[
https://issues.apache.org/jira/browse/LUCENE-6272?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Alan Woodward updated LUCENE-6272:
----------------------------------
Attachment: LUCENE-6272.patch
Patch. I also added a PhraseScorer abstract class with a phraseFreq() method
that PhraseWeight.explain() calls.
There are a couple of tests that walk the Scorer tree and check that
TermScorers return the correct freq() which I think we can just nuke? Seeing
as TermScorer.freq() is literally called nowhere else in the codebase.
Alternatively, we could keep the existing freq() implementations for
backwards-compatibility purposes, deprecating them in 5x and removing in trunk.
> Scorer should not extend PostingsEnum
> -------------------------------------
>
> Key: LUCENE-6272
> URL: https://issues.apache.org/jira/browse/LUCENE-6272
> Project: Lucene - Core
> Issue Type: Improvement
> Reporter: Alan Woodward
> Priority: Minor
> Attachments: LUCENE-6272.patch
>
>
> Scorer currently has to implement a whole bunch of methods that are never
> called. The only method that Scorer uses in addition to the methods on
> DocIdSetIterator is freq(), and as currently implemented this means different
> things on different Scorers:
> * TermScorer returns its underlying termfreq
> * MinShouldMatchScorer returns how many of its subscorers are matching
> * {Exact|Sloppy}PhraseScorer returns how many phrases it has found on a
> document
> In addition, freq() is never actually called on TermScorer, and it's only
> used in explain() on the phrase scorers.
> We should make Scorer extend DocIdSetIterator instead. In place of freq(),
> Scorer would have a coord() method that by default returns 1, and for boolean
> scorers returns how many subscorers are matching.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
