Consider page dependencies?

[maul-esel]

I have several pages in my site, whose content is generated based on YAML data in _data, or which display the items of a collection. Currently, only changes to the page source itself are reflected by page.last_modified_at. I would like page.last_modified_at to reflect when the page content changes due to changes in the underlying data / collection (such that the jekyll-sitemap plugin generates correct <lastmod> tags).

Is this feature currently supported, or planned?

If it's not yet supported, I would be willing to take a look into this issue myself, if you could give me some guidance.

• A relatively simple solution might be to have users explicitly declare dependencies in the frontmatter. I'd be happy with this solution, but others may find it tedious. On the other hand, it gives you complete control over which dependencies you want to be considered, and which should be ignored.
• Alternatively, jekyll seems to do some dependency tracking itself: https://github.com/jekyll/jekyll/blob/6855200ebda6c0e33f487da69e4e02ec3d8286b7/lib/jekyll/regenerator.rb#L106-L114 I don't know how difficult it would be to access this information from a plugin.

[maul-esel]

I have a minimal draft of the first variant (explicitly specifying dependencies): https://github.com/gjtorikian/jekyll-last-modified-at/compare/master...maul-esel:jekyll-last-modified-at:specify-dependencies

Usage:

---
lastmod_dependencies:
  collections: ['people'] # names of collections
  data: ['bla'] # names of files in `_data`, not including the .yml extension
---

The second variant (using jekyll's dependency tracking) seems a bit farther off, after some investigation:

• Dependency tracking is only enabled for incremental builds (that is the only application so far)
• Dependencies on data files are currently not tracked:
  • jekyll/jekyll#7682
  • jekyll/jekyll#8771 implemented the feature, but it was reverted in jekyll/jekyll#9170
• Based on some experimentation, it seems dependencies on documents in collections are also currently untracked.
• Fundamentally, if dependencies are only discovered during rendering, the last_modified_at time of the current page can not easily be determined (though it'd probably be fine for my use case, i.e., <lastmod> in the sitemap)