Sven Panne writes:
> GHC-2.10's "overlapping pattern matches"-warnings are sometimes quite puzzling:
>
> ------------------------
> module Foo where
>
> bar "-" = 1
> bar ('-':xs) = 2
> bar _ = 3
>
> baz ('-':[]) = 1
> baz ('-':xs) = 2
> baz _ = 3
> ------------------------
>
> Transcript:
>
> panne@liesl:/tmp > ghc -c Foo.hs
> Foo.hs:3:
> Warning: Pattern match(es) are overlapped
> in the definition of function `bar'
> `('-' : xs) = ...'
> ghc: module version changed to 1; reason: no old .hi file
>
>
> Why is there a warning for bar but none for baz?
>
Hi,
this relates to Alex' query on pattern matching on strings and how GHC
handles them. String literals aren't in general expanded into list
form pre-checking for overlapped/exhaustive patterns; maybe they
should..
>
> BTW, what do the "discarding polymorphic case"-warnings mean and
> under what circumstances do they occur? I think this has been asked
> before, but I could not find the answer in my mail folders...
Here's my take on it: The code generator isn't currently capable of
dealing with expressions of the form
case (expr::a) of
x -> ...
where `a' is an abstract type (i.e., data constructors not visible.)
Why? Because there's no way of knowing how `expr' is going to return
its result (in the heap or via registers.)
--Sigbjorn
