On Mon, 2005-08-15 at 16:41 -0600, Matt Gushee wrote:
> Anu Engineer wrote:
> 
>  > Please forgive me if my question is very naïve, I am very new to 
> Ocaml. I was
> 
>  > wondering why snd returns an  error when I use it with more than 2 
> elements, why
> 
>  > not return the rest of the list when I apply to something larger than 
> a pair ?
> 
> 
> Your question shows your misunderstanding: snd operates on tuples, not 
> lists. 

I think the original question really meant:

Why aren't "fst" and "snd" properly generic??

For example this simply *should* work:

	snd (1,2,3)

and even pattern matching can't handle it:

	match e with (h,t,...) -> 

also should work. Felix actually supports this ugly generic
tuple projection operator:

	(1,2.1).(1)   // generic n'th of tuple
	(1,2.1,3).(1) // n must be compile time constant

and even:

var i:int;
for_each {i=0;} {i<3} {++i;} {
  print (1,2,3).[i]; // run time indexing
  endl;
};

because a tuple of all the same type is an array,
but as yet I haven't found a way to make ... work in 
pattern matches. 

It would actually be nice to have more general support
for polyadic tuple management: for example thinking about
obtaining a slice of a tuple, or concatenating two tuples,
suggests that just getting the n'th component is special case.

This kind of thing can actually be done in a hacky way
within limits with C++ template meta-programming.

A more general mapping from a list of types (at compile time)
representing tuples should be possible: any list operation
on types is a run-time operation on values. Unfortunately
I doubt camlp4 can handle that, since it is dealing with untyped
terms. Perhaps MetaOcaml can do it?

I would love to do this in Felix .. but it requires dynamic loading.

-- 
John Skaller <skaller at users dot sourceforge dot net>