Thursday, November 29, 2007

Something like Lisp

Ever since compilers class in college, I've toyed with the idea of creating a new programming language. Every year or so a spend a few days thinking about one and get off to a short start. The problem is I often try to bite off more than I can chew and start dealing with the most difficult parts first. So I recently decided to get a minimal language up and running. Something with simple syntax, no concern for efficiency, and an easy to build base.

Lisp has some of the simplest syntax of any of the languages I've worked with (perhaps with the exception of BF). I was looking for something simple which I could quickly implement, so I used it as a model and created an interpreter in Python. The whole project took just a couple of hours, but I was able to write an interpreter that would run the following program:
set[x 5]
set[x add[1 2 3 4 5]]
set[y add[1 2]]

As you might have guessed, this program displays the following in the terminal:
I implemented a simple parser using Python's regular expression module re (The group method can come in quite handy), and I defined each of the above functions in a dictionary which maps function calls in my custom language to functions in a Python script which I import. You might have noticed that each of the methods is designed to take a variable number of arguments (thought only they may not all be used), and all function calls are made recursively, with nested functions being evaluated first. The syntax looks slightly like s-expressions but it's a bit more accessible for someone who has worked with C-like languages (except there's no need to reach for the shift key for those tricky parenthesis). I haven't added a mechanism for lambda expressions yet, but defining new functions might look something like this:
function[my_fun group[x y] group[multiply[x add[x y]]]]
The above would define a function named my_fun which takes two arguments, and multiplies the first by the sum of the two arguments.

I just made up this micro language, and haven't given this a great deal of thought. The idea was to create something quick and dirty to create a working interpreter, rather than the usual of intense planning without a working prototype. This language will probably never see any further development. It was a mental exercise, but I found it quite enjoyable. If you'd like to see the source code for the interpreter, please let me know (I get the feeling Matt might be somewhat interested).

No comments: