Sequential DrillSideways skips the two-phase matches() check of the second-cheapest drill-down dimension on segments without deletions, so documents that fail that dimension but are in its approximation are returned as drill-down hits and counted in the facets (released since 5.3.0)

Finding LFV-358Area: facet / DrillSideways (sequential DrillSidewaysScorer, drill-down-advance strategy, two-phase drill-down dimensions)Priority: mediumImpact: silent wrong resultsNovelty: new

Finding page for LFV-358

AI-assisted finding, confirmed by automated checks. Upstream status: Fix in progress, by reugn (open PR #16981; an independent fix, later than our listing). Open PR #16981 (created 2026-11-04) fixes the precedence bug as a side part of DrillSideways timeout and intra-segment concurrency support; its own brute-force TestDrillSidewaysScorer found it. It does not cite this list. Novelty stays new. Source. This site files nothing upstream.

What it shows

You run DrillSideways without an executor on a DrillDownQuery with two or more dimensions, one of which is a two-phase query such as a distance or range drill-down with a fast-match query or a PhraseQuery, and expect the hits to match every dimension and each dimension's sideways counts to cover the documents that match all the other dimensions. Instead, on a segment with no deleted documents, when that dimension is the second-cheapest by approximation cost and the base query is broad, DrillSidewaysScorer never calls its TwoPhaseIterator#matches(), because the condition 'acceptDocs == null || acceptDocs.get(docID) && (dc.twoPhase == null || dc.twoPhase.matches())' parses as 'acceptDocs == null || (get && matches)'. So every document in that dimension's approximation (inside the bounding box but outside the radius, the phrase words in another order) is treated as matching it, returned as a hit and counted in the facets, while concurrent DrillSideways and plain boolean queries on the same searcher give the correct sets.

The test shows that sequential DrillSideways (no executor) on a deletion-free segment never calls the two-phase matches() of the second-cheapest drill-down dimension (here a PhraseQuery) in the drill-down-advance strategy, so a doc that is only in the phrase approximation is returned as a hit.

Expected vs actual

Expected: drill-down hits [0] (only doc 0 has a:1 and the phrase "x y"). Actual: [0, 1] (doc 1 has "y x"), with -ea and -da.

Files

Run it

# Lucene checkout at the tested commit, built with: ./gradlew --no-daemon assemble   (JDK 25)
LUCENE=/path/to/lucene; V=12.0.0-SNAPSHOT
CP=$LUCENE/lucene/core/build/libs/lucene-core-$V.jar
CP=$CP:$LUCENE/lucene/codecs/build/libs/lucene-codecs-$V.jar
CP=$CP:$LUCENE/lucene/test-framework/build/libs/lucene-test-framework-$V.jar
CP=$CP:$LUCENE/lucene/facet/build/libs/lucene-facet-$V.jar
CP=$CP:randomizedtesting-runner-2.9.2.jar:junit-4.13.2.jar:hamcrest-3.0.jar   # test-framework dependencies (Maven Central)
CP=$CP:junit-platform-console-standalone-6.1.3.jar   # JUnit Platform console launcher (Maven Central)

javac -d classes -cp "$CP" $(find src -name '*.java')

# run 1: assertions on, seed DEADBEEF
java -ea -esa --add-modules jdk.incubator.vector --enable-native-access=ALL-UNNAMED -Dtests.seed=DEADBEEF \
     -cp "classes:$CP" org.junit.platform.console.ConsoleLauncher \
     execute --select-class org.apache.lucene.facet.TestDrillSidewaysTwoPhaseSecondDim

The module(s) facet also need their third-party runtime jars; ./gradlew -p lucene/<module> dependencies lists them.

Gradle alternative: copy the test into the matching module's src/test/ tree of the checkout and run ./gradlew -p lucene/<module> test --tests <class> -Ptests.seed=<seed> (add -Ptests.asserts=false for a run without assertions).

Feedback

Open an issue in lucene-ai-bug-hunt. Please do not open issues or pull requests in apache/lucene about this finding unless a Lucene committer agrees it should be filed.

Snapshot of 2026-11-04. AI-assisted; triaged by a Lucene maintainer as noted above. Feedback: open an issue in lucene-ai-bug-hunt.