build(deps): bump tree-sitter and treesitter-c to 0.20.1

[clason]

This bumps tree-sitter from a fixed pre-release commit to v0.20.1, and similarly for treesitter-c (the bundled C parser).

Two tests had to be updated to the latest query API.

[clason]

Two test failures:

• test/functional/treesitter/parser_spec.lua @ 230: treesitter parser API can match special regex characters like \ * + ( with vim-match?
• test/functional/treesitter/parser_spec.lua @ 320: treesitter parser API allow loading query with escaped quotes and capture them with lua-match? and vim-match?

[theHamsta]

The tests fail because the pattern index changed (treesitter/parser_spec.lua)

      for pattern, match in cquery:iter_matches(tree:root(), 0) do
        -- can't transmit node over RPC. just check the name and range
        local mrepr = {}
        for cid,node in pairs(match) do
          table.insert(mrepr, {cquery.captures[cid], node:type(), node:range()})
        end
        table.insert(res, {pattern, mrepr})

Probably the pattern index should be removed in the last line of both tests. Sorry, for adding it in the first place!

[clason]

Thanks for looking at this!

Probably the pattern index should be removed in the last line of both tests. Sorry, for adding it in the first place!

I don't know anything about these tests, so a very explicit change suggestion would be helpful. Do you mean just removing the line table.insert(res, {pattern, mrepr}), or something else?

[theHamsta]

Sorry, there seems to be a real problem here

Passed in:
  (table: 0x7fb6c0b62aa8) { }

Thought at first sight that only the indices were switched. Can look into this tonight.

[clason]

Yep, looks like some nodes are just missing (the second failing test makes that clearer): quote, paren, plus, times

But string and escape are there and correct.

[theHamsta]

Pretty strange. I did a clean build and the tests are failing the same way as on CI. But tree-sitter itself seems to be working normally and also the exact queries from the test still seem to be working.

Apparently, we're loosing all anonymous in the tests (does (_) really capture anonymous nodes or do we have to use something else @stsewd you know more about wild cards, right?). Related?: tree-sitter/tree-sitter#1496

[theHamsta]

This continues to work

❯ tree-sitter query foo.scm bar.c
Warning: You have not configured any parser directories!
Please run `tree-sitter init-config` and edit the resulting
configuration file to indicate where we should look for
language grammars.

bar.c
  pattern: 0
    capture: quote, start: (0, 9), text: "\""
  pattern: 0
    capture: quote, start: (0, 15), text: "\""

tree-sitter-c on  master [?] is 📦 v0.20.1 via  v17.0.1 via 🦀 v1.58.0-nightly 
❯ cat foo.scm
((_) @quote (match? @quote "^\"$"))

tree-sitter-c on  master [?] is 📦 v0.20.1 via  v17.0.1 via 🦀 v1.58.0-nightly 
❯ cat bar.c
#include "bar.h"

[stsewd]

Apparently, we're loosing all anonymous in the tests (does (_) really capture anonymous nodes or do we have to use something else @stsewd you know more about wild cards, right?). Related?: tree-sitter/tree-sitter#1496

(_) matches all named nodes, and _ matches named and anonymous nodes.

[theHamsta]

But using _ will result in a syntax error. I tried that before. Maybe one would have to put the #match statements in a different place then. Can you fix the tests?

[stsewd]

So, these queries

neovim/test/functional/treesitter/parser_spec.lua

Line 234 in fd6df74

cquery = vim.treesitter.parse_query("c", '((_) @plus (vim-match? @plus "^\\\\+$"))'..

should be written as

([_] @plus (#vim-match? @plus "^\\\\+$"))

That way we are able to match any anonymous nodes, otherwise (_ @plus) is invalid or (_) will only match named nodes. Also, we should use the new syntax for predicates (prefixing with #), the other syntax is deprecated.

[clason]

Looks like changing all (_) to [_] for these tests did the trick, thanks @stsewd !

(I can still change to the new syntax if somebody tells me exactly what to write ;))

[stsewd]

(I can still change to the new syntax if somebody tells me exactly what to write ;))

the other syntax update would be to change vim-match? and friends to #vim-match? (pre-fixing #), isn't a problem now, but that syntax could be removed in the future.

[clason]

Like so?

[stsewd]

@clason yes, that's it!

[clason]

@stsewd Will this bump have ramifications for nvim-treesitter? Will those queries need to be updated, too?

[clason]

All checks pass -- even Windows! Now we have to merge ;)

[stsewd]

@stsewd Will this bump have ramifications for nvim-treesitter? Will those queries need to be updated, too?

AFAIK we don't have queries like those, but if we did, it should be easy to fix. (I run the nightly version of tree-sitter locally, and haven't seen any errors)