[ https://issues.apache.org/jira/browse/LUCENE-8113?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16307943#comment-16307943 ]
Robert Muir commented on LUCENE-8113: ------------------------------------- Can we fold this into TermContext directly rather than subclassing? I think this class is already too complicated for any performance benefits it brings us. If we add subclassing it sets it over the edge a bit. Can we also try to think of a better name for it? I really like this change in the patch: {noformat} - public TermState get(int ord) { + public TermState get(LeafReaderContext ctx) throws IOException { {noformat} But because both things are named *Context (which to me is a meaningless name), it leads to confusing looking code such as: {noformat} TermState termState = termContext.get(context); {noformat} Also there are quite a few changes to how ScoreModes are passed during query construction in the patch. Looks correct, but is that supposed to be here? > Allow terms dictionary lookups to be lazy when scores are not needed > -------------------------------------------------------------------- > > Key: LUCENE-8113 > URL: https://issues.apache.org/jira/browse/LUCENE-8113 > Project: Lucene - Core > Issue Type: Improvement > Reporter: Alan Woodward > Attachments: LUCENE-8113.patch > > > LUCENE-7311 made it possible to avoid loading TermStates in cached > TermQueries. It would be useful to extend this to other queries that use the > terms dictionary. -- This message was sent by Atlassian JIRA (v6.4.14#64029) --------------------------------------------------------------------- To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org For additional commands, e-mail: dev-h...@lucene.apache.org