[
https://issues.apache.org/jira/browse/LUCENE-3332?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13069565#comment-13069565
]
Olivier Favre commented on LUCENE-3332:
---------------------------------------
The advantage of the generating multiple PhraseQuerys, is that highlighting
outputs a single highlighted fragment, instead of multiple highlighted words,
one next to the other.
I think looking at WeightedSpanTermExtractor's handling of MultiPhraseQuery may
help, but considering the job of FieldQuery.expand(), it may need deep
modifications.
The cons of generating the combination is that it generates lots of other
PhraseQuerys (bad complexity class), and that they will all be processed by
expand() which itself is O(n^2).
If each term is doubled in the source MultiTermQuery (say once stemmed, once
intact), for a query having 5 words we will end up with n=32 generated
PhraseQuerys, thus making 1024 comparison inside expand().
Fortunately, this process can be done only a single time per query, no need to
do it all over again for each field or for each doc.
> FastVectorHighlighter ignores MultiPhraseQuery (and more)
> ---------------------------------------------------------
>
> Key: LUCENE-3332
> URL: https://issues.apache.org/jira/browse/LUCENE-3332
> Project: Lucene - Java
> Issue Type: Bug
> Components: modules/highlighter
> Affects Versions: 3.3, 4.0
> Environment: Tested against Lucene trunk revision 1149488 (4.0), but
> seen under Lucene 3.3 (tested through ElasticSearch 0.17.1, but the code is
> clear: this version, and surely other prior (3.1-3.2) are impacted)
> Reporter: Olivier Favre
> Priority: Minor
> Labels: highlighting
> Attachments: TestMultiPhraseQueryHighlighting.java,
> UsingCombinationAndFallback.patch, UsingFallback.patch
>
> Original Estimate: 3h
> Remaining Estimate: 3h
>
> Similar to LUCENE-495.
> Create a MultiPhraseQuery, use FastVectorHighlighter, you'll have no fragment.
> Using PhraseQuery and/or Highlighter works, but not the previous combination.
--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscr...@lucene.apache.org
For additional commands, e-mail: dev-h...@lucene.apache.org
