Nested objects not deleted on "delete by query", closes #1537.

This commit is contained in:
Shay Banon 2011-12-25 13:33:02 +02:00
parent 0ea58cacfd
commit aa078788f9
4 changed files with 400 additions and 2 deletions

View File

@ -140,6 +140,10 @@ public class MapperService extends AbstractIndexComponent implements Iterable<Do
}
}
public boolean hasNested() {
return this.hasNested;
}
@Override
public UnmodifiableIterator<DocumentMapper> iterator() {
return mappers.values().iterator();

View File

@ -0,0 +1,274 @@
package org.elasticsearch.index.search.nested;
import org.apache.lucene.index.IndexReader;
import org.apache.lucene.index.Term;
import org.apache.lucene.search.*;
import org.apache.lucene.util.FixedBitSet;
import org.elasticsearch.common.lucene.docset.FixedBitDocSet;
import java.io.IOException;
import java.util.Set;
/**
* A special query that accepts a top level parent matching query, and returns all the children of that parent as
* well. This is handy when deleting by query.
*/
public class IncludeAllChildrenQuery extends Query {
private final Filter parentFilter;
private final Query parentQuery;
// If we are rewritten, this is the original childQuery we
// were passed; we use this for .equals() and
// .hashCode(). This makes rewritten query equal the
// original, so that user does not have to .rewrite() their
// query before searching:
private final Query origParentQuery;
public IncludeAllChildrenQuery(Query parentQuery, Filter parentFilter) {
this.origParentQuery = parentQuery;
this.parentQuery = parentQuery;
this.parentFilter = parentFilter;
}
IncludeAllChildrenQuery(Query origParentQuery, Query parentQuery, Filter parentFilter) {
this.origParentQuery = origParentQuery;
this.parentQuery = parentQuery;
this.parentFilter = parentFilter;
}
@Override
public Weight createWeight(Searcher searcher) throws IOException {
return new IncludeAllChildrenWeight(parentQuery, parentQuery.createWeight(searcher), parentFilter);
}
static class IncludeAllChildrenWeight extends Weight {
private final Query parentQuery;
private final Weight parentWeight;
private final Filter parentsFilter;
IncludeAllChildrenWeight(Query parentQuery, Weight parentWeight, Filter parentsFilter) {
this.parentQuery = parentQuery;
this.parentWeight = parentWeight;
this.parentsFilter = parentsFilter;
}
@Override
public Query getQuery() {
return parentQuery;
}
@Override
public float getValue() {
return parentWeight.getValue();
}
@Override
public float sumOfSquaredWeights() throws IOException {
return parentWeight.sumOfSquaredWeights() * parentQuery.getBoost() * parentQuery.getBoost();
}
@Override
public void normalize(float norm) {
parentWeight.normalize(norm * parentQuery.getBoost());
}
@Override
public Scorer scorer(IndexReader reader, boolean scoreDocsInOrder, boolean topScorer) throws IOException {
final Scorer parentScorer = parentWeight.scorer(reader, true, false);
// no matches
if (parentScorer == null) {
return null;
}
final int firstParentDoc = parentScorer.nextDoc();
if (firstParentDoc == DocIdSetIterator.NO_MORE_DOCS) {
// No matches
return null;
}
DocIdSet parents = parentsFilter.getDocIdSet(reader);
if (parents == null) {
// No matches
return null;
}
if (parents instanceof FixedBitDocSet) {
parents = ((FixedBitDocSet) parents).set();
}
if (!(parents instanceof FixedBitSet)) {
throw new IllegalStateException("parentFilter must return OpenBitSet; got " + parents);
}
return new IncludeAllChildrenScorer(this, parentScorer, (FixedBitSet) parents, firstParentDoc);
}
@Override
public Explanation explain(IndexReader reader, int doc) throws IOException {
return null;
}
@Override
public boolean scoresDocsOutOfOrder() {
return false;
}
}
static class IncludeAllChildrenScorer extends Scorer {
private final Scorer parentScorer;
private final FixedBitSet parentBits;
private int currentChildPointer = -1;
private int currentParentPointer = -1;
private int currentDoc = -1;
IncludeAllChildrenScorer(Weight weight, Scorer parentScorer, FixedBitSet parentBits, int currentParentPointer) {
super(weight);
this.parentScorer = parentScorer;
this.parentBits = parentBits;
this.currentParentPointer = currentParentPointer;
if (currentParentPointer == 0) {
currentChildPointer = 0;
} else {
this.currentChildPointer = parentBits.prevSetBit(currentParentPointer - 1);
if (currentChildPointer == -1) {
// no previous set parent, we delete from doc 0
currentChildPointer = 0;
} else {
currentChildPointer++; // we only care about children
}
}
currentDoc = currentChildPointer;
}
@Override
protected void visitSubScorers(Query parent, BooleanClause.Occur relationship, ScorerVisitor<Query, Query, Scorer> visitor) {
super.visitSubScorers(parent, relationship, visitor);
parentScorer.visitScorers(visitor);
}
@Override
public int nextDoc() throws IOException {
if (currentParentPointer == NO_MORE_DOCS) {
return (currentDoc = NO_MORE_DOCS);
}
if (currentChildPointer == currentParentPointer) {
// we need to return the current parent as well, but prepare to return
// the next set of children
currentDoc = currentParentPointer;
currentParentPointer = parentScorer.nextDoc();
if (currentParentPointer != NO_MORE_DOCS) {
currentChildPointer = parentBits.prevSetBit(currentParentPointer - 1);
if (currentChildPointer == -1) {
// no previous set parent, just set the child to the current parent
currentChildPointer = currentParentPointer;
} else {
currentChildPointer++; // we only care about children
}
}
} else {
currentDoc = currentChildPointer++;
}
assert currentDoc != -1;
return currentDoc;
}
@Override
public int advance(int target) throws IOException {
if (target == NO_MORE_DOCS) {
return (currentDoc = NO_MORE_DOCS);
}
if (target == 0) {
return nextDoc();
}
currentParentPointer = parentScorer.advance(target);
if (currentParentPointer == NO_MORE_DOCS) {
return (currentDoc = NO_MORE_DOCS);
}
if (currentParentPointer == 0) {
currentChildPointer = 0;
} else {
currentChildPointer = parentBits.prevSetBit(currentParentPointer - 1);
if (currentChildPointer == -1) {
// no previous set parent, just set the child to 0 to delete all up to the parent
currentChildPointer = 0;
} else {
currentChildPointer++; // we only care about children
}
}
currentDoc = currentChildPointer;
return currentDoc;
}
@Override
public float score() throws IOException {
return parentScorer.score();
}
@Override
public int docID() {
return currentDoc;
}
}
@Override
public void extractTerms(Set<Term> terms) {
parentQuery.extractTerms(terms);
}
@Override
public Query rewrite(IndexReader reader) throws IOException {
final Query parentRewrite = parentQuery.rewrite(reader);
if (parentRewrite != parentQuery) {
Query rewritten = new IncludeAllChildrenQuery(parentQuery, parentRewrite, parentFilter);
rewritten.setBoost(getBoost());
return rewritten;
} else {
return this;
}
}
@Override
public String toString(String field) {
return "IncludeAllChildrenQuery (" + parentQuery.toString() + ")";
}
@Override
public boolean equals(Object _other) {
if (_other instanceof IncludeAllChildrenQuery) {
final IncludeAllChildrenQuery other = (IncludeAllChildrenQuery) _other;
return origParentQuery.equals(other.origParentQuery) &&
parentFilter.equals(other.parentFilter);
} else {
return false;
}
}
@Override
public int hashCode() {
final int prime = 31;
int hash = 1;
hash = prime * hash + origParentQuery.hashCode();
hash = prime * hash + parentFilter.hashCode();
return hash;
}
@Override
public Object clone() {
return new IncludeAllChildrenQuery((Query) origParentQuery.clone(),
parentFilter);
}
}

View File

@ -52,6 +52,8 @@ import org.elasticsearch.index.merge.MergeStats;
import org.elasticsearch.index.merge.scheduler.MergeSchedulerProvider;
import org.elasticsearch.index.query.IndexQueryParserService;
import org.elasticsearch.index.refresh.RefreshStats;
import org.elasticsearch.index.search.nested.IncludeAllChildrenQuery;
import org.elasticsearch.index.search.nested.NonNestedDocsFilter;
import org.elasticsearch.index.search.stats.SearchStats;
import org.elasticsearch.index.search.stats.ShardSearchService;
import org.elasticsearch.index.settings.IndexSettings;
@ -359,6 +361,11 @@ public class InternalIndexShard extends AbstractIndexShardComponent implements I
@Override
public void deleteByQuery(Engine.DeleteByQuery deleteByQuery) throws ElasticSearchException {
writeAllowed();
if (mapperService.hasNested()) {
// we need to wrap it to delete nested docs as well...
IncludeAllChildrenQuery nestedQuery = new IncludeAllChildrenQuery(deleteByQuery.query(), indexCache.filter().cache(NonNestedDocsFilter.INSTANCE));
deleteByQuery = new Engine.DeleteByQuery(nestedQuery, deleteByQuery.source(), deleteByQuery.filteringAliases(), deleteByQuery.aliasFilter(), deleteByQuery.types());
}
if (logger.isTraceEnabled()) {
logger.trace("delete_by_query [{}]", deleteByQuery.query());
}

View File

@ -25,7 +25,7 @@ import org.elasticsearch.action.get.GetResponse;
import org.elasticsearch.action.search.SearchResponse;
import org.elasticsearch.action.search.SearchType;
import org.elasticsearch.client.Client;
import org.elasticsearch.common.settings.ImmutableSettings;
import org.elasticsearch.index.query.QueryBuilders;
import org.elasticsearch.search.facet.FacetBuilders;
import org.elasticsearch.search.facet.termsstats.TermsStatsFacet;
import org.elasticsearch.test.integration.AbstractNodesTests;
@ -35,6 +35,7 @@ import org.testng.annotations.Test;
import java.util.Arrays;
import static org.elasticsearch.common.settings.ImmutableSettings.settingsBuilder;
import static org.elasticsearch.common.xcontent.XContentFactory.jsonBuilder;
import static org.elasticsearch.index.query.FilterBuilders.nestedFilter;
import static org.elasticsearch.index.query.QueryBuilders.*;
@ -178,6 +179,118 @@ public class SimpleNestedTests extends AbstractNodesTests {
assertThat(searchResponse.hits().totalHits(), equalTo(1l));
}
@Test
public void simpleNestedDeletedByQuery1() throws Exception {
simpleNestedDeleteByQuery(3, 0);
}
@Test
public void simpleNestedDeletedByQuery2() throws Exception {
simpleNestedDeleteByQuery(3, 1);
}
@Test
public void simpleNestedDeletedByQuery3() throws Exception {
simpleNestedDeleteByQuery(3, 2);
}
private void simpleNestedDeleteByQuery(int total, int docToDelete) throws Exception {
client.admin().indices().prepareDelete().execute().actionGet();
client.admin().indices().prepareCreate("test")
.setSettings(settingsBuilder().put("index.number_of_shards", 1).put("index.referesh_interval", -1).build())
.addMapping("type1", jsonBuilder().startObject().startObject("type1").startObject("properties")
.startObject("nested1")
.field("type", "nested")
.endObject()
.endObject().endObject().endObject())
.execute().actionGet();
client.admin().cluster().prepareHealth().setWaitForGreenStatus().execute().actionGet();
for (int i = 0; i < total; i++) {
client.prepareIndex("test", "type1", Integer.toString(i)).setSource(jsonBuilder().startObject()
.field("field1", "value1")
.startArray("nested1")
.startObject()
.field("n_field1", "n_value1_1")
.field("n_field2", "n_value2_1")
.endObject()
.startObject()
.field("n_field1", "n_value1_2")
.field("n_field2", "n_value2_2")
.endObject()
.endArray()
.endObject()).execute().actionGet();
}
client.admin().indices().prepareFlush().setRefresh(true).execute().actionGet();
IndicesStatusResponse statusResponse = client.admin().indices().prepareStatus().execute().actionGet();
assertThat(statusResponse.index("test").docs().numDocs(), equalTo(total * 3));
client.prepareDeleteByQuery("test").setQuery(QueryBuilders.idsQuery("type1").ids(Integer.toString(docToDelete))).execute().actionGet();
client.admin().indices().prepareFlush().setRefresh(true).execute().actionGet();
statusResponse = client.admin().indices().prepareStatus().execute().actionGet();
assertThat(statusResponse.index("test").docs().numDocs(), equalTo((total * 3) - 3));
for (int i = 0; i < total; i++) {
assertThat(client.prepareGet("test", "type1", Integer.toString(i)).execute().actionGet().exists(), equalTo(i != docToDelete));
}
}
@Test
public void noChildrenNestedDeletedByQuery1() throws Exception {
noChildrenNestedDeleteByQuery(3, 0);
}
@Test
public void noChildrenNestedDeletedByQuery2() throws Exception {
noChildrenNestedDeleteByQuery(3, 1);
}
@Test
public void noChildrenNestedDeletedByQuery3() throws Exception {
noChildrenNestedDeleteByQuery(3, 2);
}
private void noChildrenNestedDeleteByQuery(int total, int docToDelete) throws Exception {
client.admin().indices().prepareDelete().execute().actionGet();
client.admin().indices().prepareCreate("test")
.setSettings(settingsBuilder().put("index.number_of_shards", 1).put("index.referesh_interval", -1).build())
.addMapping("type1", jsonBuilder().startObject().startObject("type1").startObject("properties")
.startObject("nested1")
.field("type", "nested")
.endObject()
.endObject().endObject().endObject())
.execute().actionGet();
client.admin().cluster().prepareHealth().setWaitForGreenStatus().execute().actionGet();
for (int i = 0; i < total; i++) {
client.prepareIndex("test", "type1", Integer.toString(i)).setSource(jsonBuilder().startObject()
.field("field1", "value1")
.endObject()).execute().actionGet();
}
client.admin().indices().prepareFlush().setRefresh(true).execute().actionGet();
IndicesStatusResponse statusResponse = client.admin().indices().prepareStatus().execute().actionGet();
assertThat(statusResponse.index("test").docs().numDocs(), equalTo(total));
client.prepareDeleteByQuery("test").setQuery(QueryBuilders.idsQuery("type1").ids(Integer.toString(docToDelete))).execute().actionGet();
client.admin().indices().prepareFlush().setRefresh(true).execute().actionGet();
statusResponse = client.admin().indices().prepareStatus().execute().actionGet();
assertThat(statusResponse.index("test").docs().numDocs(), equalTo((total) - 1));
for (int i = 0; i < total; i++) {
assertThat(client.prepareGet("test", "type1", Integer.toString(i)).execute().actionGet().exists(), equalTo(i != docToDelete));
}
}
@Test
public void multiNested() throws Exception {
client.admin().indices().prepareDelete().execute().actionGet();
@ -267,7 +380,7 @@ public class SimpleNestedTests extends AbstractNodesTests {
client.admin().indices().prepareDelete().execute().actionGet();
client.admin().indices().prepareCreate("test")
.setSettings(ImmutableSettings.settingsBuilder().put("number_of_shards", numberOfShards))
.setSettings(settingsBuilder().put("number_of_shards", numberOfShards))
.addMapping("type1", jsonBuilder().startObject().startObject("type1").startObject("properties")
.startObject("nested1")
.field("type", "nested").startObject("properties")