Possible to pre-process md for recent updates module?

[isuftin]

Our issue is that we use GitLab and publish to GitLab pages. Before we do, we run a custom python hook

def on_page_markdown(markdown, **kwargs):
    return markdown.replace("[[_TOC_]]", "")

This removes that string from the finalized docs. However, it looks like the recent updates module runs, it does so with the "[[TOC]]" content still in there so our recent updates look like..

...which also seems to want to include the raw markdown for links.

I assume there's no way to have this module run on the resultant HTML instead of the raw input markdown?

[jaywhj]

You can make the replacement directly within the on_page_markdown function.
The data fetching for the recent update module is performed in the on_env function, which runs after on_page_markdown.

    def on_page_markdown(self, markdown, page: Page, config, files):
        markdown = markdown.replace("[[_TOC_]]", "")

[isuftin]

Hi @jaywhj following up on the [[ TOC ]] question: moving the replacement into an on_files hook (setting the file content there) fixed it, thanks.

One related issue with the recently-updated cards: summaries can contain raw Markdown, and in one case the card HTML actually breaks.

Minimal repro (document-dates 3.8.7, mkdocs-materialx 10.2.0). A page starting with:

# Title

Reference style [ref link][r1] here.

[r1]: https://example.com/ref

Autolink <https://example.com/auto> here.

produces this card summary (in config.extra.recently_updated_docs):

Reference style [ref link][r1] here.  Autolink <https://example.com/auto> here.

Two problems:

1. Reference-style links survive verbatim. analyze_markdown's LINK_RE only matches the inline text form, so [text][ref] is never flattened.
2. Angle-bracket autolinks corrupt the card HTML. The Jinja environment is created without autoescape (jinja2.Environment(loader=loader, auto_reload=False) in the theme), so the {{ summary }} in the card
   template passes < through. The served HTML ends up like this:

<div class="card-summary">Autolink <https: example.com="" auto=""> here.  Rest of summary...</https:></div>

The browser parses https://... as an unknown element and renders everything after it inside that element, so the tail of the summary gets swallowed/mangled.

Suggested fixes (small, I think):

• Bug fix: HTML-escape the summary in analyze_markdown (escape & first, then < and >), which stops raw <…> from becoming markup.
• Optional feature: instead of reducing text to plain text, emit text — since the environment has no autoescape, it would render as-is. Happy to take that as a separate request if you'd
  rather not mix it with the fix.

Happy to send a PR for the escaping fix if that's the direction you'd like. Otherwise we'll work around it on our side in the meantime.

[jaywhj]

1. Yes, performing the replacement in the on_files hook is also acceptable, since both hooks run before the on_env hook.
2. For card summary extraction, we currently only handle block-level tags that significantly impact readability. Other edge cases indeed haven’t been fully accounted for — for example, reference-style links like [ref link][r1].
3. PRs are very welcome for fixing the escaping issues!

[jaywhj]

@isuftin
I've applied a fix like this. Not sure if it resolves the escaping issue you mentioned above. You can test it out in recently_updated_group.html (line 283) — modify it as follows:

<div class="card-summary">{{ summary | e }}</div>