[
https://issues.apache.org/jira/browse/LUCENE-8682?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17376769#comment-17376769
]
Michael Gibney commented on LUCENE-8682:
----------------------------------------
Yes, good catch! In most cases this likely wouldn't be a problem, but WDGF's
introduction of the PositionLengthAttribute {{>1}} -- formerly static {{1}} in
WDF -- [would trigger the building
of|https://github.com/apache/lucene/blob/baceb1690442c2cdd6164f1faa34d65b54786a04/lucene/core/src/java/org/apache/lucene/util/QueryBuilder.java#L328-L344]
"graph phrases" in cases that would under WDF have built MultiPhraseQueries.
The former has the potential for exponential expansion where the latter does
not; in the case referred to on the solr users list, the
[solution/workaround|https://lists.apache.org/thread.html/ra25ec5cb8bdfcd3d911f4d17a301a43ccf4ee02bf42cea59d1b0b9c0%40%3Cusers.solr.apache.org%3E]
involved explicitly setting {{enableGraphPhrase=false}} on the fieldType.
There may be other subtle differences in the order of tokens returned by the
two filters \(?), and the non-graph version of the phrase query has its own
problems I think (likely false negatives, false positives, in some cases). But
I doubt there's much we can do to address the more subtle differences. The
potential performance impact of migrating configs from WDF=>WDGF perhaps merits
an upgrade note in CHANGES.txt mentioning the {{enableGraphQueries=false}}
workaround.
It'd also be possible to add general documentation (javadocs, refguide),
independent of mentioning WDF, about the potential graph phrase query expansion
issues with heavily branching token streams, and the (however sub-optimal)
workaround of explicitly disabling graph phrase queries. Though that's pretty
edge-casey, and might end up raising more questions than it answers, for the
majority of users!
> Remove WordDelimiterFilter
> --------------------------
>
> Key: LUCENE-8682
> URL: https://issues.apache.org/jira/browse/LUCENE-8682
> Project: Lucene - Core
> Issue Type: Task
> Affects Versions: main (9.0)
> Reporter: Alan Woodward
> Assignee: Alan Woodward
> Priority: Major
> Attachments: LUCENE-8682.patch
>
> Time Spent: 0.5h
> Remaining Estimate: 0h
>
> WordDelimiterFilter was deprecated a while back. We can remove it entirely
> from the master branch.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org
For additional commands, e-mail: issues-h...@lucene.apache.org
