From: MikkelFJ <mikkelfj-anti-spam@...>
Date: 2002-12-18T06:47:07+09:00
Subject: Re: Functional Language Recommendation


"Wai-Sun Chia" <waisun.chia@hp.com> wrote in message
news:3DFF3E53.1030305@hp.com...
> Sorry for the OT post, but I need some advise from some like-minded
> folks who appreciate the finer gems in life ;-)
>
> I'm planning to study a functional language in the comming year (new
> year's resolution, as I have zero experience/knowledge on functional
> languages) and have narrowed it down to either Ocaml and Haskell. It'd
> be great if any of you could share your experiences with either one or
> both of them.

If you are interested in the language as such, there are many choices. If
you are interested in practical applications with multiple OS support, it
narrows down a bit.

LISP and variants are dynamically typed eagerly evaluated languages - not
very portably between Windows / Unix, at least not without paying big bucks.
Clean, Haskell are lazily evaluated statically typed functional languages.
Clean works very well on Windows. Haskell works well on several platforms
and has shown good ability to interoperate, for instance with COM on
windows. However, if for no other reason, I can't live with a slow huge Perl
startup script as in the GHC compiler - it is officially tagged as research
project.
The ML family are eagerly evaluated statically typed functional languages.
They deduce type info so they are easy to type like LISP and Ruby even with
the type information.

Personally I have chosen OCaml because I need a language that is cross
platform and useful in real life. Few languages come close to this language
in implementation quality. Clean is also interesting, but is effectively
Windows only. I also considered eager versus lazy evaluation. I found the in
real life you mostly don't need that pretty fancy things you can do with
lazy evaluation - and when you do need it, it is available in OCaml via lazy
streams. Lazy evaluation makes it difficult to predict runtime efficiency
and often especially newcomers have difficulties figuring out why a two-line
program takes 15 minutes to execute and consume all memory. One of the most
useful features of lazy evalutions are combinator parsers - but they are
very difficult to get to perform efficiently, although the can work magic.
Lisp and variants (such as Dylan) are interesting, but you don't get any
cheap compiler (or, I guess, expensive for that matter) with same qualities
as the OCaml compiler. I would give Dylan a second look though - it may be
the definitive Ruby / Functional bridging language, dynamic and static
language as you desire.

In conclusion, there are really many good choices with good implementations
around - but if you seek a power language to replace C++, I'd say the choice
would be OCaml. With OCaml you can  build a 200K standalone executable or a
much smaller cross platform bytecode program that requires a reasonably
sized runtime dll. You can link with .lib, .c and .obj files as you like.

Mikkel