.
      Frame can contain other frames with the text boxes.
    + Add `fig:` to title for Image with a caption (as expected
      by pandoc's writers).
    + Basic support for images in ODT documents.
    + Don't duplicate text for anchors (#3143).  When creating an anchor
      element we were adding its representation as well as the original
      content, leading to text duplication.

  * DocBook writer:

    + Include an anchor element when a div or span has an id (#3102).
      Note that DocBook does not have a class attribute, but at least this
      provides an anchor for internal links.

  * LaTeX writer:

    + Don't use * for unnumbered paragraph, subparagraph.  The starred
      variants don't exist.  This helps with part of #3058...it gets rid of
      the spurious `*`s.  But we still have numbers on the 4th and 5th level
      headers.
    + Properly escape backticks in verbatim (#3121, Jesse Rosenthal).
      Otherwise they can cause unintended ligatures like `` ?` ``.
    + Handle NARRAOW NO-BREAK SPACE into LaTeX (Vaclav Zeman) as `\,`.
    + Don't include `[htbp]` placement for figures (#3103, Václav Haisman).
      This allows figure placement defaults to be changed by the user
      in the template.

  * HTML writer (slide show formats): In slide shows, don't change slide title
    to level 1 header (#2221).

  * TEI writer: remove heuristic to detect book template (Albert Krewinkel).
    TEI doesn't have `<book>` elements but only generic `<divN>` division
    elements. Checking the template for a trailing `</book>` is nonsensical.

  * MediaWiki writer:  transform filename with underscores in images (#3052).
    `foo bar.jpg` becomes `foo_bar.jpg`. This was already done
    for internal links, but it also needs to happen for images.

  * ICML writer: replace partial function (!!) in table handling (#3175,
    Mauro Bieg).

  * Man writer: allow section numbers that are not a single digit (#3089).

  * AsciiDoc writer: avoid unnecessary use of "unconstrained" emphasis
    (#3068).  In AsciiDoc, you must use a special form of emphasis
    (double `__`) for intraword emphasis.  Pandoc was previously using
    this more than necessary.

  * EPUB writer:  use stringify instead of plain writer for metadata
    (#3066).  This means that underscores won't be used for emphasis,
    or CAPS for bold.  The metadata fields will just have unadorned
    text.

  * Docx Writer:

    + Implement user-defined styles (Jesse Rosenthal).  Divs and Spans
      with a `custom-style` key in the attributes will apply the corresponding
      key to the contained blocks or inlines.
    + Add ReaderT env to the docx writer (Jesse Rosenthal).
    + Clean up and streamline RTL behavior (Jesse Rosenthal, #3140).
      You can set `dir: rtl` in YAML metadata, or use `-M dir=rtl`
      on the command line.  For finer-grained control, you can set
      the `dir` attribute in Div or Span elements.

  * Org writer (Albert Krewinkel):

    + Remove blank line after figure caption.  Org-mode only treats an image
      as a figure if it is directly preceded by a caption.
    + Ensure blank line after figure.  An Org-mode figure should be surrounded
      by blank lines.  The figure would be recognized regardless, but images
      in the following line would unintentionally be treated as figures as
      well.
    + Ensure link targets are paths or URLs.  Org-mode treats links as
      document internal searches unless the link target looks like a URL or
      file path, either relative or absolute.  This change ensures that this
      is always the case.
    + Translate language identifiers.  Pandoc and Org-mode use different
      programming language identifiers.  An additional translation between
      those identifiers is added to avoid unexpected behavior.  This fixes a
      problem where language specific source code would sometimes be output
      as example code.
    + Drop space before footnote markers (Albert Krewinkel, #3162).
      The writer no longer adds an extra space before footnote markers.

  * Markdown writer:

    + Don't emit HTML for tables unless `raw_html` extension is set (#3154).
      Emit `[TABLE]` if no suitable table formats are enabled and raw HTML
      is disabled.
    + Check for the `raw_html` extension before emiting a raw HTML block.
    + Abstract out note/ref function (Jesse Rosenthal).
    + Add ReaderT monad for environment variables (Jesse Rosenthal).

  * HTML, EPUB, slidy, revealjs templates: Use `<p>` instead of `<h1>` for
    subtitle, author, date (#3119).  Note that, as a result of this change,
    authors may need to update CSS.

  * revealjs template:  Added `notes-server` option
    (jgm/pandoc-templates#212, Yoan Blanc).

  * Beamer template:

    + Restore whitespace between paragraphs. This was
      a regression in the last release (jgm/pandoc-templates#207).
    + Added `themeoptions` variable (Carsten Gips).
    + Added `beamerarticle` variable.  This causes the `beamerarticle`
      package to be loaded in beamer, to produce an article from beamer
      slides.  (Carsten Gips)
    + Added support for `fontfamilies` structured variable
      (Artem Klevtsov).
    + Added hypersetup options (Jake Zimmerman).

  * LaTeX template:

    + Added dummy definition for `\institute`.
      This isn't a standard command, and we want to avoid a crash when
      `institute` is used with the default template.
    + Define default figure placement (Václav Haisman), since pandoc
      no longer includes `[htbp]` for figures.  Users with custom templates
      will want to add this. See #3103.
    + Use footnote package to fix notes in tables (jgm/pandoc-templates#208,
      Václav Haisman).

  * Moved template compiling/rendering code to a separate library.
    `doctemplates`.  This allows the pandoc templating system to be
    used independently.

  * Text.Pandoc.Error: Fix out of index error in `handleError`
    (Matthew Pickering).  The fix is to not try to show the exact line when
    it would cause an out-of-bounds error as a result of included files.

  * Text.Pandoc.Shared: Add `linesToBlock` function (Albert Krewinkel).

  * Text.Pandoc.Parsing.emailAddress: tighten up parsing of email
    addresses.  Technically `**@user` is a valid email address, but if we
    allow things like this, we get bad results in markdown flavors
    that autolink raw email addresses (see #2940).  So we exclude a few
    valid email addresses in order to avoid these more common bad cases.

  * Text.Pandoc.PDF:  Don't crash with nonexistent image (#3100).  Instead,
    emit the alt text, emphasized.  This accords with what the ODT writer
    currently does.  The user will still get a warning about a nonexistent
    image.

  * Fix example in API documentation (#3176, Thomas Weißschuh).

  * Tell where to get tarball in INSTALL (#3062).

  * Rename README to MANUAL.txt and add GitHub-friendly README.md
    (Albert Krewinkel, Kolen Cheung).

  * Replace COPYING with Markdown version COPYING.md from GNU (Kolen Cheung).

  * MANUAL.txt:

    + Put note on structured vars in separate paragraph (#2148, Albert
      Krewinkel).  Make it clearer that structured author variables require a
      custom template
    + Note that `--katex` works best with `html5` (#3077).
    + Fix the LaTeX and EPUB links in manual (Morton Fox).
    + Document `biblio-title` variable.

  * Improve spacing of footnotes in `--help` output (Waldir Pimenta).

  * Update KaTeX to v0.6.0 (Kolen Cheung).

  * Allow latest dependencies.

  * Use texmath 0.8.6.6 (#3040).

  * Allow http-client 0.4.30, which is the version in stackage lts.
    Previously we required 0.5.
    Remove CPP conditionals for earlier versions.

  * Remove support for GHC < 7.8 (Jesse Rosenthal).

    + Remove Compat.Monoid.
    + Remove an inline monad compatibility macro.
    + Remove Text.Pandoc.Compat.Except.
    + Remove directory compat.
    + Change constraint on mtl.
    + Remove unnecessary CPP condition in UTF8.
    + Bump base lower bound to 4.7.
    + Remove 7.6 build from .travis.yaml.
    + Bump supported ghc version in CONTRIBUTING.md.
    + Add note about GHC version support to INSTALL.
    + Remove GHC 7.6 from list of tested versions (Albert Krewinkel).
    + Remove TagSoup compat.
    + Add EOL note to time compat module.  Because time 1.4 is a boot library
      for GHC 7.8, we will support the compatibility module as long as we
      support 7.8. But we should be clear about when we will no longer need
      it.
    + Remove blaze-html CPP conditional.
    + Remove unnecessary CPP in custom Prelude.

pandoc (1.17.2)

  * Added Zim Wiki writer, template and tests. `zimwiki` is now
    a valid output format. (Alex Ivkin)

  * Changed email-obfuscation default to no obfuscation (#2988).
    + `writerEmailObfuscation` in `defaultWriterOptions` is now
      `NoObfuscation`.
    + the default for the command-line `--email-obfuscation` option is
      now `none`.

  * Docbook writer: Declare xlink namespace in Docbook5 output (Ivo Clarysse).

  * Org writer:

    + Support arbitrary raw inlines (Albert Krewinkel).
      Org mode allows arbitrary raw inlines ("export snippets" in Emacs
      parlance) to be included as `@@format:raw foreign format text@@`.
    + Improve Div handling (Albert Krewinkel).  Div blocks handling is
      changed to make the output look more like idiomatic org mode:
        - Div-wrapped content is output as-is if the div's attribute is the
          null attribute.
        - Div containers with an id but neither classes nor key-value pairs
          are unwrapped and the id is added as an anchor.
        - Divs with classes associated with greater block elements are
          wrapped in a `#+BEGIN`...`#+END` block.
        - The old behavior for Divs with more complex attributes is kept.

  * HTML writer:

    + Better support for raw LaTeX environments (#2758).
      Previously we just passed all raw TeX through when MathJax was used for
      HTML math.  This passed through too much.  With this patch, only raw
      LaTeX environments that MathJax can handle get passed through.
      This patch also causes raw LaTeX environments to be treated
      as math, when possible, with MathML and WebTeX output.

  * Markdown writer:  use raw HTML for simple, pipe tables with linebreaks
    (#2993).  Markdown line breaks involve a newline, and simple and pipe
    tables can't contain one.

  * Make --webtex work with the Markdown writer (#1177).
    This is a convenient option for people using
    websites whose Markdown flavors don't provide for math.

  * Docx writer:

    + Set paragraph to FirstPara after display math (Jesse Rosenthal).
      We treat display math like block quotes, and apply FirstParagraph style
      to paragraphs that follow them. These can be styled as the user
      wishes. (But, when the user is using indentation, this allows for
      paragraphs to continue after display math without indentation.)
    + Use actual creation time as doc prop (Jesse Rosenthal).
      Previously, we had used the user-supplied date, if available, for Word's
      document creation metadata. This could lead to weird results, as in
      cases where the user post-dates a document (so the modification might be
      prior to the creation). Here we use the actual computer time to set the
      document creation.

  * LaTeX writer:

    + Don't URI-escape image source (#2825).  Usually this is a local file,
      and replacing spaces with `%20` ruins things.
    + Allow 'standout' as a beamer frame option (#3007).
      `## Slide title {.standout}`.

  * RST reader: Fixed links with no explicit link text.  The link
    `` `<foo>`_ `` should have `foo` as both its link text and its URL.
    See RST spec at <http://docutils.sourceforge.net/docs/ref/rst/restructuredtext.html#embedded-uris-and-aliases>
    Closes Debian #828167 -- reported by Christian Heller.

  * Textile reader:

    + Fixed attributes (#2984).  Attributes can't be followed by
      a space.  So, `_(class)emph_` but `_(noclass) emph_`.
    + Fixed exponential parsing bug (#3020).
    + Fix overly aggressive interpretation as images (#2998).
      Spaces are not allowed in the image URL in textile.

  * LaTeX reader:

    + Fix `\cite` so it is a NormalCitation not AuthorInText.
    + Strip off double quotes around image source if present (#2825).
      Avoids interpreting these as part of the literal filename.

  * Org reader:

    + Add semicolon to list of special chars (Albert Krewinkel)
      Semicolons are used as special characters in citations syntax.  This
      ensures the correct parsing of Pandoc-style citations: `[prefix; @key;
      suffix]`.  Previously, parsing would have failed unless there was a space
      or other special character as the last <prefix> character.
    + Add support for "Berkeley-style" cites (Albert Krewinkel, #1978).
      A specification for an official Org-mode citation syntax was drafted by
      Richard Lawrence and enhanced with the help of others on the orgmode
      mailing list.  Basic support for this citation style is added to the
      reader.
    + Support arbitrary raw inlines (Albert Krewinkel).
      Org mode allows arbitrary raw inlines ("export snippets" in Emacs
      parlance) to be included as `@@format:raw foreign format text@@`.
    + Remove partial functions (Albert Krewinkel, #2991).
      Partial functions like `head` lead to avoidable errors and should be
      avoided.  They are replaced with total functions.
    + Support figure labels (Albert Krewinkel, #2496, #2999).
      Figure labels given as `#+LABEL: thelabel` are used as the ID of the
      respective image.  This allows e.g. the LaTeX to add proper `\label`
      markup.
    + Improve tag and properties type safety (Albert Krewinkel).
      Specific newtype definitions are used to replace stringly typing of tags
      and properties.  Type safety is increased while readability is improved.
    + Parse as headlines, convert to blocks (Albert Krewinkel).
      Emacs org-mode is based on outline-mode, which treats documents as trees
      with headlines are nodes.  The reader is refactored to parse into a
      similar tree structure.  This simplifies transformations acting on
      document (sub-)trees.
    * Refactor comment tree handling (Albert Krewinkel).
      Comment trees were handled after parsing, as pattern matching on lists
      is easier than matching on sequences.  The new method of reading
      documents as trees allows for more elegant subtree removal.
    * Support archived trees export options (Albert Krewinkel).
      Handling of archived trees can be modified using the `arch` option.
      Archived trees are either dropped, exported completely, or collapsed to
      include just the header when the `arch` option is nil, non-nil, or
      `headline`, respectively.
    * Put export setting parser into module (Albert Krewinkel).
      Export option parsing is distinct enough from general block parsing to
      justify putting it into a separate module.
    * Support headline levels export setting (Albert Krewinkel).
      The depths of headlines can be modified using the `H` option.  Deeper
      headlines will be converted to lists.
    * Replace ugly code with view pattern (Albert Krewinkel).
      Some less-than-smart code required a pragma switching of overlapping
      pattern warnings in order to compile seamlessly.  Using view patterns
      makes the code easier to read and also doesn't require overlapping
      pattern checks to be disabled.
    * Fix parsing of verbatim inlines (Albert Krewinkel, #3016).
      Org rules for allowed characters before or after markup chars were not
      checked for verbatim text.  This resultet in wrong parsing outcomes of
      if the verbatim text contained e.g. space enclosed markup characters as
      part of the text (`=is_substr = True=`).  Forcing the parser to update
      the positions of allowed/forbidden markup border characters fixes this.

  * LaTeX template: fix for obscure hyperref/xelatex issue.
    Here's a minimal case:

        \documentclass[]{article}
        \usepackage{hyperref}
        \begin{document}
        \section{\%á}
        \end{document}

    Without this change, this fails on the second invocation of xelatex.
    This affects inputs this like `# %á` with pdf output via xelatex.

  * trypandoc:  call results 'html' instead of 'result'.
    This is for better compatibility with babelmark2.

  * Document MultiMarkdown as input/output format (Albert Krewinkel, #2973).
    MultiMarkdown was only mentioned as a supported Markdown dialect but not
    as a possible input or output format.  A brief mention is added
    everywhere the other supported markdown dialects are mentioned.

  * Document Org mode as a format containing raw HTML (Albert Krewinkel)
    Raw HTML is kept when the output format is Emacs Org mode.

  * Implement `RawInline` and `RawBlock` in sample lua custom writer (#2985).

  * Text.Pandoc.Shared:

    + Introduce blocksToInlines function (Jesse Rosenthal).
      This is a lossy function for converting `[Block] -> [Inline]`. Its main
      use, at the moment, is for docx comments, which can contain arbitrary
      blocks (except for footnotes), but which will be converted to spans.
      This is, at the moment, pretty useless for everything but the basic
      `Para` and `Plain` comments. It can be improved, but the docx reader
      should probably emit a warning if the comment contains more than this.
    + Add BlockQuote to blocksToInlines (Jesse Rosenthal).
    + Add further formats for `normalizeDate` (Jesse Rosenthal).
      We want to avoid illegal dates -- in particular years with greater than
      four digits. We attempt to parse series of digits first as `%Y%m%d`, then
      `%Y%m`, and finally `%Y`.
    + `normalizeDate` should reject illegal years (Jesse Rosenthal).
      We only allow years between 1601 and 9999, inclusive. The ISO 8601
      actually says that years are supposed to start with 1583, but MS Word
      only allows 1601-9999. This should stop corrupted word files if the date
      is out of that range, or is parsed incorrectly.
    + Improve year sanity check in normalizeDate (Jesse Rosenthal).
      Previously we parsed a list of dates, took the first one, and then
      tested its year range. That meant that if the first one failed, we
      returned nothing, regardless of what the others did. Now we test for
      sanity before running `msum` over the list of Maybe values. Anything
      failing the test will be Nothing, so will not be a candidate.

  * Docx reader:

    + Add simple comment functionality. (Jesse Rosenthal).
      This adds simple track-changes comment parsing to the docx reader. It is
      turned on with `--track-changes=all`. All comments are converted to
      inlines, which can list some information. In the future a warning will be
      added for comments with formatting that seems like it will be excessively
      denatured.  Note that comments can extend across blocks. For that reason
      there are two spans: `comment-start` and `comment-end`.  `comment-start`
      will contain the comment. `comment-end` will always be empty.  The two
      will be associated by a numeric id.
    + Enable warnings in top-level reader (Jesse Rosenthal).
      Previously we had only allowed for warnings in the parser. Now we allow
      for them in the `Docx.hs` as well. The warnings are simply concatenated.
    + Add warning for advanced comment formatting. (Jesse Rosenthal).
      We can't guarantee we'll convert every comment correctly, though we'll
      do the best we can. This warns if the comment includes something other
      than Para or Plain.
    + Add tests for warnings. (Jesse Rosenthal).
    + Add tests for comments (Jesse Rosenthal).
      We test for comments, using all track-changes options. Note that we
      should only output comments if `--track-changes=all`. We also test for
      emitting warnings if there is complicated formatting.

  * README: update to include track-changes comments. (Jesse Rosenthal)

  * Improved Windows installer - don't ignore properties set on command-line.
    See #2708.  Needs testing to see if this resolves the issue.
    Thanks to @nkalvi.

  * Process markdown extensions on command line in L->R order (#2995).
    Previously they were processed, very unintuitively, in R->L
    order, so that `markdown-tex_math_dollars+tex_math_dollars`
    had `tex_math_dollars` disabled.

  * Added `secnumdepth` variable to LaTeX template (#2920).

  * Include table of contents in README.html in Windows package.

  * Writers: treat SoftBreak as space for stripping (Jesse Rosenthal)
    In Writers.Shared, we strip leading and trailing spaces for display
    math. Since SoftBreak's are treated as spaces, we should strip those
    too.

  * beamer, latex templates:  pass biblatexoptions directly in package load.
    This allows runtime optinos to be used.  Fixes jgm/pandoc-citeproc#201

  * CPP workaround for deprecation of `parseUrl` in http-client.

  * Removed some redundant class constraints.

  * make_oxs_package.sh - use OSX env variable.

  * Replaced INSTALL with INSTALL.md, incorporating INSTALL and the
    old installing page from website.

  * Added `winpkg` target to Makefile.  This downloads the windows package
    from appveyor and signs it using the key.

  * Document Org mode as a format containing raw TeX (Albert Krewinkel).
    Raw TeX is kept verbatim when the output format is Emacs Org mode.

  * Support math with haddock-library >= 1.4.

  * Removed `-rtsopts` from library stanza.  It has no effect, and Hackage
    wouldn't accept the package.

  * Update library dependency versions.

pandoc (1.17.1)

  * New output format: `docbook5` (Ivo Clarysse).

  * `Text.Pandoc.Options`: Add `writerDocBook5` to `WriterOptions`
    (API change).

  * Org writer:

    + Add :PROPERTIES: drawer support (Albert Krewinkel, #1962).
      This allows header attributes to be added to org documents in the form
      of `:PROPERTIES:` drawers.  All available attributes are stored as
      key/value pairs.  This reflects the way the org reader handles
      `:PROPERTIES:` blocks.
    + Add drawer capability (Carlos Sosa).  For the implementation of the
      Drawer element in the Org Writer, we make use of a generic Block
      container with attributes.  The presence of a `drawer` class defines
      that the `Div` constructor is a drawer. The first class defines the
      drawer name to use.  The key-value list in the attributes defines
      the keys to add inside the Drawer.  Lastly, the list of Block elements
      contains miscellaneous blocks elements to add inside of the Drawer.
    + Use `CUSTOM_ID` in properties (Albert Krewinkel).  The `ID` property is
      reserved for internal use by Org-mode and should not be used.
      The `CUSTOM_ID` property is to be used instead, it is converted to the
      `ID` property for certain export format.

  * LaTeX writer:

    + Ignore `--incremental` unless output format is beamer (#2843).
    + Fix polyglossia to babel env mapping (Mauro Bieg, #2728).
      Allow for optional argument in square brackets.
    + Recognize `la-x-classic` as Classical Latin (Andrew Dunning).
      This allows one to access the hyphenation patterns in CTAN's
      hyph-utf8.
    + Add missing languages from hyph-utf8 (Andrew Dunning).
    + Improve use of `\strut` with `\minipage` inside tables
      (Jose Luis Duran).  This improves spacing in multiline
      tables.
    + Use `{}` around options containing special chars (#2892).
    + Avoid lazy `foldl`.
    + Don't escape underscore in labels (#2921).  Previously they were
      escaped as `ux5f`.
    + brazilian -> brazil for polyglossia (#2953).

  * HTML writer: Ensure mathjax link is added when math appears in footnote
    (#2881).  Previously if a document only had math in a footnote, the
    MathJax link would not be added.

  * EPUB writer: set `navpage` variable on nav page.
    This allows templates to treat it differently.

  * DocBook writer:

    + Use docbook5 if `writerDocbook5` is set (Ivo Clarysse).
    + Properly handle `ulink`/`link` (Ivo Clarysse).

  * EPUB reader:

    + Unescape URIs in spine (#2924).
    + EPUB reader:  normalise link id (Mauro Bieg).

  * Docx Reader:

    + Parse `moveTo` and `moveFrom` (Jesse Rosenthal).
      `moveTo` and `moveFrom` are track-changes tags that are used when a
      block of text is moved in the document. We now recognize these tags and
      treat them the same as `insert` and `delete`, respectively. So,
      `--track-changes=accept` will show the moved version, while
      `--track-changes=reject` will show the original version.
    + Tests for track-changes moving (Jesse Rosenthal).

  * ODT, EPUB, Docx readers: throw `PandocError` on unzip failure
    (Jesse Rosenthal) Previously, `readDocx`, `readEPUB`, and `readOdt`
    would error out if zip-archive failed. We change the archive extraction
    step from `toArchive` to `toArchiveOrFail`, which returns an Either value.

  * Markdown, HTML readers:  be more forgiving about unescaped `&` in
    HTML (#2410).  We are now more forgiving about parsing invalid HTML with
    unescaped `&` as raw HTML.  (Previously any unescaped `&`
    would cause pandoc not to recognize the string as raw HTML.)

  * Markdown reader:

    + Added bracket syntax for native spans (#168).
    + Fix pandoc title blocks with lines ending in 2 spaces (#2799).
    + Added `-s` to markdown-reader-more test.

  * HTML reader: fixed bug in `pClose`.  This caused exponential parsing
    behavior in documnets with unclosed tags in `dl`, `dd`, `dt`.

  * MediaWiki reader: Allow spaces before `!` in MediaWiki table header
    (roblabla).

  * RST reader: Support `:class:` option for code block in RST reader
    (Sidharth Kapur).

  * Org reader (all Albert Krewinkel, except where noted otherwise):

    + Stop padding short table rows.
      Emacs Org-mode doesn't add any padding to table rows.  The first
      row (header or first body row) is used to determine the column count,
      no other magic is performed.
    + Refactor rows-to-table conversion.  This refactors
      the codes conversing a list table lines to an org table ADT.
      The old code was simplified and is now slightly less ugly.
    + Fix handling of empty table cells, rows (Albert Krewinkel, #2616).
      This fixes Org mode parsing of some corner cases regarding empty cells
      and rows.  Empty cells weren't parsed correctly, e.g. `|||` should be
      two empty cells, but would be parsed as a single cell containing a pipe
      character.  Empty rows where parsed as alignment rows and dropped from
      the output.
    + Fix spacing after LaTeX-style symbols.
      The org-reader was droping space after unescaped LaTeX-style symbol
      commands: `\ForAll \Auml` resulted in `∀Ä` but should give `∀ Ä`
      instead.  This seems to be because the LaTeX-reader treats the
      command-terminating space as part of the command.  Dropping the trailing
      space from the symbol-command fixes this issue.
    + Print empty table rows.  Empty table rows should not
      be dropped from the output, so row-height is always set to be at least 1.
    + Move parser state into separate module.
      The org reader code has become large and confusing.  Extracting smaller
      parts into submodules should help to clean things up.
    + Add support for sub/superscript export options.
      Org-mode allows to specify export settings via `#+OPTIONS` lines.
      Disabling simple sub- and superscripts is one of these export options,
      this options is now supported.
    + Support special strings export option Parsing of special strings
      (like `...` as ellipsis or `--` as en dash) can be toggled using the `-`
      option.
    + Support emphasized text export option.  Parsing of emphasized text can
      be toggled using the `*` option.  This influences parsing of text marked
      as emphasized, strong, strikeout, and underline.  Parsing of inline math,
      code, and verbatim text is not affected by this option.
    + Support smart quotes export option.  Reading of smart quotes can be
      toggled using the `'` option.
    + Parse but ignore export options. All known export options are parsed
      but ignored.
    + Refactor block attribute handling.  A parser state attribute was used
      to keep track of block attributes defined in meta-lines.  Global state
      is undesirable, so block attributes are no longer saved as part of the
      parser state.  Old functions and the respective part of the parser state
      are removed.
    + Use custom `anyLine`.  Additional state changes need to be made after
      a newline is parsed, otherwise markup may not be recognized correctly.
      This fixes a bug where markup after certain block-types would not be
      recognized.
    + Add support for `ATTR_HTML` attributes (#1906).
      Arbitrary key-value pairs can be added to some block types using a
      `#+ATTR_HTML` line before the block.  Emacs Org-mode only includes these
      when exporting to HTML, but since we cannot make this distinction here,
      the attributes are always added.  The functionality is now supported
      for figures.
    + Add `:PROPERTIES:` drawer support (#1877).
      Headers can have optional `:PROPERTIES:` drawers associated with them.
      These drawers contain key/value pairs like the header's `id`.  The
      reader adds all listed pairs to the header's attributes; `id` and
      `class` attributes are handled specially to match the way `Attr` are
      defined.  This also changes behavior of how drawers of unknown type
      are handled.  Instead of including all unknown drawers, those are not
      read/exported, thereby matching current Emacs behavior.
    + Use `CUSTOM_ID` in properties.  See above on Org writer changes.
    + Respect drawer export setting.  The `d` export option can be used
      to control which drawers are exported and which are discarded.
      Basic support for this option is added here.
    + Ignore leading space in org code blocks (Emanuel Evans, #2862).
      Also fix up tab handling for leading whitespace in code blocks.
    + Support new syntax for export blocks.  Org-mode version 9
      uses a new syntax for export blocks.  Instead of `#+BEGIN_<FORMAT>`,
      where `<FORMAT>` is the format of the block's content, the new
      format uses `#+BEGIN_export <FORMAT>` instead.  Both types are
      supported.
    + Refactor `BEGIN...END` block parsing.
    + Fix handling of whitespace in blocks, allowing content to be indented
      less then the block header.
    + Support org-ref style citations.  The *org-ref* package is an
      org-mode extension commonly used to manage citations in org
      documents.  Basic support for the `cite:citeKey` and
      `[[cite:citeKey][prefix text::suffix text]]` syntax is added.
    + Split code into separate modules, making for cleaner code and
      better decoupling.

  * Added `docbook5` template.

  * `--mathjax` improvements:

    + Use new CommonHTML output for MathJax (updated default MathJax URL,
      #2858).
    + Change default mathjax setup to use `TeX-AMS_CHTML` configuration.
      This is designed for cases where the input is always TeX and maximal
      conformity with TeX is desired.  It seems to be smaller and load faster
      than what we used before.  See #2858.
    + Load the full MathJax config to maximize loading speed (KolenCheung).

  * Bumped upper version bounds to allow use of latest packages
    and compilation with ghc 8.

  * Require texmath 0.8.6.2.  Closes several texmath-related bugs (#2775,
    #2310, #2310, #2824).  This fixes behavior of roots, e.g.
    `\sqrt[3]{x}`, and issues with sub/superscript positioning
    and matrix column alignment in docx.

  * README:

    + Clarified documentation of `implicit_header_references` (#2904).
    + Improved documentation of `--columns` option.

  * Added appveyor setup, with artefacts (Jan Schulz).

  * stack.yaml versions: Use proper flags used for texmath, pandoc-citeproc.

  * LaTeX template: support for custom font families (vladipus).
    Needed for correct polyglossia operation with Cyrillic fonts and perhaps
    can find some other usages.  Example usage in YAML metadata:

          fontfamilies:
          - name: \cyrillicfont
            font: Liberation Serif
          - name: \cyrillicfonttt
            options: Scale=MatchLowercase
            font: Liberation

  * Create unsigned msi as build artifact in appveyor build.

  * On travis, test with ghc 8.0.1; drop testing for ghc 7.4.1.

pandoc (1.17.0.3)

  * LaTeX writer: Fixed position of label in figures (#2813).
    Previously the label wasn't in the right place, and `\ref`
    wouldn't work properly.
  * Added .tei test files to pandoc.cabal so they'll be included
    in tarball (#2811).
  * Updated copyright dates.

pandoc (1.17.0.2)

  * Fixed serious regression in `htmlInBalanced`, which caused
    newlines to be omitted in some raw HTML blocks in Markdown
    (#2804).

pandoc (1.17.0.1)

  * File scope is no longer used when there are no input files (i.e.,
    when input comes from stdin).  Previously file scope was triggered
    when the `json` reader was specified and input came from `stdin`,
    and this caused no output to be produced.  (Fix due to Jesse Rosenthal;
    thanks to Fedor Sheremetyev for calling the bug to our attention.)
  * Improved documentation of templates (#2797).

pandoc (1.17)

  * Added `--file-scope` option (Jesse Rosenthal).
    By default pandoc operates on multiple files by first concatenating
    them (around extra line breaks) and then processing the joined file. So
    it only parses a multi-file document at the document scope. This has the
    benefit that footnotes and links can be in different files, but for
    some purposes it is useful to parse the individual files first
    and then combine their outputs (e.g. when the files use footnotes
    or links with the same labels).  The `--file-scope` option causes
    pandoc to parse the files first, and then combine the parsed output,
    instead of combining before parsing. `--file-scope` is selected
    automatically for binary input files (which cannot be concatenated)
    and for pandoc json.

  * Add TEI Writer (Chris Forster) and `tei` output format.

  * Added a general `ByteStringReader` with warnings, used by the docx
    reader (API change, Jesse Rosenthal).

  * Add `readDocxWithWarnings` (API change, Jesse Rosenthal).

  * Changed type of `Shared.uniqueIdent`'s argument from `[String]`
    to `Set String.` This avoids performance problems in documents with
    many identically named headers (API change, #2671).

  * Removed `tex_math_single_backslash` from `markdown_github` options
    (#2707).

  * Make language extensions as well as full language names
    trigger syntax highlighting.  For example, `py` will now work as
    well as `python` (jgm/highlighting-kate#83).

  * Added `institute` variable to latex, beamer templates (Fraser
    Tweedale, Josef Svenningsson).

  * Docx reader (Jesse Rosenthal):

    + Handle alternate content.  Some word functions (especially graphics)
      give various choices for content so there can be backwards compatibility.
    + Don't turn numbered headers into lists.
    + Docx Reader: Add state to the parser, for warnings
    + Update feature checklist in source code.
    + Get rid of `Modifiable` typeclass.
    + Add tests for adjacent hyperlinks.
    + Add a "Link" modifier to `Reducible`. We want to make sure that
      links have their spaces removed, and are appropriately smushed
      together (#2689).

  * HTML reader:

    + Fixed behavior of base tag (#2777).
      If the base path does not end with slash, the last component
      will be replaced.  E.g. base = `http://example.com/foo`
      combines with `bar.html` to give `http://example.com/bar.html`.
      If the href begins with a slash, the whole path of the base
      is replaced.  E.g. base = `http://example.com/foo/` combines
      with `/bar.html` to give `http://example.com/bar.html`.
    + Rewrote `htmlInBalanced`.  This version avoids an exponential
      performance problem with `<script>` tags, and it should be faster
      in general (#2730).
    + Properly handle an empty cell in a simple table (#2718).
    + Handle multiple `<meta>` tags with same name.  Put them in a list
      in the metadata so they are all preserved, rather than (as before)
      throwing out all but one..

  * Markdown reader:

    + Improved pipe table parsing (#2765).
    + Allow `+` separators in pipe table cells.  We already allowed
      them in the header, but not in the body rows, for some reason.
      This gives compatibility with org-mode tables.
    + Don't cross line boundary parsing pipe table row.
      Previously an Emph element could be parsed across the newline
      at the end of the pipe table row.
    + Use `htmlInBalanced` for `rawVerbatimBlock`, for better
      performance (#2730).
    + Fixed bug with smart quotes around tex math.

  * LaTeX reader:

    + Handle interior `$` characters in math (#2743).  For example,
      `$$\hbox{$i$}$$`.
    + `inlineCommand` now gobbles an empty `{}` after any command (#2687).
      This gives better results when people write e.g. `\TeX{}` in Markdown.
    + Properly handle LaTeX "math" environment as inline math (#2171).

  * Textile reader: Support `>`, `<`, `=`, `<>` text alignment attributes.
    Closes #2674.

  * Org reader (Albert Krewinkel):

    + Prefix even empty figure names with "fig:" (#2643).  The
      convention used by pandoc for figures is to mark them by prefixing
      the name with `fig:`.  The org reader failed to do this if a figure
      had no name.
    + Refactor link-target processing (#2684).

  * ConTeXt writer: Fix whitespace at line beginning in line blocks (#2744).
    Thanks to @c-foster.

  * HTML writer: Don't include alignment attribute for default table columns.
    Previously these were given "left" alignment.  Better to leave off
    alignment attributes altogether (#2694).

  * Markdown writer: Use hyphens for YAML metadata block bottom line, for
    better compatibility with other Markdown flavors (Henrik Tramberend).

  * LaTeX writer:

    + Use image identifier to create a label and hypertarget for
      figures (Mauro Bieg).
    + Avoid double toprule in headerless table with caption (#2742).
    + Clean up options parser (Jesse Rosenthal).
    + Treat `memoir` template with `article` option as article, instead
      of treating all `memoir` templates as books.
    + Allow more flexible table alignment (Henrik Tramberend, #2665).
      New default is not to include `[c]` option (which is the default
      anyway if no positioning is specified).  Now LaTeX emplates can
      control the overall table alignment in a document by setting the
      longtable length variables `LTleft` and `LTright`.  For example,
      `\setlength\LTleft\parindent\setlength\LTright\fill`
      will create left-aligned tables that respect paragraph indentation.

  * Docx writer: Handle image alt text (#2754, Mauro Bieg).

  * Org writer - pass through RawInline with format "org".

  * DokuWiki writer: use `$$` for display math.

  * Custom writer: Pass attributes parameter to CaptionedImage (#2697).

  * Make protocol-relative URIs work again (#2737).

  * make_osx_package.sh:  Use env variable for developer id certs.

  * Raise `tagsoup` lower bound to 0.13.7 to fix entity-related
    problems (#2734).

  * Allow `zip-archive` 0.3.

  * Allow `aeson` 0.11.

pandoc (1.16.0.2)

  * Depend on deepseq rather than deepseq-generics (fpco/stackage#1096).

  * Fixed regression in latex smart quote parsing (#2645).
    In cases where a match was not found for a quote, everything
    from the open quote to the end of the paragraph was being dropped.

pandoc (1.16.0.1)

  * Fixed regression with `--latex-engine` (#2618).  In 1.16 `--latex-engine`
    raises an error if a full path is given.

  * Org reader: Fix function dropping subtrees tagged `:noexport`
    (Albert Krewinkel, #2628):

  * Markdown reader: renormalize table column widths if they exceed 100%
    (#2626).

  * Textile reader:  don't allow block HTML tags in inline contexts.
    The reader previously did allow this, following redcloth,
    which happily parses

        Html blocks can be <div>inlined</div> as well.

    as

        <p>Html blocks can be <div>inlined</div> as well.</p>

    This is invalid HTML.  The above sample now produces;

        <p>Html blocks can be</p>
        <div>
        <p>inlined</p>
        </div>
        <p>as well.</p>

  * Improved default template lookup for custom lua scripts (#2625).
    Previously, if you tried to do `pandoc -s -t /path/to/lua/script.lua`,
    pandoc would look for the template in
    `~/.pandoc/templates/default./path/to/lua/script.lua`.
    With this change it will look in the more reasonable
    `~/.pandoc/templates/default.script.lua`.  This makes it possible to
    store default templates for custom writers.

  * RST, Markdown writers: Fixed rendering of grid tables with blank rows
    (#2615).

  * LaTeX writer: restore old treatment of Span (#2624).  A Span is
    now rendered with surrounding `{}`, as it was before 1.16.

  * Entity handling fixes: improved handling of entities like
    `&lang;` that require a trailing semicolon.  Allow uppercase
    `x` in numerical hexidecimal character references, working
    around a tagsoup bug.

  * `stack.yaml` - use lts-4.0, but with older aeson to avoid excessive
    memory use on compile.  With aeson 0.10 we were getting an out of
    memory error on a 2GB Ubuntu 64-bit VM.

  * Improved deb package creation script. Made `DPKGVER` work.
    Renamed `COMMIT` to `TREE`.  You should now be able to do
    `TREE=1.16.0.1 DPKGVER=2 make deb`.


pandoc (1.16)

  * Added `Attr` field to `Link` and `Image` (Mauro Bieg, #261, API change).

    + Added syntax for link and image attributes to pandoc's Markdown.
    + Updated readers and writers to use link and image attributes
      when appropriate.
    + Support image attributes in Docx, Textile, RST readers.

  * Renamed link attribute extensions.  The old `link_attributes` is
    now `mmd_link_attributes`, and `link_attributes` now enables the
    new pandoc-style link and image attributes (API change).
    Note: this change could break some existing workflows.

  * Implemented `SoftBreak` and new `--wrap` option (#1701, API change).
    Added threefold wrapping option.

    + Command line option: deprecated `--no-wrap`, added
      `--wrap=[auto|none|preserve]`
    + Added `WrapOption`, exported from `Text.Pandoc.Options`
    + Changed type of `writerWrapText` in `WriterOptions` from
      `Bool` to `WrapOption`.
    + Modified `Text.Pandoc.Shared` functions to allow `SoftBreak`.
    + Supported `SoftBreak` in readers and writers.

  * Text.Pandoc.Options:  Added `writerDpi` to `WriterOptions` (API
    change, Mauro Bieg).

  * Added `--dpi` command-line option (Mauro Bieg).

  * Rationalized behavior of `--no-tex-ligatures` and `--smart` (#2541).
    This change makes `--no-tex-ligatures` affect the LaTeX reader
    as well as the LaTeX and ConTeXt writers.  If it is used,
    the LaTeX reader will parse characters `` ` ``, `'`, and `-`
    literally, rather than parsing ligatures for quotation marks
    and dashes.  And the LaTeX writer will print unicode quotation
    mark and dash characters literally, rather than converting
    them to the standard ASCII ligatures.  Note that `--smart` has
    no effect on the LaTeX reader.  `--smart` is still the default
    for all input formats when LaTeX or ConTeXt is the output format,
    *unless* `--no-tex-ligatures` is used.

    Some examples to illustrate the logic:

    ```
    % echo "'hi'" | pandoc -t latex
    `hi'
    % echo "'hi'" | pandoc -t latex --no-tex-ligatures
    'hi'
    % echo "'hi'" | pandoc -t latex --no-tex-ligatures --smart
    ‘hi’
    % echo "'hi'" | pandoc -f latex --no-tex-ligatures
    <p>'hi'</p>
    % echo "'hi'" | pandoc -f latex
    <p>’hi’</p>
    ```

  * Removed deprecated options `--offline` and `--html5`.

  * Fixed language code for Czech (`cs` not `cz`) (#2597).

  * Implemented `east_asian_line_breaks` extension (#2586).
    In `Text.Pandoc.Options`, added `Ext_east_asian_line_breaks` constructor
    to `Extension` (API change).  This extension is like
    `ignore_line_breaks`, but smarter -- it only ignores line breaks
    between two East Asian wide characters.  This makes it better suited
    for writing with a mix of East Asian and non-East Asian scripts.

  * Added support for PDF creation via `wkhtmltopdf`.
    To use this: `pandoc -t html5 -o result.pdf` (and add `--mathjax`
    if you have math.) Margins can be set using the variables
    `margin-top`, `margin-bottom`, `margin-left`, `margin-right`.
    Other styling can be done through CSS.

  * Fixed cite key parsing regression (jgm/pandoc-citeproc#201).
    We were capturing final colons as in `[@foo: bar]`; the citation id
    was being parsed as `@foo:`.

  * ICML writer:

    + Fixed image syntax for local files (#2589).
    + Changed type of `writeICML` (Mauro Bieg).
      API change:  It is now `WriterOptions -> Pandoc -> IO String`.
      Also handle new image attributes.
    + Intersperse line breaks instead of appending them to
      every `ParagraphStyleRange` (Mauro Bieg, #2501).
    + Add `Cite` style to citations (Mauro Bieg).
    + Added figure handling (#2590, Mauro Bieg).
    + Better handling of math.  Instead of just printing the raw tex,
      we now try to fake it with unicode characters.

  * HTML writer: Include `example` class for example lists (#2524).

  * ODT/OpenDocument writer: improved image attributes (Mauro Bieg).

    + Support for percentage widths/heights
    + Use `Attr` instead of title to get dimensions from ODT walker
      to `writeOpenDocument`.

  * AsciiDoc writer:

    + Support anchors in spans and divs with id elements
      (jgm/pandoc-citeproc#143).
    + Fixed code blocks (#1861).

  * Haddock writer:  omit formatting inside links, which isn't supported
    by Haddock (#2515).

  * MediaWiki writer:  Fixed spacing issues in table cells.

    + Start cell on new line unless it's a single Para or Plain
      (#2606).
    + For single Para or Plain, insert a space after the `|` to
      avoid problems when the text begins with a character like
      `-` (#2604).

  * Beamer writer: mark frame as fragile when it contains verbatim (#1613).

  * LaTeX writer:

    + Add support for GAP highlighting using listings (Raniere Silva).
    + Consider `header-includes` content as well as templates
      when determining whether to use csquotes (Andreas Lööw).
    + Create defaults for geometry using `margin-left` etc.
      If `geometry` has no value, but `margin-left`, `margin-right`,
      `margin-top`, and/or `-margin-bottom` are given, a default value
      for `geometry` is created from these.  Note that these variables
      already affect PDF production via HTML5 with `wkhtmltopdf`.

  * ConTeXt writer: set default layout based on `margin-left`, etc.
    This sets up `\setuplayout` based on the variables `margin-left`,
    `margin-right`, `margin-bottom`, and `margin-top`, if no layout
    is given.

  * Docx writer:  better handling of PDF images.  Previously we tried
    to get the image size from the image even if an explicit size was
    specified.  Since we still can't get image size for PDFs, this made
    it impossible to use PDF images in docx.  Now we don't try to get
    the image size when a size is already explicitly specified.

  * Markdown writer:  use raw HTML for link/image attributes when
    the `link_attributes` extension is unset and `raw_html` is set (#2554).

  * MediaWiki reader: interpret markup inside `<tt>`, `<code>` (#2607).

  * LaTeX reader:

    + Improved smart quote parsing (#2555). This fixes redering of
      unmatched quotes.
    + Use curly quotes for unmatched ` (#2555).
    + Allow blank space between braced arguments of commands (#2592).

  * Markdown reader:

    + Improved pipe table relative widths.  Previously pipe table
      columns got relative widths (based on the header underscore lines)
      when the source of one of the rows was greater in width than the
      column width.  This gave bad results in some cases where much of
      the width of the row was due to nonprinting material (e.g. link
      URLs).  Now pandoc only looks at printable width (the width of a
      plain string version of the source), which should give better results.
      Thanks to John Muccigrosso for bringing up the issue.
    + Fixed parsing bug with macros.  Previously macro definitions in
      indented code blocks were being parsed as macro definitions, not code.

  * Textile reader:  skip over attribute in image source (#2515).
    We don't have a place yet for styles or sizes on images, but
    we can skip the attributes rather than incorrectly taking them
    to be part of the filename.

  * Docx reader: Handle dummy list items (Jesse Rosenthal).
    These come up when people create a list item and then delete the
    bullet. It doesn't refer to any real list item, and we used to ignore
    it.

  * CommonMark reader/writer rewritten to use latest `cmark`.

  * Fixed Emoji character definitions (#2523).  There were many bugs in the
    definitions.

  * `Text.Pandoc.CSS`:

    + Added `pickStylesToKVs` function to extract multiple properties at
      once (API change, Mauro Bieg).
    + Parse CSS that doesn't contain the optional semicolon (Mauro Bieg).

  * `trypandoc`: sort drop-down lists.

  * Beamer template:

    + Made `\euro` conditional on presence of character.
      for xelatex and lualatex, as it is for pdflatex (Andrew Dunning).
    + Moved `header-includes` before setting of title (Thomas Hodgson),
      to match the LaTeX template (jgm/pandoc-templates#168).
    + Added `section-titles` variable (defaults to true)
      to enable/suppress section title pages in beamer
      slide shows (Thomas Hodgson).
    + Moved beamer themes after fonts, so that themes can
      change fonts.  (Previously the fonts set were being
      clobbered by lmodern.sty.) (Thomas Hodgson).

  * Beamer/LaTeX template changes (Thomas Hodgson):
    + Added `thanks` variable
    + Use `parskip.sty` when `indent` isn't set (fall back to using
      `setlength` as before if `parskip.sty` isn't available).
    + Use `biblio-style` with biblatex.
    + Added `biblatexoptions` variable.

  * LaTeX template changes:

    + Added `paper` after `$papersize$` variable in latex template.
      Thus you can say `papersize: a4` and the latex will contain
      `a4paper`.  This change may break some existing workflows; if
      you currently specify `a4paper`, you'll get `a4paperpaper` which
      is meaningless.  However, the change seems worth it, as it will
      make the `papersize` variable work uniformly across ConTeXt, LaTeX,
      and html->pdf via wkhtmltopdf.
    + Only pass options to color package if `colorlinks` is set
      (Andrew Dunning).
    + Make definition of `\euro` conditional in xelatex/lualatex,
      as it is already for pdflatex (Andrew Dunning).
    + Removed setting of `subject` in PDF metadata.
      This used to be set to the subtitle, but really the subtitle
      need not give the subject.  Also, `subtitle` can contain formatting,
      so we'd need, at least, a plain text version for this.
    + Moved `header-includes` before setting of `\title`, `\author`,
      etc.  This allows these macros to be redefined.
    + Use `\subtitle` command for `subtitle`, instead of tacking it
      on to the title as before.  We give a no-op fallback definition if it
      is not defined.  This change should produce much better results
      in classes that support `\subtitle`.  With the default article
      class, which does not define `\subtitle`, subtitles will no
      longer be printed unless the user defines `\subtitle` and
      redefines `\maketitle`.
    + Moved redefinitions of `\paragraph` and `\subparagraph` to
      before header-includes.

  * Context template:

    + Use `simplefonts` for font loading (Paolo Rodríguez).  This is
      needed for things to work on ConTeXt stable from TeXLive 2015.
    + Revert use of `\setuphead` in title block (Andrew Dunning,
      Rik Kabel).

  * Update LaTeX/ConTeXt link colour usage (Andrew Dunning).

  * Fixed man template so disabling hyphenation actually works.
    The command needs to come after .TH.

  * Added 'navigation' variable to beamer template (#2543).
    Valid values are `empty` (the default), `horizontal`, `vertical`,
    and `frame`.  Note that this changes the default behavior from
    `horizontal` to `empty`.  Closes #2543.

  * Added `toc` to HTML slide format templates (Andrew Dunning),
    so that `--toc` creates a contents slide.

  * Added `stack.full.yaml` to build `pandoc-citeproc` as well.

  * Allow pipe tables with no body rows (#2556).
    Previously this raised a runtime error.

  * Shared: Improved `fetchItem` so that `C:/Blah/Blah.jpg` isn't treated
    as URL.  The Haskell URI parsing routines will accept "C:" as a
    scheme, so we rule that out manually.  This helps with
    `--self-contained` and absolute Windows paths.

  * Define a `meta-json` variable for all writers (#2019).  This contains
    a JSON version of all the metadata, in the format selected for the
    writer.  So, for example, to get just the YAML metadata, you can run
    pandoc with the following custom template: `$meta-json$`.  The intent
    is to make it easier for static site generators and other tools to get
    at the metadata.

  * Document limitations of --self-contained (#2553).

  * Improved Citations section of README (#2551).  Added information
    about `link-citations` and a link to the pandoc-citeproc man page.

  * `ImageSize`: use `safeRead` instead of `readMaybe`, which isn't
    in base < 4.6.

  * Allow .adoc file extension for AsciiDoc (Andrew Dunning).

  * Improved implicit pandoc-citeproc inclusion.
    The filter pandoc-citeproc is automatically used when
    `--bibliography` is specified on the command line, unless
    `--natbib` or `--biblatex` is used.  However, previously this
    only worked if `--bibliography` was spelled out in full, and not
    if `--biblio` was used.

  * reveal.js: Interpret pauses correctly for all headers (#2530).
    Previously, when using headers below the slide level, pauses are left
    uninterpreted into pauses. In my opinion, unexpected behavior but
    intentional looking at the code.

  * Remove redundant `center` variable for reveal.js (Andrew Dunning).

  * Parsing: Add `extractIdClass`, modified type of `KeyTable` (Mauro
    Bieg, API change).

  * ImageSize:  Added functions for converting between image dimensions
    (Mauro Bieg).

  * Use lts-3.18 in stack.yaml.  This avoids Windows build
    issues with the HTTP library.

  * Bump version bounds for dependencies.

pandoc (1.15.2.1)

  * Added two missing test files, and `stack.yaml`, to
    `extra-source-files` so they're included in the source tarball.

  * reveal.js template: Fixed parallaxBackground options.
    `parallaxBackgroundHorizontal` and `parallaxBackgroundVertical`
    need integer values, not strings.  (Vaughn Iverson)

pandoc (1.15.2)

  * `pandoc my.md -t context -o my.pdf` will now create a PDF using
    ConTeXt rather than LaTeX (#2463).

  * Fixed omitted `url(...)` in CSS data-uri with `--self-contained` (#2489).

  * Added `emoji` Markdown extension, enabled by default in `markdown_github`
    (#2523).  Added `Ext_emoji` to `Extension` in `Text.Pandoc.Options`
    (API change).

  * `Text.Pandoc.Readers.HTML.parseTags`: Fixed over-eager raw HTML inline
    parsing (#2469).  Tightened up the inline HTML parser so it disallows
    TagWarnings.

  * Derive `Generic` instances for the types in `Text.Pandoc.Options`.

  * Org reader:

    + Fix paragraph/list interaction (Albert Krewinkel, #2464).
      Paragraphs can be followed by lists, even if there is no blank line
      between the two blocks.  However, this should only be true if the
      paragraph is not within a list, were the preceding block should be
      parsed as a plain instead of paragraph (to allow for compact lists).
      Thanks to @rgaiacs for bringing this up.
    + Allow toggling header args (Albert Krewinkel, #2269).
      Org-mode allows to skip the argument of a code block header argument if
      it's toggling a value.  Argument-less headers are now recognized,
      avoiding weird parsing errors.
    + Fix markup parsing in headers (Albert Krewinkel, #2504).
      Markup as the very first item in a header wasn't recognized.  This was
      caused by an incorrect parser state: positions at which inline markup
      can start need to be marked explicitly by changing the parser state.
      This wasn't done for headers.  The proper function to update the state
      is now called at the beginning of the header parser, fixing this issue.
    + Fix emphasis rules for smart parsing (Albert Krewinkel, #2513).
      Smart quotes, ellipses, and dashes should behave like normal quotes,
      single dashes, and dots with respect to text markup parsing.
    + Require whitespace around definition list markers (#2518).
      This rule was not checked before, resulting in bugs with footnotes
      and some link types.

  * Markdown reader:

    + Pipe tables with long lines now get relative cell widths (#2471).
      If a pipe table contains a line longer than the column width (as set by
      `--columns` or 80 by default), relative widths are computed based on the
      widths of the separator lines relative to the column width.  This should
      solve persistent problems with long pipe tables in LaTeX/PDF output, and
      give more flexibility for determining relative column widths in other
      formats, too.  For narrower pipe tables, column widths of 0 are used,
      telling pandoc not to specify widths explicitly in output formats that
      permit this.
    + Improved parser for `mmd_title_block`.  We now allow blank metadata
      fields.  These were explicitly disallowed before.
    + Citation keys can now contain `://`, so URLs and DOIs can be used
      as citation keys (jgm/pandoc-citeproc#166).

  * Beamer template:  fix incompatibility of section slides with natbib.
    Natbib (and presumably biblatex) bibliography commands create
    their own section.  Since these are in frame environments,
    we have an incompatibility with the `\AtBeginSection` macro
    which creates a special frame when a new section occurs.
    (We can't have a frame inside another frame.) This change disables
    `\AtBeginSection` inside bibliography slides.  Thinks to Yihui Xie for
    bringing the problem to my attention.  This supersedes #145.  See
    discussion there.

  * Textile reader:  don't do smart punctuation unless explicitly asked
    (#2480).  Note that although smart punctuation is part of the textile
    spec, it's not always wanted when converting from textile
    to, say, Markdown.  So it seems better to make this an option.

  * LaTeX reader: Handle `comment` environment (Arata Mizuki).
    The `comment` environment is handled in a similar way to the
    `verbatim` environment, except that its content is discarded.

  * Docx reader:  Follow relationships correctly in foot/endnotes (#2258,
    Jesse Rosenthal).  This fixes a problem with links in notes.

  * LaTeX and ConTeXt writers: support `lang` attribute on divs and spans
    (Mauro Bieg).  For LaTeX, also collect `lang` and `dir` attributes on
    spans and divs to set the `lang`, `otherlangs` and `dir` variables if
    they aren’t set already.  See #895.

  * LaTeX writer:

    + Use proper command for `\textarabic` (Mauro Bieg).
    + Added `de-CH-1901`, fixed `el-polyton` in `toPloyglossia` (Nick Bart).
    + Use `\hypertarget` and `\hyperlink` for links.  This works correctly
      to link to Div or Span elements.  We now don't bother defining `\label`
      for Div or Span elements.  Closes jgm/pandoc-citeproc#174.
    + Avoid footnotes in list of figures (#1506).
    + Properly handle footnotes in captions (#1506).
    + Add `\protect` to `\hyperlink` (#2490).  Thanks to Hadrien Mary.
    + Set `colorlinks` if `linkcolor`, `urlcolor`, `citecolor`, or
      `toccolor` is set (#2508).

  * Textile writer: support start number in ordered lists (#2465).

  * OpenDocument writer:  Allow customization of opendocument
    automatic styles.  Automatic styles can now be inserted in the
    template, which now provides the enclosing `<office:automatic-styles>`
    tags (#2520).

  * Docx writer:  insert space between footnote reference and note (#2527).
    This matches Word's default behavior.

  * EPUB writer:  don't download linked media when `data-external` attribute
    set (#2473).  By default pandoc downloads all linked media and includes it
    in the EPUB container.  This can be disabled by setting `data-external` on
    the tags linking to media that should not be downloaded.  Example:

        <audio controls="1">
          <source src="http://example.com/music/toccata.mp3"
                  data-external="1" type="audio/mpeg">
          </source>
        </audio>

  * HTML writer:  use width on whole table if col widths sum to < 100%.
    Otherwise some browsers display the table with the columns
    separated far apart.

  * AsciiDoc template:  Fix `author` and `date`; add `keywords`,
    `abstract` (Andrew Dunning).

  * HTML-based templates (Andrew Dunning):

    + Use en dash instead of hyphen between title prefix and title.
    + Add `keywords` to metadata.
    + Add `lang`, `dir`, `quotes` where missing.
    + Always make author and date display conditional.
    + Updated dzslides template from source.

  * Man template: make "generated by" comment conditional.

  * LaTeX, Beamer templates:

    + Add `babel-otherlangs` for language divs/spans; `babel-newcommands`,
      filled by commands that make babel understand the polyglossia-style
      language directives (Mauro Bieg, #137).
    + Improved formatting of conditionals; `$for$` is always provided to allow
      multiple options (Andrew Dunning, #141).
    + Use `Ligatures=TeX` rather than `Mapping=tex-text` with `fontspec`
      to improve support for LuaTeX (Andrew Dunning, #135).
    + Revise `hyperref` usage (Andrew Dunning, #139, #141):
        - use same options for all LaTeX engines;
        - add `subtitle` and `keywords` to PDF metadata;
        - do not override `hyperref` link coloring without user input, effectively making
          the `hidelinks` option the default (removed as a separate variable);
        - link colors can be enabled (using a slightly darker version of the old
          defaults) using a new `colorlinks` variable, automatically used by
          the LaTeX writer when custom colors are specified;
        - `pdfborder={0 0 0}` is automatically set by `hyperref` with
          `colorlinks`, and is only applied if `colorlinks` is disabled.

  * ConTeXt template (Andrew Dunning):

    + New variables for controlling styles: `linkstyle`, `linkcolor`,
      `linkcontrastcolor`, `layout`, `pagenumbering`, `whitespace`, `indenting`,
      `interlinespace`, `headertext`, `footertext`, `mainfont`, `sansfont`,
      `monofont`, `mathfont`, `fontsize`.
    + Default template no longer supports MkII.
    + Improve writing of title block (suppressing numbering of first page).
    + Add `title` `subtitle`, `author`, `date`, `keywords` to PDF metadata.
    + Support `subtitle`, `abstract`.
    + Support list of figures (`lof`), list of tables (`lot`).
    + Disable link styling by default.
    + Define styles for all section types.
    + Enable microtype.
    + Improved formatting of conditionals.

  * Beamer template:  added code to prevent slide breaks inside paragraphs
    (#2422, thanks to Nick Bart).  This will matter, in practice, only when
    `allowframebreaks` is used.  It is especially helpful for bibliography
    slides.

  * OpenDocument template:  Add `<office:automatic-styles>` tag around
    automatic styles.  The writer now longer provides this (see #2520).

  * Restored Text.Pandoc.Compat.Monoid.

  * Do not export (<>) from custom Prelude.  The Prelude now matches
    base 4.8 Prelude's API.

  * Don't use custom prelude with ghc 7.10.  Use the custom prelude
    only for earlier versions.  This change makes `stack ghci` and
    `cabal repl` work (#2503), at least with ghc 7.10.

  * Changed § to % in operators from Odt.Arrows.Utils (#2457).
    This prevents problems building haddocks with "C" locale.

  * Change default for old-locale flag to False.

  * Use stack in deb, osx, and Windows package generators.

  * Added Vagrantfile for building deb in vm.
    This should help in automating binary package creation.  'make package'
    will make the package.  'make package COMMIT=blah' will make the package
    from commit blah.

  * README:

    + Consistent capitalization for pandoc and Markdown.
    + Fixed `auto_identifiers` examples (Benoit Schweblin).
    + Improved documentation of template variables (Andrew Dunning).

pandoc (1.15.1.1)

  * `Text.Pandoc.Data`:  store paths in dataFiles using posix separators.
    This way we have uniform separators, whether on Windows or Linux.
    This should solve a problem where on some Windows versions
    the data files weren't being found (#2459).

pandoc (1.15.1)

  * `pandocVersion` is now defined in `Text.Pandoc.Shared`
    and reexported from `Text.Pandoc` (Alex Vong).  This allows
    writers to access it.  (Alex Vong) (API change)

  * For `markdown_mmd`, add: `implicit_figures`, `superscripts`,
    `subscripts` (#2401).

  * Added `odt` as input format (Martin Linnemann).  Added new module
    `Text.Pandoc.Reader.ODT` (API change). Fully implemented features:
    Paragraphs, Headers, Basic styling, Unordered lists, Ordered lists,
    External Links, Internal Links, Footnotes, Endnotes, Blockquotes.
    Partly implemented features: Citations, Tables.

  * Markdown Reader:

    + Add basic tests for each header style (Ophir Lifshitz).
    + Add implicit header ref tests for headers with spaces (Ophir Lifshitz).
    + Skip spaces in headers (Ophir Lifshitz).
    + Handle 'id' and 'class' in parsing key/value attributes (#2396).
      `# Header {id="myid" class="foo bar"}`
      is now equivalent to `# Header {#myid .foo .bar}`.
    + Use '=' instead of '#' for atx-style headers in markdown+lhs.
      (Kristof Bastiaensen)
    + Pipe tables: allow indented columns.  Previously the left-hand column
      could not start with 4 or more spaces indent.  This was inconvenient
      for right-aligned left columns.  Note that the first (header column)
      must still have 3 or fewer spaces indentation, or the table will be
      treated as an indented code  block.
    + Fix regression:  allow HTML comments containing `--`.
      Technically this isn't allowed in an HTML comment, but
      we've always allowed it, and so do most other implementations.
      It is handy if e.g. you want to put command line arguments
      in HTML comments.

  * LaTeX reader:

    + Don't eat excess whitespace after macros with only optional
      arguments (#2446).
    + Support longtable (#2411).
    + Implement `\Cite` (#2335).
    + Support abstract environment.  The abstract populates an
      `abstract` metadata field.
    + Properly handle booktabs lines.  Lines aren't part of the
      pandoc table model, so we just ignore them (#2307).

  * HTML reader:

    + Handle type attribute on ol, e.g. `<ol type="i">` (#2313).
    + Updated for new automatic header attributes.
    + Add auto identifiers if not present on headers.  This makes
      TOC linking work properly.
    + Detect `font-variant` with `pickStyleAttrProps` (Ophir Lifshitz).
    + Test `<ol>` type, class, and inline list-style(-type) CSS
      (Ophir Lifshitz).
    + Better handling of "section" elements (#2438).  Previously
      `<section>` tags were just parsed as raw HTML blocks.  With
      this change, section elements are parsed as Div elements with
      the class "section".

  * MediaWiki reader:  handle unquoted table attributes (#2355).

  * DocBook reader:

    + Added proper support for DocBook `xref` elements (Frerich Raabe).
      Added `dbContent` field to reader state, so we can lookup
      cross refs.
    + Handle `informalexample` (#2319).

  * Docx Reader:

    + Create special punctuation test (Ophir Lifshitz).
    + Parse soft, no-break hyphen elements (Ophir Lifshitz).
    + Updated headers test (Ophir Lifshitz). Replaced `styles.xml`
      in `headers.docx` with pandoc's current `styles.xml`, which
      contains styles for Heading 1 through 6. Added Heading 4
      through 7 to the test document. Note that Heading 7 is not
      parsed as a Heading because there is no Heading 7 style.

  * RST reader:  better handling of indirect roles.
    Previously the parser failed on this kind of case

        .. role:: indirect(code)

        .. role:: py(indirect)
           :language: python

        :py:`hi`

    Now it correctly recognizes `:py:` as a code role.

  * Org reader:

    + Add auto identifiers if not present on headers
      (#2354, Juliusz Gonera).
    + Allow verse blocks to contain empty lines (#2402,
      Albert Krewinkel).

  * EPUB reader:  stop mangling external URLs (#2284).

  * RST writer:

    + Don't insert `\ ` when complex expression in matched pairs.
      E.g. `` [:sup:`3`] `` is okay; you don't need `` [:sup:`3`\ ] ``.
    + Ensure that `\ ` is inserted when needed before Cite and Span
      elements that begin with a "complex" element (jgm/pandoc-citeproc#157).
    + Normalize headers only in "standalone" mode (#2394).

  * Haddock writer: escape `*` and `^` (G. Bataille).

  * Markdown writer:

    + In TOC, add links to headers (#829).
    + Use unicode super/subscripts for digits in plain output
      (when the `superscripts` and `subscripts` extensions are
      not enabled).

  * Docx writer:

    + Moved invalid character stripping to `formattedString`.
      This avoids an inefficient generic traversal (#2356).
    + Use user data directory for `reference.docx` archive.
      This allows the test suite to work without installing pandoc first.
      It also brings the docx writer in line with the odt writer.
    + Tests:  docx writer tests now use `../data` for data directory.
      This allows tests to be run without installing first.
    + Tests: Use real jpg (not empty) for docx tests to avoid warning.

  * LaTeX writer:

    + Fixed detection of 'chapters' from template.
      If a documentclass isn't specified in metadata, but the
      template has a hardwired bookish documentclass, act as if
      `--chapters` was used.  This was the default in earlier
      versions, but it has been broken for a little while.
    + Correctly recognize book documentclass in metadata (#2395).
    + Set language-related variables automatically, depending
      on the value of the `lang` field, which is now always
      assumed to be in BCP47 format (Mauro Bieg, #1614, #2437).
    + Add `\protect` to `\hyperdef` in inline context.  This way we
      don't get an error when this is used as a moveable argument (#2136).
    + Support all frame attributes in Beamer.
    + Percent-encode more special characters in URLs (#1640, #2377).
      The special characters are '<','>','|','"','{','}','[',']','^', '`'.

  * HTML writer:

    + Update KaTeX JS and CSS versions (Emily Eisenberg).
    + For dzslides, add `role="note"` for speaker notes (#1693).
    + Percent-encode more special characters in URLs (#1640, #2377).
      The special characters are '<','>','|','"','{','}','[',']','^', '`'.
    + Render Div with class `section` as `<section>` in HTML5.

  * EPUB writer:

    + In TOC, replace literal `<br/>` with space (#2105).
    + With `--webtex`, include image file rather than `data:` URI (#2363).

  * Native writer: format Div properly, with blocks separated.

  * Support bidirectional text output with XeLaTeX, ConTeXt and HTML
    (#2191, Mauro Bieg).

  * Reference Docx:

    + Add missing Header 6 style (steel blue) (Ophir Lifshitz).
    + Correct `outlineLvl` for Header styles (Ophir Lifshitz).

  * Templates

    + Beamer:  Add `innertheme`, `outertheme` variables
      (Guilhem Bonnefille, #121). Add space after colon in figure caption.
      Integrate recent font and language updates from LaTeX template;
      allow use of `mainfont` variable for changing the slide text
      in XeTeX and LuaTeX (Andrew Dunning, #131).
    + LaTeX:  Add `mainfontoptions`, `sansfontoptions`,
      `monofontoptions`, `mathfontoptions`, `fontfamilyoptions`
      (Andrew Dunning, #122).  Support handling of bidirectional
      text (Mauro Bieg, #120). Improve reliability of superscripts/subscripts
      under XeTeX and prevent letters and numbers from appearing on a
      different baseline by removing use of the `realscripts` package
      (via `xltxtra`).  To restore use of OpenType characters for these
      features under XeTeX or LuaTeX, add `\usepackage{realscripts}` to
      `header-includes` (Andrew Dunning, #130).  Remove redundant
      reference to `xunicode` (Andrew Dunning, #130).  Add `fontenc`,
      `indent`, `subparagraph` variables (Andrew Dunning).
      Allow use of `hidelinks` variable for `hyperref` package (Hugo Roy,
      #113).  Prevent package clash with `tufte-latex` and other classes that
      include `hyperref` or `color` (Xavier Olive, #115).
    + ConTeXt:  Support handling of bidirectional text (Mauro Bieg, #120).
    + LaTeX and ConTeXt: Use more specific language variables.
      Instead of directly using `lang`, we now use `babel-lang` and
      `polyglossia-lang` and `context-lang`.  These variables are set by
      the writers to the necessary values, based on the `lang` variable
      (which now always takes a value in BCP47 format). (Mauro Bieg, #114,
      #129).
    + HTML:  Support handling of bidirectional text (Mauro Bieg, #120).
      Move HTML5 shiv after CSS and fix URL (Andrew Dunning).
      Add dir attribute in html5 (Andrew Dunning).
    + reveal.js: Add `controls`, `progress` variables (Grégoire Pineau, #127).
      Add `width`, `height` variables (Andrew Dunning).  Update template
      from 3.1 source (Andrew Dunning).  All configuration options are now
      available as variables, but are only be included if set (reveal.js
      uses defaults otherwise).
    + man: Added comment stating that the page is autogenerated by pandoc,
      giving version.  Added `adjusting` and `hyphenate` variables
      (Alex Vong, #123).

  * epub.css: added selectors for nested emphasis (Pablo Rodriguez).

  * MediaBag:  ensure that `/` is always used as path separator.

  * `sample.lua`: define `CaptionedImage`, add newline at end (#2393).

  * Added `--bash-completion` option.  This generates a bash completion
    script.  To use: `eval "$(pandoc --bash-completion)"`.

  * Text.Pandoc.Error: Define Typeable and Exception instances
    for PandocError (#2386).

  * Text.Pandoc.Parsing: `toKey`: strip off outer brackets.
    This makes keys with extra space at the beginning and end
    work:  e.g.

        [foo]: bar

        [ foo ]

    will now be a link to bar (it wasn't before).

  * Text.Pandoc: disable `auto_identifiers` for epub.
    The epub writer inserts its own auto identifiers;
    this is more complex due to splitting into "chapter" files.

  * Renamed Text.Pandoc.Compat.Locale -> Text.Pandoc.Compat.Time.
    It now reexports Data.Time.

  * Use custom Prelude to avoid compiler warnings.

    + The (non-exported) prelude is in prelude/Prelude.hs.
    + It exports Monoid and Applicative, like base 4.8 prelude,
      but works with older base versions.
    + It exports (<>) for mappend.
    + It hides 'catch' on older base versions.

  * Added a `stack.ymal` and stack install instructions to INSTALL.

  * Clarified what is "out of scope" in README and CONTRIBUTING.md.

  * Added note to CONTRIBUTING.md about ghc versions and travis.

  * Clarify docs on block quotes.  The space after `>` is optional (#2346).

  * Removed obsolete reference to default.csl (#2372).

  * List all styles in manual for `--reference-docx` (Chris Black)

  * Don't capitalize header links in man page.

  * Added section on repl to CONTRIBUTING.md.

  * README:  Added space after backslash in image example (#2329).

  * Document details of citation locator terms (Nick Bart).

  * Fixed some internal links in README (#2309).

  * Improve CSL documentation, variables documentations,
    links, and cross-references in README. (Andrew Dunning)

  * Fix build failure with `--flags=-https` (Sergei Trofimovich).

  * Use `newManager` instead of `withManager` in recent `http-client`.
    This avoids a deprecation warning.

  * Allow building with latest versions of http-types,
    HUnit, criterion, syb, aeson.

  * Updated benchmark program for new criterion API.

  * Setup.hs: rewrite so as not to use process, directory, filepath.
    Using anything outside base is dangerous, since older
    versions of ghc may link against two different versions.

  * Added appveyor (Windows continuous integration) builds.

  * New `.travis.yml`.  Autgenerated using `make_travis_yml.hs`.
    This script has been modified in a few ways, e.g. to add `GHCOPTS`.
    `make .travis.yml` regenerates it based on the tested-with
    field of the cabal file.

pandoc (1.15.0.6)

  * `--self-contained`:  Fixed overaggressive CSS minimization (#2301, 2286).
    Previously `--self-contained` wiped out all spaces in CSS,
    including semantically significant spaces.  This was a regression
    from 1.14.x.

  * Markdown reader: don't allow bare URI links or autolinks in link
    label (#2300).  Added test cases.

  * `Text.Pandoc.Parsing`, `uri`: Improved bare autolink detection (#2299).
    Previously we disallowed `-` at the end of an autolink,
    and disallowed the combination `=-`.  This commit liberalizes the
    rules for allowing punctuation in a bare URI, and adds test cases.
    One potential drawback is that you can no longer put a bare
    URI in em dashes like this:
    `this uri---http://example.com---is an example.`
    But in this respect we now match github's treatment of bare URIs.

  * HTML writer:  support speaker notes in dzslides.
    With this change `<div class="notes">` and also `<div class="notes"
    role="note">` will be output if `-t dzslides` is used. So we can
    have speaker notes in dzslides too.  Thanks to maybegeek.

  * Updated dzslides template.

  * Improved documentation of options to print system default files (#2298).
    `--print-default-data-file` and `--print-default-template`.

  * DokuWiki writer: use `$..$` for Math instead of `<math>..</math>`
    (Tiziano Müller).  MathJax seems currently to be the only maintained
    math rendering extension for DokuWiki.

  * `Text.Pandoc.Shared`: Changed `hierarchicalize` so it treats references
    div as top-level header (#2294).  This fixes a bug with `--section-divs`,
    where the final references section added by pandoc-citeproc, enclosed in
    its own div, got nested in the div for the section previous to it.

  * Allow vector 0.11.

  * Require cmark > 0.4.

pandoc (1.15.0.5)

  * HTML writer: Fixed email javascript obfuscation with `mailto:`
    URLs (#2280).  This fixes a potential security issue.  Because
    single quotes weren't being escaped in the link portion, a
    specially crafted email address could allow javascript code injection.

  * Markdown/HTML readers:  Avoid parsing partial URLs like
    `<www.pandoc.org/blah#foo>` as HTML tags (#2277).

  * RST reader:  allow inline formatting in definition list field
    names (Lars-Dominik Braun).

  * PDF:  Make sure `--latex-engine-opt` goes before the filename
    on the command line.  LaTeX needs the argument to come after
    the options (#1779).

  * CommonMark writer: fixed tags used for super/subscript.

  * ConTeXt template:  activate hanging indent for definition lists
    (Mauro Bieg).

  * Make cabal require `hsb2hs` >= 0.3.1 if `embed_data_files` specified.
    This is done by adding `hookedPrograms` in `Setup.hs`, which allows us
    to include `hsb2hs` in Build-Tools in cabal.

  * Improved Windows installer (thanks to nkalvi).

    + When per-machine installation is chosen, the system path
      is updated instead of the user's.
    + An appropriate default is used for per-machine installation
      directory.
    + Admin privileges are no longer required for a per-user install

  * Travis:  unpack sdist for build to catch packaging bugs.

  * Improved documentation on where user templates go (#2272).

pandoc (1.15.0.4)

  * Added pandoc.1 man page to the repository.  It is no longer
    built as part of the cabal build process. (This proved too
    fragile.)  pandoc.1 can be regenerated (`make man/pandoc.1`)
    when `README` is changed.

  * Copying of the man page now respects `--destdir` (#2262).

  * Improved error messages for filters.  User is now informed if
    the filter requires an interpreter that isn't found in the path,
    or if the filter returns an error status.

pandoc (1.15.0.3)

  * Ensure target directory is created when installing man page.

pandoc (1.15.0.2)

  * Added files needed for building man page to Extra-Source-Files.

pandoc (1.15.0.1)

  * Man page is now built and installed as part of the cabal build
    process. Removed Makefile target for man page.

pandoc (1.15)

  * Man page changes:

    + Removed `--man1`, `--man5` options (breaking change).
    + Removed `Text.Pandoc.ManPages` module (breaking API change).
    + Makefile target for `man/man1/pandoc.1`.  This uses pandoc to
      create the man page from README using a custom template and filters.
    + Added `man/` directory with template and filters needed to build
      man page.
    + We no longer have two man pages: `pandoc.1` and `pandoc_markdown.5`.
      Now there is just pandoc.1, which has all the content from README.
      This change was needed because of the extensive cross-references
      between parts of the README.
    + Removed old `data/pandoc.1.template` and
      `data/pandoc_markdown.5.template`.

  * OpenDocument writer: Do not add a carriage return after a hard
    line break (Michael Chladek).

  * ConTeXt writer:

    + use `\goto` for internal links.
    + Added a `%` at end for `\reference` to avoid spurious space.

  * Ignore sandbox on 'make quick'

pandoc (1.14.1)

  * Added `--man1` and `--man5` options to pandoc, allowing pandoc
    to generate its own man pages.  Man pages are no longer automatically
    generated in the build process (the process for this was too complex
    and prone to failure, #2190).  The `make-pandoc-man-pages` executable
    has been removed.  The `man/` directory has been removed, and man page
    templates have been moved to `data/`.  NOTE TO PACKAGERS:  You will no
    longer find pandoc's man pages in `man/`, but you can generate them using
    `pandoc --man1 > pandoc.1` and `pandoc --man5 > pandoc_markdown.5`.

  * Added new unexported module:  `Text.Pandoc.ManPages`.

  * `README` now acts like a data file (even though it isn't in
    `data/`).  So, for example, `pandoc --print-default-data-file README`
    will produce the README.)  This change was required for the `--man1`
    and `--man5` options, since the man pages are produced from the
    README, but it may be useful for other purposes as well.

  * Allow `reference.docx` and `reference.odt` to be used with
    `--print-default-data-file` and to shadow defaults if placed in
    the user data directory.  Note that as of 1.14, we no longer
    include these files as data files; instead, we include their
    components.  This change causes pandoc to behave as if it has
    these data files; they are constructed on demand when needed
    using `getDefaultReferenceDocx` and `getDefaultReferenceODT`.

  * Fixed regression in CSS parsing with `--self-contained` (#2224).
    Pandoc 1.14.0.x used css-text to parse the CSS, but its parser
    silently drops big sections of CSS.  This commit replaces the
    use of css-text with a small but principled CSS preprocessor,
    which removes whitespace and comments and replaces `url()` with
    base 64 data when possible.

  * Use `https://` instead of `//` for MathJax and KaTeX CDN URLs (#1920).
    This will allow math to work when pages are being viewed locally.

  * `Text.Pandoc.Options`:  Export `plainExtensions`.
    These are the extensions used in `plain` output.

  * LaTeX reader: Don't parse `_` and `^` as sub/superscript outside of
    math mode; treat them as regular inline text.  Normally these will
    cause an error in LaTeX, but there are contexts (e.g. `alltt`
    environments) where they are allowed.

  * HTML reader:  allow `<body>` to close `<head>`.

  * DocBook reader: support `mediaobject`s and `figures` (#2184, Mauro Bieg).

  * RST reader: Fix reference names with special characters
    (Lars-Dominik Braun).

  * Textile writer:  escape `+` and `-` as entities (#2225).

  * DokuWiki writer: Use proper `<code>` tags for code blocks (#2213).

  * Plain writer:  don't use symbols for super/subscript (#2237).
    Simplified code by using `plainExtensions`.

  * InDesign writer: Properly escape URLs containing more than one
    colon character (gohai).

  * Docx writer: Make sure we use dist version of `reference.docx`
    (and not the user's version) for certain settings.  Taking some
    settings values from a user-supplied reference.docx can lead to
    corruption.  This fixes a regression from the last release (#2249).

  * `Text.Pandoc.Shared`: exports `getDefaultReferenceDocx` and
    `getDefaultReferenceODT` (API change).  These functions have been
    removed from the Docx and ODT writers.

  * LaTeX template (Xavier Olive):
    + Added `CJKmainfont` and `CJKoptions` variables.
    + Allow dvipsnames (e.g. `MidnightBlue`) for colors (Xavier Olive).

  * Epub templates:  use `author.role`, not `author.type`.

  * Bump cmark version to >= 0.3.4.

  * Improved Windows installer (#2205, thanks to nkalvi).
    Users can now select a per-user or systemwide install, and can set
    the installation path.  At the end of installation, the install location
    is given.  The install location is also now given in the list of
    installed programs in Control Panel.  Cleaner WiX syntax is used for
    setting the path.

  * Added `download_stats` target to Makefile.

pandoc (1.14.0.4)

  * Added missing commonmark template.

  * Improved try pandoc (moved button, show raw command).

pandoc (1.14.0.3)

  * Allow compilation with syb 0.5.*.

  * Custom writer:  fixed some compiler warnings for ghc < 7.10.

pandoc (1.14.0.2)

  * Allow building with hslua 0.4.

pandoc (1.14.0.1)

  * Fixed problem with building of `reference.docx` and `reference.odt`
    when the `embed_data_files` flag is used.  Instead of having a phase
    of the build where `reference.docx` and `reference.odt` are created
    from their constituent data files, we now construct these archives
    from their constituents when a `docx` or `odt` is built.  The
    constituent files have been moved from `extra-source-files` to
    `data-files`, and `reference.docx` and `reference.odt` have been
    removed. Users can create their own `reference.docx` or
    `reference.odt` by using pandoc to create a simple `docx` or `odt`.
    `make-reference-files.hs` has been removed, simplifying the build
    process (#2187)

  * Don't include generated man pages in extra-source-files (#2189).

  * Bumped upper bound for aeson.

  * ConTeXt writer:  create internal link anchors for Div elements with
    identifiers.  (This is needed for linked citations to work.)

pandoc (1.14)

  [new features]

  * Added `commonmark` as input and output format.

  * Added `--verbose` flag for debugging output in PDF production (#1840,
    #1653).

  * Allow wildcards in `--epub-embed-font` arguments (#1939).

  * Added `--latex-engine-opt` option (#969, #1779, Sumit Sahrawat).

  * Added `shortcut_reference_links` extension (Konstantin Zudov, #1977).
    This is enabled by default for those markdown flavors that
    support reading shortcut reference links, namely: `markdown`,
    `markdown_strict`, `markdown_github`, `markdown_php`.
    If the extension is enabled, the reader parses shortcut reference
    links like `[foo]`, and the writer creates such links unless doing
    so would cause problems.  Users of markdown flavors that support
    shortcut reference links should not notice a difference in reading
    markdown, but the markdown pandoc produces may differ.
    If shortcut links are not desired, the extension can be disabled
    in the normal way.

  [behavior changes]

  * `--toc` is now supported for `docx` output (#458, Nikolay Yakimov).
    A "dirty" TOC is created at the beginning of document.
    It can be regenerated after the document has been opened.

  * An implicit `--filter pandoc-citeproc` is now triggered only when the
    `--bibliography` option is used, and not when the `bibliography`
    field in metadata is specified (#1849).

  * Markdown reader:

    + Reference links with `implicit_header_references` are no longer
      case-sensitive (#1606).
    + Definition lists no longer require indentation for first line (#2087).
      Previously the body of the definition (after the `:` or `~` marker)
      needed to be in column 4.  This commit relaxes that requirement,
      to better match the behavior of PHP Markdown Extra.  So, now
      this is a valid definition list:

            foo
            : bar
    + Resolve a potentially ambiguity with table captions:

            foo

              : bar

              -----
              table
              -----

      Is "bar" a definition, or the caption for the table?  We'll count
      it as a caption for the table.
    + Disallow headerless pipe tables (#1996), to conform to GFM and PHP
      Markdown Extra.  Note:  If you have been using headerless pipe tables,
      this change may cause existing tables to break.
    + Allow pipe tables with header but no body (#2017).
    + Allow a digit as first character of a citation key (Matthias Troffaes).
      See https://github.com/jgm/pandoc-citeproc/issues/97

  * LaTeX reader:

    + Don't limit includes to `.tex` extension (#1882).
      If the extension is not `.tex`, it must be given explicitly in
      the `\input` or `\include`.

  * Docx reader:

    + Allow numbering in the style file.  This allows inherited styles
      with numbering (lists) (Jesse Rosenthal).

  * Org reader:

    + Support smart punctuation (Craig Bosma).
    + Drop trees with a :noexport: tag (Albert Krewinkel). Trees having a
      `:noexport:` tag set are not exported.  This mirrors org-mode.
    + Put header tags into empty spans (Albert Krewinkel, #2160).
      Org mode allows headers to be tagged: `* Headline  :TAG1:TAG2`.
      Instead of being interpreted as part of the headline, the tags are now
      put into the attributes of empty spans.  Spans without textual content
      won't be visible by default, but they are detectable by filters.  They
      can also be styled using CSS when written as HTML.
    + Generalize code block result parsing (Albert Krewinkel).
      Previously, only code blocks were recognized as result blocks;
      now, any kind of block can be the result.

  * Append newline to the LineBreak in Dokuwiki, HTML, EPUB,
    LaTeX, MediaWiki, OpenDocument, Texinfo writers (#1924, Tim Lin).

  * HTML writer:

    + Add "inline" or "display" class to math spans (#1914).
      This allows inline and display math to be styled differently.
    + Include raw latex blocks if `--mathjax` specified (#1938).
    + Require highlighting-kate >= 0.5.14 (#1903).
      This ensures that all code blocks will be wrapped in a `div`
      with class `sourceCode`.  Also, the default highlighting CSS
      now adds `div.sourceCode { x-overflow: auto; }`, which means
      that code blocks (even with line numbers) will acquire a scroll
      bar on screens too small to display them (e.g. mobile phones).
      See also jgm/highlighting-kate#65.

  * LaTeX writer:

    + Use a declaration for tight lists (Jose Luis Duran, Joseph
      Harriott). Previously, pandoc hard-coded some commands to make
      tight lists in LaTeX.  Now we use a custom command instead,
      allowing the styling to be changed in a macro in the header.
      (Note:  existing templates may need to be modified to include
      the definition of this macro.  See the current template.)
    + Beamer output: if the header introducing a slide has the
      class `fragile`, add the `[fragile]` option to the slide (#2119).

  * MediaWiki writer:

    + Use `File:` instead of the deprecated `Image:` for images and
      other media files (Greg Rundlett).

  * DocBook writer:

    + Render a `Div (id,_,_) [Para _]` element as a `para` element
      with an `id` attribute.  This makes links to citations work in
      DocBook with pandoc-citeproc.

  * RST writer:

    + Normalize headings to sequential levels (Nikolay Yakimov).
      This is pretty much required by docutils.
    + Treat headings in block quotes, etc as rubrics (Nikolay Yakimov).
    + Better handling of raw latex inline (#1961).  We use
      `` :raw-latex:`...` `` and add a definition for this role to
      the template.

  * EPUB writer:

    + Remove `linear=no` from cover `itemref` (#1609).
    + Don't use `sup` element for epub footnotes (#1995).
      Instead, just use an a element with class `footnoteRef`.
      This allows more styling options, and provides better results
      in some readers (e.g. iBooks, where anything inside the a
      tag breaks popup footnotes).
    + Take TOC title from `toc-title` metadata field.

  * Docx writer:

    + Implemented `FirstParagraph` style (Jesse Rosenthal).
      Following the ODT writer, we add the `FirstParagraph` style to the
      first text paragraph following an image, blockquote, table, heading,
      or beginning of document.  This allows it to be styled differently.
      The default is for it to be the same as `Normal`.
    + Added `BodyText` style (Jesse Rosenthal).
      We apply a `BodyText` style to all unstyled paragraphs. This is,
      essentially, the same as `Normal`, except that since not everything
      inherits from `BodyText` (the metadata won't, for example, or
      the headers or footnote numbers), we can change the text in the body
      without having to make exceptions for everything.  If we do want to
      change *everything*, we can still do it through `Normal`.
    + Altered `Blockquote` style slightly (Jesse Rosenthal).
      Since `BlockQuote` derives from `BodyText`, we just want to specify
      by default that it won't indent, regardless of what `BodyText` does.
      Note that this will not produce any visible difference in the default
      configuration.
    + Take TOC title from `toc-title` metadata field (Nikolay Yakimov).
    + Added a style to figure images (Nikolay Yakimov).
      Figures with empty captions use style `Figure`.
      Figures with nonempty captions use style `Figure with Caption`, which
      is based on `Figure`, and additionally has `keepNext` set.

  * ODT writer:

    + Added figure captions (Nikolay Yakimov). The following styles are
      used for figures:
      `Figure` -- for figure with empty caption),
      `FigureWithCaption` (based on `Figure`) -- for figure with caption,
      `FigureCaption` (based on `Caption`) -- for figure captions.
      Also, `TableCaption` (based on `Caption`) is used for table captions.

  [API changes]

  * New `Text.Pandoc.Error` module with `PandocError` type
    (Matthew Pickering).

  * All readers now return `Either PandocError Pandoc` instead of `Pandoc`
    (Matthew Pickering).  This allows better handling of errors.

  * Added `Text.Pandoc.Writers.CommonMark`, exporting `writeCommonMark`.

  * Added `Text.Pandoc.Readers.CommonMark`, exporting `readCommonMark`.

  * Derive `Data` and `Typeable` instances for `MediaBag`, `Extension`,
    `ReaderOptions`, `EPUBVersion`, `CiteMethod`, `ObfuscationMethod`,
    `HTMLSlideVariant`, `TrackChanges`, `WriterOptions` (Shabbaz
    Youssefi).

  * New `Ext_shortcut_reference_links` constructor for `Extension`
    (Konstantin Zudov).

  [bug fixes]

  * Markdown reader:

    + Allow smart `'` after inline math (#1909, Nikolay Yakimov).
    + Check for tex macros after indented code (#1973).
    + Rewrote `charsInBalancedBrackets` for efficiency.
    + Make sure a closing `</div>` doesn't get included in a
      definition list item (#2127).
    + Don't parse bracketed text as citation if it might be a link,
      image, or footnote (Nikolay Yakimov).
    + Require space after key in mmd title block (#2026, Nikolay
      Yakimov).  Require space after key-value delimiter colon in mmd title
      block.
    + Require nonempty value in mmd title block (Nikolay Yakimov).
    + Disable all metadata block extensions when parsing
      metadata field values (#2026, Nikolay Yakimov).  Otherwise we
      could get a mmd title block inside YAML metadata, for example.

  * HTML reader:

    + Improve self-closing tag detection in `htmlInBalanced` (#2146).
    + Handle tables with `<th>` in body rows (#1859, Mauro Bieg).
    + Fixed `htmlTag` (#1820).  If the tag parses as a comment, we check
      to see if the input starts with `<!--`. If not, it's bogus comment
      mode and we fail `htmlTag`.
    + Handle `base` tag; if it has an `href` value, this is added to
      all relative URLs in links and images.

  * DocBook reader:

    + Look inside "info" elements for section titles (#1931).

  * Docx reader:

    + Parse images in deprecated vml format (Jesse Rosenthal).
    + Allow sub/superscript verbatims (Jesse Rosenthal).
      Verbatim usually shuts off all other run styles, but we don't want it
      to shut off sub/superscript.

  * LaTeX reader:

    + Handle `tabular*` environment (#1850).
      Note that the table width is not actually parsed or taken into
      account, but pandoc no longer chokes on it.
    + Ignore options in `\lstinline` rather than raising error (#1997).
    + Add some test cases for simple tables (Mathias Schenner).
    + Handle valign argument in tables (Mathias Schenner) (currently
      we just ignore this).
    + Allow non-empty colsep in tables (Mathias Schenner).
      The `tabular` environment allows non-empty column separators
      with the "@{...}" syntax. Previously, pandoc would fail to
      parse tables if a non-empty colsep was present. With this
      commit, these separators are still ignored, but the table gets
      parsed. A test case is included.
    + Recognize `\newpage` as a block command.
    + Allow block content in \title{} (#2001).
    + Check for block-level newcommand aliases in blockCommand (Nikolay
      Yakimov).
    + Guard against paragraph starting with inline macro (Nikolay Yakimov).
    + Properly gobble spaces after `\\` (#2007).

  * Textile reader:

    +  Handle newlines in table cells, and empty cells (#1919).

  * Org reader:

    + Allow image links with non-image targets (Hans-Peter Deifel).
      This matches behavior of Org-Mode for links like
      `[[http://example.com][https://www.haskell.org/static/img/logo.png]]`.

  * Docbook writer:

    + Don't print empty id attributes (thanks to Steve Horne).

  * HTML writer:

    + Fixed list-style-type for numbered example lists.
      Should be "decimal," not "example" (#1902).
    + Do not omit missing `alt` attribute on `img` tag (#1131,
      Konstantin Zudov).
    + Allow multiple colgroups in table (#2122).
    + In revealjs, ensure that lists in speaker notes don't add "fragment"
      classes, which can cause additional keypresses to be needed to
      advance a slide (#1394).

  * LaTeX writer:

    + Don't escape `$` in URL (#1913).
    + Don't use listings in headers (Matthew Pickering, #1963).
    + Recognize book documentclass if set in metadata (#1971).
      This sets `--chapters` implicitly if the documentclass in metadata
      is a book documentclass.  Previously this was done only if a book
      documentclass was set in a variable.
    + Add a `\label` in `\hyperdef` for Div, Span (or links don't work).
    + Make `mainlang` work when `lang` is in metadata (#2174).

  * Texinfo writer:

    + Fix wrapping by using breakable spaces (Tim Lin).

  * RST writer:

    + Fixed toc depth in RST writer.  Previously the depth was being
      rendered as a floating point number with a decimal point.

  * Markdown writer:

    + Improved escaping (#2086).  `<` should not be escaped as `\<`, for
      compatibility with original Markdown.  We now escape `<` and `>`
      with entities.  Also, we now backslash-escape square brackets.
    + Avoid introducing spurious list items through wrapping (#1946).
    + Don't emit span tags if plain or raw HTML disabled.

  * MediaWiki writer:

    + Convert spaces to underscores in wikilink URL (#1982), like MediaWiki.

  * AsciiDoc writer:

    + Insert some needed blank lines (#1860).
    + Avoid wrapping after list marker (#1858).

  * EPUB writer:

    + Properly handle internal links to IDs in spans, divs (#1884).
    + Use plain writer for metadata dc: fields (#2121).
      This gives better results when we have, e.g. multiple paragraphs.
      Note that tags aren't allowed in these fields.
    + Properly handle image links without an extension (#1855).
    + Improved chapter splitting and internal link rewriting (#1887,
      #2162, #2163).  This will ensure that internal links work and
      that the references section produced by pandoc-citeproc is
      in its own chapter.
    + Fixed handling of svg images (#2183).

  * ICML writer:

    + Better handling of raw blocks and inlines (#1951).
      Previously these were always escaped and printed verbatim.
      Now they are ignored unless the format is `icml`, in which
      case they are passed through unescaped.
    + Fixed image URIs in ICML output (gohai).

  * Custom writer:

    + Raise error if loadstring returns an error status.
    + Raise `PandocLuaException` instead of using 'error'.
      Eventually we'll change the return type so that no exception
      is involved, but at least this can be trapped.
    + Use UTF-8 aware bytestring conversion.
    + Set foreign encoding to UTF-8 (Nikolay Yakimov, #2101, #1634).
      Also factored out ByteString, since it's only used as an intermediate
      representation.

  * Docx writer:

    + Copy hyphenation settings from reference.docx (Nikolay Yakimov).
    + Filter out illegal XML characters (#1992, Matthew Pickering).
    + Added `noProof` to docx syntax highlighting `SourceCode` style.
    + Added footnotes id -1 and 0 (Jesse Rosenthal).
      Word uses, by default, footnotes with id -1 and 0 for separators. If a
      user modifies `reference.docx`, they will end up with a `settings.xml`
      file that references these footnotes, but no such footnotes in the
      document. This will produce a corruption error. Here we add these to the
      document and `settings.xml` file, so future modifications won't break
      the file.
    + Handle lists correctly inside table cells (Jesse Rosenthal).
      Previously we didn't transform lists inside table cells.
    + Set firstRow information in tables (Nikolay Yakimov).
    + Don't replace `SourceCode` style in `reference.docx` if it is defined
      there (Nikolay Yakimov, #1872).  If `--no-highlight` specified, remove
      any `SourceCode` and `*Tok` styles in `reference.docx`.
    + Attempt to match international style names (#1607, Nikolay Yakimov).
    + Set these styles as custom (Nikolay Yakimov): `Author`, `Abstract`,
      `Compact`, `Image Caption`, `Table Caption`, `Definition Term`,
      `Definition`, `First Paragraph`.
    + Rename these styles to correspond with Word `Normal.dotm` (Nikolay
      Yakimov): `Block Quote -> Block Text`, `Link -> Hyperlink`,
      `Footnote Ref -> Footnote Reference`.
    + Added `Caption` style (Nikolay Yakimov).
    + Changed these styles' inheritance (Nikolay Yakimov):
      `Image Caption <- Caption`, `Table Caption <- Caption`.
    + Remove `SourceCode` style from `reference.docx` (#1872).
      This is added automatically by the docx writer.
    + Added toc heading style to `reference.docx` (Nikolay Yakimov).

  * `Text.Pandoc.PDF`

    + Don't suggest "Try xelatex" if xelatex already in use (Mauro Bieg,
      #1832).
    + More comprehensible errors on image conversion (#2067).
      EPS can't be supported without shelling out to something like
      ImageMagick, but at least we can avoid mysterious error messages.

  * `Text.Pandoc.Shared`:

    + Make safeRead safe (#1801, Matthew Pickering).
    + Addded `mapLeft`, `hush` (Matthew Pickering).

  * `Text.Pandoc.Pretty`:

    + Remove partial function (Matthew Pickering).

  * `Text.Pandoc.SelfContained`:

    + Add `;charset=utf-8` to script mime type if missing (#1842).
    + Improved building of data URIs (#1940).  Now base64 is used except
      for `text/*` mime types.
    + `cssURLs` no longer tries to fetch fragment URLs (#2121).
    + Properly handle data URIs in css urls (#2129).
      Use a proper CSS parser (adds dependency on `text-css`).

  * `Text.Pandoc.UTF8`:

    + Better handling of bare CRs in input files (#2132).
      Previously we just stripped them out; now we convert
      other line ending styles to LF line endings.

  * `Text.Pandoc.ImageSize`:

    + Fixed some exif header parsing bugs (#1834).
    + Make imageSize return an Either, not a Maybe (#1834).
      Use `runGetOrFail` (with `binary >= 0.7`) to return `Left` on
      parse failure (rather than `error`).
    + Improved warnings when image size can't be determined.
    + Removed error landmines (Matthew Pickering).

  * Added woff2 to MIME types (Alfred Wechselberger).

  * pandoc:  When a binary input format is used, warn that file
    arguments past the first one are being ignored (Matthew Pickering).

  [template changes]

  * LaTeX template:

    + Degrade gracefully if `\paragraph` not defined.
    + Include `grffile` together with `graphicx` (#2074).
      This properly handles filenames containing spaces and dots.
    + Redefine `\paragraph`, `\subparagraph`...  to behave more
      like section headers (#1658).
    + Import hyperref before polyglossia to avoid an error with xelatex,
      "please load package hyperref before bidi package" (Nick Bart).
    + Added `toccolor` variable to control link color in toc (Kaixhin).

  * LaTeX, Beamer templates:

    + Provide `\tightlist`, which is now used by the LaTeX writer.
    + Use polyglossia in beamer (#85).
    + Use `bibliography` instead of `biblio-files`
      (#1661).  Also use `\addbibresource` instead of `\bibliography` for
      biblatex.
    + Added `setotherlanguages` in polyglossia. This uses an `otherlang`
      variable that is derived from a comma-separated list in `lang`;
      the last language is `mainlang` and the others are `otherlang`.

  * EPUB templates:

    + Use `div`, not `p`, for "rights" on title page.
    + Added header-includes, include-before, include-after (#1987).

  * OpenDocument template:

    + Use `text:p` instead of `text:h` for title.
      Using `text:h` causes problems with numbering.  Closes #2059.
      Thansk to @nkalvi for diagnosing this.

  * reveal.js template:

    + Link to non-minified css, js.  The minified versions no longer
      ship with the library.
    + Correctly include style CSS (#1949).
    + New configurable options options: `center`, `maxScale`, `slideNuber`
      (Dmitry Smirnov, pandoc-templates#89).
    + Moved custom CSS after theme.  This allows custom CSS to modify
      themes, instead of being replaced by them.
    + Allow `center` to be set to false.

  [under the hood improvements]

  * Removed pre-built `reference.docx` and `reference.odt` (Nikolay
    Yakimov).  Instead the repository now includes the component text files,
    and the zipped binaries are built from these using a helper
    program, `make-reference-files`.  This should make maintenance of
    these components easier going forward.

  * `Text.Pandoc.Parsing`:

    + Added new `<+?>` combinator (Nikolay Yakimov).
    + Added `stateHeaderKeys` to `ParserState`.

  * `make_deb.sh` fixes:

    + Detect architecture.
    + Add Installed-Size to debian package control file (#1900).
    + Use `fakeroot` to get permissions right.
    + Use `mkdir` and `cp` instead of `install`.
    + Set permissions of directories to 755.
    + Install in `/usr` rather than `/usr/local`.
    + Compress man pages.
    + Combine copyright files for `pandoc`, `pandoc-citeproc`.

  * Added `Text.Pandoc.Compat.Locale` and `old-locale` flag
    to assist with transition to `time` 1.5.

  * Updated CONTRIBUTING.md with information about issue tags (Matthew
    Pickering).

  * Updated travis installs to the new sudo-less syntax (Tim Lin).

  * Updated dependency version bounds.

  * EPUB tests:  don't use `joinPath`, which varies across platforms.
    Instead, use a forward-slash to join paths, regardless of the
    platform. This matches the way `MediaBag` now works.

  * Clarify JSON input and output in usage message (Caleb McDaniel).

  * Improved INSTALL instructions.

  * Always build man pages.  Removed make-pandoc-man-pages flag.

  * Makefile:  removed man target, now that we generate man pages by default.

  * README:

    + Fixed typos (J. Lewis Muir).
    + Added documentation on backtick_code_blocks (#2135, Nikolay Yakimov).
    + Added note on in-field markup in biblio databases (Nick Bart).
    + Fixed misleading example of raw HTML block.
    + Various minor formatting and consistency fixes for the program
      options (Andreas Lööw).
    + Made definition lists for options all "loose" for consistency.
    + Added YAML biblio format to table, and note on `pandoc-citeproc`'s
      `--bib2json` and `--bib2yaml` options (Nick Bart).
    + Removed obsolete reference to `mods2yaml` (Nick Bart).
    + Added section on syntax highlighting.
    + Documented `toccolor` variable.

pandoc (1.13.2.1)

  * Updated to build with ghc 7.10.1.

  * Bumped package upper bounds for filepath, blaze-html,
    blaze-markup.

pandoc (1.13.2)

  * TWiki Reader: add new new twiki reader (API change, Alexander Sulfrian).

  * Markdown reader:

    + Better handling of paragraph in div (#1591).
      Previously text that ended a div would be parsed as Plain
      unless there was a blank line before the closing div tag.
    + Don't treat a citation as a reference link label (#1763).
    + Fix autolinks with following punctuation (#1811).
      The price of this is that autolinked bare URIs can no longer
      contain `>` characters, but this is not a big issue.
    + Fix `Ext_lists_without_preceding_blankline` bug (#1636, Artyom).
    + Allow `startnum` to work without `fancy_lists`. Formerly
      `pandoc -f markdown-fancy_lists+startnum` did not work properly.

  * RST reader (all Daniel Bergey):

    + Parse quoted literal blocks (#65).  RST quoted literal blocks are
      the same as indented literal blocks (which pandoc already supports)
      except that the quote character is preserved in each line.
    + Parse RST class directives. The class directive accepts one or more
      class names, and creates a Div value with those classes.  If the
      directive has an indented body, the body is parsed as the children of
      the Div.  If not, the first block folowing the directive is made a
      child of the Div. This differs from the behavior of rst2xml, which
      does not create a Div element.  Instead, the specified classes are
      applied to each child of the directive.  However, most Pandoc Block
      constructors to not take an Attr argument, so we can't duplicate this
      behavior.
    + Warn about skipped directives.
    + Literal role now produces Code. Code role should have "code" class.
    + Improved support for custom roles

        - Added `sourceCode` to classes for `:code:` role, and anything
          inheriting from it.
        - Add the name of the custom role to classes if the Inline
          constructor supports Attr.
        - If the custom role directive does not specify a parent role,
          inherit from the `:span:` role.

      This differs somewhat from the `rst2xml.py` behavior.  If a custom
      role inherits from another custom role, Pandoc will attach both
      roles' names as classes.  `rst2xml.py` will only use the class of
      the directly invoked role (though in the case of inheriting from a
      `:code:` role with a `:language:` defined, it will also provide the
      inherited language as a class).
    + Warn about ignored fields in role directives.

  * LaTeX reader:

    + Parse label after caption into a span instead of
     inserting an additional paragraph of bracketed text (#1747).
    + Parse math environments as inline when possible (#1821).
    + Better handling of `\noindent` and `\greektext` (#1783).
    + Handle `\texorpdfstring` more gracefully.
    + Handle `\cref` and `\sep` (Wikiwide).
    + Support `\smartcite` and `\Smartcite` from biblatex.

  * HTML reader:

    + Retain display type of MathML output (#1719, Matthew Pickering).
    + Recognise `<br>` tags inside `<pre>` blocks (#1620, Matthew Pickering).
    + Make `embed` tag either block or inline (#1756).

  * DocBook reader:

    + Handle `keycombo`, `keycap` (#1815).
    + Get string content in inner tags for literal elements (#1816).
    + Handle `menuchoice` elements better, with a `>` between (#1817).
    + Include `id` on section headers (#1818).
    + Document/test "type" as implemented (Brian O'Sullivan).
    + Add support for calloutlist and callout (Brian O'Sullivan).
      We treat a calloutlist as a bulleted list. This works well in practice.
    + Add support for `classname` (Bryan O'Sullivan).

  * Docx reader:

    + Fix window path for image lookup (Jesse Rosenthal).
      Don't use os-sensitive "combine", since we always want the paths in our
      zip-archive to use forward-slashes.
    + Single-item headers in ordered lists are headers (Jesse Rosenthal).
      When users number their headers, Word understands that as a single item
      enumerated list. We make the assumption that such a list is, in fact,
      a header.
    + Rewrite rewriteLink to work with new headers (Jesse Rosenthal).
      There could be new top-level headers after making lists, so we have to
      rewrite links after that.
    + Use polyglot header list (Jesse Rosenthal).
      We're just keeping a list of header formats that different languages
      use as their default styles. At the moment, we have English, German,
      Danish, and French. We can continue to add to this.
      This is simpler than parsing the styles file, and perhaps less
      error-prone, since there seems to be some variations, even within a
      language, of how a style file will define headers.
    + Remove header class properly in other langs (Jesse Rosenthal).
      When we encounter one of the polyglot header styles, we want to remove
      that from the par styles after we convert to a header. To do that, we
      have to keep track of the style name, and remove it appropriately.
    + Account for external link URLs with anchors. Previously, if a URL
      had an anchor, the reader would incorrectly identify it as an
      internal link and return only the anchor as URL. (Caleb McDaniel)
    + Fix for Issue #1692 (i18n styles) (Nikolay Yakimov).

  * Org reader:

    + Added state changing blanklines (Jesse Rosenthal).
      This allows us to emphasize at the beginning of a new paragraph (or, in
      general, after blank lines).
    + Fixed bug with bulleted lists:

        - a
        - b
        * c

      was being parsed as a list, even though an unindented `*`
      should make a heading.  See
      <http://orgmode.org/manual/Plain-lists.html#fn-1>.
    + Org reader: absolute, relative paths in link (#1741, Albert
      Krewinkel). The org reader was too restrictive when parsing links;
      some relative links and links to files given as absolute paths
      were not recognized correctly.
    + Org reader:  allow empty links (jgm/gitit#471, Albert Krewinkel).
      This is important for use in gitit, which uses empty links
      for wikilinks.
    + Respect indent when parsing Org bullet lists (#1650, Timothy
      Humphries).  Fixes issue with top-level bullet list parsing.
    + Fix indent issue for definition lists (Timothy Humphries,
      see #1650, #1698, #1680).
    + Parse multi-inline terms correctly in definition list (#1649,
      Matthew Pickering).
    + Fix rules for emphasis recognition (Albert Krewinkel).
      Things like `/hello,/` or `/hi'/` were falsy recognized as emphasised
      strings.  This is wrong, as `,` and `'` are forbidden border chars and
      may not occur on the inner border of emphasized text.
    + Drop COMMENT document trees (Albert Krewinkel).
      Document trees under a header starting with the word `COMMENT` are
      comment trees and should not be exported.  Those trees are dropped
      silently (#1678).
    + Properly handle links to `file:target` (Albert Krewinkel).
      Org links like `[[file:target][title]]` were not handled correctly,
      parsing the link target verbatim.  The org reader is changed such that
      the leading `file:` is dropped from the link target (see #756, #1812).
    + Parse LaTeX-style MathML entities (#1657, Albert Krewinkel).
      Org supports special symbols which can be included using LaTeX syntax,
      but are actually MathML entities.  Examples for this are
      `\nbsp` (non-breaking space), `\Aacute` (the letter A with accent acute)
      or `\copy` (the copyright sign ©)

  * EPUB reader:

    + URI handling improvements. Now we outsource most of the work to
      `fetchItem'`. Also, do not include queries in file extensions (#1671).

  * LaTeX writer:

    + Use `\texorpdfstring` for section captions when needed (Vaclav Zeman).
    + Handle consecutive linebreaks (#1733).
    + Protect graphics in headers (Jesse Rosenthal).
      Graphics in `\section`/`\subsection` etc titles need to be `\protect`ed.
    + Put `~` before header in list item text (Jesse Rosenthal).
      Because of the built-in line skip, LaTeX can't handle a section header
      as the first element in a list item.
    + Avoid using reserved characters as `\lstinline` delimiters (#1595).
    + Better handling of display math in simple tables (#1754).
      We convert display math to inline math in simple tables,
      since LaTeX can't deal with display math in simple tables.
    + Escape spaces in code (#1694, Bjorn Buckwalter).

  * MediaWiki writer:

    + Fixed links with URL = text. Previously these were rendered as bare
      words, even if the URL was not an absolute URL (#1825).

  * ICML writer:

    + Don't force all citations into footnotes.

  * RTF writer:

    + Add blankline at end of output (#1732, Matthew Pickering).

  * RST writer:

    + Ensure blank line after figure.
    + Avoid exces whitespace after last list item (#1777).
    + Wrap line blocks with spaces before continuations (#1656).
    + Fixed double-rendering of footnotes in RST tables (#1769).

  * DokuWiki writer:

    + Better handling of block quotes. This change ensures that
      multiple paragraph blockquotes are rendered using native `>`
      rather than as HTML (#1738).
    + Fix external images (#1739). Preface relative links with ":",
      absolute URIs without. (Timothy Humphries)

  * HTML writer:

    + Use protocol-relative URL for mathjax.
    + Put newline btw img and caption paragraph.
    + MathML now outputted with tex annotation (#1635, Matthew Pickering).
    + Add support for KaTeX HTML math (#1626, Matthew Pickering).
      This adds `KaTeX` to `HTMLMathMethod` (API change).
    + Don't double render when `email-obfuscation=none` (#1625, Matthew
      Pickering).
    + Make header attributes work outside top level (#1711).
      Previously they only appeared on top level header elements.
      Now they work e.g. in blockquotes.

  * ODT writer:

    + Correctly handle images without extensions (#1729).
    + Strip querystring in ODT write (#1682, Todd Sifleet).

  * FB2 writer:

    + Add newline to output.

  * EPUB writer:

    + Don't add `sourceURL` to absolute URIs (#1669).
    + Don't use unsupported `opf:title-type` for epub2.
    + Include "landmarks" section in nav document for epub3 (#1757).
    + Removed playOrder from navpoint elements in ncx file (#1760).
      These aren't required, and they make manual modification of epubs
      difficult.
    + Extract title even from structured title.
    + Don't include nav node in spine unless `--toc` was requested.
      Previously we included it in the spine with `linear="no"`, leading
      to odd results in some readers (#1593).
    + Fixed absolute URI detection (#1672).
    + Correctly resolve relative URIs (#1671).
    + Use regular page template for `nav.xhtml`, including doctype (#1759).

  * Docx writer:

    + Put docx table captions above tables (#1641, Nikolay Yakimov).
    + Get the page width from the reference docx file, and use
      it to scale images that are too large to fit (Grégory Bataille).
    + Partial fix for #1607 (Nikolay Yakimov). International heading styles
      are inferred based on `<w:name val="heading #">` fallback, if there
      are no en-US "Heading#" styles
    + Look in user data dir for archive `reference.docx`.
    + Renumber header and footer relationships to avoid collisions (Jesse
      Rosenthal). We previously took the old relationship names of the
      headers and footer in secptr. That led to collisions. We now make
      a map of availabl names in the relationships file, and then rename
     in secptr.

  * ConTeXt writer:

    + Add function toLabel (Mark Szepieniec).
      This function can be used to sanitize reference labels so that
      they do not contain any of the illegal characters \#[]",{}%()|= .
      Currently only Links have their labels sanitized, because they
      are the only Elements that use passed labels.

  * `Text.Pandoc.Shared`:

    + Moved import of `toChunks` outside of CPP conditional (#1590).
    + Fix `inDirectory` to reset to the original directory in case
      an exception occurs (Freiric Barral).

  * Templates:

    + LaTeX template: load polyglossia before bibtex (jgm/pandoc-templates#70).
      Thanks to bluebirch.
    + LaTeX template: Added `\VerbatimFootnotes` if there is verbatim in notes
      (#1616).
    + LaTeX template:  Add shorthands=off to babel options (#1648).
    + EPUB, EPUB3 templates: Added `id="cover"` to body of cover page.
      This aids styling, making it possible for example to set 0 margins
      on the title page (#1758).
    + EPUB, EPUB3 templates:  Handle structured metadata on titlepage.
      Previously we just expected 'title', 'subtitle', 'author', 'date'.
      Now we still support those, but also support the format recommended
      for epub metadata in the pandoc README:

            ---
            title:
            - type: main
              text: My Book
            - type: subtitle
              text: An investigation of metadata
            creator:
            - role: author
              text: John Smith
            - role: editor
              text: Sarah Jones
            identifier:
            - scheme: DOI
              text: doi:10.234234.234/33
            publisher:  My Press
            rights:  (c) 2007 John Smith, CC BY-NC
            ...

  * `Text.Pandoc.Templates.getDefaultTemplate`:
    don't fail when called with "fb2" (#1660).

  * `Text.Pandoc.Parsing`:

    + Fixed `inlineMath` so it handles `\text{..}` containing `$`.
      For example: `$x = \text{the $n$th root of $y$}` (#1677).
    + Change `parseFromString` to fail if not all input is consumed.
      (Matthew Pickering)
    + Moved `addWarning` from Markdown reader to `Parsing`, so it can be
      used by more readers (API change, Daniel Bergey).

  * `Text.Pandoc.Pretty`:

    +  Improve performance of `realLength` (Matthew Pickering).
    +  Make CR + BLANKLINE = BLANKLINE. This fixes an extra blank line we
       were getting at the end of markdown fragments (as well as rst, org,
       etc.) (#1705).

  * `Text.Pandoc.MIME`:

    + Add mime type for WebVTT (Jason Ronallo).
    + Changed mime type for `otf` to `application/vnd.ms-opentype` (#1761).
      This is needed for epub3 validation.

  * `Text.Pandoc.MediaBag`:

    + Fix Windows specific path problems (#1597).

  * `Text.Pandoc.Shared`:

    + Make `collapseFilePath` OS-agnostic (Matthew Pickering).

  * Link the test suite using `-threaded`.
    This allows the test suite to be run using `+RTS -N`.

  * Added `network` dependency under `network-uri` flag in test section.

  * Give better error messages when someone tries to convert from
    pdf, doc, odt (#1683).

  * Added `track` to list of tags treated by `--self-contained` (#1664).

pandoc (1.13.1)

  * Fixed `--self-contained` with Windows paths (#1558).
    Previously `C:\foo.js` was being wrongly interpreted as a URI.

  * HTML reader:  improved handling of tags that can be block or inline.
    Previously a section like this would be enclosed in a paragraph,
    with RawInline for the video tags (since video is a tag that can
    be either block or inline):

        <video controls="controls">
           <source src="../videos/test.mp4" type="video/mp4" />
           <source src="../videos/test.webm" type="video/webm" />
           <p>
              The videos can not be played back on your system.<br/>
              Try viewing on Youtube (requires Internet connection):
              <a href="http://youtu.be/etE5urBps_w">Relative Velocity on
        Youtube</a>.
           </p>
        </video>

    This change will cause the video and source tags to be parsed
    as RawBlock instead, giving better output.
    The general change is this:  when we're parsing a "plain" sequence
    of inlines, we don't parse anything that COULD be a block-level tag.

  * Docx reader:

    + Be sensitive to user styles.  Note that "Hyperlink" is
      "blacklisted," as we don't want the default underline styling to be
      inherited by all links by default (Jesse Rosenthal).
    + Read single paragraph in table cell as `Plain` (Jesse Rosenthal).
      This makes to docx reader's native output fit with the way the markdown
      reader understands its markdown output.

  * Textile writer:  Extended the range of cases where native textile
    tables will be used (as opposed to raw HTML):  we now handle any
    alignment type, but only for simple tables with no captions.

  * Txt2Tags reader:

    + Header is now parsed only if standalone flag is set (Matthew Pickering).
    + The header is now parsed as meta information. The first line is the
      `title`, the second is the `author` and third line is the `date`
      (Matthew Pickering).
    + Corrected formatting of `%%mtime` macro (Matthew Pickering).
    + Fixed crash when reading from stdin.

  * EPUB writer:  Don't use page-progression-direction in EPUB2, which
    doesn't support it.  Also, if page-progression-direction not specified
    in metadata, don't include the attribute even in EPUB3; not including it
    is the same as including it with the value "default", as we did before.
    (#1550)

  * Org writer: Accept example lines with indentation at the beginning
    (Calvin Beck).

  * DokuWiki writer:

    + Refactor to use Reader monad (Matthew Pickering).
    + Avoid using raw HTML in table cells; instead, use `\\`
      instead of newlines (Jesse Rosenthal).
    + Properly handle HTML table cell alignments, and use spacing
      to make the tables look prettier (#1566).

  * Docx writer:

    + Bibliography entries get `Bibliography` style (#1559).
    + Implement change tracking (Jesse Rosenthal).

  * LaTeX writer:

    + Fixed a bug that caused a table caption to repeat across all pages
      (Jose Luis Duran).
    + Improved vertical spacing in tables and made it customizable using
      standard lengths set by booktab.  See
      <https://groups.google.com/forum/#!msg/pandoc-discuss/qMu6_5lYy0o/ZAU7lzAIKw0J>
      (Jose Luis Duran).
    + Added `\strut` to fix spacing in multiline tables (Jose Luis Duran).
    + Use `\tabularnewline` instead of `\\` in table cells (Jose Luis Duran).
    + Made horizontal rules more flexible (Jose Luis Duran).

  * Text.Pandoc.MIME:

    + Added `MimeType` (type synonym for `String`) and `getMimeTypeDef`.
      Code cleanups (Artyom Kazak).

  * Templates:

    + LaTeX template: disable microtype protrusion for typewriter font (#1549,
      thanks lemzwerg).

  * Improved OSX build procedure.

  * Added `network-uri` flag, to deal with split of `network-uri` from
    `network`.

  * Fix build dependencies for the `trypandoc` flag, so that they are
    ignored if `trypandoc` flag is set to False (Gabor Pali).

  * Updated README to remove outdated claim that `--self-contained`
    looks in the user data directory for missing files.

pandoc (1.13.0.1)

  * Docx writer:

    + Fixed regression which bungled list numbering (#1544), causing
      all lists to appear as basic ordered lists.
    + Include row width in table rows (Christoffer Ackelman, Viktor Kronvall).
      Added a property to all table rows where the sum of column widths
      is specified in pct (fraction of 5000).  This helps persuade Word
      to lay out the table with the widths we specify.

  * Fixed a bug in Windows 8 which caused pandoc not to find the
    `pandoc-citeproc` filter (#1542).

  * Docx reader: miscellaneous under-the-hood improvements (Jesse Rosenthal).
    Most significantly, the reader now uses Builder, leading to some
    performance improvements.

  * HTML reader:  Parse appropriately styled span as SmallCaps.

  * Markdown writer: don't escape `$`, `^`, `~` when `tex_math_dollars`,
    `superscript`, and `subscript` extensions, respectively, are
    deactivated (#1127).

  * Added `trypandoc` flag to build CGI executable used in the online
    demo.

  * Makefile:  Added 'quick', 'osxpkg' targets.

  * Updated README in templates to indicate templates license.
    The templates are dual-licensed, BSD3 and GPL2+.

pandoc (1.13)

  [new features]

  * Added `docx` as an input format (Jesse Rosenthal).  The docx
    reader includes conversion of native Word equations to pandoc
    LaTeX `Math` elements.  Metadata is taken from paragraphs at the
    beginning of the document with styles `Author`, `Title`, `Subtitle`,
    `Date`, and `Abstract`.

  * Added `epub` as an input format (Matthew Pickering).  The epub
    reader includes conversion of MathML to pandoc LaTeX `Math`
    elements.

  * Added `t2t` (Txt2Tags) as an input format (Matthew Pickering).
    Txt2tags is a lightweight markup format described at
    <http://txt2tags.org/>.

  * Added `dokuwiki` as an output format (Clare Macrae).

  * Added `haddock` as an output format.

  * Added `--extract-media` option to extract media contained in a zip
    container (docx or epub) while adjusting image paths to point to the
    extracted images.

  * Added a new markdown extension, `compact_definition_lists`, that
    restores the syntax for definition lists of pandoc 1.12.x, allowing
    tight definition lists with no blank space between items, and
    disallowing lazy wrapping.  (See below under behavior changes.)

  * Added an extension `epub_html_exts` for parsing HTML in EPUBs.

  * Added extensions `native_spans` and `native_divs` to activate
    parsing of material in HTML span or div tags as Pandoc Span
    inlines or Div blocks.

  * `--trace` now works with the Markdown, HTML, Haddock, EPUB,
    Textile, and MediaWiki readers.  This is an option intended
    for debugging parsing problems; ordinary users should not need
    to use it.

  [behavior changes]

  * Changed behavior of the `markdown_attribute` extension, to bring
    it in line with PHP markdown extra and multimarkdown.  Setting
    `markdown="1"` on an outer tag affects all contained tags,
    recursively, until it is reversed with `markdown="0"` (#1378).

  * Revised markdown definition list syntax (#1429).  Both the reader
    and writer are affected.  This change brings pandoc's definition list
    syntax into alignment with that used in PHP markdown extra and
    multimarkdown (with the exception that pandoc is more flexible about
    the definition markers, allowing tildes as well as colons).  Lazily
    wrapped definitions are now allowed.  Blank space is required
    between list items.  The space before a definition is used to determine
    whether it is a paragraph or a "plain" element.  **WARNING: This change
    may break existing documents!**  Either check your documents for
    definition lists without blank space between items, or use
    `markdown+compact_definition_lists` for the old behavior.

  * `.numberLines` now works in fenced code blocks even if no language
    is given (#1287, jgm/highlighting-kate#40).

  * Improvements to `--filter`:

    + Don't search PATH for a filter with an explicit path.
      This fixed a bug wherein `--filter ./caps.py` would run `caps.py` from
      the system path, even if there was a `caps.py` in the working directory.
    + Respect shebang if filter is executable (#1389).
    + Don't print misleading error message.
      Previously pandoc would say that a filter was not found,
      even in a case where the filter had a syntax error.

  * HTML reader:

    + Parse `div` and `span` elements even without `--parse-raw`,
      provided `native_divs` and `native_spans` extensions are set.
      Motivation:  these now generate native pandoc Div and Span
      elements, not raw HTML.
    + Parse EPUB-specific elements if the `epub_html_exts`
      extension is enabled.  These include `switch`, `footnote`,
      `rearnote`, `noteref`.

  * Org reader:

    + Support for inline LaTeX.  Inline LaTeX is now accepted and parsed by the
      org-mode reader.  Both math symbols (like `\tau`) and LaTeX commands (like
      `\cite{Coffee}`), can be used without any further escaping (Albert
      Krewinkel).

  * Textile reader and writer:

    + The `raw_tex` extension is no longer set by default.  You can
      enable it with `textile+raw_tex`.

  * DocBook reader:

    + Support `equation`, `informalequation`, `inlineequation` elements with
      `mml:math` content.  This is converted into LaTeX and put into a Pandoc
      Math inline.

  * Revised `plain` output, largely following the style of Project
    Gutenberg:

    + Emphasis is rendered with `_underscores_`, strong emphasis
      with ALL CAPS.
    + Headings are rendered differently, with space to set them off,
      not with setext style underlines. Level 1 headers are ALL CAPS.
    + Math is rendered using unicode when possible, but without the
      distracting emphasis markers around variables.
    + Footnotes use a regular `[n]` style.

  * Markdown writer:

    + Horizontal rules are now a line across the whole page.
    + Prettier pipe tables.  Columns are now aligned  (#1323).
    + Respect the `raw_html` extension.  `pandoc -t markdown-raw_html`
      no longer emits any raw HTML, including span and div tags
      generated by Span and Div elements.
    + Use span with style for `SmallCaps` (#1360).

  * HTML writer:

    + Autolinks now have class `uri`, and email autolinks have class
      `email`, so they can be styled.

  * Docx writer:

    + Document formatting is carried over from `reference.docx`.
      This includes margins, page size, page orientation, header,
      and footer, including images in headers and footers.
    + Include abstract (if present) with `Abstract` style (#1451).
    + Include subtitle (if present) with `Subtitle` style, rather
      than tacking it on to the title (#1451).

  * Org writer:

    + Write empty span elements with an id attribute as org anchors.
      For example `Span ("uid",[],[]) []` becomes `<<uid>>`.

  * LaTeX writer:

    + Put table captions above tables, to match the conventional
      standard.  (Previously they appeared below tables.)
    + Use `\(..\)` instead of `$..$` for inline math (#1464).
    + Use `\nolinkurl` in email autolinks.  This allows them to be styled
      using `\urlstyle{tt}`.  Thanks to Ulrike Fischer for the solution.
    + Use `\textquotesingle` for `'` in inline code.  Otherwise we get
      curly quotes in the PDF output (#1364).
    + Use `\footnote<.>{..}` for notes in beamer, so that footnotes
      do not appear before the overlays in which their markers appear
      (#1525).
    + Don't produce a `\label{..}` for a Div or Span element.  Do produce
      a `\hyperdef{..}` (#1519).

  * EPUB writer:

    + If the metadata includes `page-progression-direction` (which can be
      `ltr` or `rtl`, the `page-progression-direction` attribute will
      be set in the EPUB spine (#1455).

  * Custom lua writers:

    + Custom writers now work with `--template`.
    + Removed HTML header scaffolding from `sample.lua`.
    + Made citation information available in lua writers.

  * `--normalize` and `Text.Pandoc.Shared.normalize` now consolidate
    adjacent `RawBlock`s when possible.

  [API changes]

  * Added `Text.Pandoc.Readers.Docx`, exporting `readDocx` (Jesse Rosenthal).

  * Added `Text.Pandoc.Readers.EPUB`, exporting `readEPUB` (Matthew
    Pickering).

  * Added `Text.Pandoc.Readers.Txt2Tags`, exporting `readTxt2Tags` (Matthew
    Pickering).

  * Added `Text.Pandoc.Writers.DokuWiki`, exporting `writeDokuWiki`
    (Clare Macrae).

  * Added `Text.Pandoc.Writers.Haddock`, exporting `writeHaddock`.

  * Added `Text.Pandoc.MediaBag`, exporting `MediaBag`, `lookupMedia`,
    `insertMedia`, `mediaDirectory`, `extractMediaBag`.  The docx and epub
    readers return a pair of a `Pandoc` document and a `MediaBag` with
    the media resources they contain.  This can be extracted using
    `--extract-media`.  Writers that incorporate media (PDF, Docx,
    ODT, EPUB, RTF, or HTML formats with `--self-contained`) will look
    for resources in the `MediaBag` generated by the reader, in addition to
    the file system or web.

  * `Text.Pandoc.Readers.TexMath`: Removed deprecated `readTeXMath`.
    Renamed `readTeXMath'` to `texMathToInlines`.

  * `Text.Pandoc`: Added `Reader` data type (Matthew Pickering).
    `readers` now associates names of readers with `Reader`
     structures.  This allows inclusion of readers, like the docx
     reader, that take binary rather than textual input.

  * `Text.Pandoc.Shared`:

    + Added `capitalize` (Artyom Kazak), and replaced uses of
      `map toUpper` (which give bad results for many languages).
    + Added `collapseFilePath`, which removes intermediate `.` and
      `..` from a path (Matthew Pickering).
    + Added `fetchItem'`, which works like `fetchItem` but searches
      a `MediaBag` before looking on the net or file system.
    + Added `withTempDir`.
    + Added `removeFormatting`.
    + Added `extractSpaces` (from HTML reader) and generalized its type
      so that it can be used by the docx reader (Matthew Pickering).
    + Added `ordNub`.
    + Added `normalizeInlines`, `normalizeBlocks`.
    + `normalize` is now `Pandoc -> Pandoc` instead of
      `Data a :: a -> a`.  Some users may need to change their uses of
      `normalize` to the newly exported `normalizeInlines` or
      `normalizeBlocks`.

  * `Text.Pandoc.Options`:

    + Added `writerMediaBag` to `WriterOptions`.
    + Removed deprecated and no longer used `readerStrict` in
      `ReaderOptions`.  This is handled by `readerExtensions` now.
    + Added `Ext_compact_definition_lists`.
    + Added `Ext_epub_html_exts`.
    + Added `Ext_native_divs` and `Ext_native_spans`.
      This allows users to turn off the default pandoc behavior of
      parsing contents of div and span tags in markdown and HTML
      as native pandoc Div blocks and Span inlines.

  * `Text.Pandoc.Parsing`:

    + Generalized `readWith` to `readWithM` (Matthew Pickering).
    + Export `runParserT` and `Stream` (Matthew Pickering).
    + Added `HasQuoteContext` type class (Matthew Pickering).
    + Generalized types of `mathInline`, `smartPunctuation`, `quoted`,
      `singleQuoted`, `doubleQuoted`, `failIfInQuoteContext`,
      `applyMacros` (Matthew Pickering).
    + Added custom `token` (Matthew Pickering).
    + Added `stateInHtmlBlock` to `ParserState`.  This is used to keep
      track of the ending tag we're waiting for when we're parsing inside
      HTML block tags.
    + Added `stateMarkdownAttribute` to `ParserState`. This is used
      to keep track of whether the markdown attribute has been set in
      an enclosing tag.
    + Generalized type of `registerHeader`, using new type classes
      `HasReaderOptions`, `HasIdentifierList`, `HasHeaderMap` (Matthew
      Pickering).  These allow certain common functions to be reused
      even in parsers that use custom state (instead of `ParserState`),
      such as the MediaWiki reader.
    + Moved `inlineMath`, `displayMath` from Markdown reader to Parsing,
      and generalized their types (Matthew Pickering).

  * `Text.Pandoc.Pretty`:

    + Added `nestle`.
    + Added `blanklines`, which guarantees a certain number of blank lines
      (and no more).

  [bug fixes]

  * Markdown reader:

    + Fixed parsing of indented code in list items.  Indented code
      at the beginning of a list item must be indented eight spaces
      from the margin (or edge of the container), or four spaces
      from the list marker, whichever is greater.
    + Fixed small bug in HTML parsing with `markdown_attribute`, which
      caused incorrect tag nesting for input like
      `<aside markdown="1">*hi*</aside>`.
    + Fixed regression with intraword underscores (#1121).
    + Improved parsing of inline links containing quote characters (#1534).
    + Slight rewrite of `enclosure`/`emphOrStrong` code.
    + Revamped raw HTML block parsing in markdown (#1330).
      We no longer include trailing spaces and newlines in the
      raw blocks.  We look for closing tags for elements (but without
      backtracking).  Each block-level tag is its own `RawBlock`;
      we no longer try to consolidate them (though `--normalize` will do so).
    + Combine consecutive latex environments.  This helps when you have
      two minipages which can't have blank lines between them (#690, #1196).
    + Support smallcaps through span.
      `<span style="font-variant:small-caps;">foo</span>` will be
      parsed as a `SmallCaps` inline, and will work in all output
      formats that support small caps (#1360).
    + Prevent spurious line breaks after list items (#1137).  When the
      `hard_line_breaks` option was specified, pandoc would formerly
      produce a spurious line break after a tight list item.
    + Fixed table parsing bug (#1333).
    + Handle `c++` and `objective-c` as language identifiers in
      github-style fenced blocks (#1318).
    + Inline math must have nonspace before final `$` (#1313).

  * LaTeX reader:

    + Handle comments at the end of tables.  This resolves the issue
      illustrated in <http://stackoverflow.com/questions/24009489>.
    + Correctly handle table rows with too few cells.  LaTeX seems to
      treat them as if they have empty cells at the end  (#241).
    + Handle leading/trailing spaces in `\emph` better.
      `\emph{ hi }` gets parsed as `[Space, Emph [Str "hi"], Space]`
      so that we don't get things like `* hi *` in markdown output.
      Also applies to `\textbf` and some other constructions (#1146).
    + Don't assume preamble doesn't contain environments (#1338).
    + Allow (and discard) optional argument for `\caption` (James Aspnes).

  * HTML reader:

    + Fixed major parsing problem with HTML tables.  Table cells were
      being combined into one cell (#1341).
    + Fixed performance issue with malformed HTML tables.
      We let a `</table>` tag close an open `<tr>` or `<td>` (#1167).
    + Allow space between `<col>` and `</col>`.
    + Added `audio` and `source` in `eitherBlockOrInline`.
    + Moved `video`, `svg`, `progress`, `script`, `noscript`, `svg` from
      `blockTags` to `eitherBlockOrInline`.
    + `map` and `object` were mistakenly in both lists; they have been removed
      from `blockTags`.
    + Ignore `DOCTYPE` and `xml` declarations.

  * MediaWiki reader:

    + Don't parse backslash escapes inside `<source>` (#1445).
    + Tightened up template parsing.
      The opening `{{` must be followed by an alphanumeric or `:`.
      This prevents the exponential slowdown in #1033.
    + Support "Bild" for images.

  * DocBook reader:

    + Better handle elements inside code environments.  Pandoc's document
      model does not allow structure inside code blocks, but at least this way
      we preserve the text (#1449).
    + Support `<?asciidoc-br?>` (#1236).

  * Textile reader:

    + Fixed list parsing. Lists can now start without an intervening
      blank line (#1513).
    + HTML block-level tags that do not start a line are parsed as
      inline HTML and do not interrupt paragraphs (as in RedCloth).

  * Org reader:

    + Make tildes create inline code (#1345).  Also relabeled `code` and
      `verbatim` parsers to accord with the org-mode manual.
    + Respect `:exports` header argument in code blocks (Craig Bosma).
    + Fixed tight lists with sublists (#1437).

  * EPUB writer:

    + Avoid excess whitespace in `nav.xhtml`.  This should improve
      TOC view in iBooks (#1392).
    + Fixed regression on cover image.
      In 1.12.4 and 1.12.4.2, the cover image would not appear properly,
      because the metadata id was not correct.  Now we derive the id from the
      actual cover image filename, which we preserve rather than using
      "cover-image."
    + Keep newlines between block elements.  This allows
      easier diff-ability (#1424).
    + Use `stringify` instead of custom `plainify`.
    + Use `renderTags'` for all tag rendering.  This properly handles tags
      that should be self-closing.  Previously `<hr/>` would appear in EPUB
      output as `<hr></hr>` (#1420).
    + Better handle HTML media tags.
    + Handle multiple dates with OPF `event` attributes.  Note: in EPUB3 we
      can have only one dc:date, so only the first one is used.

  * LaTeX writer:

    + Correctly handle figures in notes.  Notes can't contain figures in
      LaTeX, so we fake it to avoid an error  (#1053).
    + Fixed strikeout + highlighted code (#1294).
      Previously strikeout highlighted code caused an error.

  * ConTeXt writer:

    + Improved detection of autolinks with URLs containing escapes.

  * RTF writer:

    + Improved image embedding: `fetchItem'` is now used to get the
      images, and calculated image sizes are indicated in the RTF.
    + Avoid extra paragraph tags in metadata (#1421).

  * HTML writer:

    + Deactivate "incremental" inside slide speaker notes (#1394).
    + Don't include empty items in the table of contents for
      slide shows.  (These would result from creating a slide
      using a horizontal rule.)

  * MediaWiki writer:

    + Minor renaming of `st` prefixed names.

  * AsciiDoc writer:

    + Double up emphasis and strong emphasis markers in intraword
      contexts, as required by asciidoc (#1441).

  * Markdown writer:

    + Avoid wrapping that might start a list, blockquote, or header (#1013).
    + Use Span instead of (hackish) `SmallCaps` in `plainify`.
    + Don't use braced attributes for fenced code (#1416).
      If `Ext_fenced_code_attributes` is not set, the first class
      attribute will be printed after the opening fence as a bare word.
    + Separate adjacent lists of the same kind with an HTML comment (#1458).

  * PDF writer:

    + Fixed treatment of data uris for images (#1062).

  * Docx writer:

    + Use Compact style for empty table cells (#1353).
      Otherwise we get overly tall lines when there are empty
      table cells and the other cells are compact.
    + Create overrides per-image for `media/` in reference docx.
      This should be somewhat more robust and cover more types of images.
    + Improved `entryFromArchive` to avoid an unneeded parse.
    + Section numbering carries over from reference.docx (#1305).
    + Simplified `abstractNumId` numbering.  Instead of sequential numbering,
      we assign numbers based on the list marker styles.

  * `Text.Pandoc.Options`:

    + Removed `Ext_fenced_code_attributes` from `markdown_github`
      extensions.

  * `Text.Pandoc.ImageSize`:

    + Use default instead of failing if image size not found
      in exif header (#1358).
    + ignore unknown exif header tag rather than crashing.
      Some images seem to have tag type of 256, which was causing
      a runtime error.

  * `Text.Pandoc.Shared`:

    + `fetchItem`:  unescape URI encoding before reading local file (#1427).
    + `fetchItem`:  strip a fragment like `?#iefix` from the extension before
      doing mime lookup, to improve mime type guessing.
    + Improved logic of `fetchItem`:  absolute URIs are fetched from the net;
      other things are treated as relative URIs if `sourceURL` is `Just _`,
      otherwise as file paths on the local file system.
    + `fetchItem` now properly handles links without a protocol (#1477).
    + `fetchItem` now escapes characters not allowed in URIs before trying
      to parse the URIs.
    + Fixed runtime error with `compactify'DL` on certain lists (#1452).

  * `pandoc.hs`: Don't strip path off of `writerSourceURL`: the path is
    needed to resolve relative URLs when we fetch resources (#750).

  * `Text.Pandoc.Parsing`

    + Simplified `dash` and `ellipsis` (#1419).
    + Removed `(>>~)` in favor of the equivalent `(<*)` (Matthew Pickering).
    + Generalized functions to use `ParsecT` (Matthew Pickering).
    + Added `isbn` and `pmid` to list of recognized schemes (Matthew
      Pickering).

  [template changes]

  * Added haddock template.
  * EPUB3:  Added `type` attribute to `link` tags.  They are supposed to
    be "advisory" in HTML5, but kindlegen seems to require them.
  * EPUB3:  Put title page in section with `epub:type="titlepage"`.
  * LaTeX: Made `\subtitle` work properly (#1327).
  * LaTeX/Beamer: remove conditional around date (#1321).
  * LaTeX:  Added `lot` and `lof` variables, which can be set to
    get `\listoftables` and `\listoffigures` (#1407).  Note that
    these variables can be set at the command line with `-Vlot -Vlof`
    or in YAML metadata.

  [under the hood improvements]

  * Rewrote normalize for efficiency (#1385).

  * Rewrote Haddock reader to use `haddock-library` (#1346).

    + This brings pandoc's rendering of haddock markup in line
      with the new haddock.
    + Fixed line breaks in `@` code blocks.
    + alex and happy are no longer build-depends.

  * Added `Text.Pandoc.Compat.Directory` to allow building against
    different versions of the `directory` library.

  + Added `Text.Pandoc.Compat.Except` to allow building against
    different verions of `mtl`.

  * Code cleanup in some writers, using Reader monad to avoid
    passing options parameter around (Matej Kollar).

  * Improved readability in `pandoc.hs`.

  * Miscellaneous code cleanups (Artyom Kazak).

  * Avoid `import Prelude hiding (catch)` (#1309, thanks to Michael
    Thompson).

  * Changed `http-conduit` flag to `https`.  Depend on `http-client`
    and `http-client-tls` instead of `http-conduit`.  (Note:  pandoc still
    depends on `conduit` via `yaml`.)

  * Require `highlighting-kate >= 0.5.8.5` (#1271, #1317, Debian #753299).
    This change to highlighting-kate means that PHP fragments no longer need
    to start with `<?php`.  It also fixes a serious bug causing failures with
    ocaml and fsharp.

  * Require latest `texmath`.  This fixes `\tilde{E}` and allows
    `\left` to be used with `]`, `)` etc. (#1319), among many other
    improvements.

  * Require latest `zip-archive`.  This has fixes for unicode path names.

  * Added tests for plain writer.

  * `Text.Pandoc.Templates`:

    + Fail informatively on template syntax errors.
      With the move from parsec to attoparsec, we lost good error
      reporting.  In fact, since we weren't testing for end of input,
      malformed templates would fail silently.  Here we revert back to
      Parsec for better error messages.
    + Use `ordNub` (#1022).

  * Benchmarks:

    + Made benchmarks compile again (Artyom Kazak).
    + Fixed so that the failure of one benchmark does not prevent others
      from running (Artyom Kazak).
    + Use `nfIO` instead of the `getLength` trick to force full evaluation.
    + Changed benchmark to use only the test suite, so that benchmarks
      run more quickly.

  * Windows build script:

    + Add `-windows` to file name.
    + Use one install command for pandoc, pandoc-citeproc.
    + Force install of pandoc-citeproc.

  * `make_osx_package`:  Call zip file `pandoc-VERSION-osx.zip`.
    The zip should not be named `SOMETHING.pkg.zip`, or OSX finder
    will extract it into a folder named `SOMETHING.pkg`, which it
    will interpret as a defective package (#1308).

  * `README`:

    + Made headers for all extensions so they have IDs and can be
      linked to (Beni Cherniavsky-Paskin).
    + Fixed typos (Phillip Alday).
    + Fixed documentation of attributes (#1315).
    + Clarified documentation on small caps (#1360).
    + Better documentation for `fenced_code_attributes` extension
      (Caleb McDaniel).
    + Documented fact that you can put YAML metadata in a separate file
      (#1412).


pandoc (1.12.4.2)

  * Require highlighting-kate >= 0.5.8.  Fixes a performance regression.

  * Shared:  `addMetaValue` now behaves slightly differently:
    if both the new and old values are lists, it concatenates their
    contents to form a new list.

  * LaTeX reader:

    + Set `bibliography` in metadata from `\bibliography` or
      `\addbibresource` command.
    + Don't error on `%foo` with no trailing newline.

  * Org reader:

    + Support code block headers (`#+BEGIN_SRC ...`) (Albert Krewinkel).
    + Fix parsing of blank lines within blocks (Albert Krewinkel).
    + Support pandoc citation extension (Albert Krewinkel).  This can
      be turned off by specifying `org-citations` as the input format.

  * Markdown reader:

    + `citeKey` moved to `Text.Pandoc.Parsing` so it can be used by
      other readers (Albert Krewinkel).

  * `Text.Pandoc.Parsing`:

    + Added `citeKey` (see above).
    + Added `HasLastStrPosition` type class and `updateLastStrPos`
      and `notAfterString` functions.

  * Updated copyright notices (Albert Krewinkel).

  * Added default.icml to data files so it installs with the package.

  * OSX package:

    + The binary is now built with options to ensure that it can be
      used with OSX 10.6+.
    + Moved OSX package materials to osx directory.
    + Added OSX package uninstall script, included in the zip container
      (thanks to Daniel T. Staal).

pandoc (1.12.4)

  * Made it possible to run filters that aren't executable (#1096).
    Pandoc first tries to find the executable (searching the path
    if path isn't given).  If it fails, but the file exists and has
    a `.py`, `.pl`, `.rb`, `.hs`, or `.php` extension, pandoc runs the filter
    using the appropriate interpreter.  This should make it easier to
    use filters on Windows, and make it more convenient for everyone.

  * Added Emacs org-mode reader (Albert Krewinkel).

  * Added InDesign ICML Writer (Mauro Bieg).

  * MediaWiki reader:

    + Accept image links in more languages (Jaime Marquínez Ferrándiz).
    + Fixed bug in certain nested lists (#1213).  If a level 2 list was
      followed by a level 1 list, the first item of the level 1 list
      would be lost.
    + Handle table rows containing just an HTML comment (#1230).

  * LaTeX reader:

    + Give better location information on errors, pointing to line
      numbers within included files (#1274).
    + LaTeX reader:  Better handling of `table` environment (#1204).
      Positioning options no longer rendered verbatim.
    + Better handling of figure and table with caption (#1204).
    + Handle `@{}` and `p{length}` in tabular.  The length is not actually
      recorded, but at least we get a table (#1180).
    + Properly handle `\nocite`.  It now adds a `nocite` metadata
      field.  Citations there will appear in the bibliography but not
      in the text (unless you explicitly put a `$nocite$` variable
      in your template).

  * Markdown reader:

    + Ensure that whole numbers in YAML metadata are rendered without
      decimal points.  (This became necessary with changes to aeson
      and yaml libraries.  aeson >= 0.7 and yaml >= 0.8.8.2 are now required.)
    + Fixed regression on line breaks in strict mode (#1203).
    + Small efficiency improvements.
    + Improved parsing of nested `div`s.  Formerly a closing `div` tag
      would be missed if it came right after other block-level tags.
    + Avoid backtracking when closing `</div>` not found.
    + Fixed bug in reference link parsing in `markdown_mmd`.
    + Fixed a bug in list parsing (#1154).  When reading a raw list
      item, we now strip off up to 4 spaces.
    + Fixed parsing of empty reference link definitions (#1186).
    + Made one-column pipe tables work (#1218).

  * Textile reader:

    + Better support for attributes.  Instead of being ignored, attributes
      are now parsed and included in Span inlines.  The output will be a bit
      different from stock textile: e.g. for `*(foo)hi*`, we'll get
      `<em><span class="foo">hi</span></em>` instead of
      `<em class="foo">hi</em>`.  But at least the data is not lost.
    + Improved treatment of HTML spans (%) (#1115).
    + Improved link parsing.  In particular we now pick up on attributes.
      Since pandoc links can't have attributes, we enclose the whole link in
      a span if there are attributes (#1008).
    + Implemented correct parsing rules for inline markup (#1175, Matthew
      Pickering).
    + Use Builder (Matthew Pickering).
    + Fixed list parsing bug (#1500).
    + Don't allow inline formatting to extend over newlines.
      This matches the behavior of RedCarpet, avoids some ugly bugs,
      and improves performance.

  * DocBook reader:

    + Better treatment of `formalpara`.  We now emit the title (if present)
      as a separate paragraph with boldface text (#1215).
    + Set metadata `author` not `authors`.
    + Added recognition of `authorgroup` and `releaseinfo` elements (#1214,
      Matthew Pickering).
    + Converted current meta information parsing in DocBook to a more
      extensible version which is aware of the more recent meta
      representation (Matthew Pickering).

  * HTML reader:

    + Require tagsoup 0.13.1, to fix a bug with parsing of script tags
      (#1248).
    + Treat processing instructions & declarations as block.  Previously
      these were treated as inline, and included in paragraph tags in HTML
      or DocBook output, which is generally not what is wanted (#1233).
    + Updated `closes` with rules from HTML5 spec.
    + Use Builder (Matthew Pickering, #1162).

  * RST reader:

    + Remove duplicate `http` in PEP links (Albert Krewinkel).
    + Make rst figures true figures (#1168, CasperVector)
    + Enhanced Pandoc's support for rST roles (Merijn Verstaaten).
      rST parser now supports: all built-in rST roles, new role definition,
      role inheritance, though with some limitations.
    + Use `author` rather than `authors` in metadata.
    + Better handling of directives.  We now correctly handle field
      lists that are indented more than three spaces.  We treat an
      `aafig` directive as a code block with attributes, so it can be
      processed in a filter (#1212).

  * LaTeX writer:

    + Mark span contents with label if span has an ID (Albert Krewinkel).
    + Made `--toc-depth` work well with books in latex/pdf output (#1210).
    + Handle line breaks in simple table cells (#1217).
    + Workaround for level 4-5 headers in quotes.  These previously produced
      invalid LaTeX: `\paragraph` or `\subparagraph` in a `quote` environment.
      This adds an `mbox{}` in these contexts to work around the problem.
      See <http://tex.stackexchange.com/a/169833/22451> (#1221).
    + Use `\/` to avoid en-dash ligature instead of `-{}-` (Vaclav Zeman).
      This is to fix LuaLaTeX output. The `-{}-` sequence does not avoid the
      ligature with LuaLaTeX but `\/` does.
    + Fixed string escaping in `hyperref` and `hyperdef` (#1130).

  * ConTeXt writer:  Improved autolinks (#1270).

  * DocBook writer:

    + Improve handling of hard line breaks in Docbook writer
      (Neil Mayhew).  Use a `<literallayout>` for the entire paragraph, not
      just for the newline character.
    + Don't let line breaks inside footnotes influence the enclosing
      paragraph (Neil Mayhew).
    + Distinguish tight and loose lists in DocBook output, using
      `spacing="compact"` (Neil Mayhew, #1250).

  * Docx writer:  When needed files are not present in the user's
    `reference.docx`, fall back on the versions in the `reference.docx`
    in pandoc's data files. This fixes a bug that occurs when a
    `reference.docx` saved by LibreOffice is used. (#1185)

  * EPUB writer:

    + Include extension in epub ids.  This fixes a problem with duplicate
      extensions for fonts and images with the same base name but different
      extensions (#1254).
    + Handle files linked in raw `img` tags (#1170).
    + Handle media in `audio` source tags (#1170).
      Note that we now use a `media` directory rather than `images`.
    + Incorporate files linked in `video` tags (#1170).  `src` and `poster`
      will both be incorporated into `content.opf` and the epub container.

  * HTML writer:

    + Add colgroup around col tags (#877).  Also affects EPUB writer.
    + Fixed bug with unnumbered section headings.  Unnumbered section
      headings (with class `unnumbered`) were getting numbers.
    + Improved detection of image links. Previously image links with
      queries were not recognized, causing `<embed>` to be used instead
      of `<img>`.

  * Man writer:  Ensure that terms in definition lists aren't line wrapped
    (#1195).

  * Markdown writer:

    + Use proper escapes to avoid unwanted lists (#980).  Previously we used
      0-width spaces, an ugly hack.
    + Use longer backtick fences if needed (#1206).  If the content contains a
      backtick fence and there are attributes, make sure longer fences are
      used to delimit the code.  Note:  This works well in pandoc, but github
      markdown is more limited, and will interpret the first string of three
      or more backticks as ending the code block.

  * RST writer:  Avoid stack overflow with certain tables (#1197).

  * RTF writer:  Fixed table cells containing paragraphs.

  * Custom writer:

    + Correctly handle UTF-8 in custom lua scripts (#1189).
    + Fix bugs with lua scripts with mixed-case filenames and
      paths containing `+` or `-` (#1267).  Note that `getWriter`
      in `Text.Pandoc` no longer returns a custom writer on input
      `foo.lua`.

  * AsciiDoc writer:  Handle multiblock and empty table cells
    (#1245, #1246).  Added tests.

  * `Text.Pandoc.Options`: Added `readerTrace` to `ReaderOptions`

  * `Text.Pandoc.Shared`:

    + Added `compactify'DL` (formerly in markdown reader) (Albert Krewinkel).
    + Fixed bug in `toRomanNumeral`:  numbers ending with '9' would
      be rendered as Roman numerals ending with 'IXIV' (#1249).  Thanks to
      Jesse Rosenthal.
    + `openURL`: set proxy with value of http_proxy env variable (#1211).
      Note:  proxies with non-root paths are not supported, due to
      limitations in `http-conduit`.

  * `Text.Pandoc.PDF`:

    + Ensure that temp directories deleted on Windows (#1192).  The PDF is
      now read as a strict bytestring, ensuring that process ownership will
      be terminated, so the temp directory can be deleted.
    + Use `/` as path separators in a few places, even on Windows.
      This seems to be necessary for texlive (#1151, thanks to Tim Lin).
    + Use `;` for `TEXINPUTS` separator on Windows (#1151).
    + Changes to error reporting, to handle non-UTF8 error output.

  * `Text.Pandoc.Templates`:

    + Removed unneeded datatype context (Merijn Verstraaten).

    + YAML objects resolve to "true" in conditionals (#1133).
      Note:  If `address` is a YAML object and you just have `$address$`
      in your template, the word `true` will appear, which may be
      unexpected.  (Previously nothing would appear.)

  * `Text.Pandoc.SelfContained`:

    + `mkSelfContained` now takes just two arguments, `WriterOptions` and
      the string.
    * It no longer looks in data files.  This only made sense when we
      had copies of slidy and S5 code there.
    * `fetchItem'` is used instead of the nearly duplicate `getItem`.
    + Handle `poster` attribute in `video` tags (#1188).

  * `Text.Pandoc.Parsing`:

    + Made `F` an instance of Applicative (#1138).
    + Added `stateCaption`.
    + Added `HasMacros`, simplified other typeclasses.
      Removed `updateHeaderMap`, `setHeaderMap`, `getHeaderMap`,
      `updateIdentifierList`, `setIdentifierList`, `getIdentifierList`.
    + Changed the smart punctuation parser to return `Inlines`
      rather than `Inline` (Matthew Pickering).
    + Changed `HasReaderOptions`, `HasHeaderMap`, `HasIdentifierList`
      from typeclasses of monads to typeclasses of states.  This simplifies
      the instance definitions and provides more flexibility.  Generalized
      type of `getOption` and added a default definition.  Removed
      `askReaderOption`.  Added `extractReaderOption`.  Added
      `extractHeaderMap` and `updateHeaderMap` in `HasHeaderMap`.
      Gave default definitions for `getHeaderMap`, `putHeaderMap`,
      `modifyHeaderMap`.  Added `extractIdentifierList` and
      `updateIdentifierList` in `HasIdentifierList`.  Gave defaults
      for `getIdentifierList`, `putIdentifierList`, and
      `modifyIdentifierList`.  The ultimate goal here is to allow different
      parsers to use their own, tailored parser states (instead of
      `ParserState`) while still using shared functions.

  * Template changes:

    + LaTeX template: Use `fontenc` package only with `pdflatex` (#1164).
    + LaTeX template:  Add `linestretch` and `fontfamily` variables.
    + LaTeX template:  Conditionalize author and date commands.
    + Beamer template: Consistent styles for figure and table captions
      (aaronwolen).
    + LaTeX and beamer template:  Adjust widths correctly for oversized
      images.  Use `\setkeys{Gin}{}` to set appropriate defaults for
      `\includegraphics` (Yihui Xie, Garrick Aden-Buie).  Load
      `upquote` only after `fontenc` (Yihui Xie).
    + Beamer template: Added caption package (#1200).
    + Beamer template:  changes for better unicode handling (KarolS).
    + DocBook template:  use `authorgroup` if there are authors.
    + revealjs template: Move `include-after` to end (certainlyakey).
    + revealjs template: Fixed PDF print function (#1220, kevinkenan).

  * Bumped version bounds of dependencies.

  * Added a `--trace` command line option, for debugging backtracking
    bugs.  So far this only works with the markdown reader.

  * MathMLinHTML:  Fixed deprecation warning (#362, gwern, Albert Krewinkel).

  * Updated travis script to test with multiple GHC versions.

  * Force failure of a Travis build if GHC produces warnings (Albert
    Krewinkel).

  * Add `.editorconfig` (Albert Krewinkel).
    See <http://editorconfig.org/> for details.

  * Give more useful error message if '-t pdf' is specified (#1155).

  * Added `Cite`, `SmallCaps` to `Arbitrary` instance (#1269).

  * Allow `html4` as a synonym of `html` as a reader (it already works
    as a writer).

  * README:

    + Added an explanation of how to use YAML metadata to
      force items to appear in the bibliography without citations in
      the text (like LaTeX `\nocite`).
    + Added note to `--bibtex/--natbib`: not for use in making PDF
      (#1194, thanks to nahoj).
    + Added explanatory notes about `--natbib` and `--biblatex`.
    + Added specification of legal syntax for citation keys.
    + Fixed variable defaults documentation (Albert Krewinkel).

  * Removed copyright statements for files that have been removed
    (Albert Krewinkel).

  * Moved some doc files from `data-files` to `extra-source-files` (#1123).
    They aren't needed at runtime.  We keep README and COPYRIGHT in data
    to ensure that they'll be available on all systems on which pandoc
    is installed.

  * Use cabal sandboxes in Windows build script.

pandoc (1.12.3.3)

  * To changes to source; recompiled tarball with latest alex and
    happy, so they will work with GHC 7.8.

pandoc (1.12.3.2)

  * Bumped version bounds for blaze-html, blaze-markup.

  * ImageSize:  Avoid use of lookAhead, which is not in binary >= 0.6
    (#1124).

  * Fixed mediawiki ordered list parsing (#1122).

  * HTML reader:  Fixed bug reading inline math with `$$` (#225).

  * Added support for LaTeX style literate Haskell code blocks in rST
    (Merijn Verstraaten).

pandoc (1.12.3.1)

  * Relaxed version constraint on binary, allowing the use of binary 0.5.

pandoc (1.12.3)

  * The `--bibliography` option now sets the `biblio-files` variable.
    So, if you're using `--natbib` or `--biblatex`, you can just use
    `--bibliography=foo.bib` instead of `-V bibliofiles=foo`.

  * Don't run pandoc-citeproc filter if `--bibliography` is
    used together with `--natbib` or `--biblatex` (Florian Eitel).

  * Template changes:

    + Updated beamer template to include booktabs.
    + Added `abstract` variable to LaTeX template.
    + Put `header-includes` after `title` in LaTeX template (#908).
    + Allow use of `\includegraphics[size]` in beamer.
      This just required porting a macro definition from the default
      LaTeX template to the default beamer template.

  * `reference.docx`:  Include `FootnoteText` style.
    Otherwise Word ignores the style, even when specified in the `pPr`.
    (#901)

  * `reference.odt`:  Tidied `styles.xml`.

  * Relaxed version bounds for dependencies.

  * Added `withSocketsDo` around http conduit code in `openURL`,
    so it works on Windows (#1080).

  * Added `Cite` function to `sample.lua`.

  * Markdown reader:

    + Fixed regression in title blocks (#1089).
      If author field was empty, date was being ignored.
    + Allow backslash-newline hard line breaks in grid and
      multiline table cells.
    + Citation keys may now start with underscores, and may contain
      underscores adjacent to internal punctuation.

  * LaTeX reader:

    + Add support for `Verb` macro (jrnold) (#1090).
    + Support babel-style quoting: `` "`..."' ``.

  * Properly handle script blocks in strict mode.  (That is,
    `markdown-markdown_in_html_blocks`.) Previously a spurious
    `<p>` tag was being added (#1093).

  * Docbook reader: Avoid failure if `tbody` contains no `tr` or `row`
    elements.

  * LaTeX writer:

    + Factored out function for table cell creation.
    + Better treatment of footnotes in tables.
      Notes now appear in the regular sequence, rather than in the
      table cell.  (This was a regression in 1.10.)

  * HTML reader: Parse name/content pairs from meta tags as metadata.
    Closes #1106.

  * Moved `fixDisplayMath` from Docx writer to `Writer.Shared`.

  * OpenDocument writer:  Fixed `RawInline`, `RawBlock` so they don't escape.

  * ODT writer:  Use mathml for proper rendering of formulas.
    Note:  LibreOffice's support for this seems a bit buggy.  But
    it should be better than what we had before.

  * RST writer: Ensure no blank line after def in definition list (#992).

  * Markdown writer: Don't use tilde code blocks with braced attributes in
    `markdown_github` output.  A consequence of this change is that the
    backtick form will be preferred in general if both are enabled.  That
    is good, as it is much more widespread than the tilde form.  (#1084)

  * Docx writer:  Fixed problem with some modified reference docx files.
    Include `word/_rels/settings.xml.rels` if it exists, as well as other
    `rels` files besides the ones pandoc generates explicitly.

  * HTML writer:

    + With `--toc`, headers no longer link to themselves (#1081).
    + Omit footnotes from TOC entries.  Otherwise we get doubled
      footnotes when headers have notes!

  * EPUB writer:

    + Avoid duplicate notes when headings contain notes.
      This arose because the headings are copied into the metadata
      "title" field, and the note gets rendered twice.  We strip the
      note now before putting the heading in "title".
    + Strip out footnotes from toc entries.
    + Fixed bug with `--epub-stylesheet`.  Now the contents of
      `writerEpubStylesheet` (set by `--epub-stylesheet`)
      should again work, and take precedence over a stylesheet specified
      in the metadata.

  * `Text.Pandoc.MIME`: Added `wmf`, `emf`.

  * `Text.Pandoc.Shared`:  `fetchItem` now handles image URLs beginning
    with `//`.

  * `Text.Pandoc.ImageSize`:  Parse EXIF format JPEGs.  Previously
    we could only get size information for JFIF format, which led
    to squished images in Word documents. Closes #976.

  * Removed old `MarkdownTest_1.0.3` directory (#1104).

pandoc (1.12.2.1)

  * Markdown reader:  Fixed regression in list parser, involving
    continuation lines containing raw HTML (or even verbatim raw HTML).

pandoc (1.12.2)

  * Metadata may now be included in YAML blocks in a markdown document.
    For example,

        ---
        title:
        - type: main
          text: My Book
        - type: subtitle
          text: An investigation of metadata
        creator:
        - role: author
          text: John Smith
        - role: editor
          text: Sarah Jones
        identifier:
        - scheme: DOI
          text: doi:10.234234.234/33
        publisher:  My Press
        rights:  (c) 2007 John Smith, CC BY-NC
        cover-image: img/mypic.jpg
        ...

    Metadata may still be provided using `--epub-metadata`; it will
    be merged with the metadata in YAML blocks.

  * EPUB writer:

    + `meta` tags are now used instead of `opf` attributes for EPUB3.
    + Insert "svg" property as needed in opf (EPUB 3).
    + Simplify `imageTypeOf` using `getMimeType`.
    + Add properties attribute to `cover-image` item for EPUB 3.
    + Don't include node for `cover.xhtml` if no cover!
    + Ensure that same identifier is used throughout (#1044).
      If an identifier is given in metadata, we use that; otherwise
      we generate a random uuid.
    + Add cover reference to guide element (EPUB 2) (Shaun Attfield).
      Fixes an issue with Calibre putting the cover at the end of the book
      if the spine has `linear="no"`.  Apparently this is best practice
      for other converters as well:
      <http://www.idpf.org/epub/20/spec/OPF_2.0.1_draft.htm#Section2.6>.
    + Allow `stylesheet` in metadata.  The value is a path to the stylesheet.
    + Allow partial dates:  `YYYY`, `YYYY-MM`.

  * Markdown writer:  Fix rendering of tight sublists (#1050).
    Previously a spurious blank line was included after a tight sublist.

  * ODT writer:  Add `draw:name` attribute to `draw:frame` elements (#1069).
    This is reported to be necessary to avoid an error from recent
    versions of Libre Office when files contain more than one image
    Thanks to wmanley for reporting and diagnosing the problem.

  * ConTeXt writer:  Don't hardcode figure/table placement and numbering.
    Instead, let this be set in the template, using `\setupfloat`.
    Thanks to on4aa and Aditya Mahajan for the suggestion (#1067).

  * Implemented CSL flipflopping spans in DOCX, LaTeX, and HTML writers.

  * Fixed bug with markdown intraword emphasis.  Closes #1066.

  * Docbook writer:  Hierarchicalize block content in metadata.
    Previously headers just disappeared from block-level metadata
    when it was used in templates.  Now we apply the 'hierarchicalize'
    transformation.  Note that a block headed by a level-2 header will
    turn into a `<sect1>` element.

  * OpenDocument writer:  Skip raw HTML (#1035).
    Previously it was erroneously included as verbatim text.

  * HTML/EPUB writer, footnotes:  Put `<sup>` tag inside `<a>` tags.
    This allows better control of formatting, since the `<a>`
    tags have a distinguishing class (#1049).

  * Docx writer:

    + Use mime type info returned by fetchItem.
    + Fixed core metadata (#1046).
      Don't create empty date nodes if no date given.
      Don't create multiple `dc:creator` nodes; instead separate by
      semicolons.
    + Fix URL for core-properties in `_rels/.rels` (#1046).

  * Plain writer: don't print `<span>` tags.

  * LaTeX writer:

    + Fix definition lists with internal links in terms (#1032).
      This fix puts braces around a term that contains an internal
      link, to avoid problems with square brackets.
    + Properly escape pdftitle, pdfauthor (#1059).
    + Use booktabs package for tables (thanks to Jose Luis Duran).

  * Updated beamer template.  Now references should work properly
    (in a slide) when `--biblatex` or `--natbib` is used.

  * LaTeX reader:

    + Parse contents of curly quotes or matched `"` as quotes.
    + Support `\textnormal` as span with class `nodecor`.
      This is needed for pandoc-citeproc.
    + Improved citation parsing.  This fixes a run-time error that occured
      with `\citet{}` (empty list of keys).  It also ensures that empty keys
      don't get produced.

  * MediaWiki reader:  Add automatic header identifiers.

  * HTML reader:

    + Use pandoc `Div` and `Span` for raw `<div>`, `<span>` when
      `--parse-raw`.
    + Recognize `svg` tags as block level content (thanks to MinRK).
    + Parse LaTeX math if appropriate options are set.

  * Markdown reader:

    + Yaml block must start immediately after `---`.  If there's a blank
      line after `---`, we interpreted it as a horizontal rule.
    + Correctly handle empty bullet list items.
    + Stop parsing "list lines" when we hit a block tag.
      This fixes exponential slowdown in certain input, e.g.
      a series of lists followed by `</div>`.

  * Slides:  Preserve `<div class="references">` in references slide.

  * `Text.Pandoc.Writer.Shared`:

    + Fixed bug in `tagWithAttrs`.  A space was omitted before key-value
      attributes, leading to invalid HTML.
    + `normalizeDate`: Allow dates with year only (thanks to Shaun Attfield).
    + Fixed bug in `openURL` with `data:` URIs.  Previously the base-64
      encoded bytestring was returned.  We now decode it so it's a proper
      image!

  * DocBook reader:  Handle numerical attributes starting with decimal.
    Also use `safeRead` instead of `read`.

  * `Text.Pandoc.Readers.TexMath`: Export `readTeXMath'`, which attends
    to display/inline.  Deprecate `readTeXMath`, and use `readTeXMath'`
    in all the writers.  Require `texmath >= 0.6.5.2`.

  * `Text.Pandoc.MIME`:

    + Add entry for `jfif`.
    + In looking up extensions, drop the encoding info.
      E.g. for 'image/jpg;base64' we should lookup 'image/jpg'.

  * Templates:  Changed how array variables are resolved.  Previously if
    `foo` is an array (which might be because multiple values were set on
    the command line), `$foo$` would resolve to the concatenation of the
    elements of foo.  This is rarely useful behavior.  It has been changed
    so that the first value is rendered.  Of course, you can still iterate
    over the values using `$for(foo)$`.  This has the result that you can
    override earlier settings using `-V` by putting new values later on the
    command line, which is useful for many purposes.

  * `Text.Pandoc`:  Don't default to `pandocExtensions` for all writers.

  * Allow "epub2" as synonym for "epub", "html4" for "html".

  * Don't look for slidy files in data files with `--self-contained`.

  * Allow `https:` command line arguments to be downloaded.

  * Fixed `make_osx_package.sh` so data files embedded in `pandoc-citeproc`.

pandoc (1.12.1)

  * `Text.Pandoc.Definition`:  Changed default JSON serialization format.
    Instead of `{"Str": "foo"}`, for example, we now have `{"t": "Str",
    "c": "foo"}`.  This new format is easier to work with outside of Haskell.
    Incidentally, "t" stands for "tag", "c" for "contents".

  * MediaWiki reader: Trim contents of `<math>` tags, to avoid problems
    when converting to markdown (#1027).

  * LaTeX reader:

    + Ensure that preamble doesn't contribute to the text of
      the document.
    + Fixed character escaping in \url{}.  Previously `\~` wasn't handled
      properly, among others.
    + Parse `{groups}` as `Span`.  This is needed for accurate conversion of
      bibtex titles, since we need to know what was protected from
      titlecase conversions.

  * LaTeX writer:

    + Specially escape non-ascii characters in labels.
      Otherwise we can get compile errors and other bugs when
      compiled with pdflatex (#1007).  Thanks to begemotv2718 for the fix.
    + Add link anchors for code blocks with identifiers (#1025).

  * Throughout the code, use `isURI` instead of `isAbsoluteURI`.
    It allows fragments identifiers.

  * Slide formats:

    + A Div element with class "notes" is treated as speaker
      notes.  Currently beamer goes to `\note{}`, revealjs to
      `<aside class="notes">`, and the notes are simply suppressed in
      other formats  (#925).
    + Fixed `. . .` (pause) on HTML slide formats.  Closes #1029.
      The old version caused a pause to be inserted before the first
      material on a slide. This has been fixed.
    + Removed data files for s5, slideous, slidy.
      Users of s5 and slideous will have to download the needed
      files, as has been documented for some time in the README.
      By default, slidy code will be sought on the web, as before.

  * HTML writer: Insert command to typeset mathjax only in slideous output
    (#966, #1012).

  * RST writer:  Skip spaces after display math.  Otherwise we get indentation
    problems, and part of the next paragraph may be rendered as part of the
    math.

  * OpenDocument writer:  Fix formatting of strikeout code (#995),
    thanks to wilx.  don't use `font-face-decls` variable.

  * Fixed test suite so it works with cabal sandboxes.

pandoc (1.12.0.2)

  * Removed `stringable` dependency.

pandoc (1.12.0.1)

  * Allow `--metadata` to be repeated for the same key to form a list.
    This also has the effect that `--bibliography` can be repeated,
    as before.

  * Handle boolean values in `--metadata`.  Note that anything not parseable
    as a YAML boolean or string is treated as a literal string.
    You can get a string value with "yes", or any of the strings interpretable
    as booleans, by quoting it:

        -M boolvalue=yes -M stringvalue='"yes"'

  * LaTeX writer: Don't print references if `--natbib` or `--biblatex`
    option used.

  * DOCX writer: Add `settings.xml` to the zip container.  Fixes a bug
    in which docx files could not be read by some versions of Word
    and LibreOffice (#990).

  * Fixed a regression involving slide shows with bibliographies.
    The Div container around references messed up the procedure for carving
    a document into slides.  So we now remove the surrounding Div in
    `prepSlides`.

  * More informative error message when a filter is not found in path.

  * Depend on pandoc-types 1.12.1.  This provide `ToJSONFilter`
    instances for `Data a => a -> [a]` and `Data a => a -> IO [a]`.

  * Don't use unicode_collation in building OSX package:
    it adds something like 50MB of dependencies to the package.

  * Declare alex and happy as build-tools (#986).

pandoc (1.12)

  [new features]

  * Much more flexible metadata, including arbitrary fields and structured
    values.  Metadata can be specified flexibly in pandoc markdown using
    YAML metadata blocks, which may occur anywhere in the document:

        ---
        title: Here is my title.
        abstract: |
          This is the abstract.

          1. It can contain
          2. block content
             and *inline markup*

        tags: [cat, dog, animal]
        ...

    Metadata fields automatically populate template variables.

  * Added `opml` (OPML) as input and output format.  The `_note` attribute,
    used in OmniOutliner and supported by multimarkdown, is supported.
    We treat the contents as markdown blocks under a section header.

  * Added `haddock` (Haddock markup) as input format (David Lazar).

  * Added `revealjs` output format, for reveal.js HTML 5 slide shows.
    (Thanks to Jamie F. Olson for the initial patch.)
    Nested vertical stacks are used for hierarchical structure.
    Results for more than one level of nesting may be odd.

  * Custom writers can now be written in lua.

        pandoc -t data/sample.lua

    will load the script sample.lua and use it as a custom writer.
    (For a sample, do `pandoc --print-default-data-file sample.lua`.)
    Note that pandoc embeds a lua interpreter, so lua need not be
    installed separately.

  * New `--filter/-F` option to make it easier to run "filters"
    (Pandoc AST transformations that operate on JSON serializations).
    Filters are always passed the name of the output format, so their
    behavior can be tailored to it.  The repository
    <https://github.com/jgm/pandocfilters> contains
    a python module for writing pandoc filters in python, with
    a number of examples.

  * Added `--metadata/-M` option.
    This is like `--variable/-V`, but actually adds to metadata, not
    just variables.

  * Added `--print-default-data-file` option, which allows printing
    of any of pandoc's data files. (For example,
    `pandoc --print-default-data-file reference.odt` will print
    `reference.odt`.)

  * Added syntax for "pauses" in slide shows:

        This gives

        . . .

        me pause.

  * New markdown extensions:

    + `ignore_line_breaks`:  causes intra-paragraph line breaks to be ignored,
      rather than being treated as hard line breaks or spaces.  This is useful
      for some East Asian languages, where spaces aren't used between words,
      but text is separated into lines for readability.
    + `yaml_metadata_block`:  Parse YAML metadata blocks.  (Default.)
    + `ascii_identifiers`: This will force `auto_identifiers` to use ASCII
       only. (Default for `markdown_github`.) (#807)
    + `lists_without_preceding_blankline`:  Allow lists to start without
      preceding blank space.  (Default for `markdown_github`.) (#972)

  [behavior changes]

  * `--toc-level` no longer implies `--toc`.
    Reason: EPUB users who don't want a visible TOC may still want
    to set the TOC level for in the book navigation.

  * `--help` now prints in and out formats in alphabetical order, and
    says something about PDF output (#720).

  * `--self-contained` now returns less verbose output (telling you
    which URLs it is fetching, but not giving the full header).  In
    addition, there are better error messages when fetching a URL fails.

  * Citation support is no longer baked in to core pandoc. Users who
    need citations will need to install and use a separate filter
    (`--filter pandoc-citeproc`).  This filter will take `bibliography`,
    `csl`, and `citation-abbreviations` from the metadata, though it
    may still be specified on the command line as before.

  * A `Cite` element is now created in parsing markdown whether or not
    there is a matching reference.

  * The `pandoc-citeproc` script will put the bibliography at the
    end of the document, as before.  However, it will be put inside a `Div`
    element with class "references", allowing users some control
    over the styling of references.  A final header, if any, will
    be included in the `Div`.

  * The markdown writer will not print a bibliography if the
    `citations` extension is enabled.  (If the citations are formatted
    as markdown citations, it is redundant to have a bibliography,
    since one will be generated automatically.)

  * Previously we used to store the directory of the first input file,
    even if it was local, and used this as a base directory for finding
    images in ODT, EPUB, Docx, and PDF.  This has been confusing to many
    users.  So we now look for images relative to the current
    working directory, even if the first file argument is in another
    directory.   Note that this change may break some existing workflows.
    If you have been assuming that relative links will be interpreted
    relative to the directory of the first file argument, you'll need
    to make that the current directory before running pandoc. (#942)

  * Better error reporting in some readers, due to changes in `readWith`:
    the line in which the error occured is printed, with a caret pointing
    to the column.

  * All slide formats now support incremental slide view for definition lists.

  * Parse `\(..\)` and `\[..\]` as math in MediaWiki reader.
    Parse `:<math>...</math>` as display math.  These notations are used with
    the MathJax MediaWiki extension.

  * All writers: template variables are set automatically from metadata
    fields.  However, variables specified on the command line with
    `--variable` will completely shadow metadata fields.

  * If `--variable` is used to set many variables with the same name,
    a list is created.

  * Man writer:  The `title`, `section`, `header`, and `footer` can now
    all be set individually in metadata.  The `description` variable has been
    removed.  Quotes have been added so that spaces are allowed in the
    title.  If you have a title that begins

        COMMAND(1) footer here | header here

    pandoc will still parse it into a title, section, header, and
    footer.  But you can also specify these elements explicitly (#885).

  * Markdown reader

    + Added support for YAML metadata blocks, which can come anywhere
      in the document (not just at the beginning).  A document can contain
      multiple YAML metadata blocks.
    + HTML span and div tags are parsed as pandoc Span and Div elements.

  * Markdown writer

    + Allow simple tables to be printed as grid tables,
      if other table options are disabled.  This means you can do
      `pandoc -t markdown-pipe_tables-simple_tables-multiline_tables`
      and all tables will render as grid tables.
    + Support YAML title block (render fields in alphabetical order
      to make output predictable).

  [API changes]

  * `Meta` in `Text.Pandoc.Definition` has been changed to allow
    structured metadata.  (Note:  existing code that pattern-matches
    on `Meta` will have to be revised.)  Metadata can now contain
    indefinitely many fields, with content that can be a string,
    a Boolean, a list of `Inline` elements, a list of `Block`
    elements, or a map or list of these.

  * A new generic block container (`Div`) has been added to `Block`,
    and a generic inline container (`Span`) has been added to `Inline`.
    These can take attributes.  They will render in HTML, Textile,
    MediaWiki, Org, RST and and Markdown (with `markdown_in_html`
    extension) as HTML `<div>` and `<span>` elements; in other formats
    they will simply pass through their contents.  But they can be
    targeted by scripts.

  * `Format` is now a newtype, not an alias for String.
    Equality comparisons are case-insensitive.

  * Added `Text.Pandoc.Walk`, which exports hand-written tree-walking
    functions that are much faster than the SYB functions from
    `Text.Pandoc.Generic`.  These functions are now used where possible
    in pandoc's code.  (`Tests.Walk` verifies that `walk` and `query`
    match the generic traversals `bottomUp` and `queryWith`.)

  * Added `Text.Pandoc.JSON`, which provides `ToJSON` and `FromJSON`
    instances for the basic pandoc types. They use GHC generics and
    should be faster than the old JSON serialization using
    `Data.Aeson.Generic`.

  * Added `Text.Pandoc.Process`, exporting `pipeProcess`.
    This is a souped-up version of `readProcessWithErrorcode` that
    uses lazy bytestrings instead of strings and allows setting
    environment variables.  (Used in `Text.Pandoc.PDF`.)

  * New module `Text.Pandoc.Readers.OPML`.

  * New module `Text.Pandoc.Writers.OPML`.

  * New module `Text.Pandoc.Readers.Haddock` (David Lazar).
    This is based on Haddock's own lexer/parser.

  * New module `Text.Pandoc.Writers.Custom`.

  * In `Text.Pandoc.Shared`, `openURL` and `fetchItem` now return an
    Either, for better error handling.

  * Made `stringify` polymorphic in `Text.Pandoc.Shared`.

  * Removed `stripTags` from `Text.Pandoc.XML`.

  * `Text.Pandoc.Templates`:

    + Simplified `Template` type to a newtype.
    + Removed `Empty`.
    + Changed type of `renderTemplate`: it now takes a JSON context
      and a compiled template.
    + Export `compileTemplate`.
    + Export `renderTemplate'` that takes a string instead of a compiled
      template.
    + Export `varListToJSON`.

  * `Text.Pandoc.PDF` exports `makePDF` instead of `tex2pdf`.
    The signature of `makePDF` has changed and now contains
    an additional argument for pdf engine options.

  * `Text.Pandoc`:

    + Made `toJsonFilter` an alias for `toJSONFilter` from `Text.Pandoc.JSON`.
    + Removed `ToJsonFilter` typeclass.  `ToJSONFilter` from
      `Text.Pandoc.JSON` should be used instead.  (Compiling against
      pandoc-types instead of pandoc will also produce smaller executables.)
    * Removed the deprecated `jsonFilter` function.
    + Added `readJSON`, `writeJSON` to the API (#817).

  * `Text.Pandoc.Options`:

    + Added `Ext_lists_without_preceding_blankline`,
      `Ext_ascii_identifiers`, `Ext_ignore_line_breaks`,
      `Ext_yaml_metadataBlock` to `Extension`.
    + Changed `writerSourceDirectory` to `writerSourceURL` and changed the
      type to a `Maybe`.  `writerSourceURL` is set to 'Just url' when the
      first command-line argument is an absolute URL.  (So, relative links
      will be resolved in relation to the first page.)  Otherwise, 'Nothing'.
    + All bibliography-related fields have been removed from
      `ReaderOptions` and `WriterOptions`: `writerBiblioFiles`,
      `readerReferences`, `readerCitationStyle`.
    + Removed `writerPdfArgs` from `WriterOptions`.  These are now
      passed in as a parameter to `makePDF`.

  * The `Text.Pandoc.Biblio` module has been removed.  Users of the
    pandoc library who want citation support will need to use
    `Text.CSL.Pandoc` from `pandoc-citeproc`.


  [bug fixes]

  * In markdown, don't autolink a bare URI that is followed by `</a>`
    (#937).

  * `Text.Pandoc.Shared`

    + `openURL` now follows redirects (#701), properly handles `data:`
      URIs, and prints diagnostic output to stderr rather than stdout.
    + `readDefaultDataFile`: normalize the paths.  This fixes bugs in
      `--self-contained` on pandoc compiled with `embed_data_files` (#833).
    + Fixed `readDefaultDataFile` so it works on Windows.
    + Better error messages for `readDefaultDataFile`.  Instead of
      listing the last path tried, which can confuse people who are
      using `--self-contained`, so now we just list the data file name.
    + URL-escape pipe characters.  Even though these are legal, `Network.URI`
      doesn't regard them as legal in URLs.  So we escape them first (#535).

  * Mathjax in HTML slide shows:  include explicit "Typeset" call.
    This seems to be needed for some formats (e.g. slideous) and won't
    hurt in others (#966).

  * `Text.Pandoc.PDF`

    + On Windows, create temdir in working directory, since the system
      temp directory path may contain tildes, which can cause
      problems in LaTeX (#777).
    + Put temporary output directory in `TEXINPUTS` (see #917).
    + `makePDF` tries to download images that are not found locally,
      if the first argument is a URL (#917).
    + If compiling with `pdflatex` yields an encoding error, offer
      the suggestion to use `--latex-engine=xelatex`.

  * Produce automatic header identifiers in parsing textile, RST,
    and LaTeX, unless `auto_identifiers` extension is disabled (#967).

  * `Text.Pandoc.SelfContained`:  Strip off fragment, query of relative URL
     before treating as a filename.  This fixes `--self-contained` when used
     with CSS files that include web fonts using the method described here:
      <http://paulirish.com/2009/bulletproof-font-face-implementation-syntax/>
      (#739).  Handle `src` in `embed`, `audio`, `source`, `input` tags.

  * `Text.Pandoc.Parsing`: `uri` parser no longer treats punctuation before
    percent-encoding, or a `+` character, as final punctuation.

  * `Text.Pandoc.ImageSize`:  Handle EPS (#903).  This change will make
    EPS images properly sized on conversion to Word.

  * Slidy:  Use slidy.js rather than slidy.js.gz.
    Reason:  some browsers have trouble with the gzipped js file,
    at least on the local file system (#795).

  * Markdown reader

    + Properly handle blank line at beginning of input (#882).
    + Fixed bug in unmatched reference links.  The input
      `[*infile*] [*outfile*]` was getting improperly parsed:
      "infile" was emphasized, but "*outfile*" was literal (#883).
    + Allow internal `+` in citation identifiers (#856).
    + Allow `.` or `)` after `#` in ATX headers if no `fancy_lists`.
    + Do not generate blank title, author, or date metadata elements.
      Leave these out entirely if they aren't present.
    + Allow backtick code blocks not to be preceded by blank line (#975).

  * Textile reader:

    + Correctly handle entities.
    + Improved handling of `<pre>` blocks (#927). Remove internal HTML tags
      in code blocks, rather than printing them verbatim. Parse attributes
      on `<pre>` tag for code blocks.

  * HTML reader: Handle non-simple tables (#893).  Column widths are read from
    `col` tags if present, otherwise divided equally.

  * LaTeX reader

    + Support alltt environment (#892).
    + Support `\textasciitilde`, `\textasciicircum` (#810).
    + Treat `\textsl` as emphasized text reader (#850).
    + Skip positional options after `\begin{figure}`.
    + Support `\v{}` for hacek (#926).
    + Don't add spurious ", " to citation suffixes.
      This is added when needed in pandoc-citeproc.
    + Allow spaces in alignment spec in tables, e.g. `{ l r c }`.
    + Improved support for accented characters (thanks to Scott Morrison).
    + Parse label after section command and set id (#951).

  * RST reader:

    + Don't insert paragraphs where docutils doesn't.
      `rst2html` doesn't add `<p>` tags to list items (even when they are
      separated by blank lines) unless there are multiple paragraphs in the
      list.  This commit changes the RST reader to conform more closely to
      what docutils does (#880).
    + Improved metadata.  Treat initial field list as metadata when
      standalone specified.  Previously ALL fields "title", "author",
      "date" in field lists were treated as metadata, even if not at
      the beginning.  Use `subtitle` metadata field for subtitle.
    + Fixed 'authors' metadata parsing in reST.  Semicolons separate
      different authors.

  * MediaWiki reader

    + Allow space before table rows.
    + Fixed regression for `<ref>URL</ref>`.
      `<` is no longer allowed in URLs, according to the uri parser
      in `Text.Pandoc.Parsing`.  Added a test case.
    + Correctly handle indented preformatted text without preceding
      or following blank line.
    + Fixed `|` links inside table cells.  Improved attribute parsing.
    + Skip attributes on table rows.  Previously we just crashed if
      rows had attributes, now we ignore them.
    + Ignore attributes on headers.
    + Allow `Image:` for images (#971).
    + Parse an image with caption in a paragraph by itself as a figure.

  * LaTeX writer

    + Don't use ligatures in escaping inline code.
    + Fixed footnote numbers in LaTeX/PDF tables.  This fixes a bug
      wherein notes were numbered incorrectly in tables (#827).
    + Always create labels for sections.  Previously the labels were only
      created when there were links to the section in the document (#871).
    + Stop escaping `|` in LaTeX math.
      This caused problems with array environments (#891).
    + Change `\` to `/` in paths.  `/` works even on Windows in LaTeX.
      `\` will cause major problems if unescaped.
    + Write id for code block to label attribute in LaTeX when listings
      is used (thanks to Florian Eitel).
    + Scale LaTeX tables so they don't exceed columnwidth.
    + Avoid problem with footnotes in unnumbered headers (#940).

  * Beamer writer: when creating beamer slides, add `allowframebreaks` option
      to the slide if it is one of the header classes.  It is recommended
      that your bibliography slide have this attribute:

        # References {.allowframebreaks}

    This causes multiple slides to be created if necessary, depending
    on the length of the bibliography.

  * ConTeXt writer: Properly handle tables without captions.  The old output
    only worked in MkII. This should work in MkIV as well (#837).

  * MediaWiki writer: Use native mediawiki tables instead of HTML (#720).

  * HTML writer:

    + Fixed `--no-highlight` (Alexander Kondratskiy).
    + Don't convert to lowercase in email obfuscation (#839).
    + Ensure proper escaping in `<title>` and `<meta>` fields.

  * AsciiDoc writer:

    + Support `--atx-headers` (Max Rydahl Andersen).
    + Don't print empty identifier blocks `([[]])` on headers (Max
      Rydahl Andersen).

  * ODT writer:

    + Fixing wrong numbered-list indentation in open document format
      (Alexander Kondratskiy) (#369).
    + `reference.odt`: Added pandoc as "generator" in `meta.xml`.
    + Minor changes for ODF 1.2 conformance (#939). We leave the
      nonconforming `contextual-spacing` attribute, which is provided by
      LibreOffice itself and seems well supported.

  * Docx writer:

    + Fixed rendering of display math in lists.
      In 1.11 and 1.11.1, display math in lists rendered as a new list
      item.  Now it always appears centered, just as outside of lists,
      and in proper display math style, no matter how far indented the
      containing list item is (#784).
    + Use `w:br` with `w:type` `textWrapping` for linebreaks.
      Previously we used `w:cr` (#873).
    + Use Compact style for Plain block elements, to
      differentiate between tight and loose lists (#775).
    + Ignore most components of `reference.docx`.
      We take the `word/styles.xml`, `docProps/app.xml`,
      `word/theme/theme1.xml`, and `word/fontTable.xml` from
      `reference.docx`, ignoring everything else.  This should help
      with the corruption problems caused when different versions of
      Word resave the reference.docx and reorganize things.
    +  Made `--no-highlight` work properly.

  * EPUB writer

    + Don't add `dc:creator` tags if present in EPUB metadata.
    + Add `id="toc-title"` to `h1` in `nav.xhtml` (#799).
    + Don't put blank title page in reading sequence.
      Set `linear="no"` if no title block.  Addresses #797.
    + Download webtex images and include as data URLs.
      This allows you to use `--webtex` in creating EPUBs.
      Math with `--webtex` is automatically made self-contained.
    + In `data/epub.css`, removed highlighting styles (which
      are no longer needed, since styles are added by the HTML
      writer according to `--highlighting-style`).  Simplified
      margin fields.
    + If resource not found, skip it, as in Docx writer (#916).

  * RTF writer:

    + Properly handle characters above the 0000-FFFF range.
      Uses surrogate pairs.  Thanks to Hiromi Ishii for the patch.
    + Fixed regression with RTF table of contents.
    + Only autolink absolute URIs.  This fixes a regression, #830.

  * Markdown writer:

    + Only autolink absolute URIs.  This fixes a regression, #830.
    + Don't wrap attributes in fenced code blocks.
    + Write full metadata in MMD style title blocks.
    + Put multiple authors on separate lines in pandoc titleblock.
      Also, don't wrap long author entries, as new lines get treated
      as new authors.

  * `Text.Pandoc.Templates`:

    + Fixed bug retrieving default template for markdown variants.
    + Templates can now contain "record lookups" in variables;
      for example, `author.institution` will retrieve the `institution`
      field of the `author` variable.
    + More consistent behavior of `$for$`.  When `foo` is not a list,
      `$for(foo)$...$endfor$` should behave like $if(foo)$...$endif$.
      So if `foo` resolves to "", no output should be produced.
      See pandoc-templates#39.

  * Citation processing improvements (now part of pandoc-citeproc):

    + Fixed `endWithPunct` The new version correctly sees a sentence
      ending in '.)' as ending with punctuation.  This fixes a bug which
      led such sentences to receive an extra period at the end: '.).'.
      Thanks to Steve Petersen for reporting.
    + Don't interfere with Notes that aren't citation notes.
      This fixes a bug in which notes not generated from citations were
      being altered (e.g. first letter capitalized) (#898).
    + Only capitalize footnote citations when they have a prefix.
    + Changes in suffix parsing.  A suffix beginning with a digit gets 'p'
      inserted before it before passing to citeproc-hs, so that bare numbers
      are treated as page numbers by default.  A suffix not beginning with
      punctuation has a space added at the beginning (rather than a comma and
      space, as was done before for not-author-in-text citations).
      The result is that `\citep[23]{item1}` in LaTeX will be interpreted
      properly, with '23' treated as a locator of type 'page'.
    + Many improvements to citation rendering, due to fixes in citeproc-hs
      (thanks to Andrea Rossato).
    + Warnings are issued for undefined citations, which are rendered
      as `???`.
    + Fixed hanging behavior when locale files cannot be found.

  [template changes]

  * DocBook:  Use DocBook 4.5 doctype.

  * Org: '#+TITLE:' is inserted before the title.
    Previously the writer did this.

  * LaTeX:  Changes to make mathfont work with xelatex.
    We need the mathspec library, not just fontspec, for this.
    We also need to set options for setmathfont (#734).

  * LaTeX: Use `tex-ansi` mapping for `monofont`.
    This ensures that straight quotes appear as straight, rather than
    being treated as curly.  See #889.

  * Made `\includegraphics` more flexible in LaTeX template.
    Now it can be used with options, if needed.  Thanks to Bernhard Weichel.

  * LaTeX/Beamer: Added `classoption` variable.
    This is intended for class options like `oneside`; it may
    be repeated with different options.  (Thanks to Oliver Matthews.)

  * Beamer: Added `fonttheme` variable.  (Thanks to Luis Osa.)

  * LaTeX: Added `biblio-style` variable (#920).

  * DZSlides: title attribute on title section.

  * HTML5: add meta tag to allow scaling by user (Erik Evenson)

  [under-the-hood improvements]

  * Markdown reader:Improved strong/emph parsing, using the strategy of
    <https://github.com/jgm/Markdown>.  The new parsing algorithm requires
    no backtracking, and no keeping track of nesting levels.  It will give
    different results in some edge cases, but these should not affect normal
    uses.

  * Added `Text.Pandoc.Compat.Monoid`.
    This allows pandoc to compile with `base` < 4.5, where `Data.Monoid`
    doesn't export `<>`.  Thanks to Dirk Ullirch for the patch.

  * Added `Text.Pandoc.Compat.TagSoupEntity`.
    This allows pandoc to compile with `tagsoup` 0.13.x.
    Thanks to Dirk Ullrich for the patch.

  * Most of `Text.Pandoc.Readers.TeXMath` has been moved to the
    `texmath` module (0.6.4).  (This allows `pandoc-citeproc` to
    handle simple math in bibliography fields.)

  * Added `Text.Pandoc.Writers.Shared` for shared functions used
    only in writers.  `metaToJSON` is used in writers to create a
    JSON object for use in the templates from the pandoc metadata
    and variables.  `getField`, `setField`, and `defField` are
    for working with JSON template contexts.

  * Added `Text.Pandoc.Asciify` utility module.
    This exports functions to create ASCII-only versions of identifiers.

  * `Text.Pandoc.Parsing`

    + Generalized state type on `readWith` (API change).
    + Specialize readWith to `String` input. (API change).
    + In `ParserState`, replace `stateTitle`, `stateAuthors`, `stateDate`
      with `stateMeta` and `stateMeta'`.

  * `Text.Pandoc.UTF8`: use strict bytestrings in reading.  The use of lazy
     bytestrings seemed to cause problems using pandoc on 64-bit Windows
     7/8 (#874).

  * Factored out `registerHeader` from markdown reader, added to
    `Text.Pandoc.Parsing`.

  * Removed `blaze_html_0_5` flag, require `blaze-html` >= 0.5.
    Reason:  < 0.5 does not provide a monoid instance for Attribute,
    which is now needed by the HTML writer (#803).

  * Added `http-conduit` flag, which allows fetching https resources.
    It also brings in a large number of dependencies (`http-conduit`
    and its dependencies), which is why for now it is an optional flag
    (#820).

  * Added CONTRIBUTING.md.

  * Improved INSTALL instructions.

  * `make-windows-installer.bat`: Removed explicit paths for executables.

  * `aeson` is now used instead of `json` for JSON.

  * Set default stack size to 16M.  This is needed for some large
    conversions, esp. if pandoc is compiled with 64-bit ghc.

  * Various small documentation improvements.
    Thanks to achalddave and drothlis for patches.

  * Removed comment that chokes recent versions of CPP (#933).

  * Removed support for GHC version < 7.2, since pandoc-types now
    requires at least GHC 7.2 for GHC generics.


pandoc (1.11.1)

  * Markdown reader:

    + Fixed regression in which parentheses were lost in link URLs.
      Added tests.  Closes #786.
    + Better handling of unmatched double quotes in `--smart` mode.
      These occur frequently in fiction, since it is customary not to
      close quotes in dialogue if the speaker does not change between
      paragraphs.  The unmatched quotes now get turned into literal
      left double quotes. (No `Quoted` inline is generated, however.)
      Closes #99 (again).

  * HTML writer: Fixed numbering mismatch between TOC and sections.
    `--number-offset` now affects TOC numbering as well
    as section numbering, as it should have all along.  Closes #789.

  * Markdown writer: Reverted 1.11 change that caused citations to be rendered
    as markdown citations, even if `--bibliography` was specified, unless
    `citation` extension is disabled.  Now, formatted citations are always
    printed if `--bibliography` was specified.  If you want to reformat
    markdown keeping pandoc markdown citations intact, don't use
    `--bibliography`.  Note that citations parsed from LaTeX documents will
    be rendered as pandoc markdown citations when `--bibliography` is not
    specified.

  * ODT writer: Fixed regression leading to corrupt ODTs.
    This was due to a change in the `Show` instance for
    `Text.Pandoc.Pretty.Doc`.  Closes #780.

  * Fixed spacing bugs involving code block attributes in
    RST reader and Markdown writer. Closes #763.

  * Windows package:  Various improvements due to Fyodor Sheremetyev.

    + Automatically set installation path (Program Files or Local App Data).
    + Set system PATH environment variable when installing for all users.
    + Pandoc can installed for all users using the following command.
      `msiexec /i pandoc-1.11.msi ALLUSERS=1`.

  * Bumped QuickCheck version bound.

pandoc (1.11)

  * Added `--number-offset` option.  (See README for description.)

  * Added `--default-image-extension` option.  (See README for description.)

  * `--number-sections` behavior change: headers with class `unnumbered`
    will not be numbered.

  * `--version` now reports the default data directory.

  * `Text.Pandoc.Parsing` is no longer exposed. (API change.)

  * `Text.Pandoc.Highlighting` is no longer exposed. (API change.)

  * `Text.Pandoc.Shared`:  Changed type of `Element`.  `Sec` now includes
    a field for `Attr` rather than just `String`.  (API change.)

  * Added `markdown_github` as input format.  This was an accidental
    omission in 1.10.

  * Added `readerDefaultImageExtension` field to `ReaderOptions`.  (API
    change.)

  * Added `writerNumberOffset` field in `WriterOptions`. (API change.)

  * Beamer template:

    + Fixed captions with longtable.  Thanks to Joost Kremers.
    + Provide `\Oldincludegraphics` as in LaTeX template (Benjamin Bannier).

  * LaTeX template:

    + Load microtype after fonts.  Microtype needs to know
      what fonts are being used.  Thanks to dfc for the patch.
    + Set `secnumdepth` to 5 if `--number-sections` specified.
      This yields behavior equivalent to the other writers, numbering
      level 4 and 5 headers too.  Closes #753.

  * HTML reader:

    + Handle `<colgroup>` tag.
    + Preserve all header attributes.

  * LaTeX reader:

    + Parse `\hrule` as `HorizontalRule`.  Closes #746.
    + Parse starred variants of `\section` etc. as headers with
      attribute `unnumbered`.
    + Read optional attributes in `lstlisting` and `Verbatim` environments.
      We convert these to pandoc standard names, e.g. `numberLines`
      for `numbers=left`, `startFrom=100` from `firstnumber=100`.
    + Handle language attribute for lstlistings.
    + Better support for Verbatim and minted environments.  Closes #763.

  * Markdown reader:

    + `-` in an attribute context = `.unnumbered`.  The point of this
      is to provide a way to specify unnumbered headers in non-English
      documents.
    + Fixed bug parsing key/value attributes.  Parsing failed if you
      had an unquoted attribute immediately before the final '}'.
    + Make backslash escape work in attributes.
    + Fix title block parsing.  Now if `mmd_title_blocks` is specified,
      pandoc will parse a MMD title block if it sees one, even if
      `pandoc_title_blocks` is enabled.
    + Refactoring: `litChar` now includes entities, so we don't need
      to use `fromEntities` e.g. on titles.
    + Allow spaces around borders in pipe tables.  Closes #772.
    + Allow all punctuation in angle-bracket autolinks.  Previously
      things like `----` were disallowed, because the uri parser
      treated them as trailing punctuation.  Closes #768.
    + Make `implicit_header_references` work properly when
      headers are given explicit identifiers.
    + Check for tables before line blocks.  Otherwise some pipe
      tables get treated as line blocks.
    + Allow `&` in emails (for entities).
    + Properly handle entities in titles and links.  A markdown link
      `<http://g&ouml;ogle.com>` should be a link to `http://göogle.com`.
      Closes #723.

  * Textile reader:

    + Handle attributes on headers.

  * LaTeX reader:

    + Add `fig:` as title for images with captions.
      This is needed for them to be rendered as figures.  Closes #766.
    + Never emit an empty paragraph.  See #761.
    + Handle `\caption` for images in figures.  Closes #766.
    + Parse `\section*`, etc. as unnumbered sections.

  * HTML writer:

    + Support header attributes.  The attributes go on
      the enclosing `section` or `div` tag if `--section-divs` is specified.
    + Fixed a regression (only now noticed) in html+lhs output.
      Previously the bird tracks were being omitted.

  * LaTeX writer:

    + Omit lists with no items to avoid LaTeX errors.
    + Support line numbering with `--listings`.
      If `numberLines` class is present, we add `numbers=left`;
      if `startFrom` is present, we add `firstnumber=`. (#763)

  * ConTeXt writer:

    + Removed `\placecontent`.  This produced a duplicate toc,
      in conjunction with `\placelist`.
    + Use `\title`, `\subject` etc. for headers with
      `unnumbered` class.

  * Textile writer:

    + Support header attributes.

  * Markdown writer:

    + Use grid tables when needed, and if enabled.  Closes #740.
    + Render citations as pandoc-markdown citations.
      Previously citations were rendered as citeproc-formatted citations
      by default.  Now we render them as pandoc citations, e.g. `[@item1]`,
      unless the `citations` extension is disabled.
      If you still want formatted citations in your markdown output,
      use `pandoc -t markdown-citations`.

  * RST writer:

    + Support `:number-lines:` in code blocks.

  * Docx writer:

    + Better treatment of display math.  Display math inside a
      paragraph is now put in a separate paragraph, so it will render
      properly (centered and without extra blank lines around it).
      Partially addresses #742.
    + Content types and document rels xml files are now created from
      scratch, rather than being taken over from `reference.docx`.
      This fixes problems that arise when you edit the `reference.docx`
      with Word.
    + We also now encode mime types for each individual image rather
      than using defaults.  This should allow us to handle a wider
      range of image types (including PDF).   Closes #414.
    + Changed style names in `reference docx`.
      `FootnoteReference` -> `FootnoteRef`, `Hyperlink` -> `Link`.
      The old names got changed by Word when the `reference.docx` was
      edited.  Closes #414.

  * EPUB writer:

    + Fix section numbering.  Previously the numbering restarted from 1
      in each chapter (with `--number-sections`), though the numbers in
      the table of contents were correct.
    + Headers with "unnumbered" attribute are not numbered.  (Nor do they
      cause an increment in running numbering.) Section numbers now work
      properly, even when there is material before the first numbered section.
    + Include HTML TOC, even in epub2.  The TOC is included in `<spine>`,
      but `linear` is set to `no` unless the `--toc` option is specified.
      Include `<guide>` element in OPF.  This should allow the TOC to
      be useable in Kindles when converted with kindlegen. Closes #773.

  * `Text.Pandoc.Parsing`: Optimized `oneOfStringsCI`.
    This dramatically reduces the speed penalty that comes from enabling the
    `autolink_bare_uris` extension.  The penalty is still substantial (in one
    test, from 0.33s to 0.44s), but nowhere near what it used to be.
    The RST reader is also much faster now, as it autodetects URIs.

  * `Text.Pandoc.Shared`:  `hierarchicalize` will not number section
    with class "unnumbered".  Unnumbered sections get `[]` for their
    section number.

  * `Text.Pandoc.Pretty`:

    + Fixed `chomp` so it works inside `Prefixed` elements.
    + Changed `Show` instance so it is better for debugging.

  * `Text.Pandoc.ImageSize`:  Added `Pdf` to `ImageType`.

  * `Text.Pandoc.UTF8`:  Strip off BOM if present.  Closes #743.

  * Windows installer improvements:

    + The installer is now signed with a certificate (thanks to
      Fyodor Sheremetyev).
    + WiX is used instead of InnoSetup.  The installer is now a
      standard msi file.
    + The version number is now auto-detected, and need not be
      updated separately.

  * OSX installer improvements:

    + The package and pandoc executable are now signed with a
      certificate (thanks to Fyodor Sheremetyev).
    + RTF version of license is used.
    + Use full path for sysctl in `InstallationCheck` script (jonahbull).
      Closes #580.

  * Converted COPYING to markdown.

  * pandoc.cabal:  Require latest versions of highlighting-kate,
    texmath, citeproc-hs, zip-archive.

pandoc (1.10.1)

  * Markdown reader:  various optimizations, leading to a
    significant performance boost.

  * RST reader:  Allow anonymous form of inline links:
    `` `hello <url>`__ `` Closes #724.

  * Mediawiki reader: Don't require newlines after tables.
    Thanks to jrunningen for the patch. Closes #733.

  * Fixed LaTeX macro parsing.  Now LaTeX macro definitions are preserved
    when output is LaTeX, and applied when it is another format.
    Partially addresses #730.

  * Markdown and RST readers:  Added parser to `block` that skips blank
    lines.  This fixes a subtle regression involving grid tables with
    empty cells.  Also added test for grid table with empty cells.
    Closes #732.

  * RST writer:  Use `.. code:: language` for code blocks with language.
    Closes #721.

  * DocBook writer:  Fixed output for hard line breaks, adding a newline
    between `<literallayout>` tags.

  * Markdown writer:  Use an autolink when link text matches url.
    Previously we also checked for a null title, but this
    test fails for links produced by citeproc-hs in bibliographies.
    So, if the link has a title, it will be lost on conversion
    to an autolink, but that seems okay.

  * Markdown writer:  Set title, author, date variables as before.
    These are no longer used in the default template, since we use
    titleblock, but we set them anyway for those who use custom templates.

  * LaTeX writer:  Avoid extra space at start/end of table cell.
    Thanks to Nick Bart for the suggestion of using @{}.

  * `Text.Pandoc.Parsing`:

    + More efficient version of `anyLine`.
    + Type of `macro` has changed; the parser now returns `Blocks`
      instead of `Block`.

  * Relaxed old-time version bound, allowing 1.0.*.

  * Removed obsolete `hsmarkdown` script.  Those who need `hsmarkdown`
    should create a symlink as described in the README.

pandoc (1.10.0.5)

  * Markdown reader: Try `lhsCodeBlock` before `rawTeXBlock`.  Otherwise
    `\begin{code}...\end{code}` isn't handled properly in markdown+lhs.
    Thanks to Daniel Miot for noticing the bug and suggesting the fix.

  * Markdown reader: Fixed bug with headerless grid tables.
    The 1.10 code assumed that each table header cell contains exactly one
    block. That failed for headerless tables (0) and also for tables with
    multiple blocks in a header cell.  The code is fixed and tests provided.
    Thanks to Andrew Lee for pointing out the bug.

  * Markdown reader: Fixed regressions in fenced code blocks. Closes #722.

    + Tilde code fences can again take a bare language string
      (`~~~ haskell`), not just curly-bracketed attributes
      (`~~~ {.haskell}`).
    + Backtick code blocks can take the curly-bracketed attributes.
    + Backtick code blocks don't *require* a language.
    + Consolidated code for the two kinds of fenced code blocks.

  * LaTeX template: Use `\urlstyle{same}` to avoid monospace URLs.

  * Markdown writer: Use proportional font for email autolinks with
    obfuscation.  Closes #714.

  * Corrected name of `blank_before_blockquote` in README.  Closes #718.

  * `Text.Pandoc.Shared`: Fixed bug in `uri` parser.
    The bug prevented an autolink at the end of a string (e.g.
    at the end of a line block line) from counting as a link.  Closes #711.

  * Use the `hsb2hs` preprocessor instead of TH for embed_data_files.
    This should work on Windows, unlike the TH solution with
    file-embed.

  * Eliminated use of TH in test suite.

  * Added `Text.Pandoc.Data` (non-exported) to hold the association
    list of embedded data files, if the `embed_data_files` flag is selected.
    This isolates the code that needs special treatment with file-embed or
    `hsb2hs`.

  * Changes to `make-windows-installer.bat`.

    + Exit batch file if any of the cabal-dev installs fail.
    + There's no longer any need to reinstall `highlighting-kate`.
    + Don't start with a `cabal update`; leave that to the user.
    + Force reinstall of pandoc.

  * Fixed EPUB writer so it builds with blaze-html 0.4.x. Thanks to
    Jens Petersen.

pandoc (1.10.0.4)

  * Fixed bug with escaped % in LaTeX reader. Closes #710.

pandoc (1.10.0.3)

  * Added further missing fb2 tests to cabal file.

pandoc (1.10.0.2)

  * Added fb2 tests to cabal file's extra-source-files.

pandoc (1.10.0.1)

  * Bump version bounds on test-framework packages.

pandoc (1.10)

  [new features]

  * New input formats:  `mediawiki` (MediaWiki markup).

  * New output formats:  `epub3` (EPUB v3 with MathML),
    `fb2` (FictionBook2 ebooks).

  * New `--toc-depth` option, specifying how many levels of
    headers to include in a table of contents.

  * New `--epub-chapter-level` option, specifying the header
    level at which to divide EPUBs into separate files.
    Note that this normally affects only performance, not the
    visual presentation of the EPUB in a reader.

  * Removed the `--strict` option.  Instead of using `--strict`,
    one can now use the format name `markdown_strict` for either input
    or output.  This gives more fine-grained control that `--strict`
    did, allowing one to convert from pandoc's markdown to strict
    markdown or vice versa.

  * It is now possible to enable or disable specific syntax
    extensions by appending them (with `+` or `-`) to the writer
    or reader name.  For example,

        pandoc -f markdown-footnotes+hard_line_breaks

    disables footnotes and enables treating newlines as hard
    line breaks.  The literate Haskell extensions are now implemented
    this way as well, using either `+lhs` or `+literate_haskell`.
    For a list of extension names, see the README under
    "Pandoc's Markdown."

  * The following aliases have been introduced for specific
    combinations of markdown extensions:  `markdown_phpextra`,
    `markdown_github`, `markdown_mmd`, `markdown_strict`.  These aliases
    work just like regular reader and writer names, and can be modified
    with extension modifiers as described above. (Note that conversion
    from one markdown dialect to another does not work perfectly,
    because there are differences in markdown parsers besides
    just the extensions, and because pandoc's internal document model is
    not rich enough to capture all of the extensions.)

  * New `--html-q-tags` option.  The previous default was to use `<q>`
    tags for smart quotes in HTML5.  But `<q>` tags are also valid HTML4.
    Moreover, they are not a robust way of typesetting quotes, since
    some user agents don't support them, and some CSS resets (e.g.
    bootstrap) prevent pandoc's quotes CSS from working properly.
    We now just insert literal quote characters by default in both
    `html` and `html5` output, but this option is provided for
    those who still want `<q>` tags.

  * The markdown reader now prints warnings (to stderr) about
    duplicate link and note references.  Closes #375.

  * Markdown syntax extensions:

    + Added pipe tables.  Thanks to François Gannaz for the initial patch.
      These conform to PHP Markdown Extra's pipe table syntax. A subset
      of org-mode table syntax is also supported, which means that you can
      use org-mode's nice table editor to create tables.

    + Added support for RST-style line blocks. These are
      useful for verse and addresses.

    + Attributes can now be specified for headers, using the same
      syntax as in code blocks.  (However, currently only the
      identifier has any effect in most writers.)  For example,

            # My header {#foo}

            See [the header above](#foo).

    + Pandoc will now act as if link references have been defined
      for all headers without explicit identifiers.
      So, you can do this:

            # My header

            Link to [My header].
            Another link to [it][My header].

      Closes #691.

  * LaTeX reader:

    + Command macros now work everywhere, including non-math.
      Environment macros still not supported.
    + `\input` now works, as well as `\include`.  TEXINPUTS is used.
      Pandoc looks recursively into included files for more included files.

  [behavior changes]

  * The Markdown reader no longer puts the text of autolinks in a
    `Code` inline.  This means that autolinks will no longer appear
    in a monospace font.

  * The character `/` can now appear in markdown citation keys.

  * HTML blocks in strict_markdown are no longer required to begin
    at the left margin.  Technically this is required, according to
    the markdown syntax document, but `Markdown.pl` and other markdown
    processors are more liberal.

  * The `-V` option has been changed so that if there are duplicate
    variables, those specified later on the command line take precedence.

  * Tight lists now work in LaTeX and ConTeXt output.

  * The LaTeX writer no longer relien on the `enumerate` package.
    Instead, it uses standard LaTeX commands to change the list numbering
    style.

  * The LaTeX writer now uses `longtable` instead of `ctable`. This allows
    tables to be split over page boundaries.

  * The RST writer now uses a line block to render paragraphs containing
    linebreaks (which previously weren't supported at all).

  * The markdown writer now applies the `--id-prefix` to footnote IDs.
    Closes #614.

  * The plain writer no longer uses backslash-escaped line breaks
    (which are not very "plain").

  * `Text.Pandoc.UTF8`: Better error message for invalid UTF8.
    Read bytestring and use `Text`'s decodeUtf8 instead of using
    `System.IO.hGetContents`.  This way you get a message saying
    "invalid UTF-8 stream" instead of "invalid byte sequence."
    You are also told which byte caused the problem.

  * Docx, ODT, and EPUB writers now download images specified by a URL
    instead of skipping them or raising an error.

  * EPUB writer:

    + The default CSS now left-aligns headers by default, instead of
      centering.  This is more consistent with the rest of the writers.
    + A proper multi-level table of contents is now used in `toc.ncx`.
      There is no longer a subsidiary table of contents at the beginning
      of each chapter.
    + Code highlighting now works by default.
    + Section divs are used by default for better semantic markup.
    + The title is used instead of "Title Page" in the table of contents.
      Otherwise we have a hard-coded English string, which looks
      strange in ebooks written in other languages.  Closes #572.

  * HTML writer:

    + Put mathjax in span with class "math".  Closes #562.
    + Put citations in a span with class "citation." In HTML5, also include
      a `data-cite` attribute with a space-separated list of citation
      keys.

  * `Text.Pandoc.UTF8`:  use universalNewlineMode in reading.
    This treats both `\r\n` and `\n` as `\n` on input, no matter
    what platform we're running on.

  * Citation processing is now done in the Markdown and LaTeX
    readers, not in `pandoc.hs`.  This makes it easier for library users
    to use citations.

  [template changes]

  * HTML: Added css to template to preserve spaces in `<code>` tags.
    Thanks to Dirk Laurie.

  * Beamer:  Remove English-centric strings in section pages.
    Section pages used to have "Section" and a number as well as the
    section title. Now they just have the title.  Similarly for part
    and subsection.  Closes #566.

  * LaTeX, ConTeXt: Added papersize variable.

  * LaTeX, Beamer templates: Use longtable instead of ctable.

  * LaTeX, Beamer templates: Don't require 'float' package for tables.
    We don't actually seem to use the '[H]' option.

  * LaTeX:  Use `upquote` package if it is available.  This fixes
    straight quotes in verbatim environments.

  * Markdown, plain: Fixed titleblock so it is just a single string.
    Previously separate title, author, and date variables were used,
    but this didn't allow different kinds of title blocks.

  * EPUB:

    + Rationalized templates.  Previously there were three different
      templates involved in epub production. There is now just one
      template, `default.epub` or `default.epub3`. It can now be
      overridden using `--template`, just like other templates.
      The titlepage is now folded into the default template.
      A `titlepage` variable selects it.
    + UTF-8, lang tag, meta tags, title element.

  * Added scale-to-width feature to beamer template

  [API changes]

  * `Text.Pandoc.Definition`: Added `Attr` field to `Header`.
    Previously header identifers were autogenerated by the writers.
    Now they are added in the readers (either automatically or explicitly).

  * `Text.Pandoc.Builder`:

    + `Inlines` and `Blocks` are now synonyms for `Many Inline` and
      `Many Block`.  `Many` is a newtype wrapper around `Seq`, with
      custom Monoid instances for `Many Inline` and `Many Block.  This
      allows `Many` to be made an instance of `Foldable` and `Traversable`.
    + The old `Listable` class has been removed.
    + The module now exports `isNull`, `toList`, `fromList`.
    + The old `Read` and `Show` instances have been removed; derived
      instances are now used.
    + Added `headerWith`.

  * The readers now take a `ReaderOptions` rather than a `ParserState`
    as a parameter.  Indeed, not all parsers use the `ParserState` type;
    some have a custom state.  The motivation for this change was to separate
    user-specifiable options from the accounting functions of parser state.

  * New module `Text.Pandoc.Options`.  This includes the `WriterOptions`
    formerly in `Text.Pandoc.Shared`, and its associated
    data types.  It also includes a new type `ReaderOptions`, which
    contains many options formerly in `ParserState`, and its associated
    data types:

    + `ParserState.stateParseRaw` -> `ReaderOptions.readerParseRaw`.
    + `ParserState.stateColumns` -> `ReaderOptions.readerColumns`.
    + `ParserState.stateTabStop` -> `ReaderOptions.readerTabStop`.
    + `ParserState.stateOldDashes` -> `ReaderOptions.readerOldDashes`.
    + `ParserState.stateLiterateHaskell` -> `ReaderOptions.readerLiterateHaskell`.
    + `ParserState.stateCitations` -> `ReaderOptions.readerReferences`.
    + `ParserState.stateApplyMacros` -> `ReaderOptions.readerApplyMacros`.
    + `ParserState.stateIndentedCodeClasses` ->
      `ReaderOptions.readerIndentedCodeClasses`.
    + Added `ReaderOptions.readerCitationStyle`.

  * `WriterOptions` now includes `writerEpubVersion`, `writerEpubChapterLevel`,
    `writerEpubStylesheet`, `writerEpubFonts`, `writerReferenceODT`,
    `writerReferenceDocx`, and `writerTOCDepth`.  `writerEPUBMetadata` has
    been renamed `writerEpubMetadata` for consistency.

  * Changed signatures of `writeODT`, `writeDocx`, `writeEPUB`, since they no
    longer stylesheet, fonts, reference files as separate parameters.

  * Removed `writerLiterateHaskell` from `WriterOptions`, and
    `readerLiterateHaskell` from `ReaderOptions`.  LHS is now handled
    by an extension (`Ext_literate_haskell`).

  * Removed deprecated `writerXeTeX`.

  * Removed `writerStrict` from `WriterOptions`.  Added `writerExtensions`.
    Strict is now handled through extensions.

  * `Text.Pandoc.Options` exports `pandocExtensions`, `strictExtensions`,
    `phpMarkdownExtraExtensions`, `githubMarkdownExtensions`,
    and `multimarkdownExtensions`, as well as the `Extensions` type.

  * New `Text.Pandoc.Readers.MediaWiki` module, exporting
    `readMediaWiki`.

  * New `Text.Pandoc.Writers.FB2` module, exporting `writeFB2`
    (thanks to Sergey Astanin).

  * `Text.Pandoc`:

    + Added `getReader`, `getWriter` to `Text.Pandoc`.
    + `writers` is now an association list `(String, Writer)`.
      A `Writer` can be a `PureStringWriter`, an `IOStringWriter`, or
      an `IOByteStringWriter`.  ALL writers are now in the 'writers'
      list, including the binary writers and FB2 writer.  This allows
      code in `pandoc.hs` to be simplified.
    + Changed type of `readers`, so all readers are in IO.
      Users who want pure readers can still get them form the reader
      modules; this just affects the function `getReader` that looks up
      a reader based on the format name.  The point of this change is to
      make it possible to print warnings from the parser.

  * `Text.Pandoc.Parsing`:

    + `Text.Parsec` now exports all Parsec functions used in pandoc code.
      No other module directly imports Parsec.  This will make it easier
      to change the parsing backend in the future, if we want to.
    + `Text.Parsec` is used instead of `Text.ParserCombinators.Parsec`.
    + Export the type synonym `Parser`.
    + Export `widthsFromIndices`, `NoteTable'`, `KeyTable'`, `Key'`, `toKey'`,
     `withQuoteContext`, `singleQuoteStart`, `singleQuoteEnd`,
     `doubleQuoteStart`, `doubleQuoteEnd`, `ellipses`, `apostrophe`,
     `dash`, `nested`, `F(..)`, `askF`, `asksF`, `runF`, `lineBlockLines`.
    + `ParserState` is no longer an instance of `Show`.
    + Added `stateSubstitutions` and `stateWarnings` to `ParserState`.
    + Generalized type of `withQuoteContext`.
    + Added `guardEnabled`, `guardDisabled`, `getOption`.
    + Removed `failIfStrict`.
    + `lookupKeySrc` and `fromKey` are no longer exported.

  * `Data.Default` instances are now provided for `ReaderOptions`,
    `WriterOptions`, and `ParserState`.  `Text.Pandoc` re-exports `def`.
    Now you can use `def` (which is re-exported by `Text.Pandoc`) instead
    of `defaultWriterOptions` (which is still defined).  Closes #546.

  * `Text.Pandoc.Shared`:

    + Added `safeRead`.
    + Renamed `removedLeadingTrailingSpace` to `trim`,
      `removeLeadingSpace` to `triml`, and `removeTrailingSpace` to `trimr`.
    + Count `\r` as space in `trim` functions.
    + Moved `renderTags'` from HTML reader and `Text.Pandoc.SelfContained`
      to `Shared`.
    + Removed `failUnlessLHS`.
    + Export `compactify'`, formerly in Markdown reader.
    + Export `isTightList`.
    + Do not export `findDataFile`.
    + `readDataFile` now returns a strict ByteString.
    + Export `readDataFileUTF8` which returns a String, like the
      old `readDataFile`.
    + Export `fetchItem` and `openURL`.

  * `Text.Pandoc.ImageSize`: Use strict, not lazy bytestrings.
    Removed `readImageSize`.

  * `Text.Pandoc.UTF8`: Export `encodePath`, `decodePath`,
     `decodeArg`, `toString`, `fromString`, `toStringLazy`,
     `fromStringLazy`.

  * `Text.Pandoc.UTF8` is now an exposed module.

  * `Text.Pandoc.Biblio`:

    + csl parameter now a `String` rather than a `FilePath`.
    + Changed type of `processBiblio`.  It is no longer in the IO monad.
      It now takes a `Maybe Style` argument rather than parameters for CSL
      and abbrev filenames.  (`pandoc.hs` now calls the functions to parse
      the style file and add abbreviations.)

  * Markdown reader now exports `readMarkdownWithWarnings`.

  * `Text.Pandoc.RTF` now exports `writeRTFWithEmbeddedImages` instead of
    `rtfEmbedImage`.

  [bug fixes]

  * Make `--ascii` work properly with `--self-contained`.  Closes #568.

  * Markdown reader:

    + Fixed link parser to avoid exponential slowdowns.  Closes #620.
      Previously the parser would hang on input like this:

            [[[[[[[[[[[[[[[[[[hi

      We fixed this by making the link parser parser characters
      between balanced brackets (skipping brackets in inline code spans),
      then parsing the result as an inline list.  One change is that

            [hi *there]* bud](/url)

      is now no longer parsed as a link.  But in this respect pandoc behaved
      differently from most other implementations anyway, so that seems okay.

    + Look for raw html/latex blocks before tables.
      Otherwise the following gets parsed as a table:

            \begin{code}
            --------------
            -- My comment.
            \end{code}

      Closes #578.

  * RST reader:

    + Added support for `:target:` on `.. image::` blocks
      and substitutions.
    + Field list fixes:

        - Fixed field lists items with body beginning after a new line
          (Denis Laxalde).
        - Allow any char but ':' in names of field lists in RST reader
          (Denis Laxalde).
        - Don't allow line breaks in field names.
        - Require whitespace after field list field names.
        - Don't create empty definition list for metadata field lists.
          Previously a field list consisting only of metadata fields (author,
          title, date) would be parsed as an empty DefinitionList, which is
          not legal in LaTeX and not needed in any format.

    + Don't recognize inline-markup starts inside words.
      For example, `2*2 = 4*1` should not contain an emphasized
      section.  Added test case for "Literal symbols".  Closes #569.
    + Allow dashes as separator in simple tables.  Closes #555.
    + Added support for `container`, `compound`, `epigraph`,
      `rubric`, `highlights`, `pull-quote`.
    + Added support for `.. code::`.
    + Made directive labels case-insensitive.
    + Removed requirement that directives begin at left margin.
      This was (correctly) not in earlier releases; docutils doesn't
      make the requirement.
    + Added support for `replace::` and `unicode::` substitutions.
    + Ignore unknown interpreted roles.
    + Renamed image parser to `subst`, since it now handles all
      substitution references.

  * Textile reader:

    + Allow newlines before pipes in table.  Closes #654.
    + Fixed bug with list items containing line breaks.
      Now pandoc correctly handles hard line breaks inside list items.
      Previously they broke list parsing.
    + Implemented comment blocks.
    + Fixed bug affected words ending in hyphen.
    + Properly handle links with surrounding brackets.
      Square brackets need to be used when the link isn't surrounded by
      spaces or punctuation, or when the URL ending may be ambiguous.
      Closes #564.
    + Removed nullBlock.  Better to know about parsing problems than
      to skip stuff when we get stuck.
    + Allow ID attributes on headers.
    + Textile reader:  Avoid parsing dashes as strikeout.
      Previously the input

            text--
            text--
            text--
            text--

      would be parsed with strikeouts rather than dashes. This fixes
      the problem by requiring that a strikeout delimiting - not be
      followed by a -.  Closes #631.
    + Expanded list of `stringBreakers`.
      This fixes a bug on input like "(_hello_)" which should
      be a parenthesized emphasized "hello".
      The new list is taken from the PHP source of textile 2.4.
    + Fixed autolinks.  Previously the textile reader and writer
      incorrectly implented RST-style autolinks for URLs and email
      addresses.  This has been fixed.  Now an autolink is done this way:
      `"$":http://myurl.com`.
    + Fixed footnotes bug in textile.  This affected notes occuring
      before punctuation, e.g. `foo[1].`.  Closes #518.

  * LaTeX reader:

    + Better handling of citation commands.
    + Better handling of `\noindent`.
    + Added a 'try' in rawLaTeXBlock, so we can handle `\begin` without `{`.
      Closes #622.
    + Made `rawLaTeXInline` try to parse block commands as well.  This
      is usually what we want, given how `rawLaTeXInline` is used in
      the markdown and textile readers.  If a block-level LaTeX command
      is used in the middle of a paragraph (e.g. `\subtitle` inside a title),
      we can treat it as raw inline LaTeX.
    + Handle \slash command.  Closes #605.
    + Basic `\enquote` support.
    + Fixed parsing of paragraphs beginning with a group.  Closes #606.
    + Use curly quotes for bare straight quotes.
    + Support obeylines environment.  Closes #604.
    + Guard against "begin", "end" in inlineCommand and
      blockCommand.
    + Better error messages for environments.  Now it should tell you that
      it was looking for \end{env}, instead of giving "unknown parse error."

  * HTML reader:

    + Added HTML 5 tags to list of block-level tags.
    + HTML reader: Fixed bug in `htmlBalanced`, which
      caused hangs in parsing certain markdown input using
      strict mode.
    + Parse `<q>` as `Quoted DoubleQuote`.
    + Handle nested `<q>` tags properly.
    + Modified `htmlTag` for fewer false positives.
      A tag must start with `<` followed by `!`,`?`, `/`, or a letter.
      This makes it more useful in the wikimedia and markdown parsers.

  * DocBook reader: Support title in "figure" element.  Closes #650.

  * MediaWiki writer:

    + Remove newline after `<br/>` in translation of `LineBreak`
      There's no particular need for a newline (other than making the
      generated MediaWiki source look nice to a human), and in fact
      sometimes it is incorrect: in particular, inside an enumeration, list
      items cannot have embedded newline characters. (Brent Yorgey)
    + Use `<code>` not `<tt>` for Code.

  * Man writer: Escape `-` as `\-`.
    Unescaped `-`'s become hyphens, while `\-`'s are left as ascii minus
    signs.  That is preferable for use with command-line options.  See
    <http://lintian.debian.org/tags/hyphen-used-as-minus-sign.html>.  Thanks
    to Andrea Bolognani for bringing the issue to our attention.

  * RST writer:

    + Improved line block output. Use nonbreaking spaces for
      initial indent (otherwise lost in HTML and LaTeX).
      Allow multiple paragraphs in a single line block.
      Allow soft breaks w continuations in line blocks.
    + Properly handle images with no alt text.  Closes #678.
    + Fixed bug with links with duplicate text.  We now (a) use anonymous
      links for links with inline URLs, and (b) use an inline link instead
      of a reference link if the reference link would require a label that
      has already been used for a different link.  Closes #511.
    + Fixed hyperlinked images. Closes #611. Use `:target:`
      field when you have a simple linked image.
    + Don't add `:align: center` to figures.

  * Texinfo writer:  Fixed internal cross-references.
    Now we insert anchors after each header, and use `@ref` instead of `@uref`
    for links.  Commas are now escaped as `@comma{}` only when needed;
    previously all commas were escaped.  (This change is needed, in part,
    because `@ref` commands must be followed by a real comma or period.) Also
    insert a blank line in from of `@verbatim` environments.

  * DocBook writer:

    + Made --id-prefix work in DocBook as well as HTML.
      Closes #607.
    + Don't include empty captions in figures.  Closes #581.

  * LaTeX writer:

    + Use `\hspace*` for nonbreaking space after line break,
      since `~` spaces after a line break are just ignored.
      Closes #687.
    + Don't escape `_` in URLs or hyperref identifiers.
    + Properly escape strings inside \url{}.  Closes #576.
    + Use `[fragile]` only for slides containing code rendered
      using listings. Closes #649.
    + Escape `|` as `\vert` in LaTeX math.  This avoids a clash with
      highlighting-kate's macros, which redefine `|` as a short verbatim
      delimiter.  Thanks to Björn Peemöller for raising this issue.
    + Use minipage rather than parbox for block containers in tables.
      This allows verbatim code to be included in grid tables.
      Closes #663.
    + Prevent paragraphs containing only linebreaks or spaces.

  * HTML writer:

    + Included `highlighting-css` for code spans, too.
      Previously it was only included if used in a code block.  Closes #653.
    + Improved line breaks with `<dd>` tags.  We now put a newline between
      `</dd>` and `<dd>` when there are multiple definitions.
    + Changed mathjax cdn url so it doesn't use https.  (This caused
      problems when used with `--self-contained`.) See #609.

  * EPUB writer:

    + `--number-sections` now works properly.
    + Don't strip meta and link elements in epub metadata.
      Patch from aberrancy. Closes #589.
    + Fixed a couple validation bugs.
    + Use ch001, ch002, etc. for chapter filenames.  This improves sorting
      of chapters in some readers, which apparently sort ch2 after ch10.
      Closes #610.

  * ODT writer: properly set title property (Arlo O'Keeffe).

  * Docx writer:

    + Fixed bug with nested lists.  Previously a list like

            1. one
                - a
                - b
            2. two

      would come out with a bullet instead of "2."
      Thanks to Russell Allen for reporting the bug.
    + Use `w:cr` in `w:r` instead of `w:br` for linebreaks.
      This seems to fix a problem viewing pandoc-generated
      docx files in LibreOffice.
    + Use integer ids for bookmarks.  Closes #626.
    + Added nsid to abstractNum elements.  This helps when merging
      word documents with numbered or bulleted lists.  Closes #627.
    + Use separate footnotes.xml for notes.
      This seems to help LibreOffice convert the file, even though
      it was valid docx before.  Closes #637.
    + Use rIdNN identifiers for r:embed in images.
    + Avoid reading image files again when we've already processed them.
    + Fixed typo in `referenc.docx` that prevented image captions from
      working. Thanks to Huashan Chen.

  * `Text.Pandoc.Parsing`:

    + Fixed bug in `withRaw`, which didn't correctly handle the case
      where nothing is parsed.
    + Made `emailAddress` parser more correct.  Now it is based on RFC 822,
      though it still doesn't implement quoted strings in email addresses.
    + Revised URI parser.  It now allows many more schemes, allows
      uppercase URIs, and better handles trailing punctuation and
      trailing slashes in bare URIs.  Added many tests.
    + Simplified and improved singleQuoteStart.  This makes `'s'`, `'l'`,
      etc. parse properly.  Formerly we had some English-centric heuristics,
      but they are no longer needed. Closes #698.

  * `Text.Pandoc.Pretty`:  Added wide punctuation range to `charWidth`.
    This fixes bug with Chinese commas in markdown and reST tables, and
    a bug that caused combining characters to be dropped.

  * `Text.Pandoc.MIME`: Added MIME types for .wof and .eot.  Closes #640.

  * `Text.Pandoc.Biblio`:

    + Run `mvPunc` and `deNote` on metadata too.
      This fixed a bug with notes on titles using footnote styles.
    + Fixed bug in fetching CSL files from CSL data directory.

  * `pandoc.hs`:  Give correct value to `writerSourceDirectory` when a URL
    is provided.  It should be the URL up to the path.

  * Fixed/simplified diff output for tests.
    Biblio: Make sure mvPunc and deNote run on metadata too.
    This fixed a bug with notes on titles using footnote styles.

  [under the hood improvements]

  * We no longer depend on `utf8-string`.  Instead we use functions
    defined in `Text.Pandoc.UTF8` that use `Data.Text`'s conversions.

  * Use `safeRead` instead of using `reads` directly (various modules).

  * "Implicit figures" (images alone in a paragraph) are now handled
    differently.  The markdown reader gives their titles the prefix `fig:`; the
    writers look for this before treating the image as a figure.  Though this
    is a bit of a hack, it has two advantages: (i) implicit figures can be
    limited to the markdown reader, and (ii) they can be deactivated by turning
    off the `implicit_figures` extension.

  * `catch` from `Control.Exception` is now used instead of the
    old Preface `catch`.

  * `Text.Pandoc.Shared`:  Improved algorithm for `normalizeSpaces`
    and `oneOfStrings` (which is now non-backtracking).

  * `Text.Pandoc.Biblio`: Remove workaround for `toCapital`.
    Now citeproc-hs is fixed upstream, so this is no longer needed.
    Closes #531.

  * Textile reader: Improved speed of `hyphenedWords`.
    This speeds up the textile  reader by about a factor of 4.

  * Use `Text.Pandoc.Builder` in RST reader, for more flexibility,
    better performance, and automatic normalization.

  * Major rewrite of markdown reader:

    + Use `Text.Pandoc.Builder` instead of lists.  This also
      means that everything is normalized automatically.
    + Move to a one-pass parsing strategy, returning values in the reader
      monad, which are then run (at the end of parsing) against the final
      parser state.

  * In HTML writer, we now use `toHtml` instead of pre-escaping.
    We work around the problem that blaze-html unnecessarily escapes `'`
    by pre-escaping just the `'` characters, instead of the whole string.
    If blaze-html later stops escaping `'` characters, we can simplify
    `strToHtml` to `toHtml`.  Closes #629.

  * Moved code for embedding images in RTFs from `pandoc.hs` to the
    RTF writer (which now exports `writeRTFWithEmbeddedImages`).

  * Moved citation processing from `pandoc.hs` into the readers.
    This makes things more convenient for library users.

  * The man pages are now built by an executable `make-pandoc-man-pages`,
    which has its own stanza in the cabal file so that dependencies can be
    handled by Cabal. Special treatment in `Setup.hs` ensures that this
    executable never gets installed; it is only used to create the man pages.

  * The cabal file has been modified so that the pandoc library is used
    in building the pandoc executable.  (This required moving `pandoc.hs`
    from `src` to `.`.)  This cuts compile time in half.

  * `-O2` is no longer used in building pandoc.  The performance improvement
    it yields is so slight that it is not worth it.  (Measured with
    benchmarks on ghc 7.4.)

  * The `executable` and `library` flags have been removed.

  * `-threaded` has been removed from ghc-options.

  * Version bounds of dependencies have been raised, and the
    `blaze_html_0_5` flag now defaults to True.  Pandoc now compiles on
    GHC 7.6.

  * We now require base >= 4.2.

  * Integrated the benchmark program into cabal.  One can now do:

        cabal configure --enable-benchmarks && cabal build
        cabal bench --benchmark-option='markdown' --benchmark-option='-s 20'

    The benchmark now uses README + testsuite, so benchmark results
    from older versions aren't comparable.

  * Integrated test suite with cabal.
    To run tests, configure with `--enable-tests`, then `cabal test`.
    You can specify particular tests using `--test-options='-t markdown'`.
    No output is shown unless tests fail.  The Haskell test modules
    have been moved from `src/` to `tests/`.

  * Moved all data files and templates to the `data/` subdirectory.

  * Added an `embed_data_files` cabal flag.  This causes all
    data files to be embedded in the binary, so that the binary
    is self-sufficient and can be relocated anywhere, copied on
    a USB key, etc.  The Windows installer now uses this.
    (Since we no longer have the option to build the executable
    without the library, this is the only way to get a relocatable
    binary on Windows.)

  * Removed pcre3.dll from windows package.
    It isn't needed unless highlighting-kate is compilled with the
    `pcre-light` flag. By default, regex-prce-builtin is used.


pandoc (1.9.4.2)

  * Don't encode/decode file paths if base >= 4.4.
    Prior to base 4.4, filepaths and command line arguments were treated
    as unencoded lists of bytes, not unicode strings, so we had to work
    around that by encoding and decoding them. This commit adds CPP
    checks for the base version that intelligibly enable encoding/decoding
    when needed. Fixes a bug with multilingual filenames when pandoc was
    compiled with ghc 7.4 (#540).

  * Don't generate an empty H1 after hrule slide breaks.
    We now use a slide-level header with contents `[Str "\0"]` to mark
    an hrule break.  This avoids creation of an empty H1 in these
    contexts.  Closes #484.

  * Docbook reader: Added support for "bold" emphasis.  Thanks to Mauro Bieg.

  * In make_osx_package.sh, ensure citeproc-hs is built with the
    embed_data_files flag.

  * MediaWiki writer: Avoid extra blank lines after sublists (Gavin Beatty).

  * ConTeXt writer: Don't escape `&`, `^`, `<`, `>`, `_`,
    simplified escapes for `}` and `{` to `\{` and `\}` (Aditya Mahajan).

  * Fixed handling of absolute URLs in CSS imports with `--self-contained`.
    Closes #535.

  * Added webm to mime types. Closes #543.

  * Added some missing exports and tests to the cabal file
    (Alexander V Vershilov).

  * Compile with `-rtsopts` and `-threaded` by default.

pandoc (1.9.4.1)

  * Markdown reader: Added `cf.` and `cp.` to list of likely abbreviations.

  * LaTeX template: Added `linkcolor`, `urlcolor` and `links-as-notes`
    variables.  Make TOC links black.

  * LaTeX template improvements.

    + Don't print date unless one is given explicitly in the document.
    + Simplified templates.
    + Use fontenc [T1] by default, and lmodern.
    + Use microtype if available.

  * Biblio:

    + Add comma to beginning of bare suffix, e.g. `@item1 [50]`.
      Motivation: `@item1 [50]` should be as close as possible to
      `[@item1, 50]`.
    + Added workaround for a bug in citeproc-hs 0.3.4 that causes footnotes
      beginning with a citation to be empty.  Closes #531.

  * Fixed documentation on mixed lists.  Closes #533.

pandoc (1.9.4)

  * Simplified `Text.Pandoc.Biblio` and fixed bugs with citations inside
    footnotes and captions.  We now handle note citations by inserting
    footnotes during initial citation processing, and doing a separate
    pass later to remove notes inside notes.

  * Added 'zenburn' highlight style from highlighting-kate.

  * Added Slideous writer. Slideous is an HTML + javascript slide show
    format, similar to Slidy, but works with IE 7. (Jonas Smedegaard)

  * LaTeX writer:

    + Ensure we don't have extra blank lines at ends of cells.
      This can cause LaTeX errors, as they are interpreted as new paragraphs.
    + More consistent interblock spacing.
    + Require highlighting-kate >= 0.5.1, for proper highlighted inline
      code in LaTeX.  Closes #527.
    + Ensure that a Verbatim at the end of a footnote is followed by
      a newline. (Fixes a regression in the previous version.)
    + In default template, use black for internal links and TOC.
      Added commented-out code to use footnotes for links, as would
      be suitable in print output.

  * Beamer writer:  When `--incremental` is used, lists inside
    a block quote should appear all at once.  (This makes Beamer
    output consistent with the HTML slide show formats.)

  * ConTeXt writer:

    + Escape `%` as `\letterpercent{}` not `\letterpercent `,
      to avoid gobbling spaces after the `%` sign.
    + Ensure space after `\stopformula`.

  * Markdown writer:

    + Use `:` form instead of `~` in definition lists, for better
      compatibility with other markdown implementations.
    + Don't wrap the term, because it breaks definition lists.
    + Use a nonzero space to prevent false recognition
      of list marker in ordered lists.  Closes #516.

  * Org writer: Add space before language name.  Closes #523.

  * Docx writer: Simplified bullet characters so they work properly
    with Word 2007. Closes #520.

  * LaTeX reader: Support `\centerline`.

  * RST reader:  handle figures.  Closes #522.

  * Textile reader: fix for `<notextile>` and `==`.  Closes #517.
    (Paul Rivier)

pandoc (1.9.3)

  * Fixed bug in `fromEntities`.  The previous version would turn
    `hi & low you know;` into `hi &`.

  * HTML reader:

    + Don't skip nonbreaking spaces.
      Previously a paragraph containing just `&nbsp;` would be rendered
      as an empty paragraph. Thanks to Paul Vorbach for pointing out the bug.
    + Support `<col>` and `<caption>` in tables. Closes #486.

  * Markdown reader:

    + Don't recognize references inside delimited code blocks.
    + Allow list items to begin with lists.

  * Added basic docbook reader (John MacFarlane and Mauro Bieg).

  * LaTeX reader:

    + Handle `\bgroup`, `\egroup`, `\begingroup`, `\endgroup`.
    + Control sequences can't be followed by a letter.
      This fixes a bug where `\begingroup` was parsed as `\begin`
      followed by `group`.
    + Parse 'dimension' arguments to unknown commands.  e.g. `\parindent0pt`
    + Make `\label` and `\ref` sensitive to `--parse-raw`.
      If `--parse-raw` is selected, these will be parsed as raw latex
      inlines, rather than bracketed text.
    + Don't crash on unknown block commands (like `\vspace{10pt}`)
      inside `\author`; just skip them.  Closes #505.

  * Textile reader:

    + Implemented literal escapes with `==` and `<notextile>`.  Closes #473.
    + Added support for LaTeX blocks and inlines (Paul Rivier).
    + Better conformance to RedCloth inline parsing (Paul Rivier).
    + Parse '+text+' as emphasized (should be underlined, but this
      is better than leaving literal plus characters in the output.

  * Docx writer: Fixed multi-paragraph list items.  Previously they each
    got a list marker.  Closes #457.

  * LaTeX writer:

    + Added `--no-tex-ligatures` option to avoid replacing
      quotation marks and dashes with TeX ligatures.
    + Use `fixltx2e` package to provide '\textsubscript'.
    + Improve spacing around LaTeX block environments:
      quote, verbatim, itemize, description, enumerate.
      Closes #502.
    + Use blue instead of pink for URL links in latex/pdf output.

  * ConTeXt writer: Fixed escaping of `%`.
    In text, `%` needs to be escaped as `\letterpercent`, not `\%`
    Inside URLs, `%` needs to be escaped as `\%`
    Thanks to jmarca and adityam for the fix.  Closes #492.

  * Texinfo writer:  Escape special characters in node titles.
    This fixes a problem pointed out by Joost Kremers.  Pandoc used
    to escape an '@' in a chapter title, but not in the corresponding
    node title, leading to invalid texinfo.

  * Fixed document encoding in texinfo template.
    Resolves Debian Bug #667816.

  * Markdown writer:

    + Don't force delimited code blocks to be flush left.
      Fixes bug with delimited code blocks inside lists etc.
    + Escape `<` and `$`.

  * LaTeX writer: Use `\hyperref[ident]{text}` for internal links.
    Previously we used `\href{\#ident}{text}`, which didn't work on
    all systems. Thanks to Dirk Laurie.

  * RST writer: Don't wrap link references.  Closes #487.

  * Updated to use latest versions of blaze-html, mtl.


pandoc (1.9.2)

  * LaTeX reader:

    + Made `lstlisting` work as a proper verbatim environment.
    + Fixed bug parsing LaTeX tables with one column.

  * LaTeX writer:

    + Use `{}` around `ctable` caption, so that formatting can be used.
    + Don't require eurosym package unless document has a €.

  * LaTeX template: Added variables for `geometry`, `romanfont`,
    `sansfont`, `mathfont`, `mainfont` so users can more easily
    customize fonts.

  * PDF writer:

    + Run latex engine at least two times, to ensure
      that PDFs will have hyperlinked bookmarks.
    + Added PDF metadata (title,author) in LaTeX standalone + PDF output.

  * Texinfo writer: retain directories in image paths.  (Peter Wang)

  * RST writer:  Better handling of inline formatting, in accord
    with docutils' "inline markup recognition rules" (though we don't
    implement the unicode rules fully). Now `hi*there*hi` gets
    rendered properly as `hi\ *there*\ hi`, and unnecessary
    `\ ` are avoided around `:math:`, `:sub:`, `:sup:`.

  * RST reader:

    + Parse `\ ` as null, not escaped space.
    + Allow `` :math:`...` `` even when not followed by blank
      or `\`.  This does not implement the complex rule docutils follows,
      but it should be good enough for most purposes.
    + Add support for the rST default-role directive. (Greg Maslov)

  * Text.Pandoc.Parsing: Added `stateRstDefaultRole` field to `ParserState`.
    (Greg Maslov)

  * Markdown reader: Properly handle citations nested in other inline
    elements.

  * Markdown writer:  don't replace empty alt in image with "image".

  * DZSlides:  Updated template.html and styles in default template.
    Removed bizarre CSS for `q` in dzslides template.

  * Avoid repeated `id` attribute in section and header in HTML slides.

  * README improvements:  new instructions on internal links,
    removed misleading note on reST math.

  * Build system:

    + Fixed Windows installer so that dzslides works.
    + Removed stripansi.sh.
    + Added .travis.yml for Travis continuous integration support..
    + Fixed upper bound for zlib (Sergei Trofimovich).
    + Fixed upper bound for test-framework.
    + Updated haddocks for haddock-2.10 (Sergei Trofimovich).

pandoc (1.9.1.2)

  * Added `beamer+lhs` as output format.

  * Don't escape `<` in `<style>` tags with `--self-contained`.
    This fixes a bug which prevented highlighting from working
    when using `--self-contained`.

  * PDF: run latex engine three times if `--toc` specified.
    This fixes page numbers in the table of contents.

  * Docx writer: Added TableNormal style to tables.

  * LaTeX math environment fixes. `aligned` is now used instead of
    the nonexistent `aligned*`. `multline` instead of the nonexistent
    `multiline`.

  * LaTeX writer: Use `\textasciitilde` for literal `~`.

  * HTML writer: Don't escape contents of EQ tags with --gladtex.
    This fixes a regression from 1.8.

  * Use `<q>` tags for Quoted items for HTML5 output.
    The quote style can be changed by modifying the template
    or including a css file. A default quote style is included.

  * LaTeX reader: Fixed accents (\~{a}, `\c{c}`).
    Correctly handle \^{}. Support "minted" as a LaTeX verbatim block.

  * Updated LaTeX template for better language support.
    Use `polyglossia` instead of `babel` with xetex.
    Set `lang` as documentclass option.
    `\setmainlanguage` will use the last of a comma-separated
    list of languages. Thanks to François Gannaz.

  * Fixed default LaTeX template so `\euro` and `€` work.  The
    `eurosym` package is needed if you are using pdflatex.

  * Fixed escaping of period in man writer (thanks to Michael Thompson).

  * Fixed list label positions in beamer.

  * Set `mainlang` variable in context writer.
    This parallels behavior of latex writer.  `mainlang` is the last
    of a comma-separated list of languages in lang.

  * EPUB language metadat: convert e.g. `en_US` from locale to `en-US`.

  * Changed `-V` so that you can specify a key without a value.
    Such keys get the value `true`.

  * Fixed permissions on installed man pages - thanks Magnus Therning.

  * Windows installer: require XP or higher.  The installer is
    now compiled on a Windows 7 machine, which fixes a problem
    using citation functions on Windows 7.

  * OSX package: Check for 64-bit Intel CPU before installing.

pandoc (1.9.1.1)

  * Better handling of raw latex environments in markdown.  Now

        \begin{equation}
        a_1
        \end{equation}

    turns into a raw latex block as expected.

  * Improvements to LaTeX reader:

    + Skip options after block commands.
    + Correctly handle `{\\}` in braced.
    + Added a needed 'try'.
    + Citations: add `, ` to suffix if it doesn't start with space or
      punctuation. Otherwise we get no space between the year and the
      suffix in author-date styles.

  * Added two needed data files for S5.  This fixes a problem with
    `pandoc -t s5 --self-contained`.  Also removed `slides.min.js`,
    which was no longer being used.

  * Fixed some minor problems in `reference.docx`:
    name on "Date" style, `xCs` instead of `xIs`.

  * Fixed a problem creating docx files using a reference docx
    modified using Word.  The problem seems to be that Word
    modifies `_rels/.rels`, changing the Type of the Relationship to
    `docProps/core.xml`. Pandoc now changes this back to the correct
    value if it has been altered, fixing the problem.

  * Fixed html5 template so it works properly with highlighting.

pandoc (1.9.1)

  * LaTeX reader:

    + Fixed regression in 1.9; properly handle escaped $ in latex math.
    + Put LaTeX verse environments in blockquotes.

  * Markdown reader:

    + Limit nesting of strong/emph.  This avoids exponential lookahead
      in parasitic cases, like `a**a*a**a*a**a*a**a*a**a*a**a*a**a*a**`.
    + Improved attributes syntax (inn code blocks/spans):
      (1) Attributes can contain line breaks. (2) Values in key-value
      attributes can be surrounded by either double or single quotes, or
      left unquoted if they contain no spaces.

  * Don't wrap headers in markdown or RST writers.

  * Added `stateMaxNestingLevel` to `ParserState`.
    We set this to 6, so you can still have `Emph` inside `Emph`,
    just not indefinitely.

  * More efficient implementation of `nowrap` in `Text.Pandoc.Pretty`.

  * `Text.Pandoc.PDF`:  Only run latex twice if `\tableofcontents`
    is present.

  * Require highlighting-kate >= 0.5.0.2, texmath >= 0.6.0.2.

pandoc (1.9.0.5)

  * Changed cabal file so that build-depends for the test program
    are not required unless the tests flag is used.

  * LaTeX writer:  insert `{}` between adjacent hyphens so they don't
    form ligatures (dashes) in code spans.

pandoc (1.9.0.4)

  * Raised version bound on test-framework to avoid problems
    compiling tests on GHC 7.4.1.

  * LaTeX reader: Use raw LaTeX as fallback inline text for Cites,
    so citations don't just disappear unless you process with
    citeproc.  Ignore `\bibliographystyle`, `\nocite`.

  * Simplified tex2pdf; it will always run latex twice to
    resolve table of contents and hyperrefs.

pandoc (1.9.0.3)

  * Require Cabal >= 1.10.
  * Tweaked cabal file to meet Cabal 1.10 requirements.

pandoc (1.9.0.2)

  * Allow build with json 0.4 or 0.5.  Otherwise we can't build with
    ghc 6.12.

pandoc (1.9)

  [new features]

  * Added a Microsoft Word `docx` writer. The writer includes support
    for highlighted code and for math (which is converted from TeX to OMML,
    Office's native math markup language, using texmath's new OMML module).
    A new option `--reference-docx` allows the user to customize the
    styles.

  * Added an `asciidoc` writer (<http://www.methods.co.nz/asciidoc/>).

  * Better support for slide shows:

    + Added a `dzslides` writer. DZSlides is a lightweight HTML5/javascript
      slide show format due to Paul Rouget (<http://paulrouget.com/dzslides/>).

    + Added a LaTeX `beamer` writer. Beamer is a LaTeX package for creating
      slide presentations.

    + New, flexible rules for dividing documents into sections and slides
      (see the "Structuring the slide show" in the User's Guide).  These
      are backward-compatible with the old rules, but they allow slide
      shows to be organized into sections and subsections containing
      multiple slides.

    + A new `--slide-level` option allows users to override defaults
      and select a slide level below the first header level with content.

  * A new `--self-contained` option produces HTML output that does not
    depend on an internet connection or the presence of any external
    files. Linked images, CSS, and javascript is downloaded (or fetched
    locally) and encoded in `data:` URIs. This is useful for making portable
    `HTML slide shows. The --offline` option has been deprecated and is now
    `treated as a synonym or --self-contained`.

  * Support for PDF output:

    + Removed the old `markdown2pdf`.
    + `pandoc` can now create PDFs (assuming you have latex and a set of
      appropriate packages installed): just specify an output file with the
      `.pdf` extension.
    + A new option `--latex-engine` allows you to specify `pdflatex`,
      `xelatex`, or `lualatex` as the processor.

  * Highlighting changes:

    + Syntax highlighting is now a standard feature; the `highlighting`
      flag is no longer needed when compiling.
    + A new `--no-highlight` option allows highlighting to be disabled.
    + Highlighting now works in `docx`, `latex`, and `epub`, as well as
      `html`, `html5`, `dzslides`, `s5`, and `slidy`.
    + A new `--highlight-style` option selects between various highlighting
      color themes.

  * Internal links to sections now work in ConTeXt and LaTeX as well as HTML.

  * LaTeX `\include` and `\usepackage` commands are now processed,
    provided the files are in the working directory.

  * EPUB improvements:

    + Internal and external links now work in EPUB.
    + Raw HTML is allowed.
    + New `--epub-embed-font` option.
    + Customizable templates for EPUB pages offer more control over
      formatting: `epub-page.html`, `epub-coverimage.html`,
      `epub-titlepage.html`.

  * `--mathml` now works with DocBook.

  * Added support for math in RST reader and writer.  Inline math uses the
    `` :math:`...` `` construct.  Display math uses

        .. math:: ...

    or if the math is multiline,

        .. math::

           ...

    These constructions are now supported now by `rst2latex.py`.

  * GitHub syntax for fenced code blocks is supported in pandoc's
    markdown.  You can now write

        ```ruby
        x = 2
        ```

    instead of

        ~~~ {.ruby}
        x = 2
        ~~~~

  * Easier scripting:  a new `toJsonFilter` function makes it easier to
    write Haskell scripts to manipulate the Pandoc AST.

  [behavior changes]

  * Fixed parsing of consecutive lists in markdown.
    Pandoc previously behaved like Markdown.pl for consecutive
    lists of different styles. Thus, the following would be parsed
    as a single ordered list, rather than an ordered list followed
    by an unordered list:

        1. one
        2. two

        - one
        - two

    This change makes pandoc behave more sensibly, parsing this as
    two lists.  Any change in list type (ordered/unordered) or in
    list number style will trigger a new list. Thus, the following
    will also be parsed as two lists:

        1. one
        2. two

        a. one
        b. two

    Since we regard this as a bug in Markdown.pl, and not something
    anyone would ever rely on, we do not preserve the old behavior
    even when `--strict` is selected.

  * Dashes work differently with `--smart`: `---` is always em-dash,
    and `--` is always en-dash.  Pandoc no longer tries to guess when
    `-` should be en-dash.  *Note:* This may change how existing documents
    look when processed with pandoc. A new option, `--old-dashes`,
    is provided for legacy documents.

  * The markdown writer now uses setext headers for levels 1-2.
    The old behavior (ATX headers for all levels) can be restored
    using the new `--atx-headers` option.

  * Links are now allowed in markdown image captions.  They are also
    allowed in links, but will appear there as regular text. So,

        [link with [link](/url)](/url)

    will turn into

        <p><a href="/url">link with link</a></p>

  * Improved handling of citations using `citeproc-hs-0.3.4`.
    Added `--citation-abbreviations` option.

  * Citation keys can no longer end with a punctuation character.
    This means that `@item1.` will be parsed as a citation with key
    'item1', followed by a period, instead of a citation with key
    'item1.', as was the case previously.

  * In HTML output, citations are now put in a span with class `citation`.

  * The markdown reader now recognizes DocBook block and inline tags.
    It was always possible to include raw DocBook tags in a markdown
    document, but now pandoc will be able to distinguish block from
    inline tags and behave accordingly. Thus, for example,

        <sidebar>
        hello
        </sidebar>

    will not be wrapped in `<para>` tags.

  * The LaTeX parser has been completely rewritten; it is now much more
    accurate, robust, and extensible. However, there are two important
    changes in how it treats unknown LaTeX. (1) Previously, unknown
    environments became BlockQuote elements; now, they are treated
    as "transparent", so `\begin{unknown}xyz\end{unknown}` is the
    same as `xyz`.  (2) Previously, arguments of unknown commands
    were passed through with their braces; now the braces are stripped
    off.

  * `--smart` is no longer selected automatically with `man` output.

  * The deprecated `--xetex` option has been removed.

  * The `--html5`/`-5` option has been deprecated. Use `-t html5`
    instead. `html5` and `html5+lhs` are now separate output formats.

  * Single quotes are no longer escaped in HTML output.  They do not
    need to be escaped outside of attributes.

  * Pandoc will no longer transform leading newlines in code
    blocks to `<br/>` tags.

  * The ODT writer now sizes images appropriately, using the image
    size and DPI information embedded in the image.

  * `--standalone` is once again implicitly for a non-text output format
    (ODT, EPUB).  You can again do `pandoc test.txt -o test.odt`
    and get a standalone ODT file.

  * The Docbook writer now uses `<sect1>`, `<sect2>`, etc. instead of
    `<section>`.

  * The HTML writer now uses `<del>` for strikeout.

  * In HTML output with `--section-divs`, the classes `section` and
    `level[1,2,..6]` are put on the `div` tags so they can be styled.
    In HTML 5 output with `--section-divs`, the classes
    `level[1,2,...6]` are put on `section` tags.

  * EPUB writer changes:

    + The `lang` variable now sets the language
      in the metadata (if it is not set, we default to the locale).
    + EPUB:  UTF-8 is used rather than decimal entities.

  * Added `titleslide` class to title slide in S5 template.

  * In HTML, EPUB, and docx metadata, the date is normalized into
    YYYY-MM-DD format if possible. (This is required for validation.)

  * Attributes in highlighted code blocks are now preserved in HTML.
    The container element will have the classes, id, and key-value attributes
    you specified in the delimited code block. Previously these were stripped
    off.

  * The reference backlink in the HTML writer no longer has a special
    `footnoteBacklink` class.

  * The HTML template has been split into `html` and `html5` templates.

  * Author and date are treated more consistently in HTML templates.
    Authors are now `<h2>`, date `<h3>`.

  * URLs are hyphenated in the ConTeXt writer (B. Scott Michel).

  * In `Text.Pandoc.Builder`, `+++` has been replaced by `<>`.

  [bug fixes]

  * Better support for combining characters and East Asian wide characters
    in markdown and reST.

  * Better handling of single quotes with `--smart`.
    Previously `D'oh l'*aide*` would be parsed with left and right single
    quotes instead of apostrophes. This kind of error is now fixed.

  * Highlighting: Use `reads` instead of `read` for better error handling.
    Fixes crash on `startNum="abc"`.

  * Added blank comment after directives in rst template.

  * Unescape entities in citation `refId`. The `refId`s coming
    from citeproc contain XML numeric entities, and these don't match with the
    citation keys parsed by pandoc. Solution is to unescape them.

  * HTML reader: Fixed bug parsing tables with both thead and tbody.

  * Markdown reader:

    + Better handling of escapes in link URLs and titles.
    + Fixed backslash escapes in reference links.
    + Fixed bug in table/hrule parsing, by checking that the top
      line of a table is not followed by a blank line. This bug caused
      slowdowns on some files with hrules and tables, as pandoc tried to
      interpret the hrules as the tops of multiline tables.
    + Fixed bug in code block attribute parser. Previously the ID attribute
      got lost if it didn't come first. Now attributes can come in any order.

  * RST reader: allow footnotes followed by newline without space characters.

  * LaTeX reader:

    + Ignore empty groups {}, { }.
    + LaTeX reader: Handle \@.
    + LaTeX reader:  Don't crash on commands like `\itemsep`.
    + LaTeX reader:  Better handling of letter environments.

  * RST writer: Fixed bug involving empty table cells. isSimple was being
    calculated in a way that assumed there were no non-empty cells.

  * ConTeXt writer:

    + Made `--toc` work even without `--number-sections`.
    + Escape # in link URLs.
    + Use buffering for footnotes containing code blocks.
    + Changed 'descr' to 'description', fixed alignment.

  * LaTeX writer:

    + Escape euro character.
    + Don't escape ~ inside href{...}.
    + Escape # in href URLs.
    + Improved detection of book classes.  We now check the
      `documentclass` variable, and if that is not set, we look through
      the template itself.  Also, we have added the KOMA classes scrreprt
      and scrbook.  You can now make a book using
      `pandoc -V documentclass:book mybook.txt -o mybook.pdf`
    + LHS files now set the "listings" variable, so that the definition
      of the `code` environment will be included in the template.
    + Links are colored blue by default (this can be changed by
      modifying `hyperref` settings in the template).
    + Added `lang` variable to LaTeX template.

  * HTML writer:

    + Fixed bug in HTML template with html5 and mathml.
    + Don't use self-closing img, br, hr tags for HTML5.
    + Use `<section>` for footnotes if HTML5.
    + Update HTML templates to use Content-Style-Type meta tag.
    + Use separate variables for meta-date, meta-author.
      This makes footnotes work in author and date fields.
    + Use 'vertical-align:middle' in WebTeX math for better alignment.

  * S5/slidy writer:  Make footnotes appear on separate slide at end.

  * MIME: Added 'layout-cache' to getMimeType. This ensures that the
    META-INF/manifest.xml for ODT files will have everything it needs, so
    that ODT files modified by LibreOffice can be used as `--reference-odt`.

  * `Text.Pandoc.Templates`: Return empty string for json template.

  * `Text.Pandoc.Biblio`:

    + Expand citations recursively inside nested inlines.
    + Treat `\160` as space when parsing locator and suffix.
      This fixes a bug with "p. 33" when `--smart` is used. Previously
      the whole "p. 33" would be included in the suffix, with no locator.
    + Put whole author-in-text citation in a Cite.  Previously just the
      date and other info went in the Cite.
    + Don't add comma+space to prefix if it ends in punctuation.

  * Updated chicago-author-date.csl.  The old version did not work
    properly for edited volumes with no author.

  * EPUB writer:

    + Add date to EPUB titlepage and metadata.
    + Added TOC identifier in EPUB page template.
    + Don't generate superfluous file `cover-image.jpg`.

  [under the hood improvements]

  * Modified `make_osx_package.sh` to use cabal-dev.
    Items are no longer installed as root.
    Man pages are zipped and given proper permissions.

  * Modified windows installer generater to use cabal-dev.

  * Setup: Making man pages now works with cabal-dev (at least on OSX). In
    Setup.hs we now invoke 'runghc' in a way that points it to the correct
    package databases, instead of always falling back to the default user
    package db.

  * Updated to work with GHC 7.4.1.

  * Removed dependency on old-time.

  * Removed dependency on dlist.

  * New slidy directory for "self-contained."

  * TeXMath writer:  Use unicode thin spaces for thin spaces.

  * Markdown citations: don't strip off initial space in locator.

  [API changes]

  * Removed `Apostrophe`, `EmDash`, `EnDash`, and `Ellipses`
    from the native `Inline` type in pandoc-types.  Now we use `Str`
    elements with unicode.

  * Improvements to `Text.Pandoc.Builder`:

    + `Inlines` and `Blocks` are now newtypes (not synonyms for
      sequences).
    + Instances are defined for `IsString`, `Show`, `Read`, `Monoid`,
      and a new `Listable` class, which allows these to be manipulated
      to some extent like lists. Monoid append includes automatic
      normalization.
    + `+++` has been replaced by `<>` (mappend).

  * Use blaze-html instead of xhtml for HTML generation.
    This changes the type of `writeHtml`.

  * `Text.Pandoc.Shared`:

    + Added `warn` and `err`.
    + Removed `unescapeURI`, modified `escapeURI`.
      (See under [behavior changes], above.)

  * Changes in URI escaping:  Previously the readers escaped URIs by
    converting unicode characters to octets and then percent encoding.
    Now unicode characters are left as they are, and `escapeURI` only
    percent-encodes space characters.  This gives more readable
    URIs, and works well with modern user agents. URIs are no longer unescaped
    at all on conversion to `markdown`, `asciidoc`, `rst`, `org`.

  * New module `Text.Pandoc.SelfContained`.

  * New module `Text.Pandoc.Docx`.

  * New module `Text.Pandoc.PDF`.

  * Added `writerBeamer` to `WriterOptions`.

  * Added `normalizeDate` to `Text.Pandoc.Shared`.

  * Added `splitStringWithIndices` in `Text.Pandoc.Shared`.
    This is like `splitWithIndices`, but it is sensitive to distinctions
    between wide, combining, and regular characters.

  * `Text.Pandoc.Pretty`:

    + Added `chomp` combinator.
    + Added `beforeNonBreak` combinator.  This allows you to include
      something conditionally on it being before a nonblank.
      Used for RST inline math.
    + Added `charWidth` function. All characters marked W or F in the unicode
      spec EastAsianWidth.txt get width 2.
    + Added `realLength`, based on `charWidth`. `realLength` is now
      used in calculating offsets.

  * New module `Text.Pandoc.Slides`, for common functions for breaking
    a document into slides.

  * Removed `Text.Pandoc.S5`, which is no longer needed.

  * Removed `Text.Pandoc.CharacterReferences`.  Moved
    `characterReference` to `Text.Pandoc.Parsing`.
    `decodeCharacterReferences` is replaced by `fromEntities`
    in `Text.Pandoc.XML`.

  * Added `Text.Pandoc.ImageSize`.  This is intened for use
    in `docx` and `odt` writers, so the size and dpi of images
    can be calculated.

  * Removed `writerAscii` in `WriterOptions`.

  * Added `writerHighlight` to `WriterOptions`.

  * Added `DZSlides` to `HTMLSlideVariant`.

  * `writeEPUB` has a new argument for font files to embed.

  * Added `stateLastStrPos` to `ParserState`. This lets us keep track
    of whether we're parsing the position immediately after a regular
    (non-space, non-symbol) string, which is useful for distinguishing
    apostrophes from single quote starts.

  * `Text.Pandoc.Parsing`:

    + `escaped` now returns a `Char`.
    + Removed `charsInBalanced'`, added a character parser as
      a parameter of `charsInBalanced`.  This is needed for
      proper handling of escapes, etc.
    + Added `withRaw`.

  * Added `toEntities` to `Text.Pandoc.XML`.

  * `Text.Pandoc.Readers.LaTeX`:

    + Export `handleIncludes`.
    + Export `rawLaTeXBlock` instead of `rawLaTeXEnvironment'`.

  * Added `ToJsonFilter` class and `toJsonFilter` function to
    `Text.Pandoc`, deprecating the old `jsonFilter` function.

  * `Text.Pandoc.Highlighting`:

     + Removed `highlightHtml`, `defaultHighlightingCss`.
     + Export `formatLaTeXInline`, `formatLaTeXBlock`, and `highlight`, plus
       key functions from highlighting-kate.
     + Changed types of highlighting function.  `highlight` returns a
       `Maybe`, not an `Either`.

pandoc (1.8.2.1)

  * Relaxed cabal consntraints for test-framework (S. Trofimovich).

  * Relaxed cabal constraints for pandoc-types.

  * Adjusted Arbitrary instance to help avoid timeouts in tests.

  * Added `Tests.Writers.Markdown` to cabal file.

pandoc (1.8.2)

  * Added script to produce OS X package.

  * Made `templates` directory a git submodule.  This should make it
    easier for people to revise their custom templates when the default
    templates change.

  * Changed template naming scheme: `FORMAT.template` -> `default.FORMAT`.
    **Note:** If you have existing templates in `~/.pandoc/templates`, you
    must rename them to conform to the new scheme!

  * Fixed smart quotes bug, now handling `'...hi'` properly.

  * RST reader:

    + Partial support for labeled footnotes.
    + Improved accuracy of `simpleReferenceName` parser.

  * HTML reader:

    + Substitute correct unicode characters for
      characters in the 128..159 range, which are often found even in
      HTML that purports to be UTF-8.

  * LaTeX reader:  Handle `\subtitle` command (a subtitle is added
    to the title, after a colon and linebreak). Closes #280.

  * Leaner `reference.odt`.

  * Added unexported module `Text.Pandoc.MIME` for use in
    the ODT writer.

  * ODT writer:  Construct `manifest.xml` based on archive contents.
    This fixes a bug in ODTs containing images. Recent versions of
    LibreOffice would reject these as corrupt, because `manifest.xml`
    did not contain a reference to the image files.

  * LaTeX writer:

    + Make verbatim environments flush to avoid spurious
      blank lines.  Closes #277.
    + Use `\texttt` and escapes insntead of `\verb!...!`, which
      is too fragile (doesn't work in command arguments).
    + Use `\enquote{}` for quotes if the template includes
      the `csquotes` package. This provides better support for
      local quoting styles. (Thanks to Andreas Wagner for the idea.)

  * ConTeXt writer:  Make `\starttyping`/`\stoptyping` flush with
    margin, preventing spurious blank lines.

  * Slidy writer:

    + Use non-minimized version of `slidy.css` with `--offline`
      option, so users can more easily edit it.
    + Also fixed a bug in the CSS that prevented proper centering
      of title (now reported and fixed upstream).

  * S5 writer:

    + Replaced `s5/default/slides.js.{comment,packed}` with
      new compressed `s5/default/slides.min.js`.
    + Use `data:` protocol to embed S5 CSS in `<link>` tags,
      when `--offline` is specified. Using inline CSS didn't
      work with Chrome or Safari.  This fixes offline
      S5 on those browsers.

  * HTML writer:  Removed English title on footnote backlinks.
    This is incongrous in non-English documents.

  * Docbook writer:

    + Use CALS tables.  (Some older docbook software does not work
      well with XHTML tables.)  Closes #77.
    + Use `programlisting` tags (instead of `screen`) for code blocks.

  * `markdown2pdf`:

    + Calls latex with `-halt-on-error -interaction nonstopmode` instead
      of `-interaction=batchmode`, which essentially just ignored errors,
      leading to bad results. Better to know when something is wrong.
    + Fixed issues with non-UTF-8 output of `pdflatex`.
    + Better error reporting.

  * `--mathjax` now takes an optional URL argument. If it is not
    provided, pandoc links directly to the (secure) mathjax CDN,
    as now recommended (thanks to dsanson).

  * Deprecated `--xetex` option in `pandoc`.  It is no longer needed,
    since the LaTeX writer now produces a file that can be processed by
    `latex`, `pdflatex`, `lualatex`, or `xelatex`.

  * Introduced `--luatex` option to `markdown2pdf`. This causes `lualatex`
    to be used to create the PDF.

  * If a template specified with `--template` is not found, look for it
    in `datadir`.  Also, if no extension is provided, supply one based
    on the writer.  So now you can put your `special.latex` template in
    `~/.pandoc/templates`, and use it from any directory via
    `pandoc -t latex --template special`.

  * Default template improvements:

    + HTML:  Display author and date after title.
    + HTML:  Made table of contents more customizable.  The container
      for the TOC is now in the template, so users can insert a header
      or other styling. (Thanks to Bruce D'Arcus for the suggestion.)
    + HTML, Slidy, S5:  Enclose scripts in CDATA tags.
    + Slidy, S5: Added `s5-url` and `slidy-url` variables, instead of
      hard-coding.  If you want to put your slidy files in the slidy
      subdirectory, for example, you can do
      `pandoc -t slidy -V slidy-url=slidy -s`.
    + LaTeX: Use `\and` to separate authors in LaTeX documents (reader
      & writer).  Closes #279.
    + LaTeX: Use different `hyperref` options for `xetex`, fixing
      problems with unicode bookmarks (thanks to CircleCode).
    + LaTeX: Removed `ucs` package, use `utf8` rather than `utf8x`
      with `inputenc`.  This covers fewer characters but is more
      robust with other packages, and `ucs` is unmaintained.  Users
      who need better unicode support should use xelatex or lualatex.

pandoc (1.8.1.2)

  * Added `--epub-cover-image` option.

  * Documented `--biblatex` and `--natbib` options.

  * Allow `--section-divs` with slidy output.  Resolves Issue #296.

  * Disallow notes within notes in reST and markdown.
    These previously caused infinite looping and stack overflows.
    For example:

        [^1]

        [^1]: See [^1]

    Note references are allowed in reST notes, so this isn't a full
    implementation of reST. That can come later. For now we need to
    prevent the stack overflows.  Partially resolves Issue #297.

  * EPUB writer: Allow non-plain math methods.

  * Forbid ()s in citation item keys.  Resolves Issue #304: problems with
    `(@item1; @item2)` because the final paren was being parsed as part of
    the item key.

  * Changed URI parser so it doesn't include trailing punctuation.
    So, in RST, `http://google.com.` should be parsed as a link followed by a
    period. The parser is smart enough to recognize balanced parentheses, as
    often occur in wikipedia links: `http://foo.bar/baz_(bam)`.

  * Markdown+lhs reader: Require space after inverse bird tracks, so that
    HTML tags can be used freely at the left margin of a markdown+lhs document.
    Thanks to Conal Elliot for the suggestion.

  * Markdown reader:

    + Improved emph/strong parsing; fixes bug found by Perry Wagle.
    + Fixed bug in footnote order (reported by CircleCode).

  * RST reader:
      + Fixed bug in in field lists with multi-line items at the
        end of the list.
      + Added parentheses to RST `specialChars`, so
        `(http://google.com)` will be parsed as a link in parens.
        Resolves Issue #291.
      + Allow `|` followed by newline in RST line block.

  * LaTeX reader:
      + Support `\dots`.
      + Gobble option & space after linebreak `\\[10pt]`.

  * Textile reader:
      + Make it possible to have colons after links.  (qerub)
      + Make it possible to have colons after links.  (Christoffer Sawicki)

  * HTML reader:
      + Skip spaces after `<b>`, `<emph>`, etc.
      + Handle tbody, thead in simple tables.  Closes #274.
      + Implicit `Para`s instead of `Plains` in some contexts.

  * OpenDocument writer:  Use special `First paragraph` style for
    first paragraph after most non-paragraph blocks. This allows users to
    specify e.g. that only paragraphs after the first paragraph of a block are
    to be indented. Thanks to Andrea Rossato for the patch. Closes #20.

  * LaTeX writer:  use `deVerb` on table and picture captions.
    Otherwise LaTeX complains about `\verb` inside command argument.
    Thanks to bbanier for reporting the bug.

  * Markdown writer: Insert HTML comment btw list and indented code block.
    This prevents the code block from being interpreted as part of the list.

  * EPUB writer: Add a meta element specify the cover.
    Some EPUB e-readers, such as the Nook, require a meta element inside the
    OPF metadata block to ensure the cover image is properly displayed.
    (Kelsey Hightower)

  * HTML writer: Use embed tag for images with non-image extensions.
    (e.g.  PDFs).  Closes #264.

  * LaTeX writer: Improved tables.

      + More space between lines, top-align cells.
      + Use ctable package, which allows footnotes and
        provides additional options.
      + Made cell alignments work in multiline tables.
      + Closes #271, #272.

  * Un-URI-escape image filenames in LaTeX, ConTeXt, RTF, Texinfo.
    Also do this when copying image files into EPUBs and ODTs.
    Closes #263.

  * Changed to github issue tracker.

  * Added failing emph/strong markdown test case due to Perry Wagle.

  * Slidy improvements:
      + Updated to use Slidy2.
      + Fixed bug, unclosed div tag.
      + Added `duration` variable in template.
        Setting this activates the timer.
      + Use 'titlepage' instead of 'cover' for title div.

pandoc (1.8.1.1)

  * `markdown2pdf`:  Removed some debugging lines accidentally included
    in the 1.8.1 release. With those lines, the temp directory is created
    in the working directory, and it is not deleted.  This fix restores
    the original behavior.

pandoc (1.8.1)

  * Added `--ascii` option.  Currently supported only in HTML writer,
    which it causes to use numerical entities instead of UTF-8.

  * EPUB writer: `--toc` now works to provide a table of contents
    at the beginning of each chapter.

  * LaTeX writer:  Change figure defaults to `htbp`.
    This prevents "too many unprocessed floats."  Resolves
    Issue #285.

  * `Text.Pandoc.UTF8`:  Encode filenames even when using recent
    base.

  * `markdown2pdf`: Fixed filename encoding issues. With help from Paulo
    Tanimoto. Resolves Issue #286.

  * HTML writer: Put line breaks in section divs.

  * `Text.Pandoc.Shared`: Make `writerSectionDivs` default to False.

pandoc (1.8.0.3)

  * Fixed Source-repository stanza in cabal file.

pandoc (1.8.0.2)

  * HTML writer:

    + Stringify alt text instead of converting to HTML.
    + Break lines after block elements, not inside tags.
      HTML output now closely resembles that of tidy. Resolves Issue #134.

  * Markdown reader: Fixed bug in footnote block parser (pointed out
    by Jesse Rosenthal).  The problem arose when the blank line
    at the end of a footnote block contained indenting spaces.

  * Shared: Improved 'normalize' function so it normalizes Spaces too.
    In normal form, Space elements only occur to separate two non-Space
    elements.  So, we never have [Space], or [, ..., Space].

  * Tests:

    + Improved Arbitrary instance.
    + Added timeout for test instances.

  * README:

    + Added section on four-space rule for lists.  Resolves Issue #283.
    + Clarified optional arguments on math options.

  * markdown2pdf: Fixed bug with output file extensions.
    Previously `markdown2pdf test.txt -o test.en.pdf` would produce
    `test.pdf`, not `test.en.pdf`.  Thanks to Paolo Tanimoto for the fix.

pandoc (1.8.0.1)

  * Revised Interact.hs so that it works with the CPP macros
    in the UTF8 module.

  * Revised Setup.hs so that we don't call MakeManPage.hs unless
    the man pages are out of date.

pandoc (1.8)

  [new features]

  * Support for citations using Andrea Rossato's `citeproc-hs` 0.3.
    You can now write, for example,

        Water is wet [see @doe99, pp. 33-35; also @smith04, ch. 1].

    and, when you process your document using `pandoc`, specifying
    a citation style using `--csl` and a bibliography using `--bibliography`,
    the citation will be replaced by an appropriately formatted
    citation, and a list of works cited will be added to the end
    of the document.

    This means that you can switch effortlessly between different citation
    and bibliography styles, including footnote, numerical, and author-date
    formats. The bibliography can be in any of the following formats: MODS,
    BibTeX, BibLaTeX, RIS, EndNote, EndNote XML, ISI, MEDLINE, Copac, or JSON.
    See the README for further details.

    Citations are supported in the markdown reader, using a special
    syntax, and in the LaTeX reader, using natbib or biblatex syntax.
    (Thanks to Nathan Gass for the natbib and biblatex support.)

  * New `textile` reader and writer.  Thanks to Paul Rivier for contributing
    the `textile` reader, an almost complete implementation of the textile
    syntax used by the ruby [RedCloth library](http://redcloth.org/textile).
    Resolves Issue #51.

  * New `org` writer, for Emacs Org-mode, contributed by Puneeth Chaganti.

  * New `json` reader and writer, for reading and writing a JSON
    representation of the native Pandoc AST.  These are much faster
    than the `native` reader and writer, and should be used for
    serializing Pandoc to text.  To convert between the JSON representation
    and native Pandoc, use `encodeJSON` and `decodeJSON` from
    `Text.JSON.Generic`.

  * A new `jsonFilter` function in `Text.Pandoc` makes it easy
    to write scripts that transform a JSON-encoded pandoc document.
    For example:

        -- removelinks.hs - removes links from document
        import Text.Pandoc
        main = interact $ jsonFilter $ bottomUp removeLink
                 where removeLink (Link xs _) = Emph xs
                       removeLink x = x

    To use this to remove links while translating markdown to LaTeX:

        pandoc -t json | runghc removelinks.hs | pandoc -f json -t latex

  * Attributes are now allowed in inline `Code` elements, for example:

        In this code, `ulist ! [theclass "special"] << elts`{.haskell} is...

    The attribute syntax is the same as for delimited code blocks.
    `Code` inline has an extra argument place for attributes, just like
    `CodeBlock`. Inline code will be highlighted in HTML output, if pandoc
    is compiled with highlighting support. Resolves Issue #119.

  * New `RawBlock` and `RawInline` elements (replacing `RawHtml`,
    `HtmlInline`, and `TeX`) provide lots of flexibility in writing
    scripts to transform Pandoc documents. Scripts can now change
    how each element is rendered in each output format.

  * You can now define LaTeX macros in markdown documents, and pandoc
    will apply them to TeX math.  For example,

        \newcommand{\plus}[2]{#1 + #2}
        $\plus{3}{4}$

    yields `3+4`.  Since the macros are applied in the reader, they
    will work in every output format, not just LaTeX.

  * LaTeX macros can also be used in LaTeX documents (both in math
    and in non-math contexts).

  * A new `--mathjax` option has been added for displaying
    math in HTML using MathJax.  Resolves issue #259.

  * Footnotes are now supported in the RST reader. (Note, however,
    that unlike docutils, pandoc ignores the numeral or symbol used in
    the note; footnotes are put in an auto-numbered ordered list.)
    Resolves Issue #258.

  * A new `--normalize` option causes pandoc to normalize the AST
    before writing the document.  This means that, for example,
    `*hi**there*` will be rendered as `<em>hithere</em>`
    instead of `<em>hi</em><em>there</em>`.  This is not the default,
    because there is a significant performance penalty.

  * A new `--chapters` command-line option causes headers
    in DocBook, LaTeX, and ConTeXt to start with "chapter" (level one).
    Resolves Issue #265.

  * In DocBook output, `<chapter>` is now used for top-level
    headers if the template contains `<book>`. Resolves Issue #265.

  * A new `--listings` option in `pandoc` and `markdown2pdf` causes
    the LaTeX writer to use the listings package for code blocks.
    (Thanks to Josef Svennigsson for the pandoc patch, and Etienne
    Millon for the markdown2pdf patch.)

  * `markdown2pdf` now supports `--data-dir`.

  * URLs in autolinks now have class "url" so they can be styled.

  * Improved prettyprinting in most formats.  Lines will be wrapped
    more evenly and duplicate blank lines avoided.

  * New `--columns` command-line option sets the column width for
    line wrapping and relative width calculations for tables.

  * Made `--smart` work in HTML, RST, and Textile readers, as well
    as markdown.

  * Added `--html5` option for HTML5 output.

  * Added support for listings package in LaTeX reader
    (Puneeth Chaganti).

  * Added support for simple tables in the LaTeX reader.

  * Added support for simple tables in the HTML reader.

  * Significant performance improvements in many readers and writers.

  [API and program changes]

  * Moved `Text.Pandoc.Definition` from the `pandoc` package to a new
    auxiliary package, `pandoc-types`. This will make it possible for other
    programs to supply output in Pandoc format, without depending on the whole
    pandoc package.

  * Added `Attr` field to `Code`.

  * Removed `RawHtml`, `HtmlInline`, and `TeX` elements; added generic
    `RawBlock` and `RawInline`.

  * Moved generic functions to `Text.Pandoc.Generic`. Deprecated
    `processWith`, replacing it with two functions, `bottomUp` and `topDown`.
    Removed previously deprecated functions `processPandoc` and `queryPandoc`.

  * Added `Text.Pandoc.Builder`, for building `Pandoc` structures.

  * `Text.Pandoc` now exports association lists `readers` and `writers`.

  * Added `Text.Pandoc.Readers.Native`, which exports `readNative`.
    `readNative` can now read full pandoc documents, block lists, blocks,
    inline lists, or inlines.  It will interpret `Str "hi"`
    as if it were `Pandoc (Meta [] [] []) [Plain [Str "hi"]]`.
    This should make testing easier.

  * Removed deprecated `-C/--custom-header` option.
    Use `--template` instead.

  * `--biblio-file` has been replaced by `--bibliography`.
    `--biblio-format` has been removed; pandoc now guesses the format
    from the file extension (see README).

  * pandoc will treat an argument as a URI only if it has an
    `http(s)` scheme.  Previously pandoc would treat some
    Windows pathnames beginning with `C:/` as URIs.

  * The `--sanitize-html` option and the `stateSanitize` field in
    `ParserState` have been removed. Sanitization is better done in the
    resulting HTML using `xss-sanitize`, which is based on pandoc's
    sanitization, but improved.

  * pandoc now adds a newline to the end of its output in fragment
    mode (= not `--standalone`).

  * Added support for `lang` in `html` tag in the HTML template,
    so you can do `pandoc -s -V lang=es`, for example.

  * `highlightHtml` in `Text.Pandoc.Highlighting` now takes
    a boolean argument that selects between "inline" and
    "block" HTML.

  * `Text.Pandoc.Writers.RTF` now exports `rtfEmbedImage`.
    Images are embedded in RTF output when possible (png, jpeg).
    Resolves Issue #275.

  * Added `Text.Pandoc.Pretty`. This is better suited for pandoc than the
    `pretty` package.  Changed all writers that used
    `Text.PrettyPrint.HughesPJ` to use `Text.Pandoc.Pretty` instead.

  * Rewrote `writeNative` using the new prettyprinting module. It is
    now much faster. The output has been made more consistent and compressed.
    `writeNative` is also now sensitive to writerStandalone`, and will simply
    `print a block list if writerStandalone` is False.

  * Removed `Text.Pandoc.Blocks`. `Text.Pandoc.Pretty` allows you to define
    blocks and concatenate them, so a separate module is no longer needed.

  * `Text.Pandoc.Shared`:

    + Added `writerColumns`, `writerChapters`, and `writerHtml5` to
      `WriterOptions`.
    + Added `normalize`.
    + Removed unneeded prettyprinting functions:
      `wrapped`, `wrapIfNeeded`, `wrappedTeX`, `wrapTeXIfNeeded`, `hang'`,
      `BlockWrapper`, `wrappedBlocksToDoc`.
    + Made `splitBy` take a test instead of an element.
    + Added `findDataFile`, refactored `readDataFile`.
    + Added `stringify`. Rewrote `inlineListToIdentifier` using `stringify`.
    + Fixed `inlineListToIdentifier` to treat '\160' as ' '.

  * `Text.Pandoc.Readers.HTML`:

    + Removed `rawHtmlBlock`, `anyHtmlBlockTag`, `anyHtmlInlineTag`,
      `anyHtmlTag`, `anyHtmlEndTag`, `htmlEndTag`, `extractTagType`,
      `htmlBlockElement`, `htmlComment`
    + Added `htmlTag`, `htmlInBalanced`, `isInlineTag`, `isBlockTag`,
      `isTextTag`

  * Moved `smartPunctuation` from `Text.Pandoc.Readers.Markdown`
    to `Text.Pandoc.Readers.Parsing`, and parameterized it with
    an inline parser.

  * Added `nonspaceChar` to `Text.Pandoc.Parsing`.

  * Ellipses are no longer allowed to contain spaces.
    Previously we allowed '. . .', ' . . . ', etc.  This caused
    too many complications, and removed author's flexibility in
    combining ellipses with spaces and periods.

  * Allow linebreaks in URLs (treat as spaces). Also, a string of
    consecutive spaces or tabs is now parsed as a single space. If you have
    multiple spaces in your URL, use `%20%20`.

  * `Text.Pandoc.Parsing`:

    + Removed `refsMatch`.
    + Hid `Key` constructor.
    + Removed custom `Ord` and `Eq` instances for `Key`.
    + Added `toKey` and `fromKey` to convert between `Key` and `[Inline]`.
    + Generalized type on `readWith`.

  * Small change in calculation of relative widths of table columns.
    If the size of the header > the specified column width, use
    the header size as 100% for purposes of calculating
    relative widths of columns.

  * Markdown writer now uses some pandoc-specific features when `--strict`
    is not specified: `\` newline is used for a hard linebreak instead of
    two spaces then a newline. And delimited code blocks are used when
    there are attributes.

  * HTML writer:  improved gladTeX output by setting ENV appropriately
    for display or inline math (Jonathan Daugherty).

  * LaTeX writer: Use `\paragraph`, `\subparagraph` for level 4,5 headers.

  * LaTeX reader:

    + `\label{foo}` and `\ref{foo}` now become `{foo}` instead of `(foo)`.
    + `\index{}` commands are skipped.

  * Added `fontsize` variable to default LaTeX template.
    This makes it easy to set the font size using `markdown2pdf`:
    `markdown2pdf -V fontsize=12pt input.txt`.

  * Fixed problem with strikeout in LaTeX headers when using
    hyperref, by adding a command to the default LaTeX template
    that disables `\sout` inside pdf strings. Thanks to Joost Kremers
    for the fix.

  * The `COLUMNS` environment variable no longer has any effect.

  [under-the-hood improvements]

  * Pandoc now compiles with GHC 7. (This alone leads to a
    significant performance improvement, 15-20%.)

  * Completely rewrote HTML reader using tagsoup as a lexer. The
    new reader is faster and more accurate.  Unlike the
    old reader, it does not get bogged down on some input
    (Issues #277, 255). And it handles namespaces in tags
    (Issue #274).

  * Replaced `escapeStringAsXML` with a faster version.

  * Rewrote `spaceChar` and some other parsers in Text.Pandoc.Parsing
    for a significant performance boost.

  * Improved performance of all readers by rewriting parsers.

  * Simplified Text.Pandoc.CharacterReferences by using
    entity lookup functions from TagSoup.

  * `Text.Pandoc.UTF8` now uses the unicode-aware IO functions
    from `System.IO` if base >= 4.2.  This gives support for
    windows line endings on windows.

  * Remove duplications in documentation by generating the
    pandoc man page from README, using `MakeManPage.hs`.

  * README now includes a full description of markdown syntax,
    including non-pandoc-specific parts.  A new `pandoc_markdown`
    man page is extracted from this, so you can look up markdown
    syntax by doing `man pandoc_markdown`.

  * Completely revised test framework (with help from Nathan Gass).
    The new test framework is built when the `tests` Cabal flag is set. It
    includes the old integration tests, but also some new unit and quickcheck
    tests. Test output has been much improved, and you can now specify a glob
    pattern after `cabal test` to indicate which tests should be run;
    for example `cabal test citations` will run all the citation tests.

  * Added a shell script, `stripansi.sh`, for filtering ANSI control
    sequences from test output: `cabal test | ./stripansi.sh > test.log`.

  * Added `Interact.hs` to make it easier to use ghci while developing.
    `Interact.hs` loads `ghci` from the `src` directory, specifying
    all the options needed to load pandoc modules (including
    specific package dependencies, which it gets by parsing
    dist/setup-config).

  * Added `Benchmark.hs`, testing all readers + writers using criterion.

  * Added `stats.sh`, to make it easier to collect and archive
    benchmark and lines-of-code stats.

  * Added upper bounds to all cabal dependencies.

  * Include man pages in extra-source-files.  This allows users to
    install pandoc from the tarball without needing to build the man
    pages.

  [bug fixes]

  * Filenames are encoded as UTF8.  Resolves Issue #252.

  * Handle curly quotes better in `--smart` mode. Previously, curly quotes
    were just parsed literally, leading to problems in some output formats.
    Now they are parsed as `Quoted` inlines, if `--smart` is specified.
    Resolves Issue #270.

  * Text.Pandoc.Parsing: Fixed bug in grid table parser.
    Spaces at end of line were not being stripped properly,
    resulting in unintended LineBreaks.

  * Markdown reader:

    + Allow HTML comments as inline elements in markdown.
      So, `aaa <!-- comment --> bbb` can be a single paragraph.
    + Fixed superscripts with links: `^[link](/foo)^` gets
      recognized as a superscripted link, not an inline note followed by
      garbage.
    + Fixed regression, making markdown reference keys case-insensitive again.
      Resolves Issue #272.
    + Properly handle abbreviations (like `Mr.`) at the end of a line.
    + Better handling of intraword underscores, avoiding exponential
      slowdowns in some cases.  Resolves Issue #182.
    + Fixed bug in alignments in tables with blank rows in the header.

  * RST reader:

    + Field lists now allow spaces in field names, and
      block content in field values. (Thanks to Lachlan Musicman
      for pointing out the bug.)
    + Definition list items are now always `Para` instead of
      `Plain`, matching behavior of `rst2xml.py`.
    + In image blocks, the description is parsed properly and
      used for the alt attribute, not also the title.
    + Skip blank lines at beginning of file. Resolves
      Debian #611328.

  * LaTeX reader:

    + Improved parsing of preamble.
      Previously you'd get unexpected behavior on a document that
      contained `\begin{document}` in, say, a verbatim block.
    + Allow spaces between `\begin` or `\end` and `{`.
    + Support `\L` and `\l`.
    + Skip comments inside paragraphs.

  * LaTeX writer:

    + Escape strings in `\href{..}`.
    + In nonsimple tables, put cells in `\parbox`.

  * OpenDocument writer:  don't print raw TeX.

  * Markdown writer:

    + Fixed bug in `Image`.  URI was getting unescaped twice!
    + Avoid printing extra blank lines at the end if there are
      no notes or references.

  * LaTeX and ConTeXt: Escape `[` and `]` as `{[}` and `{]}`.
    This avoids unwanted interpretation as an optional argument.

  * ConTeXt writer: Fixed problem with inline code.  Previously
    `}` would be rendered `\type{}}`. Now we check the string for '}' and '{'.
    If it contains neither, use `\type{}`; otherwise use `\mono{}`
    with an escaped version of the string.

  * `:` now allowed in HTML tags. Resolves Issue #274.

pandoc (1.6)

  [ John MacFarlane ]

  * New EPUB and HTML Slidy writers. (Issue #122)

      - EPUB is a standard ebook format, used in Apple's iBooks for
        the iPad and iPhone, Barnes and Noble's nook reader, the Sony
        reader, and many other devices.
      - Slidy, like S5, is a system for producing HTML+javascript slide
        shows.

  * All input is assumed to be UTF-8, no matter what the locale and ghc
    version, and all output is UTF-8. This reverts to pre-1.5 behavior.
    Also, a BOM, if present, is stripped from the input.

  * Markdown now supports grid tables, whose cells can contain
    arbitrary block elements. (Issue #43)

  * Sequentially numbered example lists in markdown with `@` marker.

  * Markdown table captions can begin with a bare colon and no longer need
    to include the English word "table." Also, a caption can now occur
    either before or after the table. (Issue #227)

  * New command-line options:

      - `--epub-stylesheet` allows you to specify a CSS file that will
        be used to style your ebook.
      - `--epub-metadata` allows you to specify metadata for the ebook.
      - `--offline` causes the generated HTML slideshow to include all
        needed scripts and stylesheets.
      - `--webtex` causes TeX math to be converted to images using the
        Google Charts API (unless a different URL is specified).
      - `--section-divs` causes div tags to be added around each section
        in an HTML document. (Issue #230, 239)

  * Default behavior of S5 writer in standalone mode has changed:
    previously, it would include all needed scripts and stylesheets
    in the generated HTML; now, only links are included unless
    the `--offline` option is used.

  * Default behavior of HTML writer has changed. Between 1.2 and 1.5,
    pandoc would enclose sections in div tags with identifiers on the
    div tags, so that the sections can be manipulated in javascript.
    This caused undesirable interactions with raw HTML div tags. So,
    starting with 1.6, the default is to put the identifiers directly
    on the header tags, and not to include the divs.  The `--section-divs`
    option selects the 1.2-1.5 behavior.

  * API changes:

      - `HTMLMathMethod`: Added `WebTeX`, removed `MimeTeX`.
      - `WriterOptions`: Added `writerUserDataDir`, `writerSourceDirectory`,
        `writerEPUBMetadata` fields. Removed `writerIncludeBefore`,
        `writerIncludeAfter`.
      - Added `headerShift` to `Text.Pandoc.Shared`.
      - Moved parsing code and `ParserState` from `Text.Pandoc.Shared`
        to a new module, `Text.Pandoc.Parsing`.
      - Added `stateHasChapters` to `ParserState`.
      - Added `HTMLSlideVariant`.
      - Made `KeyTable` a map instead of an association list.
      - Added accessors for `Meta` fields (`docTitle`, `docAuthors`,
        `docDate`).
      - `Pandoc`, `Meta`, `Inline`, and `Block` have been given `Ord`
        instances.
      - Reference keys now have a type of their own (`Key`), with its
        own `Ord` instance for case-insensitive comparison.
      - Added `Text.Pandoc.Writers.EPUB`.
      - Added `Text.Pandoc.UUID`.
      - Removed `Text.Pandoc.ODT`, added `Text.Pandoc.Writers.ODT`.
        Removed `saveOpenDocumentAsODT`, added `writeODT`.
      - Added `Text.Pandoc.Writers.Native` and `writeNative`.
        Removed `prettyPandoc`.
      - Added `Text.Pandoc.UTF8` for portable UTF8 string IO.
      - Removed `Text.Pandoc.Writers.S5` and the `writeS5` function.
        Moved `s5Includes` to a new module, `Text.Pandoc.S5`.
        To write S5, you now use `writeHtml` with `writerSlideVariant`
        set to `S5Slides` or `SlidySlides`.

  * Template changes.  If you use custom templates, please update them,
    particularly if you use syntax highlighting with pandoc. The old HTML
    templates hardcoded highlighting CSS that will no longer work with
    the most recent version of highlighting-kate.

      - HTML template: avoid empty meta tag if no date.
      - HTML template: Use default highlighting CSS from highlighting-kate
        instead of hard-coding the CSS into the template.
      - HTML template: insert-before text goes before the title, and
        immediately after the <body> tag, as documented. (Issue #241)
      - Added slidy and s5 templates.
      - Added amssymb to preamble of latex template. (github Issue 1)

  * Removed excess newlines at the end of output. Note: because output
    will not contain an extra newline, you may need to make adjustments
    if you are inserting pandoc's output into a template.

  * In S5 and slidy, horizontal rules now cause a new slide, so you
    are no longer limited to one slide per section.

  * Improved handling of code in man writer. Inline code is now monospace,
    not bold, and code blocks now use .nf (no fill) and .IP (indented para).

  * HTML reader parses `<tt>` as Code. (Issue #247)

  * html+lhs output now contains bird tracks, even when compiled without
    highlighting support. (Issue #242)

  * Colons are now no longer allowed in autogenerated XML/HTML identifiers,
    since they have a special meaning in XML.

  * Code improvements in ODT writer.  Remote images are now replaced with
    their alt text rather than a broken link.

  * LaTeX reader improvements:

      - Made latex `\section`, `\chapter` parsers more forgiving of
        whitespace.
      - Parse `\chapter{}` in latex.
      - Changed `rawLaTeXInline` to accept `\section`, `\begin`, etc.
      - Use new `rawLaTeXInline'` in LaTeX reader, and export `rawLaTeXInline`
        for use in markdown reader.
      - Fixes bug wherein `\section{foo}` was not recognized as raw TeX
        in markdown document.

  * LaTeX writer:  images are automatically shrunk if they would extend
    beyond the page margin.

  * Plain, markdown, RST writers now use unicode for smart punctuation.

  * Man writer converts math to unicode when possible, as in other writers.

  * `markdown2pdf` can now recognize citeproc options.

  * Command-line arguments are converted to UTF-8. (Issue #234)

  * `Text.Pandoc.TeXMath` has been rewritten to use texmath's parser.
    This allows it to handle a wider range of formulas. Also, if a formula
    cannot be converted, it is left in raw TeX; formulas are no longer
    partially converted.

  * Unicode curly quotes are left alone when parsing smart quotes. (Issue
    #143)

  * Cabal file changes:

      - Removed parsec < 3 restriction.
      - Added 'threaded' flag for architectures where GHC lacks a threaded
        runtime.
      - Use 'threaded' only for markdown2pdf; it is not needed for pandoc.
      - Require highlighting-kate 0.2.7.

  * Use explicit imports from `Data.Generics`. Otherwise we have a
    conflict with the 'empty' symbol, introduced in syb >= 0.2. (Issue #237)

  * New data files:  slidy/slidy.min.js, slidy/slidy.min.css, epub.css.

pandoc (1.5.1.1)

  [ John MacFarlane ]

  * Fixed uniqueIdent in Shared so that header identifiers work as
   advertized in the README and are are valid XHTML names.

pandoc (1.5.1)

  [ John MacFarlane ]

  * Fixed treatment of unicode characters in URIs.
    + Shared now exports escapeURI and unescapeURI. These handle
      UTF8 encoding and decoding as well as URI escaping/unescaping.
    + Shared: uri and emailAddress now return a pair of the original
      parsed text and the escaped URI (in the latter case, with
      the mailto: prefix).
    + HTML reader: unsanitaryURI has been modified to allow unicode
      high characters in a URI.
    + Readers:  All link and image URIs are now escaped using
      escapeURI.
    + Markdown and RST writers:  unescapeURI is used so that URIs
      in these formats are human-readable.

  * Setup.hs: Don't assume that the build directory is "dist".
    Instead, get it from localBuildInfo.

  * OpenDocument writer: Use a Map for stTextStyleAttr.
    This avoids duplicates (and invalid xml). Resolves Issue #222.

pandoc (1.5.0.1)

  [ John MacFarlane ]

  * HTML writer: Fixed error in math writer (with MathML option)
    that caused an infinite loop for unparsable MathML.

pandoc (1.5)

  [ John MacFarlane ]

  * Added --mathml option.  When this is selected, pandoc will convert
    TeX math into MathML.
    + Added data/MathMLinHTML.js, which is included when no URL is
      provided for --mathml.  This allows MathML to be displayed (in
      better browsers) as text/html.
    + Removed Text.Pandoc.LaTeXMathML.  The module was no longer
      necessary; it was replaced by two lines in pandoc.hs.
    + Replaced LaTeXMathML.js.commend and LaTeXMathML.js.packed with a
      single combined file, LaTeXMathML.js.

  * Added --data-dir option.
    This specifies a user data directory. If not specified, will default
    to ~/.pandoc on unix or Application Data\pandoc on Windows.
    Files placed in the user data directory will override system default
    data files.

  * Added Maybe datadir parameter to readDataFile, saveOpenDocumentAsODT,
    latexMathMLScript, s5HeaderIncludes, and getDefaultTemplate. If
	  Nothing, no user directory is searched for an override.

  * Added 'plain' output format. This is similar to markdown, but
    removes links, pictures, inline formatting, and most anything that
    looks even vaguely markupish. The function writePlain is exported by
    Text.Pandoc.Writers.Markdown, with which it shares most of its code.

  * Allow multi-line titles and authors in meta block.
    Titles may span multiple lines, provided continuation lines
    begin with a space character.  Separate authors may be put on
    multiple lines, provided each line after the first begins with
    a space character.  Each author must fit on one line. Multiple
    authors on a single line may still be separated by a semicolon.
    Based on a patch by Justin Bogner.

  * When given an absolute URI as parameter, pandoc will try to fetch
    the content via HTTP.  So you can do:
    'pandoc -r html -w markdown http://www.fsf.org'
    Adds dependency on HTTP.

  * Made HTML reader much more forgiving.
    + Incorporated idea (from HXT) that an element can be closed
      by an open tag for another element.
    + Javascript is partially parsed to make sure that a <script>
      section is not closed by a </script> in a comment or string.
    + More lenient non-quoted attribute values.
      Now we accept anything but a space character, quote, or <>.
      This helps in parsing e.g. www.google.com!
    + Bare & signs are now parsed as a string.  This is a common
      HTML mistake.
    + Skip a bare < in malformed HTML.

  * Removed html2markdown and hsmarkdown.
    + html2markdown is no longer needed, since you can now pass URI
      arguments to pandoc and directly convert web pages. (Note,
      however, that pandoc assumes the pages are UTF8. html2markdown
      made an attempt to guess the encoding and convert them.)
    + hsmarkdown is pointless -- a large executable that could be
      replaced by 'pandoc --strict'.

  * In most writers, an image in a paragraph by itself is now rendered
    as a figure, with the alt text as the caption. (Texinfo, HTML, RST,
    MediaWiki, Docbook, LaTeX, ConTeXt, HTML.) Other images are
    rendered inline.

  * Depend on extensible-exceptions.  This allows pandoc to be compiled
    on GHC 6.8.

  * Added --base-header-level option.  For example, --base-header-level=2
    will change level 1 headers to level 2, level 2 to level 3, etc.
    Closes Debian #563416.

  * Incomplete support for RST tables (simple and grid).
    Thanks to Eric Kow. Colspans and rowspans not yet supported.

  * Added accessors (docTitle, docAuthors, docDate) to Meta type.

  * MediaWiki writer:  format links with relative URLs as wikilinks.
    The new rule:  If the link target is an absolute URL, an external
    link is created. Otherwise, a wikilink is created.

  * Text.Pandoc.Shared: Export uniqueIdent, and don't allow tilde in
    identifier.  Note:  This may break links to sections that involve
    tildes.

  * Markdown(+lhs) reader:  handle "inverse bird tracks."
    Inverse bird tracks (<) are used for haskell example code that is not
    part of the literate Haskell program.  Resolves Issue #211.

  * LaTeX reader:
    + Recognize '\ ' (interword space).
    + Recognize nonbreaking space '~'.
    + Ignore \section, \pdfannot, \pdfstringdef.  Ignore alt title in
      section headers.  Don't treat \section as inline LaTeX.
      Resolves Issue #202.
    + LaTeX reader:  allow any special character to be escaped.
      Resolves Issue #221.
    + LaTeX reader: treat \paragraph and \subparagraph as level 4, 5
      headers.  Resolves Issue #207.

  * Use template variables for include-before/after.
    + These options now imply -s; previously they worked also in fragment
      mode.
    + Users can now adjust position of include-before and include-after
      text in the templates.
    + Default position of include-before moved back (as it was before 1.4)
      before table of contents.
    + Resolves Issue #217.

  * Don't print an empty table header: (all writers).
    Resolves Issue #210.

  * HTML, Docbook writer: Use tbody, thead, and cols in tables.

  * HTML writer: Don't include TOC div if table of contents is empty.

  * Markdown writer:  Fixed citations.
    Previously the markdown writer printed raw citation codes, e.g.
    [geach1970], rather than the expanded citations provided by
    citeproc, e.g. (Geach 1970). Now it prints the expanded citations.
    This means that the document produced can be processed as a markdown
    document without citeproc. Thanks to dsanson for reporting, and
    Andrea Rossato for the patch.

  * Improved and simplified title block in context template.
    Previously it caused an error if there was no title.
    This method should also be easier for users to customize.

  * Markdown reader:
    + Treat p., pp., sec., ch., as abbreviations in smart mode.
    + Disallow blank lines in inline code span.
    + Allow footnotes to be indented < 4 spaces.
      This fixes a regression.  A test case has been added.
    + Escape spaces in URLs as %20. Previously they were incorrectly
		  escaped as +, which is appropriate only for the query part of
			a URL. Resolves Issue #220.
    + Require two spaces after capital letter + period for list item.
      Otherwise "E. coli" starts a list. This might change the semantics
      of some existing documents, since previously the two-space
      requirement was only enforced when the second word started
      with a capital letter. But it is consistent with the existing
      documentation and follows the principle of least surprise.
      Resolves Issue #212.

  * LaTeX template: redefine labelwidth when using enumerate package.
    Otherwise the list labels (numbers) often extend past the left
    margin, which looks bad.

  * Mediawiki writer: Don't print a "== Notes ==" header before
    references.  This is too English-centric. Writers can provide their
    own header at the end of the document.

  * Promoted mediawiki headers.  '= head =' is now level 1, '== head =='
    level 2, etc.  This seems to be correct; it's only by convention
    that wikipedia articles have level 2 headers at most.
    Patch due to Eric Kow.

  * RunTests.hs: Set LANG to a UTF-8 locale. Use 'pandoc --data-dir=' so
    data files don't need to have been installed. This removes the need to
    set HOME.

  * HTML reader:
    + Handle spaces before <html>.  Resolves Issue #216.
    + Be forgiving in parsing a bare list within a list.
      The following is not valid xhtml, but the intent is clear:
      <ol>
      <li>one</li>
      <ol><li>sub</li></ol>
      <li>two</li>
      </ol>
      We'll treat the <ol> as if it's in a <li>.  Resolves Issue #215.

  * Updated INSTALL instructions.  cabal method is now promoted.

  * Updated markdown2pdf man page. It no longer says all pandoc options
    are accepted.

  * README/man pages: Removed advice to pipe through tidy before HTML
    reader.  This is obsolete, now that we have a forgiving HTML parser.

  * LaTeX writer: set numbersections template variable, so
    the section numbering options work again.

  * Removed obsolete Makefile.

  * Website: renamed index.txt.in -> index.txt.

  * New batch file to make-windows-installer.
    + Removed old Makefile.windows
    + Added make-windows-installer.bat
    + Modified default installer name in pandoc-setup.iss

  * Removed freebsd and macports directories.
    They are no longer up to date.

  * Setup.hs:
    + Made man page building sensitive to build verbosity.
    + Improved detection of highlighting support in test hook.
    + Install wrapper scripts into cabal bin directory.
    + Also simplified installManpages.
    + Setup.hs: install manpages to mandir.  Code borrowed from darcs.

  * Changed default of writerXeTeX to False.

  * HTML writer: don't include empty UL if --toc but no sections.
    Resolves Issue #199.

  * LaTeX writer:

    + If book, report, or memoir documentclass, use \chapter{}
      for first-level headers. Otherwise use \section{}.
    + Removed stLink, link template variable. Reason: we now always
      include hyperref in the template.

  * LaTeX template:

    + Only show \author if there are some.
    + Always include hyperref package. It is used not just for links but
      for toc, section heading bookmarks, footnotes, etc. Also added
      unicode=true on hyperref options.

  * markdown2pdf: always do at least two runs. hyperref bookmarks
    require this.

  * cabal file: Removed unneeded dependency on template-haskell.

  * Windows installer - fixed bug in data file locations.
    Resolves Issue #197.

  * Deprecated --custom-header in documentation.
    Removed old "Custom Headers" section in README.

pandoc (1.4)

  [ John MacFarlane ]

  * Pandoc will now compile with either GHC 6.10 or 6.12.
    + Don't use System.IO.UTF8 when compiling with 6.12
    + Use -fno-warn-unused-do-bind option when compiling with 6.12

  * Replaced old headers with templates.  Now users have much more
    control over the way documents appear in --standalone mode,
    and writer code is simplified.  Resolves Issues #59, 147.
    Every effort has been made to retain backwards compatibilty.
    So, the --custom-header option should still work as before.

    + Added Text.Pandoc.Templates.  This provides functions for
      retrieving default templates and for rendering templates.
    + System templates (in the pandoc data directory) can be
      overridden by user templates in $HOME/.pandoc/templates.
    + Removed Text.Pandoc.DefaultHeaders.
    + Removed data/headers directory.
    + Added templates directory.
    + Added writerTemplate and writerVariables fields to WriterOptions.
    + Removed writerTitlePrefix, writerHeader fields from WriterOptions.
    + Changed --print-default-header to --print-default-template.
    + Added --template option.
    + Added -V/--variable option to set custom template variables.

  * Pandoc no longer requires Template Haskell.  Resolves Issue #186.

    + Removed need for TH in ODT module. Instead get reference.odt from
	  data file at run time.
    + Removed TH dependency from S5 module. S5 module now exports
      s5HeaderIncludes, which pandoc.hs includes if writer is s5 and
      standalone.
    + Refactored LaTeXMathML not to use TH.

  * Meta is now Meta [Inline] [[Inline]] [Inline] rather than
    Meta [Inline] [String] String. Authors and date in Meta are now lists
    of Inline elements rather than raw strings. This means that they can
    be formatted and can include footnotes. NOTE: This may be a breaking
    change for those using pandoc as a library.

  * Added readDataFile to Text.Pandoc.Shared.  This retrieves
    a data file from the user pandoc data directory (~/.pandoc
    on unix), or, if not found there, from the system data
    directory ($CABALDIR/shared/pandoc-VERSION/). All data
    files, including templates, LaTeXMathML.js, s5 styles,
	and reference.odt, can be overridden by the user.

  * s5 files moved from data/ui/default to s5/default.

  * Use unicode instead of entities in HTML and XML output.  Resolves
    Issue #163.

  * Prettier HTML footnote references:  put anchor inside sup,
    instead of other way.  Resolves Issue #191. Thanks to
    infinity0x.

  * Added --xetex option to pandoc and markdown2pdf.
    If --xetex is specified, pandoc produces latex suitable for
    processing by xelatex, and markdown2pdf uses xelatex to create
    the PDF. Resolves Issue #185.

  * RTF writer: multiple authors now occupy multiple paragraphs rather
    than using a line break.

  * Man writer: now the "--after-body" will come after the "AUTHORS"
    section, whereas before it would come before it.  This is a
    slight break from backwards compatibility.

  * Added --reference-odt option, so users may customize the styles
    used in pandoc-generated ODT files.  Users may also place a
	default reference.odt in the ~\.pandoc directory.

  * ODT writer:
    + Indented and line-broke styles.xml so it can be modified more easily.
    + Omitted some unnecessary style declarations.
    + Don't wrap text in OpenDocument writer.  The tags are too long, making
      wrapping ugly and pointless.

  * LaTeX reader: use \\ to separate multiple authors.

  * Markdown reader: use ; as separator between authors.
    This allows you to use ',' within author names: e.g. "John Jones, Jr."

  * S5 writer: use linebreak to separate authors in title page.

  * RST reader: Allow :: before lhs code block. The RST spec requires the
    :: before verbatim blocks. This :: should not be treated as literal
    colons. Resolves Issue #189.

  * Documented pandoc 1.3's new definition list syntax in README.
    (An oversight in the last release.)

  * markdown2pdf.hs:
    + interpret ! in a log as an error line.
    + --toc now works properly.

  * Changes in RunTests.hs:
    + Use the Diff library rather than a local copy of Diff.hs.
      (This vastly increases performance.) This change means that 'cabal
      test' presupposes that the Diff library is installed.
    + Removed tests/Diff.hs from cabal file.
    + Changed RunTests to use local environment. We need at least HOME, so
      pandoc can find its data directory.

  * Updated windows installer to install data files in the app directory.

  * Windows installer now installs portable wrappers hsmarkdown and
    markdown2pdf.

pandoc (1.3)

  [ John MacFarlane ]

  * Added --id-prefix option (Issue #41). This adds a prefix to all
    automatically generated HTML identifiers, which helps prevent
    duplicate identifiers when you're generating a fragment (say a blog
    post).

  * Added --indented-code-classes option.  This specifies classes
    to use for indented code blocks.  (Patch due to buttock; Issue #87.)

  * --number-sections now affects HTML output as well as ConTeXt and LaTeX
    (Issue #150).

  * Improved syntax for markdown definition lists (Issue #24).
    Definition lists are now more compatible with PHP Markdown Extra.
    + You can have multiple definitions for a term (but still not
      multiple terms).
    + Multi-block definitions no longer need a column before each block
      (indeed, this will now cause multiple definitions).
    + The marker no longer needs to be flush with the left margin,
      but can be indented at or two spaces.  Also, ~ as well as :
      can be used as the marker (this suggestion due to David
      Wheeler.)
    + There can now be a blank line between the term and the
      definitions.

  * Better looking simple tables.  Resolves Issue #180.
    + Markdown reader: simple tables are now given column widths of 0.
    + Column width of 0 is interpreted as meaning: use default column width.
    + Writers now include explicit column width information only
      for multiline tables.  (Exception:  RTF writer, which requires
      column widths.  In this case, columns are given equal widths,
      adding up to the text width.)
    + Simple tables should now look better in most output formats.

  * Allow markdown tables without headers (Issue #50).
    The new syntax is described in README. Also allow optional line of
    dashes at bottom of simple tables.

  * Compensate for width of final table column (Issue #144).

  * Treat a backslash followed by a newline as a hard line break
    in markdown.  Resolves Issue #154.  This is a nice alternative
	to markdown's "invisible" way of indicating hardline breaks
	using lines that end with two spaces.

  * Improved performance of markdown reader by ~10% by eliminating the
    need for a separate parsing pass for notes. Raw notes are now stored
    on the first pass (which parses references), then parsed when the
    note is inserted into the AST. The stateNotes field in ParserState
    is now a list of [(String, String)] pairs instead of [(String,
    [Block])].

  * In markdown reader, treat 4 or more * or _ in a row as literal
    text.  (Trying to parse long strings of * or _ as strong or emph
	leads to exponential performance problems.)

  * Markdown reader: Use + rather than %20 for spaces in URLs.

  * Fixed htmlComment parser, adding a needed 'try'.

  * Don't print raw HTML in man output.

  * Allow . _ and ~ in header identifiers.

  * Specially mark code blocks that were "literate" in the input.
    They can then be treated differently in the writers.  This allows
    authors to distinguish bits of the literate program they are writing
    from source code examples, even if the examples are marked as
    Haskell for highlighting.  (Issue #174.)

  * Modified html+lhs output to use "haskell" highlighter instead
    of "literateHaskell". The highlighting module now adds bird tracks
	after highlighting (for HTML output), if the code block has the
	"literate" class.  This gives better results, because kate's
    haskell highlighter is much better than the literateHaskell
    highlighter.

  * Fixed handling of footnotes in titles (HTML) and headers (LaTeX).
    (Issue #137.)

  * Support for "..code-block" directive in RST reader. Not core
    RST, but used in Sphinx for code blocks annotated with syntax
    information. Thanks to Luke Plant for the patch.

  * Added "head" to list of block-level HTML tags. Resolves
    Issue #108.

  * Added stripTags to Text.Pandoc.XML.  This is used in the HTML writer.

  * Set utf-8 encoding in texinfo headers.

  * Docbook writer: add ids to sections.  Use link for internal links.
    (Issue #60.)

  * Blank lines after lists in MediaWiki writer.

  * Properly handle commented-out list items in markdown.
    Resolves Issue #142.  Example:

    - a
    <!--
    - b
    -->
    - c

  * Changed heuristic in compactify. compactify has to decide whether a
    Para that ends a list is a Para intentionally, or just because of
    the blank lines at the end of every list. In the latter case the
    Para is turned to a Plain. The old heuristic was: change final Para
    to Plain iff the other items all end in Plain. This produces bad
    results when, for example, an item contains just a Plain and an HTML
    comment, as it does in the list above. The new heuristic: change
    final Para to Plain iff the other items don't contain a Para.

  * Added % as an rst underline character.  Resolves Issue #173.

  * Fix inline math parser so that \$ is allowed in math.
    Resolves Issue #169.

  * Translate \int (integral) into unicode when using unicode math
    method. Resolves Issue #177.

  * markdown2pdf.hs improvements:
    + Use System.IO.UTF8.
    + Print error messages on last attempt.
    + Do not create a backup when overwriting a PDF (Issue #166).
    + Accept --longopt=val options.
    + Added man/man1/markdown2pdf.1 to extra-tmp-files in cabal, so that
      it is properly cleaned.

  * Added haddock comments warning that readers assume \n line endings.

  * Updated COPYRIGHT file.

  * Makefile: Changed EXECSBASE so it doesn't pull in hsmarkdown &
    markdown2pdf. Otherwise strip tries to strip shell scripts when you
    install using 'make'.

  * Changed Makefile so it doesn't build Haskell wrappers.

  * Fixed Makefile so it doesn't try to build man pages in build-doc.

  * Install pcre3.dll in Windows install script; this allows us to
    package a version of pandoc with highlighting support.

pandoc (1.2.1)

  [ John MacFarlane ]

  * Fixed regression with --preserveTabs. Brought back optPreserveTabs.
    The trick of setting tabStop to 0 to mean "preserve tabs" had a bad
    side effect: strings of 0 spaces were interpreted as indentation.
    So, with --preserve-tabs, unindented paragraphs were treated as
    code. Resolves Issue #138.

  * HTML writer:  wrap sections in divs.  Resolves Issue #70.

    + hierarchicalize has been rationalized; it builds a hierarchical
      representation of the document from the headers, and simultaneously
      gives each section a unique identifier based on the heading title.
    + Identifiers are now attached to the divs rather than
      to the headers themselves.
    + Table of content backlinks go to the beginning of the table, rather
      than to the section reference that was clicked.
    + Code for constructing identifiers has been moved to Text.Pandoc.Shared
      from the HTML writer, since it is now consumed only by
      hierarchicalize.
    + In --strict mode, pandoc just prints bare headings, as before
      (unless --toc has been specified).
    + In s5 output, it does not wrap sections in divs, as that seems to
      confuse the s5 javascript.

  * Man writer: break lines at end of each sentence. groff expects this
    and treats '.' and '?' differently when followed by line ending as
    opposed to ordinary space. Also, don't escape periods. Instead, use
    zero-width character \& to avoid unwanted interpretation of periods
    at start of line. Resolves Issue #148.

  * Markdown writer:  Added '#' and '>' to list of characters to be
    escaped in markdown output.  Removed '<', as it is not an officially
    escapable character. This partially resolves Issue #96.

  * Make --smart the default for man output format. Otherwise we have
    trouble dividing lists of endlines into sentences.

  * DocBook writer: Use language attribute to indicate source language
    in code blocks.

  * RST reader:

    + Allow # to continue list, even if the list was started with an
      explicit marker. For example:

        A. my list
		#. continued

      Resolves Issue #140.
    + Allow continuation lines in line blocks. Also added test cases for
      line blocks for RST reader. Resolves Issue #149.
    + Allow explicit links with spaces in URL: `link <to this>`_

  * Improved LaTeX reader's coverage of math modes. Remove displaymath*
    (which is not in LaTeX) and recognize all the amsmath environments
    that are alternatives to eqnarray, namely equation, equation*,
    gather, gather*, gathered, multline, multline*, align, align*,
    alignat, alignat*, aligned, alignedat, split. Resolves Issue #103.
    Thanks to shreevatsa.public for the patch.

  * Markdown reader:

    + Allow -, _, :, . in markdown attribute names. These are legal in
      XML attribute names.
    + Use non-breaking spaces in abbreviations.
    + Markdown reader: improved efficiency of abbreviation parsing.
      Instead of a separate abbrev parser, we just check for
      abbreviations each time we parse a string. This gives a huge
      performance boost with -S. Resolves Issue #141.

  * Improved efficiency of shared parsers:  hexNum, htmlComment,
    whitespace, indentSpaces.

  * Export HTMLMathMethod in Text.Pandoc.

  * Export languagesByExtension in Text.Pandoc.Highlighting.

  * Added new Haskell version of markdown2pdf, due to
    Paulo Tanimoto.  This should be more portable than the old
	shell script.

  * Made 'pandoc -v' more explicit about compiler options.
    Resolves Issue #139.

  * pandoc.hs:  Made --strict compatible with --standalone, --toc.

  * Use Paths_pandoc to get version number, instead of hard-coding it
    into Text/Pandoc.hs.

pandoc (1.2)

  [ John MacFarlane ]

  * Added support for literate Haskell. lhs support is triggered by
    '+lhs' suffixes in formats. For example, 'latex+lhs' is literate
    Haskell LaTeX. '.lhs' files are treated by default as literate
    markdown.

    + Added stateLiterateHaskell to parser state.
    + Added parser for lhsCodeBlock to Markdown, RST, LaTeX readers.
    + Added parser for |inline lhs| to LaTeX reader.
    + Added writerLiterateHaskell to WriterOptions.
    + Added lhs support to Markdown, RST, LaTeX, HTML writers.
    + Added definition of code environment to LaTeX header.
    + Added tests (run only if highlighting support compiled in).
    + Documented lhs features in man page and README.

  * In Text.Pandoc.Definition, added processWith, processWithM,
    and queryWith, and deprecated processPandoc and queryPandoc
    for these more general functions.

  * Fixed bug in mediawiki writer:  improper closing tags in tables.
    Thanks to Benct Philip Jonsson for reporting the bug.

  * Added --email-obfuscation option.

    + Added writer option for email obfuscation.
    + Implemented email obfuscation options in HTML writer.
    + Added option to option parser.
    + Documented in README and pandoc man page.
    + Resolves Issue #97.

  * LaTeX writer: fixed bug with empty table cells.
    Resolves Issue #107.  Thanks to rodja.trappe for the patch.

  * Fixed bug with header spacing in Markdown and RST writers.
    A null header (Meta [] [] []) should not cause a blank line
    at the beginning of output.  But a blank line is needed between
    a non-null header and the main text.

  * Markdown reader: Relax spacing rules for $$ in display math. Now
    space and newlines are allowed after the opening $$ and before the
    closing $$. However, the display math cannot contain an entirely
    blank line. Resolves Issue #105.

  * Markdown reader: Gobble space after Plain blocks containing only
    raw html inline. Otherwise following header blocks are not parsed
    correctly, since the parser sees blank space before them. Resolves
    Issue #124.

  * Markdown reader: Allow " as well as '' to end a latex double-quote.

  * Conditionally depend on syb and base >= 4 if ghc >= 6.10.
    Resolves Issue #109.

  * Fixed problems in RST and markdown output due to bug in pretty-1.0.1.0

    + Added hang' function to Text.Pandoc.Shared; this will be used instead
      of hang, which doesn't work properly in pretty-1.0.1.0.  When pretty
      is upgraded, we can go back to hang.
      See  http://article.gmane.org/gmane.comp.lang.haskell.general/16687
    + Use hang' (and some different techniques) in RST and markdown writers.
      Some output is now a bit different.

  * Brought citeproc support up to date for citeproc-hs-0.2.
    (Patch by Andrea Rossato.)

  * Moved all haskell source to src subdirectory.  Renamed Main.hs to
    pandoc.hs.

  * Rewrote hsmarkdown in Haskell for portability (src/hsmarkdown.hs).
    For now, keeping the old shell script too.

  * Added TemplateHaskell to Extensions for executable, removed
    -threaded for library. Thanks to duncan.coutts for the bug report.
    Resolves Issue #121.

  * Moved some Extra-Source-Files to Data-Files.

  * Moved tabFilter to Shared.

  * In pandoc.hs, removed optPreserveTabs; instead, tabstop of 0 means
    preserve tabs.

  * Minor code cleanup based on hlint suggestions.

pandoc (1.1)

  [ John MacFarlane ]

  * Main.hs:

    + Changed date on copyright message in Main.hs.
    + Have the '-v' option print syntax highlighting languages
      separated by commas, and wrapped in lines, instead of in five
      columns as before.

  * Added --jsmath option.  Resolves Issue #68.

    + Added --jsmath option to Main.hs
    + Added JsMath to HTMLMathMethod in Text.Pandoc.Shared.
    + Handle math appropriately in HTML writer when JsMath selected.
    + Documented the option in README and man page.

  * Text.Pandoc.Shared:  Changed compactify to use a better heuristic
    for tight and loose lists. Final Para is changed to Plain if all
    other list items *end* with a Plain block. Addresses Issue #99.

  * HTML reader:

    + Added colons to protocols in unsanitaryURI. Closes Issue #88.
    + HTML reader: Don't interpret contents of <pre> blocks as markdown.
      Added rawVerbatimBlock parser.  Resolves Issue #94.

  * Markdown reader:

    + Allow URLs with spaces in them in links and references, but escape
      them as "%20".
    + Allow blank space at the end of horizontal rules.

  * RST reader: Modified 'unknownDirective' parser to handle comment
    blocks correctly, and added tests for comment blocks. Resolves Issue
    #86. Closes Debian Bug #500662.

  * HTML writer:

    + Include classes on tr elements in HTML output:
      "header", "odd", "even".  This allows tables to be styled with
      lines in alternating colors.  Resolves Issue #91.
    + Enclose all LaTeXMathML bits in <span class="LaTeX">.
      This prevents parts of the document that are not math from being
      interpreted as math by LaTeXMathML.js.

  * OpenDocument and ODT writers:  Added support for HorizontalRule elements,
    which were formerly ignored.  Resolves Issue #95.

  * Text.Pandoc.Shared:  Modified wrappedTeX to eliminate the line break
    between a footnote and immediately following nonspace characters in
    LaTeX and ConTeXt output. (This gets interpreted as a space, which
    is not desired in cases like `text^[note]---`.)  Resolves Issue #93.

  * Windows installer: Don't require admin privileges to run
    installer.  Modified pandoc-setup.iss, and changed modpath.iss to
    modify HKCU path if user lacks admin privileges.  Also fixed case
    where oldpath is empty (previously this led to the new path
    beginning with a semicolon).

  * Updated INSTALL instructions for Arch packages and OS X install using
    cabal-install.

  * Removed the (now unneeded) debian directory.
    Removed empty Codec and System directories.

  * Moved odt-styles/ to data/.  Removed unneeded variable in Makefile.

  * Modified Setup.hs so that the "test" target returns an error status
    when tests fail, and "build" returns a success status if
    the build succeeds.  Resolves Issue #100.

  * Added BUGS to files in tarball.


pandoc (1.0.0.1)

  [ John MacFarlane ]

  * Removed spurious reference to pdf output format from pandoc(1) man page.

pandoc (1.0)

  [ Andrea Rossato ]

  * Added new OpenDocument writer.

  * Added support for SmallCaps inline element.

  * Added support for integrating pandoc with citeproc-hs.

    + Added Cite element to definition and writers.
    + Added Text.Pandoc.Biblio module
    + Note: This support is included only if the 'citeproc'
      Cabal configuration flag is set.

  * Made Pandoc data structure an instance of Typeable.
    Added new processPandoc and queryPandoc functions, to query
    or transform matching elements in a Pandoc structure.

  [ Peter Wang ]

  * Added new Texinfo writer.

  [ John MacFarlane ]

  * Changes to Texinfo writer:

    + No space between paragraph and following @verbatim (provides more
      pleasing appearance in text formats)
    + Blank line consistently after list environments.
    + Removed deVerb.
    + Use @code instead of @verb for inline code (this solves the character
      escaping problem for texi2dvi and texi2pdf).
    + Added news of Texinfo writer to README.
    + Added Texinfo to list of formats in man page, and removed extra 'groff'.
    + Added texi & texinfo extensions to Main.hs, and fixed bug in determining
      default output extension.
    + Modified disallowedInNode in Texinfo writer to correct list of disallowed characters.

  * Added tests for OpenDocument writer.

  * Added ODT writer (using zip-archive library to package output of
    OpenDocument writer).  Added odt-styles directory with default ODT styles.

  * Added new mediawiki writer and tests.

  * Markdown reader: Added support for delimited code blocks, with optional
    syntax highlighting using highlighting-kate (if the 'highlighting'
    configuration option is selected).

    + Currently highlighting is supported only in the HTML writer.
    + Delimited code blocks can have attributes; using the language name as
      class triggers highlighting.
    + New Attributes parameter in CodeBlock structure.
    + --version now indicates whether syntax highlighting support is compiled
      in, and prints a list of supported languages

  * Removed debian directory. Pandoc is no longer a native debian package.

  * Changes to build process:  pandoc can now be built from the repository
    using Cabal.  No unix tools are needed (so, pandoc can be built on Windows
    without Cygwin).

    + Include shell scripts themselves in repo, rather than generating from wrappers.
      Removed wrappers directory and wrappers Makefile target.
    + Text/Pandoc/ASCIIMathML.hs, Text/Pandoc/DefaultHeaders.hs,
      and Text/Pandoc/Writers/S5.hs are no longer built in Makefile
      from templates in the templates/ directory. Instead, they use template
      haskell to read data at compile time from the relevant files in data/.
      Template haskell functions go in a new module, Text.Pandoc.TH.
    + man pages are now generated in Setup.hs hook, not by Makefile
    + Makefile 'tarball' target now calls Cabal's 'sdist'
    + Added "Extra-Source-Files" to pandoc.cabal, so sdist contains everything needed
    + Added "Build-Type" field to pandoc.cabal to avoid warning.
    + Added to "Extra-source-files" and "Extra-tmp-files" in pandoc.cabal,
      so 'sdist' and 'clean' will work properly.
    + Setup.hs now generates man pages in a postbuild hook.
    + Added dependency-checking to Setup.hs, so it only rebuilds things
      that need rebuilding.
    + Added 'library' and 'executable' configuration flags.
      Cabal can now be told to build just the library or just the executable.
    + CABALOPTS may now be specified with 'make' to pass Cabal configuration flags.
      For example:  CABALOPTS=-fhighlighting make

  * Rewrote test suite so it doesn't depend on perl or unix tools.

    + Replaced old runtests.pl with a Haskell script RunTests.hs.
    + Added Diff.hs module to be used by RunTests.hs instead of unix 'diff'.
    + Added test hook to Setup.hs, so tests may be run from cabal.
    + Changed Makefile's 'test' target to run tests via cabal.
    + Removed old generate.sh.
    + Since we no longer have 'sed' to filter out raw HTML sections
      from the docbook writer test, or raw LaTeX sections from the
      context writer test, we now just include these sections.
      They can be taken out if it is necessary to process the files.
    + Updated latex and context writer tests to remove extra spaces
      after '\\item'
    + Added a markdown table reader test.
    + Added markdown-reader-more.txt to test suite, for additional test cases
      for raw ConTeXt environments and more.

  * Compatibility fixes for CPP, Cabal, and haddock:

    + Use CPP in "Extensions" field in pandoc.cabal.
    + Removed use of backslash string continuations in source files.

  * Removed pandoc.cabal.ghc66.  We now require Cabal >= 1.2, GHC >= 6.8,
    base >= 3.

  * Require parsec < 3.
    The compatibility module in parsec 3.0.0 gives far worse performance than
    parsec 2.1.  Eventually pandoc will be upgraded to use the new bytestring
    version of parsec, and then we'll go to parsec 3.0.0.

  * Removed Text.Regex dependencies by rewriting using plain Haskell
    (Text.Pandoc.Writers.RTF, Text.Pandoc.Writers.HTML, Main.hs)

  * Moved Text.Pandoc.Writers.DefaultHeaders -> Text.Pandoc.DefaultHeaders.

  * Makefile:

    + Added 'configure' as dependency of 'uninstall-all'.
      (It uses the Cabal build program.)
    + Makefile:  only use --with-hc-pkg if GHC_PKG is defined.
      Note that Cabal will automatically choose the ghc-pkg appropriate
      for the compiler selected, so normally specifying GHC by itself
      is sufficient.

  * Removed Text.Pandoc.UTF8 module; instead, depend on utf8-string and use
    its IO and conversion functions.

  * Added -Wall to ghc-options in pandoc.cabal.  Cleaned up modules so that
    everything is -Wall clean.

    + Added pragma to HTML writer to avoid deprecation warning for use of "start" attribute.
    + Added pragma to Text/Pandoc/Shared.hs to get rid of "orphan instance" warnings.
      (These are caused by the Lift instance for ByteString.)

  * Changed the comment used to replace unsafe HTML if sanitize-html option
    selected.

  * Made -c/--css option repeatable on the command line (like -H, -A, -B).

  * Moved XML-formatting functions to new unexported module Text.Pandoc.XML.

  * Escape '\160' as "&#160;", not "&nbsp;" in XML.
    "nbsp" isn't a predefined XML entity.

  * Fixed bug in RST reader, which would choke on: "p. one\ntwo\n".
    Added some try's in ordered list parsers.

  * Man writer:  don't escape " as \".

  * Allow newline before URL in markdown link references.  Resolves Issue #81.
    Added tests for this issue in new "markdown-reader-more" tests.
    Changed RunTests.hs to run these tests.

  * Support for display math.  Resolves Issue #47.

    + Added a DisplayMath/InlineMath selector to Math inlines.
    + Markdown parser yields DisplayMath for $$...$$.
    + LaTeX parser yields DisplayMath when appropriate.  Removed
      mathBlock parsers, since the same effect is achieved by the math
      inline parsers, now that they handle display math.
    + Writers handle DisplayMath as appropriate for the format.
    + Modified tests accordingly; added new tests for display math.

  * Use LaTeXMathML instead of ASCIIMathML.  LaTeXMathML is closer
    to LaTeX in its display of math, and supports many non-math LaTeX environments.

    + Changed -m option to use LaTeXMathML rather than ASCIIMathML.
    + Modified HTML writer to print raw TeX when LaTeXMathML is
      being used instead of suppressing it.
    + Removed ASCIIMathML files from data/ and added LaTeXMathML.
    + Replaced ASCIIMathML with LaTeXMathML in source files.
    + Modified README and pandoc man page source.
    + Added --latexmathml option (kept --asciimathml as a synonym
      for backwards compatibility)

  * Markdown reader: Parse setext headers before atx headers.
    Test case:
       # hi
       ====
    parsed by Markdown.pl as an H1 header with contents "# hi".

  * Markdown reader: Treat "mixed" lists the same way as Markdown.pl does.
    The marker on the first list item determines the type of the whole
    list.  Thus, a list like
       1. one
       -  two
       *  three
    gets parsed as a single ordered list.  (Previous versions of pandoc
    treated this as an ordered list with an unordered sublist.)

  * Markdown smart typography:

    + Em dashes no longer eat surrounding whitespace.  Resolves Issue #69.
    + Use nonbreaking spaces after known abbreviations in markdown parser.
      Thus, for example, "Mr. Brown" comes out as "Mr.~Brown" in LaTeX, and does
      not produce a sentence-separating space.  Resolves Issue #75.

  * Markdown writer: Print unicode \160 literally, rather than as &nbsp;.

  * Treat '\ ' in (extended) markdown as nonbreaking space.
    Print nonbreaking space appropriately in each writer (e.g. ~ in LaTeX).

  * The '--sanitize-html' option now examines URIs in markdown links
    and images, and in HTML href and src attributes.  If the URI scheme
    is not on a whitelist of safe schemes, it is rejected.  The main point
    is to prevent cross-site scripting attacks using 'javascript:' URIs.
    See http://www.mail-archive.com/markdown-discuss@six.pairlist.net/msg01186.html
    and http://ha.ckers.org/xss.html.  Resolves Issue #62.

  * HTML writer:

    + Override Text.XHtml's stringToHtml function,
      so that characters below 0xff are not converted to numerical entity
      references. Also convert '\160' to "&nbsp;". This should aid readability
      and editability of the HTML source. It does presuppose that the HTML
      will be served as UTF-8.
    + In code blocks, change leading newlines to <br /> tags.
      (Some browsers ignore them.)  Resolves Issue #71.
      See http://six.pairlist.net/pipermail/markdown-discuss/2008-May/001297.html
    + Use style attributes rather than css classes for strikethrough
      and ordered list styles.  This works better when fragments, rather than
      standalone documents, are generated.

  * HTML reader: Count anything that isn't a known block (HTML) tag as an
    inline tag (rather than the other way around).  Added "html", "head", and
    "body" to list of block tags.  Resolves Issue #66, allowing
    <lj> to count as an inline tag.

  * RTF writer: Fixed bug.  Extra spaces were being printed after emphasized,
    boldface, and other inline elements.  Resolves Issue #64.

  * LaTeX reader:  improvements in raw LaTeX parsing.

    + "loose punctuation" (like {}) parsed as Space
    + Para elements must contain more than Str "" and Space elements
    + Added parser for "\ignore" command used in literate haskell.
    + Reworked unknownCommand and rawLaTeXInline: when not in "parse raw"
      mode, these parsers simply strip off the command part and allow
      the arguments to be parsed normally.  So, for example,
      \blorg{\emph{hi}} will be parsed as Emph "hi" rather than
      Str "{\\emph{hi}}".
    + Parse lhs "code" environments as verbatim.
      Refactored parsers for verbatim environments.
    + Removed specialEnvironment parser.
    + parse '{}', if present, after \textless, \textgreater,
      \textbar, \textbackslash, \ldots.
    + Parse unescaped special characters verbatim rather than
      changing them to spaces.  This way arguments of unknown
      commands will appear in braces.

  * Parse raw ConTeXt environments as TeX in markdown reader.
    Resolves Issue #73.

  * Moved BlockWrapper and wrappedBlocksToDoc from ConTeXt writer to Shared.

  * Made some structural changes to parsing of raw LaTeX environments.
    Previously there was a special block parser for LaTeX environments.
    It returned a Para element containing the raw TeX inline. This has
    been removed, and the raw LaTeX environment parser is now used in the
    rawLaTeXInline parser. The effect is exactly the same, except that we
    can now handle consecutive LaTeX and ConTeXt environments not separated
    by spaces.  This new flexibility is required by the example in
    Issue #73:

        \placeformula \startformula
             L_{1} = L_{2}
        \stopformula

    API change: The LaTeX reader now exports rawLaTeXEnvironment' (which
    returns a string) rather than rawLaTeXEnvironment (which returns a block
    element). This is more likely to be useful in other applications.

  * Use \textsubscr instead of \textsubscript for LaTeX subscript macro.
    \textsubscript conflicts with a definition in the memoir class.
    Resolves Issue #65.

  * Removed unneeded space after "\\item" in LaTeX and ConTeXt output.

  * Added amsmath package to default LaTeX header. Resolves Issue #48.

  * Added \setupitemize[autointro] to ConTeXt header, to prevent orphaned
    list introduction lines.

  * Changed Float to Double in definition of Table element.
    (Double is more efficient in GHC.)

  * Fixed bug in Markdown parser:  regular $s triggering math mode.
    For example:  "shoes ($20) and socks ($5)."
    The fix consists in two new restrictions:

    + the $ that ends a math span may not be directly followed by a digit.
    + no blank lines may be included within a math span.

    Thanks to Joseph Reagle for noticing the bug.

  * Use Data.List's 'intercalate' instead of custom 'joinWithSep'.
    Removed 'joinWithSep' from Text.Pandoc.Shared.

  * Updated README and man pages.  Acknowledge contributors in README.
    Added paragraph to README about producing S5 with separate CSS/javascript.

  * Updated INSTALL to reflect new build system (including configuration
    options) and document new dependencies.  Added note to INSTALL that
    Cabal >= 1.2 is required for build.  Resolves Issue #74.

  * Fixed some haddock documentation errors.

  * Small fix to markdown2pdf man page: only input needs to be piped through iconv.

pandoc (0.46) unstable; urgency=low

  [ John MacFarlane ]

  * Made -H, -A, and -B options cumulative: if they are specified
    multiple times, multiple files will be included.

  * Added optional HTML sanitization using a whitelist.
    When this option is specified (--sanitize-html on the command line),
    unsafe HTML tags will be replaced by HTML comments, and unsafe HTML
    attributes will be removed.  This option should be especially useful
    for those who want to use pandoc libraries in web applications, where
    users will provide the input.

    + Main.hs:  Added --sanitize-html option.

    + Text.Pandoc.Shared:  Added stateSanitizeHTML to ParserState.

    + Text.Pandoc.Readers.HTML:
      - Added whitelists of sanitaryTags and sanitaryAttributes.
      - Added parsers to check these lists (and state) to see if a given
        tag or attribute should be counted unsafe.
      - Modified anyHtmlTag and anyHtmlEndTag to replace unsafe tags
        with comments.
      - Modified htmlAttribute to remove unsafe attributes.
      - Modified htmlScript and htmlStyle to remove these elements if
        unsafe.

    + Modified README and man pages to document new option.

  * Improved handling of email addresses in markdown and reStructuredText.
    Consolidated uri and email address parsers.  (Resolves Issue #37.)

    + New emailAddress and uri parsers in Text.Pandoc.Shared.
      - uri parser uses parseURI from Network.URI.
      - emailAddress parser properly handles email addresses with periods
        in them.

    + Removed uri and emailAddress parsers from Text.Pandoc.Readers.RST
      and Text.Pandoc.Readers.Markdown.

  * Markdown reader:

    + Fixed emph parser so that "*hi **there***" is parsed as a Strong
      nested in an Emph. (A '*' is only recognized as the end of the
      emphasis if it's not the beginning of a strong emphasis.)

    + Moved blockQuote parser before list parsers for performance.

    + Modified 'source' parser to allow backslash-escapes in URLs.
      So, for example, [my](/url\(1\)) yields a link to /url(1).
      Resolves Issue #34.

    + Disallowed links within links. (Resolves Issue #35.)
      - Replaced inlinesInBalanced with inlinesInBalancedBrackets, which
        instead of hard-coding the inline parser takes an inline parser
        as a parameter.
      - Modified reference and inlineNote to use inlinesInBalancedBrackets.
      - Removed unneeded inlineString function.
      - Added inlineNonLink parser, which is now used in the definition of
        reference.
      - Added inlineParsers list and redefined inline and inlineNonLink parsers
        in terms of it.
      - Added failIfLink parser.

    + Better handling of parentheses in URLs and quotation marks in titles.
      - 'source' parser first tries to parse URL with balanced parentheses;
        if that doesn't work, it tries to parse everything beginning with
        '(' and ending with ')'.
      - source parser now uses an auxiliary function source'.
      - linkTitle parser simplified and improved, under assumption that it
        will be called in context of source'.

    + Make 'block' conditional on strictness state, instead of using
      failIfStrict in block parsers. Use a different ordering of parsers
      in strict mode (raw HTML block before paragraph) for performance.
      In non-strict mode use rawHtmlBlocks instead of htmlBlock.
      Simplified htmlBlock, since we know it's only called in strict
      mode.

    + Improved handling of raw HTML.  (Resolves Issue #36.)
      - Tags that can be either block or inline (e.g. <ins>) should
        be treated as block when appropriate and as inline when
        appropriate. Thus, for example,
        <ins>hi</ins>
        should be treated as a paragraph with inline <ins> tags, while
        <ins>
        hi
        </ins>
        should be treated as a paragraph within <ins> tags.
      - Moved htmlBlock after para in list of block parsers.  This ensures
        that tags that can be either block or inline get parsed as inline
        when appropriate.
      - Modified rawHtmlInline' so that block elements aren't treated as
        inline.
      - Modified para parser so that paragraphs containing only HTML tags and
        blank space are not allowed.  Treat these as raw HTML blocks instead.

    + Fixed bug wherein HTML preceding a code block could cause it to
      be parsed as a paragraph.  The problem is that the HTML parser
      used to eat all blank space after an HTML block, including the
      indentation of the code block.  (Resolves Issue #39.)
      - In Text.Pandoc.Readers.HTML, removed parsing of following space
        from rawHtmlBlock.
      - In Text.Pandoc.Readers.Markdown, modified rawHtmlBlocks so that
        indentation is eaten *only* on the first line after the HTML
        block.  This means that in
        <div>
             foo
        <div>
        the foo won't be treated as a code block, but in
        <div>

            foo

        </div>
        it will.  This seems the right approach for least surprise.

  * RST reader:

    + Fixed bug in parsing explicit links (resolves Issue #44).
      The problem was that we were looking for inlines until a '<' character
      signaled the start of the URL; so, if you hit a reference-style link,
      it would keep looking til the end of the document.  Fix:  change
      inline => (notFollowedBy (char '`') >> inline).  Note that this won't
      allow code inlines in links, but these aren't allowed in resT anyway.

    + Cleaned up parsing of reference names in key blocks and links.
      Allow nonquoted reference links to contain isolated '.', '-', '_', so
      so that strings like 'a_b_' count as links.

    + Removed unnecessary check for following link in str.
      This is unnecessary now that link is above str in the definition of
      'inline'.

  * HTML reader:

    + Modified rawHtmlBlock so it parses </html> and </body> tags.
      This allows these tags to be handled correctly in Markdown.
      HTML reader now uses rawHtmlBlock', which excludes </html> and </body>,
      since these are handled in parseHtml.  (Resolves Issue #38.)

    + Fixed bug (emph parser was looking for `<IT>` tag, not `<I>`).

    + Don't interpret contents of style tags as markdown.
      (Resolves Issue #40.)
      - Added htmlStyle, analagous to htmlScript.
      - Use htmlStyle in htmlBlockElement and rawHtmlInline.
      - Moved "script" from the list of tags that can be either block or
        inline to the list of block tags.

    + Modified rawHtmlBlock to use anyHtmlBlockTag instead of anyHtmlTag
      and anyHtmlEndTag.  This fixes a bug in markdown parsing, where
      inline tags would be included in raw HTML blocks.

    + Modified anyHtmlBlockTag to test for (not inline) rather than
      directly for block.  This allows us to handle e.g. docbook in
      the markdown reader.

  * LaTeX reader:  Properly recognize --parse-raw in rawLaTeXInline.
    Updated LaTeX reader test to use --parse-raw.

  * HTML writer:

    + Modified rules for automatic HTML header identifiers to
      ensure that identifiers begin with an alphabetic character.
      The new rules are described in README.  (Resolves Issue #33.)

    + Changed handling of titles in HTML writer so you don't get
      "titleprefix - " followed by nothing.

  * ConTeXt writer: Use wrappers around Doc elements to ensure proper
    spacing.  Each block element is wrapped with either Pad or Reg.
    Pad'ed elements are guaranteed to have a blank line in between.

  * RST writer:

    + Refactored RST writer to use a record instead of a tuple for state,
      and to include options in state so it doesn't need to be passed as
      a parameter.

    + Use an interpreted text role to render math in restructuredText.
      See http://www.american.edu/econ/itex2mml/mathhack.rst for the
      strategy.

  [ Recai Oktaş ]

  * Debian packaging changes:

    + Remove the empty 'include' directory in -dev package, which lintian
      complains about.
    + Bump Standarts-Version to 3.7.3.
    + Use new 'Homepage:' field to specify the upstream URL on suggestion of
      lintian.

 -- Recai Oktaş <roktas@debian.org>  Tue, 08 Jan 2008 05:13:31 +0200

pandoc (0.45) unstable; urgency=low

  [ John MacFarlane ]

  * Simplified parsing of reference keys and notes in markdown and RST
    readers: The Reference data structure from Text.Pandoc.Shared is no
    longer needed, since referenceKey and noteBlock parses return strings
    (as many blank lines as were occupied by the key or note) and update
    state themselves.  getPosition and setPosition are now used to ensure
    that error messages will give the correct line number. This yields
    cleaner (and slightly faster) code, with more accurate parsing error
    messages.

  * Added new Math inline element:

    + Markdown and LaTeX readers now convert TeX math into Math elements,
      not TeX.
    + This allows math to be treated differently from raw TeX in output.
      TeX elements are no longer printed in output formats other than
      Markdown, LaTeX, and ConTeXt.  But Math elements are always printed.

  * New default handling of math in writers:

    + New module Text.Pandoc.Readers.TeXMath exports readTeXMath, which
      parses raw TeX math and outputs a string of Pandoc inlines that
      tries to render it as far as possible using unicode characters,
      lapsing into literal TeX when needed.
    + readTeXMath is now used for default HTML output in HTML, S5, RTF,
      and Docbook, if no other method for displaying math in HTML is
      specified.  Enclosing $'s are no longer printed by default.
    + By default, math is put inside `<span class="math">`. This way it can be
      distinguished from the surrounding text, e.g. put in a different
      font.

  * New --gladtex and --mimetex options for display of math in HTML:

    + If --gladtex is specified, math is output between `<eq>` tags, so
      it can be processed by gladTeX.
    + If --mimetex is specified, math is put in `<img>` tags with a link
      to the mimetex CGI script (or any other script that takes TeX math
      as input and outputs an image).  The URL of the script may be
      specified, but defaults to /cgi-bin/mimetex.cgi.
    + HTMLMathMethod structure in WriterOptions keeps track of how to
      display math in HTML output.
    + Updated README with a description of the four options for displaying
      math in HTML.

  * HTML reader:

    + Fixed bug: parser for minimized attributes should not swallow
      trailing spaces.
    + Simplified HTML attribute parsing.
    + Changed parsing of code blocks in HTML reader:  `<code>` tag is no
      longer needed. `<pre>` suffices. All HTML tags in the code block
      (e.g. for syntax highlighting) are skipped, because they are not
      portable to other output formats. A `<code>...</code>` block not
      surrounded by `<pre>` now counts as inline HTML, not a code block.
    + Remove just one leading and one trailing newline from contents of
      `<pre>...</pre>` in codeBlock parser.

  * Markdown reader:

    + Removed support for box-style block quotes.
    + Require space before title in links and references.
      This fixes a bug in parsing URLs like http://silly/url(withparen).
    + Improved and simplified setextHeader parser.
    + Fixed logic in smart quote parsing, adding some needed 'try'
      statements.
    + Fixed smart quote parsing so that unicode characters 8216 and 8217
      are recognized as single quotes, and 8220 and 8221 as double quotes.

  * RST reader:

    + Fixed bug in parsing of code blocks. Previously a full tab indent
      was required, but RST allows code to be indented any amount.
      Resolves Issue #27.
    + Allow field lists to be indented.
    + Parse the contents of field lists instead of treating as a raw string.
    + Represent field lists as definition lists instead of blockquotes.
    + Fixed bug in which metadata would be overridden if the document
      contained multiple field lists.
    + Parse fields associated with '.. image::' blocks, and use 'alt'
      field, if given, for image alt and title attributes.

  * LaTeX reader:

    + Modified specialChar so that '"' characters are parsed.
    + Fixed a bug in parsing of \[ \] math blocks (thanks to Mark Kalderon).

  * HTML writer:

    + Changes in handling of math (see above).
    + Don't produce HTML for table of contents if there are
      no headers. (This would be an empty list, which is invalid XHTML.)

  * Markdown writer:

    + Don't print title attribute if title is empty. (This differs from
      the behavior of Markdown.pl, and agrees with PHP Markdown. But John
      Gruber has indicated that he prefers this behavior.) Adjusted test
      suite accordingly.
    + Fixed incorrect line wrapping in paragraphs including hard line
      breaks. Resolves Issue #25.
    + Fixed bug in markdown writer:  If an ordered list item began with
      a marker greater than 3 characters in width, and the item took more
      than one line, it would appear on the line after the list marker,
      e.g.:

          (12)
              My list item.
              Multiline.

      Now it works as follows:

          (12) My list item.
              Multiline.

  * RST writer

    + Fixed bug in RST writer's handling of ordered lists. Previously,
      list items with multiple lines would not always line up with
      single-line list items. Now, list items are nested the length of
      the list marker + 1. This looks better and ensures that list items
      all line up. (Note that list markers are padded to the length of
      the longest list marker in the series.)
    + Use 3-space indent for unordered lists.
    + If label for a link reference contains a colon, surround it by `
      signs so it won't be interpreted as the end of the link label.

  * LaTeX writer:

    + Cleaner output for footnotes. Footnotes now always begin on a new
      line, and the final } is on a line by itself only when it needs to
      be (i.e. only when the note ends with a Verbatim environment).
    + Added writer options to state, so state doesn't need to be passed as
      a parameter.
    + Text wrapping now provided, using wrapTeXIfNeeded.

  * ConTeXt writer:  many improvements for more idiomatic ConTeXt output
    (thanks to Idris Samawi Hamid for suggestions).

    + PrettyPrint module now used for output.
    + Writer options are now in state, so they don't have to be passed as
      a parameter.
    + Text wrapping now provided, using wrapTeXIfNeeded.
    + Better treatment of footnotes:  footnotes are always on lines by
      themselves, and the final } is on a line by itself only when
      it needs to be (after \stoptyping).
    + Use \subject, \subsubject, ... or \section, \subsection, ... for headings,
      depending on whether --number-sections option is selected.
    + Extra blank line inserted after \stopitemize
    + Use new, "official" definition of blockquote environment. Also, use
      blank line after \startblockquote to balance blank line at the end.
    + Both itemized and enumerated lists are now generated using
      \start-stopitemize, with appropriate options. Removed definitions
      of ltxenum and ltxitem, which are no longer needed. Provided
      defaults for itemized lists in the preamble. State keeps track of
      ordered list level, so that when default numbering is specified,
      the appropriate scheme can be used.
    + Changed \useurl to \useURL.
    + Changed link color from red to blue.
    + Use \subsubsubsubsection etc., since these are supported
      (up to at least sub x 5).

  * Text.Pandoc.Shared:

    + Save and restore position in parseFromString, so that accurate
      error messages can be given.
    + Improved efficiency of romanNumeral parser.
    + Added wrappedTeX and wrapTeXIfNeeded functions.  These ensure
      that footnotes occur on lines by themselves (to make them
      easier to see and move) and do not screw up line wrapping.

  * Text.Pandoc.UTF8:  modified fromUTF8 to strip out the BOM
    if present.  Windows Notepad and other applications insert a
    BOM at the beginning of a UTF8 file.

  * Main.hs (tabFilter): Treat '\r' at end of line as newline (in
    addition to "\r\n" and '\n').

  * Added a writer option for wrapped text and a command-line option
    '--no-wrap', which disables text wrapping and minimizes whitespace
    in HTML. (Resolves Issue #26.)

    + Added support for '--no-wrap' to Main.hs.
    + Added wrapIfNeeded function to Text.Pandoc.Shared.
    + Use wrapIfNeeded instead of wrapped in the RST, Man, Docbook, and
      Markdown writers.
    + Added render and renderFragment helpers to HTML writer.

  * Modified html2markdown to run tidy only if the HTML cannot be
    parsed. Previously html2markdown piped all input through tidy
    before passing it to pandoc. This caused problems on certain pages
    (e.g. http://daringfireball.com/markdown) which have well-formed
    XHTML that causes tidy to choke. The solution is to pipe through
    tidy only if pandoc cannot parse the input by itself. This means
    that a temp file is now always used, even when input comes from a
    local file or standard input.

  * Removed 'version' constant from Main.hs; added 'pandocVersion' to
    Text.Pandoc library.

  * pandoc.cabal:

    + Modified to work with GHC 6.8 and Cabal configurations. (For GHC
      6.8, pretty and containers must be added to Build-Depends, and it
      is desirable to use the -O2 compiler option.)  Cabal configurations
      allows one to select options depending on the compiler version.
      For GHC 6.6, the splitBase option can be disabled.
    + pandoc.cabal.ghc66 is provided for users with older versions of
      Cabal, which do not support configurations.
    + Use Ghc-Prof-Options to ensure that '-auto-all' is used when
      '--enable-(executable|library)-profiling' is specified. Updated
      PROFILING instructions accordingly.

  * Makefile:

    + Makefile now checks GHC version. If GHC is 6.6, pandoc.cabal.ghc66
      is copied to pandoc.cabal, and the old pandoc.cabal is copied
      to pandoc.cabal.orig. Otherwise, pandoc.cabal is copied to
      pandoc.cabal.orig but otherwise unmodified. This way, the Makefile
      will work properly with either GHC 6.6 or 6.8.
    + Changed BUILDCONF to point to dist/setup-config, not .setup-config.
      This is where current versions of Cabal put it.
    + Added $(BUILDCMD) target, so setup doesn't get compiled every time.
    + Removed dependency of templates on ./templates, which is circular
      now that templates is a subdirectory of the top-level.

  * MacPorts Portfile:

    + Modified to install the pandoc library in addition to programs.
    + Installation must be done manually rather than using Makefile's
      install-all.
    + Note that the library must be registered in the activate phase,
      after the library files have been copied out of the destroot.
      Cabal generates a 'register.sh' script that will do this.

  * debian/control: Added libghc6-network-dev, libghc6-xhtml-dev, and
    libghc6-mtl-dev as dependencies for libghc6-pandoc-dev.
    Closes: #445235

  * debian/rules:  Converted to UTF-8.

  * Changed pandoc home page to http://johnmacfarlane.net/pandoc/.

  * Updated ASCIIMathML.js to latest version.

  * Directory structure:

    + Moved everything from src into the top-level directory.
    + Changed references to source directory in Makefile and
      pandoc.cabal.*.
    + Moved ASCIIMathML.js, headers, and ui into templates directory.
    + Modified fillTemplates.pl to reflect new paths.

  [ Recai Oktaş ]

  * Makefile: Fixed the issue of having two copies of the library
    documentation under some usage scenarios.

  * Replaced 'ghc' with '$(GHC)' in Makefile, and made GHC
    and GHC_PKG configurable through the environment, to support
    unusual ghc installations.  For example:
    GHC=/opt/ghc/bin/ghc GHC_PKG=/opt/ghc/bin/ghc-pkg make

 -- Recai Oktaş <roktas@debian.org>  Sun, 07 Oct 2007 20:51:43 +0300

pandoc (0.44) unstable; urgency=low

  [ John MacFarlane ]

  * Fixed bug in HTML writer:  when --toc was used, anchors were put around
    headers, which is invalid XHTML (block content within inline element).
    Now the anchors are put inside the header tags.  Resolves Issue #23.

  * Added xmlns attribute to html element in html writer tests.
    This attribute is added by more recent versions of the
    xhtml library (>= 3000), and is required for valid XHTML.

  [ Recai Oktaş ]

  * On configure, compile 'Setup.hs' to 'setup' and use 'setup' as the build
    command instead of 'runhaskell', which, on some platforms (such as s390,
    alpha, m68k), throws the following error:

        runhaskell Setup.hs configure --prefix=/usr
        ghc-6.6.1: not built for interactive use

    This causes a serious FTBFS bug.  Closes: #440668.

 -- Recai Oktaş <roktas@debian.org>  Mon, 03 Sep 2007 18:24:02 +0300

pandoc (0.43) unstable; urgency=low

  [ John MacFarlane ]

  * The focus of this release is performance.  The markdown parser
    is about five times faster than in 0.42, based on benchmarks
    with the TextMate manual.

  * Main.hs: Replaced CRFilter and tabFilter with single function
    tabFilter, which operates on the whole string rather than breaking
    it into lines, and handles dos-style line-endings as well as tabs.

  * Added separate LaTeX reader and native reader tests; removed
    round-trip tests.

  * Text.Pandoc.Shared:

    + Removed tabsToSpaces and tabsInLine (they were used only in Main.hs.)
    + General code cleanup (to elimante warnings when compiling with -Wall.)
    + Added 'wrapped' function, which helps wrap text into paragraphs,
      using the prettyprinting library.
    + Rewrote charsInBalanced and charsInBalanced'.
        - Documented restriction: open and close must be distinct characters.
        - Rearranged options for greater efficiency.
        - Bug fix: Changed inner call to charsInBalanced inside
          charsInBalanced' to charsInBalanced'.
    + anyLine now requires that the line end with a newline (not eof).
      This is a harmless assumption, since we always add newlines to the
      end of a block before parsing with anyLine, and it yields a 10% speed
      boost.
    + Removed unnecessary 'try' in anyLine.
    + Removed unneeded 'try' from romanNumeral parser.
    + Use notFollowedBy instead of notFollowedBy' in charsInBalanced.
    + Removed unneeded 'try' in parseFromString.
    + Removed unneeded 'try' from stringAnyCase.  (Now it behaves
      like 'string'.)
    + Changed definition of 'enclosed' in Text.Pandoc.Shared so that
      'try' is not automatically applied to the 'end' parser. Added
      'try' in calls to 'enclosed' where needed. Slight speed increase.

  * Writers:

    + Replaced individual wrapping routines in RST, Man, and Markdown
      writers with 'wrapped' from Text.Pandoc.Shared.
    + Rewrote LaTeX writer to use the prettyprinting library,
      so we get word wrapping, etc.
    + Modified latex writer tests for new latex writer using prettyprinter.
    + Fixed bug in LaTeX writer: autolinks would not cause
      '\usepackage{url}' to be put in the document header. Also, changes
      to state in enumerated list items would be overwritten.
    + In Markdown writer, escape paragraphs that begin with ordered list
      markers, so they don't get interpreted as ordered lists.

  * Text.Pandoc.Reades.LaTeX:

    + Fixed bug in LaTeX reader, which wrongly assumed that the roman
      numeral after "enum" in "setcounter" would consist entirely of
      "i"s. 'enumiv' is legitimate.
    + LaTeX command and environment names can't contain numbers.
    + Rearranged order of parsers in inline for slight speed improvement.
    + Added '`' to special characters and 'unescapedChar'.

  * Text.Pandoc.Readers.RST:

    + Removed unneeded try's in RST reader; also minor code cleanup.
    + Removed tabchar.
    + Rearranged parsers in inline (doubled speed).

  * Text.Pandoc.Readers.Markdown:

    + Skip notes parsing if running in strict mode. (This yields a nice
      speed improvement in strict mode.)
    + Simplify autolink parsing code, using Network.URI to test for
      URIs. Added dependency on network library to debian/control and
      pandoc.cabal.
    + More perspicuous definition of nonindentSpaces.
    + Removed unneeded 'try' in 'rawLine'.
    + Combined linebreak and whitespace into a new whitespace parser, to
      avoid unnecessary reparsing of space characters.
    + Removed unnecessary 'try' in 'codeBlock', 'ellipses', 'noteMarker',
      'multilineRow', 'dashedLine', 'rawHtmlBlocks'.
    + Use lookAhead in parsers for setext headers and definition lists
      to see if the next line begins appropriately; if not, don't waste
      any more time parsing.
    + Don't require blank lines after code block. (It's sufficient to
      end code block with a nonindented line.)
    + Changed definition of 'emph': italics with '_' must not
      be followed by an alphanumeric character. This is to help
      prevent interpretation of e.g. `[LC_TYPE]: my_type` as
      `[LC<em>TYPE]:my</em>type`.
    + Improved Markdown.pl-compatibility in referenceLink: the two parts
      of a reference-style link may be separated by one space, but not
      more... [a] [link], [not]    [a link].
    + Fixed markdown inline code parsing so it better accords with
      Markdown.pl: the marker for the end of the code section is a clump
      of the same number of `'s with which the section began, followed
      by a non-` character. So, for example,
         ` h ``` i ` -> `<code>h ``` i</code>`.
    + Split 'title' into 'linkTitle' and 'referenceTitle', since the
      rules are slightly different.
    + Rewrote 'para' for greater efficiency.
    + Rewrote link parsers for greater efficiency.
    + Removed redundant 'referenceLink' in definition of inline (it's
      already in 'link').
    + Refactored escapeChar so it doesn't need 'try'.
    + Refactored hrule for performance in Markdown reader.
    + More intelligent rearranging of 'inline' so that most frequently
      used parsers are tried first.
    + Removed tabchar parser, as whitespace handles tabs anyway.

  * Text.Pandoc.CharacterReferences:

    + Refactored.
    + Removed unnecessary 'try's for a speed improvement.
    + Removed unnecessary '&' and ';' from the entity table.

  * Build process:

    + Makefile: Get VERSION from cabal file, not Main.hs.
    + Modified MacPorts Portfile:
        - Depend on haddock
        - Build and install libraries and library documentation in
          addition to pandoc executable
        - Added template item for md5 sum in Portfile.in.
        - Incorporated changes from MacPorts repository (r28278).
    + FreeBSD port:  Don't try to generate distinfo in Makefile.
      It can be made using 'make makesum' in FreeBSD.
    + Make both freebsd and macports targets depend on tarball.

  * Website and documentation:

    + Updated INSTALL instructions.
    + Added pandocwiki demo to website.
    + Removed local references to Portfile, since pandoc is now in the
      MacPorts repository.

 -- Recai Oktaş <roktas@debian.org>  Sun, 02 Sep 2007 15:50:11 +0300

pandoc (0.42) unstable; urgency=low

  [ John MacFarlane ]

  * Main.hs: Use utf8 conversion on the extra files loaded with
    the -H, -C, -B, and -A options.  This fixes problems with unicode
    characters in these files.

  * Exposed Text.Pandoc.ASCIIMathML, since it is imported in
    Text.Pandoc.Readers.HTML and without it we get a linking error when
    using the library.

  * Markdown reader:

    + Added new rule for enhanced markdown ordered lists: if the list
      marker is a capital letter followed by a period (including a
      single-letter capital roman numeral), then it must be followed by
      at least two spaces. The point of this is to avoid accidentally
      treating people's initials as list markers: a paragraph might begin,
      "B. Russell was an English philosopher," and this shouldn't be
      treated as a list.  Documented change in README.
    + Blocks that start with "p. " and a digit are no longer treated
      as ordered lists (it's a page number).
    + Added a needed 'try' to listItem.
    + Removed check for a following setext header in endline.
      A full test is too inefficient (doubles benchmark time), and the
      substitute we had before is not 100% accurate.
    + Don't use Code elements for autolinks if --strict specified.

  * LaTeX writer:  When a footnote ends with a Verbatim environment, the
    close } of the footnote cannot occur on the same line or an error occurs.
    Fixed this by adding a newline before the closing } of every footnote.

  * HTML writer:
    + Removed incorrect "{}" around style information in HTML tables.
      Column widths now work properly in HTML.
    + If --strict option is specified (and --toc is not), don't include
      identifiers in headers, for better Markdown compatibility.

  * Build process:

    + Separated $(web_dest) and website targets.
    + In website, index.txt is now constructed from template index.txt.in.
    + Added freebsd target to Markefile. This creates the freebsd Makefile
      from Makefile.in, and creates distinfo.  Removed Makefile and distinfo
      from the repository.
    + Added macport target to Makefile. Portfile is built from template
      Portfile.in.
    + Removed OSX package targets.  (Too many difficulties involving
      dependencies on dynamic libraries.)
    + More complete INSTALL instructions for all architectures.

  * Website:
    + Added a programming demo, pandocwiki.

  [ Recai Oktaş ]

  * Do not forget to close pandoc's ITP.  Closes: #391666

 -- Recai Oktaş <roktas@debian.org>  Sun, 26 Aug 2007 22:51:32 +0300

pandoc (0.41) unstable; urgency=low

  [ John MacFarlane ]

  * Fixed bugs in HTML reader:
    + Skip material at end *only if* `</html>` is present (previously,
      only part of the document would be parsed if an error was
      found; now a proper error message is given).
    + Added new constant eitherBlockOrInline with elements that may
      count either as block-level or as inline. Modified isInline and
      isBlock to take this into account.
    + Modified rawHtmlBlock to accept any tag (even an inline tag):
      this is innocuous, because rawHtmlBlock is tried only if a regular
      inline element can't be parsed.
    + Added a necessary 'try' in definition of 'para'.

  * Fixed bug in markdown ordered list parsing.  The problem was that
    anyOrderedListStart did not check for a space following the
    ordered list marker.  So in 'A.B. 2007' the parser would be
    expecting a list item, but would not find one, causing an error.
    Fixed a similar bug in the RST reader.  Resolves Issue #22.

  * Refactored RST and Markdown readers using parseFromString.

  * LaTeX reader will now skip anything after \end{document}.

  * Fixed blockquote output in markdown writer: previously, block
    quotes in indented contexts would be indented only in the first
    line.

  * Added note to INSTALL about variations in versions of the xhtml
    library that can lead to failed tests (thanks to Leif LeBaron).

 -- Recai Oktaş <roktas@debian.org>  Sun, 19 Aug 2007 23:26:07 +0300

pandoc (0.4) unstable; urgency=low

  [ John MacFarlane ]

  * Added two new output formats: groff man pages and ConTeXt. By
    default, output files with extensions ".ctx" and ".context" are
    assumed to be ConTeXt, and output files with single-digit extensions
    are assumed to be man pages.

  * Enhanced ordered lists (documented in README, under Lists):
    + The OrderedList block element now stores information about
      list number style, list number delimiter, and starting number.
    + The readers parse this information when possible.
    + The writers use this information to style ordered lists.
    + The enhancement can be disabled using the --strict option.

  * Added support for tables (with a new Table block element). Two kinds
    of tables are supported: a simple table with one-line rows, and a
    more complex variety with multiline rows. All output formats are
    supported, but only markdown tables are parsed at the moment. The
    syntax is documented in README.

  * Added support for definition lists (with a new DefinitionList block
    element). All output and input formats are supported. The syntax is
    documented in README.

  * Added support for superscripts and subscripts (with new Superscript
    and Subscript inline elements).  All input and output
    formats.  The syntax is documented in README.

  * Added support for strikeout (with a new Strikeout inline element).
    All input and output formats are supported. Thanks to Bradley Kuhn,
    who contributed a patch. The syntax is documented in README. Resolves
    Issue #18.

  * Added a --toc|--table-of-contents option.  This causes an automatically
    generated table of contents (or an instruction that creates one) to
    be inserted at the beginning of the document. Not supported in S5,
    DocBook, or man page writers.

  * Modified the -m|--asciimathml option:

    + If an optional URL argument is provided, a link is inserted
      instead of the contents of the ASCIIMathML.js script.
    + Nothing is inserted unless the document actually contains
      LaTeX math.

  * Removed Blank block element as unnecessary.

  * Removed Key and Note blocks from the Pandoc data structure. All
    links are now stored as explicit links, and note contents are
    stored with the (inline) notes.

    + All link Targets are now explicit (URL, title) pairs; there
      is no longer a 'Ref' target.
    + Markdown and RST parsers now need to extract data from key and
      note blocks and insert them into the relevant inline elements.
      Other parsers have been simplified, since there is no longer any need
      to construct separate key and note blocks.
    + Markdown, RST, and HTML writers need to construct lists of
      notes; Markdown and RST writers need to construct lists of link
      references (when the --reference-links option is specified); and
      the RST writer needs to construct a list of image substitution
      references. All writers have been rewritten to use the State monad
      when state is required.
    + Several functions (generateReference, keyTable,
      replaceReferenceLinks, replaceRefLinksBlockList, and some auxiliaries
      used by them) have been removed from Text.Pandoc.Shared, since
      they are no longer needed. New functions and data structures
      (Reference, isNoteBlock, isKeyBlock, isLineClump) have been
      added. The functions inTags, selfClosingTag, inTagsSimple, and
      inTagsIndented have been moved to the DocBook writer, since that
      is now the only module that uses them. NoteTable is now exported
      in Text.Pandoc.Shared.
    + Added stateKeys and stateNotes to ParserState; removed stateKeyBlocks,
      stateKeysUsed, stateNoteBlocks, stateNoteIdentifiers, stateInlineLinks.
    + Added writerNotes and writerReferenceLinks to WriterOptions.

  * Added Text.Pandoc module that exports basic readers, writers,
    definitions, and utility functions. This should export everything
    needed for most uses of Pandoc libraries. The haddock documentation
    includes a short example program.

  * Text.Pandoc.ASCIIMathML is no longer an exported module.

  * Added Text.Pandoc.Blocks module to help in printing markdown
    and RST tables.  This module provides functions for working with
    fixed-width blocks of text--e.g., placing them side by side, as
    in a table row.

  * Refactored to avoid reliance on Haskell's Text.Regex library, which
    (a) is slow, and (b) does not properly handle unicode.  This fixed
    some strange bugs, e.g. in parsing S-cedilla, and improved performance.

    + Replaced 'gsub' with a general list function  'substitute'
      that does not rely on Text.Regex.
    + Rewrote extractTagType in HTML reader so that it doesn't use
      regexs.
    + In Markdown reader, replaced email regex test with a custom email
      autolink parser (autoLinkEmail). Also replaced selfClosingTag regex
      with a custom function isSelfClosingTag.
    + Modified Docbook writer so that it doesn't rely on Text.Regex for
      detecting 'mailto' links.
    + Removed escapePreservingRegex and reamped entity-handling
      functions in Text.Pandoc.Shared and Text.Pandoc.CharacterReferences to
      avoid reliance on Text.Regex (see below on character reference
      handling changes).

  * Renamed Text.Pandoc.Entities as Text.Pandoc.CharacterReferences.

  * Changed handling of XML entities.  Entities are now parsed (and unicode
    characters returned) in the Markdown and HTML readers, rather than being
    handled in the writers.  In HTML and Docbook writers, UTF-8 is now used
    instead of entities for characters above 128.  This makes the HTML and
    DocBook output much more readable and more easily editable.

    + Removed sgmlHexEntity, sgmlDecimalEntity, sgmlNamedEntity, and
      sgmlCharacterEntity regexes from Text.Pandoc.Shared.
    + Renamed escapeSGMLChar to escapeCharForXML.  Added escapeStringForXML.
      Moved both functions to Text.Pandoc.Writers.Docbook.
    + Added characterReference parser to Text.Pandoc.CharacterReferences.
      This parses a string and return a unicode character.
    + Rewrote decodeCharacterReferences to use the new parser instead of
      Text.Regex.
    + Added new charRef parser for Markdown and HTML, which replaces the
      old 'entity' parser. Added '&' as a special character in Markdown reader.
    + Modified HTML and Markdown readers to call decodeEntities on all raw
      strings (e.g. authors, dates, link titles), to ensure that no
      unprocessed entities are included in the native representation of
      the document.  (In the HTML reader, most of this work is done by a
      change in extractAttributeName.)
    + In XML and Markdown output, escape unicode nonbreaking space as '&nbsp;',
      since a unicode non-breaking space is impossible to distinguish visually
      from a regular space.  (Resolves Issue #3.)
    + Removed encodeEntitiesNumerical.
    + Use Data.Map for entityTable and (new) reverseEntityTable, for a
      slight performance boost over the old association list.
    + Removed unneeded decodeEntities from 'str' parser in HTML and
      Markdown readers.

  * Text.Pandoc.UTF8:  Renamed encodeUTF8 to toUTF8, decodeUTF8 to
    fromUTF8, for clarity.

  * Replaced old haskell98 module names replaced by hierarchical module
    names, e.g. List by Data.List.  Removed haskell98 from dependencies
    in pandoc.cabal, and added mtl (needed for state monad). Substituted
    xhtml for html.

  * Refactored and cleaned up character escaping in writers, using
    backslashEscapes and escapeStringUsing functions.

  * Instead of adding "\n\n" to the end of an input string in Main.hs,
    this is now done in the readers. This makes the libraries behave
    the way you'd expect from the pandoc program. Resolves Issue #10.

  * URLs and email addresses in autolinks are now typeset as Code.

  * In Main.hs, changed putStr to putStrLn -- mainly because MacOS X
    doesn't display the whole output unless there's a line ending.

  * Major code cleanup in all modules, for greater consistency, concision,
    and readability.

  * HTML reader:

    + Fixed several bugs (extractTagType, attribute parsing).
    + Remove Null blocks in lists of blocks when possible.
    + Allow HTML comments as raw HTML inline.

  * Markdown reader:

    + Ordered list items may no longer begin with uppercase letters, or
      letters greater than 'n'.  (This prevents first initials and page
      reference, e.g. 'p. 400', from being parsed as beginning lists.)
      Also, numbers beginning list items may no longer end with ')',
      which is now allowed only after letters.  Note: These changes
      may cause documents to be parsed differently. Users should take
      care in upgrading.
    + Changed autoLink parsing to conform better to Markdown.pl's
      behavior. `<google.com>` is not treated as a link, but
      `<http://google.com>`, `<ftp://google.com>`, and
      `<mailto:google@google.com>` are.
    + Cleaned up handling of embedded quotes in link titles.  Now these are
      stored as a '"' character, not as '&quot;'.
    + Use lookAhead parser for the 'first pass' (looking for reference keys),
      instead of parsing normally, then using setInput to reset input.  This
      yields a slight performance boost.
    + Fixed several bugs in smart quote recognition.
    + Fixed bug in indentSpaces (which didn't properly handle
      cases with mixed spaces and tabs).
    + Consolidated 'text', 'special', and 'inline' into 'inline'.
    + Fixed bug which allowed URL and title to be separated by multiple blank
      lines in links and reference keys.  They can be on separate lines but
      can't have blank lines between them.
    + Correctly handle bracketed text inside inline footnotes and links,using
      new function inlinesInBalanced.  Resolves Issue #14.
    + Fixed bug in footnotes: links in footnotes were not being
      processed. Solution: three-stage parse. First, get all the
      reference keys and add information to state. Next, get all the
      notes and add information to state. (Reference keys may be needed
      at this stage.) Finally, parse everything else.
    + Replaced named constants like 'emphStart' with literals.
    + Removed an extra occurance of escapedChar in definition of inline.

  * RST reader:

    + Allow the URI in a RST hyperlink target to start on the line
      after the reference key.
    + Added 'try' in front of 'string', where needed, or used a different
      parser.  This fixes a bug where ````` would not be correctly parsed as
      a verbatim `.
    + Fixed slow performance in parsing inline literals in RST reader.  The
      problem was that ``#`` was seen by 'inline' as a potential link or image.
      Fix:  inserted 'notFollowedBy (char '`')' in link parsers.
      Resolves Issue #8.
    + Use lookAhead instead of getInput/setInput in RST reader.  Removed
      unneeded getState call, since lookAhead automatically saves and
      restores the parser state.
    + Allow hyperlink target URIs to be split over multiple lines, and
      to start on the line after the reference. Resolves Issue #7.
    + Fixed handling of autolinks.

  * LaTeX reader:

    + Replaced 'choice [(try (string ...), ...]' idiom with 'oneOfStrings',
      for clarity.
    + Added clauses for tilde and caret. Tilde is \ensuremath{\sim}, and
      caret is \^{}, not \^ as before.
    + Added parsing for \url.
    + Parse \texttt{} as code, provided there's nothing fancy inside.

  * HTML writer:

    + Modified HTML writer to use the Text.XHtml library. This results
      in cleaner, faster code, and it makes it easier to use Pandoc in
      other projects, like wikis, which use Text.XHtml. Two functions are
      now provided, writeHtml and writeHtmlString: the former outputs an
      Html structure, the latter a rendered string. The S5 writer is also
      changed, in parallel ways (writeS5, writeS5String).
    + The Html header is now written programmatically, so it has been
      removed from the 'headers' directory. The S5 header is still
      needed, but the doctype and some of the meta declarations have
      been removed, since they are written programatically. This change
      introduces a new dependency on the xhtml package.
    + Fixed two bugs in email obfuscation involving improper escaping
      of '&' in the `<noscript>` section and in `--strict` mode. Resolves
      Issue #9.
    + Fixed another bug in email obfuscation: If the text to be obfuscated
      contains an entity, this needs to be decoded before obfuscation.
      Thanks to thsutton for the patch. Resolves Issue #15.
    + Changed the way the backlink is displayed in HTML footnotes.
      Instead of appearing on a line by itself, it now generally
      appears on the last line of the note.  (Exception:  when the
      note does not end with a Plain or Para block.) This saves space
      and looks better.
    + Added automatic unique identifiers to headers:
      - The identifier is derived from the header via a scheme
        documented in README.
      - WriterState now includes a list of header identifiers and a table
        of contents in addition to notes.
      - The function uniqueIdentifiers creates a list of unique identifiers
        from a list of inline lists (e.g. headers).
      - This list is part of WriterState and gets consumed by blockToHtml
        each time a header is encountered.
    + Include CSS for .strikethrough class in header only if strikethrough
      text appears in the document.
    + If the 'strict' option is specified, elements that do not appear in
      standard markdown (like definition lists) are passed through as
      raw HTML.
    + Simplified treatment of autolinks, using pattern matching instead of
      conditionals.

  * Markdown writer:

    + Links in markdown output are now printed as inline links by default,
      rather than reference links.  A --reference-links option has been added
      that forces links to be printed as reference links.  Resolves Issue #4.
    + Use autolinks when possible.  Instead of `[site.com](site.com)`,
      use `<site.com>`.

  * LaTeX writer:

    + Rewrote to use the State monad. The preamble now includes only those
      packages that are actually required, given the document's content.
      Thus, for example, if strikeout is not used, ulem is not required.
      Modified LaTeXHeader accordingly.
    + Modified LaTeX writer to insert '\,' between consecutive quotes.
    + Removed unused function tableRowColumnWidths.
    + Simplified code for escaping special characters.
    + Leave extra blank line after \maketitle.
    + Include empty '\author{}' when no author specified to avoid LaTeX
      errors.
    + Include fancyvrb code in header only if needed -- that is, only
      if there is actually code in a footnote.
    + Use \url{} for autolinks.
    + Include [mathletters] option in ucs package, so that basic unicode
      Greek letters will work correctly.

  * RST writer:  Force blank line before lists, so that sublists will
    be handled correctly.

  * Docbook writer:  Fixed a bug:  email links with text, like
    [foo](me@bar.baz), were being incorrectly treated as autolinks.

  * Removed Text.ParserCombinators.Pandoc and moved all its functions to
    Text.Pandoc.Shared.

  * Text.Pandoc.Shared:

    + Added defaultWriterOptions.
    + Added writerTableOfContents to WriterOptions.
    + Added writerIgnoreNotes option to WriterOptions.  This is needed
      for processing header blocks for a table of contents, since notes on
      headers should not appear in the TOC.
    + Added prettyprinting for native Table format.
    + Removed some unneeded imports.
    + Moved escape and nullBlock parsers from
      Text.ParserCombinators.Pandoc, since the latter is for
      general-purpose parsers that don't depend on Text.Pandoc.Definition.
    + Moved isHeaderBlock from Text.Pandoc.Writers.HTML.
    + Moved Element, headerAtLeast, and hierarchicalize from Docbook
      writer, because HTML writer now uses these in constructing a table
      of contents.
    + Added clauses for new inline elements (Strikeout, Superscript,
      Subscript) to refsMatch.
    + Removed backslashEscape; added new functions escapeStringUsing and
      backslashEscapes.
    + Moved failIfStrict from markdown reader, since it is now used also
      by the HTML reader.
    + Added a 'try' to the definition of indentSpaces.
    + In definition of 'reference', added check to make sure it's not a note
      reference.
    + Added functions: camelCaseToHyphenated, toRomanNumeral,
      anyOrderedListMarker, orderedListmarker, orderedListMarkers,
      charsInBalanced', withHorizDisplacement, romanNumeral
    + Fixed a bug in the anyLine parser. Previously it would parse an empty
      string "", but it should fail on an empty string, or we get an error
      when it is used inside "many" combinators.
    + Removed followedBy' parser, replacing it with the lookAhead parser from
      Parsec.
    + Added some needed 'try's before multicharacter parsers, especially in
      'option' contexts.
    + Removed the 'try' from the 'end' parser in 'enclosed', so that
      'enclosed' behaves like 'option', 'manyTill', etc.
    + Added lineClump parser, which parses a raw line block up to and
      including any following blank lines.
    + Renamed parseFromStr to parseFromString.
    + Added a 'try' to the 'end' parser in 'enclosed'.  This makes errors in
      the use of 'enclosed' less likely. Removed some now-unnecessary 'try's
      in calling code.
    + Removed unneeded 'try' in blanklines.
    + Removed endsWith function and rewrote calling functions to use
      isSuffixOf instead.
    + Added >>~ combinator.
    + Fixed bug in normalizeSpaces:  Space:Str "":Space should compress to
      Space.

  * Refactored runtests.pl; added separate tests for tables.

  * Shell scripts:

    + Added -asxhtml flag to tidy in html2markdown. This will
      perhaps help the parser, which expects closing tags.
    + Modified markdown2pdf to run pdflatex a second time if --toc or
      --table-of-contents was specified; otherwise the table of
      contents won't appear.
    + Modified markdown2pdf to print a helpful message if the 'ulem'
      LaTeX package is required and not found.

  * Changes to build process:

    + Dropped support for compilation with GHC 6.4.  GHC 6.6 or higher
      is now required.
    + Removed cabalize and Pandoc.cabal.in. The repository now contains
      pandoc.cabal itself.
    + Pandoc.cabal has been changed to pandoc.cabal, because HackageDB
      likes the cabal file to have the same name as the tarball.
    + Expanded and revised the package description in pandoc.cabal.
      Revised the package synopsis.
    + The tarball built by 'make tarball' now contains files built from
      templates (including man pages and shell scripts), so pandoc can
      be built directly using Cabal tools, without preprocessing.
    + Executable binaries are now stripped before installing.
    + Man pages are now generated from markdown sources, using pandoc's
      man page writer.
    + Use HTML version of README (instead of RTF) in Mac OS X installer.
    + Instead of testing for the existence of a pandoc symlink in build-exec,
      use ln -f.

  * Documentation:

    + Updated README and man pages with information on new features.
    + Updated INSTALL instructions with some useful clarifications and
      links.
    + Updated web content.

  * Added FreeBSD port.

  [ Recai Oktaş ]

  * debian/control:

    + Changed pandoc's Build-Depends to include libghc6-mtl-dev and
      libghc6-xhtml-dev.  Removed libghc6-html-dev.
    + Suggest texlive-latex-recommended | tetex-extra instead of
      tetex-bin.  This brings in fancyvrb and unicode support.

 -- Recai Oktaş <roktas@debian.org>  Tue, 16 Jan 2007 00:37:21 +0200

pandoc (0.3) unstable; urgency=low

  [ John MacFarlane ]

  * Changes in pandoc options:

    + Allow options to follow or precede arguments.
    + Changed '--smartypants' to '--smart' and adjusted symbols accordingly.
    + Added '--strict' option.
    + Added '-o/--output' option.
    + Added '--dump-args' and '--ignore-args' options (for use in wrappers).
    + Modified '-v' and '-h' output to go to STDERR, not STDOUT, and return
      error conditions.  This is helpful for writing wrappers.
    + Added copyright message to '-v' output, modeled after FSF messages.
    + Reformatted usage message so that it doesn't wrap illegibly.
    + Removed extra blanks after '-h' and '-D' output.

  * Added docbook writer.

  * Added implicit setting of default input and output format based
    on input and output filename extensions.  These defaults are
    overridden if explicit input and output formats are specified using
    '-t', '-f', '-r', or '-w' options.  Documented in pandoc(1) man page
    and README.

  * Allow ordered list items to begin with (single) letters, as well
    as numbers.  The list item marker may now be terminated either by
    '.' or by ')'.  This extension to standard markdown is documented
    in README.

  * Revised footnote syntax.  (See README for full details.)  The
    '[^1]' format now standard in markdown extensions is supported,
    as are inline footnotes with this syntax: `^[My note.]`.
    The earlier footnote syntax `^(1)` is no longer supported.

  * Improved HTML representation of footnotes.  All footnotes
    are now auto-numbered and appear in an ordered list at the
    end of the HTML document.  Since the default appearance is now
    acceptable, the old footnote styles have been removed from the
    HTML header.

  * Bug fixes:

    + Fixed a serious bug in the markdown, LaTeX, and RST readers.
      These readers ran 'runParser' on processed chunks of text to handle
      embedded block lists in lists and quotation blocks.  But then
      any changes made to the parser state in these chunks was lost,
      as the state is local to the parser.  So, for example, footnotes
      didn't work in quotes or list items.  The fix:  instead of calling
      runParser on some raw text, use setInput to make it the input, then
      parse it, then use setInput to restore the input to what it was
      before.  This is shorter and more elegant, and it fixes the problem.
    + Fixed bug in notFollowedBy' combinator (adding 'try' before
      'parser').  Adjusted code that uses this combinator accordingly.
    + Fixed bug in RTF writer that caused improper indentation on
      footnotes occurring in indented blocks like lists.
    + Fixed parsing of metadata in LaTeX reader.  Now the title, author,
      and date are parsed correctly.  Everything else in the preamble
      is skipped.
    + Modified escapedChar in LaTeX reader to allow a '\' at the end of a
      line to count as escaped whitespace.
    + Modified LaTeX reader to produce inline links rather than reference
      links.  Otherwise, links in footnotes aren't handled properly.
    + Fixed handling of titles in links in Markdown reader, so that
      embedded quotation marks are now handled properly.
    + Fixed Markdown reader's handling of embedded brackets in links.
    + Fixed Markdown reader so that it only parses bracketed material
      as a reference link if there is actually a corresponding key.
    + Revised inline code parsing in Markdown reader to conform to
      markdown standard.  Now any number of `s can begin inline code,
      which will end with the same number of `s.  For example, to
      have two backticks as code, write ``` `` ```.  Modified Markdown
      writer accordingly.
    + Fixed bug in text-wrapping routine in Markdown and RST writers.
      Now LineBreaks no longer cause wrapping problems.
    + Supported hexadecimal numerical entity references as well as
      decimal ones.
    + Fixed bug in Markdown reader's handling of underscores and other
      inline formatting markers inside reference labels:  for example,
      in '[A_B]: /url/a_b', the material between underscores was being
      parsed as emphasized inlines.
    + Changed Markdown reader's handling of backslash escapes so that
      only non-alphanumeric characters can be escaped.  Strict mode
      follows Markdown.pl in only allowing a select group of punctuation
      characters to be escaped.
    + Modified HTML reader to skip a newline following a `<br>` tag.
      Otherwise the newline will be treated as a space at the beginning
      of the next line.

  * Made handling of code blocks more consistent.  Previously, some
    readers allowed trailing newlines, while others stripped them.
    Now, all readers strip trailing newlines in code blocks. Writers
    insert a newline at the end of code blocks as needed.

  * Modified readers to make spacing at the end of output more consistent.

  * Minor improvements to LaTeX reader:

    + '\thanks' now treated like a footnote.
    + Simplified parsing of LaTeX command arguments and options.
      commandArgs now returns a list of arguments OR options (in
      whatever order they appear).  The brackets are included, and
      a new stripFirstAndLast function is provided to strip them off
      when needed.  This fixes a problem in dealing with \newcommand
      and \newenvironment.

  * Revised RTF writer:

    + Default font is now Helvetica.
    + An '\f0' is added to each '\pard', so that font resizing works
      correctly.

  * Moved handling of "smart typography" from the writers to the Markdown
    and LaTeX readers.  This allows great simplification of the writers
    and more accurate smart quotes, dashes, and ellipses.  DocBook can
    now use `<quote>`.  The '--smart' option now toggles an option in
    the parser state rather than a writer option.  Several new kinds
    of inline elements have been added: Quoted, Ellipses, Apostrophe,
    EmDash, EnDash.

  * Changes in HTML writer:

    + Include title block in header even when title is null.
    + Made javascript obfuscation of emails even more obfuscatory,
      by combining it with entity obfuscation.

  * Changed default ASCIIMathML text color to black.

  * Test suite:

    + Added --strip-trailing-cr option to diff in runtests.pl, for
      compatibility with Windows.
    + Added regression tests with footnotes in quote blocks and lists.

  * Makefile changes:

    + osx-pkg target creates a Mac OS X package (directory). New osx
      directory contains files needed for construction of the package.
    + osx-dmg target creates a compressed disk image containing the package.
    + win-pkg target creates Windows binary package.
    + tarball target creates distribution source tarball.
    + website target generates pandoc's website automatically, including
      demos.  New 'web' directory containts files needed for construction
      of the website (which will be created as the 'pandoc' subdirectory
      of 'web').
    + Makefile checks to see if we're running Windows/Cygwin; if so,
      a '.exe' extension is added to each executable in EXECS.

  * Removed all wrappers except markdown2pdf and html2markdown.

  * Added new wrapper hsmarkdown, to be used as a drop-in replacement
    for Markdown.pl.  hsmarkdown calls pandoc with the '--strict'
    option and disables other options.

  * Added code to html2markdown that tries to determine the character
    encoding of an HTML file, by parsing the "Content-type" meta tag.

    + If the encoding can't be determined, then if the content is local,
      the local encoding is used; if it comes from a URL, UTF-8 is used
      by default.
    + If input is from STDIN, don't try to determine character encoding.
    + Encoding can be specified explicitly using '-e' option.

  * Improved warning messages in wrappers:

    + Print warning if iconv not available
    + More user-friendly error messages in markdown2pdf, when
      pdflatex fails.

  * Code cleanup:

    + Renamed 'Text/Pandoc/HtmlEntities' module to
      'Text/Pandoc/Entities'. Also changed function names so as
      not to be HTML-specific.
    + Refactored SGML string escaping functions from HTML and Docbook
      writers into Text/Pandoc/Shared.  (escapeSGML, stringToSGML)
    + Removed 'BlockQuoteContext' from ParserContext, as it isn't
      used anywhere.
    + Removed splitBySpace and replaced it with a general, polymorphic
      splitBy function.
    + Refactored LaTeX reader for clarity (added isArg function).
    + Converted some CR's to LF's in src/ui/default/print.css.
    + Added license text to top of source files.
    + Added module data for haddock to source files.
    + Reformatted code for consistency.

  * Rewrote documentation and man pages.  Split README into INSTALL
    and README.

  * Split LICENSE into COPYING and COPYRIGHT.

  * Removed TODO, since we now maintain ToDo on the wiki.

  * Made COPYRIGHT in top level a symlink to debian/copyright, to avoid
    duplication.

  [ Recai Oktaş ]

  * Revamped build process to conform to debian standards and created
    a proper debian package.  Closes: #391666.

  * Modified build process to support GHC 6.6.

    + The package can still be compiled using GHC 6.4.2, though because
      of dependencies the "make deb" target works only with GHC 6.6+.
    + The script 'cabalize' is used to create an appropriate
      'Pandoc.cabal' from 'Pandoc.cabal.in', depending on the GHC and
      Cabal versions.

  * Refactored template processing (fillTemplates.pl).

  * Modified wrapper scripts to make them more robust and portable.
    To avoid code duplication and ensure consistency, wrappers are
    generated via a templating system from templates in src/wrappers.

    + Wrappers now accept multiple filenames, when appropriate.
    + Spaces and tabs allowed in filenames.
    + getopts shell builtin is used for portable option parsing.
    + Improved html2markdown's web grabber code, making it more robust,
      configurable and verbose.  Added '-e', '-g' options.

 -- Recai Oktaş <roktas@debian.org>  Fri, 05 Jan 2007 09:41:19 +0200

pandoc (0.2) unstable; urgency=low

  * Fixed unicode/utf-8 translation

 -- John MacFarlane <jgm@berkeley.edu>  Mon,  14 Aug 2006 00:00:00 -0400

pandoc (0.1) unstable; urgency=low

  * Initial creation of debian package

 -- John MacFarlane <jgm@berkeley.edu>  Mon,  14 Aug 2006 00:00:00 -0400