Compiler¶

The Fourier compiler lives at fourier/ in the source tree. It is a single Python package with one public entry point:

from fourier import compile_source

bytecode: bytes = compile_source(source_text)

This constitutes the entire public API; everything else is internal.

Pipeline¶

.fou source text
      │
      │   tokenize()                  fourier/lexer.py
      ▼
list[Token]
      │
      │   parse()                     fourier/parser.py
      ▼
Contract AST
      │
      │   compile_contract()          fourier/codegen.py
      ▼
list of (op, arg) tuples + labels
      │
      │   assemble()                  vm/asm.py
      ▼
bytecode (bytes)

No optimizer, no IR, no register allocator. Each statement compiles to a fixed bytecode template.

Stage summary¶

Stage	Input	Output	Errors raised
Lex	source string	`list[Token]`	`LexError` (unexpected char, empty hex literal)
Parse	tokens	`Contract` AST	`ParseError` (expected X got Y, etc.)
Codegen	AST	assembly items	`CompileError` (unknown identifier, slot collision, type mismatch, etc.)
Assemble	items	bytecode	`ValueError` (PUSH out of range, unknown asm item)

See Error reference for the catalog of error messages each stage can produce.

CLI¶

V1 has no standalone CLI. Compilation runs from Python:

from fourier import compile_source

with open("counter.fou") as f:
    src = f.read()

bytecode = compile_source(src)

with open("counter.bin", "wb") as f:
    f.write(bytecode)

Future extensions¶

A standalone fourierc binary is a planned addition. Until it ships, the python -c '…' one-liners on the CLI usage page run from any terminal and produce identical bytecode.

Subpages¶

CLI usage — compiling a .fou file without writing Python
Python API — compile_source and the internal hooks for lex/parse/codegen access
Error reference — every error message the compiler can produce, with cause and fix