Backlinks from Jekyll Liquid loops don't appear to work?

[steinea]

Another thread for a different issue I'm having, also with the site I'm building at https://github.com/steinea/garden. As noted previously, I'm trying to get this theme to work with an existing site that has quite a lot of content. I heavily rely on Jekyll's ability to run through a set of posts or pages and generate lists or tables based on categories or tags in the front matter.

My site has a slightly different structure than the default site for this template, with two different collections of file types. Maxime's bidirectional_links_generator.rb was clear and easy to parse, and I was able to tweak it to work for my two collections instead of the one _notes collection. With the tweaked plugin, links written directly in post or page content generate backlinks appropriately, and the notes graph likewise displays these links. See the modified plugin here: https://github.com/steinea/garden/blob/main/_plugins/bidirectional_links_generator.rb.

However, I've noticed that links generated by a {% for %} loop do not generate backlinks, nor do the links appear in the notes graph. For instance, see https://github.com/steinea/garden/blob/main/collections/_pages/activity.html:

<p><a href="/playing/" class="internal-link">Playing</a></p>

This hard-coded link generates a backlink and notes graph connection. But on the same page, this code...

<ul>
{% for page in site.pages %}
  {% if page.category contains "activity" %}
  <li><a href="{{ page.url }}" class="internal-link">{{ page.title }}</a></li>
  {% endif %}
{% endfor %}
</ul>

...which generates the same link, does not generate a backlink nor a notes graph connection. If I inspect the page source, both links are being rendered by Jekyll identically in html. So I think it must be the order of operations between the plugin and the for loop that is causing the problem? If the plugin runs first in the build sequence, it will only find the hard-coded link. Then if the for loop runs second, the link will be generated after the plugin has run. So if there is no hard-coded link, then the plugin will find no link at all, even if that link is generated after the fact by the for loop. This is my suspicion as to what's the issue is, but unfortunately I don't know how to resolve it if true. Help? Ty!

[steinea]

@maximevaillancourt I've been reading the Jekyll docs, and the original issue I seem to have resolved using a hook. My modified version of the plugin looks like this, and so far it is working for my _pages collection, generating backlinks and paths in the notes graph:

# MODIFIED PLUGIN FOR MY SITE CONFIG

# frozen_string_literal: true
class BidirectionalLinksGenerator < Jekyll::Generator
  def generate(site)
    Jekyll::Hooks.register :documents, :post_convert do |docs|
      graph_nodes = []
      graph_edges = []

      all_posts = site.collections['posts'].docs
      all_pages = site.collections['pages'].docs

      all_docs = all_posts + all_pages

      link_extension = !!site.config["use_html_extension"] ? '.html' : ''

      # REMOVED WIKILINK LOGIC

      # Identify note backlinks and add them to each note
      all_docs.each do |current_note|
        # Nodes: Jekyll
        notes_linking_to_current_note = all_docs.filter do |e|
          e.url != current_note.url && e.content.include?(current_note.url)
        end

        # Nodes: Graph
        graph_nodes << {
          id: note_id_from_note(current_note),
          path: "#{site.baseurl}#{current_note.url}#{link_extension}",
          label: current_note.data['title'],
        } unless current_note.path.include?('_notes/index.html')

        # Edges: Jekyll
        current_note.data['backlinks'] = notes_linking_to_current_note

        # Edges: Graph
        notes_linking_to_current_note.each do |n|
          graph_edges << {
            source: note_id_from_note(n),
            target: note_id_from_note(current_note),
          }
        end
      end

      File.write('_includes/notes_graph.json', JSON.dump({
        edges: graph_edges,
        nodes: graph_nodes,
      }))
    end

    def note_id_from_note(note)
      note.data['title'].bytes.join
    end
  end
end

By hooking the plugin to the :post_convert event, I'm able to have the backlink generator run after the renderer has interpreted all the Liquid on the site, but before the render has actually happened. The plugin is then able to pick up links interpreted from Liquid, add backlinks, and then add the links to the notes graph.

HOWEVER! This does not work for my posts directory, for some inexplicable reason. If I try to add a post in _posts, the notes graph breaks, with all paths between nodes disappearing, and no backlinks are generated for the post in question. It seems to only occur for a file with a date in the filename (e.g., 2025-01-01-test-post.md) but I can for the life of me figure out why this would be the case. I don't have a good enough handle on the innerworkings of this plugin to diagnose what is going on.

[steinea]

Additional context, in my config, I define my collections like so:

# Collections
collections_dir: collections
collections:
  posts:
    permalink: /:year/:month/:day/:title/
  pages:
    output: true
    permalink: /:title/

[steinea]

Update here in case anyone reads this thread: still haven't solved this problem, but I'm trying a new solution. No longer using the collections above, since I was worried about collision between site.collections['posts'].docs and site.collections['pages'].docs with site.posts and site.pages. I removed the collections, and now I'm using _config.yml defaults:

# Defaults
defaults:
  - scope:
      path: ''
      type: pages
    values:
      permalink: /:title/
  - scope:
      path: ''
      type: posts
    values:
      permalink: /:year/:month/:day/:title/

I've also cut out a bunch of extra logic from the plugin to help with troubleshooting. I'm using an edited version of TerminalAddict's refactor from the Jekyll forum. Mine looks like this:

class BidirectionalLinksGenerator < Jekyll::Generator
  def generate(site)

      Jekyll::Hooks.register :pages, :post_convert do |page|
        all_posts = site.posts.docs
        all_pages = site.pages

        all_docs = all_posts + all_pages

        all_docs.each do |current_note|
          notes_linking_to_current_note = all_docs.filter do |e|
            e.content.include?(current_note.url)
          end
          current_note.data['backlinks'] = notes_linking_to_current_note
        end
      end

    end

  end

I describe my approach and issues in more detail here: https://talk.jekyllrb.com/t/trouble-with-hooks-for-a-custom-backlink-generator/9945/8. Long story short, This works for pages only (can't figure out how to do the :post_convert event for both pages and posts), and for pages, it only sporadically produces backlinks for links generated by a Liquid for loop iterating through site.pages. Very strange. Cannot figure out a pattern for which backlinks get generated for which .md files and which don't.