bf2fj — Brainfuck to FlipJump

github.com/tomhea/bf2fj

A direct Brainfuck-to-FlipJump compiler with four optimisation passes. Smaller and faster than running Brainfuck on top of an interpreter implemented in FlipJump.

Install

pip install bf2fj

Run

bf2fj hello_world.bf            # generate hello_world.fj
bf2fj hello_world.bf -r         # generate AND immediately run

-r (run) shells out to the local fj interpreter — make sure the main FlipJump toolchain is installed too.

Optimisations

The compiler applies four passes to the Brainfuck source before code generation:

  1. Consolidate consecutive data ops — +++++ becomes a single “add 5” macro call instead of five separate + operations.

  2. Consolidate pointer movements — >>>> becomes “shift right by 4” instead of four separate >s.

  3. Replace zeroing loops — the [-] idiom (loop subtracting 1 until zero) is detected and replaced with a direct “set to 0” operation.

  4. Combine zero+modify — [-]+++ (zero, then add 3) becomes a single “set to 3” operation.

The cumulative effect is large. On the BF test corpus shipped with the repo, total operation counts drop ~88% — one program went from 2,900,990 ops (naïve translation) to 337,484 ops (with all four passes).

Useful flags

Flag

Purpose

-c N / --cells N

Reserve N Brainfuck tape cells (default: enough for typical programs).

-r / --run

Compile then immediately run via the local fj interpreter.

Caveats

  • Output is a single .fj file matching the input’s base name.

  • The compiler doesn’t currently expose hooks for custom cell sizes or signed-cell semantics beyond the BF spec defaults.

See also