Release: Scotch programming language, version 0.2.0 | Hacker News Reader

Release: Scotch programming language, version 0.2.0

bendmorris.com

25 pointsbendmorris11 comments

This looks like it could be promising: "Intended to provide a fast, developer-friendly functional scripting language with increased productivity similar to object-oriented languages Python and Ruby."

What's the evaluation strategy? At first glance I thought it might be term rewriting, but it doesn't appear to be so. I don't understand what I'm allowed to pattern match. I cannot define a function like "foo(Somedata(a, b)) = a" and I don't know why.

The interpreter also became confused at one point, and stopped handling inputs.

Since this is Haskell, you should probably be using cabal to build. I didn't know where the function main was, and had to dig around until I knew which file to invoke ghc --make with.

>> foo(Somedata(a,b)) = a

>> foo(Somedata(1,2))

It is term rewriting. If I understand you correctly, this should work as you would expect.

Thanks for taking some time to play around with it. There are definitely still some bugs to work out in the interpreter; any details you can provide about the issues you find would be much appreciated.

Looks like it was because I had a space after the comma. Oops!

Why did you choose to distinguish between data constructors and functions?

Whoops, that'd be a parsing error. Thanks for catching that.

I think different syntax for data constructors and functions reduces ambiguity and adds to readability.

One reason to have separate domains for data and functions is that it allows you to have types. If you start with the presumption that you aren't going to have types, you no longer need to make this distinction.

Scotch only goes half-way. For example, I can't do this in Scotch:

    >> foo(a,b) = a
    
    >> foo(bar(a,b),bar(c,d)) = bar(a,c)
    
    >> bar(a) = "spaghetti"
    
    >> foo("baz","qux")
    
    "baz"
    
    >> foo(foo(bar("baz","qux"),bar("corge","grault")))
    
    "baz"
    
    >> foo(bar("anything"),bar("something"))
    
    "spaghetti"

Scotch distinguishes between data construction and function application, but the only benefit you get out of it is that you won't get any typing errors.

If you are interested by the sort of very powerful pattern matching demonstrated above, check out languages in the family of term rewriting, of which the newest and coolest is Pure, http://code.google.com/p/pure-lang/, powered by LLVM with influence from Haskell. It is also dynamically typed (by necessity).

Another popular example of term rewriting is the Mathematica language, used in the software package of the same name.

Forgot to also add: good job so far, but clean it up so it works with cabal/hackage :)

The syntax looks amazing. Would be great if you added information about the interpreter (is it implemented in Haskell?) and build instructions for Mac.

Thanks! The interpreter is implemented in Haskell. Better documentation is in the works. Until then, here's an earlier thread with Mac build instructions:

http://news.ycombinator.com/item?id=2037973

What's the underlying motivation?

Basically, the aim is to be developer-friendly and fast to throw things together in, but with the power of a functional language. It's a dynamic, weak typed functional language with readable syntax that makes certain things that are chores in other languages easier - I/O and concurrency being two examples.

Scotch is very new and if you haven't heard of it before, you can check the main wiki page on github for some more links to blog posts, etc. It started as a hobby project that, due to a lot of positive feedback, I'm trying to push a little harder.

Reply on news.ycombinator.com