Non-empty nondeterminism effect #198
Conversation
| mfix f = ChooseC $ \ fork leaf -> | ||
| mfix (runChoose (liftA2 Fork) (pure . Leaf) | ||
| . f . fromJust . fold (<|>) Just) | ||
| >>= fold fork leaf |
|
|
||
| -- | Nondeterministically choose between two computations. | ||
| choose :: (Carrier sig m, Member Choose sig) => m a -> m a -> m a | ||
| choose a b = send (Choose (bool b a)) |
There was a problem hiding this comment.
I didn’t want to bring in a dependency on semigroupoids (and all of the transitive dependencies it entails), or this might have been an Alt instance (tho the n² instance problem would apply to that anyway).
NB: this name conflicts with #201, but I think we might want to generalize it to a foldMapA function instead anyway.
| -- | Nondeterministically choose between two computations. | ||
| choose :: (Carrier sig m, Member Choose sig) => m a -> m a -> m a | ||
| choose a b = send (Choose (bool b a)) | ||
|
|
There was a problem hiding this comment.
We could consider defining functionality like optional against this interface, since it doesn’t require empty. Likewise, possibly something like Data.List.NonEmpty.some1.
|
Ready for review ✨ |
|
I’ve added |
|
Since this now exports identifiers that could clash with Prelude-defined names, should we indicate in the docs that users may want to import this module qualified? |
None of these names are exported by Prelude. |
|
@patrickt: any objection to me merging given the above? ☝️ |
This PR introduces a variant of
NonDetwhich lacksEmpty, and thus can only represent branching computations terminating successfully (or never terminating at all).