I'd like to revive this thread now that these new optimizations for collection of top docs are more usable and enabled by default in IndexSearcher (https://issues.apache.org/jira/browse/LUCENE-8060). Any feedback about starting to work towards releasing 8.0 and targeting October 2018?
Le jeu. 21 juin 2018 à 09:31, Adrien Grand <[email protected]> a écrit : > Hi Robert, > > I agree we need to make it more usable before 8.0. I would also like to > improve ReqOptSumScorer (https://issues.apache.org/jira/browse/LUCENE-8204) > to leverage impacts so that queries that incorporate queries on feature > fields (https://issues.apache.org/jira/browse/LUCENE-8197) in an optional > clause are also fast. > > Le jeu. 21 juin 2018 à 03:06, Robert Muir <[email protected]> a écrit : > >> How can the end user actually use the biggest new feature: impacts and >> BMW? As far as I can tell, the issue to actually implement the >> necessary API changes (IndexSearcher/TopDocs/etc) is still open and >> unresolved, although there are some interesting ideas on it. This >> seems like a really big missing piece, without a proper API, the stuff >> is not really usable. I also can't imagine a situation where the API >> could be introduced in a followup minor release because it would be >> too invasive. >> >> On Mon, Jun 18, 2018 at 1:19 PM, Adrien Grand <[email protected]> wrote: >> > Hi all, >> > >> > I would like to start discussing releasing Lucene/Solr 8.0. Lucene 8 >> already >> > has some good changes around scoring, notably cleanups to >> > similarities[1][2][3], indexing of impacts[4], and an implementation of >> > Block-Max WAND[5] which, once combined, allow to run queries faster when >> > total hit counts are not requested. >> > >> > [1] https://issues.apache.org/jira/browse/LUCENE-8116 >> > [2] https://issues.apache.org/jira/browse/LUCENE-8020 >> > [3] https://issues.apache.org/jira/browse/LUCENE-8007 >> > [4] https://issues.apache.org/jira/browse/LUCENE-4198 >> > [5] https://issues.apache.org/jira/browse/LUCENE-8135 >> > >> > In terms of bug fixes, there is also a bad relevancy bug[6] which is >> only in >> > 8.0 because it required a breaking change[7] to be implemented. >> > >> > [6] https://issues.apache.org/jira/browse/LUCENE-8031 >> > [7] https://issues.apache.org/jira/browse/LUCENE-8134 >> > >> > As usual, doing a new major release will also help age out old codecs, >> which >> > in-turn make maintenance easier: 8.0 will no longer need to care about >> the >> > fact that some codecs were initially implemented with a random-access >> API >> > for doc values, that pre-7.0 indices encoded norms differently, or that >> > pre-6.2 indices could not record an index sort. >> > >> > I also expect that we will come up with ideas of things to do for 8.0 >> as we >> > feel that the next major is getting closer. In terms of planning, I was >> > thinking that we could target something like october 2018, which would >> be >> > 12-13 months after 7.0 and 3-4 months from now. >> > >> > From a Solr perspective, the main change I'm aware of that would be >> worth >> > releasing a new major is the Star Burst effort. Is it something we want >> to >> > get in for 8.0? >> > >> > Adrien >> >> --------------------------------------------------------------------- >> To unsubscribe, e-mail: [email protected] >> For additional commands, e-mail: [email protected] >> >>
