OS X 2013Q1 lyx doesn't work #62

rmustacc · 2013-07-10T18:01:08Z

If you simply try to run lyx with no arguments, it immediately gets a sigbus:

PATH=$PATH:/usr/pkg//Applications/LyX.app/Contents/MacOS lyx
Bus error: 10

The text was updated successfully, but these errors were encountered:

Changes since 1.1.0 (from git log): issue #70: boolean flags may have true default Merge pull request #69 from Tieske/master general updates, mainly added CONTRIBUTING.md Merge pull request #1 from mark-otaris/patch-1 Caught red handed, after a shameless copy 😄. Thx for the fix This is Penlight, not Busted added - a shameless copy of busted - CONTRIBUTING.md file renamed docs to doc, to fix luarocks warning updated readme to md format changed .txt to .md files added rockspec, fixed typo in filename Merge pull request #67 from Tieske/add_executeex added utils.executeex, which also returns output of stdout and stderr updated, using readfile now, remove obsolete os.remove commands Merge pull request #66 from Tieske/fix_temp windows environment variable TMP is deprecated, use TEMP mode binary mode optional (binary would require platform specific line endings to be converted which is not very portable, and mostly just text will do) update: read files in binary mode added utils.executeex, which also returns output of stdout and stderr windows environment variable TMP is deprecated, use TEMP HTML mode skips DOCTYPE; no globals harmed outside pl.utils Merge pull request #64 from Tieske/fix_dir fixed makepath creation, added/updated some documentation fixed dirpath creation, added/updated some documentation basic parser handles comments, and HTML mode is working nicely with real-world HTML use LOM ordered attributes when stringifying if present (Danny). manual merge of John Schember's readonly table support Merge pull request #61 from user-none/isempty Thanks, John! Merge pull request #62 from CoolisTheName007/patch-2 Transform using values as keys. missing locals Set.issempty mispelling; default ctor borked; more tests Transform using values as keys. Merge remote-tracking branch 'upstream/master' into empty Add function to check if an object is empty. Merge pull request #53 from pkazmier/master Added iterators to allow traversing tables in a sorted manner Merge pull request #60 from theypsilon/master base class methods invocation possible not only in method constructor _init Adding new class method 'base' which can be used for calling base class methods easily pl.class broke with classes that redefined __newindex; OrderedMap can now use normal indexing to assign keys in order Merge pull request #54 from CoolisTheName007/patch-1 Typo fixed, libarary->library Typo fixed. Added tablex.sort and tablex.sortv to iterate over sorted elements There are times when it would be convenient to iterate over a table either by sorted keys or values. There is an example of this in the PIL book. I added the two functions to tablex module, added two test cases, and updated the PL manual. Merge pull request #51 from pkazmier/master Thanks! lapp.add_type was broken due to uninitialized table variable. The variable `types` was not initialized preventing `add_type` from appending custom types to the variable. I also added a new test to cover this test case in the future.

== 2.7.0 / 2013-08-03 Charlie Savage * Do not call Ruby code during a GC cycle (Dirkjan Bussink) * Fix thread error handling (Geoffrey Giesemann) * Set libxml error handler when a libxml-ruby error handler is set (Geoffrey Giesemann) * Fix tests for nil TypeErrors (Geoffrey Giesemann) * Fix segmentation fault for issue #62. The deregisternode function was incorrectly NULLing out a node's mark and free functions for nodes removed from a document tree (Charlie Savage) * Cleanup writer code (Charlie Savage) * Fix equality check of LibXML::XML::Error against other objects (Michał Szajbe) * Fix XML::Writer.set_quote_char UT, wrong expected string (#63). (julp) * Fix potential segfault when GC occurs while creating ns from xpath (Timothy Elliott) * Fix segmentation fault after failing to load external schema (#59). (Clifford Heath)

(benchmarks clipped from release notes) ------------------------------------------------------------------------ r80 | [email protected] | 2013-08-13 14:55:00 +0200 (Tue, 13 Aug 2013) | 6 lines Add autoconf tests for size_t and ssize_t. Sort-of resolves public issue 79; it would solve the problem if MSVC typically used autoconf. However, it gives a natural place (config.h) to put the typedef even for MSVC. R=jsbell ------------------------------------------------------------------------ r79 | [email protected] | 2013-07-29 13:06:44 +0200 (Mon, 29 Jul 2013) | 14 lines When we compare the number of bytes produced with the offset for a backreference, make the signedness of the bytes produced clear, by sticking it into a size_t. This avoids a signed/unsigned compare warning from MSVC (public issue 71), and also is slightly clearer. Since the line is now so long the explanatory comment about the -1u trick has to go somewhere else anyway, I used the opportunity to explain it in slightly more detail. This is a purely stylistic change; the emitted assembler from GCC is identical. R=jeff ------------------------------------------------------------------------ r78 | [email protected] | 2013-06-30 21:24:03 +0200 (Sun, 30 Jun 2013) | 111 lines In the fast path for decompressing literals, instead of checking whether there's 16 bytes free and then checking right afterwards (when having subtracted the literal size) that there are now 5 bytes free, just check once for 21 bytes. This skips a compare and a branch; although it is easily predictable, it is still a few cycles on a fast path that we would like to get rid of. Benchmarking this yields very confusing results. On open-source GCC 4.8.1 on Haswell, we get exactly the expected results; the benchmarks where we hit the fast path for literals (in particular the two HTML benchmarks and the protobuf benchmark) give very nice speedups, and the others are not really affected. However, benchmarks with Google's GCC branch on other hardware is much less clear. It seems that we have a weak loss in some cases (and the win for the typical win cases are not nearly as clear), but that it depends on microarchitecture and plain luck in how we run the benchmark. Looking at the generated assembler, it seems that the removal of the if causes other large-scale changes in how the function is laid out, which makes it likely that this is just bad luck. Thus, we should keep this change, even though its exact current impact is unclear; it's a sensible change per se, and dropping it on the basis of microoptimization for a given compiler (or even branch of a compiler) would seem like a bad strategy in the long run. ------------------------------------------------------------------------ r77 | [email protected] | 2013-06-14 23:42:26 +0200 (Fri, 14 Jun 2013) | 92 lines Make the two IncrementalCopy* functions take in an ssize_t instead of a len, in order to avoid having to do 32-to-64-bit signed conversions on a hot path during decompression. (Also fixes some MSVC warnings, mentioned in public issue 75, but more of those remain.) They cannot be size_t because we expect them to go negative and test for that. This saves a few movzwl instructions, yielding ~2% speedup in decompression. ------------------------------------------------------------------------ r76 | [email protected] | 2013-06-13 18:19:52 +0200 (Thu, 13 Jun 2013) | 9 lines Add support for uncompressing to iovecs (scatter I/O). Windows does not have struct iovec defined anywhere, so we define our own version that's equal to what UNIX typically has. The bulk of this patch was contributed by Mohit Aron. R=jeff ------------------------------------------------------------------------ r75 | [email protected] | 2013-06-12 21:51:15 +0200 (Wed, 12 Jun 2013) | 4 lines Some code reorganization needed for an internal change. R=fikes ------------------------------------------------------------------------ r74 | [email protected] | 2013-04-09 17:33:30 +0200 (Tue, 09 Apr 2013) | 4 lines Supports truncated test data in zippy benchmark. R=sesse ------------------------------------------------------------------------ r73 | [email protected] | 2013-02-05 15:36:15 +0100 (Tue, 05 Feb 2013) | 4 lines Release Snappy 1.1.0. R=sanjay ------------------------------------------------------------------------ r72 | [email protected] | 2013-02-05 15:30:05 +0100 (Tue, 05 Feb 2013) | 9 lines Make ./snappy_unittest pass without "srcdir" being defined. Previously, snappy_unittests would read from an absolute path /testdata/..; convert it to use a relative path instead. Patch from Marc-Antonie Ruel. R=maruel ------------------------------------------------------------------------ r71 | [email protected] | 2013-01-18 13:16:36 +0100 (Fri, 18 Jan 2013) | 287 lines Increase the Zippy block size from 32 kB to 64 kB, winning ~3% density while being effectively performance neutral. The longer story about density is that we win 3-6% density on the benchmarks where this has any effect at all; many of the benchmarks (cp, c, lsp, man) are smaller than 32 kB and thus will have no effect. Binary data also seems to win little or nothing; of course, the already-compressed data wins nothing. The protobuf benchmark wins as much as ~18% depending on architecture, but I wouldn't be too sure that this is representative of protobuf data in general. As of performance, we lose a tiny amount since we get more tags (e.g., a long literal might be broken up into literal-copy-literal), but we win it back with less clearing of the hash table, and more opportunities to skip incompressible data (e.g. in the jpg benchmark). Decompression seems to get ever so slightly slower, again due to more tags. The total net change is about as close to zero as we can get, so the end effect seems to be simply more density and no real performance change. The comment about not changing kBlockSize, scary as it is, is not really relevant, since we're never going to have a block-level decompressor without explicitly marked blocks. Replace it with something more appropriate. This affects the framing format, but it's okay to change it since it basically has no users yet. ------------------------------------------------------------------------ r70 | [email protected] | 2013-01-06 20:21:26 +0100 (Sun, 06 Jan 2013) | 6 lines Adjust the Snappy open-source distribution for the changes in Google's internal file API. R=sanjay ------------------------------------------------------------------------ r69 | [email protected] | 2013-01-04 12:54:20 +0100 (Fri, 04 Jan 2013) | 15 lines Change a few ORs to additions where they don't matter. This helps the compiler use the LEA instruction more efficiently, since e.g. a + (b << 2) can be encoded as one instruction. Even more importantly, it can constant-fold the COPY_* enums together with the shifted negative constants, which also saves some instructions. (We don't need it for LITERAL, since it happens to be 0.) I am unsure why the compiler couldn't do this itself, but the theory is that it cannot prove that len-1 and len-4 cannot underflow/wrap, and thus can't do the optimization safely. The gains are small but measurable; 0.5-1.0% over the BM_Z* benchmarks (measured on Westmere, Sandy Bridge and Istanbul). R=sanjay ------------------------------------------------------------------------ r68 | [email protected] | 2012-10-08 13:37:16 +0200 (Mon, 08 Oct 2012) | 5 lines Stop giving -Werror to automake, due to an incompatibility between current versions of libtool and automake on non-GNU platforms (e.g. Mac OS X). R=sanjay ------------------------------------------------------------------------ r67 | [email protected] | 2012-08-17 15:54:47 +0200 (Fri, 17 Aug 2012) | 5 lines Fix public issue 66: Document GetUncompressedLength better, in particular that it leaves the source in a state that's not appropriate for RawUncompress. R=sanjay ------------------------------------------------------------------------ r66 | [email protected] | 2012-07-31 13:44:44 +0200 (Tue, 31 Jul 2012) | 5 lines Fix public issue 64: Check for <sys/time.h> at configure time, since MSVC seemingly does not have it. R=sanjay ------------------------------------------------------------------------ r65 | [email protected] | 2012-07-04 11:34:48 +0200 (Wed, 04 Jul 2012) | 10 lines Handle the case where gettimeofday() goes backwards or returns the same value twice; it could cause division by zero in the unit test framework. (We already had one fix for this in place, but it was incomplete.) This could in theory happen on any system, since there are few guarantees about gettimeofday(), but seems to only happen in practice on GNU/Hurd, where gettimeofday() is cached and only updated ever so often. R=sanjay ------------------------------------------------------------------------ r64 | [email protected] | 2012-07-04 11:28:33 +0200 (Wed, 04 Jul 2012) | 6 lines Mark ARMv4 as not supporting unaligned accesses (not just ARMv5 and ARMv6); apparently Debian still targets these by default, giving us segfaults on armel. R=sanjay ------------------------------------------------------------------------ r63 | [email protected] | 2012-05-22 11:46:05 +0200 (Tue, 22 May 2012) | 5 lines Fix public bug #62: Remove an extraneous comma at the end of an enum list, causing compile errors when embedded in Mozilla on OpenBSD. R=sanjay ------------------------------------------------------------------------ r62 | [email protected] | 2012-05-22 11:32:50 +0200 (Tue, 22 May 2012) | 8 lines Snappy library no longer depends on iostream. Achieved by moving logging macro definitions to a test-only header file, and by changing non-test code to use assert, fprintf, and abort instead of LOG/CHECK macros. R=sesse

Changes since 1.1.0 (from git log): issue #70: boolean flags may have true default Merge pull request #69 from Tieske/master general updates, mainly added CONTRIBUTING.md Merge pull request #1 from mark-otaris/patch-1 Caught red handed, after a shameless copy 😄. Thx for the fix This is Penlight, not Busted added - a shameless copy of busted - CONTRIBUTING.md file renamed docs to doc, to fix luarocks warning updated readme to md format changed .txt to .md files added rockspec, fixed typo in filename Merge pull request #67 from Tieske/add_executeex added utils.executeex, which also returns output of stdout and stderr updated, using readfile now, remove obsolete os.remove commands Merge pull request #66 from Tieske/fix_temp windows environment variable TMP is deprecated, use TEMP mode binary mode optional (binary would require platform specific line endings to be converted which is not very portable, and mostly just text will do) update: read files in binary mode added utils.executeex, which also returns output of stdout and stderr windows environment variable TMP is deprecated, use TEMP HTML mode skips DOCTYPE; no globals harmed outside pl.utils Merge pull request #64 from Tieske/fix_dir fixed makepath creation, added/updated some documentation fixed dirpath creation, added/updated some documentation basic parser handles comments, and HTML mode is working nicely with real-world HTML use LOM ordered attributes when stringifying if present (Danny). manual merge of John Schember's readonly table support Merge pull request #61 from user-none/isempty Thanks, John! Merge pull request #62 from CoolisTheName007/patch-2 Transform using values as keys. missing locals Set.issempty mispelling; default ctor borked; more tests Transform using values as keys. Merge remote-tracking branch 'upstream/master' into empty Add function to check if an object is empty. Merge pull request #53 from pkazmier/master Added iterators to allow traversing tables in a sorted manner Merge pull request #60 from theypsilon/master base class methods invocation possible not only in method constructor _init Adding new class method 'base' which can be used for calling base class methods easily pl.class broke with classes that redefined __newindex; OrderedMap can now use normal indexing to assign keys in order Merge pull request #54 from CoolisTheName007/patch-1 Typo fixed, libarary->library Typo fixed. Added tablex.sort and tablex.sortv to iterate over sorted elements There are times when it would be convenient to iterate over a table either by sorted keys or values. There is an example of this in the PIL book. I added the two functions to tablex module, added two test cases, and updated the PL manual. Merge pull request #51 from pkazmier/master Thanks! lapp.add_type was broken due to uninitialized table variable. The variable `types` was not initialized preventing `add_type` from appending custom types to the variable. I also added a new test to cover this test case in the future.

== 2.7.0 / 2013-08-03 Charlie Savage * Do not call Ruby code during a GC cycle (Dirkjan Bussink) * Fix thread error handling (Geoffrey Giesemann) * Set libxml error handler when a libxml-ruby error handler is set (Geoffrey Giesemann) * Fix tests for nil TypeErrors (Geoffrey Giesemann) * Fix segmentation fault for issue #62. The deregisternode function was incorrectly NULLing out a node's mark and free functions for nodes removed from a document tree (Charlie Savage) * Cleanup writer code (Charlie Savage) * Fix equality check of LibXML::XML::Error against other objects (Michał Szajbe) * Fix XML::Writer.set_quote_char UT, wrong expected string (#63). (julp) * Fix potential segfault when GC occurs while creating ns from xpath (Timothy Elliott) * Fix segmentation fault after failing to load external schema (#59). (Clifford Heath)

(benchmarks clipped from release notes) ------------------------------------------------------------------------ r80 | [email protected] | 2013-08-13 14:55:00 +0200 (Tue, 13 Aug 2013) | 6 lines Add autoconf tests for size_t and ssize_t. Sort-of resolves public issue 79; it would solve the problem if MSVC typically used autoconf. However, it gives a natural place (config.h) to put the typedef even for MSVC. R=jsbell ------------------------------------------------------------------------ r79 | [email protected] | 2013-07-29 13:06:44 +0200 (Mon, 29 Jul 2013) | 14 lines When we compare the number of bytes produced with the offset for a backreference, make the signedness of the bytes produced clear, by sticking it into a size_t. This avoids a signed/unsigned compare warning from MSVC (public issue 71), and also is slightly clearer. Since the line is now so long the explanatory comment about the -1u trick has to go somewhere else anyway, I used the opportunity to explain it in slightly more detail. This is a purely stylistic change; the emitted assembler from GCC is identical. R=jeff ------------------------------------------------------------------------ r78 | [email protected] | 2013-06-30 21:24:03 +0200 (Sun, 30 Jun 2013) | 111 lines In the fast path for decompressing literals, instead of checking whether there's 16 bytes free and then checking right afterwards (when having subtracted the literal size) that there are now 5 bytes free, just check once for 21 bytes. This skips a compare and a branch; although it is easily predictable, it is still a few cycles on a fast path that we would like to get rid of. Benchmarking this yields very confusing results. On open-source GCC 4.8.1 on Haswell, we get exactly the expected results; the benchmarks where we hit the fast path for literals (in particular the two HTML benchmarks and the protobuf benchmark) give very nice speedups, and the others are not really affected. However, benchmarks with Google's GCC branch on other hardware is much less clear. It seems that we have a weak loss in some cases (and the win for the typical win cases are not nearly as clear), but that it depends on microarchitecture and plain luck in how we run the benchmark. Looking at the generated assembler, it seems that the removal of the if causes other large-scale changes in how the function is laid out, which makes it likely that this is just bad luck. Thus, we should keep this change, even though its exact current impact is unclear; it's a sensible change per se, and dropping it on the basis of microoptimization for a given compiler (or even branch of a compiler) would seem like a bad strategy in the long run. ------------------------------------------------------------------------ r77 | [email protected] | 2013-06-14 23:42:26 +0200 (Fri, 14 Jun 2013) | 92 lines Make the two IncrementalCopy* functions take in an ssize_t instead of a len, in order to avoid having to do 32-to-64-bit signed conversions on a hot path during decompression. (Also fixes some MSVC warnings, mentioned in public issue 75, but more of those remain.) They cannot be size_t because we expect them to go negative and test for that. This saves a few movzwl instructions, yielding ~2% speedup in decompression. ------------------------------------------------------------------------ r76 | [email protected] | 2013-06-13 18:19:52 +0200 (Thu, 13 Jun 2013) | 9 lines Add support for uncompressing to iovecs (scatter I/O). Windows does not have struct iovec defined anywhere, so we define our own version that's equal to what UNIX typically has. The bulk of this patch was contributed by Mohit Aron. R=jeff ------------------------------------------------------------------------ r75 | [email protected] | 2013-06-12 21:51:15 +0200 (Wed, 12 Jun 2013) | 4 lines Some code reorganization needed for an internal change. R=fikes ------------------------------------------------------------------------ r74 | [email protected] | 2013-04-09 17:33:30 +0200 (Tue, 09 Apr 2013) | 4 lines Supports truncated test data in zippy benchmark. R=sesse ------------------------------------------------------------------------ r73 | [email protected] | 2013-02-05 15:36:15 +0100 (Tue, 05 Feb 2013) | 4 lines Release Snappy 1.1.0. R=sanjay ------------------------------------------------------------------------ r72 | [email protected] | 2013-02-05 15:30:05 +0100 (Tue, 05 Feb 2013) | 9 lines Make ./snappy_unittest pass without "srcdir" being defined. Previously, snappy_unittests would read from an absolute path /testdata/..; convert it to use a relative path instead. Patch from Marc-Antonie Ruel. R=maruel ------------------------------------------------------------------------ r71 | [email protected] | 2013-01-18 13:16:36 +0100 (Fri, 18 Jan 2013) | 287 lines Increase the Zippy block size from 32 kB to 64 kB, winning ~3% density while being effectively performance neutral. The longer story about density is that we win 3-6% density on the benchmarks where this has any effect at all; many of the benchmarks (cp, c, lsp, man) are smaller than 32 kB and thus will have no effect. Binary data also seems to win little or nothing; of course, the already-compressed data wins nothing. The protobuf benchmark wins as much as ~18% depending on architecture, but I wouldn't be too sure that this is representative of protobuf data in general. As of performance, we lose a tiny amount since we get more tags (e.g., a long literal might be broken up into literal-copy-literal), but we win it back with less clearing of the hash table, and more opportunities to skip incompressible data (e.g. in the jpg benchmark). Decompression seems to get ever so slightly slower, again due to more tags. The total net change is about as close to zero as we can get, so the end effect seems to be simply more density and no real performance change. The comment about not changing kBlockSize, scary as it is, is not really relevant, since we're never going to have a block-level decompressor without explicitly marked blocks. Replace it with something more appropriate. This affects the framing format, but it's okay to change it since it basically has no users yet. ------------------------------------------------------------------------ r70 | [email protected] | 2013-01-06 20:21:26 +0100 (Sun, 06 Jan 2013) | 6 lines Adjust the Snappy open-source distribution for the changes in Google's internal file API. R=sanjay ------------------------------------------------------------------------ r69 | [email protected] | 2013-01-04 12:54:20 +0100 (Fri, 04 Jan 2013) | 15 lines Change a few ORs to additions where they don't matter. This helps the compiler use the LEA instruction more efficiently, since e.g. a + (b << 2) can be encoded as one instruction. Even more importantly, it can constant-fold the COPY_* enums together with the shifted negative constants, which also saves some instructions. (We don't need it for LITERAL, since it happens to be 0.) I am unsure why the compiler couldn't do this itself, but the theory is that it cannot prove that len-1 and len-4 cannot underflow/wrap, and thus can't do the optimization safely. The gains are small but measurable; 0.5-1.0% over the BM_Z* benchmarks (measured on Westmere, Sandy Bridge and Istanbul). R=sanjay ------------------------------------------------------------------------ r68 | [email protected] | 2012-10-08 13:37:16 +0200 (Mon, 08 Oct 2012) | 5 lines Stop giving -Werror to automake, due to an incompatibility between current versions of libtool and automake on non-GNU platforms (e.g. Mac OS X). R=sanjay ------------------------------------------------------------------------ r67 | [email protected] | 2012-08-17 15:54:47 +0200 (Fri, 17 Aug 2012) | 5 lines Fix public issue 66: Document GetUncompressedLength better, in particular that it leaves the source in a state that's not appropriate for RawUncompress. R=sanjay ------------------------------------------------------------------------ r66 | [email protected] | 2012-07-31 13:44:44 +0200 (Tue, 31 Jul 2012) | 5 lines Fix public issue 64: Check for <sys/time.h> at configure time, since MSVC seemingly does not have it. R=sanjay ------------------------------------------------------------------------ r65 | [email protected] | 2012-07-04 11:34:48 +0200 (Wed, 04 Jul 2012) | 10 lines Handle the case where gettimeofday() goes backwards or returns the same value twice; it could cause division by zero in the unit test framework. (We already had one fix for this in place, but it was incomplete.) This could in theory happen on any system, since there are few guarantees about gettimeofday(), but seems to only happen in practice on GNU/Hurd, where gettimeofday() is cached and only updated ever so often. R=sanjay ------------------------------------------------------------------------ r64 | [email protected] | 2012-07-04 11:28:33 +0200 (Wed, 04 Jul 2012) | 6 lines Mark ARMv4 as not supporting unaligned accesses (not just ARMv5 and ARMv6); apparently Debian still targets these by default, giving us segfaults on armel. R=sanjay ------------------------------------------------------------------------ r63 | [email protected] | 2012-05-22 11:46:05 +0200 (Tue, 22 May 2012) | 5 lines Fix public bug #62: Remove an extraneous comma at the end of an enum list, causing compile errors when embedded in Mozilla on OpenBSD. R=sanjay ------------------------------------------------------------------------ r62 | [email protected] | 2012-05-22 11:32:50 +0200 (Tue, 22 May 2012) | 8 lines Snappy library no longer depends on iostream. Achieved by moving logging macro definitions to a test-only header file, and by changing non-test code to use assert, fprintf, and abort instead of LOG/CHECK macros. R=sesse

Incompatible Changes: * can no longer use both `\' and `:' (didn't work anyway) New Features: * Handling of run-time errors (default: die) is now overridable by subclasses via signature_error_handler(). [github #54] * Can now have aliased named parameters. [github #57] * remove dependency on Devel::BeginLift [github #39] * can now use `when' to specify default conditions [github #48] * can use `//=' as a shortcut for `when undef' [github #45] * can now provide `where' constraints in addition to (or instead of) a type [github #7] * can now use `...' to disable further argument checking [github #49] * can now specify more than one alternative in type unions [github #55] * can now nest parameterized types Bug Fixes: * Removed experimental smartmatch warnings * Don't require Data::Alias for named params unless you have to [github #71] * Fixed obscure bug where an eval in Method::Signatures wouldn't be skipped when carp'ing (i.e. in carp_location_for()) [github #72] * Data::Alias is only loaded when needed avoiding a threads + eval bug in most cases and improving compile time performance. [rt.cpan.org 82922, github #62] * Compile-time errors now reporting proper line numbers. [github #61] * Trailing commas on parameter lists are now ok. [rt.cpan.org 81364] * Default condition of `when {}' now interpreted as `when { $_ ~~ {} }' (avoids parse error). [github #60] Optimizations: * better signature parsing using PPI [github #11] Distribution Fixes: * Fixed failing test in 5.10.0 (uncovered by CPAN Testers) * Fixed repo link in metadata (thanks dsteinbrunner) [github #87] * Add M::S::Parameter to MANIFEST [github #76] * Change representation of Infinity to work on Win32 [github #75] * Fixed stray detritus in MANIFEST. * Somehow my last-minute fix to the new error handler test didn't make it in; this will fix "Can't locate Moose.pm" errors. * Fixed test failing on 5.10.0 as per github #59. * Fixed subtests failing on Test::More's prior to 0.96. Docs: * Updated close parend problem to include quotes and a workaround [rt.cpan.org 85925] * Fixed some typos (thanks dsteinbrunner) [github #88] * Found and fixed missing parend * Minor clarifications here and there. * Clarified what doesn't work in Perl 5.8. * Added Function::Parameters to See Also section. * Updated copyright. * documented all new features * new ASCI-art breakdown of signature syntax * minor tweaks and corrections Misc: * Rearranged so signature is now an object [github #30] * Add hook for Travis CI [github #78] * Failure to parse parameters will now produce a more useful error.

Changes since 1.1.0 (from git log): issue #70: boolean flags may have true default Merge pull request #69 from Tieske/master general updates, mainly added CONTRIBUTING.md Merge pull request #1 from mark-otaris/patch-1 Caught red handed, after a shameless copy 😄. Thx for the fix This is Penlight, not Busted added - a shameless copy of busted - CONTRIBUTING.md file renamed docs to doc, to fix luarocks warning updated readme to md format changed .txt to .md files added rockspec, fixed typo in filename Merge pull request #67 from Tieske/add_executeex added utils.executeex, which also returns output of stdout and stderr updated, using readfile now, remove obsolete os.remove commands Merge pull request #66 from Tieske/fix_temp windows environment variable TMP is deprecated, use TEMP mode binary mode optional (binary would require platform specific line endings to be converted which is not very portable, and mostly just text will do) update: read files in binary mode added utils.executeex, which also returns output of stdout and stderr windows environment variable TMP is deprecated, use TEMP HTML mode skips DOCTYPE; no globals harmed outside pl.utils Merge pull request #64 from Tieske/fix_dir fixed makepath creation, added/updated some documentation fixed dirpath creation, added/updated some documentation basic parser handles comments, and HTML mode is working nicely with real-world HTML use LOM ordered attributes when stringifying if present (Danny). manual merge of John Schember's readonly table support Merge pull request #61 from user-none/isempty Thanks, John! Merge pull request #62 from CoolisTheName007/patch-2 Transform using values as keys. missing locals Set.issempty mispelling; default ctor borked; more tests Transform using values as keys. Merge remote-tracking branch 'upstream/master' into empty Add function to check if an object is empty. Merge pull request #53 from pkazmier/master Added iterators to allow traversing tables in a sorted manner Merge pull request #60 from theypsilon/master base class methods invocation possible not only in method constructor _init Adding new class method 'base' which can be used for calling base class methods easily pl.class broke with classes that redefined __newindex; OrderedMap can now use normal indexing to assign keys in order Merge pull request #54 from CoolisTheName007/patch-1 Typo fixed, libarary->library Typo fixed. Added tablex.sort and tablex.sortv to iterate over sorted elements There are times when it would be convenient to iterate over a table either by sorted keys or values. There is an example of this in the PIL book. I added the two functions to tablex module, added two test cases, and updated the PL manual. Merge pull request #51 from pkazmier/master Thanks! lapp.add_type was broken due to uninitialized table variable. The variable `types` was not initialized preventing `add_type` from appending custom types to the variable. I also added a new test to cover this test case in the future.

== 2.7.0 / 2013-08-03 Charlie Savage * Do not call Ruby code during a GC cycle (Dirkjan Bussink) * Fix thread error handling (Geoffrey Giesemann) * Set libxml error handler when a libxml-ruby error handler is set (Geoffrey Giesemann) * Fix tests for nil TypeErrors (Geoffrey Giesemann) * Fix segmentation fault for issue #62. The deregisternode function was incorrectly NULLing out a node's mark and free functions for nodes removed from a document tree (Charlie Savage) * Cleanup writer code (Charlie Savage) * Fix equality check of LibXML::XML::Error against other objects (Michał Szajbe) * Fix XML::Writer.set_quote_char UT, wrong expected string (#63). (julp) * Fix potential segfault when GC occurs while creating ns from xpath (Timothy Elliott) * Fix segmentation fault after failing to load external schema (#59). (Clifford Heath)

(benchmarks clipped from release notes) ------------------------------------------------------------------------ r80 | [email protected] | 2013-08-13 14:55:00 +0200 (Tue, 13 Aug 2013) | 6 lines Add autoconf tests for size_t and ssize_t. Sort-of resolves public issue 79; it would solve the problem if MSVC typically used autoconf. However, it gives a natural place (config.h) to put the typedef even for MSVC. R=jsbell ------------------------------------------------------------------------ r79 | [email protected] | 2013-07-29 13:06:44 +0200 (Mon, 29 Jul 2013) | 14 lines When we compare the number of bytes produced with the offset for a backreference, make the signedness of the bytes produced clear, by sticking it into a size_t. This avoids a signed/unsigned compare warning from MSVC (public issue 71), and also is slightly clearer. Since the line is now so long the explanatory comment about the -1u trick has to go somewhere else anyway, I used the opportunity to explain it in slightly more detail. This is a purely stylistic change; the emitted assembler from GCC is identical. R=jeff ------------------------------------------------------------------------ r78 | [email protected] | 2013-06-30 21:24:03 +0200 (Sun, 30 Jun 2013) | 111 lines In the fast path for decompressing literals, instead of checking whether there's 16 bytes free and then checking right afterwards (when having subtracted the literal size) that there are now 5 bytes free, just check once for 21 bytes. This skips a compare and a branch; although it is easily predictable, it is still a few cycles on a fast path that we would like to get rid of. Benchmarking this yields very confusing results. On open-source GCC 4.8.1 on Haswell, we get exactly the expected results; the benchmarks where we hit the fast path for literals (in particular the two HTML benchmarks and the protobuf benchmark) give very nice speedups, and the others are not really affected. However, benchmarks with Google's GCC branch on other hardware is much less clear. It seems that we have a weak loss in some cases (and the win for the typical win cases are not nearly as clear), but that it depends on microarchitecture and plain luck in how we run the benchmark. Looking at the generated assembler, it seems that the removal of the if causes other large-scale changes in how the function is laid out, which makes it likely that this is just bad luck. Thus, we should keep this change, even though its exact current impact is unclear; it's a sensible change per se, and dropping it on the basis of microoptimization for a given compiler (or even branch of a compiler) would seem like a bad strategy in the long run. ------------------------------------------------------------------------ r77 | [email protected] | 2013-06-14 23:42:26 +0200 (Fri, 14 Jun 2013) | 92 lines Make the two IncrementalCopy* functions take in an ssize_t instead of a len, in order to avoid having to do 32-to-64-bit signed conversions on a hot path during decompression. (Also fixes some MSVC warnings, mentioned in public issue 75, but more of those remain.) They cannot be size_t because we expect them to go negative and test for that. This saves a few movzwl instructions, yielding ~2% speedup in decompression. ------------------------------------------------------------------------ r76 | [email protected] | 2013-06-13 18:19:52 +0200 (Thu, 13 Jun 2013) | 9 lines Add support for uncompressing to iovecs (scatter I/O). Windows does not have struct iovec defined anywhere, so we define our own version that's equal to what UNIX typically has. The bulk of this patch was contributed by Mohit Aron. R=jeff ------------------------------------------------------------------------ r75 | [email protected] | 2013-06-12 21:51:15 +0200 (Wed, 12 Jun 2013) | 4 lines Some code reorganization needed for an internal change. R=fikes ------------------------------------------------------------------------ r74 | [email protected] | 2013-04-09 17:33:30 +0200 (Tue, 09 Apr 2013) | 4 lines Supports truncated test data in zippy benchmark. R=sesse ------------------------------------------------------------------------ r73 | [email protected] | 2013-02-05 15:36:15 +0100 (Tue, 05 Feb 2013) | 4 lines Release Snappy 1.1.0. R=sanjay ------------------------------------------------------------------------ r72 | [email protected] | 2013-02-05 15:30:05 +0100 (Tue, 05 Feb 2013) | 9 lines Make ./snappy_unittest pass without "srcdir" being defined. Previously, snappy_unittests would read from an absolute path /testdata/..; convert it to use a relative path instead. Patch from Marc-Antonie Ruel. R=maruel ------------------------------------------------------------------------ r71 | [email protected] | 2013-01-18 13:16:36 +0100 (Fri, 18 Jan 2013) | 287 lines Increase the Zippy block size from 32 kB to 64 kB, winning ~3% density while being effectively performance neutral. The longer story about density is that we win 3-6% density on the benchmarks where this has any effect at all; many of the benchmarks (cp, c, lsp, man) are smaller than 32 kB and thus will have no effect. Binary data also seems to win little or nothing; of course, the already-compressed data wins nothing. The protobuf benchmark wins as much as ~18% depending on architecture, but I wouldn't be too sure that this is representative of protobuf data in general. As of performance, we lose a tiny amount since we get more tags (e.g., a long literal might be broken up into literal-copy-literal), but we win it back with less clearing of the hash table, and more opportunities to skip incompressible data (e.g. in the jpg benchmark). Decompression seems to get ever so slightly slower, again due to more tags. The total net change is about as close to zero as we can get, so the end effect seems to be simply more density and no real performance change. The comment about not changing kBlockSize, scary as it is, is not really relevant, since we're never going to have a block-level decompressor without explicitly marked blocks. Replace it with something more appropriate. This affects the framing format, but it's okay to change it since it basically has no users yet. ------------------------------------------------------------------------ r70 | [email protected] | 2013-01-06 20:21:26 +0100 (Sun, 06 Jan 2013) | 6 lines Adjust the Snappy open-source distribution for the changes in Google's internal file API. R=sanjay ------------------------------------------------------------------------ r69 | [email protected] | 2013-01-04 12:54:20 +0100 (Fri, 04 Jan 2013) | 15 lines Change a few ORs to additions where they don't matter. This helps the compiler use the LEA instruction more efficiently, since e.g. a + (b << 2) can be encoded as one instruction. Even more importantly, it can constant-fold the COPY_* enums together with the shifted negative constants, which also saves some instructions. (We don't need it for LITERAL, since it happens to be 0.) I am unsure why the compiler couldn't do this itself, but the theory is that it cannot prove that len-1 and len-4 cannot underflow/wrap, and thus can't do the optimization safely. The gains are small but measurable; 0.5-1.0% over the BM_Z* benchmarks (measured on Westmere, Sandy Bridge and Istanbul). R=sanjay ------------------------------------------------------------------------ r68 | [email protected] | 2012-10-08 13:37:16 +0200 (Mon, 08 Oct 2012) | 5 lines Stop giving -Werror to automake, due to an incompatibility between current versions of libtool and automake on non-GNU platforms (e.g. Mac OS X). R=sanjay ------------------------------------------------------------------------ r67 | [email protected] | 2012-08-17 15:54:47 +0200 (Fri, 17 Aug 2012) | 5 lines Fix public issue 66: Document GetUncompressedLength better, in particular that it leaves the source in a state that's not appropriate for RawUncompress. R=sanjay ------------------------------------------------------------------------ r66 | [email protected] | 2012-07-31 13:44:44 +0200 (Tue, 31 Jul 2012) | 5 lines Fix public issue 64: Check for <sys/time.h> at configure time, since MSVC seemingly does not have it. R=sanjay ------------------------------------------------------------------------ r65 | [email protected] | 2012-07-04 11:34:48 +0200 (Wed, 04 Jul 2012) | 10 lines Handle the case where gettimeofday() goes backwards or returns the same value twice; it could cause division by zero in the unit test framework. (We already had one fix for this in place, but it was incomplete.) This could in theory happen on any system, since there are few guarantees about gettimeofday(), but seems to only happen in practice on GNU/Hurd, where gettimeofday() is cached and only updated ever so often. R=sanjay ------------------------------------------------------------------------ r64 | [email protected] | 2012-07-04 11:28:33 +0200 (Wed, 04 Jul 2012) | 6 lines Mark ARMv4 as not supporting unaligned accesses (not just ARMv5 and ARMv6); apparently Debian still targets these by default, giving us segfaults on armel. R=sanjay ------------------------------------------------------------------------ r63 | [email protected] | 2012-05-22 11:46:05 +0200 (Tue, 22 May 2012) | 5 lines Fix public bug #62: Remove an extraneous comma at the end of an enum list, causing compile errors when embedded in Mozilla on OpenBSD. R=sanjay ------------------------------------------------------------------------ r62 | [email protected] | 2012-05-22 11:32:50 +0200 (Tue, 22 May 2012) | 8 lines Snappy library no longer depends on iostream. Achieved by moving logging macro definitions to a test-only header file, and by changing non-test code to use assert, fprintf, and abort instead of LOG/CHECK macros. R=sesse

Incompatible Changes: * can no longer use both `\' and `:' (didn't work anyway) New Features: * Handling of run-time errors (default: die) is now overridable by subclasses via signature_error_handler(). [github #54] * Can now have aliased named parameters. [github #57] * remove dependency on Devel::BeginLift [github #39] * can now use `when' to specify default conditions [github #48] * can use `//=' as a shortcut for `when undef' [github #45] * can now provide `where' constraints in addition to (or instead of) a type [github #7] * can now use `...' to disable further argument checking [github #49] * can now specify more than one alternative in type unions [github #55] * can now nest parameterized types Bug Fixes: * Removed experimental smartmatch warnings * Don't require Data::Alias for named params unless you have to [github #71] * Fixed obscure bug where an eval in Method::Signatures wouldn't be skipped when carp'ing (i.e. in carp_location_for()) [github #72] * Data::Alias is only loaded when needed avoiding a threads + eval bug in most cases and improving compile time performance. [rt.cpan.org 82922, github #62] * Compile-time errors now reporting proper line numbers. [github #61] * Trailing commas on parameter lists are now ok. [rt.cpan.org 81364] * Default condition of `when {}' now interpreted as `when { $_ ~~ {} }' (avoids parse error). [github #60] Optimizations: * better signature parsing using PPI [github #11] Distribution Fixes: * Fixed failing test in 5.10.0 (uncovered by CPAN Testers) * Fixed repo link in metadata (thanks dsteinbrunner) [github #87] * Add M::S::Parameter to MANIFEST [github #76] * Change representation of Infinity to work on Win32 [github #75] * Fixed stray detritus in MANIFEST. * Somehow my last-minute fix to the new error handler test didn't make it in; this will fix "Can't locate Moose.pm" errors. * Fixed test failing on 5.10.0 as per github #59. * Fixed subtests failing on Test::More's prior to 0.96. Docs: * Updated close parend problem to include quotes and a workaround [rt.cpan.org 85925] * Fixed some typos (thanks dsteinbrunner) [github #88] * Found and fixed missing parend * Minor clarifications here and there. * Clarified what doesn't work in Perl 5.8. * Added Function::Parameters to See Also section. * Updated copyright. * documented all new features * new ASCI-art breakdown of signature syntax * minor tweaks and corrections Misc: * Rearranged so signature is now an object [github #30] * Add hook for Travis CI [github #78] * Failure to parse parameters will now produce a more useful error.

FITS (Flexible Image Transport System) is a data format most used in astronomy. PyFITS is a Python module for reading, writing, and manipulating FITS files. The module uses Python's object-oriented features to provide quick, easy, and efficient access to FITS files. The use of Python's array syntax enables immediate access to any FITS extension, header cards, or data items. Changes to 2.4.0 (in py-pyfits): Changelog =========== 3.2 (2013-11-26) ---------------- Highlights ^^^^^^^^^^ - Rewrote CFITSIO-based backend for handling tile compression of FITS files. It now uses a standard CFITSIO instead of heavily modified pieces of CFITSIO as before. PyFITS ships with its own copy of CFITSIO v3.35 which supports the latest version of the Tiled Image Convention (v2.3), but system packagers may choose instead to strip this out in favor of a system-installed version of CFITSIO. Earlier versions may work, but nothing earlier than 3.28 has been tested yet. (#169) - Added support for reading and writing tables using the Q format for columns. The Q format is identical to the P format (variable-length arrays) except that it uses 64-bit integers for the data descriptors, allowing more than 4 GB of variable-length array data in a single table. (#160) - Added initial support for table columns containing pseudo-unsigned integers. This is currently enabled by using the ``uint=True`` option when opening files; any table columns with the correct BZERO value will be interpreted and returned as arrays of unsigned integers. - Some refactoring of the table and ``FITS_rec`` modules in order to better separate the details of the FITS binary and ASCII table data structures from the HDU data structures that encapsulate them. Most of these changes should not be apparent to users (but see API Changes below). API Changes ^^^^^^^^^^^ - Assigning to values in ``ColDefs.names``, ``ColDefs.formats``, ``ColDefs.nulls`` and other attributes of ``ColDefs`` instances that return lists of column properties is no longer supported. Assigning to those lists will no longer update the corresponding columns. Instead, please just modify the ``Column`` instances directly (``Column.name``, ``Column.null``, etc.) - The ``pyfits.new_table`` function is marked "pending deprecation". This does not mean it will be removed outright or that its functionality has changed. It will likely be replaced in the future for a function with similar, if not subtly different functionality. A better, if not slightly more verbose approach is to use ``pyfits.FITS_rec.from_columns`` to create a new ``FITS_rec`` table--this has the same interface as ``pyfits.new_table``. The difference is that it returns a plan ``FITS_rec`` array, and not an HDU instance. This ``FITS_rec`` object can then be used as the data argument in the constructors for ``BinTableHDU`` (for binary tables) or ``TableHDU`` (for ASCII tables). This is analogous to creating an ``ImageHDU`` by passing in an image array. ``pyfits.FITS_rec.from_columns`` is just a simpler way of creating a FITS-compatible recarray from a FITS column specification. - The ``updateHeader``, ``updateHeaderData``, and ``updateCompressedData`` methods of the ``CompDataHDU`` class are pending deprecation and moved to internal methods. The operation of these methods depended too much on internal state to be used safely by users; instead they are invoked automatically in the appropriate places when reading/writing compressed image HDUs. - The ``CompDataHDU.compData`` attribute is pending deprecation in favor of the clearer and more PEP-8 compatible ``CompDataHDU.compressed_data``. - The constructor for ``CompDataHDU`` has been changed to accept new keyword arguments. The new keyword arguments are essentially the same, but are in underscore_separated format rather than camelCase format. The old arguments are still pending deprecation. - The internal attributes of HDU classes ``_hdrLoc``, ``_datLoc``, and ``_datSpan`` have been replaced with ``_header_offset``, ``_data_offset``, and ``_data_size`` respectively. The old attribute names are still pending deprecation. This should only be of interest to advanced users who have created their own HDU subclasses. - The following previously deprecated functions and methods have been removed entirely: ``createCard``, ``createCardFromString``, ``upperKey``, ``ColDefs.data``, ``setExtensionNameCaseSensitive``, ``_File.getfile``, ``_TableBaseHDU.get_coldefs``, ``Header.has_key``, ``Header.ascardlist``. If you run your code with a previous version of PyFITS (>= 3.0, < 3.2) with the ``python -Wd`` argument, warnings for all deprecated interfaces still in use will be displayed. - Interfaces that were pending deprecation are now fully deprecated. These include: ``create_card``, ``create_card_from_string``, ``upper_key``, ``Header.get_history``, and ``Header.get_comment``. - The ``.name`` attribute on HDUs is now directly tied to the HDU's header, so that if ``.header['EXTNAME']`` changes so does ``.name`` and vice-versa. - The ``pyfits.file.PYTHON_MODES`` constant dict was renamed to ``pyfits.file.PYFITS_MODES`` which better reflects its purpose. This is rarely used by client code, however. Support for the old name will be removed by PyFITS 3.4. Other Changes and Additions ^^^^^^^^^^^^^^^^^^^^^^^^^^^ - The new compression code also adds support for the ZQUANTIZ and ZDITHER0 keywords added in more recent versions of this FITS Tile Compression spec. This includes support for lossless compression with GZIP. (#198) By default no dithering is used, but the ``SUBTRACTIVE_DITHER_1`` and ``SUBTRACTIVE_DITHER_2`` methods can be enabled by passing the correct constants to the ``quantize_method`` argument to the ``CompImageHDU`` constuctor. A seed can be manually specified, or automatically generated using either the system clock or checksum-based methods via the ``dither_seed`` argument. See the documentation for ``CompImageHDU`` for more details. (#198) (spacetelescope/PYFITS#32) - Images compressed with the Tile Compression standard can now be larger than 4 GB through support of the Q format. (#159) - All HDUs now have a ``.ver`` ``.level`` attribute that returns the value of the EXTVAL and EXTLEVEL keywords from that HDU's header, if the exist. This was added for consistency with the ``.name`` attribute which returns the EXTNAME value from the header. - Then ``Column`` and ``ColDefs`` classes have new ``.dtype`` attributes which give the Numpy dtype for the column data in the first case, and the full Numpy compound dtype for each table row in the latter case. - There was an issue where new tables created defaulted the values in all string columns to '0.0'. Now string columns are filled with empty strings by default--this seems a less surprising default, but it may cause differences with tables created with older versions of PyFITS. - Improved round-tripping and preservation of manually assigned column attributes (``TNULLn``, ``TSCALn``, etc.) in table HDU headers. (astropy/astropy#996) Bug Fixes ^^^^^^^^^ - Binary tables containing compressed images may, optionally, contain other columns unrelated to the tile compression convention. Although this is an uncommon use case, it is permitted by the standard. (#159) - Reworked some of the file I/O routines to allow simpler, more consistent mapping between OS-level file modes ('rb', 'wb', 'ab', etc.) and the more "PyFITS-specific" modes used by PyFITS like "readonly" and "update". That is, if reading a FITS file from an open file object, it doesn't matter as much what "mode" it was opened in so long as it has the right capabilities (read/write/etc.) Also works around bugs in the Python io module in 2.6+ with regard to file modes. (spacetelescope/PyFITS#33) - Fixed an obscure issue that can occur on systems that don't have flush to memory-mapped files implemented (namely GNU Hurd). (astropy/astropy#968) 3.1.3 (2013-11-26) ------------------ - Disallowed assigning NaN and Inf floating point values as header values, since the FITS standard does not define a way to represent them in. Because this is undefined, the previous behavior did not make sense and produced invalid FITS files. (spacetelescope/PyFITS#11) - Added a workaround for a bug in 64-bit OSX that could cause truncation when writing files greater than 2^32 bytes in size. (spacetelescope/PyFITS#28) - Fixed a long-standing issue where writing binary tables did not correctly write the TFORMn keywords for variable-length array columns (they ommitted the max array length parameter of the format). This was thought fixed in v3.1.2, but it was only fixed there for compressed image HDUs and not for binary tables in general. - Fixed an obscure issue that can occur on systems that don't have flush to memory-mapped files implemented (namely GNU Hurd). (Backported from 3.2) 3.0.12 (2013-11-26) ------------------- - Disallowed assigning NaN and Inf floating point values as header values, since the FITS standard does not define a way to represent them in. Because this is undefined, the previous behavior did not make sense and produced invalid FITS files. (Backported from 3.1.3) - Added a workaround for a bug in 64-bit OSX that could cause truncation when writing files greater than 2^32 bytes in size. (Backported from 3.1.3) - Fixed a long-standing issue where writing binary tables did not correctly write the TFORMn keywords for variable-length array columns (they ommitted the max array length parameter of the format). This was thought fixed in v3.1.2, but it was only fixed there for compressed image HDUs and not for binary tables in general. (Backported from 3.1.3) - Fixed an obscure issue that can occur on systems that don't have flush to memory-mapped files implemented (namely GNU Hurd). (Backported from 3.2) 3.1.2 (2013-04-22) ------------------ - When an error occurs opening a file in fitsdiff the exception message will now at least mention which file had the error. (#168) - Fixed support for opening gzipped FITS files by filename in a writeable mode (PyFITS has supported writing to gzip files for some time now, but only enabled it when GzipFile objects were passed to ``pyfits.open()`` due to some legacy code preventing full gzip support. (#195) - Added a more helpful error message in the case of malformatted FITS files that contain non-float NULL values in an ASCII table but are missing the required TNULLn keywords in the header. (#197) - Fixed an (apparently long-standing) issue where writing compressed images did not correctly write the TFORMn keywords for variable-length array columns (they ommitted the max array length parameter of the format). (#199) - Slightly refactored how tables containing variable-length array columns are handled to add two improvements: Fixes an issue where accessing the data after a call to the `pyfits.getdata` convenience function caused an exception, and allows the VLA data to be read from an existing mmap of the FITS file. (#200) - Fixed a bug that could occur when opening a table containing multi-dimensional columns (i.e. via the TDIMn keyword) and then writing it out to a new file. (#201) - Added use of the console_scripts entry point to install the fitsdiff and fitscheck scripts, which if nothing else provides better Windows support. The generated scripts now override the ones explicitly defined in the scripts/ directory (which were just trivial stubs to begin with). (#202) - Fixed a bug on Python 3 where attempting to open a non-existent file on Python 3 caused a seemingly unrelated traceback. (#203) - Fixed a bug in fitsdiff that reported two header keywords containing NaN as value as different. (#204) - Fixed an issue in the tests that caused some tests to fail if pyfits is installed with read-only permissions. (#208) - Fixed a bug where instantiating a ``BinTableHDU`` from a numpy array containing boolean fields converted all the values to ``False``. (#215) - Fixed an issue where passing an array of integers into the constructor of ``Column()`` when the column type is floats of the same byte width caused the column array to become garbled. (#218) - Fixed inconsistent behavior in creating CONTINUE cards from byte strings versus unicode strings in Python 2--CONTINUE cards can now be created properly from unicode strings (so long as they are convertable to ASCII). (spacetelescope/PyFITS#1) - Fixed a couple cases where creating a new table using TDIMn in some of the columns could caused a crash. (spacetelescope/PyFITS#3) - Fixed a bug in parsing HIERARCH keywords that do not have a space after the first equals sign (before the value). (spacetelescope/PyFITS#5) - Prevented extra leading whitespace on HIERARCH keywords from being treated as part of the keyword. (spacetelescope/PyFITS#6) - Fixed a bug where HIERARCH keywords containing lower-case letters was mistakenly marked as invalid during header validation. (spacetelescope/PyFITS#7) - Fixed an issue that was ancillary to (spacetelescope/PyFITS#7) where the ``Header.index()`` method did not work correctly with HIERARCH keywords containing lower-case letters. 3.0.11 (2013-04-17) ------------------- - Fixed support for opening gzipped FITS files by filename in a writeable mode (PyFITS has supported writing to gzip files for some time now, but only enabled it when GzipFile objects were passed to ``pyfits.open()`` due to some legacy code preventing full gzip support. Backported from 3.1.2. (#195) - Added a more helpful error message in the case of malformatted FITS files that contain non-float NULL values in an ASCII table but are missing the required TNULLn keywords in the header. Backported from 3.1.2. (#197) - Fixed an (apparently long-standing) issue where writing compressed images did not correctly write the TFORMn keywords for variable-length array columns (they ommitted the max array length parameter of the format). Backported from 3.1.2. (#199) - Slightly refactored how tables containing variable-length array columns are handled to add two improvements: Fixes an issue where accessing the data after a call to the `pyfits.getdata` convenience function caused an exception, and allows the VLA data to be read from an existing mmap of the FITS file. Backported from 3.1.2. (#200) - Fixed a bug that could occur when opening a table containing multi-dimensional columns (i.e. via the TDIMn keyword) and then writing it out to a new file. Backported from 3.1.2. (#201) - Fixed a bug on Python 3 where attempting to open a non-existent file on Python 3 caused a seemingly unrelated traceback. Backported from 3.1.2. (#203) - Fixed a bug in fitsdiff that reported two header keywords containing NaN as value as different. Backported from 3.1.2. (#204) - Fixed an issue in the tests that caused some tests to fail if pyfits is installed with read-only permissions. Backported from 3.1.2. (#208) - Fixed a bug where instantiating a ``BinTableHDU`` from a numpy array containing boolean fields converted all the values to ``False``. Backported from 3.1.2. (#215) - Fixed an issue where passing an array of integers into the constructor of ``Column()`` when the column type is floats of the same byte width caused the column array to become garbled. Backported from 3.1.2. (#218) - Fixed a couple cases where creating a new table using TDIMn in some of the columns could caused a crash. Backported from 3.1.2. (spacetelescope/PyFITS#3) 3.1.1 (2013-01-02) ------------------ This is a bug fix release for the 3.1.x series. Bug Fixes ^^^^^^^^^ - Improved handling of scaled images and pseudo-unsigned integer images in compressed image HDUs. They now work more transparently like normal image HDUs with support for the ``do_not_scale_image_data`` and ``uint`` options, as well as ``scale_back`` and ``save_backup``. The ``.scale()`` method works better too. (#88) - Permits non-string values for the EXTNAME keyword when reading in a file, rather than throwing an exception due to the malformatting. Added verification for the format of the EXTNAME keyword when writing. (#96) - Added support for EXTNAME and EXTVER in PRIMARY HDUs. That is, if EXTNAME is specified in the header, it will also be reflected in the ``.name`` attribute and in ``pyfits.info()``. These keywords used to be verboten in PRIMARY HDUs, but the latest version of the FITS standard allows them. (#151) - HCOMPRESS can again be used to compress data cubes (and higher-dimensional arrays) so long as the tile size is effectively 2-dimensional. In fact, PyFITS will automatically use compatible tile sizes even if they're not explicitly specified. (#171) - Added support for the optional ``endcard`` parameter in the ``Header.fromtextfile()`` and ``Header.totextfile()`` methods. Although ``endcard=False`` was a reasonable default assumption, there are still text dumps of FITS headers that include the END card, so this should have been more flexible. (#176) - Fixed a crash when running fitsdiff on two empty (that is, zero row) tables. (#178) - Fixed an issue where opening files containing random groups HDUs in update mode could cause an unnecessary rewrite of the file even if none of the data is modified. (#179) - Fixed a bug that could caused a deadlock in the filesystem on OSX if PyFITS is used with Numpy 1.7 in some cases. (#180) - Fixed a crash when generating diff reports from diffs using the ``ignore_comments`` options. (#181) - Fixed some bugs with WCS Paper IV record-valued keyword cards: - Cards that looked kind of like RVKCs but were not intended to be were over-permissively treated as such--commentary keywords like COMMENT and HISTORY were particularly affected. (#183) - Looking up a card in a header by its standard FITS keyword only should always return the raw value of that card. That way cards containing values that happen to valid RVKCs but were not intended to be will still be treated like normal cards. (#184) - Looking up a RVKC in a header with only part of the field-specifier (for example "DP1.AXIS" instead of "DP1.AXIS.1") was implicitly treated as a wildcard lookup. (#184) - Fixed a crash when diffing two FITS files where at least one contains a compressed image HDU which was not recognized as an image instead of a table. (#187) - Fixed bugs in the backwards compatibility layer for the ``CardList.index`` and ``CardList.count`` methods. (#190) - Improved ``__repr__`` and text file representation of cards with long values that are split into CONTINUE cards. (#193) - Fixed a crash when trying to assign a long (> 72 character) value to blank ('') keywords. This also changed how blank keywords are represented--there are still exactly 8 spaces before any commentary content can begin; this *may* affect the exact display of header cards that assumed there could be fewer spaces in a blank keyword card before the content begins. However, the current approach is more in line with the requirements of the FITS standard. (#194) 3.0.10 (2013-01-02) ------------------- - Improved handling of scaled images and pseudo-unsigned integer images in compressed image HDUs. They now work more transparently like normal image HDUs with support for the ``do_not_scale_image_data`` and ``uint`` options, as well as ``scale_back`` and ``save_backup``. The ``.scale()`` method works better too. Backported from 3.1.1. (#88) - Permits non-string values for the EXTNAME keyword when reading in a file, rather than throwing an exception due to the malformatting. Added verification for the format of the EXTNAME keyword when writing. Backported from 3.1.1. (#96) - Added support for EXTNAME and EXTVER in PRIMARY HDUs. That is, if EXTNAME is specified in the header, it will also be reflected in the ``.name`` attribute and in ``pyfits.info()``. These keywords used to be verbotten in PRIMARY HDUs, but the latest version of the FITS standard allows them. Backported from 3.1.1. (#151) - HCOMPRESS can again be used to compress data cubes (and higher-dimensional arrays) so long as the tile size is effectively 2-dimensional. In fact, PyFITS will not automatically use compatible tile sizes even if they're not explicitly specified. Backported from 3.1.1. (#171) - Fixed a bug when writing out files containing zero-width table columns, where the TFIELDS keyword would be updated incorrectly, leaving the table largely unreadable. Backported from 3.1.0. (#174) - Fixed an issue where opening files containing random groups HDUs in update mode could cause an unnecessary rewrite of the file even if none of the data is modified. Backported from 3.1.1. (#179) - Fixed a bug that could caused a deadlock in the filesystem on OSX if PyFITS is used with Numpy 1.7 in some cases. Backported from 3.1.1. (#180) 3.1 (2012-08-08) ---------------- Highlights ^^^^^^^^^^ - The ``Header`` object has been significantly reworked, and ``CardList`` objects are now deprecated (their functionality folded into the ``Header`` class). See API Changes below for more details. - Memory maps are now used by default to access HDU data. See API Changes below for more details. - Now includes a new version of the ``fitsdiff`` program for comparing two FITS files, and a new FITS comparison API used by ``fitsdiff``. See New Features below. API Changes ^^^^^^^^^^^ - The ``Header`` class has been rewritten, and the ``CardList`` class is deprecated. Most of the basic details of working with FITS headers are unchanged, and will not be noticed by most users. But there are differences in some areas that will be of interest to advanced users, and to application developers. For full details of the changes, see the "Header Interface Transition Guide" section in the PyFITS documentation. See ticket #64 on the PyFITS Trac for futher details and background. Some highlights are listed below: * The Header class now fully implements the Python dict interface, and can be used interchangably with a dict, where the keys are header keywords. * New keywords can be added to the header using normal keyword assignment (previously it was necessary to use ``Header.update`` to add new keywords). For example:: >>> header['NAXIS'] = 2 will update the existing 'FOO' keyword if it already exists, or add a new one if it doesn't exist, just like a dict. * It is possible to assign both a value and a comment at the same time using a tuple:: >>> header['NAXIS'] = (2, 'Number of axes') * To add/update a new card and ensure it's added in a specific location, use ``Header.set()``:: >>> header.set('NAXIS', 2, 'Number of axes', after='BITPIX') This works the same as the old ``Header.update()``. ``Header.update()`` still works in the old way too, but is deprecated. * Although ``Card`` objects still exist, it generally is not necessary to work with them directly. ``Header.ascardlist()``/``Header.ascard`` are deprecated and should not be used. To directly access the ``Card`` objects in a header, use ``Header.cards``. * To access card comments, it is still possible to either go through the card itself, or through ``Header.comments``. For example:: >>> header.cards['NAXIS'].comment Number of axes >>> header.comments['NAXIS'] Number of axes * ``Card`` objects can now be used interchangeably with ``(keyword, value, comment)`` 3-tuples. They still have ``.value`` and ``.comment`` attributes as well. The ``.key`` attribute has been renamed to ``.keyword`` for consistency, though ``.key`` is still supported (but deprecated). - Memory mapping is now used by default to access HDU data. That is, ``pyfits.open()`` uses ``memmap=True`` as the default. This provides better performance in the majority of use cases--there are only some I/O intensive applications where it might not be desirable. Enabling mmap by default also enabled finding and fixing a large number of bugs in PyFITS' handling of memory-mapped data (most of these bug fixes were backported to PyFITS 3.0.5). (#85) * A new ``pyfits.USE_MEMMAP`` global variable was added. Set ``pyfits.USE_MEMMAP = False`` to change the default memmap setting for opening files. This is especially useful for controlling the behavior in applications where pyfits is deeply embedded. * Likewise, a new ``PYFITS_USE_MEMMAP`` environment variable is supported. Set ``PYFITS_USE_MEMMAP = 0`` in your environment to change the default behavior. - The ``size()`` method on HDU objects is now a ``.size`` property--this returns the size in bytes of the data portion of the HDU, and in most cases is equivalent to ``hdu.data.nbytes`` (#83) - ``BinTableHDU.tdump`` and ``BinTableHDU.tcreate`` are deprecated--use ``BinTableHDU.dump`` and ``BinTableHDU.load`` instead. The new methods output the table data in a slightly different format from previous versions, which places quotes around each value. This format is compatible with data dumps from previous versions of PyFITS, but not vice-versa due to a parsing bug in older versions. - Likewise the ``pyfits.tdump`` and ``pyfits.tcreate`` convenience function versions of these methods have been renamed ``pyfits.tabledump`` and ``pyfits.tableload``. The old deprecated, but currently retained for backwards compatibility. (r1125) - A new global variable ``pyfits.EXTENSION_NAME_CASE_SENSITIVE`` was added. This serves as a replacement for ``pyfits.setExtensionNameCaseSensitive`` which is not deprecated and may be removed in a future version. To enable case-sensitivity of extension names (i.e. treat 'sci' as distict from 'SCI') set ``pyfits.EXTENSION_NAME_CASE_SENSITIVE = True``. The default is ``False``. (r1139) - A new global configuration variable ``pyfits.STRIP_HEADER_WHITESPACE`` was added. By default, if a string value in a header contains trailing whitespace, that whitespace is automatically removed when the value is read. Now if you set ``pyfits.STRIP_HEADER_WHITESPACE = False`` all whitespace is preserved. (#146) - The old ``classExtensions`` extension mechanism (which was deprecated in PyFITS 3.0) is removed outright. To our knowledge it was no longer used anywhere. (r1309) - Warning messages from PyFITS issued through the Python warnings API are now output to stderr instead of stdout, as is the default. PyFITS no longer modifies the default behavior of the warnings module with respect to which stream it outputs to. (r1319) - The ``checksum`` argument to ``pyfits.open()`` now accepts a value of 'remove', which causes any existing CHECKSUM/DATASUM keywords to be ignored, and removed when the file is saved. New Features ^^^^^^^^^^^^ - Added support for the proposed "FITS" extension HDU type. See http://listmgr.cv.nrao.edu/pipermail/fitsbits/2002-April/001094.html. FITS HDUs contain an entire FITS file embedded in their data section. `FitsHDU` objects work like other HDU types in PyFITS. Their ``.data`` attribute returns the raw data array. However, they have a special ``.hdulist`` attribute which processes the data as a FITS file and returns it as an in-memory HDUList object. FitsHDU objects also support a ``FitsHDU.fromhdulist()`` classmethod which returns a new `FitsHDU` object that embeds the supplied HDUList. (#80) - Added a new ``.is_image`` attribute on HDU objects, which is True if the HDU data is an 'image' as opposed to a table or something else. Here the meaning of 'image' is fairly loose, and mostly just means a Primary or Image extension HDU, or possibly a compressed image HDU (#71) - Added an ``HDUList.fromstring`` classmethod which can parse a FITS file already in memory and instantiate and ``HDUList`` object from it. This could be useful for integrating PyFITS with other libraries that work on FITS file, such as CFITSIO. It may also be useful in streaming applications. The name is a slight misnomer, in that it actually accepts any Python object that implements the buffer interface, which includes ``bytes``, ``bytearray``, ``memoryview``, ``numpy.ndarray``, etc. (#90) - Added a new ``pyfits.diff`` module which contains facilities for comparing FITS files. One can use the ``pyfits.diff.FITSDiff`` class to compare two FITS files in their entirety. There is also a ``pyfits.diff.HeaderDiff`` class for just comparing two FITS headers, and other similar interfaces. See the PyFITS Documentation for more details on this interface. The ``pyfits.diff`` module powers the new ``fitsdiff`` program installed with PyFITS. After installing PyFITS, run ``fitsdiff --help`` for usage details. - ``pyfits.open()`` now accepts a ``scale_back`` argument. If set to ``True``, this automatically scales the data using the original BZERO and BSCALE parameters the file had when it was first opened, if any, as well as the original BITPIX. For example, if the original BITPIX were 16, this would be equivalent to calling ``hdu.scale('int16', 'old')`` just before calling ``flush()`` or ``close()`` on the file. This option applies to all HDUs in the file. (#120) - ``pyfits.open()`` now accepts a ``save_backup`` argument. If set to ``True``, this automatically saves a backup of the original file before flushing any changes to it (this of course only applies to update and append mode). This may be especially useful when working with scaled image data. (#121) Changes in Behavior ^^^^^^^^^^^^^^^^^^^ - Warnings from PyFITS are not output to stderr by default, instead of stdout as it has been for some time. This is contrary to most users' expectations and makes it more difficult for them to separate output from PyFITS from the desired output for their scripts. (r1319) Bug Fixes ^^^^^^^^^ - Fixed ``pyfits.tcreate()`` (now ``pyfits.tableload()``) to be more robust when encountering blank lines in a column definition file (#14) - Fixed a fairly rare crash that could occur in the handling of CONTINUE cards when using Numpy 1.4 or lower (though 1.4 is the oldest version supported by PyFITS). (r1330) - Fixed ``_BaseHDU.fromstring`` to actually correctly instantiate an HDU object from a string/buffer containing the header and data of that HDU. This allowed for the implementation of ``HDUList.fromstring`` described above. (#90) - Fixed a rare corner case where, in some use cases, (mildly, recoverably) malformatted float values in headers were not properly returned as floats. (#137) - Fixed a corollary to the previous bug where float values with a leading zero before the decimal point had the leading zero unnecessarily removed when saving changes to the file (eg. "0.001" would be written back as ".001" even if no changes were otherwise made to the file). (#137) - When opening a file containing CHECKSUM and/or DATASUM keywords in update mode, the CHECKSUM/DATASUM are updated and preserved even if the file was opened with checksum=False. This change in behavior prevents checksums from being unintentionally removed. (#148) - Fixed a bug where ``ImageHDU.scale(option='old')`` wasn't working at all--it was not restoring the image to its original BSCALE and BZERO values. (#162) - Fixed a bug when writing out files containing zero-width table columns, where the TFIELDS keyword would be updated incorrectly, leaving the table largely unreadable. This fix will be backported to the 3.0.x series in version 3.0.10. (#174) 3.0.9 (2012-08-06) ------------------ This is a bug fix release for the 3.0.x series. Bug Fixes ^^^^^^^^^ - Fixed ``Header.values()``/``Header.itervalues()`` and ``Header.items()``/ ``Header.iteritems()`` to correctly return the different values for duplicate keywords (particularly commentary keywords like HISTORY and COMMENT). This makes the old Header implementation slightly more compatible with the new implementation in PyFITS 3.1. (#127) .. note:: This fix did not change the existing behavior from earlier PyFITS versions where ``Header.keys()`` returns all keywords in the header with duplicates removed. PyFITS 3.1 changes that behavior, so that ``Header.keys()`` includes duplicates. - Fixed a bug where ``ImageHDU.scale(option='old')`` wasn't working at all--it was not restoring the image to its original BSCALE and BZERO values. (#162) - Fixed a bug where opening a file containing compressed image HDUs in 'update' mode and then immediately closing it without making any changes caused the file to be rewritten unncessarily. (#167) - Fixed two memory leaks that could occur when writing compressed image data, or in some cases when opening files containing compressed image HDUs in 'update' mode. (#168) 3.0.8 (2012-06-04) ------------------ Changes in Behavior ^^^^^^^^^^^^^^^^^^^ - Prior to this release, image data sections did not work with scaled data--that is, images with non-trivial BSCALE and/or BZERO values. Previously, in order to read such images in sections, it was necessary to manually apply the BSCALE+BZERO to each section. It's worth noting that sections *did* support pseudo-unsigned ints (flakily). This change just extends that support for general BSCALE+BZERO values. Bug Fixes ^^^^^^^^^ - Fixed a bug that prevented updates to values in boolean table columns from being saved. This turned out to be a symptom of a deeper problem that could prevent other table updates from being saved as well. (#139) - Fixed a corner case in which a keyword comment ending with the string "END" could, in some circumstances, cause headers (and the rest of the file after that point) to be misread. (#142) - Fixed support for scaled image data and psuedo-unsigned ints in image data sections (``hdu.section``). Previously this was not supported at all. At some point support was supposedly added, but it was buggy and incomplete. Now the feature seems to work much better. (#143) - Fixed the documentation to point out that image data sections *do* support non-contiguous slices (and have for a long time). The documentation was never updated to reflect this, and misinformed users that only contiguous slices were supported, leading to some confusion. (#144) - Fixed a bug where creating an ``HDUList`` object containing multiple PRIMARY HDUs caused an infinite recursion when validating the object prior to writing to a file. (#145) - Fixed a rare but serious case where saving an update to a file that previously had a CHECKSUM and/or DATASUM keyword, but removed the checksum in saving, could cause the file to be slightly corrupted and unreadable. (#147) - Fixed problems with reading "non-standard" FITS files with primary headers containing SIMPLE = F. PyFITS has never made many guarantees as to how such files are handled. But it should at least be possible to read their headers, and the data if possible. Saving changes to such a file should not try to prepend an unwanted valid PRIMARY HDU. (#157) - Fixed a bug where opening an image with ``disable_image_compression = True`` caused compression to be disabled for all subsequent ``pyfits.open()`` calls. (r1651) 3.0.7 (2012-04-10) ------------------ Changes in Behavior ^^^^^^^^^^^^^^^^^^^ - Slices of GroupData objects now return new GroupData objects instead of extended multi-row _Group objects. This is analogous to how PyFITS 3.0 fixed FITS_rec slicing, and should have been fixed for GroupData at the same time. The old behavior caused bugs where functions internal to Numpy expected that slicing an ndarray would return a new ndarray. As this is a rare usecase with a rare feature most users are unlikely to be affected by this change. - The previously internal _Group object for representing individual group records in a GroupData object are renamed Group and are now a public interface. However, there's almost no good reason to create Group objects directly, so it shouldn't be considered a "new feature". - An annoyance from PyFITS 3.0.6 was fixed, where the value of the EXTEND keyword was always being set to F if there are not actually any extension HDUs. It was unnecessary to modify this value. Bug Fixes ^^^^^^^^^ - Fixed GroupData objects to return new GroupData objects when sliced instead of _Group record objects. See "Changes in behavior" above for more details. - Fixed slicing of Group objects--previously it was not possible to slice slice them at all. - Made it possible to assign `np.bool_` objects as header values. (#123) - Fixed overly strict handling of the EXTEND keyword; see "Changes in behavior" above. (#124) - Fixed many cases where an HDU's header would be marked as "modified" by PyFITS and rewritten, even when no changes to the header are necessary. (#125) - Fixed a bug where the values of the PTYPEn keywords in a random groups HDU were forced to be all lower-case when saving the file. (#130) - Removed an unnecessary inline import in `ExtensionHDU.__setattr__` that was causing some slowdown when opening files containing a large number of extensions, plus a few other small (but not insignficant) performance improvements thanks to Julian Taylor. (#133) - Fixed a regression where header blocks containing invalid end-of-header padding (i.e. null bytes instead of spaces) couldn't be parsed by PyFITS. Such headers can be parsed again, but a warning is raised, as such headers are not valid FITS. (#136) - Fixed a memory leak where table data in random groups HDUs weren't being garbage collected. (#138) 3.0.6 (2012-02-29) ------------------ Highlights ^^^^^^^^^^ The main reason for this release is to fix an issue that was introduced in PyFITS 3.0.5 where merely opening a file containing scaled data (that is, with non-trivial BSCALE and BZERO keywords) in 'update' mode would cause the data to be automatically rescaled--possibly converting the data from ints to floats--as soon as the file is closed, even if the application did not touch the data. Now PyFITS will only rescale the data in an extension when the data is actually accessed by the application. So opening a file in 'update' mode in order to modify the header or append new extensions will not cause any change to the data in existing extensions. This release also fixes a few Windows-specific bugs found through more extensive Windows testing, and other miscellaneous bugs. Bug Fixes ^^^^^^^^^ - More accurate error messages when opening files containing invalid header cards. (#109) - Fixed a possible reference cycle/memory leak that was caught through more extensive testing on Windows. (#112) - Fixed 'ostream' mode to open the underlying file in 'wb' mode instead of 'w' mode. (#112) - Fixed a Windows-only issue where trying to save updates to a resized FITS file could result in a crash due to there being open mmaps on that file. (#112) - Fixed a crash when trying to create a FITS table (i.e. with new_table()) from a Numpy array containing bool fields. (#113) - Fixed a bug where manually initializing an ``HDUList`` with a list of of HDUs wouldn't set the correct EXTEND keyword value on the primary HDU. (#114) - Fixed a crash that could occur when trying to deepcopy a Header in Python < 2.7. (#115) - Fixed an issue where merely opening a scaled image in 'update' mode would cause the data to be converted to floats when the file is closed. (#119) 3.0.5 (2012-01-30) ------------------ - Fixed a crash that could occur when accessing image sections of files opened with memmap=True. (r1211) - Fixed the inconsistency in the behavior of files opened in 'readonly' mode when memmap=True vs. when memmap=False. In the latter case, although changes to array data were not saved to disk, it was possible to update the array data in memory. On the other hand with memmap=True, 'readonly' mode prevented even in-memory modification to the data. This is what 'copyonwrite' mode was for, but difference in behavior was confusing. Now 'readonly' is equivalent to 'copyonwrite' when using memmap. If the old behavior of denying changes to the array data is necessary, a new 'denywrite' mode may be used, though it is only applicable to files opened with memmap. (r1275) - Fixed an issue where files opened with memmap=True would return image data as a raw numpy.memmap object, which can cause some unexpected behaviors--instead memmap object is viewed as a numpy.ndarray. (r1285) - Fixed an issue in Python 3 where a workaround for a bug in Numpy on Python 3 interacted badly with some other software, namely to vo.table package (and possibly others). (r1320, r1337, and #110) - Fixed buggy behavior in the handling of SIGINTs (i.e. Ctrl-C keyboard interrupts) while flushing changes to a FITS file. PyFITS already prevented SIGINTs from causing an incomplete flush, but did not clean up the signal handlers properly afterwards, or reraise the keyboard interrupt once the flush was complete. (r1321) - Fixed a crash that could occur in Python 3 when opening files with checksum checking enabled. (r1336) - Fixed a small bug that could cause a crash in the `StreamingHDU` interface when using Numpy below version 1.5. - Fixed a crash that could occur when creating a new `CompImageHDU` from an array of big-endian data. (#104) - Fixed a crash when opening a file with extra zero padding at the end. Though FITS files should not have such padding, it's not explictly forbidden by the format either, and PyFITS shouldn't stumble over it. (#106) - Fixed a major slowdown in opening tables containing large columns of string values. (#111) 3.0.4 (2011-11-22) ------------------ - Fixed a crash when writing HCOMPRESS compressed images that could happen on Python 2.5 and 2.6. (r1217) - Fixed a crash when slicing an table in a file opened in 'readonly' mode with memmap=True. (r1230) - Writing changes to a file or writing to a new file verifies the output in 'fix' mode by default instead of 'exception'--that is, PyFITS will automatically fix common FITS format errors rather than raising an exception. (r1243) - Fixed a bug where convenience functions such as getval() and getheader() crashed when specifying just 'PRIMARY' as the extension to use (r1263). - Fixed a bug that prevented passing keyword arguments (beyond the standard data and header arguments) as positional arguments to the constructors of extension HDU classes. - Fixed some tests that were failing on Windows--in this case the tests themselves failed to close some temp files and Windows refused to delete them while there were still open handles on them. (r1295) - Fixed an issue with floating point formatting in header values on Python 2.5 for Windows (and possibly other platforms). The exponent was zero-padded to 3 digits; although the FITS standard makes no specification on this, the formatting is now normalized to always pad the exponent to two digits. (r1295) - Fixed a bug where long commentary cards (such as HISTORY and COMMENT) were broken into multiple CONTINUE cards. However, commentary cards are not expected to be found in CONTINUE cards. Instead these long cards are broken into multiple commentary cards. (#97) - GZIP/ZIP-compressed FITS files can be detected and opened regardless of their filename extension. (#99) - Fixed a serious bug where opening scaled images in 'update' mode and then closing the file without touching the data would cause the file to be corrupted. (#101) 3.0.3 (2011-10-05) ------------------ - Fixed several small bugs involving corner cases in record-valued keyword cards (#70) - In some cases HDU creation failed if the first keyword value in the header was not a string value (#89) - Fixed a crash when trying to compute the HDU checksum when the data array contains an odd number of bytes (#91) - Disabled an unnecessary warning that was displayed on opening compressed HDUs with disable_image_compression = True (#92) - Fixed a typo in code for handling HCOMPRESS compressed images. 3.0.2 (2011-09-23) ------------------ - The ``BinTableHDU.tcreate`` method and by extension the ``pyfits.tcreate`` function don't get tripped up by blank lines anymore (#14) - The presence, value, and position of the EXTEND keyword in Primary HDUs is verified when reading/writing a FITS file (#32) - Improved documentation (in warning messages as well as in the handbook) that PyFITS uses zero-based indexing (as one would expect for C/Python code, but contrary to the PyFITS standard which was written with FORTRAN in mind) (#68) - Fixed a bug where updating a header card comment could cause the value to be lost if it had not already been read from the card image string. - Fixed a related bug where changes made directly to Card object in a header (i.e. assigning directly to card.value or card.comment) would not propagate when flushing changes to the file (#69) [Note: This and the bug above it were originally reported as being fixed in version 3.0.1, but the fix was never included in the release.] - Improved file handling, particularly in Python 3 which had a few small file I/O-related bugs (#76) - Fixed a bug where updating a FITS file would sometimes cause it to lose its original file permissions (#79) - Fixed the handling of TDIMn keywords; 3.0 added support for them, but got the axis order backards (they were treated as though they were row-major) (#82) - Fixed a crash when a FITS file containing scaled data is opened and immediately written to a new file without explicitly viewing the data first (#84) - Fixed a bug where creating a table with columns named either 'names' or 'formats' resulted in an infinite recursion (#86) 3.0.1 (2011-09-12) ------------------ - Fixed a bug where updating a header card comment could cause the value to be lost if it had not already been read from the card image string. - Changed ``_TableBaseHDU.data`` so that if the data contain an empty table a ``FITS_rec`` object with zero rows is returned rather than ``None``. - The ``.key`` attribute of ``RecordValuedKeywordCards`` now returns the full keyword+field-specifier value, instead of just the plain keyword (#46) - Fixed a related bug where changes made directly to Card object in a header (i.e. assigning directly to card.value or card.comment) would not propagate when flushing changes to the file (#69) - Fixed a bug where writing a table with zero rows could fail in some cases (#72) - Miscellanous small bug fixes that were causing some tests to fail, particularly on Python 3 (#74, #75) - Fixed a bug where creating a table column from an array in non-native byte order would not preserve the byte order, thus interpreting the column array using the wrong byte order (#77) 3.0.0 (2011-08-23) -------------------- - Contains major changes, bumping the version to 3.0 - Large amounts of refactoring and reorganization of the code; tried to preserve public API backwards-compatibility with older versions (private API has many changes and is not guaranteed to be backwards-compatible). There are a few small public API changes to be aware of: * The pyfits.rec module has been removed completely. If your version of numpy does not have the numpy.core.records module it is too old to be used with PyFITS. * The ``Header.ascardlist()`` method is deprecated--use the ``.ascard`` attribute instead. * ``Card`` instances have a new ``.cardimage`` attribute that should be used rather than ``.ascardimage()``, which may become deprecated. * The ``Card.fromstring()`` method is now a classmethod. It returns a new ``Card`` instance rather than modifying an existing instance. * The ``req_cards()`` method on HDU instances has changed: The ``pos`` argument is not longer a string. It is either an integer value (meaning the card's position must match that value) or it can be a function that takes the card's position as it's argument, and returns True if the position is valid. Likewise, the ``test`` argument no longer takes a string, but instead a function that validates the card's value and returns True or False. * The ``get_coldefs()`` method of table HDUs is deprecated. Use the ``.columns`` attribute instead. * The ``ColDefs.data`` attribute is deprecated--use ``ColDefs.columns`` instead (though in general you shouldn't mess with it directly--it might become internal at some point). * ``FITS_record`` objects take ``start`` and ``end`` as arguments instead of ``startColumn`` and ``endColumn`` (these are rarely created manually, so it's unlikely that this change will affect anyone). * ``BinTableHDU.tcreate()`` is now a classmethod, and returns a new ``BinTableHDU`` instance. * Use ``ExtensionHDU`` and ``NonstandardExtHDU`` for making new extension HDU classes. They are now public interfaces, wheres previously they were private and prefixed with underscores. * Possibly others--please report if you find any changes that cause difficulties. - Calls to deprecated functions will display a Deprecation warning. However, in Python 2.7 and up Deprecation warnings are ignored by default, so run Python with the `-Wd` option to see if you're using any deprecated functions. If we get close to actually removing any functions, we might make the Deprecation warnings display by default. - Added basic Python 3 support - Added support for multi-dimensional columns in tables as specified by the TDIMn keywords (#47) - Fixed a major memory leak that occurred when creating new tables with the ``new_table()`` function (#49) be padded with zero-bytes) vs ASCII tables (where strings are padded with spaces) (#15) - Fixed a bug in which the case of Random Access Group parameters names was not preserved when writing (#41) - Added support for binary table fields with zero width (#42) - Added support for wider integer types in ASCII tables; although this is non- standard, some GEIS images require it (#45) - Fixed a bug that caused the index_of() method of HDULists to crash when the HDUList object is created from scratch (#48) - Fixed the behavior of string padding in binary tables (where strings should be padded with nulls instead of spaces) - Fixed a rare issue that caused excessive memory usage when computing checksums using a non-standard block size (see r818) - Add support for forced uint data in image sections (#53) - Fixed an issue where variable-length array columns were not extended when creating a new table with more rows than the original (#54) - Fixed tuple and list-based indexing of FITS_rec objects (#55) - Fixed an issue where BZERO and BSCALE keywords were appended to headers in the wrong location (#56) - ``FITS_record`` objects (table rows) have full slicing support, including stepping, etc. (#59) - Fixed a bug where updating multiple files simultaneously (such as when running parallel processes) could lead to a race condition with mktemp() (#61) - Fixed a bug where compressed image headers were not in the order expected by the funpack utility (#62)

Changes since 1.1.0 (from git log): issue #70: boolean flags may have true default Merge pull request #69 from Tieske/master general updates, mainly added CONTRIBUTING.md Merge pull request #1 from mark-otaris/patch-1 Caught red handed, after a shameless copy 😄. Thx for the fix This is Penlight, not Busted added - a shameless copy of busted - CONTRIBUTING.md file renamed docs to doc, to fix luarocks warning updated readme to md format changed .txt to .md files added rockspec, fixed typo in filename Merge pull request #67 from Tieske/add_executeex added utils.executeex, which also returns output of stdout and stderr updated, using readfile now, remove obsolete os.remove commands Merge pull request #66 from Tieske/fix_temp windows environment variable TMP is deprecated, use TEMP mode binary mode optional (binary would require platform specific line endings to be converted which is not very portable, and mostly just text will do) update: read files in binary mode added utils.executeex, which also returns output of stdout and stderr windows environment variable TMP is deprecated, use TEMP HTML mode skips DOCTYPE; no globals harmed outside pl.utils Merge pull request #64 from Tieske/fix_dir fixed makepath creation, added/updated some documentation fixed dirpath creation, added/updated some documentation basic parser handles comments, and HTML mode is working nicely with real-world HTML use LOM ordered attributes when stringifying if present (Danny). manual merge of John Schember's readonly table support Merge pull request #61 from user-none/isempty Thanks, John! Merge pull request #62 from CoolisTheName007/patch-2 Transform using values as keys. missing locals Set.issempty mispelling; default ctor borked; more tests Transform using values as keys. Merge remote-tracking branch 'upstream/master' into empty Add function to check if an object is empty. Merge pull request #53 from pkazmier/master Added iterators to allow traversing tables in a sorted manner Merge pull request #60 from theypsilon/master base class methods invocation possible not only in method constructor _init Adding new class method 'base' which can be used for calling base class methods easily pl.class broke with classes that redefined __newindex; OrderedMap can now use normal indexing to assign keys in order Merge pull request #54 from CoolisTheName007/patch-1 Typo fixed, libarary->library Typo fixed. Added tablex.sort and tablex.sortv to iterate over sorted elements There are times when it would be convenient to iterate over a table either by sorted keys or values. There is an example of this in the PIL book. I added the two functions to tablex module, added two test cases, and updated the PL manual. Merge pull request #51 from pkazmier/master Thanks! lapp.add_type was broken due to uninitialized table variable. The variable `types` was not initialized preventing `add_type` from appending custom types to the variable. I also added a new test to cover this test case in the future.

== 2.7.0 / 2013-08-03 Charlie Savage * Do not call Ruby code during a GC cycle (Dirkjan Bussink) * Fix thread error handling (Geoffrey Giesemann) * Set libxml error handler when a libxml-ruby error handler is set (Geoffrey Giesemann) * Fix tests for nil TypeErrors (Geoffrey Giesemann) * Fix segmentation fault for issue #62. The deregisternode function was incorrectly NULLing out a node's mark and free functions for nodes removed from a document tree (Charlie Savage) * Cleanup writer code (Charlie Savage) * Fix equality check of LibXML::XML::Error against other objects (Michał Szajbe) * Fix XML::Writer.set_quote_char UT, wrong expected string (#63). (julp) * Fix potential segfault when GC occurs while creating ns from xpath (Timothy Elliott) * Fix segmentation fault after failing to load external schema (#59). (Clifford Heath)

(benchmarks clipped from release notes) ------------------------------------------------------------------------ r80 | [email protected] | 2013-08-13 14:55:00 +0200 (Tue, 13 Aug 2013) | 6 lines Add autoconf tests for size_t and ssize_t. Sort-of resolves public issue 79; it would solve the problem if MSVC typically used autoconf. However, it gives a natural place (config.h) to put the typedef even for MSVC. R=jsbell ------------------------------------------------------------------------ r79 | [email protected] | 2013-07-29 13:06:44 +0200 (Mon, 29 Jul 2013) | 14 lines When we compare the number of bytes produced with the offset for a backreference, make the signedness of the bytes produced clear, by sticking it into a size_t. This avoids a signed/unsigned compare warning from MSVC (public issue 71), and also is slightly clearer. Since the line is now so long the explanatory comment about the -1u trick has to go somewhere else anyway, I used the opportunity to explain it in slightly more detail. This is a purely stylistic change; the emitted assembler from GCC is identical. R=jeff ------------------------------------------------------------------------ r78 | [email protected] | 2013-06-30 21:24:03 +0200 (Sun, 30 Jun 2013) | 111 lines In the fast path for decompressing literals, instead of checking whether there's 16 bytes free and then checking right afterwards (when having subtracted the literal size) that there are now 5 bytes free, just check once for 21 bytes. This skips a compare and a branch; although it is easily predictable, it is still a few cycles on a fast path that we would like to get rid of. Benchmarking this yields very confusing results. On open-source GCC 4.8.1 on Haswell, we get exactly the expected results; the benchmarks where we hit the fast path for literals (in particular the two HTML benchmarks and the protobuf benchmark) give very nice speedups, and the others are not really affected. However, benchmarks with Google's GCC branch on other hardware is much less clear. It seems that we have a weak loss in some cases (and the win for the typical win cases are not nearly as clear), but that it depends on microarchitecture and plain luck in how we run the benchmark. Looking at the generated assembler, it seems that the removal of the if causes other large-scale changes in how the function is laid out, which makes it likely that this is just bad luck. Thus, we should keep this change, even though its exact current impact is unclear; it's a sensible change per se, and dropping it on the basis of microoptimization for a given compiler (or even branch of a compiler) would seem like a bad strategy in the long run. ------------------------------------------------------------------------ r77 | [email protected] | 2013-06-14 23:42:26 +0200 (Fri, 14 Jun 2013) | 92 lines Make the two IncrementalCopy* functions take in an ssize_t instead of a len, in order to avoid having to do 32-to-64-bit signed conversions on a hot path during decompression. (Also fixes some MSVC warnings, mentioned in public issue 75, but more of those remain.) They cannot be size_t because we expect them to go negative and test for that. This saves a few movzwl instructions, yielding ~2% speedup in decompression. ------------------------------------------------------------------------ r76 | [email protected] | 2013-06-13 18:19:52 +0200 (Thu, 13 Jun 2013) | 9 lines Add support for uncompressing to iovecs (scatter I/O). Windows does not have struct iovec defined anywhere, so we define our own version that's equal to what UNIX typically has. The bulk of this patch was contributed by Mohit Aron. R=jeff ------------------------------------------------------------------------ r75 | [email protected] | 2013-06-12 21:51:15 +0200 (Wed, 12 Jun 2013) | 4 lines Some code reorganization needed for an internal change. R=fikes ------------------------------------------------------------------------ r74 | [email protected] | 2013-04-09 17:33:30 +0200 (Tue, 09 Apr 2013) | 4 lines Supports truncated test data in zippy benchmark. R=sesse ------------------------------------------------------------------------ r73 | [email protected] | 2013-02-05 15:36:15 +0100 (Tue, 05 Feb 2013) | 4 lines Release Snappy 1.1.0. R=sanjay ------------------------------------------------------------------------ r72 | [email protected] | 2013-02-05 15:30:05 +0100 (Tue, 05 Feb 2013) | 9 lines Make ./snappy_unittest pass without "srcdir" being defined. Previously, snappy_unittests would read from an absolute path /testdata/..; convert it to use a relative path instead. Patch from Marc-Antonie Ruel. R=maruel ------------------------------------------------------------------------ r71 | [email protected] | 2013-01-18 13:16:36 +0100 (Fri, 18 Jan 2013) | 287 lines Increase the Zippy block size from 32 kB to 64 kB, winning ~3% density while being effectively performance neutral. The longer story about density is that we win 3-6% density on the benchmarks where this has any effect at all; many of the benchmarks (cp, c, lsp, man) are smaller than 32 kB and thus will have no effect. Binary data also seems to win little or nothing; of course, the already-compressed data wins nothing. The protobuf benchmark wins as much as ~18% depending on architecture, but I wouldn't be too sure that this is representative of protobuf data in general. As of performance, we lose a tiny amount since we get more tags (e.g., a long literal might be broken up into literal-copy-literal), but we win it back with less clearing of the hash table, and more opportunities to skip incompressible data (e.g. in the jpg benchmark). Decompression seems to get ever so slightly slower, again due to more tags. The total net change is about as close to zero as we can get, so the end effect seems to be simply more density and no real performance change. The comment about not changing kBlockSize, scary as it is, is not really relevant, since we're never going to have a block-level decompressor without explicitly marked blocks. Replace it with something more appropriate. This affects the framing format, but it's okay to change it since it basically has no users yet. ------------------------------------------------------------------------ r70 | [email protected] | 2013-01-06 20:21:26 +0100 (Sun, 06 Jan 2013) | 6 lines Adjust the Snappy open-source distribution for the changes in Google's internal file API. R=sanjay ------------------------------------------------------------------------ r69 | [email protected] | 2013-01-04 12:54:20 +0100 (Fri, 04 Jan 2013) | 15 lines Change a few ORs to additions where they don't matter. This helps the compiler use the LEA instruction more efficiently, since e.g. a + (b << 2) can be encoded as one instruction. Even more importantly, it can constant-fold the COPY_* enums together with the shifted negative constants, which also saves some instructions. (We don't need it for LITERAL, since it happens to be 0.) I am unsure why the compiler couldn't do this itself, but the theory is that it cannot prove that len-1 and len-4 cannot underflow/wrap, and thus can't do the optimization safely. The gains are small but measurable; 0.5-1.0% over the BM_Z* benchmarks (measured on Westmere, Sandy Bridge and Istanbul). R=sanjay ------------------------------------------------------------------------ r68 | [email protected] | 2012-10-08 13:37:16 +0200 (Mon, 08 Oct 2012) | 5 lines Stop giving -Werror to automake, due to an incompatibility between current versions of libtool and automake on non-GNU platforms (e.g. Mac OS X). R=sanjay ------------------------------------------------------------------------ r67 | [email protected] | 2012-08-17 15:54:47 +0200 (Fri, 17 Aug 2012) | 5 lines Fix public issue 66: Document GetUncompressedLength better, in particular that it leaves the source in a state that's not appropriate for RawUncompress. R=sanjay ------------------------------------------------------------------------ r66 | [email protected] | 2012-07-31 13:44:44 +0200 (Tue, 31 Jul 2012) | 5 lines Fix public issue 64: Check for <sys/time.h> at configure time, since MSVC seemingly does not have it. R=sanjay ------------------------------------------------------------------------ r65 | [email protected] | 2012-07-04 11:34:48 +0200 (Wed, 04 Jul 2012) | 10 lines Handle the case where gettimeofday() goes backwards or returns the same value twice; it could cause division by zero in the unit test framework. (We already had one fix for this in place, but it was incomplete.) This could in theory happen on any system, since there are few guarantees about gettimeofday(), but seems to only happen in practice on GNU/Hurd, where gettimeofday() is cached and only updated ever so often. R=sanjay ------------------------------------------------------------------------ r64 | [email protected] | 2012-07-04 11:28:33 +0200 (Wed, 04 Jul 2012) | 6 lines Mark ARMv4 as not supporting unaligned accesses (not just ARMv5 and ARMv6); apparently Debian still targets these by default, giving us segfaults on armel. R=sanjay ------------------------------------------------------------------------ r63 | [email protected] | 2012-05-22 11:46:05 +0200 (Tue, 22 May 2012) | 5 lines Fix public bug #62: Remove an extraneous comma at the end of an enum list, causing compile errors when embedded in Mozilla on OpenBSD. R=sanjay ------------------------------------------------------------------------ r62 | [email protected] | 2012-05-22 11:32:50 +0200 (Tue, 22 May 2012) | 8 lines Snappy library no longer depends on iostream. Achieved by moving logging macro definitions to a test-only header file, and by changing non-test code to use assert, fprintf, and abort instead of LOG/CHECK macros. R=sesse

Incompatible Changes: * can no longer use both `\' and `:' (didn't work anyway) New Features: * Handling of run-time errors (default: die) is now overridable by subclasses via signature_error_handler(). [github #54] * Can now have aliased named parameters. [github #57] * remove dependency on Devel::BeginLift [github #39] * can now use `when' to specify default conditions [github #48] * can use `//=' as a shortcut for `when undef' [github #45] * can now provide `where' constraints in addition to (or instead of) a type [github #7] * can now use `...' to disable further argument checking [github #49] * can now specify more than one alternative in type unions [github #55] * can now nest parameterized types Bug Fixes: * Removed experimental smartmatch warnings * Don't require Data::Alias for named params unless you have to [github #71] * Fixed obscure bug where an eval in Method::Signatures wouldn't be skipped when carp'ing (i.e. in carp_location_for()) [github #72] * Data::Alias is only loaded when needed avoiding a threads + eval bug in most cases and improving compile time performance. [rt.cpan.org 82922, github #62] * Compile-time errors now reporting proper line numbers. [github #61] * Trailing commas on parameter lists are now ok. [rt.cpan.org 81364] * Default condition of `when {}' now interpreted as `when { $_ ~~ {} }' (avoids parse error). [github #60] Optimizations: * better signature parsing using PPI [github #11] Distribution Fixes: * Fixed failing test in 5.10.0 (uncovered by CPAN Testers) * Fixed repo link in metadata (thanks dsteinbrunner) [github #87] * Add M::S::Parameter to MANIFEST [github #76] * Change representation of Infinity to work on Win32 [github #75] * Fixed stray detritus in MANIFEST. * Somehow my last-minute fix to the new error handler test didn't make it in; this will fix "Can't locate Moose.pm" errors. * Fixed test failing on 5.10.0 as per github #59. * Fixed subtests failing on Test::More's prior to 0.96. Docs: * Updated close parend problem to include quotes and a workaround [rt.cpan.org 85925] * Fixed some typos (thanks dsteinbrunner) [github #88] * Found and fixed missing parend * Minor clarifications here and there. * Clarified what doesn't work in Perl 5.8. * Added Function::Parameters to See Also section. * Updated copyright. * documented all new features * new ASCI-art breakdown of signature syntax * minor tweaks and corrections Misc: * Rearranged so signature is now an object [github #30] * Add hook for Travis CI [github #78] * Failure to parse parameters will now produce a more useful error.

FITS (Flexible Image Transport System) is a data format most used in astronomy. PyFITS is a Python module for reading, writing, and manipulating FITS files. The module uses Python's object-oriented features to provide quick, easy, and efficient access to FITS files. The use of Python's array syntax enables immediate access to any FITS extension, header cards, or data items. Changes to 2.4.0 (in py-pyfits): Changelog =========== 3.2 (2013-11-26) ---------------- Highlights ^^^^^^^^^^ - Rewrote CFITSIO-based backend for handling tile compression of FITS files. It now uses a standard CFITSIO instead of heavily modified pieces of CFITSIO as before. PyFITS ships with its own copy of CFITSIO v3.35 which supports the latest version of the Tiled Image Convention (v2.3), but system packagers may choose instead to strip this out in favor of a system-installed version of CFITSIO. Earlier versions may work, but nothing earlier than 3.28 has been tested yet. (#169) - Added support for reading and writing tables using the Q format for columns. The Q format is identical to the P format (variable-length arrays) except that it uses 64-bit integers for the data descriptors, allowing more than 4 GB of variable-length array data in a single table. (#160) - Added initial support for table columns containing pseudo-unsigned integers. This is currently enabled by using the ``uint=True`` option when opening files; any table columns with the correct BZERO value will be interpreted and returned as arrays of unsigned integers. - Some refactoring of the table and ``FITS_rec`` modules in order to better separate the details of the FITS binary and ASCII table data structures from the HDU data structures that encapsulate them. Most of these changes should not be apparent to users (but see API Changes below). API Changes ^^^^^^^^^^^ - Assigning to values in ``ColDefs.names``, ``ColDefs.formats``, ``ColDefs.nulls`` and other attributes of ``ColDefs`` instances that return lists of column properties is no longer supported. Assigning to those lists will no longer update the corresponding columns. Instead, please just modify the ``Column`` instances directly (``Column.name``, ``Column.null``, etc.) - The ``pyfits.new_table`` function is marked "pending deprecation". This does not mean it will be removed outright or that its functionality has changed. It will likely be replaced in the future for a function with similar, if not subtly different functionality. A better, if not slightly more verbose approach is to use ``pyfits.FITS_rec.from_columns`` to create a new ``FITS_rec`` table--this has the same interface as ``pyfits.new_table``. The difference is that it returns a plan ``FITS_rec`` array, and not an HDU instance. This ``FITS_rec`` object can then be used as the data argument in the constructors for ``BinTableHDU`` (for binary tables) or ``TableHDU`` (for ASCII tables). This is analogous to creating an ``ImageHDU`` by passing in an image array. ``pyfits.FITS_rec.from_columns`` is just a simpler way of creating a FITS-compatible recarray from a FITS column specification. - The ``updateHeader``, ``updateHeaderData``, and ``updateCompressedData`` methods of the ``CompDataHDU`` class are pending deprecation and moved to internal methods. The operation of these methods depended too much on internal state to be used safely by users; instead they are invoked automatically in the appropriate places when reading/writing compressed image HDUs. - The ``CompDataHDU.compData`` attribute is pending deprecation in favor of the clearer and more PEP-8 compatible ``CompDataHDU.compressed_data``. - The constructor for ``CompDataHDU`` has been changed to accept new keyword arguments. The new keyword arguments are essentially the same, but are in underscore_separated format rather than camelCase format. The old arguments are still pending deprecation. - The internal attributes of HDU classes ``_hdrLoc``, ``_datLoc``, and ``_datSpan`` have been replaced with ``_header_offset``, ``_data_offset``, and ``_data_size`` respectively. The old attribute names are still pending deprecation. This should only be of interest to advanced users who have created their own HDU subclasses. - The following previously deprecated functions and methods have been removed entirely: ``createCard``, ``createCardFromString``, ``upperKey``, ``ColDefs.data``, ``setExtensionNameCaseSensitive``, ``_File.getfile``, ``_TableBaseHDU.get_coldefs``, ``Header.has_key``, ``Header.ascardlist``. If you run your code with a previous version of PyFITS (>= 3.0, < 3.2) with the ``python -Wd`` argument, warnings for all deprecated interfaces still in use will be displayed. - Interfaces that were pending deprecation are now fully deprecated. These include: ``create_card``, ``create_card_from_string``, ``upper_key``, ``Header.get_history``, and ``Header.get_comment``. - The ``.name`` attribute on HDUs is now directly tied to the HDU's header, so that if ``.header['EXTNAME']`` changes so does ``.name`` and vice-versa. - The ``pyfits.file.PYTHON_MODES`` constant dict was renamed to ``pyfits.file.PYFITS_MODES`` which better reflects its purpose. This is rarely used by client code, however. Support for the old name will be removed by PyFITS 3.4. Other Changes and Additions ^^^^^^^^^^^^^^^^^^^^^^^^^^^ - The new compression code also adds support for the ZQUANTIZ and ZDITHER0 keywords added in more recent versions of this FITS Tile Compression spec. This includes support for lossless compression with GZIP. (#198) By default no dithering is used, but the ``SUBTRACTIVE_DITHER_1`` and ``SUBTRACTIVE_DITHER_2`` methods can be enabled by passing the correct constants to the ``quantize_method`` argument to the ``CompImageHDU`` constuctor. A seed can be manually specified, or automatically generated using either the system clock or checksum-based methods via the ``dither_seed`` argument. See the documentation for ``CompImageHDU`` for more details. (#198) (spacetelescope/PYFITS#32) - Images compressed with the Tile Compression standard can now be larger than 4 GB through support of the Q format. (#159) - All HDUs now have a ``.ver`` ``.level`` attribute that returns the value of the EXTVAL and EXTLEVEL keywords from that HDU's header, if the exist. This was added for consistency with the ``.name`` attribute which returns the EXTNAME value from the header. - Then ``Column`` and ``ColDefs`` classes have new ``.dtype`` attributes which give the Numpy dtype for the column data in the first case, and the full Numpy compound dtype for each table row in the latter case. - There was an issue where new tables created defaulted the values in all string columns to '0.0'. Now string columns are filled with empty strings by default--this seems a less surprising default, but it may cause differences with tables created with older versions of PyFITS. - Improved round-tripping and preservation of manually assigned column attributes (``TNULLn``, ``TSCALn``, etc.) in table HDU headers. (astropy/astropy#996) Bug Fixes ^^^^^^^^^ - Binary tables containing compressed images may, optionally, contain other columns unrelated to the tile compression convention. Although this is an uncommon use case, it is permitted by the standard. (#159) - Reworked some of the file I/O routines to allow simpler, more consistent mapping between OS-level file modes ('rb', 'wb', 'ab', etc.) and the more "PyFITS-specific" modes used by PyFITS like "readonly" and "update". That is, if reading a FITS file from an open file object, it doesn't matter as much what "mode" it was opened in so long as it has the right capabilities (read/write/etc.) Also works around bugs in the Python io module in 2.6+ with regard to file modes. (spacetelescope/PyFITS#33) - Fixed an obscure issue that can occur on systems that don't have flush to memory-mapped files implemented (namely GNU Hurd). (astropy/astropy#968) 3.1.3 (2013-11-26) ------------------ - Disallowed assigning NaN and Inf floating point values as header values, since the FITS standard does not define a way to represent them in. Because this is undefined, the previous behavior did not make sense and produced invalid FITS files. (spacetelescope/PyFITS#11) - Added a workaround for a bug in 64-bit OSX that could cause truncation when writing files greater than 2^32 bytes in size. (spacetelescope/PyFITS#28) - Fixed a long-standing issue where writing binary tables did not correctly write the TFORMn keywords for variable-length array columns (they ommitted the max array length parameter of the format). This was thought fixed in v3.1.2, but it was only fixed there for compressed image HDUs and not for binary tables in general. - Fixed an obscure issue that can occur on systems that don't have flush to memory-mapped files implemented (namely GNU Hurd). (Backported from 3.2) 3.0.12 (2013-11-26) ------------------- - Disallowed assigning NaN and Inf floating point values as header values, since the FITS standard does not define a way to represent them in. Because this is undefined, the previous behavior did not make sense and produced invalid FITS files. (Backported from 3.1.3) - Added a workaround for a bug in 64-bit OSX that could cause truncation when writing files greater than 2^32 bytes in size. (Backported from 3.1.3) - Fixed a long-standing issue where writing binary tables did not correctly write the TFORMn keywords for variable-length array columns (they ommitted the max array length parameter of the format). This was thought fixed in v3.1.2, but it was only fixed there for compressed image HDUs and not for binary tables in general. (Backported from 3.1.3) - Fixed an obscure issue that can occur on systems that don't have flush to memory-mapped files implemented (namely GNU Hurd). (Backported from 3.2) 3.1.2 (2013-04-22) ------------------ - When an error occurs opening a file in fitsdiff the exception message will now at least mention which file had the error. (#168) - Fixed support for opening gzipped FITS files by filename in a writeable mode (PyFITS has supported writing to gzip files for some time now, but only enabled it when GzipFile objects were passed to ``pyfits.open()`` due to some legacy code preventing full gzip support. (#195) - Added a more helpful error message in the case of malformatted FITS files that contain non-float NULL values in an ASCII table but are missing the required TNULLn keywords in the header. (#197) - Fixed an (apparently long-standing) issue where writing compressed images did not correctly write the TFORMn keywords for variable-length array columns (they ommitted the max array length parameter of the format). (#199) - Slightly refactored how tables containing variable-length array columns are handled to add two improvements: Fixes an issue where accessing the data after a call to the `pyfits.getdata` convenience function caused an exception, and allows the VLA data to be read from an existing mmap of the FITS file. (#200) - Fixed a bug that could occur when opening a table containing multi-dimensional columns (i.e. via the TDIMn keyword) and then writing it out to a new file. (#201) - Added use of the console_scripts entry point to install the fitsdiff and fitscheck scripts, which if nothing else provides better Windows support. The generated scripts now override the ones explicitly defined in the scripts/ directory (which were just trivial stubs to begin with). (#202) - Fixed a bug on Python 3 where attempting to open a non-existent file on Python 3 caused a seemingly unrelated traceback. (#203) - Fixed a bug in fitsdiff that reported two header keywords containing NaN as value as different. (#204) - Fixed an issue in the tests that caused some tests to fail if pyfits is installed with read-only permissions. (#208) - Fixed a bug where instantiating a ``BinTableHDU`` from a numpy array containing boolean fields converted all the values to ``False``. (#215) - Fixed an issue where passing an array of integers into the constructor of ``Column()`` when the column type is floats of the same byte width caused the column array to become garbled. (#218) - Fixed inconsistent behavior in creating CONTINUE cards from byte strings versus unicode strings in Python 2--CONTINUE cards can now be created properly from unicode strings (so long as they are convertable to ASCII). (spacetelescope/PyFITS#1) - Fixed a couple cases where creating a new table using TDIMn in some of the columns could caused a crash. (spacetelescope/PyFITS#3) - Fixed a bug in parsing HIERARCH keywords that do not have a space after the first equals sign (before the value). (spacetelescope/PyFITS#5) - Prevented extra leading whitespace on HIERARCH keywords from being treated as part of the keyword. (spacetelescope/PyFITS#6) - Fixed a bug where HIERARCH keywords containing lower-case letters was mistakenly marked as invalid during header validation. (spacetelescope/PyFITS#7) - Fixed an issue that was ancillary to (spacetelescope/PyFITS#7) where the ``Header.index()`` method did not work correctly with HIERARCH keywords containing lower-case letters. 3.0.11 (2013-04-17) ------------------- - Fixed support for opening gzipped FITS files by filename in a writeable mode (PyFITS has supported writing to gzip files for some time now, but only enabled it when GzipFile objects were passed to ``pyfits.open()`` due to some legacy code preventing full gzip support. Backported from 3.1.2. (#195) - Added a more helpful error message in the case of malformatted FITS files that contain non-float NULL values in an ASCII table but are missing the required TNULLn keywords in the header. Backported from 3.1.2. (#197) - Fixed an (apparently long-standing) issue where writing compressed images did not correctly write the TFORMn keywords for variable-length array columns (they ommitted the max array length parameter of the format). Backported from 3.1.2. (#199) - Slightly refactored how tables containing variable-length array columns are handled to add two improvements: Fixes an issue where accessing the data after a call to the `pyfits.getdata` convenience function caused an exception, and allows the VLA data to be read from an existing mmap of the FITS file. Backported from 3.1.2. (#200) - Fixed a bug that could occur when opening a table containing multi-dimensional columns (i.e. via the TDIMn keyword) and then writing it out to a new file. Backported from 3.1.2. (#201) - Fixed a bug on Python 3 where attempting to open a non-existent file on Python 3 caused a seemingly unrelated traceback. Backported from 3.1.2. (#203) - Fixed a bug in fitsdiff that reported two header keywords containing NaN as value as different. Backported from 3.1.2. (#204) - Fixed an issue in the tests that caused some tests to fail if pyfits is installed with read-only permissions. Backported from 3.1.2. (#208) - Fixed a bug where instantiating a ``BinTableHDU`` from a numpy array containing boolean fields converted all the values to ``False``. Backported from 3.1.2. (#215) - Fixed an issue where passing an array of integers into the constructor of ``Column()`` when the column type is floats of the same byte width caused the column array to become garbled. Backported from 3.1.2. (#218) - Fixed a couple cases where creating a new table using TDIMn in some of the columns could caused a crash. Backported from 3.1.2. (spacetelescope/PyFITS#3) 3.1.1 (2013-01-02) ------------------ This is a bug fix release for the 3.1.x series. Bug Fixes ^^^^^^^^^ - Improved handling of scaled images and pseudo-unsigned integer images in compressed image HDUs. They now work more transparently like normal image HDUs with support for the ``do_not_scale_image_data`` and ``uint`` options, as well as ``scale_back`` and ``save_backup``. The ``.scale()`` method works better too. (#88) - Permits non-string values for the EXTNAME keyword when reading in a file, rather than throwing an exception due to the malformatting. Added verification for the format of the EXTNAME keyword when writing. (#96) - Added support for EXTNAME and EXTVER in PRIMARY HDUs. That is, if EXTNAME is specified in the header, it will also be reflected in the ``.name`` attribute and in ``pyfits.info()``. These keywords used to be verboten in PRIMARY HDUs, but the latest version of the FITS standard allows them. (#151) - HCOMPRESS can again be used to compress data cubes (and higher-dimensional arrays) so long as the tile size is effectively 2-dimensional. In fact, PyFITS will automatically use compatible tile sizes even if they're not explicitly specified. (#171) - Added support for the optional ``endcard`` parameter in the ``Header.fromtextfile()`` and ``Header.totextfile()`` methods. Although ``endcard=False`` was a reasonable default assumption, there are still text dumps of FITS headers that include the END card, so this should have been more flexible. (#176) - Fixed a crash when running fitsdiff on two empty (that is, zero row) tables. (#178) - Fixed an issue where opening files containing random groups HDUs in update mode could cause an unnecessary rewrite of the file even if none of the data is modified. (#179) - Fixed a bug that could caused a deadlock in the filesystem on OSX if PyFITS is used with Numpy 1.7 in some cases. (#180) - Fixed a crash when generating diff reports from diffs using the ``ignore_comments`` options. (#181) - Fixed some bugs with WCS Paper IV record-valued keyword cards: - Cards that looked kind of like RVKCs but were not intended to be were over-permissively treated as such--commentary keywords like COMMENT and HISTORY were particularly affected. (#183) - Looking up a card in a header by its standard FITS keyword only should always return the raw value of that card. That way cards containing values that happen to valid RVKCs but were not intended to be will still be treated like normal cards. (#184) - Looking up a RVKC in a header with only part of the field-specifier (for example "DP1.AXIS" instead of "DP1.AXIS.1") was implicitly treated as a wildcard lookup. (#184) - Fixed a crash when diffing two FITS files where at least one contains a compressed image HDU which was not recognized as an image instead of a table. (#187) - Fixed bugs in the backwards compatibility layer for the ``CardList.index`` and ``CardList.count`` methods. (#190) - Improved ``__repr__`` and text file representation of cards with long values that are split into CONTINUE cards. (#193) - Fixed a crash when trying to assign a long (> 72 character) value to blank ('') keywords. This also changed how blank keywords are represented--there are still exactly 8 spaces before any commentary content can begin; this *may* affect the exact display of header cards that assumed there could be fewer spaces in a blank keyword card before the content begins. However, the current approach is more in line with the requirements of the FITS standard. (#194) 3.0.10 (2013-01-02) ------------------- - Improved handling of scaled images and pseudo-unsigned integer images in compressed image HDUs. They now work more transparently like normal image HDUs with support for the ``do_not_scale_image_data`` and ``uint`` options, as well as ``scale_back`` and ``save_backup``. The ``.scale()`` method works better too. Backported from 3.1.1. (#88) - Permits non-string values for the EXTNAME keyword when reading in a file, rather than throwing an exception due to the malformatting. Added verification for the format of the EXTNAME keyword when writing. Backported from 3.1.1. (#96) - Added support for EXTNAME and EXTVER in PRIMARY HDUs. That is, if EXTNAME is specified in the header, it will also be reflected in the ``.name`` attribute and in ``pyfits.info()``. These keywords used to be verbotten in PRIMARY HDUs, but the latest version of the FITS standard allows them. Backported from 3.1.1. (#151) - HCOMPRESS can again be used to compress data cubes (and higher-dimensional arrays) so long as the tile size is effectively 2-dimensional. In fact, PyFITS will not automatically use compatible tile sizes even if they're not explicitly specified. Backported from 3.1.1. (#171) - Fixed a bug when writing out files containing zero-width table columns, where the TFIELDS keyword would be updated incorrectly, leaving the table largely unreadable. Backported from 3.1.0. (#174) - Fixed an issue where opening files containing random groups HDUs in update mode could cause an unnecessary rewrite of the file even if none of the data is modified. Backported from 3.1.1. (#179) - Fixed a bug that could caused a deadlock in the filesystem on OSX if PyFITS is used with Numpy 1.7 in some cases. Backported from 3.1.1. (#180) 3.1 (2012-08-08) ---------------- Highlights ^^^^^^^^^^ - The ``Header`` object has been significantly reworked, and ``CardList`` objects are now deprecated (their functionality folded into the ``Header`` class). See API Changes below for more details. - Memory maps are now used by default to access HDU data. See API Changes below for more details. - Now includes a new version of the ``fitsdiff`` program for comparing two FITS files, and a new FITS comparison API used by ``fitsdiff``. See New Features below. API Changes ^^^^^^^^^^^ - The ``Header`` class has been rewritten, and the ``CardList`` class is deprecated. Most of the basic details of working with FITS headers are unchanged, and will not be noticed by most users. But there are differences in some areas that will be of interest to advanced users, and to application developers. For full details of the changes, see the "Header Interface Transition Guide" section in the PyFITS documentation. See ticket #64 on the PyFITS Trac for futher details and background. Some highlights are listed below: * The Header class now fully implements the Python dict interface, and can be used interchangably with a dict, where the keys are header keywords. * New keywords can be added to the header using normal keyword assignment (previously it was necessary to use ``Header.update`` to add new keywords). For example:: >>> header['NAXIS'] = 2 will update the existing 'FOO' keyword if it already exists, or add a new one if it doesn't exist, just like a dict. * It is possible to assign both a value and a comment at the same time using a tuple:: >>> header['NAXIS'] = (2, 'Number of axes') * To add/update a new card and ensure it's added in a specific location, use ``Header.set()``:: >>> header.set('NAXIS', 2, 'Number of axes', after='BITPIX') This works the same as the old ``Header.update()``. ``Header.update()`` still works in the old way too, but is deprecated. * Although ``Card`` objects still exist, it generally is not necessary to work with them directly. ``Header.ascardlist()``/``Header.ascard`` are deprecated and should not be used. To directly access the ``Card`` objects in a header, use ``Header.cards``. * To access card comments, it is still possible to either go through the card itself, or through ``Header.comments``. For example:: >>> header.cards['NAXIS'].comment Number of axes >>> header.comments['NAXIS'] Number of axes * ``Card`` objects can now be used interchangeably with ``(keyword, value, comment)`` 3-tuples. They still have ``.value`` and ``.comment`` attributes as well. The ``.key`` attribute has been renamed to ``.keyword`` for consistency, though ``.key`` is still supported (but deprecated). - Memory mapping is now used by default to access HDU data. That is, ``pyfits.open()`` uses ``memmap=True`` as the default. This provides better performance in the majority of use cases--there are only some I/O intensive applications where it might not be desirable. Enabling mmap by default also enabled finding and fixing a large number of bugs in PyFITS' handling of memory-mapped data (most of these bug fixes were backported to PyFITS 3.0.5). (#85) * A new ``pyfits.USE_MEMMAP`` global variable was added. Set ``pyfits.USE_MEMMAP = False`` to change the default memmap setting for opening files. This is especially useful for controlling the behavior in applications where pyfits is deeply embedded. * Likewise, a new ``PYFITS_USE_MEMMAP`` environment variable is supported. Set ``PYFITS_USE_MEMMAP = 0`` in your environment to change the default behavior. - The ``size()`` method on HDU objects is now a ``.size`` property--this returns the size in bytes of the data portion of the HDU, and in most cases is equivalent to ``hdu.data.nbytes`` (#83) - ``BinTableHDU.tdump`` and ``BinTableHDU.tcreate`` are deprecated--use ``BinTableHDU.dump`` and ``BinTableHDU.load`` instead. The new methods output the table data in a slightly different format from previous versions, which places quotes around each value. This format is compatible with data dumps from previous versions of PyFITS, but not vice-versa due to a parsing bug in older versions. - Likewise the ``pyfits.tdump`` and ``pyfits.tcreate`` convenience function versions of these methods have been renamed ``pyfits.tabledump`` and ``pyfits.tableload``. The old deprecated, but currently retained for backwards compatibility. (r1125) - A new global variable ``pyfits.EXTENSION_NAME_CASE_SENSITIVE`` was added. This serves as a replacement for ``pyfits.setExtensionNameCaseSensitive`` which is not deprecated and may be removed in a future version. To enable case-sensitivity of extension names (i.e. treat 'sci' as distict from 'SCI') set ``pyfits.EXTENSION_NAME_CASE_SENSITIVE = True``. The default is ``False``. (r1139) - A new global configuration variable ``pyfits.STRIP_HEADER_WHITESPACE`` was added. By default, if a string value in a header contains trailing whitespace, that whitespace is automatically removed when the value is read. Now if you set ``pyfits.STRIP_HEADER_WHITESPACE = False`` all whitespace is preserved. (#146) - The old ``classExtensions`` extension mechanism (which was deprecated in PyFITS 3.0) is removed outright. To our knowledge it was no longer used anywhere. (r1309) - Warning messages from PyFITS issued through the Python warnings API are now output to stderr instead of stdout, as is the default. PyFITS no longer modifies the default behavior of the warnings module with respect to which stream it outputs to. (r1319) - The ``checksum`` argument to ``pyfits.open()`` now accepts a value of 'remove', which causes any existing CHECKSUM/DATASUM keywords to be ignored, and removed when the file is saved. New Features ^^^^^^^^^^^^ - Added support for the proposed "FITS" extension HDU type. See http://listmgr.cv.nrao.edu/pipermail/fitsbits/2002-April/001094.html. FITS HDUs contain an entire FITS file embedded in their data section. `FitsHDU` objects work like other HDU types in PyFITS. Their ``.data`` attribute returns the raw data array. However, they have a special ``.hdulist`` attribute which processes the data as a FITS file and returns it as an in-memory HDUList object. FitsHDU objects also support a ``FitsHDU.fromhdulist()`` classmethod which returns a new `FitsHDU` object that embeds the supplied HDUList. (#80) - Added a new ``.is_image`` attribute on HDU objects, which is True if the HDU data is an 'image' as opposed to a table or something else. Here the meaning of 'image' is fairly loose, and mostly just means a Primary or Image extension HDU, or possibly a compressed image HDU (#71) - Added an ``HDUList.fromstring`` classmethod which can parse a FITS file already in memory and instantiate and ``HDUList`` object from it. This could be useful for integrating PyFITS with other libraries that work on FITS file, such as CFITSIO. It may also be useful in streaming applications. The name is a slight misnomer, in that it actually accepts any Python object that implements the buffer interface, which includes ``bytes``, ``bytearray``, ``memoryview``, ``numpy.ndarray``, etc. (#90) - Added a new ``pyfits.diff`` module which contains facilities for comparing FITS files. One can use the ``pyfits.diff.FITSDiff`` class to compare two FITS files in their entirety. There is also a ``pyfits.diff.HeaderDiff`` class for just comparing two FITS headers, and other similar interfaces. See the PyFITS Documentation for more details on this interface. The ``pyfits.diff`` module powers the new ``fitsdiff`` program installed with PyFITS. After installing PyFITS, run ``fitsdiff --help`` for usage details. - ``pyfits.open()`` now accepts a ``scale_back`` argument. If set to ``True``, this automatically scales the data using the original BZERO and BSCALE parameters the file had when it was first opened, if any, as well as the original BITPIX. For example, if the original BITPIX were 16, this would be equivalent to calling ``hdu.scale('int16', 'old')`` just before calling ``flush()`` or ``close()`` on the file. This option applies to all HDUs in the file. (#120) - ``pyfits.open()`` now accepts a ``save_backup`` argument. If set to ``True``, this automatically saves a backup of the original file before flushing any changes to it (this of course only applies to update and append mode). This may be especially useful when working with scaled image data. (#121) Changes in Behavior ^^^^^^^^^^^^^^^^^^^ - Warnings from PyFITS are not output to stderr by default, instead of stdout as it has been for some time. This is contrary to most users' expectations and makes it more difficult for them to separate output from PyFITS from the desired output for their scripts. (r1319) Bug Fixes ^^^^^^^^^ - Fixed ``pyfits.tcreate()`` (now ``pyfits.tableload()``) to be more robust when encountering blank lines in a column definition file (#14) - Fixed a fairly rare crash that could occur in the handling of CONTINUE cards when using Numpy 1.4 or lower (though 1.4 is the oldest version supported by PyFITS). (r1330) - Fixed ``_BaseHDU.fromstring`` to actually correctly instantiate an HDU object from a string/buffer containing the header and data of that HDU. This allowed for the implementation of ``HDUList.fromstring`` described above. (#90) - Fixed a rare corner case where, in some use cases, (mildly, recoverably) malformatted float values in headers were not properly returned as floats. (#137) - Fixed a corollary to the previous bug where float values with a leading zero before the decimal point had the leading zero unnecessarily removed when saving changes to the file (eg. "0.001" would be written back as ".001" even if no changes were otherwise made to the file). (#137) - When opening a file containing CHECKSUM and/or DATASUM keywords in update mode, the CHECKSUM/DATASUM are updated and preserved even if the file was opened with checksum=False. This change in behavior prevents checksums from being unintentionally removed. (#148) - Fixed a bug where ``ImageHDU.scale(option='old')`` wasn't working at all--it was not restoring the image to its original BSCALE and BZERO values. (#162) - Fixed a bug when writing out files containing zero-width table columns, where the TFIELDS keyword would be updated incorrectly, leaving the table largely unreadable. This fix will be backported to the 3.0.x series in version 3.0.10. (#174) 3.0.9 (2012-08-06) ------------------ This is a bug fix release for the 3.0.x series. Bug Fixes ^^^^^^^^^ - Fixed ``Header.values()``/``Header.itervalues()`` and ``Header.items()``/ ``Header.iteritems()`` to correctly return the different values for duplicate keywords (particularly commentary keywords like HISTORY and COMMENT). This makes the old Header implementation slightly more compatible with the new implementation in PyFITS 3.1. (#127) .. note:: This fix did not change the existing behavior from earlier PyFITS versions where ``Header.keys()`` returns all keywords in the header with duplicates removed. PyFITS 3.1 changes that behavior, so that ``Header.keys()`` includes duplicates. - Fixed a bug where ``ImageHDU.scale(option='old')`` wasn't working at all--it was not restoring the image to its original BSCALE and BZERO values. (#162) - Fixed a bug where opening a file containing compressed image HDUs in 'update' mode and then immediately closing it without making any changes caused the file to be rewritten unncessarily. (#167) - Fixed two memory leaks that could occur when writing compressed image data, or in some cases when opening files containing compressed image HDUs in 'update' mode. (#168) 3.0.8 (2012-06-04) ------------------ Changes in Behavior ^^^^^^^^^^^^^^^^^^^ - Prior to this release, image data sections did not work with scaled data--that is, images with non-trivial BSCALE and/or BZERO values. Previously, in order to read such images in sections, it was necessary to manually apply the BSCALE+BZERO to each section. It's worth noting that sections *did* support pseudo-unsigned ints (flakily). This change just extends that support for general BSCALE+BZERO values. Bug Fixes ^^^^^^^^^ - Fixed a bug that prevented updates to values in boolean table columns from being saved. This turned out to be a symptom of a deeper problem that could prevent other table updates from being saved as well. (#139) - Fixed a corner case in which a keyword comment ending with the string "END" could, in some circumstances, cause headers (and the rest of the file after that point) to be misread. (#142) - Fixed support for scaled image data and psuedo-unsigned ints in image data sections (``hdu.section``). Previously this was not supported at all. At some point support was supposedly added, but it was buggy and incomplete. Now the feature seems to work much better. (#143) - Fixed the documentation to point out that image data sections *do* support non-contiguous slices (and have for a long time). The documentation was never updated to reflect this, and misinformed users that only contiguous slices were supported, leading to some confusion. (#144) - Fixed a bug where creating an ``HDUList`` object containing multiple PRIMARY HDUs caused an infinite recursion when validating the object prior to writing to a file. (#145) - Fixed a rare but serious case where saving an update to a file that previously had a CHECKSUM and/or DATASUM keyword, but removed the checksum in saving, could cause the file to be slightly corrupted and unreadable. (#147) - Fixed problems with reading "non-standard" FITS files with primary headers containing SIMPLE = F. PyFITS has never made many guarantees as to how such files are handled. But it should at least be possible to read their headers, and the data if possible. Saving changes to such a file should not try to prepend an unwanted valid PRIMARY HDU. (#157) - Fixed a bug where opening an image with ``disable_image_compression = True`` caused compression to be disabled for all subsequent ``pyfits.open()`` calls. (r1651) 3.0.7 (2012-04-10) ------------------ Changes in Behavior ^^^^^^^^^^^^^^^^^^^ - Slices of GroupData objects now return new GroupData objects instead of extended multi-row _Group objects. This is analogous to how PyFITS 3.0 fixed FITS_rec slicing, and should have been fixed for GroupData at the same time. The old behavior caused bugs where functions internal to Numpy expected that slicing an ndarray would return a new ndarray. As this is a rare usecase with a rare feature most users are unlikely to be affected by this change. - The previously internal _Group object for representing individual group records in a GroupData object are renamed Group and are now a public interface. However, there's almost no good reason to create Group objects directly, so it shouldn't be considered a "new feature". - An annoyance from PyFITS 3.0.6 was fixed, where the value of the EXTEND keyword was always being set to F if there are not actually any extension HDUs. It was unnecessary to modify this value. Bug Fixes ^^^^^^^^^ - Fixed GroupData objects to return new GroupData objects when sliced instead of _Group record objects. See "Changes in behavior" above for more details. - Fixed slicing of Group objects--previously it was not possible to slice slice them at all. - Made it possible to assign `np.bool_` objects as header values. (#123) - Fixed overly strict handling of the EXTEND keyword; see "Changes in behavior" above. (#124) - Fixed many cases where an HDU's header would be marked as "modified" by PyFITS and rewritten, even when no changes to the header are necessary. (#125) - Fixed a bug where the values of the PTYPEn keywords in a random groups HDU were forced to be all lower-case when saving the file. (#130) - Removed an unnecessary inline import in `ExtensionHDU.__setattr__` that was causing some slowdown when opening files containing a large number of extensions, plus a few other small (but not insignficant) performance improvements thanks to Julian Taylor. (#133) - Fixed a regression where header blocks containing invalid end-of-header padding (i.e. null bytes instead of spaces) couldn't be parsed by PyFITS. Such headers can be parsed again, but a warning is raised, as such headers are not valid FITS. (#136) - Fixed a memory leak where table data in random groups HDUs weren't being garbage collected. (#138) 3.0.6 (2012-02-29) ------------------ Highlights ^^^^^^^^^^ The main reason for this release is to fix an issue that was introduced in PyFITS 3.0.5 where merely opening a file containing scaled data (that is, with non-trivial BSCALE and BZERO keywords) in 'update' mode would cause the data to be automatically rescaled--possibly converting the data from ints to floats--as soon as the file is closed, even if the application did not touch the data. Now PyFITS will only rescale the data in an extension when the data is actually accessed by the application. So opening a file in 'update' mode in order to modify the header or append new extensions will not cause any change to the data in existing extensions. This release also fixes a few Windows-specific bugs found through more extensive Windows testing, and other miscellaneous bugs. Bug Fixes ^^^^^^^^^ - More accurate error messages when opening files containing invalid header cards. (#109) - Fixed a possible reference cycle/memory leak that was caught through more extensive testing on Windows. (#112) - Fixed 'ostream' mode to open the underlying file in 'wb' mode instead of 'w' mode. (#112) - Fixed a Windows-only issue where trying to save updates to a resized FITS file could result in a crash due to there being open mmaps on that file. (#112) - Fixed a crash when trying to create a FITS table (i.e. with new_table()) from a Numpy array containing bool fields. (#113) - Fixed a bug where manually initializing an ``HDUList`` with a list of of HDUs wouldn't set the correct EXTEND keyword value on the primary HDU. (#114) - Fixed a crash that could occur when trying to deepcopy a Header in Python < 2.7. (#115) - Fixed an issue where merely opening a scaled image in 'update' mode would cause the data to be converted to floats when the file is closed. (#119) 3.0.5 (2012-01-30) ------------------ - Fixed a crash that could occur when accessing image sections of files opened with memmap=True. (r1211) - Fixed the inconsistency in the behavior of files opened in 'readonly' mode when memmap=True vs. when memmap=False. In the latter case, although changes to array data were not saved to disk, it was possible to update the array data in memory. On the other hand with memmap=True, 'readonly' mode prevented even in-memory modification to the data. This is what 'copyonwrite' mode was for, but difference in behavior was confusing. Now 'readonly' is equivalent to 'copyonwrite' when using memmap. If the old behavior of denying changes to the array data is necessary, a new 'denywrite' mode may be used, though it is only applicable to files opened with memmap. (r1275) - Fixed an issue where files opened with memmap=True would return image data as a raw numpy.memmap object, which can cause some unexpected behaviors--instead memmap object is viewed as a numpy.ndarray. (r1285) - Fixed an issue in Python 3 where a workaround for a bug in Numpy on Python 3 interacted badly with some other software, namely to vo.table package (and possibly others). (r1320, r1337, and #110) - Fixed buggy behavior in the handling of SIGINTs (i.e. Ctrl-C keyboard interrupts) while flushing changes to a FITS file. PyFITS already prevented SIGINTs from causing an incomplete flush, but did not clean up the signal handlers properly afterwards, or reraise the keyboard interrupt once the flush was complete. (r1321) - Fixed a crash that could occur in Python 3 when opening files with checksum checking enabled. (r1336) - Fixed a small bug that could cause a crash in the `StreamingHDU` interface when using Numpy below version 1.5. - Fixed a crash that could occur when creating a new `CompImageHDU` from an array of big-endian data. (#104) - Fixed a crash when opening a file with extra zero padding at the end. Though FITS files should not have such padding, it's not explictly forbidden by the format either, and PyFITS shouldn't stumble over it. (#106) - Fixed a major slowdown in opening tables containing large columns of string values. (#111) 3.0.4 (2011-11-22) ------------------ - Fixed a crash when writing HCOMPRESS compressed images that could happen on Python 2.5 and 2.6. (r1217) - Fixed a crash when slicing an table in a file opened in 'readonly' mode with memmap=True. (r1230) - Writing changes to a file or writing to a new file verifies the output in 'fix' mode by default instead of 'exception'--that is, PyFITS will automatically fix common FITS format errors rather than raising an exception. (r1243) - Fixed a bug where convenience functions such as getval() and getheader() crashed when specifying just 'PRIMARY' as the extension to use (r1263). - Fixed a bug that prevented passing keyword arguments (beyond the standard data and header arguments) as positional arguments to the constructors of extension HDU classes. - Fixed some tests that were failing on Windows--in this case the tests themselves failed to close some temp files and Windows refused to delete them while there were still open handles on them. (r1295) - Fixed an issue with floating point formatting in header values on Python 2.5 for Windows (and possibly other platforms). The exponent was zero-padded to 3 digits; although the FITS standard makes no specification on this, the formatting is now normalized to always pad the exponent to two digits. (r1295) - Fixed a bug where long commentary cards (such as HISTORY and COMMENT) were broken into multiple CONTINUE cards. However, commentary cards are not expected to be found in CONTINUE cards. Instead these long cards are broken into multiple commentary cards. (#97) - GZIP/ZIP-compressed FITS files can be detected and opened regardless of their filename extension. (#99) - Fixed a serious bug where opening scaled images in 'update' mode and then closing the file without touching the data would cause the file to be corrupted. (#101) 3.0.3 (2011-10-05) ------------------ - Fixed several small bugs involving corner cases in record-valued keyword cards (#70) - In some cases HDU creation failed if the first keyword value in the header was not a string value (#89) - Fixed a crash when trying to compute the HDU checksum when the data array contains an odd number of bytes (#91) - Disabled an unnecessary warning that was displayed on opening compressed HDUs with disable_image_compression = True (#92) - Fixed a typo in code for handling HCOMPRESS compressed images. 3.0.2 (2011-09-23) ------------------ - The ``BinTableHDU.tcreate`` method and by extension the ``pyfits.tcreate`` function don't get tripped up by blank lines anymore (#14) - The presence, value, and position of the EXTEND keyword in Primary HDUs is verified when reading/writing a FITS file (#32) - Improved documentation (in warning messages as well as in the handbook) that PyFITS uses zero-based indexing (as one would expect for C/Python code, but contrary to the PyFITS standard which was written with FORTRAN in mind) (#68) - Fixed a bug where updating a header card comment could cause the value to be lost if it had not already been read from the card image string. - Fixed a related bug where changes made directly to Card object in a header (i.e. assigning directly to card.value or card.comment) would not propagate when flushing changes to the file (#69) [Note: This and the bug above it were originally reported as being fixed in version 3.0.1, but the fix was never included in the release.] - Improved file handling, particularly in Python 3 which had a few small file I/O-related bugs (#76) - Fixed a bug where updating a FITS file would sometimes cause it to lose its original file permissions (#79) - Fixed the handling of TDIMn keywords; 3.0 added support for them, but got the axis order backards (they were treated as though they were row-major) (#82) - Fixed a crash when a FITS file containing scaled data is opened and immediately written to a new file without explicitly viewing the data first (#84) - Fixed a bug where creating a table with columns named either 'names' or 'formats' resulted in an infinite recursion (#86) 3.0.1 (2011-09-12) ------------------ - Fixed a bug where updating a header card comment could cause the value to be lost if it had not already been read from the card image string. - Changed ``_TableBaseHDU.data`` so that if the data contain an empty table a ``FITS_rec`` object with zero rows is returned rather than ``None``. - The ``.key`` attribute of ``RecordValuedKeywordCards`` now returns the full keyword+field-specifier value, instead of just the plain keyword (#46) - Fixed a related bug where changes made directly to Card object in a header (i.e. assigning directly to card.value or card.comment) would not propagate when flushing changes to the file (#69) - Fixed a bug where writing a table with zero rows could fail in some cases (#72) - Miscellanous small bug fixes that were causing some tests to fail, particularly on Python 3 (#74, #75) - Fixed a bug where creating a table column from an array in non-native byte order would not preserve the byte order, thus interpreting the column array using the wrong byte order (#77) 3.0.0 (2011-08-23) -------------------- - Contains major changes, bumping the version to 3.0 - Large amounts of refactoring and reorganization of the code; tried to preserve public API backwards-compatibility with older versions (private API has many changes and is not guaranteed to be backwards-compatible). There are a few small public API changes to be aware of: * The pyfits.rec module has been removed completely. If your version of numpy does not have the numpy.core.records module it is too old to be used with PyFITS. * The ``Header.ascardlist()`` method is deprecated--use the ``.ascard`` attribute instead. * ``Card`` instances have a new ``.cardimage`` attribute that should be used rather than ``.ascardimage()``, which may become deprecated. * The ``Card.fromstring()`` method is now a classmethod. It returns a new ``Card`` instance rather than modifying an existing instance. * The ``req_cards()`` method on HDU instances has changed: The ``pos`` argument is not longer a string. It is either an integer value (meaning the card's position must match that value) or it can be a function that takes the card's position as it's argument, and returns True if the position is valid. Likewise, the ``test`` argument no longer takes a string, but instead a function that validates the card's value and returns True or False. * The ``get_coldefs()`` method of table HDUs is deprecated. Use the ``.columns`` attribute instead. * The ``ColDefs.data`` attribute is deprecated--use ``ColDefs.columns`` instead (though in general you shouldn't mess with it directly--it might become internal at some point). * ``FITS_record`` objects take ``start`` and ``end`` as arguments instead of ``startColumn`` and ``endColumn`` (these are rarely created manually, so it's unlikely that this change will affect anyone). * ``BinTableHDU.tcreate()`` is now a classmethod, and returns a new ``BinTableHDU`` instance. * Use ``ExtensionHDU`` and ``NonstandardExtHDU`` for making new extension HDU classes. They are now public interfaces, wheres previously they were private and prefixed with underscores. * Possibly others--please report if you find any changes that cause difficulties. - Calls to deprecated functions will display a Deprecation warning. However, in Python 2.7 and up Deprecation warnings are ignored by default, so run Python with the `-Wd` option to see if you're using any deprecated functions. If we get close to actually removing any functions, we might make the Deprecation warnings display by default. - Added basic Python 3 support - Added support for multi-dimensional columns in tables as specified by the TDIMn keywords (#47) - Fixed a major memory leak that occurred when creating new tables with the ``new_table()`` function (#49) be padded with zero-bytes) vs ASCII tables (where strings are padded with spaces) (#15) - Fixed a bug in which the case of Random Access Group parameters names was not preserved when writing (#41) - Added support for binary table fields with zero width (#42) - Added support for wider integer types in ASCII tables; although this is non- standard, some GEIS images require it (#45) - Fixed a bug that caused the index_of() method of HDULists to crash when the HDUList object is created from scratch (#48) - Fixed the behavior of string padding in binary tables (where strings should be padded with nulls instead of spaces) - Fixed a rare issue that caused excessive memory usage when computing checksums using a non-standard block size (see r818) - Add support for forced uint data in image sections (#53) - Fixed an issue where variable-length array columns were not extended when creating a new table with more rows than the original (#54) - Fixed tuple and list-based indexing of FITS_rec objects (#55) - Fixed an issue where BZERO and BSCALE keywords were appended to headers in the wrong location (#56) - ``FITS_record`` objects (table rows) have full slicing support, including stepping, etc. (#59) - Fixed a bug where updating multiple files simultaneously (such as when running parallel processes) could lead to a race condition with mktemp() (#61) - Fixed a bug where compressed image headers were not in the order expected by the funpack utility (#62)

What's New in libchewing 0.4.0 (Apr 11, 2014) --------------------------------------------------------- * Add the following APIs: - Userphrase manipulation - chewing_userphrase_enumerate - chewing_userphrase_has_next - chewing_userphrase_get - chewing_userphrase_add - chewing_userphrase_remove - chewing_userphrase_lookup - Keyboardless action - chewing_cand_open - chewing_cand_close - chewing_cand_string_by_index_static - chewing_cand_choose_by_index - chewing_cand_list_first - chewing_cand_list_last - chewing_cand_list_has_next - chewing_cand_list_has_prev - chewing_cand_list_next - chewing_cand_list_prev - chewing_commit_preedit_buf - chewing_clean_preedit_buf - chewing_clean_bopomofo_buf - Bopomofo buffer - chewing_bopomofo_Check - chewing_bopomofo_String_static - Static API - chewing_aux_String_static - chewing_buffer_String_static - chewing_cand_String_static - chewing_commit_String_static - chewing_kbtype_String_static - Misc - chewing_new2 * Dictionary format: - Remove text data support - Introduce platform independent binary data representation - Move binary data to share/libchewing/* * Userphrase format: - Support SQLite-based user phrase storage and manipulation * New coding style: - Use scripts/indent.sh to change coding style * Bug fixed: - Fix chewing_zuin_Check wrong return value #62 - Fix numlock key cannot output number #45 - Fix negative frequency problem #75 - Fix cannot select candidate when symbol exists #79 - Avoid clearing the buffer when Up arrow key is pressed and escCleanAllBuf is 1. - Fix cannot handle numlock 9 correctly #88 - Fix invalid characters are committed when pre-edit buffer is full - Fix AutoLearnPhrase cannot fill correct wordSeq #108 - Fix wrong auto learn when symbol in between #112 - Fix missing rpl_malloc for cross compilation #118 - Fix '\n' appear in symbol table #126 - Reopen candidate list for symbols with down key #127 #135 - Let up key close candidate list of symbols #131 #138 * Misc: - Add `make check' in cmake build - Change MAX_UTF8_SIZE to 4 - Update official website URL - Use CreateFileA() instead of CreateFile() to stick with the ANSI API - Do not check phrase longer than MAX_PHRASE_LEN - Implement dynamic programming phrasing algorithm #73 - Clean bopomofo buffer when mode change to English mode #89 #97 - Update README to GFM. - Replace 'INCLUDE' with 'AM_CPPFLAGS' #68 #114 #116 - Use [Shift]+[-] to input EM DASH (u+2014) as default #113 #124 - Use Backspace to close candidate window #128 - Use PgUp/PgDn to turn pages of candidates #129 - Update Windows default search path #78 - The word "zuin", "Zhuyin" are obsolete terms for "Bopomofo" - Add const whenever possible in several APIs - Toggle shape mode with shift-space #141 #142 * Successful stories: - chewing-windows-tsf project provides Text Services Framework support for Microsoft Windows.

Update DEPENDS Add test target Upstream changes: 2015-03-14 -- 1.4.3 * Remove three warnings: star-args, abstract-class-little-used, abstract-class-not-used. These warnings don't add any real value and they don't imply errors or problems in the code. * Added a new option for controlling the peephole optimizer in astroid. The option ``--optimize-ast`` will control the peephole optimizer, which is used to optimize a couple of AST subtrees. The current problem solved by the peephole optimizer is when multiple joined strings, with the addition operator, are encountered. If the numbers of such strings is high enough, Pylint will then fail with a maximum recursion depth exceeded error, due to its visitor architecture. The peephole just transforms such calls, if it can, into the final resulting string and this exhibit a problem, because the visit_binop method stops being called (in the optimized AST it will be a Const node). 2015-03-11 -- 1.4.2 * Don't require a docstring for empty modules. Closes issue #261. * Fix a false positive with `too-few-format-args` string warning, emitted when the string format contained a normal positional argument ('{0}'), mixed with a positional argument which did an attribute access ('{0.__class__}'). Closes issue #463. * Take in account all the methods from the ancestors when checking for too-few-public-methods. Closes issue #471. * Catch enchant errors and emit 'invalid-characters-in-docstring' when checking for spelling errors. Closes issue #469. * Use all the inferred statements for the super-init-not-called check. Closes issue #389. * Add a new warning, 'unichr-builtin', emitted by the Python 3 porting checker, when the unichr builtin is found. Closes issue #472. * Add a new warning, 'intern-builtin', emitted by the Python 3 porting checker, when the intern builtin is found. Closes issue #473. * Add support for editable installations. * The HTML output accepts the `--msg-template` option. Patch by Dan Goldsmith. * Add 'map-builtin-not-iterating' (replacing 'implicit-map-evaluation'), 'zip-builtin-not-iterating', 'range-builtin-not-iterating', and 'filter-builtin-not-iterating' which are emitted by `--py3k` when the appropriate built-in is not used in an iterating context (semantics taken from 2to3). * Add a new warning, 'unidiomatic-typecheck', emitted when an explicit typecheck uses type() instead of isinstance(). For example, `type(x) == Y` instead of `isinstance(x, Y)`. Patch by Chris Rebert. Closes issue #299. * Add support for combining the Python 3 checker mode with the --jobs flag (--py3k and --jobs). Closes issue #467. * Add a new warning for the Python 3 porting checker, 'using-cmp-argument', emitted when the `cmp` argument for the `list.sort` or `sorted builtin` is encountered. * Make the --py3k flag commutative with the -E flag. Also, this patch fixes the leaks of error messages from the Python 3 checker when the errors mode was activated. Closes issue #437. 2015-01-16 -- 1.4.1 * Look only in the current function's scope for bad-super-call. Closes issue #403. * Check the return of properties when checking for not-callable. Closes issue #406. * Warn about using the input() or round() built-ins for Python 3. Closes issue #411. * Proper abstract method lookup while checking for abstract-class-instantiated. Closes issue #401. * Use a mro traversal for finding abstract methods. Closes issue #415. * Fix a false positive with catching-non-exception and tuples of exceptions. * Fix a false negative with raising-non-exception, when the raise used an uninferrable exception context. * Fix a false positive on Python 2 for raising-bad-type, when raising tuples in the form 'raise (ZeroDivisionError, None)'. * Fix a false positive with invalid-slots-objects, where the slot entry was an unicode string on Python 2. Closes issue #421. * Add a new warning, 'redundant-unittest-assert', emitted when using unittest's methods assertTrue and assertFalse with constant value as argument. Patch by Vlad Temian. * Add a new JSON reporter, usable through -f flag. * Add the method names for the 'signature-differs' and 'argument-differs' warnings. Closes issue #433. * Don't compile test files when installing. * Fix a crash which occurred when using multiple jobs and the files given as argument didn't exist at all. 2014-11-23 -- 1.4.0 * Added new options for controlling the loading of C extensions. By default, only C extensions from the stdlib will be loaded into the active Python interpreter for inspection, because they can run arbitrary code on import. The option `--extension-pkg-whitelist` can be used to specify modules or packages that are safe to load. * Change default max-line-length to 100 rather than 80 * Drop BaseRawChecker class which were only there for backward compat for a while now * Don't try to analyze string formatting with objects coming from function arguments. Closes issue #373. * Port source code to be Python 2/3 compatible. This drops the need for 2to3, but does drop support for Python 2.5. * Each message now comes with a confidence level attached, and can be filtered base on this level. This allows to filter out all messages that were emitted even though an inference failure happened during checking. * Improved presenting unused-import message. Closes issue #293. * Add new checker for finding spelling errors. New messages: wrong-spelling-in-comment, wrong-spelling-in-docstring. New options: spelling-dict, spelling-ignore-words. * Add new '-j' option for running checks in sub-processes. * Added new checks for line endings if they are mixed (LF vs CRLF) or if they are not as expected. New messages: mixed-line-endings, unexpected-line-ending-format. New option: expected-line-ending-format. * 'dangerous-default-value' no longer evaluates the value of the arguments, which could result in long error messages or sensitive data being leaked. Closes issue #282 * Fix a false positive with string formatting checker, when encountering a string which uses only position-based arguments. Closes issue #285. * Fix a false positive with string formatting checker, when using keyword argument packing. Closes issue #288. * Proper handle class level scope for lambdas. * Handle 'too-few-format-args' or 'too-many-format-args' for format strings with both named and positional fields. Closes issue #286. * Analyze only strings by the string format checker. Closes issue #287. * Properly handle nested format string fields. Closes issue #294. * Don't emit 'attribute-defined-outside-init' if the attribute was set by a function call in a defining method. Closes issue #192. * Properly handle unicode format strings for Python 2. Closes issue #296. * Don't emit 'import-error' if an import was protected by a try-except, which excepted ImportError. * Fix an 'unused-import' false positive, when the error was emitted for all the members imported with 'from import' form. Closes issue #304. * Don't emit 'invalid-name' when assigning a name in an ImportError handler. Closes issue #302. * Don't count branches from nested functions. * Fix a false positive with 'too-few-format-args', when the format strings contains duplicate manual position arguments. Closes issue #310. * fixme regex handles comments without spaces after the hash. Closes issue #311. * Don't emit 'unused-import' when a special object is imported (__all__, __doc__ etc.). Closes issue #309. * Look in the metaclass, if defined, for members not found in the current class. Closes issue #306. * Don't emit 'protected-access' if the attribute is accessed using a property defined at the class level. * Detect calls of the parent's __init__, through a binded super() call. * Check that a class has an explicitly defined metaclass before emitting 'old-style-class' for Python 2. * Emit 'catching-non-exception' for non-class nodes. Closes issue #303. * Order of reporting is consistent. * Add a new warning, 'boolean-datetime', emitted when an instance of 'datetime.time' is used in a boolean context. Closes issue #239. * Fix a crash which ocurred while checking for 'method-hidden', when the parent frame was something different than a function. * Generate html output for missing files. Closes issue #320. * Fix a false positive with 'too-many-format-args', when the format string contains mixed attribute access arguments and manual fields. Closes issue #322. * Extend the cases where 'undefined-variable' and 'used-before-assignment' can be detected. Closes issue #291. * Add support for customising callback identifiers, by adding a new '--callbacks' command line option. Closes issue #326. * Add a new warning, 'logging-format-interpolation', emitted when .format() string interpolation is used within logging function calls. * Don't emit 'unbalanced-tuple-unpacking' when the rhs of the assignment is a variable length argument. Closes issue #329. * Add a new warning, 'inherit-non-class', emitted when a class inherits from something which is not a class. Closes issue #331. * Fix another false positives with 'undefined-variable', where the variable can be found as a class assignment and used in a function annotation. Closes issue #342. * Handle assignment of the string format method to a variable. Closes issue #351. * Support wheel packaging format for PyPi. Closes issue #334. * Check that various built-ins that do not exist in Python 3 are not used: apply, basestring, buffer, cmp, coerce, execfile, file, long raw_input, reduce, StandardError, unicode, reload and xrange. * Warn for magic methods which are not used in any way in Python 3: __coerce__, __delslice__, __getslice__, __setslice__, __cmp__, __oct__, __nonzero__ and __hex__. * Don't emit 'assigning-non-slot' when the assignment is for a property. Closes issue #359. * Fix for regression: '{path}' was no longer accepted in '--msg-template'. * Report the percentage of all messages, not just for errors and warnings. Closes issue #319. * 'too-many-public-methods' is reported only for methods defined in a class, not in its ancestors. Closes issue #248. * 'too-many-lines' disable pragma can be located on any line, not only the first. Closes issue #321. * Warn in Python 2 when an import statement is found without a corresponding `from __future__ import absolute_import`. * Warn in Python 2 when a non-floor division operation is found without a corresponding `from __future__ import division`. * Add a new option, 'exclude-protected', for excluding members from the protected-access warning. Closes issue #48. * Warn in Python 2 when using dict.iter*(), dict.view*(); none of these methods are available in Python 3. * Warn in Python 2 when calling an object's next() method; Python 3 uses __next__() instead. * Warn when assigning to __metaclass__ at a class scope; in Python 3 a metaclass is specified as an argument to the 'class' statement. * Warn when performing parameter tuple unpacking; it is not supported in Python 3. * 'abstract-class-instantiated' is also emitted for Python 2. It was previously disabled. * Add 'long-suffix' error, emitted when encountering the long suffix on numbers. * Add support for disabling a checker, by specifying an 'enabled' attribute on the checker class. * Add a new CLI option, --py3k, for enabling Python 3 porting mode. This mode will disable all other checkers and will emit warnings and errors for constructs which are invalid or removed in Python 3. * Add 'old-octal-literal' to Python 3 porting checker, emitted when encountering octals with the old syntax. * Add 'implicit-map-evaluation' to Python 3 porting checker, emitted when encountering the use of map builtin, without explicit evaluation. 2014-07-26 -- 1.3.0 * Allow hanging continued indentation for implicitly concatenated strings. Closes issue #232. * Pylint works under Python 2.5 again, and its test suite passes. * Fix some false positives for the cellvar-from-loop warnings. Closes issue #233. * Return new astroid class nodes when the inferencer can detect that that result of a function invocation on a type (like `type` or `abc.ABCMeta`) is requested. Closes #205. * Emit 'undefined-variable' for undefined names when using the Python 3 `metaclass=` argument. * Checkers respect priority now. Close issue #229. * Fix a false positive regarding W0511. Closes issue #149. * Fix unused-import false positive with Python 3 metaclasses (#143). * Don't warn with 'bad-format-character' when encountering the 'a' format on Python 3. * Add multiple checks for PEP 3101 advanced string formatting: 'bad-format-string', 'missing-format-argument-key', 'unused-format-string-argument', 'format-combined-specification', 'missing-format-attribute' and 'invalid-format-index'. * Issue broad-except and bare-except even if the number of except handlers is different than 1. Fixes issue #113. * Issue attribute-defined-outside-init for all cases, not just for the last assignment. Closes issue #262. * Emit 'not-callable' when calling properties. Closes issue #268. * Fix a false positive with unbalanced iterable unpacking, when encountering starred nodes. Closes issue #273. * Add new checks, 'invalid-slice-index' and 'invalid-sequence-index' for invalid sequence and slice indices. * Add 'assigning-non-slot' warning, which detects assignments to attributes not defined in slots. * Don't emit 'no-name-in-module' for ignored modules. Closes issue #223. * Fix an 'unused-variable' false positive, where the variable is assigned through an import. Closes issue #196. * Definition order is considered for classes, function arguments and annotations. Closes issue #257. * Don't emit 'unused-variable' when assigning to a nonlocal. Closes issue #275. * Do not let ImportError propagate from the import checker, leading to crash in some namespace package related cases. Closes issue #203. * Don't emit 'pointless-string-statement' for attribute docstrings. Closes issue #193. * Use the proper mode for pickle when opening and writing the stats file. Closes issue #148. * Don't emit hidden-method message when the attribute has been monkey-patched, you're on your own when you do that. * Only emit attribute-defined-outside-init for definition within the same module as the offended class, avoiding to mangle the output in some cases. * Don't emit 'unnecessary-lambda' if the body of the lambda call contains call chaining. Closes issue #243. * Don't emit 'missing-docstring' when the actual docstring uses `.format`. Closes issue #281. 2014-04-30 -- 1.2.1 * Restore the ability to specify the init-hook option via the configuration file, which was accidentally broken in 1.2.0. * Add a new warning [bad-continuation] for badly indentend continued lines. * Emit [assignment-from-none] when the function contains bare returns. Fixes BitBucket issue #191. * Added a new warning for closing over variables that are defined in loops. Fixes Bitbucket issue #176. * Do not warn about \u escapes in string literals when Unicode literals are used for Python 2.*. Fixes BitBucket issue #151. * Extend the checking for unbalanced-tuple-unpacking and unpacking-non-sequence to instance attribute unpacking as well. * Fix explicit checking of python script (1.2 regression, #219) * Restore --init-hook, renamed accidentally into --init-hooks in 1.2.0 (#211) * Add 'indexing-exception' warning, which detects that indexing an exception occurs in Python 2 (behaviour removed in Python 3). 2014-04-18 -- 1.2.0 * Pass the current python paths to pylint process when invoked via epylint. Fixes BitBucket issue #133. * Add -i / --include-ids and -s / --symbols back as completely ignored options. Fixes BitBucket issue #180. * Extend the number of cases in which logging calls are detected. Fixes bitbucket issue #182. * Improve pragma handling to not detect pylint:* strings in non-comments. Fixes BitBucket issue #79. * Do not crash with UnknownMessage if an unknown message ID/name appears in disable or enable in the configuration. Patch by Cole Robinson. Fixes bitbucket issue #170. * Add new warning 'eval-used', checking that the builtin function `eval` was used. * Make it possible to show a naming hint for invalid name by setting include-naming-hint. Also make the naming hints configurable. Fixes BitBucket issue #138. * Added support for enforcing multiple, but consistent name styles for different name types inside a single module; based on a patch written by [email protected]. * Also warn about empty docstrings on overridden methods; contributed by [email protected]. * Also inspect arguments to constructor calls, and emit relevant warnings; contributed by [email protected]. * Added a new configuration option logging-modules to make the list of module names that can be checked for 'logging-not-lazy' et. al. configurable; contributed by [email protected]. * ensure init-hooks is evaluated before other options, notably load-plugins (#166) * Python 2.5 support restored: fixed small issues preventing pylint to run on python 2.5. Bitbucket issues #50 and #62. * bitbucket #128: pylint doesn't crash when looking for used-before-assignment in context manager assignments. * Add new warning, 'bad-reversed-sequence', for checking that the reversed() builtin receive a sequence (implements __getitem__ and __len__, without being a dict or a dict subclass) or an instance which implements __reversed__. * Mark `file` as a bad function when using python2 (closes #8). * Add new warning 'bad-exception-context', checking that `raise ... from ...` uses a proper exception context (None or an exception). * Enhance the check for 'used-before-assignment' to look for 'nonlocal' uses. * Emit 'undefined-all-variable' if a package's __all__ variable contains a missing submodule (closes #126). * Add a new warning 'abstract-class-instantiated' for checking that abstract classes created with `abc` module and with abstract methods are instantied. * Do not warn about 'return-arg-in-generator' in Python 3.3+. * Do not warn about 'abstract-method' when the abstract method is implemented through assignment (#155). * Improve cyclic import detection in the case of packages, patch by Buck Golemon * Add new warnings for checking proper class __slots__: `invalid-slots-object` and `invalid-slots`. * Search for rc file in `~/.config/pylintrc` if `~/.pylintrc` doesn't exists (#121) * Don't register the newstyle checker w/ python >= 3 * Fix unused-import false positive w/ augment assignment (#78) * Fix access-member-before-definition false negative wrt aug assign (#164) * Do not attempt to analyze non python file, eg .so file (#122)

@jeremyevans

== 2.6.1 / 2015-05-25 * Bugs: * Make columnar store handle all supported extensions, not just the first. * Avoid circular require when using the columnar store. == 2.6 / 2015-05-25 * New Feature: * Columnar data storage for the MIME::Types registry, contributed by Jeremy Evans (@jeremyevans). Reduces default memory use substantially (the mail gem drops from 19 Mib to about 3 Mib). Resolves {#96}[mime-types/ruby-mime-types#96], {#94}[mime-types/ruby-mime-types#94], {#83}[mime-types/ruby-mime-types#83]. Partially addresses {#64}[mime-types/ruby-mime-types#64] and {#62}[mime-types/ruby-mime-types#62]. * Development: * Removed caching of deprecation messages in preparation for mime-types 3.0. Now, deprecated methods will always warn their deprecation instead of only warning once. * Added a logger for deprecation messages. * Renamed <tt>lib/mime.rb</tt> to <tt>lib/mime/deprecations.rb</tt> to not conflict with the {mime}[https://rubygems.org/gems/mime] gem on behalf of the maintainers of the {Praxis Framework}[http://praxis-framework.io/]. Provided by Josep M. Blanquer (@blanquer), {#100}[mime-types/ruby-mime-types#100]. * Added the columnar data conversion tool, also provided by Jeremy Evans. * Documentation: * Improved documentation and ensured that all deprecated methods are marked as such in the documentation. * Development: * Added more Ruby variants to Travis CI. * Silenced deprecation messages for internal tools. Noisy deprecations are noisy, but that's the point. == 2.5 / 2015-04-25 * Bugs: * David Genord (@albus522) fixed a bug in loading MIME::types cache where a container loaded from cache did not have the expected +default_proc+, {#86}[mime-types/ruby-mime-types#86]. * Richard Schneeman (@schneems) provided a patch that substantially reduces unnecessary allocations. * Documentation: * Tibor Szolár (@flexik) fixed a typo in the README, {#82}[mime-types/ruby-mime-types#82] * Fixed {#80}[mime-types/ruby-mime-types#80], clarifying the relationship of MIME::Type#content_type and MIME::Type#simplified, with Ken Ip (@kenips). * Development: * Juanito Fatas (@JuanitoFatas) enabled container mode on Travis CI, {#87}[mime-types/ruby-mime-types#87]. * Moved development to a mime-types organization under {mime-types/ruby-mime-types}[https://github.com/mime-types/ruby-mime-types].

Version 1.4.1 - 05 May 2015 * Fix for WAL archival stop working if first backup is EMPTY (Closes: #64) * Fix exception during error handling in Barman recovery (Closes: #65) * After a backup, limit cron activity to WAL archiving only (Closes: #62) * Improved robustness and error reporting of the backup delete command (Closes: #63) * Fix computation of WAL production ratio as reported in the show-backup command * Improved management of xlogb file, which is now correctly fsynced when updated. Also, the rebuild-xlogdb command now operates on a temporary new file, which overwrites the main one when finished. * Add unit tests for dateutil module compatibility * Modified Barman version following PEP 440 rules and added support of tests in Python 3.4

Update LICENSE HOMEPAGE and MASTER_SITES Now it is python3 compatible Complete test target Upstream changes: 3.0.8 2015-06-23 [NEW] Added a monitoring guide to the documentation. [FIX] Improved packaging (thanks to Larissa Reis). [FIX] Fixed and improved various test cases. 3.0.7 2015-03-01 [FIX] State of resources and requests were inconsistent before the request has been processed (issue #62). [FIX] Empty conditions were never triggered (regression in 3.0.6, issue #63). [FIX] Environment.run() will fail if the until event does not get triggered (issue #64). [FIX] Callback modification during event processing is now prohibited (thanks to Andreas Beham). 3.0.6 - 2015-01-30 [NEW] Guide to SimPy resources. [CHANGE] Improve performance of condition events. [CHANGE] Improve performance of filter store (thanks to Christoph Körner). [CHANGE] Exception tracebacks are now more compact. [FIX] AllOf conditions handle already processed events correctly (issue #52). [FIX] Add sync() to RealtimeEnvironment to reset its internal wall-clock reference time (issue #42). [FIX] Only send copies of exceptions into processes to prevent traceback modifications. [FIX] Documentation improvements. 3.0.5 2014-05-14 [CHANGE] Move interruption and all of the safety checks into a new event (pull request #30) [FIX] FilterStore.get() now behaves correctly (issue #49). [FIX] Documentation improvements. 3.0.4 2014-04-07 [NEW] Verified, that SimPy works on Python 3.4. [NEW] Guide to SimPy events [CHANGE] The result dictionary for condition events (AllOF / & and AnyOf / |) now is an OrderedDict sorted in the same way as the original events list. [CHANGE] Condition events now also except processed events. [FIX] Resource.request() directly after Resource.release() no longer successful. The process now has to wait as supposed to. [FIX] Event.fail() now accept all exceptions derived from BaseException instead of only Exception. 3.0.3 2014-03-06 [NEW] Guide to SimPy basics. [NEW] Guide to SimPy Environments. [FIX] Timing problems with real time simulation on Windows (issue #46). [FIX] Installation problems on Windows due to Unicode errors (issue #41). [FIX] Minor documentation issues. 3.0.2 2013-10-24 [FIX] The default capacity for Container and FilterStore is now also inf. 3.0.1 2013-10-24 [FIX] Documentation and default parameters of Store didn’t match. Its default capacity is now inf. 3.0 2013-10-11 SimPy 3 has been completely rewritten from scratch. Our main goals were to simplify the API and code base as well as making SimPy more flexible and extensible. Some of the most important changes are: Stronger focus on events. Processes yield event instances and are suspended until the event is triggered. An example for an event is a timeout (formerly known as hold), but even processes are now events, too (you can wait until a process terminates). Events can be combined with & (and) and | (or) to create condition events. Process can now be defined by any generator function. You don’t have to subclass Process anymore. No more global simulation state. Every simulation stores its state in an environment which is comparable to the old Simulation class. Improved resource system with newly added resource types. Removed plotting and GUI capabilities. Pyside and matplotlib are much better with this. Greatly improved test suite. Its cleaner, and the tests are shorter and more numerous. Completely overhauled documentation.

2015.6.21 ========= ---- * Fix #31: HTML entities stay inside link. * Fix #71: Coverage detects command line tests. * Fix #39: Documentation update. * Fix #61: Functionality added for optional use of automatic links. * Feature #80: ``title`` attribute is preserved in both inline and reference links. * Feature #82: More command line options. See docs. 2015.6.12 ========= ---- * Feature #76: Making ``pre`` blocks clearer for further automatic formatting. * Fix #71: Coverage detects tests carried out in ``subprocesses`` 2015.6.6 ======== ---- * Fix #24: ``3.200.3`` vs ``2014.7.3`` output quirks. * Fix #61. Malformed links in markdown output. * Feature #62: Automatic version number. * Fix #63: Nested code, anchor bug. * Fix #64: Proper handling of anchors with content that starts with tags. * Feature #67: Documentation all over the module. * Feature #70: Adding tests for the module. * Fix #73: Typo in config documentation.

Notable changes between 0.5 and 0.6: Options from OpenSSL 1.0.2f Use "any" protocol, but SSL. Merge pull request #20 from Zash/zash/checkissued Method for checking if one certificate issued another Merge pull request #68 from ignacio/master Enables building with LuaRocks and MS compilers Enables building with LuaRocks and MS compilers Merge pull request #56 from gleydsonsoares/Makefile-tweaks Makefile tweaks Keep 'sslv23' for compability, but deprected. (it will be removed in the next version) Merge pull request #62 from gleydsonsoares/update_protocol_samples add TLS_method / rename "sslv23" to "any" / update protocol samples. update protocol samples(bring "tlsv1_2" to clients and "any" to servers) for consistency and readability, rename "sslv23" to "any" since that it is related to {TLS, SSLv23}methods that handles all supported protocols. add TLS_method(). for now, keep SSLv23_method() for compatibility. Update samples (using 'tlsv1'). Merge pull request #55 from gleydsonsoares/ifndef-OPENSSL_NO_SSL3 guard SSLv3_method() with #ifndef OPENSSL_NO_SSL3 Add lsec_testcontext(). bump MACOSX_VERSION fix typo; s,intall,install, guard SSLv3_method() with #ifndef OPENSSL_NO_SSL3 Set flags to compile with internal inet_ntop() by default. Tag "alpha" explicit. MinGW progress. Merge pull request #53 from hishamhm/master Reuse tag in the LuaSec upstream repository. Merge pull request #26 from Tieske/master Update rockspec to fix Windows build Alternative implementation to inet_ntop() for old versions of Windows. Do not hardcode ar added batch files to generate sample certs on Windows Perform all validation before allocating structures Validate signatures too. API changes to root:issued([intermediate]*, cert) Fix inet_ntop() on Windows. Merge branch 'master' of https://github.com/brunoos/luasec Merge branch 'moteus_rock' added bindir to lib section, as mingw links against dll's to be found in bindir updated defines in rockspec Merge branch 'master' of github.com:Tieske/luasec into moteus_rock use winsock 2 Don't set globals from C. Fix unpack(). Stop using module(). Change to luaL_newlib(). Remove luaL_optint() and luaL_checkint(). BSD headers. Merge pull request #21 from Zash/zash/iPAddress-fix iPAddress encoding Stop if we don't have a string. Changed for strict compiles. Fix for LibreSSL/OPENSSL_NO_COMP Problem on Win64, since double does not represent SOCKET_INVALID exactly. - Add a parameter to server:sni(), so that we can accept an unknown name, using the initial context. - Add the method :getsniname() to retrieve the SNI hostname used. Updated (and renamed) rockspec Windows Encode iPAddress fields in human readable form Don't try to encode IP addresses as UTF-8 Return early if ASN1 string is invalid Push nil if unable to encode ASN1 string as UTF-8 Return human readable error message from cert:issued() SNI support. SNI support. Merge pull request #17 from Zash/zash/checkkey Verify that certificate and key belong together Merge pull request #19 from Zash/zash/pubkey Zash/pubkey Add cert:pubkey() to methods registry Add cert:issued(leafcert) for checking chains Check if private key matches cert only if both key and cert are set Check that certificate matches private key Add method for extracting public key, type and size from x509 objects

@savraj

NEWS: Version 2.5.3 ------------- - Updated zoneinfo to 2016d - Fixed parser bug where unambiguous datetimes fail to parse when dayfirst is set to true. (gh issue #233, pr #234) - Bug in zoneinfo file on platforms such as Google App Engine which do not do not allow importing of subprocess.check_call was reported and fixed by @savraj (gh issue #239, gh pr #240) - Fixed incorrect version in documentation (gh issue #235, pr #243) Version 2.5.2 ------------- - Updated zoneinfo to 2016c - Fixed parser bug where yearfirst and dayfirst parameters were not being respected when no separator was present. (gh issue #81 and #217, pr #229) Version 2.5.1 ------------- - Updated zoneinfo to 2016b - Changed MANIFEST.in to explicitly include test suite in source distributions, with help from @koobs (gh issue #193, pr #194, #201, #221) - Explicitly set all line-endings to LF, except for the NEWS file, on a per-repository basis (gh pr #218) - Fixed an issue with improper caching behavior in rruleset objects (gh issue #104, pr #207) - Changed to an explicit error when rrulestr strings contain a missing BYDAY (gh issue #162, pr #211) - tzfile now correctly handles files containing leapcnt (although the leapcnt information is not actually used). Contributed by @hjoukl (gh issue #146, pr #147) - Fixed recursive import issue with tz module (gh pr #204) - Added compatibility between tzwin objects and datetime.time objects (gh issue #216, gh pr #219) - Refactored monolithic test suite by module (gh issue #61, pr #200 and #206) - Improved test coverage in the relativedelta module (gh pr #215) - Adjusted documentation to reflect possibly counter-intuitive properties of RFC-5545-compliant rrules, and other documentation improvements in the rrule module (gh issue #105, gh issue #149 - pointer to the solution by @phep, pr #213). Version 2.5.0 ------------- - Updated zoneinfo to 2016a - zoneinfo_metadata file version increased to 2.0 - the updated updatezinfo.py script will work with older zoneinfo_metadata.json files, but new metadata files will not work with older updatezinfo.py versions. Additionally, we have started hosting our own mirror of the Olson databases on a github pages site (https://dateutil.github.io/tzdata/) (gh pr #183) - dateutil zoneinfo tarballs now contain the full zoneinfo_metadata file used to generate them. (gh issue #27, gh pr #85) - relativedelta can now be safely subclassed without derived objects reverting to base relativedelta objects as a result of arithmetic operations. (lp:1010199, gh issue #44, pr #49) - relativedelta 'weeks' parameter can now be set and retrieved as a property of relativedelta instances. (lp: 727525, gh issue #45, pr #49) - relativedelta now explicitly supports fractional relative weeks, days, hours, minutes and seconds. Fractional values in absolute parameters (year, day, etc) are now deprecated. (gh issue #40, pr #190) - relativedelta objects previously did not use microseconds to determine of two relativedelta objects were equal. This oversight has been corrected. Contributed by @elprans (gh pr #113) - rrule now has an xafter() method for retrieving multiple recurrences after a specified date. (gh pr #38) - str(rrule) now returns an RFC2445-compliant rrule string, contributed by @schinckel and @armicron (lp:1406305, gh issue #47, prs #50, #62 and #160) - rrule performance under certain conditions has been significantly improved thanks to a patch contributed by @dekoza, based on an article by Brian Beck (@exogen) (gh pr #136) - The use of both the 'until' and 'count' parameters is now deprecated as inconsistent with RFC2445 (gh pr #62, #185) - Parsing an empty string will now raise a ValueError, rather than returning the datetime passed to the 'default' parameter. (gh issue #78, pr #187) - tzwinlocal objects now have a meaningful repr() and str() implementation (gh issue #148, prs #184 and #186) - Added equality logic for tzwin and tzwinlocal objects. (gh issue #151, pr #180, #184) - Added some flexibility in subclassing timelex, and switched the default behavior over to using string methods rather than comparing against a fixed list. (gh pr #122, #139) - An issue causing tzstr() to crash on Python 2.x was fixed. (lp: 1331576, gh issue #51, pr #55) - An issue with string encoding causing exceptions under certain circumstances when tzname() is called was fixed. (gh issue #60, #74, pr #75) - Parser issue where calling parse() on dates with no day specified when the day of the month in the default datetime (which is "today" if unspecified) is greater than the number of days in the parsed month was fixed (this issue tended to crop up between the 29th and 31st of the month, for obvious reasons) (canonical gh issue #25, pr #30, #191) - Fixed parser issue causing fuzzy_with_tokens to raise an unexpected exception in certain circumstances. Contributed by @MichaelAquilina (gh pr #91) - Fixed parser issue where years > 100 AD were incorrectly parsed. Contributed by @Bachmann1234 (gh pr #130) - Fixed parser issue where commas were not a valid separator between seconds and microseconds, preventing parsing of ISO 8601 dates. Contributed by @RyansS (gh issue #28, pr #106) - Fixed issue with tzwin encoding in locales with non-Latin alphabets (gh issue #92, pr #98) - Fixed an issue where tzwin was not being properly imported on Windows. Contributed by @labrys. (gh pr #134) - Fixed a problem causing issues importing zoneinfo in certain circumstances. Issue and solution contributed by @alexxv (gh issue #97, pr #99) - Fixed an issue where dateutil timezones were not compatible with basic time objects. One of many, many timezone related issues contributed and tested by @labrys. (gh issue #132, pr #181) - Fixed issue where tzwinlocal had an invalid utcoffset. (gh issue #135, pr #141, #142) - Fixed issue with tzwin and tzwinlocal where DST transitions were incorrectly parsed from the registry. (gh issue #143, pr #178) - updatezinfo.py no longer suppresses certain OSErrors. Contributed by @bjamesv (gh pr #164) - An issue that arose when timezone locale changes during runtime has been fixed by @carlosxl and @mjschultz (gh issue #100, prs #107, #109) - Python 3.5 was added to the supported platforms in the metadata (@tacaswell gh pr #159) and the test suites (@moreati gh pr #117). - An issue with tox failing without unittest2 installed in Python 2.6 was fixed by @moreati (gh pr #115) - Several deprecated functions were replaced in the tests by @moreati (gh pr #116) - Improved the logic in Travis and Appveyor to alleviate issues where builds were failing due to connection issues when downloading the IANA timezone files. In addition to adding our own mirror for the files (gh pr #183), the download is now retried a number of times (with a delay) (gh pr #177) - Many failing doctests were fixed by @moreati. (gh pr #120) - Many fixes to the documentation (gh pr #103, gh pr #87 from @radarhere, gh pr #154 from @gpoesia, gh pr #156 from @awsum, gh pr #168 from @ja8zyjits) - Added a code coverage tool to the CI to help improve the library. (gh pr #182) - We now have a mailing list - [email protected], graciously hosted by Python.org. Version 2.4.2 ------------- - Updated zoneinfo to 2015b. - Fixed issue with parsing of tzstr on Python 2.7.x; tzstr will now be decoded if not a unicode type. gh #51 (lp:1331576), gh pr #55. - Fix a parser issue where AM and PM tokens were showing up in fuzzy date stamps, triggering inappropriate errors. gh #56 (lp: 1428895), gh pr #63. - Missing function "setcachesize" removed from zoneinfo __all__ list by @RyansS, fixing an issue with wildcard imports of dateutil.zoneinfo. (gh pr #66). - (PyPi only) Fix an issue with source distributions not including the test suite. Version 2.4.1 ------------- - Added explicit check for valid hours if AM/PM is specified in parser. (gh pr #22, issue #21) - Fix bug in rrule introduced in 2.4.0 where byweekday parameter was not handled properly. (gh pr #35, issue #34) - Fix error where parser allowed some invalid dates, overwriting existing hours with the last 2-digit number in the string. (gh pr #32, issue #31) - Fix and add test for Python 2.x compatibility with boolean checking of relativedelta objects. Implemented by @nimasmi (gh pr #43) and Cédric Krier (lp: 1035038) - Replaced parse() calls with explicit datetime objects in unit tests unrelated to parser. (gh pr #36) - Changed private _byxxx from sets to sorted tuples and fixed one currently unreachable bug in _construct_byset. (gh pr #54) - Additional documentation for parser (gh pr #29, #33, #41) and rrule. - Formatting fixes to documentation of rrule and README.rst. - Updated zoneinfo to 2015a.

New Features ============ Configuration ------------- - CMake: Added NAMESPACE hdf5:: to package configuration files to allow projects using installed HDF5 binaries built with CMake to link with them without specifying the HDF5 library location via IMPORTED_LOCATION. (ADB, 2016/10/17, HDFFV-10003) - CMake: Changed the CTEST_BUILD_CONFIGURATION option to CTEST_CONFIGURATION_TYPE as recommended by the CMake documentation. (ADB, 2016/10/17, HDFFV-9971) - CMake: Added support for GIT (ADB, 2016/07/12) Bug Fixes since HDF5-1.8.17 =========================== Configuration ------------- - Fixed a problem preventing HDF5 to be built on 32-bit CYGWIN by condensing cygwin configuration files into a single file and removing outdated compiler settings. (ABD, 2016/07/12, HDFFV-9946) - CMake: Fixed a command length overflow error by converting custom commands inside CMakeTest.cmake files into regular dependencies and targets. (ABD, 2016/07/12, HDFFV-9939) - CMake: Fixed a timeout error that would occasionally occur when running the virtual file driver tests simultaneously due to test directory and file name collisions. (ABD, 2016/09/19, HDFFV-9431) Library ------- - Fixed a memory leak that would occur when the library allocated memory for an external file prefix (H5Pset_efile_prefix) and failed to free it. (DER, 2016/04/29) - Fixed an error that would occur when calling H5Adelete on an attribute which is attached to an externally linked object in the target file and whose datatype is a committed datatype in the main file. (VC, 2016-07-04, HDFFV-9940) - Fixed a problem where a plugin compiled into a DLL in the default plugin directory could not be found by the HDF5 library at runtime on Windows when the HDF5_PLUGIN_PATH environment variable was not set. (ABD, 2016/08/01, HDFFV-9706) - Fixed an issue where H5Pset_alignment could result in misaligned blocks with some input combinations, causing an assertion failure in debug mode. (NAF, 2016/08/11, HDFFV-9948) - A number of issues were fixed when reading/writing from/to corrupted files to ensure that the library fails gracefully in these cases: * Writing to a corrupted file that has an object message which is incorrectly marked as sharable on disk results in a buffer overflow / invalid write instead of a clean error message. * Decoding data from a corrupted file with a dataset encoded with the H5Z_NBIT decoding can result in a code execution vulnerability under the context of the application using the HDF5 library. * When decoding an array datatype from a corrupted file, the HDF5 library fails to return an error in production if the number of dimensions decoded is greater than the maximum rank. * When decoding an "old style" array datatype from a corrupted file, the HDF5 library fails to return an error in production if the number of dimensions decoded is greater than the maximum rank. (NAF, 2016/10/06, HDFFV-9950, HDFFV-9951, HDFFV-9992, HDFFV-9993) - Fixed an error that would occur when copying an object with an attribute which is a compound datatype consisting of a variable length string. (VC, 2016-10-17, HDFFV-7991) Parallel Library ---------------- - Fixed a bug that could occur when allocating a chunked dataset in parallel with an alignment set and an alignment threshold greater than the chunk size but less than or equal to the raw data aggregator size. (NAF, 2016/08/11, HDFFV-9969) Performance ------------- - None Tools ----- - Fixed an error in the compiler wrapper scripts (h5cc, h5fc, et al.) in which they would erroneously drop the file argument specified via the -o flag when the -o flag was specified before the -c flag on the command line, resulting in a failure to compile. (LRK, 2016/06/08, HDFFV-9938, HDFFV-9530) - h5repack User Defined (UD) filter parameters were not parsed correctly. The UD filter parameters were not being parsed correctly. Reworked coding section to parse the correct values and verify number of parameters. (ADB, 2016/10/19, HDFFV-9996, HDFFV-9974, HDFFV-9515, HDFFV-9039) Fortran API ----------- - Fortran library fails to compile and fails tests with NAG compiler. * Removed the non-standard assumption that KIND=SIZEOF, in the HDF5 configure programs. * Removed Fortran 66 character/integer conversions from tests. * Removed the use of C_SIZEOF in the test programs * Changed to using STORAGE_SIZE in the test programs if available. Otherwise, uses C_SIZEOF or SIZEOF. (MSB, 2016/9/22, HDFFV-9973) - Fortran segfaults for F03 tests with NAG compiler * Removed INTENT(OUT) from 'fillvalue' in F2003 interface for H5Pget_fill_value_f. (MSB, 2016/9/22, HDFFV-9980) C++ API ------- - The macro H5_NO_NAMESPACE is deprecated from the HDF5 C++ API library. In future releases, the macros H5_NO_STD and OLD_HEADER_FILENAME may also be removed. (BMR, 2016/10/27, HDFFV-9532) High-Level APIs: --------------- - The high-level API Packet Table (PT) did not write data correctly when the datatype is a compound type that has string type as one of the members. This problem started in 1.8.15, after the fix of HDFFV-9042 was applied, which caused the Packet Table to use native type to access the data. It should be up to the application to specify whether the buffer to be read into memory in the machine’s native architecture. Thus, the PT is fixed to not use native type but to make a copy of the user's provided datatype during creation or the packet table's datatype during opening. If an application wishes to use native type to read the data, then the application will request that. However, the Packet Table doesn't provide a way to specify memory datatype in this release. This feature will be available in future releases, HDFFV-10023. (BMR, 2016/10/27, HDFFV-9758) Fortran High-Level APIs: ------------------------ - None Testing ------- - None Supported Platforms =================== The following platforms are supported and have been tested for this release. They are built with the configure process unless specified otherwise. Linux 2.6.32-573.22.1.el6 GNU C (gcc), Fortran (gfortran), C++ (g++) #1 SMP x86_64 GNU/Linux compilers: (platypus/mayll) Version 4.4.7 20120313 Versions 4.8.4, 4.9.3, 5.2.0 PGI C, Fortran, C++ for 64-bit target on x86-64; Version 15.7-0 Intel(R) C (icc), C++ (icpc), Fortran (icc) compilers: Version 15.0.3.187 Build 20150407 MPICH 3.1.4 compiled with GCC 4.9.3 Linux 2.6.32-573.18.1.el6 gcc (GCC) 4.4.7 20120313 (Red Hat 4.4.7-16) #1 SMP ppc64 GNU/Linux g++ (GCC) 4.4.7 20120313 (Red Hat 4.4.7-16) (ostrich) GNU Fortran (GCC) 4.4.7 20120313 (Red Hat 4.4.7-16) IBM XL C/C++ V13.1 IBM XL Fortran V15.1 Linux 3.10.0-327.10.1.el7 GNU C (gcc), Fortran (gfortran), C++ (g++) #1 SMP x86_64 GNU/Linux compilers: (kituo/moohan) Version 4.8.5 20150623 (Red Hat 4.8.5-4) Versions 4.9.3, 5.2.0 Intel(R) C (icc), C++ (icpc), Fortran (icc) compilers: Version 15.0.3.187 Build 20150407 MPICH 3.1.4 compiled with GCC 4.9.3 SunOS 5.11 32- and 64-bit Sun C 5.12 SunOS_sparc (emu) Sun Fortran 95 8.6 SunOS_sparc Sun C++ 5.12 SunOS_sparc Windows 7 Visual Studio 2012 w/ Intel Fortran 15 (cmake) Visual Studio 2013 w/ Intel Fortran 15 (cmake) Visual Studio 2015 w/ Intel Fortran 16 (cmake) Cygwin(CYGWIN_NT-6.1 2.2.1(0.289/5/3) gcc(4.9.3) compiler and gfortran) (cmake and autotools) Windows 7 x64 Visual Studio 2012 w/ Intel Fortran 15 (cmake) Visual Studio 2013 w/ Intel Fortran 15 (cmake) Visual Studio 2015 w/ Intel Fortran 16 (cmake) Windows 10 Visual Studio 2015 w/ Intel Fortran 16 (cmake) Windows 10 x64 Visual Studio 2015 w/ Intel Fortran 16 (cmake) Mac OS X Mt. Lion 10.8.5 Apple LLVM version 5.1 (clang-503.0.40) 64-bit gfortran GNU Fortran (GCC) 4.8.2 (swallow/kite) Intel icc/icpc/ifort version 15.0.3 Mac OS X Mavericks 10.9.5 Apple LLVM version 6.0 (clang-600.0.57) 64-bit gfortran GNU Fortran (GCC) 4.9.2 (wren/quail) Intel icc/icpc/ifort version 15.0.3 Mac OS X Yosemite 10.10.5 Apple LLVM version 6.1 (clang-602.0.53) 64-bit gfortran GNU Fortran (GCC) 4.9.2 (osx1010dev/osx1010test) Intel icc/icpc/ifort version 15.0.3 Mac OS X El Capitan 10.11.4 Apple LLVM version 7.3.0 (clang-703.0.29) 64-bit gfortran GNU Fortran (GCC) 5.2.0 (VM osx1011dev/osx1011test) Intel icc/icpc/ifort version 16.0.2 Tested Configuration Features Summary ===================================== In the tables below y = tested n = not tested in this release C = Cluster W = Workstation x = not working in this release dna = does not apply ( ) = footnote appears below second table <blank> = testing incomplete on this feature or platform Platform C F90/ F90 C++ zlib SZIP parallel F2003 parallel SunOS 5.11 32-bit n y/y n y y y SunOS 5.11 64-bit n y/y n y y y Windows 7 y y/y n y y y Windows 7 x64 y y/y n y y y Windows 7 Cygwin n y/y n y y n Windows 10 n y/y n y y y Windows 10 x64 n y/y n y y y Mac OS X Mountain Lion 10.8.5 64-bit n y/y n y y y Mac OS X Mavericks 10.9.5 64-bit n y/y n y y y Mac OS X Yosemite 10.10.5 64-bit n y/y n y y y AIX 6.1 32- and 64-bit n y/n n y y y CentOS 6.7 Linux 2.6.32 x86_64 GNU y y/y y y y y CentOS 6.7 Linux 2.6.32 x86_64 Intel n y/y n y y y CentOS 6.7 Linux 2.6.32 x86_64 PGI n y/y n y y y CentOS 7.1 Linux 3.10.0 x86_64 GNU y y/y y y y y CentOS 7.1 Linux 3.10.0 x86_64 Intel n y/y n y y y Linux 2.6.32-431.11.2.el6.ppc64 n y/n n y y y Platform Shared Shared Shared Thread- C libs F90 libs C++ libs safe SunOS 5.11 32-bit y y y y SunOS 5.11 64-bit y y y y Windows 7 y y y y Windows 7 x64 y y y y Windows 7 Cygwin n n n y Windows 10 y y y y Windows 10 x64 y y y y Mac OS X Mountain Lion 10.8.5 64-bit y n y y Mac OS X Mavericks 10.9.5 64-bit y n y y Mac OS X Yosemite 10.10.5 64-bit y n y y AIX 6.1 32- and 64-bit y n n y CentOS 6.7 Linux 2.6.32 x86_64 GNU y y y y CentOS 6.7 Linux 2.6.32 x86_64 Intel y y y y CentOS 6.7 Linux 2.6.32 x86_64 PGI y y y y CentOS 7.1 Linux 3.10.0 x86_64 GNU y y y y CentOS 7.1 Linux 3.10.0 x86_64 Intel y y y y Linux 2.6.32-431.11.2.el6.ppc64 y y y y Compiler versions for each platform are listed in the preceding "Supported Platforms" table. More Tested Platforms ===================== The following platforms are not supported but have been tested for this release. Linux 2.6.32-573.22.1.el6 g95 (GCC 4.0.3 (g95 0.94!) #1 SMP x86_64 GNU/Linux (platypus) Debian8.4.0 3.16.0-4-amd64 #1 SMP Debian 3.16.36-1 x86_64 GNU/Linux gcc (Debian 4.9.2-10) 4.9.2 GNU Fortran (Debian 4.9.2-10) 4.9.2 (cmake and autotools) Fedora24 4.7.2-201.fc24.x86_64 #1 SMP x86_64 x86_64 x86_64 GNU/Linux gcc (GCC) 6.1.1 20160621 (Red Hat 6.1.1-3) GNU Fortran (GCC) 6.1.1 20160621 (Red Hat 6.1.1-3) (cmake and autotools) CentOS 7.2 3.10.0-327.28.2.el7.x86_64 #1 SMP x86_64 x86_64 x86_64 GNU/Linux gcc (GCC) 4.8.5 20150623 (Red Hat 4.8.5-4) GNU Fortran (GCC) 4.8.5 20150623 (Red Hat 4.8.5-4) (cmake and autotools) Ubuntu 16.04 4.4.0-38-generic #62-Ubuntu SMP x86_64 GNU/Linux gcc (Ubuntu 5.4.0-6ubuntu1~16.04.2) 5.4.0 GNU Fortran (Ubuntu 5.4.0-6ubuntu1~16.04.2) 5.4.0 (cmake and autotools) Known Problems ============== * On windows platforms in debug configurations, the VFD flush1 tests will fail with the split and multi VFD drivers. These tests will display a modal debug dialog which must be answered or wait for the test timeout to expire. (ADB - 2014/06/23 - HDFFV-8851) * CLANG compiler with the options -fcatch-undefined-behavior and -ftrapv catches some undefined behavior in the alignment algorithm of the macro DETECT_I in H5detect.c (Issue 8147). Since the algorithm is trying to detect the alignment of integers, ideally the flag -fcatch-undefined-behavior shouldn't to be used for H5detect.c. In the future, we can separate flags for H5detect.c from the rest of the library. (SLU - 2013/10/16) * Make provided by Solaris fails in "make check". Solaris users should use gmake to build and install the HDF5 software. (AKC - 2013/10/08 - HDFFV-8534) * The C++ and FORTRAN bindings are not currently working on FreeBSD with the native release 8.2 compilers (4.2.1), but are working with gcc 4.6 from the ports (and probably gcc releases after that). (QAK - 2012/10/19) * The following h5dump test case fails in BG/P machines (and potentially other machines that use a command script to launch executables): h5dump --no-compact-subset -d "AHFINDERDIRECT::ah_centroid_t[0] it=0 tl=0" tno-subset.h5 This is due to the embedded spaces in the dataset name being interpreted by the command script launcher as meta-characters, thus passing three arguments to h5dump's -d flag. The command passes if run by hand, just not via the test script. (AKC - 2012/05/03) * The STDIO VFD does not work on some architectures, possibly due to 32/64 bit or large file issues. The basic STDIO VFD test is known to fail on 64-bit SunOS 5.10 on SPARC when built with -m64 and 32-bit OS X/Darwin 10.7.0. The STDIO VFD test has been disabled while we investigate and a fix should appear in a future release. (DER - 2011/10/14 - HDFFV-8235) * h5diff can report inconsistent results when comparing datasets of enum type that contain invalid values. This is due to how enum types are handled in the library and will be addressed in a future release. (DER - 2011/10/14 - HDFFV-7527) * The links test can fail under the stdio VFD due to some issues with external links. This will be investigated and fixed in a future release. (DER - 2011/10/14 - HDFFV-7768) * After the shared library support was fixed for some bugs, it was discovered that "make prefix=XXX install" no longer works for shared libraries. It still works correctly for static libraries. Therefore, if you want to install the HDF5 shared libraries in a location such as /usr/local/hdf5, you need to specify the location via the --prefix option during configure time. E.g, ./configure --prefix=/usr/local/hdf5 ... (AKC - 2011/05/07 - HDFFV-7583) * The parallel test, t_shapesame, in testpar/, may run for a long time and may be terminated by the alarm signal. If that happens, one can increase the alarm seconds (default is 1200 seconds = 20 minutes) by setting the environment variable, $HDF5_ALARM_SECONDS, to a larger value such as 3600 (60 minutes). Note that the t_shapesame test may fail in some systems (see the "While working on the 1.8.6 release..." problem below). If it does, it will waste more time if $HDF5_ALARM_SECONDS is set to a larger value. (AKC - 2011/05/07) * Shared Fortran libraries are not quite working on AIX. While they are generated when --enable-shared is specified, the fortran and hl/fortran tests fail. the issue. HL and C++ shared libraries should now be working as intended, however. (MAM - 2011/04/20) * While working on the 1.8.6 release of HDF5, a bug was discovered that can occur when reading from a dataset in parallel shortly after it has been written to collectively. The issue was exposed by a new test in the parallel HDF5 test suite, but had existed before that. We believe the problem lies with certain MPI implementations and/or file systems. We have provided a pure MPI test program, as well as a standalone HDF5 program, that can be used to determine if this is an issue on your system. They should be run across multiple nodes with a varying number of processes. These programs can be found at: http://www.hdfgroup.org/ftp/HDF5/examples/known_problems/ (NAF - 2011/01/19) * All the VFL drivers aren't backward compatible. In H5FDpublic.h, the structure H5FD_class_t changed in 1.8. There is new parameter added to get_eoa and set_eoa callback functions. A new callback function get_type_map was added in. The public function H5FDrealloc was taken out in 1.8. The problem only happens when users define their own driver for 1.6 and try to plug in 1.8 library. Because there's only one user complaining about it, we (Elena, Quincey, and I) decided to leave it as it is (see bug report #1279). Quincey will make a plan for 1.10. (SLU - 2010/02/02) * The --enable-static-exec configure flag will only statically link libraries if the static version of that library is present. If only the shared version of a library exists (i.e., most system libraries on Solaris, AIX, and Mac, for example, only have shared versions), the flag should still result in a successful compilation, but note that the installed executables will not be fully static. Thus, the only guarantee on these systems is that the executable is statically linked with just the HDF5 library. (MAM - 2009/11/04) * A dataset created or rewritten with a v1.6.3 library or after cannot be read with the v1.6.2 library or before when the Fletcher32 EDC filter is enabled. There was a bug in the calculation of the Fletcher32 checksum in the library before v1.6.3; the checksum value was not consistent between big- endian and little-endian systems. This bug was fixed in Release 1.6.3. However, after fixing the bug, the checksum value was no longer the same as before on little-endian system. Library releases after 1.6.4 can still read datasets created or rewritten with an HDF5 library of v1.6.2 or before. (SLU - 2005/06/30)

v0.9.8 * fsevent_watch : Fix "zombie" processes not dying (#62)

## Changes between 2.0.0 and 2.1.0 ### Ruby Warnings Squashed Contributed by Akira Matsuda. GitHub issue: [#62](ruby-amqp/amq-protocol#62) ### Byte Array Decoding Byte array values in types now can be decoded (to the extent Ruby type system permits) by this library. GitHub issue: [#58](ruby-amqp/amq-protocol#58)

jperkin closed this as completed in f3c0369 Jan 22, 2016

jperkin pushed a commit that referenced this issue Mar 20, 2017

Update rb-fsevent to 0.9.8.

5989f29

v0.9.8 * fsevent_watch : Fix "zombie" processes not dying (#62)

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

OS X 2013Q1 lyx doesn't work #62

OS X 2013Q1 lyx doesn't work #62

rmustacc commented Jul 10, 2013

OS X 2013Q1 lyx doesn't work #62

OS X 2013Q1 lyx doesn't work #62

Comments

rmustacc commented Jul 10, 2013