tree-sitter: make tree-sitter-grammars an overridable scope

[khaneliman]

Turn pkgs.tree-sitter-grammars into the single, override-aware source of tree-sitter grammars, and route the existing consumers through it.

Alternative solution to (technically we could combine them and use the other as a more "backwards-compatible" drop in.
Closes #536475

Things done

• Built on platform:
  • x86_64-linux
  • aarch64-linux
  • aarch64-darwin
• Tested, as applicable:
  • NixOS tests in nixos/tests.
  • Package tests at passthru.tests.
  • Tests in lib/tests or pkgs/test for functions and "core" functionality.
• Ran nixpkgs-review on this PR. See nixpkgs-review usage.
• Tested basic functionality of all binary files, usually in ./result/bin/.
• Nixpkgs Release Notes
  • Package update: when the change is major or breaking.
• NixOS Release Notes
  • Module addition: when adding a new NixOS module.
  • Module update: when the change is significant.
• Fits CONTRIBUTING.md, pkgs/README.md, maintainers/README.md and other READMEs.
• Follows the automation/AI policy.

[RossSmyth]

I understand better on how it is structured, there was/is is scope, but it has no relationship to allGrammars and grammars so overrides don't actually do anything. Few observations and questions.

[RossSmyth]

Should probably just be called grammars, no need to make people wonder "what is a scope" if they have no need to be aware.

[RossSmyth]

The problem is that grammars and grammarScope are not actually related here. If I were making it, I would only expose grammar, which would be a scope.

Then you need to put a fix between allGrammars and grammars so that when grammars is overriden, allGrammars changes too. Currently it does not as you just make a copy of the unoveridden scope.

The above is similar to how python3Packges and typstPackages works, just without allGrammars.

Can alias grammarsScope to grammars so it's not a breaking change.

Suggested change

	passthru = {
	inherit
	buildGrammar
	builtGrammars
	withPlugins
	allGrammars
	;

[khaneliman]

I think there are two reasonable priorities here.

Using tree-sitter.grammars for the built scope matches the Python/Typst package-set shape and makes the public extension point tidy. The cost is that tree-sitter.grammars currently exposes source specs, so external consumers using that passthru could break.

Keeping tree-sitter.grammars as source specs and putting the built scope at something like tree-sitter.grammarPackages is a little less elegant, but it preserves the existing public meaning while still letting tree-sitter-grammars = recurseIntoAttrs tree-sitter.grammarPackages and making overrides compose.

Given the chance of external consumers, I lean toward the backwards-compatible shape unless we think the tree-sitter.grammars source-spec passthru is safe to repurpose.

[aciceri]

✅ nixpkgs-review result

Generated using nixpkgs-review-gha (da739d6)

Command: nixpkgs-review pr 537058
Commit: 0a53614dbbd241a3f99ddb0eadaa7655092a8bd9 (subsequent changes)
Merge: f3d0cb9d4c38915de2e7222da6861a9a2eefa6bb

Logs: https://github.com/aciceri/nixpkgs-review-gha/actions/runs/28519406230/attempts/1

---

x86_64-linux

✅ No rebuilds

---

aarch64-linux

✅ No rebuilds

---

x86_64-darwin

✅ No rebuilds

---

aarch64-darwin

✅ No rebuilds

[aciceri]

I like this, I wish it had been already here when I packages helix's grammars, happy to see that someone implemented it!
No derivations are rebuilt so we are sure that it wouldn't cause any harm to grammars consumers, in my opinion it's already good enough to be merged and further adjustments can be done in separate PRs.