Treesitter Patterns

Tree-sitter Patterns Skill

Universal patterns for working with tree-sitter in any project. Covers AST parsing, query patterns, visitors, and language plugin development.

Design Principle

This skill is framework-generic. It provides universal tree-sitter patterns:

• NOT tailored to Code-Index-MCP, treesitter-chunker, or any specific project
• Covers common patterns applicable across all tree-sitter projects
• Project-specific queries go in project-specific skills

Variables

Instructions

MANDATORY - Follow the Workflow steps below in order.

1. Identify languages to parse
2. Install appropriate language parsers
3. Write queries for extraction needs
4. Handle edge cases and errors

Red Flags - STOP and Reconsider

If you're about to:

• Parse without error handling (syntax errors are common)
• Assume all files parse successfully
• Write queries without testing on sample code
• Ignore performance for large files

STOP -> Add error handling -> Test on edge cases -> Then proceed

Cookbook

Language Plugin Development

• IF: Creating a new language parser
• THEN: Read and execute ./cookbook/language-plugin.md

AST Visitor Patterns

• IF: Walking the AST to extract information
• THEN: Read and execute ./cookbook/ast-visitor.md

Query Patterns

• IF: Writing tree-sitter queries
• THEN: Read and execute ./cookbook/query-patterns.md

Quick Reference

Python Setup

import tree_sitter_python as tspython
from tree_sitter import Language, Parser

# Create parser
parser = Parser(Language(tspython.language()))

# Parse code
source = b"def hello(): pass"
tree = parser.parse(source)

# Access root node
root = tree.root_node
print(root.sexp())

Node Navigation

# Get children
for child in node.children:
    print(child.type, child.text)

# Named children only (skip punctuation)
for child in node.named_children:
    print(child.type)

# Find by type
def find_all(node, type_name):
    results = []
    if node.type == type_name:
        results.append(node)
    for child in node.children:
        results.extend(find_all(child, type_name))
    return results

functions = find_all(root, "function_definition")

Query Language

; Match function definitions
(function_definition
  name: (identifier) @function.name
  parameters: (parameters) @function.params
  body: (block) @function.body)

; Match class definitions
(class_definition
  name: (identifier) @class.name
  body: (block) @class.body)

; Match imports
(import_statement
  (dotted_name) @import.module)

; Match decorated functions
(decorated_definition
  (decorator) @decorator
  definition: (function_definition
    name: (identifier) @function.name))

Running Queries

from tree_sitter import Query

query = Query(Language(tspython.language()), """
(function_definition
  name: (identifier) @name
  body: (block) @body)
""")

captures = query.captures(root)
for node, name in captures:
    print(f"{name}: {node.text.decode()}")

Common Node Types

Error Handling

def safe_parse(source: bytes) -> tuple[Tree | None, list[str]]:
    """Parse with error collection."""
    tree = parser.parse(source)
    errors = []

    def collect_errors(node):
        if node.type == "ERROR" or node.is_missing:
            errors.append(f"Error at {node.start_point}: {node.text[:50]}")
        for child in node.children:
            collect_errors(child)

    collect_errors(tree.root_node)
    return tree, errors

tree, errors = safe_parse(source)
if errors:
    print(f"Parse errors: {errors}")

Visitor Pattern

from abc import ABC, abstractmethod

class ASTVisitor(ABC):
    """Base visitor for tree-sitter AST."""

    def visit(self, node):
        method_name = f"visit_{node.type}"
        visitor = getattr(self, method_name, self.generic_visit)
        return visitor(node)

    def generic_visit(self, node):
        for child in node.named_children:
            self.visit(child)

    @abstractmethod
    def visit_function_definition(self, node):
        pass

class FunctionExtractor(ASTVisitor):
    def __init__(self):
        self.functions = []

    def visit_function_definition(self, node):
        name_node = node.child_by_field_name("name")
        if name_node:
            self.functions.append(name_node.text.decode())
        self.generic_visit(node)

extractor = FunctionExtractor()
extractor.visit(tree.root_node)
print(extractor.functions)

Performance Tips

1. Incremental parsing: For edits, use parser.parse(new_source, old_tree)
2. Lazy evaluation: Don't traverse entire tree if you only need specific nodes
3. Query optimization: Use more specific queries to reduce matches
4. Memory management: Large files can use significant memory
5. Batch processing: Process multiple files in parallel

Integration

With Code Analysis

def analyze_file(path: Path) -> CodeAnalysis:
    source = path.read_bytes()
    tree = parser.parse(source)

    return CodeAnalysis(
        functions=extract_functions(tree),
        classes=extract_classes(tree),
        imports=extract_imports(tree),
        complexity=calculate_complexity(tree)
    )

With BAML

class CodeStructure {
  functions FunctionInfo[]
  classes ClassInfo[]
  imports string[]
}

class FunctionInfo {
  name string
  parameters string[]
  return_type string?
  line_start int
  line_end int
}

Best Practices

1. Error tolerance: Always handle parse errors gracefully
2. Use queries: Prefer queries over manual traversal
3. Test on real code: Test with actual codebases, not just samples
4. Document node types: Reference language grammar for node types
5. Version parsers: Pin tree-sitter language versions