update languages.toml: tree-sitter-lua grammar

[briandorsey]

repo has moved, use new URL and the rev of the latest release (v0.0.19)

There is a more recent release of the tree-sitter-lua grammar (and new repo url, which appears to be the same repo, just moved):
tree-sitter-grammars/tree-sitter-lua@88e4464

I'm a first-time contributor, opening this PR per request from @archseer on #9726.

I tested locally by adding a .config/helix/languages.toml file with the following content (the source = line is the same content as this PR):

use-grammars = { only = [ "lua" ] }

[[grammar]]
name = "lua"
source = { git = "https://github.com/tree-sitter-grammars/tree-sitter-lua", rev = "88e446476a1e97a8724dff7a23e2d709855077f2" }

Then updated the grammar:

% hx --grammar fetch
Fetching 1 grammars
1 updated grammars
	lua now on 88e446476a1e97a8724dff7a23e2d709855077f2
% hx --grammar build
Building 1 grammars
1 grammars built now
	["lua"]

Then confirmed that the new grammar is being used. It is, yea!

fixes: #9726

[woojiq]

Very often, when you update a version of the Tree-sitter grammar, the node names change, causing the current helix queries to become inaccurate. Check the changelog and update our queries accordingly.
For example, I see that their highlight queries have changed several times in the last 2 years: https://github.com/tree-sitter-grammars/tree-sitter-lua/commits/main/queries/highlights.scm

[briandorsey]

Are there any tools or docs to help validate that changes to these queries correctly map to helix expectations? I assume it's not ok to just copy the query files over from the tree-sitter-lua repo, correct?

I spent about an hour trying to learn about this. I started by poking around at the files in the tree-sitter-lua repo vs. what is in the helix repo... they are very different, and I don't really know where to start changing them with any confidence that it wouldn't break something. Later, started looking at tree-sitter docs and trying to run tree-sitter query (some progress) or playground (blocked on wasm/Docker troubleshooting)... and then ran out of the time I have available to look at it today.

At the moment, this looks like it'll take a while to learn enough to update the queries with any confidence.

[the-mikedavis]

The queries that are in that repo are ok to copy but they'll need some small modifications. We use a custom set of captures https://docs.helix-editor.com/master/themes.html#syntax-highlighting, so @conditional for example should become @keyword.control.conditional, @number should become @constant.numeric, etc.

We will also need to re-order the patterns a little because neovim currently has the opposite precedence compared to us (though that will change with #9436). So this identifier pattern needs to move to basically the bottom of the file: after all other uses of (identifier). Also this function pattern should move above all of the other patterns that highlight functions.

In general you can use :tree-sitter-subtree on a selection to see what the tree looks like and then write queries based on that

[briandorsey]

Thanks for the guidance! I attempted to follow it. I'm not very confident that I got the ordering or captures correct.

The :tree-sitter-* commands were very helpful for visualizing/understanding what's happening with the mapping.

Aside: I was a bit confused how the :tree-sitter-* commands decided what to show. -subtree seemed to perhaps parse the whole file when a full line was selected (with x)

[the-mikedavis]

Thanks!

[the-mikedavis]

-subtree seemed to perhaps parse the whole file when a full line was selected (with x)

:tree-sitter-subtree tries to find the largest node that fully contains the current selection. Usually newlines are part of the root of the syntax tree so :tree-sitter-subtree on a full line gives the root. You can use trim_selections (_) before to trim the leading/trailing whitespace of the selection