Draft: Give a better implementation for (*>)

[L0neGamer]

~~The default implementation of (*>) uses (<*>) which tries to perform each branch in parallel. However, since we only get the value from the right argument, we should just return that argument.

I don't implement (>>) similarly due to my attempts with haskell/core-libraries-committee#333, but if that proposal falls through I'll be back.~~

Simply cleanup some instance definitions and add a comment.

[ocharles]

This doesn't look sound - what if I'm using GenT State? You're not going to run the left potentially-stateful computation at all!

[L0neGamer]

Damn, you've a point... I'll have another go

[L0neGamer]

@ocharles if you don't mind, why is mzip and (<*>) different for TreeT?

I'm not sure why NodeT's applicative wasn't implemented like below. Could you shed some light?

diff --git a/hedgehog/src/Hedgehog/Internal/Tree.hs b/hedgehog/src/Hedgehog/Internal/Tree.hs
index 0970ec1..d95aa62 100644
--- a/hedgehog/src/Hedgehog/Internal/Tree.hs
+++ b/hedgehog/src/Hedgehog/Internal/Tree.hs
@@ -428,9 +428,9 @@ instance (Eq1 m, Eq a) => Eq (TreeT m a) where
 instance Applicative m => Applicative (NodeT m) where
   pure x =
     NodeT x []
-  (<*>) (NodeT ab tabs) na@(NodeT a tas) =
-    NodeT (ab a) $
-      map (<*> (fromNodeT na)) tabs ++ map (fmap ab) tas
+  (<*>) nf@(NodeT f tfs) na@(NodeT a tas) =
+    NodeT (f a) $
+      map (<*> fromNodeT na) tfs ++ map (fromNodeT nf <*>) tas
 
 instance Applicative m => Applicative (TreeT m) where
   pure =

[ocharles]

Sorry, I don't know why they are different - I'm not familiar with the underlying code.

[moodmosaic]

I think this makes (*>) behavior consistent with do notation, where you'd expect sequential execution.

[L0neGamer]

I've changed this to instead just cleanup some instances and add a comment. I can adjust the changelog if wanted, but this seems minor.