Use Data.Bits.unsafeShift* instead of shift*?

[sjakobi]

Now that containers has base >= 4.6 it can unconditionally use unsafeShiftR and unsafeShiftL from Data.Bits which might be faster shift{R,L}.

I'm also wondering whether the many shifts by 1 should be replaced by calls to two helper functions double and half.

[AndreasPK]

Performance of shiftL/R might soon get worse than the unsafe variants. See https://gitlab.haskell.org/ghc/ghc/merge_requests/113

[sjakobi]

Related: #216

[sjakobi]

It appears that GHC optimizes

x `shift{R,L}` 1

to

x `unsafeShift{R,L}` 1

Nonetheless it would be clearer and safer to remove the usage of shift{R,L}. That is all in Data.Map.Internal, Data.Map.Strict.Internal and Data.Set.Internal.

How about we add two functions half and double to Data.Set.Internal and use them in Data.Map.* too @treeowl?

[sjakobi]

I stumbled across this issue again, and I would like to get it finally resolved.

How do you feel about my comment above, @treeowl?

[meooow25]

There are no more shiftL/shiftR usages after #1042.