2021-10-13 00:23:56 +02:00
|
|
|
/*
|
|
|
|
* Licensed to the Apache Software Foundation (ASF) under one or more
|
|
|
|
* contributor license agreements. See the NOTICE file distributed with
|
|
|
|
* this work for additional information regarding copyright ownership.
|
|
|
|
* The ASF licenses this file to You under the Apache License, Version 2.0
|
|
|
|
* (the "License"); you may not use this file except in compliance with
|
|
|
|
* the License. You may obtain a copy of the License at
|
|
|
|
*
|
|
|
|
* http://www.apache.org/licenses/LICENSE-2.0
|
|
|
|
*
|
|
|
|
* Unless required by applicable law or agreed to in writing, software
|
|
|
|
* distributed under the License is distributed on an "AS IS" BASIS,
|
|
|
|
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
|
|
* See the License for the specific language governing permissions and
|
|
|
|
* limitations under the License.
|
|
|
|
*/
|
|
|
|
|
|
|
|
package it.cavallium.dbengine.lucene;
|
|
|
|
|
|
|
|
import java.util.Objects;
|
|
|
|
import java.util.concurrent.atomic.LongAccumulator;
|
|
|
|
|
2021-11-16 23:19:13 +01:00
|
|
|
/**
|
|
|
|
* Maintains the maximum score and its corresponding document id concurrently
|
|
|
|
*
|
|
|
|
* This class must mirror this changes:
|
|
|
|
* <a href="https://github.com/apache/lucene/commits/94b66c0ed279fe23656d451fecd56fdfd106e1ea/lucene/core/src/java/org/apache/lucene/search/MaxScoreAccumulator.java">
|
|
|
|
* Lucene MaxScoreAccumulator changes on GitHub</a>
|
|
|
|
*/
|
2021-10-13 00:23:56 +02:00
|
|
|
public final class MaxScoreAccumulator {
|
|
|
|
// we use 2^10-1 to check the remainder with a bitwise operation
|
|
|
|
static final int DEFAULT_INTERVAL = 0x3ff;
|
|
|
|
|
|
|
|
// scores are always positive
|
2021-11-16 23:19:13 +01:00
|
|
|
final LongAccumulator acc = new LongAccumulator(MaxScoreAccumulator::maxEncode, Long.MIN_VALUE);
|
2021-10-13 00:23:56 +02:00
|
|
|
|
|
|
|
// non-final and visible for tests
|
|
|
|
public long modInterval;
|
|
|
|
|
|
|
|
public MaxScoreAccumulator() {
|
|
|
|
this.modInterval = DEFAULT_INTERVAL;
|
|
|
|
}
|
|
|
|
|
2021-11-16 23:19:13 +01:00
|
|
|
/**
|
|
|
|
* Return the max encoded DocAndScore in a way that is consistent with {@link
|
|
|
|
* DocAndScore#compareTo}.
|
|
|
|
*/
|
|
|
|
private static long maxEncode(long v1, long v2) {
|
|
|
|
float score1 = Float.intBitsToFloat((int) (v1 >> 32));
|
|
|
|
float score2 = Float.intBitsToFloat((int) (v2 >> 32));
|
|
|
|
int cmp = Float.compare(score1, score2);
|
|
|
|
if (cmp == 0) {
|
|
|
|
// tie-break on the minimum doc base
|
|
|
|
return (int) v1 < (int) v2 ? v1 : v2;
|
|
|
|
} else if (cmp > 0) {
|
|
|
|
return v1;
|
|
|
|
}
|
|
|
|
return v2;
|
|
|
|
}
|
|
|
|
|
|
|
|
public void accumulate(int docBase, float score) {
|
|
|
|
assert docBase >= 0 && score >= 0;
|
|
|
|
long encode = (((long) Float.floatToIntBits(score)) << 32) | docBase;
|
2021-10-13 00:23:56 +02:00
|
|
|
acc.accumulate(encode);
|
|
|
|
}
|
|
|
|
|
|
|
|
public DocAndScore get() {
|
|
|
|
long value = acc.get();
|
|
|
|
if (value == Long.MIN_VALUE) {
|
|
|
|
return null;
|
|
|
|
}
|
|
|
|
float score = Float.intBitsToFloat((int) (value >> 32));
|
2021-11-16 23:19:13 +01:00
|
|
|
int docBase = (int) value;
|
|
|
|
return new DocAndScore(docBase, score);
|
2021-10-13 00:23:56 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
public static class DocAndScore implements Comparable<DocAndScore> {
|
2021-11-16 23:19:13 +01:00
|
|
|
public final int docBase;
|
2021-10-13 00:23:56 +02:00
|
|
|
public final float score;
|
|
|
|
|
2021-11-16 23:19:13 +01:00
|
|
|
public DocAndScore(int docBase, float score) {
|
|
|
|
this.docBase = docBase;
|
2021-10-13 00:23:56 +02:00
|
|
|
this.score = score;
|
|
|
|
}
|
|
|
|
|
|
|
|
@Override
|
|
|
|
public int compareTo(DocAndScore o) {
|
|
|
|
int cmp = Float.compare(score, o.score);
|
|
|
|
if (cmp == 0) {
|
2021-11-16 23:19:13 +01:00
|
|
|
// tie-break on the minimum doc base
|
|
|
|
// For a given minimum competitive score, we want to know the first segment
|
|
|
|
// where this score occurred, hence the reverse order here.
|
|
|
|
// On segments with a lower docBase, any document whose score is greater
|
|
|
|
// than or equal to this score would be competitive, while on segments with a
|
|
|
|
// higher docBase, documents need to have a strictly greater score to be
|
|
|
|
// competitive since we tie break on doc ID.
|
|
|
|
return Integer.compare(o.docBase, docBase);
|
2021-10-13 00:23:56 +02:00
|
|
|
}
|
|
|
|
return cmp;
|
|
|
|
}
|
|
|
|
|
|
|
|
@Override
|
|
|
|
public boolean equals(Object o) {
|
|
|
|
if (this == o) return true;
|
|
|
|
if (o == null || getClass() != o.getClass()) return false;
|
|
|
|
DocAndScore result = (DocAndScore) o;
|
2021-11-16 23:19:13 +01:00
|
|
|
return docBase == result.docBase && Float.compare(result.score, score) == 0;
|
2021-10-13 00:23:56 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
@Override
|
|
|
|
public int hashCode() {
|
2021-11-16 23:19:13 +01:00
|
|
|
return Objects.hash(docBase, score);
|
2021-10-13 00:23:56 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
@Override
|
|
|
|
public String toString() {
|
2021-11-16 23:19:13 +01:00
|
|
|
return "DocAndScore{" + "docBase=" + docBase + ", score=" + score + '}';
|
2021-10-13 00:23:56 +02:00
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|