Join-Type implementation

Spring Data Elasticsearch supports the Join data type for creating the corresponding index mappings and for storing the relevant information.

Setting up the data

For an entity to be used in a parent child join relationship, it must have a property of type JoinField which must be annotated. Let’s assume a Statement entity where a statement may be a question, an answer, a comment or a vote (a Builder is also shown in this example, it’s not necessary, but later used in the sample code):

@Document(indexName = "statements")
@Routing("routing")                                                                       (1)
public class Statement {
    @Id
    private String id;

    @Field(type = FieldType.Text)
    private String text;

    @Field(type = FieldType.Keyword)
    private String routing;

    @JoinTypeRelations(
        relations =
            {
                @JoinTypeRelation(parent = "question", children = {"answer", "comment"}), (2)
                @JoinTypeRelation(parent = "answer", children = "vote")                   (3)
            }
    )
    private JoinField<String> relation;                                                   (4)

    private Statement() {
    }

    public static StatementBuilder builder() {
        return new StatementBuilder();
    }

    public String getId() {
        return id;
    }

    public void setId(String id) {
        this.id = id;
    }

    public String getRouting() {
        return routing;
    }

    public void setRouting(Routing routing) {
        this.routing = routing;
    }

    public String getText() {
        return text;
    }

    public void setText(String text) {
        this.text = text;
    }

    public JoinField<String> getRelation() {
        return relation;
    }

    public void setRelation(JoinField<String> relation) {
        this.relation = relation;
    }

    public static final class StatementBuilder {
        private String id;
        private String text;
        private String routing;
        private JoinField<String> relation;

        private StatementBuilder() {
        }

        public StatementBuilder withId(String id) {
            this.id = id;
            return this;
        }

        public StatementBuilder withRouting(String routing) {
            this.routing = routing;
            return this;
        }

        public StatementBuilder withText(String text) {
            this.text = text;
            return this;
        }

        public StatementBuilder withRelation(JoinField<String> relation) {
            this.relation = relation;
            return this;
        }

        public Statement build() {
            Statement statement = new Statement();
            statement.setId(id);
            statement.setRouting(routing);
            statement.setText(text);
            statement.setRelation(relation);
            return statement;
        }
    }
}
1 for routing related info see Routing values
2 a question can have answers and comments
3 an answer can have votes
4 the JoinField property is used to combine the name (question, answer, comment or vote) of the relation with the parent id. The generic type must be the same as the @Id annotated property.

Spring Data Elasticsearch will build the following mapping for this class:

{
  "statements": {
    "mappings": {
      "properties": {
        "_class": {
          "type": "text",
          "fields": {
            "keyword": {
              "type": "keyword",
              "ignore_above": 256
            }
          }
        },
        "routing": {
          "type": "keyword"
        },
        "relation": {
          "type": "join",
          "eager_global_ordinals": true,
          "relations": {
            "question": [
              "answer",
              "comment"
            ],
            "answer": "vote"
          }
        },
        "text": {
          "type": "text"
        }
      }
    }
  }
}

Storing data

Given a repository for this class the following code inserts a question, two answers, a comment and a vote:

void init() {
    repository.deleteAll();

    Statement savedWeather = repository.save(
        Statement.builder()
            .withText("How is the weather?")
            .withRelation(new JoinField<>("question"))                          (1)
            .build());

    Statement sunnyAnswer = repository.save(
        Statement.builder()
            .withText("sunny")
            .withRelation(new JoinField<>("answer", savedWeather.getId()))      (2)
            .build());

    repository.save(
        Statement.builder()
            .withText("rainy")
            .withRelation(new JoinField<>("answer", savedWeather.getId()))      (3)
            .build());

    repository.save(
        Statement.builder()
            .withText("I don't like the rain")
            .withRelation(new JoinField<>("comment", savedWeather.getId()))     (4)
            .build());

    repository.save(
        Statement.builder()
            .withText("+1 for the sun")
            ,withRouting(savedWeather.getId())
            .withRelation(new JoinField<>("vote", sunnyAnswer.getId()))         (5)
            .build());
}
1 create a question statement
2 the first answer to the question
3 the second answer
4 a comment to the question
5 a vote for the first answer, this needs to have the routing set to the weather document, see Routing values.

Retrieving data

Currently native queries must be used to query the data, so there is no support from standard repository methods. Custom Repository Implementations can be used instead.

The following code shows as an example how to retrieve all entries that have a vote (which must be answers, because only answers can have a vote) using an ElasticsearchOperations instance:

SearchHits<Statement> hasVotes() {

	Query query = NativeQuery.builder()
		.withQuery(co.elastic.clients.elasticsearch._types.query_dsl.Query.of(qb -> qb
			.hasChild(hc -> hc
				.queryName("vote")
				.query(matchAllQueryAsQuery())
				.scoreMode(ChildScoreMode.None)
			)))
		.build();

	return operations.search(query, Statement.class);
}