proposed extenion to patterns

proposed extenion to patterns

[skaller]

Pattern matches in Ocaml now allow alternatives within 
a branch subject to a restriction that all alternatives
supply the same set of pattern variables (and of the same type).

This is useful but still quite restrictive, for example:

| A (i,j) | B i with j = 1 -> i + j

cannot be coded, because there's no way to introduce
the variable j.

However this isn't really general enough either,
because it can't HIDE variables!

This form looks more general:

| in A (h,k) let i = h and k = k
| in B s let i = s + 1 let k = i

These forms are inverted let/in constructions,
with all the equivalent options (and, rec, and sequence).

The variables h,k in the first alternative are hidden.

The implementation is roughly syntactic sugar:

match x with A (h,k) -> let i = h and k = k in i,k

The 'with' form is a simplification, as is the
usual form. 

Any comments on this idea?


-- 
John Skaller <skaller at users dot sf dot net>
Felix, successor to C++: http://felix.sf.net

Re: [Caml-list] proposed extenion to patterns

[Tom]

[-- Attachment #1: Type: text/plain, Size: 902 bytes --]

On 09/03/07, skaller <skaller@users.sourceforge.net> wrote:
>
> Pattern matches in Ocaml now allow alternatives within
> a branch subject to a restriction that all alternatives
> supply the same set of pattern variables (and of the same type).
>
> This is useful but still quite restrictive, for example:
>
> | A (i,j) | B i with j = 1 -> i + j


 I think this is a good idea.

This form looks more general:
>
> | in A (h,k) let i = h and k = k
> | in B s let i = s + 1 let k = i
>


But I find this too complicated (for a programmer to memorize the syntax,
write it and someone else read it). You could simply say:

  | A (h as i, k as k)
  | B s with i = s + 1 and k = i

I know that this would introduce some problems (not equal variable in both
branches, but this could easily be solved (the compiler only complaining
when a non-common variable is actually used (in this example, h and s)).

- Tom

[-- Attachment #2: Type: text/html, Size: 1408 bytes --]

Re: [Caml-list] proposed extenion to patterns

[Jon Harrop]

On Friday 09 March 2007 02:02, skaller wrote:
> Any comments on this idea?

Every now and again I want that, but I think there are much bigger gains to be 
had elsewhere and I'd rather those were addressed first.

-- 
Dr Jon D Harrop, Flying Frog Consultancy Ltd.
OCaml for Scientists
http://www.ffconsultancy.com/products/ocaml_for_scientists

Re: [Caml-list] proposed extenion to patterns

[skaller]

On Fri, 2007-03-09 at 10:08 +0000, Jon Harrop wrote:
> On Friday 09 March 2007 02:02, skaller wrote:
> > Any comments on this idea?
> 
> Every now and again I want that, but I think there are much bigger gains to be 
> had elsewhere

Agreed.

>  and I'd rather those were addressed first.

Development in sequence of importance isn't the only model.

When working with Stroustrup and the C++ extensions subgroup,
on the (long list) of proposals, we'd vote.

Then we'd deal with the 3 proposals at the top of the list --
and the three at the bottom :)

-- 
John Skaller <skaller at users dot sf dot net>
Felix, successor to C++: http://felix.sf.net