search-data.json does not resolve included content

[PhilLab]

Describe the bug
If one of your content files includes other content via {% include %} the search data JSON file does not reflect this.

To Reproduce
Steps to reproduce the behavior:

1. Have a content file like this:

---
title: My title
---
{% include IncludeMe.md %}

2. Inspect the generated search-data.json at /_size/assets/js/search_data.json
3. The file contains:

{"0": {
    "doc": "My title",
    "title": "My title",
    "content": "{% include IncludeMe.md %} ",
    "url": "/10.html",
    "relUrl": "/10.html"
  }
...

4. As a result, the search does not return the file when searching for the included content

Expected behavior
The content included from IncludeMe.md should turn up in the search-data.json

[pdmosses]

It appears that Jekyll renders pages in alphabetical order, see also this clarification. @SgtSilvio wrote:

Jekyll evaluates files in alphabetical order.
So before when it was called search-data.json pages that started with a letter t or alphabetically greater were not rendered > to html yet, so their plain md was included in the search-data.json

That is why the generating file has the prefix zzzz-. I know very little about Jekyll or Ruby, but it's conceivable that more than one thread is used for rendering pages. Then the rendering of zzzz-search-data.json may start before that of other pages has finished. It's unclear to me whether there is an alternative (thread-safe) way of generating the search data using Jekyll.

On a moderately large site that makes use of included md files, the reported bug affects my generated search data when the site is built locally with jekyll@4.2, but not when using jekyll@3.8.7. By default, GH Pages still uses jekyll@3.8 🙄, so the bug might not appear on most published sites. However, I'm using GH Actions and jekyll@4.2 to build on GH Pages, and indeed, the bug shows up on the published site.

[pdmosses]

This was due to a Jekyll regression, not a bug in Just the Docs.

See my Jekyll issue comment for further details.

[pdmosses]

The Jekyll v4.2.0 regression that caused this issue has been fixed.

If it reappears, please reopen, mentioning the Jekyll version used.