Proposal: Remove Semigroup and Monoid instances for Data.Map, Data.IntMap, Data.HashMap

Bardur Arantsson spam at scientician.net
Thu Feb 15 21:13:57 UTC 2018


[ blahblah :) ]

+1 for the original suggestion.

+2 for Kris' newtype-only suggestion -- especially if we can actually
prevent orphan instances effectively. (I think I got that right, it's a
big thread at this point.). IMO, it's fine to force people to at least
*consider* which instance they want absent a "best" answer. I do think
it's important that we are actually able to effectively prevent people
from just declaring orphans willy-nilly.

+0 for _hvr's dual-incompatible releases. It would be a +1, but my worry
here is (as usual) the worry about *forcing* the whole ecosystem to
upgrade all at once. Usually this is not actually a big deal, but
"containers" is used by almost *everything*, so it's effectively a
"split" until everything upgrades. (Obviously, things are not quite are
dramatic as that since, presumably, relatively few packages actually use
the instance in question, but...)

Cheers,




More information about the Libraries mailing list