a >>= \p -> f <$> b -- 'free p' and 'free b' disjoint

-->
((\p -> f) <$> a) <*> b

Will there also be an optimisation for some sort of simple patterns? I.e., where we could rewrite this to:

liftA2 (\pa pb -> f ...) a b

I think I remember someone saying that the one-at-a-time application of <*> inhibits certain optimisations.

liftA2 is defined via one-at-a-time application and cannot be redefined because it is no method of Applicative. Do you remember more details?

I find (a << b) confusing. The intended semantics seem to be "effect a", then "effect b", return result of "a".

Sorry, I didn't know that << doesn't exist. I meant an operator with the meaning of <* .

Sebastian