View source for Composing functions with multiple values

== Question ==

I see a certain imbalance between multiple function arguments
and multiple function values.
On the one hand
functions with multiple arguments are often in [[Currying|curried]] form,
like
<haskell>
(++) :: [a] -> [a] -> [a]
</haskell>
which allows for [[partial application]].
On the other hand,
functions cannot emit values in a compatible way,
and thus must return pairs instead.
Example:
<haskell>
partition :: (a -> Bool) -> [a] -> ([a], [a])
</haskell>

How can I convert the result of <hask>partition</hask>
such that it meets the needs of <hask>(++)</hask>?

== Answer ==

You can't convert a pair to something that fits
a function that accepts two arguments in curried form.
But you can convert the curried function to one that accepts pairs using the standard function <hask>uncurry</hask>!

It is
<haskell>
uncurry (++) :: ([a], [a]) -> [a]
</haskell>
and consequently
<haskell>
uncurry (++) . partition p :: [a] -> [a]
</haskell>
is what you are after.

In fact, there is no imbalance.
Functions always accept exactly one input
and emit exactly one output.
You can always choose tuples as both input and output.
Using tuples as input is close to the way
other languages handle multiple inputs.
If you choose the curried form
you do that because of partial application.
In this case you should also think intensively about
how to [[Parameter order|order]] the parameters
in order to make partial application useful.

[[Category:FAQ]]
[[Category:Idioms]]