Thin wrapper for interacting with elasticsearch using the REST client. Besides making it easier from the spring framework to interact with elasticsearch it gives you some additional functionality. It contains methods to help while interacting with indexes, transform objects into searchable documents and create queries using templates that are easy to read without using complete java builder structures.
The client comes in two flavors:
- Single cluster configuration ** Multi cluster configuration
Artifacts are deployed in the luminis nexus repository. This repository cannot act as a mirror. You can only download our own artifacts over there.
Add this repository to your pom:
<repository>
<id>luminis</id>
<name>Snapshots</name>
<url>https://repository.luminis.amsterdam/repository/maven-public</url>
</repository>
<dependency>
<groupId>eu.luminis</groupId>
<artifactId>elastic-rest-spring-wrapper</artifactId>
<version>0.12.0</version>
</dependency>
<dependency>
<groupId>eu.luminis</groupId>
<artifactId>elastic-rest-spring-wrapper</artifactId>
<version>0.13.0-SNAPSHOT</version>
</dependency>
As mentioned before, there are two ways to use the library. If you just connect to one custer you can use the following configuration in spring.
@Import(SingleClusterRestClientConfig.class)
If however you want to specify the cluster yourself and be able to switch clusters in you application. You want to use the following configuration. In this case you do have to register one or more clusters yourself. Watch for messages that no cluster was found if you are having problems.
@Import(RestClientConfig.class)
To query for a document by ID we just need to provide a TypeReference that is required to obtain a typed object.
package nl.gridshore.recommend;
import com.fasterxml.jackson.core.type.TypeReference;
import eu.luminis.elastic.document.response.GetByIdResponse;
import nl.gridshore.kafka.JobSession;
public class JobSessionTypeReference extends TypeReference<GetByIdResponse<JobSession>> {
}
With this type reference we can use the DocumentService to obtain an object of type JobSession
QueryByIdRequest queryRequest = new QueryByIdRequest();
queryRequest.setId(jobSession.getKey());
queryRequest.setType(TYPE);
queryRequest.setIndex(INDEX);
queryRequest.setTypeReference(TYPE_REFERENCE);
try {
JobSession currentJobSession = this.documentService.querybyId(queryRequest);
jobSession.getIds().forEach(currentJobSession::addId);
} catch (QueryByIdNotFoundException e) {
indexRequest.setEntity(jobSession);
}
Again we use the TypeReference as mentioned by the query
IndexRequest indexRequest = new IndexRequest();
indexRequest.setIndex(INDEX);
indexRequest.setType(TYPE);
indexRequest.setId(jobSession.getKey());
indexRequest.setEntity(aJobSession);
In the next sample code we use another index containing employees, we want to search for employees using their name and email address. For creating a query you can use the twig json templates. Such a template looks like this:
{
"query":{
{% if (length(searchText)==0) %}
"match_all": {}
{% else %}
"multi_match": {
"query": "{{ searchText }}",
"operator": "{{ default(operator, 'or') }}",
"fields": ["name","email"]
}
{% endif %}
}
}
In the template we check if you entered a search text, if not, we execute the match_all query. If you do provide a text we give the option to provide another operator than the default OR and we enter the searchText. With the following method we load the template, set the parameters and we execute the service call. We do need another TypeReference for a query in contrast with the query by id.
package nl.gridshore.employees;
import com.fasterxml.jackson.core.type.TypeReference;
import eu.luminis.elastic.document.response.QueryResponse;
public class EmployeeTypeReference extends TypeReference<QueryResponse<Employee>> {
}
public List<Employee> queryForEmployeesByNameAndEmail(String searchString) {
Map<String, Object> params = new HashMap<>();
params.put("searchText", searchString);
params.put("operator", "and");
QueryByTemplateRequest request = QueryByTemplateRequest.create()
.setAddId(true)
.setTypeReference(new EmployeeTypeReference())
.setIndexName(INDEX)
.setModelParams(params)
.setTemplateName("find_employee_by_email.twig");
return documentService.queryByTemplate(request);
}
This is mainly the same as for searching for documents. We do expect a Jackson ObjectMapper bean to be present. Usually spring takes care of this. If not, you have to provide one by yourself.
The aggregation responses can be divided into two groups. The Bucket aggregations and the Metric Aggregations. Not all aggregations are supported at the moment. Below is a list we do support at the moment: Bucket Aggregations
- Terms
- Histogram
- Date Histogram
Metric Aggregations
- Avg
- Sum
- Max
- Min
- Cardinality
- Value Count
Aggregations can now also be nested. You can nest a bucket aggregation with a metric aggregation. You can also nest two bucket aggregations.
The command to upload an artifact is:
mvn clean deploy -DskipTests
Of course this can only be done with the appropriate rights, so change your settings.xml accordingly.
If you remove the -SNAPHOT from the version, it will become a release. ANd if you add it again it will become a snapshot again.
These are some interesting not so obvious blog posts that helped me during the creation of this library:
Help with using Jackson for object deserialization http://www.robinhowlett.com/blog/2015/03/19/custom-jackson-polymorphic-deserialization-without-type-metadata/
At the moment the tests need to have a running elasticsearch server on port 19200. You can start one yourself and run the tests from within your IDE. It this is to much hassle, you can run the build with Maven. A plugin is used to download and start elasticsearch from maven. Below a link to the used plugin:
https://github.com/alexcojocaru/elasticsearch-maven-plugin
Another approach that is interesting is to start an embedded elasticsearch instance from Java. Need some time to evaluate this option. https://github.com/allegro/embedded-elasticsearch