benwtrent commented on code in PR #12253: URL: https://github.com/apache/lucene/pull/12253#discussion_r1226828445
########## lucene/queries/src/java/org/apache/lucene/queries/function/valuesource/KnnVectorFieldSource.java: ########## @@ -0,0 +1,112 @@ +/* + * 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 org.apache.lucene.queries.function.valuesource; + +import java.io.IOException; +import java.util.Map; +import org.apache.lucene.index.ByteVectorValues; +import org.apache.lucene.index.FloatVectorValues; +import org.apache.lucene.index.LeafReaderContext; +import org.apache.lucene.queries.function.FunctionValues; +import org.apache.lucene.queries.function.ValueSource; +import org.apache.lucene.search.DocIdSetIterator; + +/** An implementation for retrieving {@link FunctionValues} instances for knn vectors fields. */ +public class KnnVectorFieldSource extends ValueSource { + private final String fieldName; + + public KnnVectorFieldSource(String fieldName) { + this.fieldName = fieldName; + } + + @Override + public FunctionValues getValues(Map<Object, Object> context, LeafReaderContext readerContext) + throws IOException { + + final FloatVectorValues floatValues = readerContext.reader().getFloatVectorValues(fieldName); + final ByteVectorValues byteValues = readerContext.reader().getByteVectorValues(fieldName); + + if (floatValues == null && byteValues == null) { + throw new IllegalArgumentException( + "no vector value is indexed for field '" + fieldName + "'"); + } + return new FunctionValues() { + int lastDocID; + + @Override + public float[] floatVectorVal(int doc) throws IOException { + float[] floatVector = null; + if (floatValues != null && exists(floatValues, doc)) { + floatVector = floatValues.vectorValue(); + } else { + byte[] byteVector = byteVectorVal(doc); + if (byteVector != null) { + floatVector = new float[byteVector.length]; + for (int i = 0; i < byteVector.length; i++) { + floatVector[i] = byteVector[i]; + } + } + } + return floatVector; + } Review Comment: > I understand your concerns regarding quantization, and I agree that in the majority of cases, casting from byte to float may not be meaningful. However, it's important to note that the single API we proposed does not restrict users from utilizing the API appropriately in the case of quantization. API's by definition restrict users to the correct use case. I think auto-casting is a bad idea. It gives users the ability to shoot themselves in the foot at runtime. > Currently, we have two encodings, but there may be a need to introduce additional encodings in the future. Duplicating the same code for each encoding increases the likelihood of errors in future contributions. I say we cross that bridge when we get to it. There are ways to deduplicate the code while still giving the users a nice, safe, and correct API. > I want to emphasize that I have not a strong opinion about which approach is the best. I acknowledge that there are advantages and disadvantages to both options. We need to do a Float version and a Byte version. They can rely on similar code to cover your duplication concerns (as long as this doesn't break the design). But, we should have consistency with the existing API design and protect users from making easy mistakes. -- 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: issues-unsubscr...@lucene.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org --------------------------------------------------------------------- To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For additional commands, e-mail: issues-h...@lucene.apache.org
