Tweaked tree-sitter query classification

[mkatychev]

Updated Tree-sitter Query filename detection and heuristics such that all previously classified as Scheme languages are now properly identified as Tree-sitter Query in https://github.com/nvim-treesitter/nvim-treesitter:

Description

Checklist:

• I am fixing a misclassified language
  • I have included a new sample for the misclassified language:
    • Sample source(s):
      • https://github.com/nvim-treesitter/nvim-treesitter/tree/master/queries
    • Sample license(s):
      • Apache-2.0 license
  • I have included a change to the heuristics to distinguish my language from others using the same extension.

[mkatychev]

Moving TSQ above Scheme somehow increased classification hit-rate 🤷

[lildude]

You shouldn't need these as they all share the .scm extension which should be sufficient.

Suggested change

	filenames:
	- folds.scm
	- highlights.scm
	- indents.scm
	- injections.scm
	- locals.scm
	filenames:
	- folds.scm
	- highlights.scm
	- indents.scm
	- injections.scm
	- locals.scm

[mkatychev]

I tested it out, it seems those lines change classification, before deletion for nvim-tresitter:

$ ./bin/github-linguist $doc/nvim-treesitter | rg '(Tree-sitter|Scheme)'
69.59%  1128396    Tree-sitter Query

After deleting those lines:

$ ./bin/github-linguist $doc/nvim-treesitter | rg '(Tree-sitter|Scheme)'
69.52%  1127317    Tree-sitter Query
0.07%   1079       Scheme

[lildude]

That would then indicate the heuristics are missing a scenario specifically in your repo or the samples are not sufficiently teaching the classifier. Keep in mind we’re aiming for good enough, not perfection as every repo is different. It’s also why we have overrides.

The use of specific files is for cases where a specific filename is always used by that language, and generally doesn’t have an extension or has a generic or very common extension associated with another language. Using a filename will mean every instance of that filename will be classified as that language, even if the file is legitimately written in another language. So if a Scala developer creates a file called locals.scm, they’ll be forced to use an override because you decided all files with this name should be Tree-sitter query.

[mkatychev]

Thanks for the explanation, the changed turned out to be a rather small tweak.