-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Cale Gibbard wrote:
I agree regarding Maybe, but in the long run, I really think we should bring back MonadZero. This issue comes up all the time because there's not really a proper class for monads that do handle failure gracefully any more. I personally consider 'fail' to be a wart, and mostly try to ignore that it even exists. Monad really isn't the class for expressing computations which might fail.
agree about Monad fail.... but does everything that can fail have to be a monad? /me thinks about non-monadic applicative and arrows (at least)... there is a class ArrowZero even! There is a Monad=>MonadPlus, an Arrow=>ArrowZero=>ArrowPlus, an Applicative=>Alternative, a Monoid. (and the Monoid kind of has the same issue as the two kinds of MonadPlus instances, considering the proposal to change Data.Map's Monoid instance, perhaps) There is a MonadFix and a ArrowLoop, but no ApplicativeFix. The arrows are somewhat special cases since they involve different-looking types. (I wonder how much things are going to change when we start using associated-types... after some years...) This feels like somewhat of an ad-hoc mess at the moment :-/ for now, I wholeheartedly support the Maybe version of dropPrefix for the purpose of going into the existing Data.List Isaac -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.6 (GNU/Linux) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org iD8DBQFGhXn3HgcxvIWYTTURAorzAJ9GlcLCk2P0/NHrNddKeQivSU3vZQCgybiq FTfTp6zdJjJfqPFPbH5O1A0= =EcsX -----END PGP SIGNATURE-----