[ 
https://issues.apache.org/jira/browse/LUCENE-9744?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17280950#comment-17280950
 ] 

Alan Woodward commented on LUCENE-9744:
---------------------------------------

> There is no time pressure on this at all, Alan

I found a fix already though :). This is in fact due to the degeneracy; in a 
'normal' min-should-match scenario, the leading interval gets moved on once a 
match is found to see if the interval can be minimized, in the same way that an 
unordered interval does.  In order to return the correct sub-matches, we cache 
this leading interval before it is moved on.  Then when getSubMatches is 
called, we collect the cached lead interval and then all the other intervals 
that are behind the end of the current interval, which will filter out the 
*current* position of the lead.  In the degenerate case, though, the interval 
is already minimized (because it only has one sub-interval), and so the lead 
variable never gets set.  And so when we collect sub matches, collecting the 
lead returns null.

I'm attaching a patch with a) a test for this case, and b) a fix in the factory 
method that ensures we don't ever actually get a degenerate min-should-match 
interval.

PS I see that your test uses the matches highlighter, this was helpful for 
finding the bug but of course couldn't be backported along with a fix because 
the matches highlighter still isn't in 8.x, hint hint :)

> NPE on a degenerate query in 
> MinimumShouldMatchIntervalsSource$MinimumMatchesIterator.getSubMatches()
> -----------------------------------------------------------------------------------------------------
>
>                 Key: LUCENE-9744
>                 URL: https://issues.apache.org/jira/browse/LUCENE-9744
>             Project: Lucene - Core
>          Issue Type: Bug
>    Affects Versions: master (9.0), 8.7
>            Reporter: Dawid Weiss
>            Priority: Major
>         Attachments: LUCENE-9744.patch
>
>
> This query:
> {code}
>               Query q1 =
>                   new IntervalQuery(
>                       FLD_TEXT1,
>                       Intervals.ordered(
>                           Intervals.atLeast(1, Intervals.term("foo")),
>                           Intervals.atLeast(1, Intervals.term("bar"))));
> {code}
> produces an NPE when retrieving match regions. Note degenerate atLeast with 
> only one argument - this may play a role but it's not signaled anywhere 
> before (the query itself does select the right documents, NPE blows up when 
> matches are retrieved).
> Stack:
> {code}
> java.lang.NullPointerException
>       at 
> __randomizedtesting.SeedInfo.seed([2DDAAECF04D90D2F:EB69410397F8761E]:0)
>       at 
> org.apache.lucene.queries.intervals.MinimumShouldMatchIntervalsSource$MinimumMatchesIterator.getSubMatches(MinimumShouldMatchIntervalsSource.java:389)
>       at 
> org.apache.lucene.queries.intervals.CachingMatchesIterator.cache(CachingMatchesIterator.java:39)
>       at 
> org.apache.lucene.queries.intervals.CachingMatchesIterator.next(CachingMatchesIterator.java:68)
>       at 
> org.apache.lucene.queries.intervals.IntervalMatches$2.nextInterval(IntervalMatches.java:137)
>       at 
> org.apache.lucene.queries.intervals.OrderedIntervalsSource$OrderedIntervalIterator.nextInterval(OrderedIntervalsSource.java:166)
>       at 
> org.apache.lucene.queries.intervals.ConjunctionIntervalsSource.matches(ConjunctionIntervalsSource.java:89)
>       at 
> org.apache.lucene.queries.intervals.IntervalQuery$IntervalWeight.lambda$matches$0(IntervalQuery.java:177)
>       at org.apache.lucene.search.MatchesUtils.forField(MatchesUtils.java:115)
>       at 
> org.apache.lucene.queries.intervals.IntervalQuery$IntervalWeight.matches(IntervalQuery.java:174)
>       at 
> org.apache.lucene.search.matchhighlight.MatchRegionRetriever.highlightDocument(MatchRegionRetriever.java:210)
>       at 
> org.apache.lucene.search.matchhighlight.MatchRegionRetriever.highlightDocuments(MatchRegionRetriever.java:193)
>       at 
> org.apache.lucene.search.matchhighlight.MatchRegionRetriever.highlightDocuments(MatchRegionRetriever.java:146)
>       at 
> org.apache.lucene.search.matchhighlight.MatchHighlighter.highlight(MatchHighlighter.java:236)
>       at 
> org.apache.lucene.search.matchhighlight.TestMatchHighlighter.lambda$testNpeOnIntervalQuery$7(TestMatchHighlighter.java:540)
>       at 
> org.apache.lucene.search.matchhighlight.IndexBuilder.build(IndexBuilder.java:100)
>       at 
> org.apache.lucene.search.matchhighlight.TestMatchHighlighter.testNpeOnIntervalQuery(TestMatchHighlighter.java:510)
>       at 
> java.base/jdk.internal.reflect.NativeMethodAccessorImpl.invoke0(Native Method)
>       at 
> java.base/jdk.internal.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62)
>       at 
> java.base/jdk.internal.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43)
>       at java.base/java.lang.reflect.Method.invoke(Method.java:564)
> {code}
> Repro test attached. 
> [~romseygeek], help! :)



--
This message was sent by Atlassian Jira
(v8.3.4#803005)

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to