API conventions

The Java API Client uses a very consistent code structure, using modern code patterns that make complex requests easier to write and complex responses easier to process. This page explains these so that you quickly feel at home.

The Elasticsearch API is large and is organized into feature groups, as can be seen in the Elasticsearch API documentation.

The Java API Client follows this structure: feature groups are called “namespaces”, and each namespace is located in a subpackage of co.elastic.clients.elasticsearch.

Each of the namespace clients can be accessed from the top level Elasticsearch client. The only exceptions are the “search” and “document” APIs which are located in the core subpackage and can be accessed on the main Elasticsearch client object.

The snippet below shows how to use the indices namespace client to create an index (the lambda syntax is explained below):

// Create the "products" index
ElasticsearchClient client = ...
client.indices().create(c -> c.index("products"));

Namespace clients are very lightweight objects that can be created on the fly.

Classes in the Java API Client contain two kinds of methods and properties:

All data types in the Java API Client are immutable. Object creation uses the builder pattern that was popularized in Effective Java in 2008.

ElasticsearchClient client = ...
CreateIndexResponse createResponse = client.indices().create(
    new CreateIndexRequest.Builder()
        .index("my-index")
        .aliases("foo",
            new Alias.Builder().isWriteIndex(true).build()
        )
        .build()
);

Note that a builder should not be reused after its build() method has been called.

Although this works nicely, having to instantiate builder classes and call the build() method is a bit verbose. So every property setter in the Java API Client also accepts a lambda expression that takes a newly created builder as a parameter and returns a populated builder. The snippet above can also be written as:

ElasticsearchClient client = ...
CreateIndexResponse createResponse = client.indices()
    .create(createIndexBuilder -> createIndexBuilder
        .index("my-index")
        .aliases("foo", aliasBuilder -> aliasBuilder
            .isWriteIndex(true)
        )
    );

This approach allows for much more concise code, and also avoids importing classes (and even remembering their names) since types are inferred from the method parameter signature.

Note in the above example that builder variables are only used to start a chain of property setters. The names of these variables are therefore unimportant and can be shortened to improve readability:

ElasticsearchClient client = ...
CreateIndexResponse createResponse = client.indices()
    .create(c -> c
        .index("my-index")
        .aliases("foo", a -> a
            .isWriteIndex(true)
        )
    );

Builder lambdas become particularly useful with complex nested queries like the one below, taken from the intervals query API documentation.

This example also highlights a useful naming convention for builder parameters in deeply nested structures. For lambda expressions with a single argument, Kotlin provides the implicit it parameter and Scala allows use of _. This can be approximated in Java by using an underscore prefix followed by a number representing the depth level (i.e. _0, _1, and so on). Not only does this remove the need to create throw-away variable names, but it also improves code readability. Correct indentation also allows the structure of the query to stand out.

ElasticsearchClient client = ...
SearchResponse<SomeApplicationData> results = client
    .search(_0 -> _0
        .query(_1 -> _1
            .intervals(_2 -> _2
                .field("my_text")
                .allOf(_3 -> _3
                    .ordered(true)
                    .intervals(_4 -> _4
                        .match(_5 -> _5
                            .query("my favorite food")
                            .maxGaps(0)
                            .ordered(true)
                        )
                    )
                    .intervals(_4 -> _4
                        .anyOf(_5 -> _5
                            .intervals(_6 -> _6
                                .match(_7 -> _7
                                    .query("hot water")
                                )
                            )
                            .intervals(_6 -> _6
                                .match(_7 -> _7
                                    .query("cold porridge")
                                )
                            )
                        )
                    )
                )
            )
        ),
    SomeApplicationData.class 
);

Properties of type List and Map are exposed by object builders as a set of overloaded additive-only methods that update the property value, by appending to lists and adding new entries to maps (or replacing existing ones).

Object builders create immutable objects, and this also applies to list and map properties that are made immutable at object construction time.

// Prepare a list of index names
List<String> names = Arrays.asList("idx-a", "idx-b", "idx-c");

// Prepare cardinality aggregations for fields "foo" and "bar"
Map<String, Aggregation> cardinalities = new HashMap<>();
cardinalities.put("foo-count", Aggregation.of(a -> a.cardinality(c -> c.field("foo"))));
cardinalities.put("bar-count", Aggregation.of(a -> a.cardinality(c -> c.field("bar"))));

// Prepare an aggregation that computes the average of the "size" field
final Aggregation avgSize = Aggregation.of(a -> a.avg(v -> v.field("size")));

SearchRequest search = SearchRequest.of(r -> r
    // Index list:
    // - add all elements of a list
    .index(names)
    // - add a single element
    .index("idx-d")
    // - add a vararg list of elements
    .index("idx-e", "idx-f", "idx-g")

    // Sort order list: add elements defined by builder lambdas
    .sort(s -> s.field(f -> f.field("foo").order(SortOrder.Asc)))
    .sort(s -> s.field(f -> f.field("bar").order(SortOrder.Desc)))

    // Aggregation map:
    // - add all entries of an existing map
    .aggregations(cardinalities)
    // - add a key/value entry
    .aggregations("avg-size", avgSize)
    // - add a key/value defined by a builder lambda
    .aggregations("price-histogram",
        a -> a.histogram(h -> h.field("price")))
);

The Elasticsearch API has a lot of optional properties. For single-valued properties, the Java API Client represents missing optional values as null. Applications therefore have to null-check optional values before using them.

For lists and maps however, applications often only care about whether they’re empty or not, or even just iterate on their content. Using null values is then cumbersome. To avoid this, Java API Client collection properties are never null, and missing optional collections are returned as an empty collection.

If you ever need to distinguish between a missing (undefined) optional collection and an effectively-empty collection returned by Elasticsearch, the ApiTypeHelper class provides a utility method to distinguish them:

NodeStatistics stats = NodeStatistics.of(b -> b
    .total(1)
    .failed(0)
    .successful(1)
);

// The `failures` list was not provided.
// - it's not null
assertNotNull(stats.failures());
// - it's empty
assertEquals(0, stats.failures().size());
// - and if needed we can know it was actually not defined
assertFalse(ApiTypeHelper.isDefined(stats.failures()));

The Elasticsearch API has a lot of variant types: queries, aggregations, field mappings, analyzers, and so on. Finding the correct class name in such large collections can be challenging.

The Java API Client builders make this easy: the builders for variant types, such as Query, have methods for each of the available implementations. We’ve seen this in action above with intervals (a kind of query) and allOf, match and anyOf (various kinds of intervals).

This is because variant objects in the Java API Client are implementations of a “tagged union”: they contain the identifier (or tag) of the variant they hold and the value for that variant. For example, a Query object can contain an IntervalsQuery with tag intervals, a TermQuery with tag term, and so on. This approach allows writing fluent code where you can let the IDE completion features guide you to build and navigate complex nested structures:

Variant builders have setter methods for every available implementation. They use the same conventions as regular properties and accept both a builder lambda expression and a ready-made object of the actual type of the variant. Here’s an example to build a term query:

Query query = new Query.Builder()
    .term(t -> t                          
        .field("name")                    
        .value(v -> v.stringValue("foo"))
    )
    .build();                             

Variant objects have getter methods for every available implementation. These methods check that the object actually holds a variant of that kind and return the value downcasted to the correct type. They throw an IllegalStateException otherwise. This approach allows writing fluent code to traverse variants.

assertEquals("foo", query.term().value().stringValue());

Variant objects also provide information on the variant kind they currently hold:

This information can then be used to navigate down into specific variants after checking their actual kind:

if (query.isTerm()) { 
    doSomething(query.term());
}

switch(query._kind()) { 
    case Term:
        doSomething(query.term());
        break;
    case Intervals:
        doSomething(query.intervals());
        break;
    default:
        doSomething(query._kind(), query._get()); 
}

API clients come in two flavors: blocking and asynchronous. All methods on asynchronous clients return a standard CompletableFuture.

Both flavors can be used at the same time depending on your needs, sharing the same transport object:

ElasticsearchTransport transport = ...

// Synchronous blocking client
ElasticsearchClient client = new ElasticsearchClient(transport);

if (client.exists(b -> b.index("products").id("foo")).value()) {
    logger.info("product exists");
}

// Asynchronous non-blocking client
ElasticsearchAsyncClient asyncClient =
    new ElasticsearchAsyncClient(transport);

asyncClient
    .exists(b -> b.index("products").id("foo"))
    .thenAccept(response -> {
        if (response.value()) {
            logger.info("product exists");
        }
    });

Client methods can throw two kinds of exceptions:

There are five kinds of objects in the Java API Client with different life cycles: