MultiSimilarity does not document that only sims[0] computes the field norm, so the other sub-similarities' index-time norm settings (discountOverlaps, a computeNorm override) are silently ignored and the array order decides the stored norm, and its inherited getDiscountOverlaps() always returns true (documentation defect; on one index the score is an exact sum in either order)
Finding LFV-903Area: core / search similarities (MultiSimilarity: undocumented norm ownership, inherited getDiscountOverlaps)Priority: lowImpact: minorNovelty: known upstream
AI-assisted finding, confirmed by automated checks, not yet triaged by a human Lucene committer. This site files nothing upstream.
What it shows
You combine two sub-similarities in a MultiSimilarity, for example BM25 with discountOverlaps=true and BM25 with discountOverlaps=false, and use it for indexing and searching; the javadoc says only that it 'will sum the scores of the provided sims' (CombSUM), so you expect both sub-similarities' settings to count. In fact computeNorm returns sims[0].computeNorm(state), so only the first sub-similarity writes the field's single norm and every sub-scorer decodes that norm: the second one's discountOverlaps (or its own computeNorm encoding) is silently ignored, swapping the array order changes the stored norms and therefore the scores (0.7534208 for [discount, noDiscount] vs 0.6862843 for [noDiscount, discount] in an independent repro), and the inherited getDiscountOverlaps() returns true even when sims[0] counts overlap tokens. The single norm is the intended one-norm-per-field design and on any one index the score is an exact sum in either order; what is missing is the documentation (and a correct getter), so the fix is a javadoc sentence, optionally plus passing sims[0].getDiscountOverlaps() to the super constructor.
Expected vs actual
- Expected: You expect both sub-similarities' settings to count.
- Actual: While the reported behaviour is present the test fails; the failure message states expected against actual.
Files
Run it
- Tested on Lucene main e1faeba08bde (2026-11-02): still reproduces.
- Tested on Lucene main a9779ef3c193 (2026-10-26): reproduced.
- How to read the outcome: a failing test is the finding reproducing (the failure message states expected against actual).
- Seeds: a rerun with the same seed replays the same random choices. Assertions off (-da -dsa -Dtests.asserts=false) is the production-like setting.
# 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=$CP:randomizedtesting-runner-2.9.2.jar:junit-4.13.2.jar:hamcrest-3.0.jar # test-framework dependencies (Maven Central)
CP=$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.search.similarities.TestMultiSimilarityNormsGradle 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; not yet triaged by a human Lucene committer. Feedback: open an issue in lucene-ai-bug-hunt.