On Fri, Dec 30, 2011 at 1:24 PM, Daniel Fischer
<daniel.is.fisc...@googlemail.com> wrote:
> On Thursday 29 December 2011, 23:52:46, Omari Norman wrote:
>> [...]
>
> 'fail' doesn't properly belong in the Monad class, it was added for the
> purpose of dealing with pattern-match failures, but most monads can't do
> anything better than abort with an error in such cases.
> 'fail' is widely considered a wart.
I thought I'd add my own reason why I don't like fail.
Take these two functions, for example:
test :: Maybe Int
test = do
(Right v) <- Just (Left 1)
return v
test' :: Maybe Int
test' = do
let (Right v) = Left 1
return v
The first returns Nothing. The second crashes with a pattern match failure.
Why should a pattern failure cause a crash everywhere *except* a do
binding? It makes no sense. It violates the principle of least
surprise by behaving differently to every other occurrence of pattern
matching in the whole language.
As for custom failures, I'd recommend either Michael Snoyman's Failure
class or MonadPlus, which were both designed for this sort of thing.
But I'd stay away from using fail, since as Omari Norman said, it's a
wart.
> _______________________________________________
> Haskell-Cafe mailing list
> Haskell-Cafe@haskell.org
> http://www.haskell.org/mailman/listinfo/haskell-cafe
_______________________________________________
Haskell-Cafe mailing list
Haskell-Cafe@haskell.org
http://www.haskell.org/mailman/listinfo/haskell-cafe
