Re: From dynamic to static

[David Given <dg@...>]

On Monday 31 October 2005 17:06, PA wrote:
> http://www.mindview.net/WebLog/log-0025

He wants to use OCaml, he does.

OCaml is a strong, polymorphic, statically typed language where, ideally, the 
compiler guesses what types you're using automatically. If you say:

let fnord x = x

...then it knows that the function fnord will work for all xes, regardless 
what type they are. If you say:

let fnord x = x + 1

...then it knows that x must be an integer (because + only works on integers).

But the really cunning bit is that this applies to objects too. I can't 
remember the syntax, so I'll switch to Lua-esque pseudocode. This:

o:method1()

...tells the compiler that o must be an object that implements method1. This 
means that you can pass in *any object* that happens to implement method1 --- 
it'll work, and be statically typed. Using his example, you could implement 
Cat, Dog and Bob has three seperate classes, and provided they all 
implemented the 'speak()' method, this function:

function command(pet)
	pet:speak()
end

...would work on all of them *and be type-safe*. If you like, the compiler has 
generated an implicit interface containing the 'speak' method.

It's very, very neat, and manages to give you the best of both worlds; strong, 
statically checked types, and easy-to-write, polymorphic code.

Unfortunately, OCaml doesn't quite live up to this. There's a whole bunch of 
issues with the language design which means you have to help the type checker 
along, and it's very hard to produce 'nil' object pointers --- when I last 
tried to write a real program in it, I eventually gave up in disgust.

I don't see any reason why you couldn't implement a Lua type checker that used 
the same concepts. It would have to analyse a whole program at once, of 
course, and some of Lua's more dynamic features would manage to bypass the 
type checker... but it should be possible to define a subset of the Lua 
semantics that could be usefully typechecked, but would also run on a 
standard interpreter (which means no explicit type annotations). 
Unfortunately, I suspect this is beyond my skills...

-- 
+- David Given --McQ-+ "If God spammed rasfw with the answer to Life, the
|  dg@cowlark.com    | Universe, and Everything, I'd report him to
| (dg@tao-group.com) | abuse@heaven.org." --- David Bilek
+- www.cowlark.com --+

Attachment: pgpv5asc69GbN.pgp
Description: PGP signature