Skip to main content

Compilation Pipeline

Source Code (.mt)
    ↓
Lexer (Tokenization)
    ↓
Parser (AST Construction)
    ↓
AST Optimizer (optional, by optimization level)
    ├─ Constant Folding Pass
    ├─ Dead Code Elimination Pass
    └─ Unused Declaration Elimination Pass
    ↓
Bytecode Compiler
    ↓
Bytecode Program (.mtc)
    ↓
Virtual Machine (stack-based execution)
    ↓
Result

Stages​

1. Lexer​

Translates source text to a token stream. Tracks source locations for diagnostics; balances brackets so the parser can recover from mismatched delimiters.

2. Parser​

Constructs the AST from tokens. Specialized sub-parsers handle classes, interfaces, expressions, statements, lambdas, and types. Errors are accumulated rather than aborting on the first failure.

3. AST Optimizer​

Optional, pluggable passes:

  • Constant Folding — reduces 1 + 2 to 3 at compile time.
  • Dead Code Elimination — removes unreachable branches. Preserves @Script and @Throw annotations.
  • Unused Declaration Elimination — drops symbols nothing references.

Optimization level selects which passes run. Debug compiles skip optimization to keep source-line mapping faithful.

4. Bytecode Compiler​

Lowers AST to a stack-based instruction stream. Emits:

  • Instructions and their operands.
  • A constant pool (strings, numbers).
  • Class metadata (fields, method tables, generic parameters).
  • Source locations for debugger and stack-trace use.

Output is either an in-memory BytecodeProgram (for mType script.mt) or a serialized .mtc file (for mType --compile).

5. Virtual Machine​

Executes the bytecode. See VM.

Execution Modes​

  1. Direct — mType script.mt. Compiles AST → bytecode in memory and executes.
  2. Compile — mType --compile script.mt. Emits script.mtc.
  3. Cached — mType --run-cached script.mtc. Loads bytecode and executes.
  4. Project Build — mType --build [project.mtproj]. Compiles every source file to .mtc, optionally bundling into a .mtcLib or native .exe.

See Also​