Mirror/MarginaliaSearch
1
0
Fork 0
mirror of https://github.com/MarginaliaSearch/MarginaliaSearch.git synced 2025-02-23 21:18:58 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

(index) Fix term coherence evaluation

Browse Source 
The code was incorrectly using the documentId instead of the combined id, resulting in almost all result sets being incorrectly seen as zero.
This commit is contained in:
Viktor Lofgren 2024-04-16 18:07:43 +02:00
parent 1748fcc5ac
commit adf846bfd2
2 changed files with 5 additions and 5 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
code/index/java/nu/marginalia/index/results/IndexResultValuationContext.java
View File

@ -64,7 +64,7 @@ public class IndexResultValuationContext {
long docId = UrlIdCodec.removeRank(combinedId);
if (!searchTerms.coherences.test(termMetadataForCombinedDocumentIds, docId))
if (!searchTerms.coherences.test(termMetadataForCombinedDocumentIds, combinedId))
return null;
long docMetadata = statefulIndex.getDocumentMetadata(docId);

8
code/index/java/nu/marginalia/index/results/model/TermCoherenceGroupList.java
View File

@ -15,9 +15,9 @@ public record TermCoherenceGroupList(List<TermCoherenceGroup> words) {
this.words = Collections.unmodifiableList(words);
}
public boolean test(TermMetadataForCombinedDocumentIds documents, long docId) {
public boolean test(TermMetadataForCombinedDocumentIds documents, long combinedId) {
for (var coherenceSet : words()) {
if (!coherenceSet.test(documents, docId)) {
if (!coherenceSet.test(documents, combinedId)) {
return false;
}
}
@ -36,11 +36,11 @@ public record TermCoherenceGroupList(List<TermCoherenceGroup> words) {
this(coh.stream().mapToLong(SearchTermsUtil::getWordId).toArray());
}
public boolean test(TermMetadataForCombinedDocumentIds documents, long docId) {
public boolean test(TermMetadataForCombinedDocumentIds documents, long combinedId) {
long overlap = 0xFF_FFFF_FFFF_FFFFL;
for (var word : words) {
overlap &= documents.getTermMetadata(word, docId);
overlap &= documents.getTermMetadata(word, combinedId);
}
return WordMetadata.decodePositions(overlap) != 0L;