I firmly believe that making pattern matches in let and where clauses lazy by default was a mistake in the Haskell Report. It's inconsistent with how pattern matching works elsewhere in the language, and also makes a strange distinction between outer and inner pattern matches. Unfortunately, it's way too late to fix that mistake.

On Feb 24, 2017 12:09 AM, "Harendra Kumar" <harendra.kumar@gmail.com> wrote:

CCing the list. I guess you intended to cc but forgot.

On 24 February 2017 at 09:27, <ok@cs.otago.ac.nz> wrote:
In Erlang, the equivalent of a let fails.
1> 1=2.
** exception error: no match of right hand side value 2

In SML, the equivalent of a let fails.
- val 1 = 1;
- val 1 = 2;

uncaught exception Bind [nonexhaustive binding failure]
raised at: stdIn:2.5-2.10

The problem is not that let 1 = 2 ... is *legal* but that
- the compiler is *silent* about it
- the runtime is *silent* about it.
Compiling the little program

main = let 1 = 2 in print "hi"

I expected that the compiler would be silent but that
there would be some sort of "matched failed" error at
run time. Silly me.

The thing is, it is not just bindings that bind no variables
that act as if they were not there.

main = let [x] = [1,2] in print "hi"

also compiles silently and runs without error. Change it to

main = let [x] = [1,2] in print ("hi" ++ show x)

and you get a runtime error

<object>: <source>: Irrefutable pattern failed for pattern [x].

I wish the compiler would report an error something like

"<location>: possibly failing match deleted
because it binds no live variables"

_______________________________________________
Haskell-Cafe mailing list
To (un)subscribe, modify options or view archives go to:
http://mail.haskell.org/cgi-bin/mailman/listinfo/haskell-cafe
Only members subscribed via the mailman list are allowed to post.