(<*>) apply vs. ($) apply

Chat
#1

The multiple definitions of apply can be pretty confusing for beginners (at least it was for me), and now it’s a bit awkward when referring to apply in documentation, since I feel like additional clarification is often needed.

Control.Apply
apply :: forall f a b. Apply f => f (a -> b) -> f a -> f b
<*>

Data.Function
apply :: forall a b. (a -> b) -> a -> b
$

It also seems like it would be easier to remember the infix notation if <*> was swapped with map's <$>, although that would be a departure from Haskell. I guess Haskell doesn’t have this issue with overlapping function names, since it just defines these infix operators without a named function.

One way to add clarity is to rename $'s apply to something like noParens. It seems like only the infix operator is used (and exported by Prelude), so changing the underlying function name wouldn’t be too annoying of a breaking change.

1 Like
#2

I think it’s hard to justify a breaking change like this. I’d recommend referring to $ using the symbol rather than the name apply in documentation to avoid this confusion; I’m not sure I’ve ever seen people referring to $ using the name “apply”.

#3

…but by the same token, if no-one is using the underlying function name for ($) it’s not that bad a breaking change to consider at some point.

I can’t imagine why or how anyone ever would use the function by name, has it ever been seen in the wild?

#4

It’s in Prelude, so you have to be pessimistic. There probably are cases of people referring to it by name in code; I just haven’t seen it in any documentation or learning resources.