Resolve ambiguities based on closer common ancestor

[srid]

• Prototype
• Do the reverse:
  • "backlinks" should not show all the places linking to [[foo]] (if that foo was an ambiguity resolved)
  • what about uptree? folgezettel parents, etc.
• refactor

[edrex]

I was just looking into how it might be possible to add resolution (nearest ancestor), since I have same-name notes in different directories all over the place, making the current behavior a blocker for using emanote for my big repo. So great to see a patch in progress Srid!

I would prefer that the behavior be to omit descendant search entirely, and search only ancestors (that's what Ikiwiki does), but I don't need that to get unblocked.

I just did a test build and it's still showing ambiguous link errors for me:

(out of scope here, the rendering for these inline errors has always seemed broken for me. Missing style or something?)

LMK if you want review when you're ready.

[edrex]

Previous discussion, starting at #115 (reply in thread)

[srid]

I would prefer that the behavior be to omit descendant search entirely, and search only ancestors (that's what Ikiwiki does), but I don't need that to get unblocked.

Without descendant search, it would not be possible to "combine" multiple sites though (see #494), and thus it would break https://community.flake.parts/ (the various docs come from different repos; and guide.md for example exists in both haskell-flake and nixos-flake repos).

[edrex]

I was too ambigious with "... omit descendant search entirely". I meant a ref Foo shouldn't resolve to /A/Foo.md, but (A/Foo) still would. So a ref R from path A would resolve to B if and only if B = Apre + Ref for some Apre in Prefixes(A).

This would require all descending path elements to be explicit in refs. This would allow matching on subdirs for flexible reference remapping (eg, a local site registry mapping ref prefixes to URL prefixes for external sites).

I'm only guessing at the use-case you're working on. Am I missing part of the requirement?

[srid]

@edrex What is the value of Prefixes(A) if A = /Foo/Bar/Qux.md?

If Prefixes == ["/", "/Foo", "/Foo/Bar"], then by the rule you stated above, Qux.md can link to any note in the notebook.

Could you provide a more rigorous logic so I can understand what you are proposing?

[srid]

I'm gonna merge this now; but will always be happy to revisit the ambiguity resolution algorithm based on convincing rationale.