Handle leniency for phrase query on a field indexed without positions (#26388)

This change rewrite phrase query built on a field indexed without positions
to match_no_docs query when the `lenient` option is set to true.
This change affects all full text queries.
This commit is contained in:
Jim Ferenczi 2017-08-25 16:41:01 +02:00 committed by GitHub
parent ad8f359deb
commit 74cd32942a
4 changed files with 54 additions and 9 deletions

View File

@ -22,6 +22,7 @@ package org.elasticsearch.index.search;
import org.apache.lucene.analysis.Analyzer;
import org.apache.lucene.analysis.miscellaneous.DisableGraphAttribute;
import org.apache.lucene.analysis.TokenStream;
import org.apache.lucene.index.IndexOptions;
import org.apache.lucene.index.Term;
import org.apache.lucene.queries.ExtendedCommonTermsQuery;
import org.apache.lucene.search.BooleanClause;
@ -239,6 +240,10 @@ public class MatchQuery {
}
}
private boolean hasPositions(MappedFieldType fieldType) {
return fieldType.indexOptions().compareTo(IndexOptions.DOCS_AND_FREQS_AND_POSITIONS) >= 0;
}
public Query parse(Type type, String fieldName, Object value) throws IOException {
MappedFieldType fieldType = context.fieldMapper(fieldName);
if (fieldType == null) {
@ -263,7 +268,11 @@ public class MatchQuery {
assert analyzer != null;
MatchQueryBuilder builder = new MatchQueryBuilder(analyzer, fieldType);
builder.setEnablePositionIncrements(this.enablePositionIncrements);
builder.setAutoGenerateMultiTermSynonymsPhraseQuery(this.autoGenerateSynonymsPhraseQuery);
if (hasPositions(fieldType)) {
builder.setAutoGenerateMultiTermSynonymsPhraseQuery(this.autoGenerateSynonymsPhraseQuery);
} else {
builder.setAutoGenerateMultiTermSynonymsPhraseQuery(false);
}
Query query = null;
switch (type) {
@ -331,6 +340,20 @@ public class MatchQuery {
return blendTermsQuery(terms, mapper);
}
@Override
protected Query analyzePhrase(String field, TokenStream stream, int slop) throws IOException {
if (hasPositions(mapper) == false) {
IllegalStateException exc =
new IllegalStateException("field:[" + field + "] was indexed without position data; cannot run PhraseQuery");
if (lenient) {
return newLenientFieldQuery(field, exc);
} else {
throw exc;
}
}
return super.analyzePhrase(field, stream, slop);
}
/**
* Checks if graph analysis should be enabled for the field depending
* on the provided {@link Analyzer}

View File

@ -437,8 +437,12 @@ public class MatchQueryBuilderTests extends AbstractQueryTestCase<MatchQueryBuil
@Override
protected void initializeAdditionalMappings(MapperService mapperService) throws IOException {
mapperService.merge("doc", new CompressedXContent(PutMappingRequest.buildFromSimplifiedDef("doc",
"string_boost", "type=text,boost=4").string()), MapperService.MergeReason.MAPPING_UPDATE, false);
mapperService.merge("doc", new CompressedXContent(PutMappingRequest.buildFromSimplifiedDef(
"doc",
"string_boost", "type=text,boost=4", "string_no_pos",
"type=text,index_options=docs").string()
),
MapperService.MergeReason.MAPPING_UPDATE, false);
}
public void testMatchPhrasePrefixWithBoost() throws Exception {
@ -463,6 +467,16 @@ public class MatchQueryBuilderTests extends AbstractQueryTestCase<MatchQueryBuil
Query query = builder.toQuery(context);
assertThat(query, instanceOf(MultiPhrasePrefixQuery.class));
}
}
public void testLenientPhraseQuery() throws Exception {
assumeTrue("test runs only when at least a type is registered", getCurrentTypes().length > 0);
QueryShardContext context = createShardContext();
MatchQuery b = new MatchQuery(context);
b.setLenient(true);
Query query = b.parse(Type.PHRASE, "string_no_pos", "foo bar");
assertThat(query, instanceOf(MatchNoDocsQuery.class));
assertThat(query.toString(),
containsString("field:[string_no_pos] was indexed without position data; cannot run PhraseQuery"));
}
}

View File

@ -242,16 +242,24 @@ public class QueryStringIT extends ESIntegTestCase {
}
@LuceneTestCase.AwaitsFix(bugUrl="currently can't perform phrase queries on fields that don't support positions")
public void testPhraseQueryOnFieldWithNoPositions() throws Exception {
List<IndexRequestBuilder> reqs = new ArrayList<>();
reqs.add(client().prepareIndex("test", "doc", "1").setSource("f1", "foo bar", "f4", "eggplant parmesan"));
reqs.add(client().prepareIndex("test", "doc", "2").setSource("f1", "foo bar", "f4", "chicken parmesan"));
indexRandom(true, false, reqs);
SearchResponse resp = client().prepareSearch("test").setQuery(queryStringQuery("\"eggplant parmesan\"")).get();
assertHits(resp.getHits(), "1");
assertHitCount(resp, 1L);
SearchResponse resp = client().prepareSearch("test")
.setQuery(queryStringQuery("\"eggplant parmesan\"").lenient(true)).get();
assertHitCount(resp, 0L);
Exception exc = expectThrows(Exception.class,
() -> client().prepareSearch("test").setQuery(
queryStringQuery("f4:\"eggplant parmesan\"").lenient(false)
).get()
);
IllegalStateException ise = (IllegalStateException) ExceptionsHelper.unwrap(exc, IllegalStateException.class);
assertNotNull(ise);
assertThat(ise.getMessage(), containsString("field:[f4] was indexed without position data; cannot run PhraseQuery"));
}
public void testBooleanStrictQuery() throws Exception {

View File

@ -177,8 +177,8 @@ public class SearchQueryIT extends ESIntegTestCase {
assertHitCount(searchResponse, 1L);
assertFailures(client().prepareSearch().setQuery(matchQuery("field1", "quick brown").type(Type.PHRASE).slop(0)),
RestStatus.INTERNAL_SERVER_ERROR,
containsString("field \"field1\" was indexed without position data; cannot run PhraseQuery"));
RestStatus.BAD_REQUEST,
containsString("field:[field1] was indexed without position data; cannot run PhraseQuery"));
}
// see #3521