Apache
/
lucene
mirror of https://github.com/apache/lucene.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Remove statement that SSDV facets aren't hierarchical (#12232)

This commit is contained in:
Stefan Vodita 2023-04-21 23:40:08 +01:00 committed by GitHub
parent 60c9039d9f
commit 2e7426961b
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
2 changed files with 3 additions and 7 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
lucene/facet/src/java/org/apache/lucene/facet/sortedset/SortedSetDocValuesFacetCounts.java
View File

@ -43,8 +43,6 @@ import org.apache.lucene.util.LongValues;
* Compute facets counts from previously indexed {@link SortedSetDocValuesFacetField}, without
* require a separate taxonomy index. Faceting is a bit slower (~25%), and there is added cost on
* every {@link IndexReader} open to create a new {@link SortedSetDocValuesReaderState}.
* Furthermore, this does not support hierarchical facets; only flat (dimension + label) facets, but
* it uses quite a bit less RAM to do so.
*
* <p><b>NOTE</b>: this class should be instantiated and then used from a single thread, because it
* holds a thread-private instance of {@link SortedSetDocValues}.

8
lucene/facet/src/java/org/apache/lucene/facet/sortedset/SortedSetDocValuesReaderState.java
View File

@ -28,11 +28,9 @@ import org.apache.lucene.util.FixedBitSet;
/**
* Wraps a {@link IndexReader} and resolves ords using existing {@link SortedSetDocValues} APIs
* without a separate taxonomy index. This only supports flat facets (dimension + label), and it
* makes faceting a bit slower, adds some cost at reopen time, but avoids managing the separate
* taxonomy index. It also requires less RAM than the taxonomy index, as it manages the flat
* (2-level) hierarchy more efficiently. In addition, the tie-break during faceting is now
* meaningful (in label sorted order).
* without a separate taxonomy index. This makes faceting a bit slower, adds some cost at reopen
* time, but avoids managing the separate taxonomy index. In addition, the tie-break during faceting
* is now meaningful (in label sorted order).
*
* <p><b>NOTE</b>: creating an instance of this class is somewhat costly, as it computes per-segment
* ordinal maps, so you should create it once and re-use that one instance for a given {@link