feat(lang): Update Nix grammar & improve queries

[nrdxp]

Update grammar and queiries to the latest upstream. Also add improved indents thanks to @oxalica.

[the-mikedavis]

Helix uses custom scopes (see here), so some of the scopes here need to be renamed. For example: @property becomes @variable.other.member. The changes for booleans and numbers can be reverted except that it looks like integer and float nodes have been renamed to integer_expression and float_expression. It looks like most of the changes in the grammar came from renaming nodes to have a _expression suffix.

[nrdxp]

@the-mikedavis, I believe I addressed all your comments with a force push.

[the-mikedavis]

We should probably keep these unless the injects work better without them

[nrdxp]

Not sure exactly why tbh, but I was getting extra indents with the previous definitions. These ones seem to work more accurately

[nrdxp]

@the-mikedavis I think I was able to resolve everything. I just recently stumbled on this comment though: nix-community/tree-sitter-nix#13 (comment) and now I'm wondering if I should rebase on oxalica's fork to add the injections. Although what I'd really like to do is get those changes incooperated into upstream properly.

I think I'll give this a shot later. Maybe we should hold off on merging for a bit just in case I'm successful?

[the-mikedavis]

I think this is a good increment even without shell injections - that can always be a follow-up PR. Would you mind rebasing?

[nrdxp]

I went ahead and force pushed a rebase, and added this back:

diff --git a/runtime/queries/nix/highlights.scm b/runtime/queries/nix/highlights.scm
index d15fb9c2..a171a39b 100644
--- a/runtime/queries/nix/highlights.scm
+++ b/runtime/queries/nix/highlights.scm
@@ -72,9 +72,15 @@
 (binary_expression
   operator: _ @operator)
 
+(variable_expression (identifier) @variable)
+
 (binding
   attrpath: (attrpath (identifier)) @variable.other.member)
 
+(identifier) @variable.other.member
+
+(inherit_from attrs: (inherited_attrs attr: (identifier) @variable.other.member) )
+
 [
   ";"
   "."

Without it, I experienced a regression in path type highlighting.

[the-mikedavis]

The

(variable_expression (identifier) @variable)

stanza is a good catch: it highlights variables in inherit-from statements:

{
  inherit (common) pkgs;
  #         ^ variable
  #                 ^ variable.other.member
}

The

(identifier) @variable.other.member

stanza takes priority over the

(identifier) @variable

on the last line, so let's remove the (identifier) @variable one since it's now redundant - the highlights look much better with the @variable.other.member capture.

The

(inherit_from attrs: (inherited_attrs attr: (identifier) @variable.other.member))

doesn't appear to have an effect because of the (identifier) @variable.other.member above it, I think we're safe to remove it. Based on the old highlights, it looks like the recent refactor upstream combined the old attr_identifier and identifier nodes which is what I think led to the redundant stanzas in the upstream queries like the above one. I'll make a PR to remove the redundant ones upstream.

By path highlights did you mean like ./shell.nix in import ./shell.nix or in attribute paths? Those stanzas shouldn't interact with the file-system path highlights so I think you may have seen #2484 if that's what you're talking about.

[the-mikedavis]

Looks great, thanks for working on this!

[nrdxp]

Thanks for the feedback. Also, just to confirm, it is #2484 that is the problem. After opening the same file several times I noticed sometimes paths are highlighted correcly, and sometimes they are just hightlighted as regular strings.