code-search-efficient

Efficient code search minimizes token waste by applying a tool selection decision tree, using LSP for Java definitions and usages, ast-grep for structural matching, and ripgrep fallbacks with counts-first narrowing to analyze large repositories accurately.

AST-based structural matching with ast-grep identifies refactoring candidates accurately across languages by finding method and call patterns, and can apply batch transformations using the --rewrite flag to reduce manual edit workload.

Java code search prioritizes an LSP-first strategy to map goToDefinition, findReferences, and incomingCalls accurately, preventing incorrect grep-based exploration, while providing ripgrep and fd as fallback tools for simpler text queries.

Prevent context explosion by using token-saving workflows such as checking counts and file lists first, utilizing documentSymbol for partial reads of large files, and saving intermediate search results to /tmp to strictly control read scope.

Use ripgrep for simple text matching and fd for file name or extension filtering, while reserving ast-grep for AST-accurate structural matching where regular expressions would be too short, inaccurate, or produce false positives.