JoinSpec

This class represents the specifications for the joining operation in search.

Joins are only possible for matching on the qualified id of an outer document and a property value within a subquery document. In the subquery documents, these values may be referred to with a property path such as "email.recipient.id" or "entityId" or a property expression. One such property expression is "this.qualifiedId()", which refers to the document's combined package, database, namespace, and id.

Note that in order for perform the join, the property referred to by getChildPropertyExpression() has to be a property with AppSearchSchema.StringPropertyConfig.getJoinableValueType() set to AppSearchSchema.StringPropertyConfig.JOINABLE_VALUE_TYPE_QUALIFIED_ID. Otherwise no documents will be joined to any SearchResult.

Take these outer query and subquery results for example:

Outer result {
   id: id1
   score: 5
 }
 Subquery result 1 {
   id: id2
   score: 2
   entityId: pkg$db/ns#id1
   notes: This is some doc
 }
 Subquery result 2 {
   id: id3
   score: 3
   entityId: pkg$db/ns#id2
   notes: This is another doc
 }
 

In this example, subquery result 1 contains a property "entityId" whose value is "pkg$db/ns#id1", referring to the outer result. If you call JoinSpec.Builder with "entityId", we will retrieve the value of the property "entityId" from the child document, which is "pkg$db#ns/id1". Let's say the qualified id of the outer result is "pkg$db#ns/id1". This would mean the subquery result 1 document will be matched to that parent document. This is done by adding a SearchResult containing the child document to the top-level parent SearchResult.getJoinedResults().

If getChildPropertyExpression() is "notes", we will check the values of the notes property in the subquery results. In subquery result 1, this values is "This is some doc", which does not equal the qualified id of the outer query result. As such, subquery result 1 will not be joined to the outer query result.

It's possible to define an advanced ranking strategy in the nested SearchSpec and also use SearchSpec.RANKING_STRATEGY_JOIN_AGGREGATE_SCORE in the outer SearchSpec. In this case, the parents will be ranked based on an aggregation, such as the sum, of the signals calculated by scoring the joined documents with the advanced ranking strategy.

In terms of scoring, if SearchSpec.RANKING_STRATEGY_JOIN_AGGREGATE_SCORE is set in SearchSpec.getRankingStrategy(), the scores of the outer SearchResults can be influenced by the ranking signals of the subquery results. For example, if the getAggregationScoringStrategy() is set to:

• AGGREGATION_SCORING_MIN_RANKING_SIGNAL, the ranking signal of the outer SearchResult will be set to the minimum of the ranking signals of the subquery results. In this case, it will be the minimum of 2 and 3, which is 2.
• AGGREGATION_SCORING_MAX_RANKING_SIGNAL, the ranking signal of the outer SearchResult will be 3.
• AGGREGATION_SCORING_AVG_RANKING_SIGNAL, the ranking signal of the outer SearchResult will be 2.5.
• AGGREGATION_SCORING_RESULT_COUNT, the ranking signal of the outer SearchResult will be 2 as there are two joined results.
• AGGREGATION_SCORING_SUM_RANKING_SIGNAL, the ranking signal of the outer SearchResult will be 5, the sum of 2 and 3.
• AGGREGATION_SCORING_OUTER_RESULT_RANKING_SIGNAL, the ranking signal of the outer SearchResult will stay as it is.

Referring to "this.childrenRankingSignals()" in the ranking signal of the outer query will return the signals calculated by scoring the joined documents using the scoring strategy in the nested SearchSpec, as in SearchResult.getRankingSignal().