Grammars

jint uses tree-sitter to parse your code and embeds ten grammars in the binary.

A grammar is a directory named for its id. It holds three files:

<id>.wasm parses the language. Its tree_sitter_<id> export must match the directory name.
extensions.json routes files to the grammar. It is a bare JSON array, such as [".md"].
queries.scm holds the extraction patterns and the shape directives.

The walker knows no node types. The .scm assigns captures, and the rest of jint sees only captures.

Your own grammars live in $JINT_GRAMMAR_DIR, default ~/.jint/grammars. Each subdirectory is one grammar. See Bring your own grammar.

Embedded grammars

grammar	extensions
`typescript`	`.ts`, `.mts`, `.cts`
`tsx`	`.tsx`
`python`	`.py`
`go`	`.go`
`c`	`.c`, `.h`
`cpp`	`.cpp`, `.cc`, `.cxx`, `.hpp`, `.h`
`css`	`.css`
`markdown`	`.md`
`php`	`.php`
`bash`	`.sh`, `.bash`

Both c and cpp claim .h. Registration order decides, and cpp serves it.

css captures comments only, because the code captures have no honest reading of css.

Walk

The walk turns one parsed file into elements. Each element carries a capture, a span, a positional label (d0, c2, i0), an optional name, its parent label and its source.

Parsing is demand-driven. jint parses a file only when a rule targets elements in it. A files rule reads the raw source and skips the parse.
Unparsed files are a note. Without a grammar, a file is measured as raw source by a files rule. Otherwise the run notes it once, for example not parsed (no grammar): .json, .sql.
A named file with no grammar is an error. An explicit path, an element-target rule and no grammar give a MISSING_GRAMMAR setup error that names the supported grammars.