Headerless tables from DOCX

[denismaier]

Hi, is it possible to read a headerless table from DOCX? Converting from docx always reads the first line as a header, but in some cases this isn't what I want. Is it possible to handle different kinds of tables differently? (In this particular case, I can just run some xsl on the result html as I'm using a xproc post-processing pipeline already. But a generic solution would be much better and also more robust and flexible.)

[massifrg]

I tried to follow the hint for headerless tables in markdown, and it works somehow.

I wrote a docx file with a table like this (the first line is made of blank cells):

+---+---+---+
|   |   |   |
+---+---+---+
| a | b | c |
+---+---+---+
| d | e | f |
+---+---+---+

Then this is the result of pandoc -f docx -t markdown:

  ----------------------- ----------------------- -----------------------
  a                       b                       c

  d                       e                       f
  ----------------------- ----------------------- -----------------------

So the markdown Writer produces a headerless table.

This is the result of pandoc -f docx -t html:

<table>
<colgroup>
<col style="width: 33%" />
<col style="width: 33%" />
<col style="width: 33%" />
</colgroup>
<tbody>
<tr>
<td>a</td>
<td>b</td>
<td>c</td>
</tr>
<tr>
<td>d</td>
<td>e</td>
<td>f</td>
</tr>
</tbody>
</table>

Also the html Writer produces a headerless table.

And this is the result of pandoc -f docx -t native:

[ Table
    ( "" , [] , [] )
    (Caption Nothing [])
    [ ( AlignDefault , ColWidth 0.3333333333333333 )
    , ( AlignDefault , ColWidth 0.3333333333333333 )
    , ( AlignDefault , ColWidth 0.3333333333333333 )
    ]
    (TableHead
       ( "" , [] , [] )
       [ Row
           ( "" , [] , [] )
           [ Cell
               ( "" , [] , [] ) AlignDefault (RowSpan 1) (ColSpan 1) []
           , Cell
               ( "" , [] , [] ) AlignDefault (RowSpan 1) (ColSpan 1) []
           , Cell
               ( "" , [] , [] ) AlignDefault (RowSpan 1) (ColSpan 1) []
           ]
       ])
    [ TableBody
        ( "" , [] , [] )
        (RowHeadColumns 0)
        []
        [ Row
            ( "" , [] , [] )
            [ Cell
                ( "" , [] , [] )
                AlignDefault
                (RowSpan 1)
                (ColSpan 1)
                [ Plain [ Str "a" ] ]
            , Cell
                ( "" , [] , [] )
                AlignDefault
                (RowSpan 1)
                (ColSpan 1)
                [ Plain [ Str "b" ] ]
            , Cell
                ( "" , [] , [] )
                AlignDefault
                (RowSpan 1)
                (ColSpan 1)
                [ Plain [ Str "c" ] ]
            ]
        , Row
            ( "" , [] , [] )
            [ Cell
                ( "" , [] , [] )
                AlignDefault
                (RowSpan 1)
                (ColSpan 1)
                [ Plain [ Str "d" ] ]
            , Cell
                ( "" , [] , [] )
                AlignDefault
                (RowSpan 1)
                (ColSpan 1)
                [ Plain [ Str "e" ] ]
            , Cell
                ( "" , [] , [] )
                AlignDefault
                (RowSpan 1)
                (ColSpan 1)
                [ Plain [ Str "f" ] ]
            ]
        ]
    ]
    (TableFoot ( "" , [] , [] ) [])
]

Looking at the native output, you can see that the first row of blank cell is still there, so I think the headerless table is a result of the Writers (markdown and html).

And, of course, you must provide docx documents with tables that have the first line made of blank cells.
Maybe a filter that moves the head rows to the first body is the best solution.

[jgm]

I think that in the case of the markdown writer, it is, effectively, emitting the empty row but it's being collapsed by some logic that collapses excess blank lines. If you do pandoc -f markdown -t markdown-simple_tables-multiline_tables you'll see that the blank row is preserved in a grid table.

In the case of the HTML writer, there is some logic that emits empty rows:
https://github.com/jgm/pandoc/blob/main/src/Text/Pandoc/Writers/HTML.hs#L1175-L1178
I'm not sure why exactly, the code seems to go back very far and can be found in many writers.

[denismaier]

Would it make sense to assume a header would be formatted differently, e.g. in bold? So, a table without a first row formatted in bold would be headerless?

[jgm]

The docx reader can distinguish between tables with and without a header row: see test/docx/tables.docx, and run that through pandoc. What criteria it uses, I don't recall exactly.

[denismaier]

Thanks – I'll have a look.

[denismaier]

Ok, I found it. There's a setting that sets the first row as a header.

Who could have guessed that... Sorry for the noise.