A compiler frontend for a subset of the Cangjie programming language that performs comprehensive static analysis and validation.
This project implements the frontend stages of a Cangjie compiler, transforming source code through multiple representations to validate syntax, types, and initialization semantics. Valid programs are pretty-printed for compilation; invalid programs generate detailed, actionable error messages.
Converts source text into tokens with kind, value, and position information for error reporting.
Builds an Abstract Syntax Tree (AST) from tokens using a grammar subset that excludes packages, type aliases, extensions, foreign declarations, annotations, and macros.
Transforms the AST into an Abstract Syntax Graph (ASG) where variables are represented as edges rather than nodes, enabling precise variable resolution and scope tracking through environment structures.
Performs two-pass type inference and validation:
- Validates type declarations form a valid DAG (no cycles)
- Infers types for expressions, considering context and overload resolution
- Handles complex cases like function calls with multiple candidates and conditional branches
- Ensures assignments and operations use compatible types
Verifies proper variable and class initialization:
- Variables are initialized before use
- Immutable variables are not mutated
- Class constructors initialize all members
- Static/global variables respect declaration order
- Constructor functions follow proper calling conventions
- Incremental grammar implementation: Developed layer-by-layer for maintainability
- ASG representation: Solves variable resolution problems inherent in ASTs
- Sophisticated type inference: Handles overloading, subtyping, and contextual type requirements
- Comprehensive initialization tracking: Scope-aware analysis with special handling for classes and closures
- Code refactoring and cleanup
- Grammar extension to cover more language features
- Enhanced error handling across all compilation stages