Repository navigation
Getting rid of Data.Monoid.DecidablyEmpty #44
Description
There's a comment at the top of Data.Monoid.DecidablyEmpty which reads: -- TODO upstream somwhere else?
That "somewhere else" already exists: Data.Monoid.Null from monoid-subclasses.
I had a look at the instances it provides, and it is missing the following ones:
instance MonoidNull a => MonoidNull (Identity a)
instance MonoidNull a => MonoidNull (WrappedMonoid a)
instance (Ord a, Bounded a) => MonoidNull (Max a)
instance (Ord a, Bounded a) => MonoidNull (Min a)
instance MonoidNull (Proxy a)
instance MonoidNull a => MonoidNull (Const a b)
instance MonoidNull a => MonoidNull (Down a)
instance MonoidNull p => MonoidNull (Par 1 p)
instance MonoidNull (U1 p)
instance MonoidNull (f p) => MonoidNull (Rec1 f p)
instance MonoidNull (f p) => MonoidNull (M1 i c f p)
instance MonoidNull c => MonoidNull (K1 i c p)
instance (MonoidNull (f p), MonoidNull (g p)) => MonoidNull ((f :*: g) p)
instance (MonoidNull (f p), MonoidNull (g p)) => MonoidNull ((f :.: g) p)
instance (...) => MonoidNull (a, b, c, d, e)
-- Should go to dependent-map
instance GCompare k => MonoidNull (DMap k v)This would solve #42 but is too much work to hold up GHC 9.2 support. Tasks, as I see them:
- Add the missing instances to
monoid-subclassesand release - Add the missing instance to
dependent-mapand release - Replace
Data.Monoid.DecidablyEmpty.DecidablyEmptywith type alias, setisEmpty = Data.Monoid.Null.null, deprecate both, and release - Remove deprecated
Data.Monoid.DecidablyEmpty.
None of this seems particularly controversial. We might want monoid-subclasses anyway because of #37. It's quite a light package, so depending on it from dependent-map shouldn't be a problem either.
Activity
endgame commentedon May 21, 2022
@Ericson2314 Let me know if you're happy with this plan, and I will begin making PRs.
Ericson2314 commentedon Sep 15, 2022
@endgame this one I have more mixed feelings about because
monoid-subclassesis such overkill for our use-case.Ericson2314 commentedon Sep 15, 2022
Also, if we we had more "non-empty" data types, like haskell/containers#616, we could do this in a more "type directed" way which would be more satisfying.
(If you have spare cycles, I think they would be more usefully spent getting haskell/containers#616 over the finish line. I hope the conflicts are not too bad! The only thing remaining before that I remember was dealing with some perf regressions, which I think is just a matter of tweaking the inlining.)
endgame commentedon Jan 10, 2024
It's not that big a package, and its library component has a pretty reasonable dependency footprint. Regardless, I opened blamario/monoid-subclasses#50 .
ryantrinkle commentedon Sep 26, 2024
@endgame This is awesome, thank you!
endgame commentedon Sep 29, 2024
Since I got the github notification, I thought I'd do
dependent-mapwhile I had a couple of minutes.