array: generalize S.head, S.last, S.tail, and S.init

[syves]

Closes #570

Initial attempt to generalize head, tail, last and init functions for Foldables.

[davidchambers]

Good progress, @syves! I will wait for you to address the outstanding issues before I comment on your changes.

[syves]

I've addressed the above changes, but for some reason doctest is not running.

20 error code ELIFECYCLE
21 error errno 1
22 error [email protected] doctest: sanctuary-doctest
22 error Exit status 1
23 error Failed at the [email protected] doctest script.

[davidchambers]

Something doesn't feel right about using Z.map in conjunction with the isNothing check. If we're to keep the isNothing check we could safely use maybe.value in the Just case:

return maybe.isNothing ?
       Just (Pair (Z.empty (foldable.constructor)) (x)) :
       Just (Pair (Z.append (maybe.value.snd, maybe.value.fst)) (x));

This could be improved by removing the common code from the branches:

return Just (maybe.isNothing ?
             Pair (Z.empty (foldable.constructor)) (x) :
             Pair (Z.append (maybe.value.snd, maybe.value.fst)) (x));

return Just (Pair (maybe.isNothing ?
                   Z.empty (foldable.constructor) :
                   Z.append (maybe.value.snd, maybe.value.fst))
                  (x));

We could even remove the explicit conditional logic by using Z.reduce:

return Just (Pair (Z.reduce (function(_, pair) { return Z.append (pair.snd, pair.fst); },
                             Z.empty (foldable.constructor),
                             maybe))
                  (x));

The problem with this approach is that Z.empty (foldable.constructor) would be evaluated once for each element of the structure. Can you think of a solution to this problem?

[davidchambers]

The comment above applies to init rather than to tail. I'm sorry for causing confusion.

[syves]

I stored the empty in a var.

[davidchambers]

Let's complete this pull request together. I will push some commits to your branch. :)

Issues have been addressed. :)

[davidchambers]

I just added fast paths for arrays, so these functions will continue to operate efficiently on arrays. @syves, if you're happy with these changes please squash the commits in preparation for merging. :)

[davidchambers]

Thank you, @syves!