gsmiller commented on code in PR #13974: URL: https://github.com/apache/lucene/pull/13974#discussion_r1931291804
########## lucene/sandbox/src/java/org/apache/lucene/sandbox/search/SortedSetMultiRangeQuery.java: ########## @@ -0,0 +1,300 @@ +/* + * 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.sandbox.search; + +import java.io.IOException; +import java.util.ArrayList; +import java.util.List; +import java.util.Objects; +import org.apache.lucene.document.SortedSetDocValuesField; +import org.apache.lucene.index.DocValues; +import org.apache.lucene.index.DocValuesSkipper; +import org.apache.lucene.index.LeafReaderContext; +import org.apache.lucene.index.SortedSetDocValues; +import org.apache.lucene.index.TermsEnum; +import org.apache.lucene.search.BooleanQuery; +import org.apache.lucene.search.ConstantScoreScorer; +import org.apache.lucene.search.ConstantScoreWeight; +import org.apache.lucene.search.DocIdSetIterator; +import org.apache.lucene.search.DocValuesRangeIterator; +import org.apache.lucene.search.IndexSearcher; +import org.apache.lucene.search.Query; +import org.apache.lucene.search.QueryVisitor; +import org.apache.lucene.search.ScoreMode; +import org.apache.lucene.search.Scorer; +import org.apache.lucene.search.ScorerSupplier; +import org.apache.lucene.search.TwoPhaseIterator; +import org.apache.lucene.search.Weight; +import org.apache.lucene.util.ArrayUtil; +import org.apache.lucene.util.BytesRef; +import org.apache.lucene.util.LongBitSet; + +/** A union multiple ranges over SortedSetDocValuesField */ +public class SortedSetMultiRangeQuery extends Query { + private final String field; + private final int bytesPerDim; + private final ArrayUtil.ByteArrayComparator comparator; + List<MultiRangeQuery.RangeClause> rangeClauses; + + SortedSetMultiRangeQuery( + String name, + List<MultiRangeQuery.RangeClause> clauses, + int bytes, + ArrayUtil.ByteArrayComparator comparator) { + this.field = name; + this.rangeClauses = clauses; + this.bytesPerDim = bytes; + this.comparator = comparator; + } + + /** Builder for creating a SortedSetMultiRangeQuery. */ + public static class Builder { + private final String name; + protected final List<MultiRangeQuery.RangeClause> clauses = new ArrayList<>(); + private final int bytes; + private final ArrayUtil.ByteArrayComparator comparator; + + public Builder(String name, int bytes) { + this.name = Objects.requireNonNull(name); + this.bytes = bytes; // TODO assrt positive + this.comparator = ArrayUtil.getUnsignedComparator(bytes); + } + + public Builder add(BytesRef lowerValue, BytesRef upperValue) { + byte[] low = lowerValue.clone().bytes; + byte[] up = upperValue.clone().bytes; + if (this.comparator.compare(low, 0, up, 0) > 0) { + throw new IllegalArgumentException("lowerValue must be <= upperValue"); + } else { + clauses.add(new MultiRangeQuery.RangeClause(low, up)); + } + return this; + } + + public Query build() { + if (clauses.isEmpty()) { + return new BooleanQuery.Builder().build(); + } + if (clauses.size() == 1) { + return SortedSetDocValuesField.newSlowRangeQuery( + name, + new BytesRef(clauses.getFirst().lowerValue), + new BytesRef(clauses.getFirst().upperValue), + true, + true); + } + return new SortedSetMultiRangeQuery(name, clauses, this.bytes, comparator); + } + } + + @Override + public Query rewrite(IndexSearcher indexSearcher) throws IOException { + ArrayList<MultiRangeQuery.RangeClause> sortedClauses = new ArrayList<>(this.rangeClauses); + sortedClauses.sort( + (o1, o2) -> { + // if (result == 0) { + // return comparator.compare(o1.upperValue, 0, o2.upperValue, 0); + // } else { + return comparator.compare(o1.lowerValue, 0, o2.lowerValue, 0); + // } + }); + if (!this.rangeClauses.equals(sortedClauses)) { + return new SortedSetMultiRangeQuery( + this.field, sortedClauses, this.bytesPerDim, this.comparator); + } else { + return this; + } + } + + @Override + public String toString(String fld) { + return "SortedSetMultiRangeQuery{" + + "field='" + + fld + + '\'' + + ", rangeClauses=" + + rangeClauses + + // TODO better toString + '}'; + } + + // what TODO with reverse ranges ??? + @Override + public Weight createWeight(IndexSearcher searcher, ScoreMode scoreMode, float boost) + throws IOException { + return new ConstantScoreWeight(this, boost) { + @Override + public ScorerSupplier scorerSupplier(LeafReaderContext context) throws IOException { + if (context.reader().getFieldInfos().fieldInfo(field) == null) { + return null; + } + DocValuesSkipper skipper = context.reader().getDocValuesSkipper(field); + SortedSetDocValues values = DocValues.getSortedSet(context.reader(), field); + // implement ScorerSupplier, since we do some expensive stuff to make a scorer + return new ScorerSupplier() { + @Override + public Scorer get(long leadCost) throws IOException { + if (rangeClauses.isEmpty()) { + return empty(); + } + TermsEnum termsEnum = values.termsEnum(); + LongBitSet matchingOrdsShifted = null; + long minOrd = 0, maxOrd = values.getValueCount() - 1; + long matchesAbove = + values.getValueCount(); // it's last range goes to maxOrd, by default - no match + long maxSeenOrd = values.getValueCount(); + TermsEnum.SeekStatus seekStatus = TermsEnum.SeekStatus.NOT_FOUND; + for (int r = 0; r < rangeClauses.size(); r++) { + MultiRangeQuery.RangeClause range = rangeClauses.get(r); + long startingOrd; + seekStatus = termsEnum.seekCeil(new BytesRef(range.lowerValue)); + if (matchingOrdsShifted == null) { // first iter + if (seekStatus == TermsEnum.SeekStatus.END) { + return empty(); // no bitset yet, give up + } + minOrd = termsEnum.ord(); + if (skipper != null) { + minOrd = Math.max(minOrd, skipper.minValue()); + maxOrd = Math.min(maxOrd, skipper.maxValue()); + } + if (maxOrd < minOrd) { + return empty(); + } + startingOrd = minOrd; + } else { + if (seekStatus == TermsEnum.SeekStatus.END) { + break; // ranges - we are done, terms are exhausted + } else { + startingOrd = termsEnum.ord(); + } + } + byte[] upper = range.upperValue; // TODO ignore reverse ranges + // looking for overlap + for (int overlap = r + 1; overlap < rangeClauses.size(); overlap++, r++) { + MultiRangeQuery.RangeClause mayOverlap = rangeClauses.get(overlap); + assert comparator.compare(range.lowerValue, 0, mayOverlap.lowerValue, 0) <= 0 + : "since they are sorted"; + // TODO it might be contiguous ranges, it's worth to check but I have no idea how + if (comparator.compare(mayOverlap.lowerValue, 0, upper, 0) <= 0) { + // overlap, expand if needed + if (comparator.compare(upper, 0, mayOverlap.upperValue, 0) < 0) { + upper = mayOverlap.upperValue; + } + // continue; // skip overlapping rng + } else { + break; // no r++ + } + } + seekStatus = termsEnum.seekCeil(new BytesRef(upper)); + + if (seekStatus == TermsEnum.SeekStatus.END) { + maxSeenOrd = maxOrd; // perhaps it's worth to set for skipper + matchesAbove = startingOrd; + break; // no need to create bitset + } + maxSeenOrd = + seekStatus == TermsEnum.SeekStatus.FOUND + ? termsEnum.ord() + : termsEnum.ord() - 1; // floor + + if (matchingOrdsShifted == null) { + matchingOrdsShifted = new LongBitSet(maxOrd + 1 - minOrd); + } + matchingOrdsShifted.set( + startingOrd - minOrd, maxSeenOrd - minOrd + 1); // up is exclusive + } + /// ranges are over, there might be no set!! + TwoPhaseIterator iterator; + long finalMatchesAbove = matchesAbove; + LongBitSet finalMatchingOrdsShifted = matchingOrdsShifted; + long finalMinOrd = minOrd; + iterator = + new TwoPhaseIterator(values) { + // TODO unwrap singleton? + @Override + public boolean matches() throws IOException { + for (int i = 0; i < values.docValueCount(); i++) { + long ord = values.nextOrd(); + if (ord >= finalMinOrd + && ((finalMatchesAbove < values.getValueCount() + && ord >= finalMatchesAbove) + || finalMatchingOrdsShifted.get(ord - finalMinOrd))) { Review Comment: Thanks for your thoughts @mkhludnev. I'd suggest one-of-two options. I'd be happy to move forward with either: 1. Keep it as simple as possible and drop all range merging, as you outline above (e.g., drop `resolveOverlaps` and `createOrdRanges`). Then consider adding these ideas back in as spin-off SIMs after the initial implementation is merged. 2. Drop `resolveOverlaps` but keep the logic in `createOrdRanges` as you currently have it. I like how you have this implemented currently. Seems like a pretty clean way of reducing unnecessary calls to `seekCeil` and bitset sets by recognizing adjacent or overlapping ranges without adding much code complexity. Apologies on my above communication. It may not have been as specific as it could have been. The complexity that I'd like to avoid (at least in the initial version of this) is `resolveOverlaps`. I don't mind `createOrdRanges` since it is immediately applicable in this implementation and isn't particularly complex to reason about. The `resolveOverlaps` method seems more future-looking to other implementations, etc., so the value vs. complexity trade-off isn't as clear to me right now. What do you think? (I'm happy to go with whatever you think will help get this merged and reduce headaches for you since we're discussed plenty at this point and I don't want to slow you down further ) -- 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
