Re: [Ocaml-lib-devel] Future and status of extlib?

[Brian H. <bri...@ql...>]

On 28 Jun 2003, Scott A Crosby wrote:

> On Fri, 27 Jun 2003 11:39:51 +0900, "Nicolas Cannasse" <war...@fr...> writes:
> 
> > - in the beginning we asked for useful functions that should be added and we
> > stated case-by-case which one should be or not. This is still "open" if you
> > have idea.
> 
> IMHO, It should collect all useful standard library functionality,
> whether this includes enhancing functionality of existing STD-lib
> classes,
> 
> For instance, from
> http://pauillac.inria.fr/cdk/newdoc/htmlman/cdk_toc.html
> 
> Why not grab the weak sets, weak arrays, weak hash tables?

I dislike the idea of "stealing" code from other projects, even when the 
license allows it.  First off, it's possible that doing so would create 
hard feelings where none need to be.  Second of it, it creates a 
maintainance problem, as we have to track another version of the code and 
import changes as needed.  Plus the possibility of divergent build 
systems, etc.

We should, of course, welcome any code anyone wants to contribute.  And I 
wouldn't be opposed to possibly "merging" projects if enough overlap 
exists.

> 
> There's also the entire 'The Extra Library', to use that as a starting
> point? Bit vectors, String2, Trie, etc.

I have some code kicking around for bit vectors.  A couple of things I 
don't like about the code in the state it's in (which is why it isn't 
being shared):
1) It's in C.
2) It's not optimized for any platform.
2) It doesn't interface with any bignum package.

Ideally, I'd like to see GMP add support for bit vector operations, in 
addition to their excellent bignum support, and then provide an ocaml 
interface to GMP.  Best of both worlds.

> Would people be against attempting to import 'The Extra Library'
> pretty much wholesale?

If the authors/maintainers of that project don't mind, I don't mind.

> > We actually need a Set in a defonctorized style , and with a signature that
> > looks more like Hashtbl ( with a default "empty" constructor that is using
> > compare).
> 
> Why?
>

Ease of use.  Functors are OK if you're doing a lot with a particular type 
of set, but using a set in a single function is clumsy.
 
Brian