protected virtual BaseQuery HandleNot(NotNode node, ElasticSearchQueryMapperState state)
{
var query = Handle(node.Operand, state);
//TODO: this works for basic term queries. While the LuceneProvider is somewhat the same, the SolrProvider had extra logic presumably to handle specific cases.
return(!query);
}
public override ElasticSearchQuery MapQuery(IndexQuery query)
{
var state = new ElasticSearchQueryMapperState();
var mappedQuery = Handle(query.RootNode, state);
return(new ElasticSearchQuery(mappedQuery, state.FilterQuery, state.AdditionalQueryMethods, state.VirtualFieldProcessors, state.FacetQueries));
}
protected virtual BaseQuery HandleOr(OrNode node, ElasticSearchQueryMapperState state)
{
//TODO: the code below specifically handles IsNullOrEmpty method, seems like a lot to handle one case... is there a better place for it?
//Lucene provider doesn't have this code, but Solr provider does
if (node.LeftNode.NodeType == QueryNodeType.Equal && node.RightNode.NodeType == QueryNodeType.Equal)
{
var leftNode = (EqualNode)node.LeftNode;
var rightNode = (EqualNode)node.RightNode;
if (leftNode.RightNode.NodeType == QueryNodeType.Constant && rightNode.RightNode.NodeType == QueryNodeType.Constant)
{
var leftNodeValue = ((ConstantNode)leftNode.RightNode).Value;
var rightNodeValue = ((ConstantNode)rightNode.RightNode).Value;
if ((string)leftNodeValue == string.Empty && rightNodeValue == null)
{
var fieldName = ((FieldNode)leftNode.LeftNode).FieldKey;
//TODO: this query works for 99% of items, however if a field contains a stopword and only a stopword, then it's treated as "missing" by ES
//For example, say you have an item whose "Title" field contains just the word "To" (which is a stopword), if you try to search for all items
//without a value in the "Title" field (i.e. the field is missing), ES will still return the item whose "Title" field contains just the word "To".
//There's likely a better query to use... or maybe not, maybe it can only be done with analyzers. who could say?
return(Query.Filtered(fq => fq.Filter(f => f.Missing(fieldName))));
}
}
}
var query1 = Handle(node.LeftNode, state);
var query2 = Handle(node.RightNode, state);
if (query1)
{
return(query1);
}
return(query1 | query2);
}
protected virtual BaseQuery HandleBetween(BetweenNode node, ElasticSearchQueryMapperState state)
{
var excludeLowerBound = !(node.Inclusion == Inclusion.Both || node.Inclusion == Inclusion.Lower);
var excludeUpperBound = !(node.Inclusion == Inclusion.Both || node.Inclusion == Inclusion.Upper);
var fieldName = GetFormattedFieldName(node.Field);
var lowerBound = ValueFormatter.FormatValueForIndexStorage(node.From);
var upperBound = ValueFormatter.FormatValueForIndexStorage(node.To);
return(Query.Range(delegate(RangeQueryDescriptor <dynamic> descriptor)
{
descriptor.From(lowerBound.ToString());
if (excludeLowerBound)
{
descriptor.FromExclusive();
}
descriptor.To(upperBound.ToString());
if (excludeUpperBound)
{
descriptor.ToExclusive();
}
descriptor.OnField(fieldName);
}));
}
protected virtual BaseQuery HandleStartsWith(StartsWithNode node, ElasticSearchQueryMapperState state)
{
var fieldName = GetFormattedFieldName(node);
var valueNode = QueryHelper.GetValueNode <string>(node);
var queryValue = ValueFormatter.FormatValueForIndexStorage(valueNode.Value);
return(Query.Prefix(fieldName, queryValue.ToString()));
}
protected BaseQuery HandleWildcardMatch(WildcardMatchNode node, ElasticSearchQueryMapperState mappingState)
{
var fieldName = GetFormattedFieldName(node);
var valueNode = QueryHelper.GetValueNode <string>(node);
var queryValue = ValueFormatter.FormatValueForIndexStorage(valueNode.Value);
//TODO: same as the HandleContains method... is there a better way to do this without wildcard queries?
return(Query.Wildcard(fieldName, queryValue.ToString()));
}
protected virtual BaseQuery HandleEndsWith(EndsWithNode node, ElasticSearchQueryMapperState state)
{
var fieldName = GetFormattedFieldName(node);
var valueNode = QueryHelper.GetValueNode <string>(node);
var queryValue = ValueFormatter.FormatValueForIndexStorage(valueNode.Value);
//TODO: same as the HandleContains method... is there a better way to do this without wildcard queries?
return(Query.Wildcard(fieldName, "*" + queryValue));
}
protected virtual BaseQuery HandleField(FieldNode node, ElasticSearchQueryMapperState state)
{
if (node.FieldType != typeof(bool))
{
throw new NotSupportedException(string.Format("The query node type '{0}' is not supported in this context.", node.NodeType));
}
var fieldName = GetFormattedFieldName(node.FieldKey);
return(Query.Term(fieldName, true.ToString()));
}
protected virtual BaseQuery HandleAnd(AndNode node, ElasticSearchQueryMapperState state)
{
var query1 = Handle(node.LeftNode, state);
var query2 = Handle(node.RightNode, state);
if (!query1)
{
return(query1);
}
return(query1 & query2);
}
protected virtual BaseQuery HandleLessThanOrEqual(LessThanOrEqualNode node, ElasticSearchQueryMapperState state)
{
BaseQuery query;
var fieldNode = node.GetFieldNode();
var valueNode = QueryHelper.GetValueNode(node, fieldNode.FieldType);
if (ProcessAsVirtualField(fieldNode, valueNode, node.Boost, ComparisonType.LessThanOrEqual, state, out query))
{
return(query);
}
return(HandleLessThanOrEqual(fieldNode.FieldKey, valueNode.Value, node.Boost));
}
protected BaseQuery HandleLike(LikeNode node, ElasticSearchQueryMapperState mappingState)
{
var fieldName = GetFormattedFieldName(node);
var valueNode = QueryHelper.GetValueNode <string>(node);
var formattedValue = ValueFormatter.FormatValueForIndexStorage(valueNode.Value);
var query =
Query.Fuzzy(
descriptor =>
descriptor.OnField(fieldName)
.MinSimilarity(node.MinimumSimilarity)
.Like(formattedValue.ToStringOrEmpty())
.Boost(node.Boost));
return(query);
}
protected virtual BaseQuery HandleContains(ContainsNode node, ElasticSearchQueryMapperState state)
{
var fieldName = GetFormattedFieldName(node);
var valueNode = QueryHelper.GetValueNode <string>(node);
//wildcard query values should be lowercase, even if the value is stored case-sensitively
//not sure why it needs to be this way - and there's probably some way to make it work either way using analyzers...
var queryValue = ValueFormatter.FormatValueForIndexStorage(valueNode.Value).ToString().ToLowerInvariant();
//TODO: evidently, wildcard queries do not scale well for large indexes
//therefore, this query should be replaced with something else.
//rumor has it the use of n-gram analyzers and regular term queries works better than wildcard queries.
//question is, how to do this in generic terms so that it works "out of the box"?
//probably an easy way, just not today...
return(Query.Wildcard(fieldName, "*" + queryValue + "*"));
}
/// <summary>
/// Regex query not currently supported in NEST, and there's no straightforward way to append raw query text to any parsed queries.
/// </summary>
/// <param name="node"></param>
/// <param name="state"></param>
/// <returns></returns>
protected virtual BaseQuery HandleMatches(MatchesNode node, ElasticSearchQueryMapperState state)
{
//var fieldName = GetFormattedFieldName(node);
//var valueNode = QueryHelper.GetValueNode<string>(node);
//var formattedValue = ValueFormatter.FormatValueForIndexStorage(valueNode.Value);
//using (var sw = new StringWriter())
//{
// using (var jw = new JsonTextWriter(sw))
// {
// jw.WriteStartObject();
// jw.WritePropertyName("regexp");
// jw.WriteStartObject();
// jw.WritePropertyName(fieldName);
// jw.WriteStartObject();
// jw.WritePropertyName("value");
// jw.WriteValue(formattedValue);
// if (Math.Abs(node.Boost - 1f) > float.Epsilon)
// {
// jw.WritePropertyName("boost");
// jw.WriteValue(node.Boost);
// }
// jw.WriteEndObject();
// if (node.RegexOptions != null && !string.IsNullOrEmpty(node.RegexOptions.ToString()))
// {
// jw.WritePropertyName("flags");
// jw.WriteValue(node.RegexOptions);
// }
// jw.WriteEndObject();
// jw.WriteEndObject();
// }
// var rawQuery = sw.ToString();
// return rawQuery;
//}
//rawQuery = "{\"regexp\" : { \"" + fieldName + "\" : { \"value\" : \"" + formattedValue + "\", \"flags\" : \"" + node.RegexOptions + "\" } }";
throw new NotImplementedException("Matches expression is not implemented (i.e. no regex queries)");
}
protected virtual BaseQuery HandleWhere(WhereNode node, ElasticSearchQueryMapperState state)
{
var query1 = Handle(node.PredicateNode, state);
var query2 = Handle(node.SourceNode, state);
if (query1 == Query.MatchAll() && query2 == Query.MatchAll())
{
return(query1);
}
if (query1 == Query.MatchAll() || query2 == Query.MatchAll())
{
if (query1 != Query.MatchAll())
{
return(query1);
}
if (query2 != Query.MatchAll())
{
return(query2);
}
}
return(query1 & query2);
}
protected virtual void StripFacetPivotOn(FacetPivotOnNode node, ElasticSearchQueryMapperState state)
{
state.FacetQueries.Add(new FacetQuery(null, node.Fields, node.MinimumNumberOfDocuments, node.FilterValues));
}
protected virtual BaseQuery HandleMatchNone(MatchNoneNode node, ElasticSearchQueryMapperState state)
{
//TODO: does this actually work? not sure about the syntax...
return(!Query.MatchAll());
//return Query.Bool(descriptor => descriptor.MustNot(new[] { Query.MatchAll() }));
}
protected virtual BaseQuery HandleMatchAll(MatchAllNode node, ElasticSearchQueryMapperState state)
{
return(Query.MatchAll());
}
protected virtual BaseQuery HandleFilter(FilterNode node, ElasticSearchQueryMapperState state)
{
return(Handle(node.PredicateNode, state));
}
protected virtual BaseQuery Handle(QueryNode node, ElasticSearchQueryMapperState state)
{
switch (node.NodeType)
{
case QueryNodeType.All:
StripAll((AllNode)node, state.AdditionalQueryMethods);
return(Handle(((AllNode)node).SourceNode, state));
case QueryNodeType.And:
return(HandleAnd((AndNode)node, state));
case QueryNodeType.Any:
StripAny((AnyNode)node, state.AdditionalQueryMethods);
return(Handle(((AnyNode)node).SourceNode, state));
case QueryNodeType.Between:
return(HandleBetween((BetweenNode)node, state));
case QueryNodeType.Cast:
StripCast((CastNode)node, state.AdditionalQueryMethods);
return(Handle(((CastNode)node).SourceNode, state));
case QueryNodeType.Contains:
return(HandleContains((ContainsNode)node, state));
case QueryNodeType.Count:
StripCount((CountNode)node, state.AdditionalQueryMethods);
return(Handle(((CountNode)node).SourceNode, state));
case QueryNodeType.ElementAt:
StripElementAt((ElementAtNode)node, state.AdditionalQueryMethods);
return(Handle(((ElementAtNode)node).SourceNode, state));
case QueryNodeType.EndsWith:
return(HandleEndsWith((EndsWithNode)node, state));
case QueryNodeType.Equal:
return(HandleEqual((EqualNode)node, state));
case QueryNodeType.Field:
return(HandleField((FieldNode)node, state));
case QueryNodeType.First:
StripFirst((FirstNode)node, state.AdditionalQueryMethods);
return(Handle(((FirstNode)node).SourceNode, state));
case QueryNodeType.GreaterThan:
return(HandleGreaterThan((GreaterThanNode)node, state));
case QueryNodeType.GreaterThanOrEqual:
return(HandleGreaterThanOrEqual((GreaterThanOrEqualNode)node, state));
case QueryNodeType.Last:
StripLast((LastNode)node, state.AdditionalQueryMethods);
return(Handle(((LastNode)node).SourceNode, state));
case QueryNodeType.LessThan:
return(HandleLessThan((LessThanNode)node, state));
case QueryNodeType.LessThanOrEqual:
return(HandleLessThanOrEqual((LessThanOrEqualNode)node, state));
case QueryNodeType.MatchAll:
return(HandleMatchAll((MatchAllNode)node, state));
case QueryNodeType.MatchNone:
return(HandleMatchNone((MatchNoneNode)node, state));
case QueryNodeType.Max:
StripMax((MaxNode)node, state.AdditionalQueryMethods);
return(Handle(((MaxNode)node).SourceNode, state));
case QueryNodeType.Min:
StripMin((MinNode)node, state.AdditionalQueryMethods);
return(Handle(((MinNode)node).SourceNode, state));
case QueryNodeType.Not:
return(HandleNot((NotNode)node, state));
case QueryNodeType.Or:
return(HandleOr((OrNode)node, state));
case QueryNodeType.OrderBy:
StripOrderBy((OrderByNode)node, state.AdditionalQueryMethods);
return(Handle(((OrderByNode)node).SourceNode, state));
case QueryNodeType.Select:
StripSelect((SelectNode)node, state.AdditionalQueryMethods);
return(Handle(((SelectNode)node).SourceNode, state));
case QueryNodeType.Single:
StripSingle((SingleNode)node, state.AdditionalQueryMethods);
return(Handle(((SingleNode)node).SourceNode, state));
case QueryNodeType.Skip:
StripSkip((SkipNode)node, state.AdditionalQueryMethods);
return(Handle(((SkipNode)node).SourceNode, state));
case QueryNodeType.StartsWith:
return(HandleStartsWith((StartsWithNode)node, state));
case QueryNodeType.Take:
StripTake((TakeNode)node, state.AdditionalQueryMethods);
return(Handle(((TakeNode)node).SourceNode, state));
case QueryNodeType.Where:
return(HandleWhere((WhereNode)node, state));
case QueryNodeType.Matches:
return(HandleMatches((MatchesNode)node, state));
case QueryNodeType.Filter:
if (state.FilterQuery != null)
{
var filterQuery = state.FilterQuery;
state.FilterQuery = filterQuery & HandleFilter((FilterNode)node, state);
break;
}
state.FilterQuery = HandleFilter((FilterNode)node, state);
break;
case QueryNodeType.GetResults:
StripGetResults((GetResultsNode)node, state.AdditionalQueryMethods);
return(Handle(((GetResultsNode)node).SourceNode, state));
case QueryNodeType.GetFacets:
StripGetFacets((GetFacetsNode)node, state.AdditionalQueryMethods);
return(Handle(((GetFacetsNode)node).SourceNode, state));
case QueryNodeType.FacetOn:
StripFacetOn((FacetOnNode)node, state);
return(Handle(((FacetOnNode)node).SourceNode, state));
case QueryNodeType.FacetPivotOn:
StripFacetPivotOn((FacetPivotOnNode)node, state);
return(Handle(((FacetPivotOnNode)node).SourceNode, state));
case QueryNodeType.WildcardMatch:
return(HandleWildcardMatch((WildcardMatchNode)node, state));
case QueryNodeType.Like:
return(HandleLike((LikeNode)node, state));
default:
throw new NotSupportedException(string.Format("Unknown query node type: '{0}'", node.NodeType));
}
return(Handle(((FilterNode)node).SourceNode, state));
}
protected virtual bool ProcessAsVirtualField(FieldNode fieldNode, ConstantNode valueNode, float boost, ComparisonType comparison, ElasticSearchQueryMapperState state, out BaseQuery query)
{
query = null;
if (_fieldQueryTranslators == null)
{
return(false);
}
var translator = _fieldQueryTranslators.GetTranslator(fieldNode.FieldKey.ToLowerInvariant());
if (translator == null)
{
return(false);
}
var formattedValue = ValueFormatter.FormatValueForIndexStorage(valueNode.Value);
var fieldQuery = translator.TranslateFieldQuery(fieldNode.FieldKey, formattedValue, comparison, FieldNameTranslator); //TODO: does the fieldNode.FieldKey value need to be formatted here?
if (fieldQuery == null)
{
return(false);
}
var queryList = new List <BaseQuery>();
if (fieldQuery.FieldComparisons != null)
{
foreach (var tuple in fieldQuery.FieldComparisons)
{
var indexFieldName = FieldNameTranslator.GetIndexFieldName(tuple.Item1);
switch (tuple.Item3)
{
case ComparisonType.Equal:
queryList.Add(HandleEqual(indexFieldName, tuple.Item2, boost));
break;
case ComparisonType.LessThan:
queryList.Add(HandleLessThan(indexFieldName, tuple.Item2, boost));
break;
case ComparisonType.LessThanOrEqual:
queryList.Add(HandleLessThanOrEqual(indexFieldName, tuple.Item2, boost));
break;
case ComparisonType.GreaterThan:
queryList.Add(HandleGreaterThan(indexFieldName, tuple.Item2, boost));
break;
case ComparisonType.GreaterThanOrEqual:
queryList.Add(HandleGreaterThanOrEqual(indexFieldName, tuple.Item2, boost));
break;
default:
throw new InvalidOperationException("Unknown comparison type: " + tuple.Item3);
}
}
foreach (var q in queryList)
{
if (query == null)
{
query = q;
}
else
{
query &= q;
}
}
}
if (fieldQuery.QueryMethods != null)
{
foreach (var method in fieldQuery.QueryMethods)
{
state.AdditionalQueryMethods.Add(method);
}
}
state.VirtualFieldProcessors.Add(translator);
return(true);
}