[ https://issues.apache.org/jira/browse/LUCENE-2130?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12787269#action_12787269 ]
Robert Muir commented on LUCENE-2130: ------------------------------------- I like the idea of fixing this problem somehow, here is the background: I noticed one very bad-performing wildcard query on an index with numbers 0000000-9999999 was the query "*NNNNNN" where N's are random numbers. the problem is that the query will enumerate the entire term dict but only return 10 results. in Constant Auto mode, this means it will use DirectoryReader's MultiTerms, even on an optimized index, with a lot of wasted priority queue usage, etc. This is a corner case in my tests, but the reason I think it would be nice to fix is because things like contrib/regex behave in this way maybe more often, they frequently scan the entire term dictionary only to return a few results. > Investigate Rewriting Constant Scoring MultiTermQueries per segment > ------------------------------------------------------------------- > > Key: LUCENE-2130 > URL: https://issues.apache.org/jira/browse/LUCENE-2130 > Project: Lucene - Java > Issue Type: Improvement > Reporter: Mark Miller > Priority: Minor > Attachments: LUCENE-2130.patch > > > This issue is likely not to go anywhere, but I thought we might explore it. > The only idea I have come up with is fairly ugly, and unless something better > comes up, this is not likely to happen. > But if we could rewrite constant score multi-term queries per segment, MTQ's > with auto, constant, or constant boolean rewrite could enum terms against a > single segment and then apply a boolean query against each segment with just > the terms that are known to be in that segment. This way, if you have a bunch > of really large segments and a lot of really small segments, you wouldn't > apply a huge booleanquery against all of the small segments which don't have > those terms anyway. How advantageous this is, I'm not sure yet. > No biggie, not likely, but what the heck. > So the ugly way to do it is to add a property to query's and weights - > lateCnstRewrite or something, that defaults to false. MTQ would return true > if its in a constant score mode. On the top level rewrite, if this is > detected, an empty ConstantScoreQuery is made, and its Weight is turned to > lateCnstRewrite and it keeps a ref to the original MTQ query. It also gets > its boost set to the MTQ's boost. Then when we are searching per segment, if > the Weight is lateCnstRewrite, we grab the orig query and actually do the > rewrite against the subreader and grab the actual constantscore weight. It > works I think - but its a little ugly. -- This message is automatically generated by JIRA. - You can reply to this email to add a comment to the issue online. --------------------------------------------------------------------- To unsubscribe, e-mail: java-dev-unsubscr...@lucene.apache.org For additional commands, e-mail: java-dev-h...@lucene.apache.org