LuXugang commented on code in PR #792:
URL: https://github.com/apache/lucene/pull/792#discussion_r864475729
##########
lucene/core/src/java/org/apache/lucene/codecs/lucene91/Lucene91HnswVectorsReader.java:
##########
@@ -507,8 +515,90 @@ public BytesRef binaryValue(int targetOrd) throws
IOException {
}
private void readValue(int targetOrd) throws IOException {
- dataIn.seek((long) targetOrd * byteSize);
- dataIn.readBytes(byteBuffer.array(), byteBuffer.arrayOffset(), byteSize);
+ slice.seek((long) targetOrd * byteSize);
+ slice.readBytes(byteBuffer.array(), byteBuffer.arrayOffset(), byteSize);
+ }
+
+ public IndexedDISI initDISI(IndexInput vectorData) throws IOException {
+ // dense
+ if (fieldEntry == null || fieldEntry.docsWithFieldOffset == -1) {
+ return null;
+ }
+ assert fieldEntry.docsWithFieldOffset != -2;
+ // sparse
+ return new IndexedDISI(
+ vectorData,
+ fieldEntry.docsWithFieldOffset,
+ fieldEntry.docsWithFieldLength,
+ fieldEntry.jumpTableEntryCount,
+ fieldEntry.denseRankPower,
+ fieldEntry.size);
+ }
+
+ public static final OffHeapVectorValues emptyOffHeapVectorValues(int
dimension)
Review Comment:
> I think it'd be clearer if we separated out the sparse and dense cases
into two different subclasses
Totally agreed! It makes code much more readable. but I still want to keep
`emptyOffHeapVectorValues` which has been renamed to `EmptyOffHeapVectorValues`
in latest PR, I would like to keep same pattern like
`Lucene90NormsProducer#getNorms` or `Lucene90DocValuesProducer#getXXX`.
Addressd in
https://github.com/apache/lucene/pull/792/commits/49f67a57e8e294fbed2a63d4887169103f05073b
.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]