Skip to content

Fix QueryPhraseMap.markTerminal() boost override due to conflicting query expansion #15434

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
dsmiley merged 3 commits into from
Jan 5, 2026
Merged

Fix QueryPhraseMap.markTerminal() boost override due to conflicting query expansion #15434

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
50e52f0
Fix qpm override
fragosoluana Nov 18, 2025
95b659a
Merge branch 'main' into highlight_qpm_boost_override_fix
fragosoluana Jan 5, 2026
e0a075d
Addressed comments
fragosoluana Jan 5, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
2 changes: 2 additions & 0 deletions lucene/CHANGES.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -112,6 +112,8 @@ Bug Fixes

* GITHUB#15125: Handle inconsistent schema on flush with index sorts (Nhat Nguyen)

* GITHUB#15434: Don't override boost of conflicting expanded phrases by taking the max boost in markTerminal for FVHighlighter (Luana Fragoso)

Changes in Runtime Behavior
---------------------
* GITHUB#14187: The query cache is now disabled by default. (Adrien Grand)
Expand Down
2 changes: 1 addition & 1 deletion lucene/highlighter/src/java/org/apache/lucene/search/vectorhighlight/FieldQuery.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -429,7 +429,7 @@ private void markTerminal(float boost) {
private void markTerminal(int slop, float boost) {
this.terminal = true;
this.slop = slop;
this.boost = boost;
this.boost = Math.max(this.boost, boost);
Copy link

@sunny6142 sunny6142 Nov 19, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Your approach seems simpler, but I think we have an opportunity to optimize this. Instead of always executing all assignments, should we have something like:

   if (!this.terminal || boost > this.boost) {
        this.terminal = true;
        this.slop = slop;
        this.boost = boost;
        this.termOrPhraseNumber = fieldQuery.nextTermOrPhraseNumber();
      }

Benefits:

Performance: Avoids unnecessary state updates when boost doesn't improve

Semantic correctness: termOrPhraseNumber only increments when meaningful changes occur

Cleaner logic: Single condition handles both initialization and duplicate prevention

Current approach with Math.max():

Always updates all fields, even when boost=1 < existing boost=100

Increments termOrPhraseNumber unnecessarily on duplicate calls

Proposed approach:

Only updates when first call (!this.terminal) or when boost improves (boost > this.boost)

More efficient for queries with many overlapping phrases

What do you think?

Copy link
Contributor

@dsmiley dsmiley Dec 26, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It's a fine suggestion but it's also a trivial matter. I'd even argue the complexity of the condition you propose makes the situation worse.... a future reader may scratching their heads for a little longer than merits the code being avoided.

this.termOrPhraseNumber = fieldQuery.nextTermOrPhraseNumber();
}

Expand Down
43 changes: 43 additions & 0 deletions lucene/highlighter/src/test/org/apache/lucene/search/vectorhighlight/TestFieldQuery.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -19,6 +19,7 @@
import java.io.IOException;
import java.util.ArrayList;
import java.util.HashSet;
import java.util.LinkedHashSet;
import java.util.List;
import java.util.Map;
import java.util.Set;
Expand Down Expand Up @@ -814,6 +815,48 @@ public void testQueryPhraseMapOverlap2gram() throws Exception {
assertEquals(0, qpm2.subMap.size());
}

public void testQueryPhraseMapDuplicate() throws IOException {
BooleanQuery.Builder query = new BooleanQuery.Builder();
Query bq = toPhraseQuery(analyze("a b c", F, analyzerB), F);
bq = new BoostQuery(bq, 100);
query.add(bq, Occur.SHOULD);

bq = toPhraseQuery(analyze("a b", F, analyzerB), F);
bq = new BoostQuery(bq, 20);
query.add(bq, Occur.SHOULD);

bq = toPhraseQuery(analyze("b c", F, analyzerB), F);
bq = new BoostQuery(bq, 50);
query.add(bq, Occur.SHOULD);

bq = query.build();
FieldQuery fq = new FieldQuery(bq, true, true);
Set<Query> flatQueries = new LinkedHashSet<>();
fq.flatten(bq, searcher, flatQueries, 1f);

assertCollectionQueries(
fq.expand(flatQueries),
pqF(100, "a", "b", "c"),
pqF(20, "a", "b"),
// "a b c": 1 -> expanded from "a b" + "b c"
new BoostQuery(pqF(1f, "a", "b", "c"), 1f),
pqF(50, "b", "c"));

Map<String, QueryPhraseMap> map = fq.rootMaps;
QueryPhraseMap a_qpm = map.get("f").subMap.get("a");
assertEquals(0, a_qpm.boost, 0.0);
QueryPhraseMap b_qpm = a_qpm.subMap.get("b");
assertEquals(20, b_qpm.boost, 0.0);
QueryPhraseMap c_qpm = b_qpm.subMap.get("c");
// make sure final boost is from the query and not the expanded boost 1
assertEquals(100, c_qpm.boost, 0.0);

b_qpm = map.get("f").subMap.get("b");
assertEquals(0, b_qpm.boost, 0.0);
c_qpm = b_qpm.subMap.get("c");
assertEquals(50, c_qpm.boost, 0.0);
}

public void testSearchPhrase() throws Exception {
Query query = pqF("a", "b", "c");

Expand Down