Inbuilt Related configuration ignoring cascaded taxonomies

support
1

I am reorganising some of my blogs that had nested sections one level deep. I decided to make them into taxonomies by removing the _index.md since I still want to retain the structure even if the parent section becomes the branch bundle for them. I then cascaded the taxonomies to match these folders in the configuration e.g

[[cascade]]
  [cascade._target]
  path = "/posts/cats/**"
  [cascade.params]
  categories = ["cats"]

But Hugo’s Configure related content is not picking these categories.

2

I’m not sure I completely understand your content structure, but here’s an example that I think is close:

git clone --single-branch -b hugo-forum-topic-56148 https://github.com/jmooring/hugo-testing hugo-forum-topic-56148
cd hugo-forum-topic-56148
hugo server
3

Thanks Joe! I just found the issue is assigning categories weight 1 and tags weight 2. When reversed, the related pages show up for those pages without tags.

[[cascade]]
  [cascade._target]
  path = "/posts/cats/**"
  [cascade.params]
  categories = ["cats"]

[related]
  includeNewer = true
  threshold = 80
  toLower = true
  [[related.indices]]
    name = 'categories'
    weight = 1
  [[related.indices]]
    name = 'tags'
    weight = 2

Which I think means the search falls back to categories? (I can reproduce the same issue with your test site. Just add tags to one page, paste the code above in hugo.toml then visit that page.)

Edit: In another project with tags and categories defined in front matter, now the opposite is happening. Giving tags priority makes categories return null. Strange! (in my case, I am sorting by date)

Edit 2: Claude.ai recommended to use larger numbers for easier ratio tuning. 80 and 100 work even when reversed.

4

There’s a somewhat opaque relationship between weight and threshold:

I have never taken the time to understand the calculations; I always set the weights in the double digits and futz with them until I get the desired result.

I suspect with very low weights you’d need to use a very low threshold.

5

Interesting…maybe a note in the docs on this just in case someone else comes across it? I thought it was a bug!