fold is bimap

[puffnfresh]

I got a question about fold, so I took a look at the signature:

@Task(α, β) => (α → γ), (β → γ) → Task(δ, γ)

This is not a fold, this is bimap.

[robotlolita]

Yeah, that was removed in the new version :)

I should update the documentation to note that here though.

[Avaq]

@puffnfresh it differs a little bit from bimap in that this fold always returns a "resolved" task. You could use the following to convert a Task a b to Task c (Either a b), for example:

task.fold(Left, Right) 

[Avaq]

Another way to view fold is (I believe) that it's similar to a proper fold, except that we cannot force the asynchronous operation to return synchronously, therefore we must return in a new Task.

[Avaq]

And for the sake of clarity, let's view the signatures side-by-side:

fold  :: Future a b ~> (a -> c, b -> c) -> Future d c
bimap :: Future a b ~> (a -> c, b -> d) -> Future c d

[puffnfresh]

@Avaq I missed that detail, thank you. It's still not fold, but it's also not bimap.

[Avaq]

It's still not fold

I know. :(

It's what I was referring to when I said similar to a proper fold. I'm open to rename the function in Fluture for the sake of correctness, but I do find that fold captures the idea of the function quite nicely.

If Fantasy Land were to adopt a spec that uses the fold name, I will definitely do the rename, just to avoid confusion.

[ivan-demchenko]

fold :: Future a b ~> (a -> c, b -> c) -> Future d c

if I'm allowed to ask an unrelated question, what is d in this signature? A unit, I suppose?

[robotlolita]

Folktale 2's task only has willMatchWith. No way of writing fold without deep coroutines/fibers or threads :(

@raqystyle an arbitrary type variable. There it just means it's not the same type as a, but it doesn't have any equivalent runtime set of values (it's not really unit or null). In that sense, it's closer to TypeScript's never.

[ivan-demchenko]

@robotlolita thank you for the explanation!