~ patterns

[Simon Marlow]

On 27 January 2006 10:50, John Hughes wrote:

> I'd be fairly horrified at the removal of ~ patterns. I've used them
> to fix very serious space-leaks that would have been awkward to fix
> in any other way.

You use them to *fix* space leaks?  Can you give an example?  I'm
genuinely surprised and interested - I don't think I've ever seen a case
of this before, but I've seen the opposite many times (lazy pattern
matching being the cause of space leaks).

The reasoning behind removing ~ patterns, for me, is that the number of
times they are used doesn't justify stealing the ~ symbol and the extra
complexity in the language design, when you can get the same effect with
let bindings.  With pattern guards there's even less justification for
keeping ~, eg.

  f (a, ~(x:xs)) = e

turns into
 
  f (a,b) | let (x:xs) = b  = e

you can translate any equation this way.  I would probably write the
second form in prefernce to the first, because the strict pattern
matches are clearly separated from the lazy.

I admit to being slightly biased though, in the type of Haskell
programming I do I have almost never needed to use ~.  Very occasionally
when doing cyclic programming, is all.

Cheers,
	Simon