cartog-languages

Tree-sitter language extractors for the cartog code graph.

Overview

Parses source code using tree-sitter grammars and extracts symbols (functions, classes, methods, etc.) and edges (calls, imports, inherits, etc.). Each language has a dedicated extractor implementing the Extractor trait.

How it works

Extractor trait

pub trait Extractor: Send {
    fn extract(&mut self, source: &str, file_path: &str) -> Result<ExtractionResult>;
}

Takes &mut self so implementations can reuse their internal tree_sitter::Parser across files, avoiding per-file allocation overhead.

Tree-sitter S-expression queries

Extractors use declarative S-expression queries (not cursor walking) to match AST patterns. Queries are compiled once in the extractor's new() constructor and reused on every extract() call via the CachedQuery helper.

Example (Python call extraction):

(call function: [(identifier) (attribute)] @callee)

Named captures (@callee, @exception_type, etc.) identify the matched nodes for symbol/edge construction.

Nested scope filtering

is_inside_nested_scope() walks up the AST from a node to a given root node, checking if any ancestor in between matches a set of scope kinds (e.g., function_definition, class_definition). This prevents extracting edges from nested function bodies as if they belong to the outer scope.

Supported languages

Code: Python, TypeScript, TSX, JavaScript, Rust, Go, Ruby, Java, PHP, Dart.

Documents: Markdown (.md) — chunked by heading for semantic search. Each heading section becomes a Document symbol. Large sections are sub-chunked at paragraph boundaries (~1500 bytes). Files without headings use fixed-size paragraph chunking.

A crate-internal js_shared module holds extraction logic shared between the JavaScript and TypeScript/TSX extractors (not part of the public API).

Public API

Crate dependencies

cartog-core