How to Write an Interpreter in One Day

While we're showing off pointless interpreters, here's a simple currying lazy lambda-calculus-like language with pattern-matching-based choice: http://hpaste.org/fastcgi/hpaste.fcgi/view?id=24800

It achieves recursive definitions using fix (teh Y combinator!! \f.(\x.f (x x)) (\x.f (x x))). Binding variables is done with lambda abstractions. Pattern matching is done with (foo).

    fix (\a.\(0) n.add n 1;m n.(\(true).a (sub m 1) 1;(false).a (sub m 1) (a m (sub n 1))) (eq n 0)) 3 2

It works out the Ackermann of 3 and 2 (29) in about the same time normal Haskell does in GHCi. Isn't it just a joy to read?

Such a toy language isn't pointless. Everyone needs to implement a language, even a toy one. If one is working in C, one should learn assembly and implement a toy C-like language. This is really a necessity, so one has a working knowledge of what happens under the covers when they compile a program.

Such a freely available kit with a toy assembly language, a provided lexical analyser, and a toy grammar with no semantics would be a great boon to the programming community at large, especially if the project could be completed by the student in one weekend. Countless lightbulbs going off in programmer's heads all around the world would result.

The one day interpreter project should be done by everyone who wants to become uber-devs in a language like Python or Ruby.

Choosing a postfix syntax stack oriented language can make things lot easier - at least for the interpreter developer, if not for anyone trying to program in your language.

He says Yacc and Lex suck, but didn't mention ANTLR which, I'm told, sucks a lot less. Does anyone here have experience enough to compare and contrast?

Related: http://antirez.com/page/picol

#lang scheme

(define my-interpreter eval)