Add Enumerable#sole

[kivikakk]

This adds the Enumerable sole definition moreorless exactly per @matthewd's comment at #26206 (comment). I've left the block default out for simplicity's sake, but we could add it if we feel the utility's there. If we do, it needs to be added to #40768 as well.

This is an add-on to #40768 "Add ActiveRecord::FinderMethods#sole and #find_sole_by"; this will slide in underneath the FinderMethod's #sole. It seems a bit ugly that the exceptions raised by the method will change from Enumerable::SoleItemExpectedError to ActiveRecord::SoleRecordExceeded, but it also seems most correct/least incorrect.

/cc @dhh @matthewd

[rafaelfranca]

Can't we use the same technique of first(2)?

[kivikakk]

I've made this change, but realised it doesn't let us distinguish these cases:

irb(main):001:0> found, undesired = [:a].first(2)
=> [:a]
irb(main):002:0> found
=> :a
irb(main):003:0> undesired
=> nil
irb(main):004:0> found, undesired = [:a, nil].first(2)
=> [:a, nil]
irb(main):005:0> found
=> :a
irb(main):006:0> undesired
=> nil
irb(main):007:0> 

As it stands, sole will now not raise if an Enumerable yields two items, the second of which is nil.

The original change per #26206 (comment) does distinguish this case, so I think I prefer it.

[kivikakk]

I've reverted the change of algorithm in 0c58f6f817.

[dhh]

That strikes me as more confusing, to be honest. That an extra element that happens to be nil isn't considered an element. I'd rather we treat nil as an element in this evaluation. So I think we should revert to the first(2) approach.

[kivikakk]

I'd rather we treat nil as an element in this evaluation. So I think we should revert to the first(2) approach.

I'm a bit confused by this comment.

I've reverted to the first(2) approach as requested, but — per the example irb session above — a second item which happens to be nil is not distinguished from no second item at all, as x, y = first(2) will put a nil into the y in either case. As a result, [:a, nil].sole == :a.

The alternative implementation (kivikakk@a7c90384c77) causes a SoleItemExpectedError in this case. I think that's more correct and less likely to cause unexpected errors, but happy to defer to you here.

[dhh]

Let's make a point to do show that rebelling against Rubocop is possible ✊. So please do add that per-line exemption, and let's put the return back in there. It's this level of finesse and subtlety that brings charm and joy to writing Ruby 😄❤️

[kivikakk]

Awesome. Done!

[kamipo]

Can we use size instead of count?
Unlike size, count on a relation will do a COUNT query even if a relation is loaded.

[dhh]

AR already has its own implementation of #sole. This is for other Enumerables.

[kamipo]

Ah I didn't know that (#40768), thank you for letting me know!

[rails-bot]

This pull request has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs.
Thank you for your contributions.

[kivikakk]

Rebased on latest main.

[dhh]

👏

[kivikakk]

@dhh Thanks for your time and care in reviewing, I really appreciate it! ❤️