DOCX Reader: Is it correct behaviour / best solution to use the last title element present in file? #10359

Aspvik · 2024-11-07T08:27:15Z

Aspvik
Nov 7, 2024

If there are multiple title elements/styles in a docx file, Pandoc will use the last one. Is this det logical solution? Would it not be better to use the first one?

jgm · 2024-11-07T15:16:10Z

jgm
Nov 7, 2024
Maintainer

I just tried converting a docx with two elements with the Title style.
Pandoc used the first as the title. What version of pandoc are you using, and can you upload a docx I can test with?

4 replies

Aspvik Nov 8, 2024
Author

Hi!

Thanks for the quick response. First of all we have been using Pandoc for a few years now, and are very grateful for the work you and the maintainers are doing.

Some background: We use Pandoc to convert from docx to epub3, and then post process the chunked xhtml. (Chunked HTML was not an available output when we started using Pandoc, so in the future we should probably use that instead). Our standard command takes advantage of the --epub-title-page=false flag, since we don't use that in the post processing, so we have been extracting the document title from the <title> element in the nav.xhtml file.

Files:

single.docx: Single title
double-empty: Second title element is empty
double-content: Second title element is non-empty

Command:
pandoc double.docx --from docx --to epub3 -o double.epub

Pandoc details:
pandoc 3.5
Features: +server +lua
Scripting engine: Lua 5.4

Results
single.docx
Warning: Non.
nav.xhtml: <title> is set to the first title element.
title_page.xhtml: <title>/<h1.title> is set to the first title element.

double-empty.docx
Warning: This document format requires a nonempty <title> element. Defaulting to 'double' as the title.
nav.xhtml: <title> is set to the filename.
title_page.xhtml: <title>/<h1.title> is set to the first title element.

double-content.docx
Warning: This document format requires a nonempty <title> element. Defaulting to 'double' as the title.
nav.xhtml: <title> is set to the filename.
title_page.xhtml:

<title> is set to a concatenated string of the two titles (First titleSecond title)
<h1.title> is the parent of two <p>, one for each of the two titles.

Summary

There seems to be a mismatch between the <title> element in nav.xhtml and the title used on the title page when there is multiple titles or empty title elements.
There seems to be a unwanted/unexpected handling of multiple titles (one after the other) in title_page.xhtml where they are concatenated in the <title> and producing non-valid HTML (<p> child of <\h1>) in the <body>.

As for my situation I could start converting without the --epub-title-page=false flag, and get the title from there instead. However the last case (double-content.docx) prevents this from being a viable option.

single.docx
double-empty.docx
double-content.docx

jgm Nov 8, 2024
Maintainer

Pandoc parses the metadata in double-content.docx as follows:

  Meta
    { unMeta =
        fromList
          [ ( "title"
            , MetaBlocks
                [ Para [ Str "First" , Space , Str "title" ]
                , Para [ Str "Second" , Space , Str "title" ]
                ]
            )
          ]
    }

So, you have a single title with a two-paragraph content.

In HTML this comes out as

<h1 class="title"><p>First title</p>
<p>Second title</p></h1>

which as you observe is malformed. The HTML writer should do something to ensure that the title variable contains only inline content. It looks like we already do this for the pagetitle, and that is why you get the two titles concatenated. (Perhaps we could do something better in this case.) It's also interesting that nav.xhtml isn't seeing the title; this is also probably related to it being a two-paragraph block, but I'll have to look into that.

jgm Nov 8, 2024
Maintainer

Or, we could address this in the docx reader, and ensure that when there are multiple blocks marked "Title", only the first gets used as metadata.

Aspvik Nov 11, 2024
Author

I don't have any strong opinions on what is the best solution. But it seems to me that addressing the issue in the docx reader and only parse the first title element as the document title would give the most consistent result.

jgm · 2024-11-11T17:48:37Z

jgm
Nov 11, 2024
Maintainer

I've closed this with a change to the docx reader.
13d37b4

0 replies

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

DOCX Reader: Is it correct behaviour / best solution to use the last title element present in file? #10359

{{title}}

Replies: 2 comments 4 replies

{{title}}

{{title}}

{{title}}

{{title}}

{{title}}

{{editor}}'s edit

{{editor}}'s edit

{{title}}

Select a reply

DOCX Reader: Is it correct behaviour / best solution to use the last title element present in file? #10359

Aspvik Nov 7, 2024

Replies: 2 comments · 4 replies

jgm Nov 7, 2024 Maintainer

Aspvik Nov 8, 2024 Author

jgm Nov 8, 2024 Maintainer

jgm Nov 8, 2024 Maintainer

Aspvik Nov 11, 2024 Author

jgm Nov 11, 2024 Maintainer

Aspvik
Nov 7, 2024

Replies: 2 comments 4 replies

jgm
Nov 7, 2024
Maintainer

Aspvik Nov 8, 2024
Author

jgm Nov 8, 2024
Maintainer

jgm Nov 8, 2024
Maintainer

Aspvik Nov 11, 2024
Author

jgm
Nov 11, 2024
Maintainer