readme.md

Code Generator

The code generator, located in malang/codegen/codegen.go, is the final phase of the compiler (in this simple design). It takes the AST and translates it into executable code. In this case, the target language is Go.

Theoretical Background:

• Intermediate Representation (IR): Many compilers use an intermediate representation (IR) between the AST and the final target code. The AST itself can be considered a high-level IR. More sophisticated compilers might use multiple IRs, gradually lowering the level of abstraction.
• Tree Traversal: Code generation typically involves a tree traversal of the AST. The Malang compiler uses a recursive, depth-first traversal.
• Target Code: The code generator's goal is to produce code that is semantically equivalent to the original Malang program.

Key Components:

• GenerateCode(program ast.Program) string: The main function.

• generateStatementCode(...): Generates code for statements.

• generateExpressionCode(...): Generates code for expressions, handling:

  • Operator Precedence: Uses operatorPrecedence() to determine the order of operations.
  • Associativity: Uses isLeftAssociative() to handle operators with the same precedence.
  • String Conversion: Uses generateExpressionCodeForParayu and the isParayu flag to conditionally convert integers to strings within parayu statements.
  • Type inference: Uses inferType to infer the type based on the operator and operands.

• declaredVars map[string]string: Tracks declared variables and their inferred types.

Design Choices:

• Target Language: The code generator produces Go code. This simplifies the compilation process, as we can leverage the Go compiler to produce native executables.
• Tree Traversal: The code generator uses a recursive, depth-first traversal of the AST to generate code.
• Dynamic Typing: Malang does not have explicit types, the compiler uses type inference based on how variables are used.