python: arguments in Callable not highlighted as types

[MithicSpirit]

Describe the highlighting problem

In python, the arguments in a Callable are highlighted as variables/constants rather than as types.

Example snippet that causes the problem

f: Callable[[t], u]

Tree-sitter parsing result

(expression_statement) ; [1:1 - 19]
 (assignment) ; [1:1 - 19]
  left: (identifier) ; [1:1 - 1]
  type: (type) ; [1:4 - 19]
   (generic_type) ; [1:4 - 19]
    (identifier) ; [1:4 - 11]
    (type_parameter) ; [1:12 - 19]
     (type) ; [1:13 - 15]
      (list) ; [1:13 - 15]
       (identifier) ; [1:14 - 14]
     (type) ; [1:18 - 18]
      (identifier) ; [1:18 - 18]

Example screenshot

Using the catppuccin colorscheme. :Inspect was called with the cursor on the t.

Expected behavior

The t should be a @type, but it is a @variable instead. If it is changed to T, then it is both a @variable and a @constant. Note that the u is correctly identified as an @type. I suspect that the issue is the type node does not "propagate into" the list node for syntax highlighting.

Output of :checkhealth nvim-treesitter

==============================================================================
nvim-treesitter: require("nvim-treesitter.health").check()

Installation ~
- OK `tree-sitter` found 0.21.0 (1c55abb5308fe3891da545662e5df7ba28ade275) (parser generator, only needed for :TSInstallFromGrammar)
- OK `node` found v18.18.2 (only needed for :TSInstallFromGrammar)
- OK `git` executable found.
- OK `cc` executable found. Selected from { vim.NIL, "cc", "gcc", "clang", "cl", "zig" }
  Version: cc (GCC) 13.2.1 20230801
- OK Neovim was compiled with tree-sitter runtime ABI version 14 (required >=13). Parsers must be compatible with runtime ABI.

OS Info:
{
  machine = "x86_64",
  release = "6.7.6-2-cachyos-lto",
  sysname = "Linux",
  version = "#1 SMP PREEMPT_DYNAMIC Sat, 24 Feb 2024 08:43:40 +0000"
} ~

Parser/Features         H L F I J
  - agda                ✓ . ✓ . .
  - awk                 ✓ . . . ✓
  - bash                ✓ ✓ ✓ . ✓
  - c                   ✓ ✓ ✓ ✓ ✓
  - commonlisp          ✓ ✓ ✓ . .
  - cpp                 ✓ ✓ ✓ ✓ ✓
  - css                 ✓ . ✓ ✓ ✓
  - diff                ✓ . . . .
  - dot                 ✓ . . . ✓
  - fennel              ✓ ✓ ✓ . ✓
  - git_config          ✓ . ✓ . ✓
  - git_rebase          ✓ . . . ✓
  - gitcommit           ✓ . . . ✓
  - gitignore           ✓ . . . .
  - go                  ✓ ✓ ✓ ✓ ✓
  - groovy              ✓ ✓ ✓ ✓ ✓
  - haskell             ✓ . ✓ . ✓
  - html                ✓ ✓ ✓ ✓ ✓
  - ini                 ✓ . ✓ . .
  - java                ✓ ✓ ✓ ✓ ✓
  - javascript          ✓ ✓ ✓ ✓ ✓
  - json                ✓ ✓ ✓ ✓ .
  - jsonc               ✓ ✓ ✓ ✓ ✓
  - latex               ✓ . ✓ . ✓
  - lua                 ✓ ✓ ✓ ✓ ✓
  - make                ✓ . ✓ . ✓
  - markdown            ✓ . ✓ ✓ ✓
  - markdown_inline     ✓ . . . ✓
  - meson               ✓ . ✓ ✓ ✓
  - org                 . . . . .
  - passwd              ✓ . . . .
  - perl                ✓ . ✓ . ✓
  - properties          ✓ ✓ . . ✓
  - proto               ✓ . ✓ . .
  - python              ✓ ✓ ✓ ✓ ✓
  - query               ✓ ✓ ✓ ✓ ✓
  - rust                ✓ ✓ ✓ ✓ ✓
  - scss                ✓ . ✓ ✓ .
  - strace              ✓ . . . ✓
  - toml                ✓ ✓ ✓ ✓ ✓
  - vim                 ✓ ✓ ✓ . ✓
  - vimdoc              ✓ . . . ✓
  - yaml                ✓ ✓ ✓ ✓ ✓
  - zig                 ✓ ✓ ✓ ✓ ✓

  Legend: H[ighlight], L[ocals], F[olds], I[ndents], In[j]ections
         +) multiple parsers found, only one will be used
         x) errors found in the query, try to run :TSUpdate {lang} ~

Output of nvim --version

NVIM v0.9.5
Build type: Release
LuaJIT 2.1.1702233742

   system vimrc file: "$VIM/sysinit.vim"
  fall-back for $VIM: "/usr/share/nvim"

Run :checkhealth for more info

Additional context

Also reported (erroneously) at tree-sitter/tree-sitter-python#258.

[lucario387]

Don't think this can be handled gracefully without bloating up the parser/slowing down parsing (#has-ancestor? shenanigan)

[MithicSpirit]

Don't think this can be handled gracefully without bloating up the parser/slowing down parsing (#has-ancestor? shenanigan)

Been thinking about this. I'm not too familiar with treesitter queries, but maybe it could only look up a few nodes rather than all the way up the tree? If it only has to look back one or two nodes, I don't think that it'll have as significant a performance impact. No clue whether there's a query for this, though.

[clason]

There is one for direct parents, but that's it.