sorting breaks when sorting on a field that has no value in some documents

This commit is contained in:
kimchy 2010-07-13 09:38:55 +03:00
parent e6bd3f2693
commit f0cf552bc5
5 changed files with 126 additions and 30 deletions

View File

@ -29,4 +29,8 @@ public class ShardFieldDocSortedHitQueue extends FieldDocSortedHitQueue {
super(size); super(size);
setFields(fields); setFields(fields);
} }
@Override public void setFields(SortField[] fields) {
super.setFields(fields);
}
} }

View File

@ -134,16 +134,18 @@ public class Lucene {
for (int j = 0; j < cFields.length; j++) { for (int j = 0; j < cFields.length; j++) {
byte type = in.readByte(); byte type = in.readByte();
if (type == 0) { if (type == 0) {
cFields[j] = in.readUTF(); cFields[j] = null;
} else if (type == 1) { } else if (type == 1) {
cFields[j] = in.readInt(); cFields[j] = in.readUTF();
} else if (type == 2) { } else if (type == 2) {
cFields[j] = in.readLong(); cFields[j] = in.readInt();
} else if (type == 3) { } else if (type == 3) {
cFields[j] = in.readFloat(); cFields[j] = in.readLong();
} else if (type == 4) { } else if (type == 4) {
cFields[j] = in.readDouble(); cFields[j] = in.readFloat();
} else if (type == 5) { } else if (type == 5) {
cFields[j] = in.readDouble();
} else if (type == 6) {
cFields[j] = in.readByte(); cFields[j] = in.readByte();
} else { } else {
throw new IOException("Can't match type [" + type + "]"); throw new IOException("Can't match type [" + type + "]");
@ -198,29 +200,33 @@ public class Lucene {
FieldDoc fieldDoc = (FieldDoc) doc; FieldDoc fieldDoc = (FieldDoc) doc;
out.writeVInt(fieldDoc.fields.length); out.writeVInt(fieldDoc.fields.length);
for (Comparable field : fieldDoc.fields) { for (Comparable field : fieldDoc.fields) {
if (field == null) {
out.writeByte((byte) 0);
} else {
Class type = field.getClass(); Class type = field.getClass();
if (type == String.class) { if (type == String.class) {
out.writeByte((byte) 0); out.writeByte((byte) 1);
out.writeUTF((String) field); out.writeUTF((String) field);
} else if (type == Integer.class) { } else if (type == Integer.class) {
out.writeByte((byte) 1); out.writeByte((byte) 2);
out.writeInt((Integer) field); out.writeInt((Integer) field);
} else if (type == Long.class) { } else if (type == Long.class) {
out.writeByte((byte) 2); out.writeByte((byte) 3);
out.writeLong((Long) field); out.writeLong((Long) field);
} else if (type == Float.class) { } else if (type == Float.class) {
out.writeByte((byte) 3); out.writeByte((byte) 4);
out.writeFloat((Float) field); out.writeFloat((Float) field);
} else if (type == Double.class) { } else if (type == Double.class) {
out.writeByte((byte) 4); out.writeByte((byte) 5);
out.writeDouble((Double) field); out.writeDouble((Double) field);
} else if (type == Byte.class) { } else if (type == Byte.class) {
out.writeByte((byte) 5); out.writeByte((byte) 6);
out.writeByte((Byte) field); out.writeByte((Byte) field);
} else { } else {
throw new IOException("Can't handle sort field value of type [" + type + "]"); throw new IOException("Can't handle sort field value of type [" + type + "]");
} }
} }
}
out.writeVInt(doc.doc); out.writeVInt(doc.doc);
out.writeFloat(doc.score); out.writeFloat(doc.score);

View File

@ -20,10 +20,7 @@
package org.elasticsearch.search.controller; package org.elasticsearch.search.controller;
import org.apache.lucene.index.Term; import org.apache.lucene.index.Term;
import org.apache.lucene.search.FieldDoc; import org.apache.lucene.search.*;
import org.apache.lucene.search.ScoreDoc;
import org.apache.lucene.search.ShardFieldDocSortedHitQueue;
import org.apache.lucene.search.TopFieldDocs;
import org.apache.lucene.util.PriorityQueue; import org.apache.lucene.util.PriorityQueue;
import org.elasticsearch.common.collect.Iterables; import org.elasticsearch.common.collect.Iterables;
import org.elasticsearch.common.collect.Lists; import org.elasticsearch.common.collect.Lists;
@ -83,8 +80,29 @@ public class SearchPhaseController {
} }
PriorityQueue queue; PriorityQueue queue;
if (queryResultProvider.queryResult().topDocs() instanceof TopFieldDocs) { if (queryResultProvider.queryResult().topDocs() instanceof TopFieldDocs) {
// sorting ... // sorting, first if the type is a String, chance CUSTOM to STRING so we handle nulls properly
queue = new ShardFieldDocSortedHitQueue(((TopFieldDocs) queryResultProvider.queryResult().topDocs()).fields, queueSize); // we need to accumulate for all and then filter the from TopFieldDocs fieldDocs = (TopFieldDocs) queryResultProvider.queryResult().topDocs();
for (int i = 0; i < fieldDocs.fields.length; i++) {
boolean resolvedField = false;
for (QuerySearchResultProvider resultProvider : results) {
for (ScoreDoc doc : resultProvider.queryResult().topDocs().scoreDocs) {
FieldDoc fDoc = (FieldDoc) doc;
if (fDoc.fields[i] != null) {
if (fDoc.fields[i] instanceof String) {
fieldDocs.fields[i] = new SortField(fieldDocs.fields[i].getField(), SortField.STRING, fieldDocs.fields[i].getReverse());
}
resolvedField = true;
break;
}
}
if (resolvedField) {
break;
}
}
}
queue = new ShardFieldDocSortedHitQueue(fieldDocs.fields, queueSize);
// we need to accumulate for all and then filter the from
for (QuerySearchResultProvider resultProvider : results) { for (QuerySearchResultProvider resultProvider : results) {
QuerySearchResult result = resultProvider.queryResult(); QuerySearchResult result = resultProvider.queryResult();
ScoreDoc[] scoreDocs = result.topDocs().scoreDocs; ScoreDoc[] scoreDocs = result.topDocs().scoreDocs;

View File

@ -20,6 +20,7 @@
package org.elasticsearch.test.integration.search.sort; package org.elasticsearch.test.integration.search.sort;
import org.elasticsearch.action.search.SearchResponse; import org.elasticsearch.action.search.SearchResponse;
import org.elasticsearch.action.search.ShardSearchFailure;
import org.elasticsearch.client.Client; import org.elasticsearch.client.Client;
import org.elasticsearch.search.builder.SearchSourceBuilder; import org.elasticsearch.search.builder.SearchSourceBuilder;
import org.elasticsearch.test.integration.AbstractNodesTests; import org.elasticsearch.test.integration.AbstractNodesTests;
@ -140,4 +141,70 @@ public class SimpleSortTests extends AbstractNodesTests {
assertThat((String) searchResponse.hits().getAt(0).field("id").value(), equalTo("2")); assertThat((String) searchResponse.hits().getAt(0).field("id").value(), equalTo("2"));
assertThat((String) searchResponse.hits().getAt(1).field("id").value(), equalTo("1")); assertThat((String) searchResponse.hits().getAt(1).field("id").value(), equalTo("1"));
} }
@Test public void testDocumentsWithNullValue() throws Exception {
try {
client.admin().indices().prepareDelete("test").execute().actionGet();
} catch (Exception e) {
// ignore
}
client.admin().indices().prepareCreate("test").execute().actionGet();
client.admin().cluster().prepareHealth().setWaitForGreenStatus().execute().actionGet();
client.prepareIndex("test", "type1").setSource(jsonBuilder().startObject()
.field("id", "1")
.field("svalue", "aaa")
.endObject()).execute().actionGet();
client.prepareIndex("test", "type1").setSource(jsonBuilder().startObject()
.field("id", "2")
.nullField("svalue")
.endObject()).execute().actionGet();
client.prepareIndex("test", "type1").setSource(jsonBuilder().startObject()
.field("id", "3")
.field("svalue", "bbb")
.endObject()).execute().actionGet();
client.admin().indices().prepareFlush().setRefresh(true).execute().actionGet();
SearchResponse searchResponse = client.prepareSearch()
.setQuery(matchAllQuery())
.addScriptField("id", "doc['id'].value")
.addSort("svalue", SearchSourceBuilder.Order.ASC)
.execute().actionGet();
if (searchResponse.failedShards() > 0) {
logger.warn("Failed shards:");
for (ShardSearchFailure shardSearchFailure : searchResponse.shardFailures()) {
logger.warn("-> {}", shardSearchFailure);
}
}
assertThat(searchResponse.failedShards(), equalTo(0));
assertThat(searchResponse.hits().getTotalHits(), equalTo(3l));
assertThat((String) searchResponse.hits().getAt(0).field("id").value(), equalTo("2"));
assertThat((String) searchResponse.hits().getAt(1).field("id").value(), equalTo("1"));
assertThat((String) searchResponse.hits().getAt(2).field("id").value(), equalTo("3"));
searchResponse = client.prepareSearch()
.setQuery(matchAllQuery())
.addScriptField("id", "doc['id'].value")
.addSort("svalue", SearchSourceBuilder.Order.DESC)
.execute().actionGet();
if (searchResponse.failedShards() > 0) {
logger.warn("Failed shards:");
for (ShardSearchFailure shardSearchFailure : searchResponse.shardFailures()) {
logger.warn("-> {}", shardSearchFailure);
}
}
assertThat(searchResponse.failedShards(), equalTo(0));
assertThat(searchResponse.hits().getTotalHits(), equalTo(3l));
assertThat((String) searchResponse.hits().getAt(0).field("id").value(), equalTo("3"));
assertThat((String) searchResponse.hits().getAt(1).field("id").value(), equalTo("1"));
assertThat((String) searchResponse.hits().getAt(2).field("id").value(), equalTo("2"));
}
} }

View File

@ -1,6 +1,7 @@
log4j.rootLogger=INFO, out log4j.rootLogger=INFO, out
log4j.logger.jgroups=WARN log4j.logger.jgroups=WARN
#log4j.logger.action=DEBUG
#log4j.logger.transport=TRACE #log4j.logger.transport=TRACE
#log4j.logger.discovery=TRACE #log4j.logger.discovery=TRACE
#log4j.logger.cluster.service=TRACE #log4j.logger.cluster.service=TRACE