Miscellaneous Elasticsearch Operation Support

This chapter covers additional support for Elasticsearch operations that cannot be directly accessed via the repository interface. It is recommended to add those operations as custom implementation as described in Custom Repository Implementations .

Index settings

When creating Elasticsearch indices with Spring Data Elasticsearch different index settings can be defined by using the @Setting annotation. The following arguments are available:

useServerConfiguration does not send any settings parameters, so the Elasticsearch server configuration determines them.
settingPath refers to a JSON file defining the settings that must be resolvable in the classpath
shards the number of shards to use, defaults to 1
replicas the number of replicas, defaults to 1
refreshIntervall, defaults to "1s"
indexStoreType, defaults to "fs"

It is as well possible to define index sorting (check the linked Elasticsearch documentation for the possible field types and values):

@Document(indexName = "entities")
@Setting(
  sortFields = { "secondField", "firstField" },                                  (1)
  sortModes = { Setting.SortMode.max, Setting.SortMode.min },                    (2)
  sortOrders = { Setting.SortOrder.desc, Setting.SortOrder.asc },
  sortMissingValues = { Setting.SortMissing._last, Setting.SortMissing._first })
class Entity {
    @Nullable
    @Id private String id;

    @Nullable
    @Field(name = "first_field", type = FieldType.Keyword)
    private String firstField;

    @Nullable @Field(name = "second_field", type = FieldType.Keyword)
    private String secondField;

    // getter and setter...
}

1	when defining sort fields, use the name of the Java property (firstField), not the name that might be defined for Elasticsearch (first_field)
2	`sortModes`, `sortOrders` and `sortMissingValues` are optional, but if they are set, the number of entries must match the number of `sortFields` elements

Index Mapping

When Spring Data Elasticsearch creates the index mapping with the IndexOperations.createMapping() methods, it uses the annotations described in Mapping Annotation Overview, especially the @Field annotation. In addition to that it is possible to add the @Mapping annotation to a class. This annotation has the following properties:

mappingPath a classpath resource in JSON format; if this is not empty it is used as the mapping, no other mapping processing is done.
enabled when set to false, this flag is written to the mapping and no further processing is done.
dateDetection and numericDetection set the corresponding properties in the mapping when not set to DEFAULT.
dynamicDateFormats when this String array is not empty, it defines the date formats used for automatic date detection.
runtimeFieldsPath a classpath resource in JSON format containing the definition of runtime fields which is written to the index mappings, for example:

{
  "day_of_week": {
    "type": "keyword",
    "script": {
      "source": "emit(doc['@timestamp'].value.dayOfWeekEnum.getDisplayName(TextStyle.FULL, Locale.ROOT))"
    }
  }
}

Filter Builder

Filter Builder improves query speed.

private ElasticsearchOperations operations;

IndexCoordinates index = IndexCoordinates.of("sample-index");

Query query = NativeQuery.builder()
	.withQuery(q -> q
		.matchAll(ma -> ma))
	.withFilter( q -> q
		.bool(b -> b
			.must(m -> m
				.term(t -> t
					.field("id")
					.value(documentId))
			)))
	.build();

SearchHits<SampleEntity> sampleEntities = operations.search(query, SampleEntity.class, index);

Using Scroll For Big Result Set

Elasticsearch has a scroll API for getting big result set in chunks. This is internally used by Spring Data Elasticsearch to provide the implementations of the <T> SearchHitsIterator<T> SearchOperations.searchForStream(Query query, Class<T> clazz, IndexCoordinates index) method.

IndexCoordinates index = IndexCoordinates.of("sample-index");

Query searchQuery = NativeQuery.builder()
    .withQuery(q -> q
        .matchAll(ma -> ma))
    .withFields("message")
    .withPageable(PageRequest.of(0, 10))
    .build();

SearchHitsIterator<SampleEntity> stream = elasticsearchOperations.searchForStream(searchQuery, SampleEntity.class,
index);

List<SampleEntity> sampleEntities = new ArrayList<>();
while (stream.hasNext()) {
  sampleEntities.add(stream.next());
}

stream.close();

There are no methods in the SearchOperations API to access the scroll id, if it should be necessary to access this, the following methods of the AbstractElasticsearchTemplate can be used (this is the base implementation for the different ElasticsearchOperations implementations):

@Autowired ElasticsearchOperations operations;

AbstractElasticsearchTemplate template = (AbstractElasticsearchTemplate)operations;

IndexCoordinates index = IndexCoordinates.of("sample-index");

Query query = NativeQuery.builder()
    .withQuery(q -> q
        .matchAll(ma -> ma))
    .withFields("message")
    .withPageable(PageRequest.of(0, 10))
    .build();

SearchScrollHits<SampleEntity> scroll = template.searchScrollStart(1000, query, SampleEntity.class, index);

String scrollId = scroll.getScrollId();
List<SampleEntity> sampleEntities = new ArrayList<>();
while (scroll.hasSearchHits()) {
  sampleEntities.addAll(scroll.getSearchHits());
  scrollId = scroll.getScrollId();
  scroll = template.searchScrollContinue(scrollId, 1000, SampleEntity.class);
}
template.searchScrollClear(scrollId);

To use the Scroll API with repository methods, the return type must defined as Stream in the Elasticsearch Repository. The implementation of the method will then use the scroll methods from the ElasticsearchTemplate.

interface SampleEntityRepository extends Repository<SampleEntity, String> {

    Stream<SampleEntity> findBy();

}

Sort options

In addition to the default sort options described in Paging and Sorting, Spring Data Elasticsearch provides the class org.springframework.data.elasticsearch.core.query.Order which derives from org.springframework.data.domain.Sort.Order. It offers additional parameters that can be sent to Elasticsearch when specifying the sorting of the result (see www.elastic.co/guide/en/elasticsearch/reference/7.15/sort-search-results.html).

There also is the org.springframework.data.elasticsearch.core.query.GeoDistanceOrder class which can be used to have the result of a search operation ordered by geographical distance.

If the class to be retrieved has a GeoPoint property named location, the following Sort would sort the results by distance to the given point:

Sort.by(new GeoDistanceOrder("location", new GeoPoint(48.137154, 11.5761247)))

Runtime Fields

From version 7.12 on Elasticsearch has added the feature of runtime fields (www.elastic.co/guide/en/elasticsearch/reference/7.12/runtime.html). Spring Data Elasticsearch supports this in two ways:

Runtime field definitions in the index mappings

The first way to define runtime fields is by adding the definitions to the index mappings (see www.elastic.co/guide/en/elasticsearch/reference/7.12/runtime-mapping-fields.html). To use this approach in Spring Data Elasticsearch the user must provide a JSON file that contains the corresponding definition, for example:

Example 1. runtime-fields.json

{
  "day_of_week": {
    "type": "keyword",
    "script": {
      "source": "emit(doc['@timestamp'].value.dayOfWeekEnum.getDisplayName(TextStyle.FULL, Locale.ROOT))"
    }
  }
}

The path to this JSON file, which must be present on the classpath, must then be set in the @Mapping annotation of the entity:

@Document(indexName = "runtime-fields")
@Mapping(runtimeFieldsPath = "/runtime-fields.json")
public class RuntimeFieldEntity {
	// properties, getter, setter,...
}

Runtime fields definitions set on a Query

The second way to define runtime fields is by adding the definitions to a search query (see www.elastic.co/guide/en/elasticsearch/reference/7.12/runtime-search-request.html). The following code example shows how to do this with Spring Data Elasticsearch :

The entity used is a simple object that has a price property:

@Document(indexName = "some_index_name")
public class SomethingToBuy {

	private @Id @Nullable String id;
	@Nullable @Field(type = FieldType.Text) private String description;
	@Nullable @Field(type = FieldType.Double) private Double price;

	// getter and setter
}

The following query uses a runtime field that calculates a priceWithTax value by adding 19% to the price and uses this value in the search query to find all entities where priceWithTax is higher or equal than a given value:

RuntimeField runtimeField = new RuntimeField("priceWithTax", "double", "emit(doc['price'].value * 1.19)");
Query query = new CriteriaQuery(new Criteria("priceWithTax").greaterThanEqual(16.5));
query.addRuntimeField(runtimeField);

SearchHits<SomethingToBuy> searchHits = operations.search(query, SomethingToBuy.class);

This works with every implementation of the Query interface.

Point In Time (PIT) API

ElasticsearchOperations supports the point in time API of Elasticsearch (see www.elastic.co/guide/en/elasticsearch/reference/8.3/point-in-time-api.html). The following code snippet shows how to use this feature with a fictional Person class:

ElasticsearchOperations operations; // autowired
Duration tenSeconds = Duration.ofSeconds(10);

String pit = operations.openPointInTime(IndexCoordinates.of("person"), tenSeconds); (1)

// create query for the pit
Query query1 = new CriteriaQueryBuilder(Criteria.where("lastName").is("Smith"))
    .withPointInTime(new Query.PointInTime(pit, tenSeconds))                        (2)
    .build();
SearchHits<Person> searchHits1 = operations.search(query1, Person.class);
// do something with the data

// create 2nd query for the pit, use the id returned in the previous result
Query query2 = new CriteriaQueryBuilder(Criteria.where("lastName").is("Miller"))
    .withPointInTime(
        new Query.PointInTime(searchHits1.getPointInTimeId(), tenSeconds))          (3)
    .build();
SearchHits<Person> searchHits2 = operations.search(query2, Person.class);
// do something with the data

operations.closePointInTime(searchHits2.getPointInTimeId());                        (4)

1	create a point in time for an index (can be multiple names) and a keep-alive duration and retrieve its id
2	pass that id into the query to search together with the next keep-alive value
3	for the next query, use the id returned from the previous search
4	when done, close the point in time using the last returned id

Search Template support

Use of the search template API is supported. To use this, it first is necessary to create a stored script. The ElasticsearchOperations interface extends ScriptOperations which provides the necessary functions. The example used here assumes that we have Person entity with a property named firstName. A search template script can be saved like this:

import org.springframework.data.elasticsearch.core.ElasticsearchOperations;
import org.springframework.data.elasticsearch.core.script.Script;

operations.putScript(                            (1)
  Script.builder()
    .withId("person-firstname")                  (2)
    .withLanguage("mustache")                    (3)
    .withSource("""                              (4)
      {
        "query": {
          "bool": {
            "must": [
              {
                "match": {
                  "firstName": "{{firstName}}"   (5)
                }
              }
            ]
          }
        },
        "from": "{{from}}",                      (6)
        "size": "{{size}}"                       (7)
      }
      """)
    .build()
);

1	Use the `putScript()` method to store a search template script
2	The name / id of the script
3	Scripts that are used in search templates must be in the mustache language.
4	The script source
5	The search parameter in the script
6	Paging request offset
7	Paging request size

To use a search template in a search query, Spring Data Elasticsearch provides the SearchTemplateQuery, an implementation of the org.springframework.data.elasticsearch.core.query.Query interface.

In the following code, we will add a call using a search template query to a custom repository implementation (see Custom Repository Implementations) as an example how this can be integrated into a repository call.

We first define the custom repository fragment interface:

interface PersonCustomRepository {
	SearchPage<Person> findByFirstNameWithSearchTemplate(String firstName, Pageable pageable);
}

The implementation of this repository fragment looks like this:

public class PersonCustomRepositoryImpl implements PersonCustomRepository {

  private final ElasticsearchOperations operations;

  public PersonCustomRepositoryImpl(ElasticsearchOperations operations) {
    this.operations = operations;
  }

  @Override
  public SearchPage<Person> findByFirstNameWithSearchTemplate(String firstName, Pageable pageable) {

    var query = SearchTemplateQuery.builder()                               (1)
      .withId("person-firstname")                                           (2)
      .withParams(
        Map.of(                                                             (3)
          "firstName", firstName,
          "from", pageable.getOffset(),
          "size", pageable.getPageSize()
          )
      )
      .build();

    SearchHits<Person> searchHits = operations.search(query, Person.class); (4)

    return SearchHitSupport.searchPageFor(searchHits, pageable);
  }
}

1	Create a `SearchTemplateQuery`
2	Provide the id of the search template
3	The parameters are passed in a `Map<String,Object>`
4	Do the search in the same way as with the other query types.

Nested sort

Spring Data Elasticsearch supports sorting within nested objects (www.elastic.co/guide/en/elasticsearch/reference/8.9/sort-search-results.html#nested-sorting)

The following example, taken from the org.springframework.data.elasticsearch.core.query.sort.NestedSortIntegrationTests class, shows how to define the nested sort.

var filter = StringQuery.builder("""
	{ "term": {"movies.actors.sex": "m"} }
	""").build();
var order = new org.springframework.data.elasticsearch.core.query.Order(Sort.Direction.DESC,
	"movies.actors.yearOfBirth")
	.withNested(
		Nested.builder("movies")
			.withNested(
				Nested.builder("movies.actors")
					.withFilter(filter)
					.build())
			.build());

var query = Query.findAll().addSort(Sort.by(order));

About the filter query: It is not possible to use a CriteriaQuery here, as this query would be converted into a Elasticsearch nested query which does not work in the filter context. So only StringQuery or NativeQuery can be used here. When using one of these, like the term query above, the Elasticsearch field names must be used, so take care, when these are redefined with the @Field(name="…") definition.

For the definition of the order path and the nested paths, the Java entity property names should be used.