[
https://issues.apache.org/jira/browse/LUCENE-9524?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17211490#comment-17211490
]
Zach Chen commented on LUCENE-9524:
-----------------------------------
So after much debugging and tracing, I found that revert-ting the one-line
change from https://issues.apache.org/jira/browse/LUCENE-7238 would resolve
this NPE issue, but not sure if this is the right approach to be taken here.
The underlying issue causing the differences in behavior is that, the *NOT
\"dolor lorem\"* in the query above would generate a _SpanNearWeight_ with
ScoreMode _COMPLETE_NO_SCORES_ for both cases. However, when _LRUQueryCache_ is
available for non MemoryIndex, this weight gets wrapped into
_LRUQueryCache$CachingWrapperWeight_, which extends _ConstantScoreWeight_ and
doesn't use (the null) simScorer for explanation.
So it seems like the use of _LRUQueryCache$CachingWrapperWeight_ can make the
underlying potentially variable-score weight to be constant-score weight for
explanation.
On the other hand, _SpanNearWeight's_ super class _SpanWight_ seems to allow
null simScorer field, but its explanation method's call to _LeafSimScorer_
constructor doesn't check for the case of null simScorer. Maybe a check is
needed there as well?
[~jpountz] Sine from git you worked on both aspects above, I'm wondering
what's your take on this? I can work on a patch if needed.
> NullPointerException in IndexSearcher.explain() when using
> ComplexPhraseQueryParser
> -----------------------------------------------------------------------------------
>
> Key: LUCENE-9524
> URL: https://issues.apache.org/jira/browse/LUCENE-9524
> Project: Lucene - Core
> Issue Type: Bug
> Components: core/queryparser, core/search
> Affects Versions: 8.6, 8.6.2
> Reporter: Michał Słomkowski
> Priority: Major
>
> I get NPE when I use {{IndexSearcher.explain()}}. Checked with Lucene 8.6.0
> and 8.6.2.
> The query: {{(lorem AND NOT "dolor lorem") OR ipsum}}
> The text: {{dolor lorem ipsum}}
> Stack trace:
> {code}
> java.lang.NullPointerException at
> java.util.Objects.requireNonNull(Objects.java:203)
> at org.apache.lucene.search.LeafSimScorer.<init>(LeafSimScorer.java:38)
> at
> org.apache.lucene.search.spans.SpanWeight.explain(SpanWeight.java:160)
> at org.apache.lucene.search.BooleanWeight.explain(BooleanWeight.java:87)
> at org.apache.lucene.search.BooleanWeight.explain(BooleanWeight.java:87)
> at
> org.apache.lucene.search.IndexSearcher.explain(IndexSearcher.java:716)
> at
> org.apache.lucene.search.IndexSearcher.explain(IndexSearcher.java:693)
> {code}
> Minimal example code:
> {code:java}
> val analyzer = new StandardAnalyzer();
> val query = new ComplexPhraseQueryParser("", analyzer).parse(queryString);
> final MemoryIndex memoryIndex = new MemoryIndex(true);
> memoryIndex.addField("", text, analyzer);
> final IndexSearcher searcher = memoryIndex.createSearcher();
> final TopDocs topDocs = searcher.search(query, 1);
> final ScoreDoc match = topDocs.scoreDocs[0];
> searcher.explain(query, match.doc);
> {code}
--
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
