Module code 

Code Analysis and Code2Vec Embeddings

This module provides statistical approximation techniques for code analysis without state explosion, following the code2vec approach.

Architecture

Source Code
     │
     ▼
┌──────────────┐
│  AST Parser  │
└──────────────┘
     │
     ▼
┌──────────────────────────────┐
│  Path Extractor              │
│  (leaf-to-leaf paths)        │
└──────────────────────────────┘
     │
     ▼
┌──────────────────────────────┐
│  Code2Vec Encoder            │
│  (path → vector embedding)   │
└──────────────────────────────┘
     │
     ▼
┌──────────────────────────────┐
│  GNN for Code Graphs         │
│  (type/lifetime propagation) │
└──────────────────────────────┘

Features

• AST Representation: Lightweight AST node types for code analysis
• Path Extraction: Extract paths between terminal nodes (code2vec style)
• Embedding Encoder: Map paths to dense vector representations
• Code Graph Processing: Use GNN layers for type inference

References

• Alon et al. (2019), “code2vec: Learning distributed representations of code”
• Allamanis et al. (2018), “A survey of machine learning for big code”

Modules

pooling

Global pooling operations for code graphs

Structs

AstNode

A node in the Abstract Syntax Tree

AstPath

A path context representing a connection between two terminals

Code2VecEncoder

Code2Vec encoder for generating embeddings from AST paths

CodeEmbedding

Code embedding representation

CodeGraph

Code graph representation

CodeGraphEdge

An edge in the code graph

CodeGraphNode

A node in the code graph

CodeMPNN

Stack of MPNN layers for deep code analysis

CodeMPNNLayer

Message Passing Neural Network layer for code graphs

PathContext

Context for a path including positional information

PathExtractor

Extracts paths from AST following the code2vec approach

Token

A token (terminal node) in the AST

Enums

AstNodeType

Types of AST nodes for code analysis

CodeEdgeType

Edge types in a code graph

TokenType

Types of tokens (terminal nodes in the AST)

Constants

DEFAULT_EMBEDDING_DIM

Default embedding dimension

MAX_PATHS_PER_METHOD

Maximum number of paths to sample per method

MAX_PATH_LENGTH

Maximum path length for code2vec paths