I love messing with the TypeScript type system. Somehow building crazy things with technologies that were definitely not supposed to be built with is exciting for me. And, building a Brainf*ck interpreter in the TypeScript type system is basically doing exactly that but twice at the same time.

I built the interpreter using only TypeScript types. I was also using Deno for easy TypeScript integration but the code in this repository will run on anywhere that can run a version of TypeScript that supports template literal string types and conditional inferration. It greatly depends on using recursive types. Here's a small example of a recursive type to build an array of N size with each item of type T:

type TupleOfNSize<N extends number, T, $Draft extends T[] = []> =
  // 'lil spacer for better formatting
  $Draft["length"] extends N ? $Draft : TupleOfNSize<N, T, [...$Draft, T]>;

You can see, we basically just build an array using a recursive type that hold the accumulator array, $Draft. I prefer to prefix private generics within types with a dollar sign but that's just my personal preference.

This basic recursive approach combined with TypeScript's inferring within conditional types allowed me to create the Brainf*ck interpreter.

Let's run a hello world Brainheck program using our interpreter.

// import type { BF } from "./mod.ts";

type Output = BF<{
  program: "        [>    [>  >   >   > <<<<-]> > >->> [<]<-]>>.>---.       ..   .>>.<-.<.   .------.--------.>> .>  .";
  outputType: "ascii";
}>;

Now, when you hover over the Output type, you will see that its type is "Hello, world!\n". Note how the BF type simply returns the output. See the API section for how you can pass input and change the output encoding.

Due to TypeScript's type complexity cap, it is impossible to run certain complex programs as we breach the recursive type limit when running our recursive types.

This is the only exported type from mod.ts and it is the type used to run programs. It takes in a single object type with the options for how to run your program within it. The pointers are Unsigned 8-bit integers and the default memory size is 256 slots. Here are the options you should know about:

• program

This is your Brainheck program.

• input

The input that the Brainheck program can ready byte-by-byte via the , command. By default this is just an empty string.

• memory

The initial memory for the program. This is 256 slots byte default initialized with 0.

• outputType

The type of the output when you use the . command. This determines the type of the output representation when the BF type finishes and returns your output. By default, this will be "buffer" which means that it will return all of the bytes stored in a tuple with each of the unsigned 8-bit integers within. However, you can specify "ascii" in which the API will return a string with the output characters encoded in the "ascii" format in a single string.

MIT