Not sure if you saw this, Greg, but Alex ran into a similar question recently from Solr. https://lists.apache.org/thread/1gs3nsv1mcns1czdtdnqyz84f31tqm2x
On Mon, Oct 9, 2023 at 10:47 AM Adrien Grand <jpou...@gmail.com> wrote: > Hi Greg, > > I agree that LeafCollector implementations should be able to assume that > finish() only gets called once. The test framework already makes this > assumption: > https://github.com/apache/lucene/blob/dfff1e635805ffc61dd6029a8060e2635bfcbdb9/lucene/test-framework/src/java/org/apache/lucene/tests/search/AssertingLeafCollector.java#L95-L100 > . > > On Mon, Oct 9, 2023 at 5:38 PM Greg Miller <gsmil...@gmail.com> wrote: > >> Hey folks- >> >> I'm curious if anyone has thoughts around idempotency concerns related to >> the LeafCollector#finish API added in GH#12380 >> <https://github.com/apache/lucene/pull/12380>. My expectation would be >> that LeafCollector implementations should be able to assume #finish will >> only get called once. In fact, it looks like FacetsCollector is already >> making that assumption. >> >> Is this inline with other folks' expectations? If so, I'm going to, 1) >> address a small bug related to drill-sideways that results #finish being >> called multiple times on one of the collectors, and 2) propose some >> additional javadoc on LeafCollector#finish clarifying this. >> >> Make sense? >> >> Cheers, >> -Greg >> > > > -- > Adrien >
