On 02/11/2006, at 9:56 PM, ajb@spamcop.net wrote:

G'day all.

Quoting Bernie Pope :

...

This is a weird example of a pattern binding, and it is surprising (to me) that the syntax is valid.

Maybe. But you wouldn't balk at this:

numzeroes xs = sum [ 1 | 0 <- xs ]

...even if you wouldn't naturally express it that way. Patterns like this are useful in places other than lets. I'd find it more surprising if lets were treated as a special case.

...

In this case you are defining the function which is bound to the variable called "+".

Though thanks to n+k patterns, this is well-understood to be a confusing and controversial case.

...

One would normally expect to see at least one variable inside the pattern on the left-hand-side of a pattern binding (otherwise what point would there be to the definition?).

If the pattern is demanded (as it may well be in a list comprehension generator, do block generator or a lambda), then it can be very useful. The programmer is asking for an error/empty list/monad "fail" if the conformality check fails.

I agree with what you say. When I said "pattern binding", I was using the terminology of the Haskell Report, which means only declarations of the form "pat = exp". Generator statements and lambdas are a different story, as you rightly point out. Nonetheless, I still believe that it is surprising for "pattern bindings" that this syntax is allowed, since there doesn't seem to be any purpose to them. (A cursory glance at the Report suggests that this kind of pattern binding is allowed). Then again, since the bindings are benign, maybe one could argue that it is better to have one less rule in the language definition? Then again, such a declaration could be the sign of a buggy program, and it might be better to give a compile time error. Cheers, Bernie.