Introduce "fold" as an alternative to "reduce"

[atilaneves]

I took the easy way out here; I have no idea why reduce throws when passed an empty range and I don't think fold should do it either. However, since it's a lot harder to make fold nothrow I'll leave it as it is right now and maybe submit a PR later to make it so. IMHO, this should compile and pass:

@safe pure nothrow unittest
{
    int[] arr;
    assert(arr.fold!((a, b) => a + b) == 0);
    assert(arr.fold!(min, max) == tuple(0, 0));
}

i.e. return Unqual!(ElementType!R).init when passed in an empty range. It gets more complicated since reduce also supports opApply.

[MetaLang]

I have no idea why reduce throws when passed an empty range

I believe this is just historical baggage. Today it would be written as assert(!range.empty) or just left up to the range to assert when front/popFront are called and empty == true.

[dcarp]

Don't break the lexical order.

[dcarp]

Changelog entry is missing.

[dcarp]

For a smoother deprecation path, it will be better to use fold in the original implementation and create reduce as a wrapper of fold.
Initially the unit tests can still use reduce. This will assure that the wrapper implementation works in all the tested cases.

[dcarp]

The reference in std.algorithm.package.d's ddoc is missing.

[atilaneves]

I don't think it's a good idea to use fold as the "real" implementation. It's a lot more work for little to no benefit, and contrary to what was discussed here.

[quickfur]

What about adding an example of a UFCS chain, which is the primary motivation for introducing this function over reduce in the first place?

[quickfur]

Looks like a rebase is in order.

[atilaneves]

Added UFCS example and rebased.

[quickfur]

Nitpick: insert space after +, i.e., + 1 instead of +1.

[quickfur]

Also, please add fold to the ddoc index in std/algorithm/package.d.

[wilzbach]

Nice! I came by this issue from #1955 and thought this will never make it into phobos. Thanks @atilaneves!

I would recommend to add some comments to reduce about fold - newbies like I am are otherwise confused how to call reduce in UCFS syntax with seeds ;-)

[atilaneves]

I'm on holiday right now, will address the most recent comments when I get back home

[atilaneves]

Comments addressed.

[quickfur]

Thanks! LGTM.

[quickfur]

ping random reviewers @JakobOvrum @andralex @yebblies @klickverbot
This has been green for a while. Any other comments that need to be addressed, or can we merge this now?

[andralex]

Would be nice to use backticks instead of $(D ...) throughout (not a blocker).

[andralex]

This looks good. Will pull as is but please there is one follow-up task for @atilaneves: please make one more pass through all documentation and replace all uses of reduce with fold. Also deemphasize reduce to send people to fold. Essentially, even though fold is implemented in terms of reduce, the documentation and examples should go the other way around. Thanks!

[andralex]

Auto-merge toggled on