ensures given iterable is interacted once per subscriber

[OlegDokuka]

close #3295

Signed-off-by: Oleh Dokuka odokuka@vmware.com

[chemicL]

I'd love to see a comment explaining why this is feasible, e.g. "If the Spliterator is finite and we know it has not been traversed, the estimate is accurate." - but this raises a question - should we check whether the Spliterator has not been traversed yet? Is there a way to do so?

[OlegDokuka]

I think it is irrelevant. At this stage, we just retrieved a new spliterator instance. So does it make sense to care about "check whether the Spliterator has not been traversed yet"?

Also, the code it selves is taken from the method Spliterator#getExactSizeIfKnown() so this avoids extra work since we knowToBeFinite

[chemicL]

Right, I failed to record the fact the spliterator was retrieved in the constructor. Makes me think though: what happens in case of -1?

[OlegDokuka]

it is not used explicitly. I tend to think that indeed we need to replace that with simplified code. e.g.

hasValues = !knownToBeFinite || it.estimateSize() != 0

[OlegDokuka]

using the above we can say: ok if it is unknown to be finite - we can do nothing, assume has values, if it is known to be finite and estimated size is not 0 then it has values

[OlegDokuka]

@chemicL good to go?

[chemicL]

Sounds good, thanks.

[OlegDokuka]

removed instanceof Collection check in checkFinite method

instanceof check brings significant overhead. All JMH benchmarks with observations are available at the following link https://gist.github.com/OlegDokuka/2a5dc81ba89d1f3b8f81153a78a4dd96

[reactorbot]

@OlegDokuka this PR seems to have been merged on a maintenance branch, please ensure the change is merge-forwarded to intermediate maintenance branches and up to main 🙇