`reduce` should be an alias of `foldl1`, not `foldl`. about ramda HOT 5 CLOSED

i disagree. First, we don't have a fold, we have foldl and foldr. foldl folds the list from the left with the provided accumulator, as opposed to foldr which folds from the right. foldl is the same behavior as reduce. foldl without an initial accumulator, i.e. using the head of the list, is sometimes called foldl1, and it's my POV that it should not be in the library, because of the problem of deciding what the behavior is when you call foldl1 with an empty list. Since it is so easy to implement foldl1, I argue it is best to leave its problems out of the library.

from ramda.

I wasn't aware that "fold" was something other than assumed shorthand for "foldl".

It seems to me like reduce generally can work without an initial value, whereas fold always requires it.

I would expect a JavaScript library to throw an exception if it was asked to do something invalid like that, but I suppose it's possible to expect null or perhaps something else entirely, and can understand requiring the library user to define it themselves if they want it.

from ramda.

I would expect a JavaScript library to throw an exception if it was asked to do something invalid like that,

exactly--and since it is so easy to wind up in a situation where you pass an empty array to foldl1 inadvertently and then the app goes boom. or worse, you wrap it in a try-catch.

Returning undefined or null is not good either. The point is to compose these functions; which means we want functions that have--as much as possible in a JS environment--clearly defined inputs and outputs. If foldl1 is in some composition of functions that winds up passing an empty list, and it returns null, then the consuming function goes boom, or has to type-check its arguments.

That's why I think foldl1 is a non-starter in this environment.

from ramda.

we covered this before: https://github.com/CrossEye/ramda/issues/20

at that time we opted to remove foldl1 and foldr1--somehow they snuck back in. I've pulled them from the lib. Thanks for the input.

from ramda.

For further reference, there is a great list of how various languages handle this in Wikipedia's Fold Entry

There are a few that do it as Brian is suggesting, including some important ones like Scheme and Scala. There is really only Haskell that uses the foldl/foldl1 distinction. But the overwhelming majority use some sort of overloading to reuse the same name for the two scenarios.

We can't really do that in Ramda with our focus on curried functions. There's no good way to combine those notions with optional parameters, and that's the only version of overloading available in Javascript.

from ramda.