On Mon, May 2, 2011 at 11:14 PM, Antoine Latter
Part of the reason for the suggestion is that bind is "more powerful" than fmap, which means that it asks more of the implementing type, which then means that fewer type satisfy bind than satisfy fmap. So your function is more general when written using fmap.
And depending on the data type, 'f <$> m' may be more efficient than 'm >>= return . f'
I might write this with the infix version of fmap, but some folks don't like the explosion of operators:
((:) ds) <$> applyOp op x y
Note that this should be "(: ds) <$> applyOp op x y". I mean, (:) ds === (ds :) flip (:) ds === (\x -> x : ds) === (: ds)
or closer to yours:
flip (:) ds <$> applyOp op x y
If the type of the function is already fixed I think it comes down to personal preference.
Applicative style is very nice, specially when your functions have multiple arguments. Consider op1 >>= \x1 -> op2 >>= \x2 -> return (f x1 x2) do {x1 <- op1 x2 <- op2 return (f x1 x2)} against f <$> op1 <*> op2 Cheers, -- Felipe.