github-mirrors/elasticsearch
1
0
Fork 0
mirror of https://github.com/elastic/elasticsearch.git synced 2025-06-28 01:22:26 -04:00
Code Issues Projects Releases Packages Wiki Activity

Optimize sparse vector stats collection (#128740) (#128806)

Browse source 
This change improves the performance of sparse vector statistics gathering by using the document count of terms directly, rather than relying on the field name field to compute stats.
By avoiding per-term disk/network reads and instead leveraging statistics already loaded into leaf readers at index opening, we expect to significantly reduce overhead.

Relates to #128583
This commit is contained in:
Jim Ferenczi 2025-06-03 11:17:34 +02:00 • committed by GitHub
parent 1b03d0956d
commit bab6e835d4
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
2 changed files with 14 additions and 10 deletions
Download patch file Download diff file Expand all files Collapse all files

5
docs/changelog/128740.yaml Normal file
View file

@ -0,0 +1,5 @@
pr: 128740
summary: Optimize sparse vector stats collection
area: Stats
type: enhancement
issues: []

19
server/src/main/java/org/elasticsearch/index/engine/Engine.java
View file

@ -25,7 +25,6 @@ import org.apache.lucene.index.SegmentCommitInfo;
import org.apache.lucene.index.SegmentInfos;
import org.apache.lucene.index.SegmentReader;
import org.apache.lucene.index.Terms;
import org.apache.lucene.index.TermsEnum;
import org.apache.lucene.search.IndexSearcher;
import org.apache.lucene.search.QueryCache;
import org.apache.lucene.search.QueryCachingPolicy;
@ -61,7 +60,6 @@ import org.elasticsearch.core.TimeValue;
import org.elasticsearch.index.IndexVersion;
import org.elasticsearch.index.VersionType;
import org.elasticsearch.index.mapper.DocumentParser;
import org.elasticsearch.index.mapper.FieldNamesFieldMapper;
import org.elasticsearch.index.mapper.LuceneDocument;
import org.elasticsearch.index.mapper.Mapper;
import org.elasticsearch.index.mapper.Mapping;
@ -337,14 +335,15 @@ public abstract class Engine implements Closeable {
private long getSparseVectorValueCount(final LeafReader atomicReader, List<BytesRef> fields) throws IOException {
long count = 0;
Terms terms = atomicReader.terms(FieldNamesFieldMapper.NAME);
if (terms == null) {
return count;
}
TermsEnum termsEnum = terms.iterator();
for (var fieldName : fields) {
if (termsEnum.seekExact(fieldName)) {
count += termsEnum.docFreq();
for (var fieldNameBR : fields) {
var fieldName = fieldNameBR.utf8ToString();
var fi = atomicReader.getFieldInfos().fieldInfo(fieldName);
if (fi == null) {
continue;
}
Terms terms = atomicReader.terms(fieldName);
if (terms != null) {
count += terms.getDocCount();
}
}
return count;