Switch from []-specific to Functor-general unzip?

[akhra]

Odd thing I just discovered: the default Prelude and RIO.Prelude both have

unzip :: [(a, b)] -> ([a], [b])

...but Data.List.NonEmpty provides

unzip :: Functor f => f (a, b) -> (f a, f b)

This feels like it just got left behind during FTP, are there any arguments against generalizing it?

[snoyberg]

Looking at the implementation, I have a feeling that the generalized unzip may have negative performance implications, and that's why it wasn't generalized in Prelude:

-- | The 'unzip' function is the inverse of the 'zip' function.
unzip :: Functor f => f (a,b) -> (f a, f b)
unzip xs = (fst <$> xs, snd <$> xs)

Before making this change in rio, I'd like to hear why it wasn't included as part of FTP. If it's just an oversight, then I'm OK with the change.

[akhra]

As a counterpoint, the specialized version is also exported from Data.List (and RIO.List) so it remains available. Granted, this doesn't negate the concern about a surprising performance cost -- even if we document it, how many people are going to check the docs before using unzip?