dict docs: document that ordering of keys/values/pairs match iterate

[cossio]

Fix #56841.

Currently the documentation states that keys(dict) and values(dict) iterate in the same order. But it is not stated whether this is the same order as that used by pairs(dict), or when looping, for (k,v) in dict.

This PR makes this guarantee explicit.

[LilithHafner]

In my interpretation, this guarantees that keys, values, and pairs iterate in the same order for all collections (not just Dicts).

I think this is a reasonable guarantee for this API.

[cossio]

I don't think the CI failure is related to this PR ?

[Seelengrab]

Do these need to be collected or is a check with any/all (consuming the iterators in the process) also good here? Would save on some allocations.

[cossio]

a check with any/all

How exactly?

[LilithHafner]

I don't think we should worry about optimizing allocations here; this code is run once with n=100 so it will be compile-time dominated anyway.

[Seelengrab]

How exactly?

Something like this:

@test all(zip(d, keys(d), pairs(d))) do (kv, k, kpair)
   first(kv) == k == first(kpair)
end

(might need some adjustment in the argument destructuring of the do block - not 100% sure this unpacks correctly)

I don't think we should worry about optimizing allocations here

I'm not worried about allocations, that would just be a sideeffect (of course, consuming fewer resources in CI is always nice to have). In my opinion, a all+zip version is just a bit cleaner and makes sure to only rely on iteration, and not incidental behavior of collect or other unrelated stuff :)

[cossio]

@Seelengrab That's a nice suggestion. I updated the tests.

[Seelengrab]

In case there's a bug that decouples keys/values/iterate/pairs, this will cause up to 300 individual failures, each showing up in the log. I think it'd be better for debugging to only have one failure through @test all(...), instead of three @test per key-value-pair in the dict.

[cossio]

Better now?