Python AST to EMF Metamodel

Metamodel-based framework for static analysis of Python code using OCL.

Transforms Python Abstract Syntax Trees into EMF model instances, enabling constraint-based static analysis using Eclipse's OCL implementation. Provides infrastructure to define custom OCL constraints for Python code analysis.

Uses Eclipse OCL to leverage:

• Standardized constraint language (OMG OCL 2.4 specification)
• Mature evaluation engine with debugging support
• Integration with EMF validation framework
• Existing tooling ecosystem

Metamodel maintains 1:1 structural correspondence with Python's ast module.

Architecture

Components

Ecore Metamodel (dk.ingi.emf.python.ast.ecore)

• Complete mapping of Python's ast module structure to EMF/Ecore
• Covers Python 3.10+ grammar including pattern matching (PEP 634)
• Preserves structural correspondence through EAnnotations

Acceleo Template (generate.mtl)

• Generates Python AST visitor from Ecore metamodel
• Produces visitor.py with full type hints
• Maintains synchronization between metamodel and implementation

AST Visitor (visitor.py)

• Traverses Python AST via ast.parse()
• Uses JPype to instantiate EMF model elements
• Serializes models as XMI for Eclipse OCL processing

Execution Flow

Python source → ast.parse() → AST nodes → Visitor → EMF instances → XMI → OCL validation

Capabilities

Supported Analysis

Syntactic constraints on AST structure:

• Control flow complexity (cyclomatic, nesting depth)
• Code structure patterns (function length, decorator count)
• Naming conventions via string operations on identifiers
• Literal value detection (magic numbers, hardcoded strings)

Example OCL Constraints

-- Maximum function body length
context FunctionDef
inv: self.body->size() <= 50

-- Limit nested comprehension depth  
context ListComp
inv: self.generators->size() <= 2

-- Require type annotations on function arguments
context FunctionDef
inv: self.args.args->forAll(a | a.annotation <> null)

Limitations

No semantic analysis in current implementation:

• Name resolution (which definition a Name refers to)
• Scope analysis (local/global/nonlocal)
• Type inference or checking
• Dataflow analysis (def-use chains, liveness)
• Call graph construction

These require symbol table construction, not provided in base metamodel.

Extension

Adding Semantic Analysis

Extend metamodel with resolution information:

<eClassifiers xsi:type="ecore:EClass" name="SymbolTable">
  <eStructuralFeatures name="symbols" eType="#//Symbol" 
                       upperBound="-1" containment="true"/>
  <eStructuralFeatures name="parent" eType="#//SymbolTable"/>
</eClassifiers>

<eClassifiers xsi:type="ecore:EClass" name="Symbol">
  <eStructuralFeatures name="name" eType="ecore:EDataType#//EString"/>
  <eStructuralFeatures name="definition" eType="#//AbstractStmt"/>
  <eStructuralFeatures name="references" eType="#//Name" upperBound="-1"/>
</eClassifiers>

Populate via second visitor pass using Python's symtable module or custom resolver.

Usage

Prerequisites

• Python 3.10+
• JPype1
• Eclipse with EMF/OCL plugins
• Java classpath including:
  • org.eclipse.emf.ecore
  • org.eclipse.emf.common
  • org.eclipse.emf.ecore.xmi

Running

python3 visitor.py  # Parses self, outputs models/visitor.ast

Configure eclipse_app and emf_project_dir paths in visitor.py.

OCL Validation

Load generated .ast files into Eclipse OCL environment. Define constraints in .ocl files referencing the http://ingi.dk/emf/python namespace.

Technical Details

Type Mapping

Python constants mapped to EMF via Constant.value (EString) and Constant.kind.

XML Sanitization

Handles Python strings containing characters invalid in XML 1.0 (control characters U+0000-U+001F except tab/LF/CR) via escape sequences.

Performance Characteristics

JPype crossing cost: O(n) where n = AST node count. Subsequent OCL evaluation runs in JVM without additional crossing overhead.

Repository Structure

dk.ingi.emf.python.ast/          # EMF project
  └── model/
      └── python.ecore           # Metamodel definition

dk.ingi.emf.python.ast.api/      # Acceleo project
  └── generate.mtl               # Visitor template