"generate_parser_in_directory" generates spurious "grammar.js" and "package.json" files when "grammar_path" is specified but not equal to "<repo_path>/grammar.js".

[Cameron Gunnin]

Problem

Spurious files are created by generate::generate_parser_in_directory under the following conditions:

• grammar_path is Some, but not equal to "{repo_path}/grammar.js".
• repo_path is a directory

It looks like this was introduced around 2 months ago by commit 9e5bf65. The fix seems to be: also check whether grammar_path was specified before generating the grammar.js and package.js.

Also note, right below the added code, there's two paths for calling load_grammar_file:

• If grammar_path is specified, it uses that file.
• If grammar_path is not specified, it uses "{repo_path}/grammar.js".

This seems to suggest that the grammar path should be used if present, otherwise fall-back to "{repo_path}/grammar.js". So, one fix would be to first compute the grammar path to use at the top of the function, then checking it later. Something like:

    // ...

    let should_generate_grammar_json = grammar_path.is_none();
    let grammar_path = match grammar_path {
        Some(path) => PathBuf::from(path),
        None => repo_path.join("grammar.js"),
    }

    if repo_path.is_dir() && !grammar_path.exists() && !path_in_ignore(&repo_path) {
        // ...
    }

    let grammar_json = load_grammar_file(&grammar_path, js_runtime)?;

    // ...

Steps to reproduce

Create the following directory structure

• top/
  • input/
    • grammar.js
  • output/

Then invoke generate_parser_in_directory:

// CWD: top
tree_sitter_cli::generate::generate_parser_in_directory(
    Path::new("output"),
    Some("input/grammar.js"),
    14,
    false,
    None,
    None,
);

Expected behavior

If grammar_path is specified and exists, skeleton grammar.js and package.json are not generated under repo_path.

Tree-sitter version (tree-sitter --version)

tree-sitter-cli 0.22.2

Operating system/version

All Versions

[Max Brunsfeld]

Do you have repro steps using the Tree-sitter CLI, or is this only a problem if you use tree_sitter_cli as a library and call this function manually?

[Cameron Gunnin]

Do you have repro steps using the Tree-sitter CLI, or is this only a problem if you use tree_sitter_cli as a library and call this function manually?

Yes. Like before, create the following directory structure:

• top/
  • input/
    • grammar.js
  • output/

Then change the CWD to any directory other than input (e.g. cd top/output).

Then run tree-sitter generate path/to/grammar.js.

grammar.js and package.json will be generated in the current directory, alongside other artifacts.

[Max Brunsfeld]

Great thanks for the explanation.