Brainfuck Research Compliler

The project is designed as a tool for learning about compiler construction in an extremely simple way. Although being kind of over-engineered for the purpose of compiling Brainfuck, this compiler features most of the architectural elements of a mature language compiler. Nevertheless, its simplicity allows beginners to implement extensions in the course of an afternoon instead of weeks of work.

This project is licensed under a liberal and easily understandable ISC license, see LICENSE file.

The repository contains several sub-projects:

bfrc-base

defines the common API used by all other projects, including the Abstract Syntax Tree (AST) structure.

The general flow through a compiler is described in the following illustration. BFRC provides Interfaces for all active elements depicted.

        1            2                           3                             4
Input -----> Lexer -----> Parser -----+----------------------+-----> Backend -----> Output
                                      |                      |
                                      +----- Optimizer <-----+

where

1. is a stream of Characters,
2. is a stream of Tokens,
3. is the AST,
4. is dependent upon the backend. It may be a stream of Bytes, a stream of Characters or nothing at all (interpreter).

bfrc-run

provides the runtime environment to execute the compiler.

You generally call the compiler by {-config} inputFile [outputFile] where

• inputFile is the mandatory parameter specifying the input file
• outputFile is optional, when omitted bfrc will generate the output according to inputFile
• -config loads the configuration config

A configuration is defined by a .config file in the java package bfrc. It is a simple Java properties file, that defines the classes to load via the keys bfrc.lexer, bfrc.parser, bfrc.optimizers and bfrc.backend. Optimizers may be multiple entries separated by whitespace, that are called in the order provided. The Configuration default.config (in bfrc-run) will always be loaded as first config and may be overridden by multiple configurations in the order they are named.

bfrc-bf

includes the default Brainfuck frontend (-bf) and an additional Brainfuck backend (-bf-out).

bfrc-int

contains a simple interpreter backend (-int), that directly traverses the AST.

bfrc-c

contains a backend (-c), that produces simple C files.

bfrc-dot

provides a backend, that creates a cool Graph representation of the AST based on the DOT-language (-dot), that may be rendered by a tool like Graphviz.

bfrc-java

uses Javassist (required library) to create Java Bytecode. There are several options available:

• -java compiles the whole input and saves it as a .class file to disk.
• -aot is just like -java, but instead of saving to disk the class is directly executed in RAM.
• -jit compiles the code just in time on a per-block (loop) basis.

bfrc-ook

provides an additional frontend, that lexes the Ook! language (-ook) while utilizing bfrc-bf's parser. It also includes a backend (-ook-out).

bfrc-opts

contains all optimizations, that may be performed on the AST independently of the front- or backend.