py-ast

Python AST Parser for TypeScript

A comprehensive TypeScript-based Python source code parser that generates Abstract Syntax Trees (AST) following the Python ASDL grammar specification. This library provides complete parsing, unparsing, and AST traversal infrastructure similar to ESPrima for JavaScript, with bidirectional Python code ↔ AST conversion.

Features

• 🔍 Complete Python lexical analysis - Tokenizes Python source code with full syntax support
• 🌳 AST generation - Creates comprehensive Abstract Syntax Trees based on Python ASDL grammar
• � Code generation - Convert AST back to Python source code with unparse()
• �🚶 AST traversal - Walk and visit all nodes in the syntax tree
• 📄 JSON serialization - Export ASTs to JSON format for analysis or storage
• 🔧 TypeScript types - Full type definitions for all AST nodes
• ⚡ ESPrima-style API - Familiar interface for JavaScript developers
• 🐍 Python-compatible - Follows Python's official AST structure

Installation

npm install py-ast

Quick Start

Basic Parsing and Code Generation

import {
  parse,
  parsePython,
  unparse,
  walk,
  NodeVisitor,
} from "py-ast";

// Parse Python source code - that's it! No mode selection needed.
const pythonCode = `
def fibonacci(n):
    if n <= 1:
        return n
    else:
        return fibonacci(n-1) + fibonacci(n-2)

result = fibonacci(10)
`;

// Generate AST - handles any Python code (expressions, statements, modules)
const ast = parse(pythonCode);
console.log(ast.nodeType); // "Module"

// Convert AST back to Python source code
const regeneratedCode = unparse(ast);
console.log(regeneratedCode);
// Output: Properly formatted Python code equivalent to the original

Parsing Different Python Constructs

import { parse, unparse } from "py-ast";

// 1. Simple expressions
const expr = parse("x + y * 2");
console.log(expr.nodeType); // "Module"
console.log(unparse(expr)); // "x + y * 2"

// 2. Function definitions
const funcCode = `
def greet(name, greeting="Hello"):
    return f"{greeting}, {name}!"
`;
const funcAst = parse(funcCode);
console.log(unparse(funcAst));

// 3. Class definitions with methods
const classCode = `
class Calculator:
    def __init__(self, precision=2):
        self.precision = precision
    
    def add(self, a, b):
        return round(a + b, self.precision)
    
    @staticmethod
    def multiply(x, y):
        return x * y
`;
const classAst = parse(classCode);
console.log(unparse(classAst));

// 4. Complex expressions with comprehensions
const complexExpr = parse(`
result = [x**2 for x in range(10) if x % 2 == 0]
data = {key: value for key, value in items.items() if value > 0}
`);
console.log(unparse(complexExpr));

// 5. Async/await patterns
const asyncCode = `
async def fetch_data(urls):
    async with aiohttp.ClientSession() as session:
        tasks = [fetch_url(session, url) for url in urls]
        return await asyncio.gather(*tasks)
`;
const asyncAst = parse(asyncCode);
console.log(unparse(asyncAst));

Working with Parse Options

import { parse, parsePython } from "py-ast";

// Basic parsing with filename for better error reporting
const ast1 = parse(pythonCode, { filename: "fibonacci.py" });

// Enable comment parsing to include hash comments in AST
const codeWithComments = `
# This is a header comment
def process_data(items):
    """This is a docstring, not a comment"""
    # This is an inline comment
    return [x * 2 for x in items]  # Another comment
`;
const astWithComments = parse(codeWithComments, {
  comments: true,
  filename: "commented_code.py",
});

// Comments are now available in the AST
console.log(astWithComments.comments?.length); // Number of hash comments found
astWithComments.comments?.forEach(comment => {
  console.log(`Line ${comment.lineno}: ${comment.value}`);
});

// Alternative explicit function name
const ast2 = parsePython(pythonCode, { filename: "fib.py" });

AST Traversal and Analysis

import { parse, walk, NodeVisitor } from "py-ast";

const code = `
class DataProcessor:
    def __init__(self, config):
        self.config = config
    
    def process_file(self, filename):
        with open(filename, 'r') as f:
            data = f.read()
        return self.transform_data(data)
    
    def transform_data(self, data):
        if self.config.uppercase:
            return data.upper()
        return data.lower()

processor = DataProcessor({'uppercase': True})
result = processor.process_file('input.txt')
`;

const ast = parse(code);

// 1. Walk all nodes in the AST
console.log("=== Walking all nodes ===");
for (const node of walk(ast)) {
  console.log(`${node.nodeType} at line ${node.lineno || 'unknown'}`);
}

// 2. Custom visitor to analyze code structure
class CodeAnalyzer extends NodeVisitor {
  functions: string[] = [];
  classes: string[] = [];
  variables: string[] = [];
  imports: string[] = [];

  visitFunctionDef(node: any) {
    this.functions.push(node.name);
    this.genericVisit(node); // Continue visiting child nodes
  }

  visitClassDef(node: any) {
    this.classes.push(node.name);
    this.genericVisit(node);
  }

  visitAssign(node: any) {
    // Extract variable names from assignments
    for (const target of node.targets) {
      if (target.nodeType === "Name") {
        this.variables.push(target.id);
      }
    }
    this.genericVisit(node);
  }

  visitImport(node: any) {
    for (const alias of node.names) {
      this.imports.push(alias.name);
    }
    this.genericVisit(node);
  }

  visitImportFrom(node: any) {
    const module = node.module || '';
    for (const alias of node.names) {
      this.imports.push(`${module}.${alias.name}`);
    }
    this.genericVisit(node);
  }
}

const analyzer = new CodeAnalyzer();
analyzer.visit(ast);

console.log("=== Code Analysis Results ===");
console.log("Functions:", analyzer.functions);
console.log("Classes:", analyzer.classes);
console.log("Variables:", analyzer.variables);
console.log("Imports:", analyzer.imports);

Safe Literal Evaluation

import { literalEval } from "py-ast";

// Safely evaluate Python literals
console.log(literalEval("42"));              // 42
console.log(literalEval('"hello world"'));   // "hello world"
console.log(literalEval("[1, 2, 3, 4]"));    // [1, 2, 3, 4]
console.log(literalEval('{"a": 1, "b": 2}')); // {a: 1, b: 2}
console.log(literalEval("(1, 2, 3)"));       // [1, 2, 3] (tuple as array)
console.log(literalEval("{1, 2, 3}"));       // [1, 2, 3] (set as array)
console.log(literalEval("True"));            // true
console.log(literalEval("None"));            // null

// Complex nested structures
const complexLiteral = `{
    "config": {
        "debug": True,
        "max_items": 100,
        "allowed_types": ["string", "number", "boolean"]
    },
    "data": [1, 2.5, "test", None]
}`;
console.log(literalEval(complexLiteral));

Code Transformation and Generation

import { parse, unparse, NodeTransformer } from "py-ast";

const originalCode = `
def calculate_total(items, tax_rate):
    subtotal = sum(item.price for item in items)
    tax = subtotal * tax_rate
    total = subtotal + tax
    return total

def process_order(order):
    total = calculate_total(order.items, 0.08)
    if total > 100:
        total *= 0.9  # 10% discount for orders over $100
    return total
`;

const ast = parse(originalCode);

// Transform the AST - rename functions and variables
class CodeRefactorer extends NodeTransformer {
  visitFunctionDef(node: any) {
    // Rename functions with a prefix
    if (node.name === "calculate_total") {
      return { ...node, name: "compute_order_total" };
    }
    if (node.name === "process_order") {
      return { ...node, name: "handle_order_processing" };
    }
    return this.genericVisit(node);
  }

  visitName(node: any) {
    // Rename variables
    const renames = {
      "subtotal": "base_amount",
      "tax_rate": "tax_percentage",
      "total": "final_amount"
    };
    
    if (renames[node.id]) {
      return { ...node, id: renames[node.id] };
    }
    return node;
  }

  visitCall(node: any) {
    // Update function calls to match renamed functions
    if (node.func.nodeType === "Name" && node.func.id === "calculate_total") {
      return {
        ...node,
        func: { ...node.func, id: "compute_order_total" }
      };
    }
    return this.genericVisit(node);
  }
}

const transformer = new CodeRefactorer();
const transformedAst = transformer.visit(ast);

// Generate the refactored code
const refactoredCode = unparse(transformedAst);
console.log("=== Refactored Code ===");
console.log(refactoredCode);

// Custom indentation
const compactCode = unparse(transformedAst, { indent: "  " });
console.log("=== Compact Version ===");
console.log(compactCode);

Advanced Usage: JSON Serialization and Analysis

import { parse, walk } from "py-ast";

const pythonCode = `
import asyncio
from dataclasses import dataclass
from typing import List, Optional

@dataclass
class User:
    id: int
    name: str
    email: Optional[str] = None

class UserService:
    def __init__(self, database_url: str):
        self.db_url = database_url
        self.users: List[User] = []
    
    async def fetch_user(self, user_id: int) -> Optional[User]:
        # Simulate async database call
        await asyncio.sleep(0.1)
        return next((u for u in self.users if u.id == user_id), None)
    
    def add_user(self, user: User) -> None:
        self.users.append(user)

# Usage
service = UserService("postgresql://localhost/mydb")
user = User(1, "John Doe", "[e-mail]")
service.add_user(user)
`;

const ast = parse(pythonCode);

// Serialize AST to JSON for analysis or storage
const astJson = JSON.stringify(ast, null, 2);
console.log("AST JSON size:", astJson.length, "characters");

// Analyze AST structure
const nodeStats = new Map<string, number>();
for (const node of walk(ast)) {
  const count = nodeStats.get(node.nodeType) || 0;
  nodeStats.set(node.nodeType, count + 1);
}

console.log("=== AST Node Statistics ===");
console.log(Object.fromEntries(
  Array.from(nodeStats.entries()).sort((a, b) => b[1] - a[1])
));

// Extract specific information
class ImportAnalyzer extends NodeVisitor {
  imports: Array<{type: string, module: string, names: string[]}> = [];

  visitImport(node: any) {
    this.imports.push({
      type: 'import',
      module: '',
      names: node.names.map((alias: any) => alias.name)
    });
  }

  visitImportFrom(node: any) {
    this.imports.push({
      type: 'from_import',
      module: node.module || '',
      names: node.names.map((alias: any) => alias.name)
    });
  }
}

const importAnalyzer = new ImportAnalyzer();
importAnalyzer.visit(ast);
console.log("=== Import Analysis ===");
console.log(importAnalyzer.imports);

API Reference

Core Functions

parse(source, options?)

Parses Python source code and returns an AST. Handles any Python code - expressions, statements, modules, classes, functions, etc.

Parameters:

• source (string): The Python source code to parse
• options (ParseOptions, optional): Parsing options

ParseOptions:

• filename (string): Source filename for error reporting (default: '<unknown>')
• comments (boolean): Include hash comments (# comment) in AST (default: false)
• feature_version (number): Python feature version

Note: Triple-quoted strings ("""text""") are parsed as regular string literals, not comments. Only hash comments (# comment) are treated as comments when comments: true is enabled.

Returns: Module - The root AST node containing all parsed statements

// Basic usage - handles any Python code
const ast = parse("x = 42");
const ast2 = parse("def func(): return 42");
const ast3 = parse("class MyClass: pass");

// With options
const ast = parse("x + y", {
  filename: "my_script.py",
  comments: true,
});

parseFile(filename, options?)

Note: Placeholder function. Read file content first and use parse().

literalEval(source)

Safely evaluates Python literal expressions by parsing and evaluating constant values.

console.log(literalEval("42")); // 42
console.log(literalEval('"hello"')); // "hello"
console.log(literalEval("[1, 2, 3]")); // [1, 2, 3]
console.log(literalEval('{"key": "value"}')); // {key: "value"}

unparse(node, options?)

Converts an AST node back to Python source code. This is the reverse operation of parse().

Parameters:

• node (ASTNodeUnion): The AST node to convert back to source code
• options (object, optional): Unparsing options
  • indent (string): Indentation string (default: " " - 4 spaces)

Returns: string - The generated Python source code

import { parse, unparse } from "py-ast";

// Basic roundtrip: parse then unparse
const originalCode = "def greet(name):\n    return f'Hello, {name}!'";
const ast = parse(originalCode);
const generatedCode = unparse(ast);

console.log(generatedCode);
// Output: def greet(name):
//     return f"Hello, {name}!"

// Perfect roundtrip for simple expressions
const simpleCode = "x = 42";
const simpleAst = parse(simpleCode);
console.log(unparse(simpleAst) === simpleCode); // true

// Custom indentation
const ast2 = parse("if True:\n    pass");
const twoSpaceCode = unparse(ast2, { indent: "  " });
console.log(twoSpaceCode);
// Output: if True:
//   pass

// Works with all Python constructs
const complexCode = `
class Calculator:
    def add(self, a, b):
        return a + b
    
    async def process(self, items):
        results = [await self.compute(x) for x in items if x > 0]
        return results
`;

const complexAst = parse(complexCode);
const regenerated = unparse(complexAst);
// regenerated contains valid Python code equivalent to the original

Quote Style Preservation:

The unparser automatically preserves the original quote styles used in string literals when parsing with comments enabled:

const codeWithMixedQuotes = `
name = 'John'
message = "Hello, world!"
multiline = '''This is
a multiline string'''
`;

const ast = parse(codeWithMixedQuotes, { comments: true });
const unparsed = unparse(ast);
console.log(unparsed);
// Output preserves original quote styles:
// name = 'John'
// message = "Hello, world!"  
// multiline = '''This is
// a multiline string'''

walk(node)

Recursively walks all nodes in an AST tree.

for (const node of walk(ast)) {
  console.log(`${node.nodeType} at line ${node.lineno}`);
}

Convenience Functions

parsePython(source, options?)

Alternative name for parse() that makes the intent clearer.

import { parsePython } from "py-ast";

const ast = parsePython("x = 42", { filename: "script.py" });

parseModule(source, filename?)

Legacy convenience function for parsing with just a filename.

import { parseModule } from "py-ast";

const ast = parseModule("def hello(): pass", "hello.py");

Lexer

Lexer

Low-level tokenizer for Python source code.

import { Lexer, TokenType } from "py-ast";

const lexer = new Lexer("x = 42 + 3.14");
const tokens = lexer.tokenize();

tokens.forEach((token) => {
  console.log(`${token.type}: ${token.value}`);
});
// Output:
// NAME: x
// EQUAL: =
// NUMBER: 42
// PLUS: +
// NUMBER: 3.14
// EOF:

Visitors

NodeVisitor

Base class for creating custom AST visitors.

class CountVisitor extends NodeVisitor {
  counts = new Map<string, number>();

  visitFunctionDef(node: any) {
    this.increment("functions");
    this.genericVisit(node);
  }

  visitClassDef(node: any) {
    this.increment("classes");
    this.genericVisit(node);
  }

  private increment(key: string) {
    this.counts.set(key, (this.counts.get(key) || 0) + 1);
  }
}

NodeTransformer

Visitor that can modify AST nodes during traversal.

class RenameTransformer extends NodeTransformer {
  visitName(node: any) {
    if (node.id === "old_name") {
      return { ...node, id: "new_name" };
    }
    return node;
  }
}

Node Types

The library provides TypeScript interfaces for all Python AST nodes based on the ASDL grammar:

• Module nodes: Module, Interactive, Expression, FunctionType
• Statement nodes: FunctionDef, ClassDef, If, For, While, With, etc.
• Expression nodes: BinOp, Call, Attribute, Subscript, List, Dict, etc.
• Supporting types: Arguments, Keyword, Alias, etc.

Usage Examples

Parse and Analyze Python Code

import { parse, walk, NodeVisitor } from "py-ast";

const code = `
class Calculator:
    def add(self, a, b):
        return a + b
    
    def multiply(self, a, b):
        return a * b

calc = Calculator()
result = calc.add(5, 3)
`;

const ast = parse(code);

// Count different node types
const nodeCounts = new Map<string, number>();
for (const node of walk(ast)) {
  const count = nodeCounts.get(node.nodeType) || 0;
  nodeCounts.set(node.nodeType, count + 1);
}

console.log("Node distribution:", Object.fromEntries(nodeCounts));

Find Function Definitions

class FunctionFinder extends NodeVisitor {
  functions: string[] = [];

  visitFunctionDef(node: any) {
    this.functions.push(node.name);
    this.genericVisit(node);
  }

  visitAsyncFunctionDef(node: any) {
    this.functions.push(node.name);
    this.genericVisit(node);
  }
}

const code = `
def sync_function():
    pass

async def async_function():
    pass
`;

const ast = parse(code);
const finder = new FunctionFinder();
finder.visit(ast);
console.log(finder.functions); // ['sync_function', 'async_function']

Code Generation and Transformation

import { parse, unparse, NodeTransformer } from "py-ast";

// Parse, modify, and regenerate Python code
const originalCode = `
def process_data(items):
    results = []
    for item in items:
        if item > 0:
            results.append(item * 2)
    return results
`;

const ast = parse(originalCode);

// Transform the AST - rename variables
class VariableRenamer extends NodeTransformer {
  visitName(node: any) {
    if (node.id === "item") {
      return { ...node, id: "element" };
    }
    if (node.id === "results") {
      return { ...node, id: "output" };
    }
    return node;
  }
}

const transformer = new VariableRenamer();
const transformedAst = transformer.visit(ast);

// Generate new Python code
const newCode = unparse(transformedAst);
console.log(newCode);
// Output: Function with renamed variables (item -> element, results -> output)

// Perfect roundtrip for unmodified code
const simpleCode = "x = [i**2 for i in range(10) if i % 2 == 0]";
const roundtripCode = unparse(parse(simpleCode));
console.log(simpleCode === roundtripCode); // true - perfect roundtrip

Test Results

=== Testing Python AST Parser ===

1. Testing Lexer:
Tokens: NAME(x) EQUAL(=) NUMBER(42) PLUS(+) NUMBER(3.14) EOF()

2. Testing Parser:
✓ Parsed successfully!
AST Type: Module
Statements: 4

3. Testing AST Walking:
Total nodes: 19
Node types: Module, Assign, Name, Load, Constant, Expr, BinOp, Add

4. Testing JSON serialization:
✓ JSON serialization successful
JSON length: 1818 characters

=== Test Complete ===

Supported Python Features

Parsing & Code Generation Support

• ✅ Functions and Classes - Function/class definitions with decorators
• ✅ Control Flow - if/elif/else, for/while loops, try/except
• ✅ Expressions - All binary/unary operations, comparisons, calls
• ✅ Literals - Numbers, strings, lists, dicts, sets, tuples
• ✅ Comprehensions - List/dict/set comprehensions and generators
• ✅ Async/Await - Async functions, async for, async with
• ✅ Context Managers - with statements
• ✅ Import Statements - import and from...import
• ✅ Exception Handling - try/except/finally

Code Generation: All parsed constructs can be converted back to Python source code using unparse(). The unparser produces clean, readable Python code that maintains semantic equivalence with the original.

Design Principles

This library is designed to be independent from Python's built-in ast module while following the same ASDL grammar specification:

• TypeScript-native - Built for TypeScript/JavaScript environments
• Bidirectional - Parse Python to AST and unparse AST back to Python
• JSON serializable - All nodes can be directly serialized to JSON
• ESPrima-style API - Familiar interface for web developers
• Custom AST format - Optimized for JavaScript object handling
• No Python runtime required - Runs entirely in Node.js/browser

Development

# Install dependencies
npm install

# Build the library
npm run build

# Run tests
npm test

License

MIT License

AI Usage

Most part of this project is built using Claude Sonnet 4 in Github Copilot. However, a comprehensive verification and testing of both the code and the features.