I have merged and backported the connected components logic for the 10.2.0 release.
Thanks y'all! On Thu, Apr 3, 2025 at 8:50 AM Ignacio Vera <iver...@gmail.com> wrote: > I agree that the downsides are much bigger than the benefits. I have > no strong opinion between reverting the change or disabling it, maybe > for 10.2 is safer to just disable it. > > > > On Thu, Apr 3, 2025 at 2:28 PM Michael Sokolov <msoko...@gmail.com> wrote: > > > > It makes sense to me. I think it's providing marginal benefits, and the > downside is bad > > > > On Thu, Apr 3, 2025, 4:58 AM Benjamin Trent <ben.w.tr...@gmail.com> > wrote: > >> > >> Hey y'all, > >> > >> Unless there is strong dissenting opinion, I think we should revert the > connected components work in HNSW for 10.2 as a bug fix. > https://github.com/apache/lucene/pull/14411 > >> > >> We found that when "connectedComponents" is most needed (e.g. a very > disconnected graph), it takes an inordinate amount of compute and can > significantly stall indexing and merging. This is for vector distributions > that are likely not well supported for HNSW anyways. > >> > >> The discussion here indicates that reversion (or disabling), is the way > to go: https://github.com/apache/lucene/issues/14214 > >> > >> Here is my PR for disabling: > https://github.com/apache/lucene/pull/14436 > >> > >> I am not doing a full revert as the original changes adjusted the APIs > in v10 (adding exception declaration, etc.). > >> > >> Let me know what y'all think, > >> > >> Ben > >> > >> On Wed, Apr 2, 2025 at 5:59 PM Mayya Sharipova < > mayya.sharip...@elastic.co.invalid> wrote: > >>> > >>> Hello everyone, > >>> We have discovered that a change introduced in 10.2 doesn't work > nicely when two fields are indexed with the same analyzer. There is a > workaround, but for now I will revert this change for the 10.2 release to > further investigation. > >>> > >>> Sorry for the trouble. > >>> > >>> On Wed, Apr 2, 2025 at 4:37 AM Ignacio Vera <iver...@gmail.com> wrote: > >>>> > >>>> Hey Tommaso, > >>>> > >>>> I have no objections if you can get the PR merged today. > >>>> > >>>> Cheers, > >>>> > >>>> Ignacio > >>>> > >>>> > >>>> On Tue, Apr 1, 2025 at 8:14 PM Tommaso Teofili > >>>> <tommaso.teof...@gmail.com> wrote: > >>>> > > >>>> > I'd like to get https://github.com/apache/lucene/pull/14094 in by > tomorrow, unless anyone objects. > >>>> > > >>>> > > >>>> > On Mon, 31 Mar 2025 at 17:15, Robert Muir <rcm...@gmail.com> wrote: > >>>> >> > >>>> >> i added some bullets for automaton/regex improvements. > >>>> >> > >>>> >> On Sun, Mar 30, 2025 at 9:47 AM Adrien Grand <jpou...@gmail.com> > wrote: > >>>> >> > > >>>> >> > I would like to highlight the performance improvements that are > coming in 10.2 in a way that sounds a bit more exciting that our usual > bullet points so I started working on the release notes: > >>>> >> > > https://cwiki.apache.org/confluence/display/LUCENE/Releases+Notes+10.2.0 > >>>> >> > > >>>> >> > On Wed, Mar 26, 2025 at 1:37 PM Ignacio Vera <iver...@gmail.com> > wrote: > >>>> >> >> > >>>> >> >> Thanks Rob for the heads up. Let's try to improve those scripts > for posterity. > >>>> >> >> > >>>> >> >> On Tue, Mar 25, 2025 at 5:50 PM Robert Muir <rcm...@gmail.com> > wrote: > >>>> >> >> > > >>>> >> >> > +1 and I am happy to assist RM with the python scripts > situation > >>>> >> >> > (release wizard / smoke tester) > >>>> >> >> > > >>>> >> >> > I am 100% sure I caused some breakage in the scripts by > introducing > >>>> >> >> > the linting and type-checking/hints. But I am convinced if we > work > >>>> >> >> > through the problems (one-time pain), it will make it easier > to > >>>> >> >> > maintain these scripts down the road. > >>>> >> >> > > >>>> >> >> > On Tue, Mar 25, 2025 at 12:35 PM Adrien Grand < > jpou...@gmail.com> wrote: > >>>> >> >> > > > >>>> >> >> > > +1 let's get these huge performance improvements out > >>>> >> >> > > > >>>> >> >> > > Thank you Ignacio > >>>> >> >> > > > >>>> >> >> > > > >>>> >> >> > > Le mar. 25 mars 2025, 17:18, Ignacio Vera < > iver...@gmail.com> a écrit : > >>>> >> >> > >> > >>>> >> >> > >> Hi, > >>>> >> >> > >> > >>>> >> >> > >> Lucene 10.1.0 was released 3 months ago and improvements > are piling up > >>>> >> >> > >> so I would like to propose a lucene 10.2.0 release. I am > proposing to > >>>> >> >> > >> cut the branch on Monday next week and build the RC by the > end of the > >>>> >> >> > >> week. > >>>> >> >> > >> > >>>> >> >> > >> I am happy to be the release manager for 10.2.0 unless > anyone else > >>>> >> >> > >> wants to do it. > >>>> >> >> > >> > >>>> >> >> > >> Thanks, > >>>> >> >> > >> > >>>> >> >> > >> Ignacio > >>>> >> >> > >> > >>>> >> >> > >> > --------------------------------------------------------------------- > >>>> >> >> > >> To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org > >>>> >> >> > >> For additional commands, e-mail: > dev-h...@lucene.apache.org > >>>> >> >> > >> > >>>> >> >> > > >>>> >> >> > > --------------------------------------------------------------------- > >>>> >> >> > To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org > >>>> >> >> > For additional commands, e-mail: dev-h...@lucene.apache.org > >>>> >> >> > > >>>> >> >> > >>>> >> >> > --------------------------------------------------------------------- > >>>> >> >> To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org > >>>> >> >> For additional commands, e-mail: dev-h...@lucene.apache.org > >>>> >> >> > >>>> >> > > >>>> >> > > >>>> >> > -- > >>>> >> > Adrien > >>>> >> > >>>> >> > --------------------------------------------------------------------- > >>>> >> To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org > >>>> >> For additional commands, e-mail: dev-h...@lucene.apache.org > >>>> >> > >>>> > >>>> --------------------------------------------------------------------- > >>>> To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org > >>>> For additional commands, e-mail: dev-h...@lucene.apache.org > >>>> > > --------------------------------------------------------------------- > To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org > For additional commands, e-mail: dev-h...@lucene.apache.org > >
