Fix duplicate removal when merging completion suggestions (#36996)
The completion suggester ignores the original weight of the suggestion when duplicates are removed. This change fixes this bug and keeps the best weighted suggestion among the duplicates. It also removes the custom implementation of the top docs suggest collector now that https://issues.apache.org/jira/browse/LUCENE-8529 is committed in Lucene. Closes #35836
This commit is contained in:
Jim Ferenczi
GitHub
parent
bea46f7b52
commit
f8d80dff7c
|
|
@ -20,7 +20,7 @@ package org.elasticsearch.search.suggest.completion;
|
|||
|
||||
import org.apache.lucene.analysis.CharArraySet;
|
||||
import org.apache.lucene.search.ScoreDoc;
|
||||
import org.apache.lucene.search.suggest.Lookup;
|
||||
import org.apache.lucene.util.PriorityQueue;
|
||||
import org.elasticsearch.Version;
|
||||
import org.elasticsearch.common.ParseField;
|
||||
import org.elasticsearch.common.io.stream.StreamInput;
|
||||
|
|
@ -36,9 +36,9 @@ import org.elasticsearch.search.suggest.Suggest.Suggestion;
|
|||
|
||||
import java.io.IOException;
|
||||
import java.util.Collections;
|
||||
import java.util.Comparator;
|
||||
import java.util.HashMap;
|
||||
import java.util.HashSet;
|
||||
import java.util.Iterator;
|
||||
import java.util.LinkedHashMap;
|
||||
import java.util.List;
|
||||
import java.util.Map;
|
||||
|
|
@ -138,31 +138,34 @@ public final class CompletionSuggestion extends Suggest.Suggestion<CompletionSug
|
|||
return suggestion;
|
||||
}
|
||||
|
||||
private static final class OptionPriorityQueue extends org.apache.lucene.util.PriorityQueue<Entry.Option> {
|
||||
|
||||
private final Comparator<Suggest.Suggestion.Entry.Option> comparator;
|
||||
|
||||
OptionPriorityQueue(int maxSize, Comparator<Suggest.Suggestion.Entry.Option> comparator) {
|
||||
private static final class OptionPriorityQueue extends PriorityQueue<ShardOptions> {
|
||||
OptionPriorityQueue(int maxSize) {
|
||||
super(maxSize);
|
||||
this.comparator = comparator;
|
||||
}
|
||||
|
||||
@Override
|
||||
protected boolean lessThan(Entry.Option a, Entry.Option b) {
|
||||
int cmp = comparator.compare(a, b);
|
||||
if (cmp != 0) {
|
||||
return cmp > 0;
|
||||
}
|
||||
return Lookup.CHARSEQUENCE_COMPARATOR.compare(a.getText().string(), b.getText().string()) > 0;
|
||||
protected boolean lessThan(ShardOptions a, ShardOptions b) {
|
||||
return COMPARATOR.compare(a.current, b.current) < 0;
|
||||
}
|
||||
}
|
||||
|
||||
private static class ShardOptions {
|
||||
final Iterator<Entry.Option> optionsIterator;
|
||||
Entry.Option current;
|
||||
|
||||
private ShardOptions(Iterator<Entry.Option> optionsIterator) {
|
||||
assert optionsIterator.hasNext();
|
||||
this.optionsIterator = optionsIterator;
|
||||
this.current = optionsIterator.next();
|
||||
}
|
||||
|
||||
Entry.Option[] get() {
|
||||
int size = size();
|
||||
Entry.Option[] results = new Entry.Option[size];
|
||||
for (int i = size - 1; i >= 0; i--) {
|
||||
results[i] = pop();
|
||||
boolean advanceToNextOption() {
|
||||
if (optionsIterator.hasNext()) {
|
||||
current = optionsIterator.next();
|
||||
return true;
|
||||
} else {
|
||||
return false;
|
||||
}
|
||||
return results;
|
||||
}
|
||||
}
|
||||
|
||||
|
|
@ -177,37 +180,43 @@ public final class CompletionSuggestion extends Suggest.Suggestion<CompletionSug
|
|||
final CompletionSuggestion leader = (CompletionSuggestion) toReduce.get(0);
|
||||
final Entry leaderEntry = leader.getEntries().get(0);
|
||||
final String name = leader.getName();
|
||||
int size = leader.getSize();
|
||||
if (toReduce.size() == 1) {
|
||||
return leader;
|
||||
} else {
|
||||
// combine suggestion entries from participating shards on the coordinating node
|
||||
// the global top <code>size</code> entries are collected from the shard results
|
||||
// using a priority queue
|
||||
OptionPriorityQueue priorityQueue = new OptionPriorityQueue(leader.getSize(), COMPARATOR);
|
||||
// Dedup duplicate suggestions (based on the surface form) if skip duplicates is activated
|
||||
final CharArraySet seenSurfaceForms = leader.skipDuplicates ? new CharArraySet(leader.getSize(), false) : null;
|
||||
OptionPriorityQueue pq = new OptionPriorityQueue(toReduce.size());
|
||||
for (Suggest.Suggestion<Entry> suggestion : toReduce) {
|
||||
assert suggestion.getName().equals(name) : "name should be identical across all suggestions";
|
||||
for (Entry.Option option : ((CompletionSuggestion) suggestion).getOptions()) {
|
||||
if (leader.skipDuplicates) {
|
||||
assert ((CompletionSuggestion) suggestion).skipDuplicates;
|
||||
String text = option.getText().string();
|
||||
if (seenSurfaceForms.contains(text)) {
|
||||
continue;
|
||||
}
|
||||
seenSurfaceForms.add(text);
|
||||
}
|
||||
if (option == priorityQueue.insertWithOverflow(option)) {
|
||||
// if the current option has overflown from pq,
|
||||
// we can assume all of the successive options
|
||||
// from this shard result will be overflown as well
|
||||
Iterator<Entry.Option> it = ((CompletionSuggestion) suggestion).getOptions().iterator();
|
||||
if (it.hasNext()) {
|
||||
pq.add(new ShardOptions(it));
|
||||
}
|
||||
}
|
||||
// Dedup duplicate suggestions (based on the surface form) if skip duplicates is activated
|
||||
final CharArraySet seenSurfaceForms = leader.skipDuplicates ? new CharArraySet(leader.getSize(), false) : null;
|
||||
final Entry entry = new Entry(leaderEntry.getText(), leaderEntry.getOffset(), leaderEntry.getLength());
|
||||
final List<Entry.Option> options = entry.getOptions();
|
||||
while (pq.size() > 0) {
|
||||
ShardOptions top = pq.top();
|
||||
Entry.Option current = top.current;
|
||||
if (top.advanceToNextOption()) {
|
||||
pq.updateTop();
|
||||
} else {
|
||||
//options exhausted for this shard
|
||||
pq.pop();
|
||||
}
|
||||
if (leader.skipDuplicates == false ||
|
||||
seenSurfaceForms.add(current.getText().toString())) {
|
||||
options.add(current);
|
||||
if (options.size() >= size) {
|
||||
break;
|
||||
}
|
||||
}
|
||||
}
|
||||
final CompletionSuggestion suggestion = new CompletionSuggestion(leader.getName(), leader.getSize(), leader.skipDuplicates);
|
||||
final Entry entry = new Entry(leaderEntry.getText(), leaderEntry.getOffset(), leaderEntry.getLength());
|
||||
Collections.addAll(entry.getOptions(), priorityQueue.get());
|
||||
suggestion.addTerm(entry);
|
||||
return suggestion;
|
||||
}
|
||||
|
|
|
|||
|
|
@ -18,14 +18,7 @@
|
|||
*/
|
||||
package org.elasticsearch.search.suggest.completion;
|
||||
|
||||
import org.apache.lucene.analysis.CharArraySet;
|
||||
import org.apache.lucene.index.LeafReaderContext;
|
||||
import org.apache.lucene.search.CollectionTerminatedException;
|
||||
import org.apache.lucene.search.TotalHits;
|
||||
import org.apache.lucene.search.suggest.Lookup;
|
||||
import org.apache.lucene.search.suggest.document.TopSuggestDocs;
|
||||
import org.apache.lucene.search.suggest.document.TopSuggestDocsCollector;
|
||||
import org.apache.lucene.util.PriorityQueue;
|
||||
|
||||
import java.io.IOException;
|
||||
import java.util.ArrayList;
|
||||
|
|
@ -36,69 +29,13 @@ import java.util.Map;
|
|||
|
||||
/**
|
||||
*
|
||||
* Custom {@link TopSuggestDocsCollector} that returns top documents from the completion suggester.
|
||||
* <p>
|
||||
* TODO: this should be refactored when https://issues.apache.org/jira/browse/LUCENE-8529 is fixed.
|
||||
* Unlike the parent class, this collector uses the surface form to tie-break suggestions with identical
|
||||
* scores.
|
||||
* <p>
|
||||
* Extension of the {@link TopSuggestDocsCollector} that returns top documents from the completion suggester.
|
||||
*
|
||||
* This collector groups suggestions coming from the same document but matching different contexts
|
||||
* or surface form together. When different contexts or surface forms match the same suggestion form only
|
||||
* the best one per document (sorted by weight) is kept.
|
||||
* <p>
|
||||
* This collector is also able to filter duplicate suggestion coming from different documents.
|
||||
* In order to keep this feature fast, the de-duplication of suggestions with different contexts is done
|
||||
* only on the top N*num_contexts (where N is the number of documents to return) suggestions per segment.
|
||||
* This means that skip_duplicates will visit at most N*num_contexts suggestions per segment to find unique suggestions
|
||||
* that match the input. If more than N*num_contexts suggestions are duplicated with different contexts this collector
|
||||
* will not be able to return more than one suggestion even when N is greater than 1.
|
||||
**/
|
||||
class TopSuggestGroupDocsCollector extends TopSuggestDocsCollector {
|
||||
private final class SuggestScoreDocPriorityQueue extends PriorityQueue<TopSuggestDocs.SuggestScoreDoc> {
|
||||
/**
|
||||
* Creates a new priority queue of the specified size.
|
||||
*/
|
||||
private SuggestScoreDocPriorityQueue(int size) {
|
||||
super(size);
|
||||
}
|
||||
|
||||
@Override
|
||||
protected boolean lessThan(TopSuggestDocs.SuggestScoreDoc a, TopSuggestDocs.SuggestScoreDoc b) {
|
||||
if (a.score == b.score) {
|
||||
// tie break by completion key
|
||||
int cmp = Lookup.CHARSEQUENCE_COMPARATOR.compare(a.key, b.key);
|
||||
// prefer smaller doc id, in case of a tie
|
||||
return cmp != 0 ? cmp > 0 : a.doc > b.doc;
|
||||
}
|
||||
return a.score < b.score;
|
||||
}
|
||||
|
||||
/**
|
||||
* Returns the top N results in descending order.
|
||||
*/
|
||||
public TopSuggestDocs.SuggestScoreDoc[] getResults() {
|
||||
int size = size();
|
||||
TopSuggestDocs.SuggestScoreDoc[] res = new TopSuggestDocs.SuggestScoreDoc[size];
|
||||
for (int i = size - 1; i >= 0; i--) {
|
||||
res[i] = pop();
|
||||
}
|
||||
return res;
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
private final SuggestScoreDocPriorityQueue priorityQueue;
|
||||
private final int num;
|
||||
|
||||
/** Only set if we are deduplicating hits: holds all per-segment hits until the end, when we dedup them */
|
||||
private final List<TopSuggestDocs.SuggestScoreDoc> pendingResults;
|
||||
|
||||
/** Only set if we are deduplicating hits: holds all surface forms seen so far in the current segment */
|
||||
final CharArraySet seenSurfaceForms;
|
||||
|
||||
/** Document base offset for the current Leaf */
|
||||
protected int docBase;
|
||||
|
||||
private Map<Integer, List<CharSequence>> docContexts = new HashMap<>();
|
||||
|
||||
/**
|
||||
|
|
@ -108,125 +45,24 @@ class TopSuggestGroupDocsCollector extends TopSuggestDocsCollector {
|
|||
* with corresponding document and weight
|
||||
*/
|
||||
TopSuggestGroupDocsCollector(int num, boolean skipDuplicates) {
|
||||
super(1, skipDuplicates);
|
||||
if (num <= 0) {
|
||||
throw new IllegalArgumentException("'num' must be > 0");
|
||||
}
|
||||
this.num = num;
|
||||
this.priorityQueue = new SuggestScoreDocPriorityQueue(num);
|
||||
if (skipDuplicates) {
|
||||
seenSurfaceForms = new CharArraySet(num, false);
|
||||
pendingResults = new ArrayList<>();
|
||||
} else {
|
||||
seenSurfaceForms = null;
|
||||
pendingResults = null;
|
||||
}
|
||||
super(num, skipDuplicates);
|
||||
}
|
||||
|
||||
/**
|
||||
* Returns the contexts associated with the provided <code>doc</code>.
|
||||
*/
|
||||
public List<CharSequence> getContexts(int doc) {
|
||||
List<CharSequence> getContexts(int doc) {
|
||||
return docContexts.getOrDefault(doc, Collections.emptyList());
|
||||
}
|
||||
|
||||
@Override
|
||||
protected boolean doSkipDuplicates() {
|
||||
return seenSurfaceForms != null;
|
||||
}
|
||||
|
||||
@Override
|
||||
public int getCountToCollect() {
|
||||
return num;
|
||||
}
|
||||
|
||||
@Override
|
||||
protected void doSetNextReader(LeafReaderContext context) throws IOException {
|
||||
docBase = context.docBase;
|
||||
if (seenSurfaceForms != null) {
|
||||
seenSurfaceForms.clear();
|
||||
// NOTE: this also clears the priorityQueue:
|
||||
for (TopSuggestDocs.SuggestScoreDoc hit : priorityQueue.getResults()) {
|
||||
pendingResults.add(hit);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
@Override
|
||||
public void collect(int docID, CharSequence key, CharSequence context, float score) throws IOException {
|
||||
int globalDoc = docID + docBase;
|
||||
boolean isDuplicate = docContexts.containsKey(globalDoc);
|
||||
boolean isNewDoc = docContexts.containsKey(globalDoc) == false;
|
||||
List<CharSequence> contexts = docContexts.computeIfAbsent(globalDoc, k -> new ArrayList<>());
|
||||
if (context != null) {
|
||||
contexts.add(context);
|
||||
}
|
||||
if (isDuplicate) {
|
||||
return;
|
||||
}
|
||||
TopSuggestDocs.SuggestScoreDoc current = new TopSuggestDocs.SuggestScoreDoc(globalDoc, key, context, score);
|
||||
if (current == priorityQueue.insertWithOverflow(current)) {
|
||||
// if the current SuggestScoreDoc has overflown from pq,
|
||||
// we can assume all of the successive collections from
|
||||
// this leaf will be overflown as well
|
||||
// TODO: reuse the overflow instance?
|
||||
throw new CollectionTerminatedException();
|
||||
contexts.add(context);
|
||||
if (isNewDoc) {
|
||||
super.collect(docID, key, context, score);
|
||||
}
|
||||
}
|
||||
|
||||
@Override
|
||||
public TopSuggestDocs get() throws IOException {
|
||||
|
||||
TopSuggestDocs.SuggestScoreDoc[] suggestScoreDocs;
|
||||
|
||||
if (seenSurfaceForms != null) {
|
||||
// NOTE: this also clears the priorityQueue:
|
||||
for (TopSuggestDocs.SuggestScoreDoc hit : priorityQueue.getResults()) {
|
||||
pendingResults.add(hit);
|
||||
}
|
||||
|
||||
// Deduplicate all hits: we already dedup'd efficiently within each segment by
|
||||
// truncating the FST top paths search, but across segments there may still be dups:
|
||||
seenSurfaceForms.clear();
|
||||
|
||||
// TODO: we could use a priority queue here to make cost O(N * log(num)) instead of O(N * log(N)), where N = O(num *
|
||||
// numSegments), but typically numSegments is smallish and num is smallish so this won't matter much in practice:
|
||||
|
||||
Collections.sort(pendingResults,
|
||||
(a, b) -> {
|
||||
// sort by higher score
|
||||
int cmp = Float.compare(b.score, a.score);
|
||||
if (cmp == 0) {
|
||||
// tie break by completion key
|
||||
cmp = Lookup.CHARSEQUENCE_COMPARATOR.compare(a.key, b.key);
|
||||
if (cmp == 0) {
|
||||
// prefer smaller doc id, in case of a tie
|
||||
cmp = Integer.compare(a.doc, b.doc);
|
||||
}
|
||||
}
|
||||
return cmp;
|
||||
});
|
||||
|
||||
List<TopSuggestDocs.SuggestScoreDoc> hits = new ArrayList<>();
|
||||
|
||||
for (TopSuggestDocs.SuggestScoreDoc hit : pendingResults) {
|
||||
if (seenSurfaceForms.contains(hit.key) == false) {
|
||||
seenSurfaceForms.add(hit.key);
|
||||
hits.add(hit);
|
||||
if (hits.size() == num) {
|
||||
break;
|
||||
}
|
||||
}
|
||||
}
|
||||
suggestScoreDocs = hits.toArray(new TopSuggestDocs.SuggestScoreDoc[0]);
|
||||
} else {
|
||||
suggestScoreDocs = priorityQueue.getResults();
|
||||
}
|
||||
|
||||
if (suggestScoreDocs.length > 0) {
|
||||
return new TopSuggestDocs(new TotalHits(suggestScoreDocs.length, TotalHits.Relation.EQUAL_TO), suggestScoreDocs);
|
||||
} else {
|
||||
return TopSuggestDocs.EMPTY;
|
||||
}
|
||||
}
|
||||
|
||||
}
|
||||
|
|
|
|||
|
|
@ -54,6 +54,7 @@ import java.io.IOException;
|
|||
import java.util.ArrayList;
|
||||
import java.util.Arrays;
|
||||
import java.util.Collection;
|
||||
import java.util.Comparator;
|
||||
import java.util.LinkedHashMap;
|
||||
import java.util.List;
|
||||
import java.util.Locale;
|
||||
|
|
@ -890,29 +891,40 @@ public class CompletionSuggestSearchIT extends ESIntegTestCase {
|
|||
int numDocs = randomIntBetween(10, 100);
|
||||
int numUnique = randomIntBetween(1, numDocs);
|
||||
List<IndexRequestBuilder> indexRequestBuilders = new ArrayList<>();
|
||||
int[] weights = new int[numUnique];
|
||||
Integer[] termIds = new Integer[numUnique];
|
||||
for (int i = 1; i <= numDocs; i++) {
|
||||
int id = i % numUnique;
|
||||
indexRequestBuilders.add(client().prepareIndex(INDEX, TYPE, "" + i)
|
||||
termIds[id] = id;
|
||||
int weight = randomIntBetween(0, 100);
|
||||
weights[id] = Math.max(weight, weights[id]);
|
||||
String suggestion = "suggestion-" + String.format(Locale.ENGLISH, "%03d" , id);
|
||||
indexRequestBuilders.add(client().prepareIndex(INDEX, TYPE)
|
||||
.setSource(jsonBuilder()
|
||||
.startObject()
|
||||
.startObject(FIELD)
|
||||
.field("input", "suggestion" + id)
|
||||
.field("weight", id)
|
||||
.field("input", suggestion)
|
||||
.field("weight", weight)
|
||||
.endObject()
|
||||
.endObject()
|
||||
));
|
||||
}
|
||||
String[] expected = new String[numUnique];
|
||||
int sugg = numUnique - 1;
|
||||
for (int i = 0; i < numUnique; i++) {
|
||||
expected[i] = "suggestion" + sugg--;
|
||||
}
|
||||
indexRandom(true, indexRequestBuilders);
|
||||
CompletionSuggestionBuilder completionSuggestionBuilder =
|
||||
SuggestBuilders.completionSuggestion(FIELD).prefix("sugg").skipDuplicates(true).size(numUnique);
|
||||
|
||||
Arrays.sort(termIds,
|
||||
Comparator.comparingInt(o -> weights[(int) o]).reversed().thenComparingInt(a -> (int) a));
|
||||
String[] expected = new String[numUnique];
|
||||
for (int i = 0; i < termIds.length; i++) {
|
||||
expected[i] = "suggestion-" + String.format(Locale.ENGLISH, "%03d" , termIds[i]);
|
||||
}
|
||||
CompletionSuggestionBuilder completionSuggestionBuilder = SuggestBuilders.completionSuggestion(FIELD)
|
||||
.prefix("sugg")
|
||||
.skipDuplicates(true)
|
||||
.size(numUnique);
|
||||
|
||||
SearchResponse searchResponse = client().prepareSearch(INDEX)
|
||||
.suggest(new SuggestBuilder().addSuggestion("suggestions", completionSuggestionBuilder)).get();
|
||||
.suggest(new SuggestBuilder().addSuggestion("suggestions", completionSuggestionBuilder))
|
||||
.get();
|
||||
assertSuggestions(searchResponse, true, "suggestions", expected);
|
||||
}
|
||||
|
||||
|
|
|
|||
|
|
@ -26,7 +26,9 @@ import org.elasticsearch.test.ESTestCase;
|
|||
import java.util.ArrayList;
|
||||
import java.util.Collections;
|
||||
import java.util.List;
|
||||
import java.util.stream.Collectors;
|
||||
|
||||
import static org.elasticsearch.search.suggest.Suggest.COMPARATOR;
|
||||
import static org.hamcrest.Matchers.equalTo;
|
||||
import static org.hamcrest.Matchers.lessThanOrEqualTo;
|
||||
|
||||
|
|
@ -61,4 +63,41 @@ public class CompletionSuggestionTests extends ESTestCase {
|
|||
count++;
|
||||
}
|
||||
}
|
||||
|
||||
public void testToReduceWithDuplicates() {
|
||||
List<Suggest.Suggestion<CompletionSuggestion.Entry>> shardSuggestions = new ArrayList<>();
|
||||
int nShards = randomIntBetween(2, 10);
|
||||
String name = randomAlphaOfLength(10);
|
||||
int size = randomIntBetween(10, 100);
|
||||
int totalResults = size * nShards;
|
||||
int numSurfaceForms = randomIntBetween(1, size);
|
||||
String[] surfaceForms = new String[numSurfaceForms];
|
||||
for (int i = 0; i < numSurfaceForms; i++) {
|
||||
surfaceForms[i] = randomAlphaOfLength(20);
|
||||
}
|
||||
List<CompletionSuggestion.Entry.Option> options = new ArrayList<>();
|
||||
for (int i = 0; i < nShards; i++) {
|
||||
CompletionSuggestion suggestion = new CompletionSuggestion(name, size, true);
|
||||
CompletionSuggestion.Entry entry = new CompletionSuggestion.Entry(new Text(""), 0, 0);
|
||||
suggestion.addTerm(entry);
|
||||
int maxScore = randomIntBetween(totalResults, totalResults*2);
|
||||
for (int j = 0; j < size; j++) {
|
||||
String surfaceForm = randomFrom(surfaceForms);
|
||||
CompletionSuggestion.Entry.Option newOption =
|
||||
new CompletionSuggestion.Entry.Option(j, new Text(surfaceForm), maxScore - j, Collections.emptyMap());
|
||||
entry.addOption(newOption);
|
||||
options.add(newOption);
|
||||
}
|
||||
shardSuggestions.add(suggestion);
|
||||
}
|
||||
List<CompletionSuggestion.Entry.Option> expected = options.stream()
|
||||
.sorted(COMPARATOR)
|
||||
.distinct()
|
||||
.limit(size)
|
||||
.collect(Collectors.toList());
|
||||
CompletionSuggestion reducedSuggestion = CompletionSuggestion.reduceTo(shardSuggestions);
|
||||
assertNotNull(reducedSuggestion);
|
||||
assertThat(reducedSuggestion.getOptions().size(), lessThanOrEqualTo(size));
|
||||
assertEquals(expected, reducedSuggestion.getOptions());
|
||||
}
|
||||
}
|
||||
|
|
|
|||
Loading…
Reference in New Issue