Run a docx validator ahead of providing output

[cscheid]

There are some docx validators out there. If we found a way to run xml validation in typescript, we could check our output before emitting it. (Avoiding situations like #6357)

[cscheid]

Hm. It turns out that this specific docx validator appears to be out of date.

A basic

| foo | bar |
|-----|-----|
| 1   | 2   |

table, rendered to .docx, produces readable input (in my case) for Office for Mac, but fails validation.

[edwintorok]

The XML validation errors are mostly due to wrong ordering of XML tags (apparently both the .xsd and RELAXNG schemas demand a particular ordering for some tags).
See jgm/pandoc#9263 where I'm attempting to fix the ordering in the reference document, which now passes validation (there are more validation errors to fix in pandoc and quarto itself).

I've also used a 2nd validator https://github.com/mikeebowen/OOXML-Validator which uses the open source https://github.com/dotnet/Open-XML-SDK, and that fails in similar ways.

I don't know whether the ordering matters for any application, but probably best to fix them so the real errors can be seen.
E.g. the reference doc had an extra > after a tag (now fixed in the PR above), and at first glance the callout tables created by quarto are missing a tblGrid tag, and have a duplicate w:pPr tag (one added by pandoc itself, the other be Quarto's Lua filter with raw opendocumentxml), which is very likely the reason for the linked bug.

[edwintorok]

FWIW here is how to run the docx validators in the CI (one using xmllint and the other using the open source officexml SDK from dotnet): https://github.com/jgm/pandoc/blob/main/.github/workflows/docx-validation.yaml

(Using the latter to validate is sufficient, but I find that the error messages from the former are easier to follow since it gives you line numbers in the XML instead of an xpath)

The latest pandoc master should now produce docx that passes validation (Quarto doesn't yet as discussed in the other linked issue)

[cscheid]

Thanks for the update! If Pandoc releases sufficiently soon we'll try to fix our docx issues as well ahead of the 1.4 release.