Getting Started With Elasticsearch in Java Spring Boot
Both Java and Elasticsearch is a popular technology stack companies use. Java is a programming language that was released back in 1996. Currently, Java is acquired by Oracle and still in active development.
Elasticsearch is a young technology when we compare it to Java, it has only released in 2010 (14 years younger than Java). It’s gaining popularity quickly and now used in many companies as a search engine.
Seeing how popular they are, I’m sure that many people and companies want to connect Java with Elasticsearch to develop their own search engine. In this article, I want to teach you how to connect Java Spring Boot 2 with Elasticsearch. We will learn how to create an API that will call Elasticsearch to produce results.
Connecting Java with Elasticsearch
The first thing we must do to connect our Spring Boot project with Elasticsearch. The easiest way to do this is to use the client library provided by Elasticsearch, which we can just add to our package manager like Maven or Gradle.
For this article, we’ll use a spring-data-elasticsearch library provided by Spring Data, which also includes Elasticsearch’s high level client library.
Starting our project
Let’s start by creating our Spring Boot project with Spring Initialzr. I’ll configure my project to be like the picture below, since we’re going to use high-level client, then we can use a convenient library provided by Spring, “Spring Data Elasticsearch”:
Adding dependency to Spring Data Elasticsearch
If you followed my Spring Initialzr configuration in the previous section, then you should already have the elasticsearch client dependency in your project. But, if you don’t, you can add:
<dependency>
<groupId>org.springframework.boot</groupId>
<artifactId>spring-boot-starter-data-elasticsearch</artifactId>
</dependency>Creating Elasticsearch client's bean
There are 2 methods to initialize the bean, you can either use the beans defined in the spring data elasticsearch library or you can create your own bean.
The first and easy one is to use the bean configured by spring data elasticsearch.
For example, you can add these properties in your application.properties:
spring.elasticsearch.rest.uris=localhost:9200
spring.elasticsearch.rest.connection-timeout=1s
spring.elasticsearch.rest.read-timeout=1m
spring.elasticsearch.rest.password=
spring.elasticsearch.rest.username=The second method is to create your own bean. You can configure the settings by creating RestHighLevelClient bean. If the bean is exist, the spring data will use it as its configuration.
@Configuration
@RequiredArgsConstructor
public class ElasticsearchConfiguration extends AbstractElasticsearchConfiguration {
private final ElasticsearchProperties elasticsearchProperties;
@Override
@Bean
public RestHighLevelClient elasticsearchClient() {
final ClientConfiguration clientConfiguration = ClientConfiguration.builder()
.connectedTo(elasticsearchProperties.getHostAndPort())
.withConnectTimeout(elasticsearchProperties.getConnectTimeout())
.withSocketTimeout(elasticsearchProperties.getSocketTimeout())
.build();
return RestClients.create(clientConfiguration).rest();
}
}Testing the connection from our Spring Boot application to Elasticsearch
Your Spring Boot app and Elasticsearch should be connected now that you’ve configured the bean. Since we’re going to test the connection, make sure that your Elasticsearch is up and running!
To test it, we can create a bean that will create an index in the Elasticsearch in the DemoApplication.class. The class would look like:
@SpringBootApplication
public class DemoApplication {
public static void main(String[] args) {
SpringApplication.run(DemoApplication.class, args);
}
@Bean
public boolean createTestIndex(RestHighLevelClient restHighLevelClient) throws Exception {
try {
DeleteIndexRequest deleteIndexRequest = new DeleteIndexRequest("hello-world");
restHighLevelClient.indices().delete(deleteIndexRequest, RequestOptions.DEFAULT); // 1
} catch (Exception ignored) {
}
CreateIndexRequest createIndexRequest = new CreateIndexRequest("hello-world");
createIndexRequest.settings(
Settings.builder().put("index.number_of_shards", 1)
.put("index.number_of_replicas", 0));
restHighLevelClient.indices().create(createIndexRequest, RequestOptions.DEFAULT); // 2
return true;
}
}Okay, in that code we called Elasticsearch twice with the RestHighLevelClient, which we will learn later on in this article. The first call is to delete the index if it’s already exists. We used try catch that because if the index doesn’t exist then the elasticsearch will throw an error and failing our app starting process.
The second call is to create an index. Since I’m only running an 1 node Elasticsearch, I configured the shards to be 1 and replicas to be 0.
If everything went fine, then you should see the indices when you check your Elasticsearch. To check it, just go to http://localhost:9200/_cat/indices?v and you can see the list of the indexes in your Elasticsearch:
health status index uuid pri rep docs.count docs.deleted store.size pri.store.size
yellow open hello-world 0NgzXS5gRxmj1eFTPMCynQ 1 1 0 0 208b 208bCongrats! You just connect your application to the Elasticsearch!!
Another ways to connect
I recommend you to use spring-data-elasticsearch library if you want to connect to Elasticsearch with Java. But, in case that you can’t use the library, there is another way to connect your apps to Elasticsearch.
High level client
As we know in the previous section, the spring-data-elasticsearch library we use also includes Elasticsearch’s high level client. If you’ve already imported spring-data-elasticsearch, then you can already use the Elasticsearch’s high level client.
If you want to, it’s also possible to use the high level client library directly without spring data’s dependency. You just need to add this dependency in your dependency manager:
<dependency>
<groupId>org.elasticsearch.client</groupId>
<artifactId>elasticsearch-rest-high-level-client</artifactId>
<version>8.0.0</version>
</dependency>We’ll also use this client in our examples because the function in high level client is more complete than the spring-data-elasticsearch.
Low level client
Elasticsearch’s low level client. You’ll have a harder time with this library, but you can customize it more. To use it, you can add the following dependency:
<dependency>
<groupId>org.elasticsearch.client</groupId>
<artifactId>elasticsearch-rest-client</artifactId>
<version>8.0.0</version>
</dependency>Transport Client
Elasticsearch also provides transport client, which will make your application identify as one of the node of Elasticsearch. I don’t recommend this method because it will be deprecated soon.
If you’re interested, you can read about transport client here.
REST call
The last way to connect to Elasticsearch is by doing a REST call. Since Elasticsearch uses REST API to connect to its client, you basically can use a REST call to connect your apps to Elasticsearch. You can use OKHTTP, Feign or Web Client to connect your apps with Elasticsearch.
I also don’t recommend this method because it’s a hassle. Since Elasticsearch already provides client libraries, it’s better to use them instead. Only use this method if you don’t have any other way to connect.
Using Spring Data Elasticsearch
First, let’s learn how to use spring-data-elasticsearch in our spring project. spring-data-elasticsearch is a very easy to use and high level library we can use to access the Elasticsearch.
Creating entity and configuring our index
After we’re done connecting your apps with Elasticsearch, it’s time to create an entity! With spring data, we can add a metadata in our entity, which will be read by the repository bean we created. This way the code will be much cleaner and faster to develop since we don’t need to create any mapping logic in our service level.
Let’s create an entity called Product:
@Data
@AllArgsConstructor
@NoArgsConstructor
@Builder
@Document(indexName = "product", shards = 1, replicas = 0, refreshInterval = "5s", createIndex = true)
public class Product {
@Id
private String id;
@Field(type = FieldType.Text)
private String name;
@Field(type = FieldType.Keyword)
private Category category;
@Field(type = FieldType.Long)
private double price;
public enum Category {
CLOTHES,
ELECTRONICS,
GAMES;
}
}So let me explain what’s going on in the code block above. First, I won’t explain about @Data @AllArgsConstructor @NoArgsConstructor @Builder . They’re annotations from Lombok library for constructor, getter, setter, builder, and other things. If you don’t know about them yet, I urge you to check it out.
Now, let’s talk about the first spring data annotation in the Entity, @Document . @Document annotation show that the class is an entity containing a metadata of the Elasticsearch index’s setup. To use spring data repository, which we’ll learn later on, the @Document annotation is mandatory.
The only annotation that is mandatory in the @Document is the indexName. It should be pretty clear from the name, we should fill it with the Index name we want to use for the entity. In this article, we’ll use the same name as the entity, product.
The second parameter of the @Document to talk about is the createIndex parameter. If you set the createIndex as true, your apps will create an index automatically when you’re starting the apps if the index doesn’t yet exist.
shards, replicas and refreshInterval parameters determine the index settings when the index is created. If you change the value of those parameters after the index is already created, the settings won’t be applied. So, the parameters will only be used when creating the index for the first time.
If you want to use a custom id in the Elasticsearch, you can use @Id annotations. If you use the @Id annotations, spring data will tell Elasticsearch to store the id in the document and the document source.
The @Field type will determine the field mapping of the field. Like shards, replicas and refreshInterval, the @Field type will only affect Elasticsearch when first creating the index. If you add a new field or change types when the index is already created, it won’t do anything.
Now that we configured the entity, let’s try out the automatic index creation by spring data! When we configure the createIndex as true, spring data will check whether the index exists in Elasticsearch. If it doesn’t exist, spring data will create the index with the configuration we created in the entity.
Let’s start our apps, after it is running, let’s check the settings and see if it’s correct:
curl --request GET \
--url http://localhost:9200/product/_settingsThe result is:
{
"product": {
"settings": {
"index": {
"routing": {
"allocation": {
"include": {
"_tier_preference": "data_content"
}
}
},
"refresh_interval": "5s",
"number_of_shards": "1",
"provided_name": "product",
"creation_date": "1607959499342",
"store": {
"type": "fs"
},
"number_of_replicas": "0",
"uuid": "iuoO8lE6QyWVSoECxa0I8w",
"version": {
"created": "7100099"
}
}
}
}
}Everything is as we configured! The refresh_interval is set to 5s, the number_of_shards is 1 and number_of_replicas is 0.
Now let's check the mappings:
curl --request GET \
--url http://localhost:9200/product/_mappingsThe result is:
{
"product": {
"mappings": {
"properties": {
"category": {
"type": "keyword"
},
"name": {
"type": "text"
},
"price": {
"type": "long"
}
}
}
}
}The mappings are also as we expected. It’s the same as we configured in the entity class!
Basic CRUD with spring data repository interface
After we created the entity, we’ve everything we need to create a repository interface in Spring Boot. Let’s create a repository called ProductRepository. When you’re creating an interface, make sure to extend ElasticsearchRepository<T, U>. In this case, the T object is your entity and U object type you want to use for the data id. In our case, since we’ll use Product entity, we created earlier than T and String as U .
public interface ProductRepository extends ElasticsearchRepository<Product, String> {
}Now, your repository interface is done, you don’t need to take care about the implementation because spring is taking care of that. Now, you can call every function in the classes that your repository extends to.
For the examples of CRUD, you can check the codes below:
@Service
@RequiredArgsConstructor
public class SpringDataProductServiceImpl implements SpringDataProductService {
private final ProductRepository productRepository;
public Product createProduct(Product product) {
return productRepository.save(product);
}
public Optional<Product> getProduct(String id) {
return productRepository.findById(id);
}
public void deleteProduct(String id) {
productRepository.deleteById(id);
}
public Iterable<Product> insertBulk(List<Product> products) {
return productRepository.saveAll(products);
}
}In the code blocks above, we created a service class called SpringDataProductServiceImpl which is autowired to ProductRepository we created before.
There are 4 basic CRUD function in it. The first one is createProduct which as its name will create a new product in the product index. The second one, getProduct is to get the product we’ve indexed by its id. The deleteProduct function can be used to delete the product in the index by id. insertBulk function will allow you to insert multiple products to Elasticsearch.
All is done! I won’t write about the API testing in this article because I want to focus about how our apps can interact with Elasticsearch. But, if you want to try the API, I left a GitHub link in the end of the article so you can clone and try this project after you’re done with this article.
Custom query methods in the spring data
In the previous section, we only take advantage of using the basic methods that are already defined in the other classes. But we can also create a custom query methods to use. What’s very convenient about spring data is that you can make a method in the repository interface and you don’t need to code any implementation. Spring data library will read the repository and automatically create the implementations for it.
Let’s try searching for products by the name field:
public interface ProductRepository extends ElasticsearchRepository<Product, String> {
List<Product> findAllByName(String name);
}Yes, that’s all you need to do to create a function in spring data repository interface.
You can also define a custom query with @Query annotation and insert a JSON query in the parameters.
public interface ProductRepository extends ElasticsearchRepository<Product, String> {
List<Product> findAllByName(String name);
@Query("{\"match\":{\"name\":\"?0\"}}")
List<Product> findAllByNameUsingAnnotations(String name);
}Both of the methods we’ve created do the same thing, use the match query with name as its parameter. If you try it, you’ll get the same results.
Using ElasticsearchRestTemplate
If you want to do a more advanced query, like aggregations, highlighting or suggestions, you can use ElasticsearchsearchRestTemplate provided by the spring data library. By using it, you can create your own query as complex as you want.
For example, let’s create a function for doing a match query to the name field like before:
public List<Product> getProductsByName(String name) {
Query query = new NativeSearchQueryBuilder()
.withQuery(QueryBuilders.matchQuery("name", name))
.build();
SearchHits<Product> searchHits = elasticsearchRestTemplate.search(query, Product.class);
return searchHits.get().map(SearchHit::getContent).collect(Collectors.toList());
}You should notice that the code above is more complex than the one we defined in the ElasticserchRepository. It is recommended to use the spring data repository if you can. But, for more advanced query like aggregation, highlighting or suggestions, you must use the ElasticsearchRestTemplate.
For example, let’s write a code that will aggregate a term:
public Map<String, Long> aggregateTerm(String term) {
Query query = new NativeSearchQueryBuilder()
.addAggregation(new TermsAggregationBuilder(term).field(term).size(10))
.build();
SearchHits<Product> searchHits = elasticsearchRestTemplate.search(query, Product.class);
Map<String, Long> result = new HashMap<>();
searchHits.getAggregations().asList().forEach(aggregation -> {
((Terms) aggregation).getBuckets()
.forEach(bucket -> result.put(bucket.getKeyAsString(), bucket.getDocCount()));
});
return result;
}Elasticsearch RestHighLevelClient
If you’re not using spring, or your spring version doesn’t support spring-data-elasticsearch, you can use a Java library developed by Elasticsearch, RestHighLevelClient.
RestHighLevelClient is a library you can use to do from basic things like CRUD to managing your Elasticsearch. Even though the name is high level, it’s actually more low level if you compare it to spring-data-elasticsearch.
The advantage of this library over spring data is that you can also manage your Elasticsearch with it. It provides index and elasticsearch configuration, which you can use more flexibility compared to spring data. It’s also has a more complete function that interact with Elasticsearch. The disadvantage of this library over spring data is this library is more low level, which means you must code more.
CRUD with RestHighLevelClient
Let’s see how we can create a simple create a function with the library so we can compare it to the previous methods we’ve used:
@Service
@RequiredArgsConstructor
@Slf4j
public class HighLevelClientProductServiceImpl implements HighLevelClientProductService {
private final RestHighLevelClient restHighLevelClient;
private final ObjectMapper objectMapper;
public Product createProduct(Product product) {
IndexRequest indexRequest = new IndexRequest("product");
indexRequest.id(product.getId());
indexRequest.source(product);
try {
IndexResponse indexResponse = restHighLevelClient.index(indexRequest, RequestOptions.DEFAULT);
if (indexResponse.status() == RestStatus.ACCEPTED) {
return product;
}
throw new RuntimeException("Wrong status: " + indexResponse.status());
} catch (Exception e) {
log.error("Error indexing, product: {}", product, e);
return null;
}
}
}As you can see, it’s now more complicated and harder to implement. Now, you need to handle the exception and also convert the JSON result to your entity. It’s recommended to use spring data instead for basic CRUD because RestHighLevelClient is more complicated.
I’ve included another CRUD functions in the GitHub project. If you’re interested, you can check it out. The link is at the end of this article.
Index Creation
This section is where the RestHighLevelClient holds a clear advantage compared to spring data elasticsearch. When we’re creating an index with its mappings and settings in the previous section, we’ve only used annotations. It’s very easy to do, but you can’t do much with it.
With RestHighLevelClient, you can create methods for index managements, or basically almost anything that Elasticsearch REST API allows.
For example, let’s write a code that will creates product index with the settings and mappings we used before:
public boolean createProductIndex() {
CreateIndexRequest createIndexRequest = new CreateIndexRequest("product");
createIndexRequest.settings(Settings.builder()
.put("number_of_shards", 1)
.put("number_of_replicas", 0)
.put("index.requests.cache.enable", false)
.build());
Map<String, Map<String, String>> mappings = new HashMap<>();
mappings.put("name", Collections.singletonMap("type", "text"));
mappings.put("category", Collections.singletonMap("type", "keyword"));
mappings.put("price", Collections.singletonMap("type", "long"));
createIndexRequest.mapping(Collections.singletonMap("properties", mappings));
try {
CreateIndexResponse createIndexResponse = restHighLevelClient.indices()
.create(createIndexRequest, RequestOptions.DEFAULT);
return createIndexResponse.isAcknowledged();
} catch (Exception e) {
e.printStackTrace();
}
return false;
}So let’s see what we did in the code:
- We initialized the
createIndexRequestwhen also determining the index name. - We added the settings in the request when calling
createIndexRequest.settings. In the settings, we also configured the fieldindex.requests.cache.enable, which is not possible with spring data library. - We made a
Mapcontaining the properties and mappings of the fields in the index. - We called the Elasticsearch with
restHighlevelClient.indices.create
As you can see, with the RestHighLevelClient we can create a more customized call for creating index to Elasticsearch compared to the annotations in spring data entity. There are also many more function in the RestHighLevelClient that aren’t exist in the spring data library. You can read Elasticsearch’s documentation for more information about the library.
Conclusion
In this article, we’ve learned two ways to connect to Elasticsearch, by using spring data and Elasticsearch client. Both are powerful library, but you should use only the spring data if it’s possible for your use case. The code with spring data elasticsearch is more readable and easy to use.
If you want a more powerful library that can basically do anything the Elasticsearch allows, though, then you can also use Elasticsearch high level client. You can also use the low level client, which we didn’t cover in this article, if you need even more powerful feature.
I hope that this article is to help you in getting started with Elasticsearch in Java Spring Boot. If you want to learn more about the libraries, you can check out spring data elasticsearch documentation and Elasticsearch’s high level client documentation.
Alas, thank you for reading until the end! Stay tuned for other articles, be sure to become member and subscribe to the newsletter!
Below is the GitHub link for the GitHub project used for this article: