/code_graph

Fast program graph generation in Python

Primary LanguagePythonMIT LicenseMIT

Code Graph

Fast program graph generation in Python

Many Programming Language Processing (PLP) exploit the fact that programming languages are highly structured. Therefore, it is easy to parse a program into an abstract syntax tree, analyse its control flow and track data flow relations between variables (the last one is a bit more harder :D).

code.graph provides easy access to graph representations of program codes for code written in Java and Python. The library is mainly designed to replicate the graph representation introduced in Self-Supervised Bug Detection and Repair for Python and published in NeurIPS21 by Allamanis et al. Therefore, the implementation is close to the implementation used by the original authors.

Note: This implementation does not compute interprocedural relations currently and its main purpose is parsing the implementation of single functions.

Installation

The package is tested under Python 3. It can be installed by cloning this repository and installing the package via:

pip install -e .

Usage

code.graph can be used to transform Java and Python program into a graph representation with a few lines of code:

import code_graph as cg

# Python
cg.codegraph(
    '''
        def my_func():
            print("Hello World")
    ''',
lang = "python")

# Output: PythonCodeGraph(19), a graph with 19 nodes

# Java
cg.codegraph(
    '''
        public static void main(String[] args){
          System.out.println("Hello World");
        }
    ''',
lang = "java", 
syntax_error = "ignore")

# Output: JavaCodeGraph(32)

Further, you can easily traverse the code graph, e.g. via depth-first search:

graph = cg.codegraph(...)

dfs_stack = [graph.root_node] # Root of the parsed AST
while len(dfs_tack) > 0:
    node = dfs_stack.pop(-1)

    for current, edge_type, next_node in node.successors():
        dfs_stack.append(next_node)

Alternatively, you can also export the graph int Dot Format by:

graph.todot("file_name.dot")

Project Info

This is currently developed as a helper library for internal research projects. Therefore, it will only be updated as needed.

Feel free to open an issue if anything unexpected happens.

Distributed under the MIT license. See LICENSE for more information.

We thank the developer of tree-sitter library. Without tree-sitter this project would not be possible.