private AggregatorFactories parseAggregators(
XContentParser parser, SearchContext context, int level) throws IOException {
Matcher validAggMatcher = VALID_AGG_NAME.matcher("");
AggregatorFactories.Builder factories = new AggregatorFactories.Builder();
XContentParser.Token token = null;
while ((token = parser.nextToken()) != XContentParser.Token.END_OBJECT) {
if (token != XContentParser.Token.FIELD_NAME) {
throw new SearchParseException(
context,
"Unexpected token "
+ token
+ " in [aggs]: aggregations definitions must start with the name of the aggregation.");
}
final String aggregationName = parser.currentName();
if (!validAggMatcher.reset(aggregationName).matches()) {
throw new SearchParseException(
context,
"Invalid aggregation name ["
+ aggregationName
+ "]. Aggregation names must be alpha-numeric and can only contain '_' and '-'");
}
token = parser.nextToken();
if (token != XContentParser.Token.START_OBJECT) {
throw new SearchParseException(
context,
"Aggregation definition for ["
+ aggregationName
+ " starts with a ["
+ token
+ "], expected a ["
+ XContentParser.Token.START_OBJECT
+ "].");
}
AggregatorFactory factory = null;
AggregatorFactories subFactories = null;
while ((token = parser.nextToken()) != XContentParser.Token.END_OBJECT) {
if (token != XContentParser.Token.FIELD_NAME) {
throw new SearchParseException(
context,
"Expected ["
+ XContentParser.Token.FIELD_NAME
+ "] under a ["
+ XContentParser.Token.START_OBJECT
+ "], but got a ["
+ token
+ "] in ["
+ aggregationName
+ "]");
}
final String fieldName = parser.currentName();
token = parser.nextToken();
if ("aggregations_binary".equals(fieldName)) {
if (subFactories != null) {
throw new SearchParseException(
context, "Found two sub aggregation definitions under [" + aggregationName + "]");
}
XContentParser binaryParser = null;
if (token == XContentParser.Token.VALUE_STRING
|| token == XContentParser.Token.VALUE_EMBEDDED_OBJECT) {
byte[] source = parser.binaryValue();
binaryParser = XContentFactory.xContent(source).createParser(source);
} else {
throw new SearchParseException(
context,
"Expected ["
+ XContentParser.Token.VALUE_STRING
+ " or "
+ XContentParser.Token.VALUE_EMBEDDED_OBJECT
+ "] for ["
+ fieldName
+ "], but got a ["
+ token
+ "] in ["
+ aggregationName
+ "]");
}
XContentParser.Token binaryToken = binaryParser.nextToken();
if (binaryToken != XContentParser.Token.START_OBJECT) {
throw new SearchParseException(
context,
"Expected ["
+ XContentParser.Token.START_OBJECT
+ "] as first token when parsing ["
+ fieldName
+ "], but got a ["
+ binaryToken
+ "] in ["
+ aggregationName
+ "]");
}
subFactories = parseAggregators(binaryParser, context, level + 1);
} else if (token == XContentParser.Token.START_OBJECT) {
switch (fieldName) {
case "aggregations":
case "aggs":
if (subFactories != null) {
throw new SearchParseException(
context,
"Found two sub aggregation definitions under [" + aggregationName + "]");
}
subFactories = parseAggregators(parser, context, level + 1);
break;
default:
if (factory != null) {
throw new SearchParseException(
context,
"Found two aggregation type definitions in ["
+ aggregationName
+ "]: ["
+ factory.type
+ "] and ["
+ fieldName
+ "]");
}
Aggregator.Parser aggregatorParser = parser(fieldName);
if (aggregatorParser == null) {
throw new SearchParseException(
context,
"Could not find aggregator type ["
+ fieldName
+ "] in ["
+ aggregationName
+ "]");
}
factory = aggregatorParser.parse(aggregationName, parser, context);
}
} else {
throw new SearchParseException(
context,
"Expected ["
+ XContentParser.Token.START_OBJECT
+ "] under ["
+ fieldName
+ "], but got a ["
+ token
+ "] in ["
+ aggregationName
+ "]");
}
}
if (factory == null) {
throw new SearchParseException(
context, "Missing definition for aggregation [" + aggregationName + "]");
}
if (subFactories != null) {
factory.subFactories(subFactories);
}
if (level == 0) {
factory.validate();
}
factories.add(factory);
}
return factories.build();
}
public void testResetRootDocId() throws Exception {
Directory directory = newDirectory();
IndexWriterConfig iwc = new IndexWriterConfig(null);
iwc.setMergePolicy(NoMergePolicy.INSTANCE);
RandomIndexWriter indexWriter = new RandomIndexWriter(random(), directory, iwc);
List<Document> documents = new ArrayList<>();
// 1 segment with, 1 root document, with 3 nested sub docs
Document document = new Document();
document.add(
new Field(UidFieldMapper.NAME, "type#1", UidFieldMapper.Defaults.NESTED_FIELD_TYPE));
document.add(
new Field(TypeFieldMapper.NAME, "__nested_field", TypeFieldMapper.Defaults.FIELD_TYPE));
documents.add(document);
document = new Document();
document.add(
new Field(UidFieldMapper.NAME, "type#1", UidFieldMapper.Defaults.NESTED_FIELD_TYPE));
document.add(
new Field(TypeFieldMapper.NAME, "__nested_field", TypeFieldMapper.Defaults.FIELD_TYPE));
documents.add(document);
document = new Document();
document.add(
new Field(UidFieldMapper.NAME, "type#1", UidFieldMapper.Defaults.NESTED_FIELD_TYPE));
document.add(
new Field(TypeFieldMapper.NAME, "__nested_field", TypeFieldMapper.Defaults.FIELD_TYPE));
documents.add(document);
document = new Document();
document.add(new Field(UidFieldMapper.NAME, "type#1", UidFieldMapper.Defaults.FIELD_TYPE));
document.add(new Field(TypeFieldMapper.NAME, "test", TypeFieldMapper.Defaults.FIELD_TYPE));
documents.add(document);
indexWriter.addDocuments(documents);
indexWriter.commit();
documents.clear();
// 1 segment with:
// 1 document, with 1 nested subdoc
document = new Document();
document.add(
new Field(UidFieldMapper.NAME, "type#2", UidFieldMapper.Defaults.NESTED_FIELD_TYPE));
document.add(
new Field(TypeFieldMapper.NAME, "__nested_field", TypeFieldMapper.Defaults.FIELD_TYPE));
documents.add(document);
document = new Document();
document.add(new Field(UidFieldMapper.NAME, "type#2", UidFieldMapper.Defaults.FIELD_TYPE));
document.add(new Field(TypeFieldMapper.NAME, "test", TypeFieldMapper.Defaults.FIELD_TYPE));
documents.add(document);
indexWriter.addDocuments(documents);
documents.clear();
// and 1 document, with 1 nested subdoc
document = new Document();
document.add(
new Field(UidFieldMapper.NAME, "type#3", UidFieldMapper.Defaults.NESTED_FIELD_TYPE));
document.add(
new Field(TypeFieldMapper.NAME, "__nested_field", TypeFieldMapper.Defaults.FIELD_TYPE));
documents.add(document);
document = new Document();
document.add(new Field(UidFieldMapper.NAME, "type#3", UidFieldMapper.Defaults.FIELD_TYPE));
document.add(new Field(TypeFieldMapper.NAME, "test", TypeFieldMapper.Defaults.FIELD_TYPE));
documents.add(document);
indexWriter.addDocuments(documents);
indexWriter.commit();
indexWriter.close();
IndexService indexService = createIndex("test");
DirectoryReader directoryReader = DirectoryReader.open(directory);
directoryReader =
ElasticsearchDirectoryReader.wrap(directoryReader, new ShardId(indexService.index(), 0));
IndexSearcher searcher = new IndexSearcher(directoryReader);
indexService
.mapperService()
.merge(
"test",
new CompressedXContent(
PutMappingRequest.buildFromSimplifiedDef("test", "nested_field", "type=nested")
.string()),
MapperService.MergeReason.MAPPING_UPDATE,
false);
SearchContext searchContext = createSearchContext(indexService);
AggregationContext context = new AggregationContext(searchContext);
AggregatorFactories.Builder builder = AggregatorFactories.builder();
NestedAggregatorBuilder factory = new NestedAggregatorBuilder("test", "nested_field");
builder.addAggregator(factory);
AggregatorFactories factories = builder.build(context, null);
searchContext.aggregations(new SearchContextAggregations(factories));
Aggregator[] aggs = factories.createTopLevelAggregators();
BucketCollector collector = BucketCollector.wrap(Arrays.asList(aggs));
collector.preCollection();
// A regular search always exclude nested docs, so we use NonNestedDocsFilter.INSTANCE here
// (otherwise MatchAllDocsQuery would be sufficient)
// We exclude root doc with uid type#2, this will trigger the bug if we don't reset the root doc
// when we process a new segment, because
// root doc type#3 and root doc type#1 have the same segment docid
BooleanQuery.Builder bq = new BooleanQuery.Builder();
bq.add(Queries.newNonNestedFilter(), Occur.MUST);
bq.add(new TermQuery(new Term(UidFieldMapper.NAME, "type#2")), Occur.MUST_NOT);
searcher.search(new ConstantScoreQuery(bq.build()), collector);
collector.postCollection();
Nested nested = (Nested) aggs[0].buildAggregation(0);
// The bug manifests if 6 docs are returned, because currentRootDoc isn't reset the previous
// child docs from the first segment are emitted as hits.
assertThat(nested.getDocCount(), equalTo(4L));
directoryReader.close();
directory.close();
}