Extension Documentation
Extension
interfaceNote: This document is evolving and is in draft state.
This document outlines how to create a new custom extension. For migration of existing plugins, see PLUGIN_MIGRATION.
For this example, you will create a CRUD extension, demonstrating the create, read, update, and delete operations on an index.
Create a new repository at a location of your choice.
In your dependency management, set up a dependency on the OpenSearch SDK for Java. Here is the required key information:
org.opensearch.sdk
opensearch-sdk-java
1.0.0-SNAPSHOT
(compatible with OpenSearch 2.x) or 2.0.0-SNAPSHOT
(compatible with OpenSearch 3.x)At general availability, dependencies will be released to the Central Repository. To use SNAPSHOT versions, add these repositories:
If you use Maven, the following POM entries will work:
<repositories>
<repository>
<id>opensearch.snapshots</id>
<name>OpenSearch Snapshot Repository</name>
<url>https://aws.oss.sonatype.org/content/repositories/snapshots/</url>
</repository>
<repository>
<id>lucene.snapshots</id>
<name>Lucene Snapshot Repository</name>
<url>https://ci.opensearch.org/ci/dbc/snapshots/lucene/</url>
</repository>
</repositories>
<dependencies>
<dependency>
<groupId>org.opensearch.sdk</groupId>
<artifactId>opensearch-sdk-java</artifactId>
<version>2.0.0-SNAPSHOT</version>
</dependency>
</dependencies>
For Gradle, specify dependencies as follows:
repositories {
mavenCentral()
maven { url "https://aws.oss.sonatype.org/content/repositories/snapshots/" }
maven { url "https://ci.opensearch.org/ci/dbc/snapshots/lucene/"}
}
dependencies {
implementation("org.opensearch.sdk:opensearch-sdk-java:2.0.0-SNAPSHOT")
}
An extension requires host and port information for both the extension and OpenSearch.
You may either define these in code in an ExtensionSettings
object, or in a YAML file. The following are equivalent:
import org.opensearch.sdk.ExtensionSettings;
new ExtensionSettings("crud", "127.0.0.1", "4532", "127.0.0.1", "9200")
crud.yml
file:extensionName: crud
hostAddress: 127.0.0.1
hostPort: 4532
opensearchAddress: 127.0.0.1
opensearchPort: 9200
Extension
interfaceCreate a class that implements Extension
. You may prefer to create a class that extends BaseExtension
, which provides some helper methods.
Implementing the Extension
interface requires you to implement the getExtensionSettings()
and setExtensionsRunner()
methods. The BaseExtension
class implements these and only requires that you call super()
with either the ExtensionSettings
object you created or a path to the YAML file (either absolute or classpath-based).
Implement a main()
method that instantiates your object and passes an instance of itself to ExtensionsRunner
. You will need to either handle or throw an IOException
from this method.
The following Java code accomplishes the preceding steps:
import java.io.IOException;
import org.opensearch.sdk.BaseExtension;
import org.opensearch.sdk.ExtensionSettings;
import org.opensearch.sdk.ExtensionsRunner;
public class CRUDExtension extends BaseExtension {
public CRUDExtension() {
// Optionally, pass a String path to a YAML file with these settings
super(new ExtensionSettings("crud", "127.0.0.1", "4532", "127.0.0.1", "9200"));
}
public static void main(String[] args) throws IOException {
ExtensionsRunner.run(new CRUDExtension());
}
}
At this point, you have a working extension! Start it by executing the main()
method, and then start your OpenSearch cluster.
But it doesn’t do anything yet. Here is where you can start defining your own functionality.
If you want to handle REST requests, implement the ActionExtension
interface and override the getExtensionRestHandlers()
method. Pass a list of classes that will handle those requests:
import org.opensearch.sdk.api.ActionExtension;
public class CRUDExtension extends BaseExtension implements ActionExtension {
// keep the constructor and main method from before and add the following code
@Override
public List<ExtensionRestHandler> getExtensionRestHandlers() {
// you need to create this class next!
return List.of(new CrudAction());
}
}
These classes must implement ExtensionRestHandler
, which is a functional interface that requires the implementation of the handleRequest()
method with the signature public ExtensionRestResponse handleRequest(RestRequest request)
.
The BaseExtensionRestHandler
class provides many useful methods for exception handling in requests.
For the CRUD extension example, you’ll implement one REST route for each option and delegate it to the appropriate handler function. Each route is an instance of NamedRoute
and requires at least a method, path, and globally unique name.
import java.util.List;
import java.util.function.Function;
import org.opensearch.rest.NamedRoute;
import org.opensearch.rest.RestRequest;
import org.opensearch.rest.RestRequest.Method;
import org.opensearch.rest.RestResponse;
import org.opensearch.rest.RestStatus;
import org.opensearch.sdk.rest.BaseExtensionRestHandler;
public class CrudAction extends BaseExtensionRestHandler {
@Override
public List<NamedRoute> routes() {
return List.of(
new NamedRoute.Builder().method(Method.PUT)
.path("/sample")
.uniqueName("crud_extension:sample/create")
.handler(createHandler)
.build(),
new NamedRoute.Builder().method(Method.GET)
.path("/sample/{id}")
.uniqueName("crud_extension:sample/get")
.handler(readHandler)
.build(),
new NamedRoute.Builder().method(Method.POST)
.path("/sample/{id}")
.uniqueName("crud_extension:sample/post")
.handler(updateHandler)
.build(),
new NamedRoute.Builder().method(Method.DELETE)
.path("/sample/{id}")
.uniqueName("crud_extension:sample/delete")
.handler(deleteHandler)
.build()
);
}
Function<RestRequest, RestResponse> createHandler = (request) -> {
return new ExtensionRestResponse(request, RestStatus.OK, "To be implemented");
};
Function<RestRequest, RestResponse> readHandler = (request) -> {
return new ExtensionRestResponse(request, RestStatus.OK, "To be implemented");
};
Function<RestRequest, RestResponse> updateHandler = (request) -> {
return new ExtensionRestResponse(request, RestStatus.OK, "To be implemented");
};
Function<RestRequest, RestResponse> deleteHandler = (request) -> {
return new ExtensionRestResponse(request, RestStatus.OK, "To be implemented");
};
}
To use the OpenSearch REST API, you will need an instance of the OpenSearch Java client.
Refer to the OpenSearch Java client documentation for either Apache HttpClient 5 Transport or OpenSearch RestClient Transport.
The remainder of this example assumes you have implemented one of the preceding options in your constructor:
private final OpenSearchClient client;
public CrudAction() {
final OpenSearchTransport transport = // implement per documentation
this.client = new OpenSearchClient(transport);
}
For your CRUD sample you will create a simple Java class with a single field:
public static class CrudData {
private String value;
public String getValue() {
return value;
}
public void setValue(String value) {
this.value = value;
}
}
Now in the create handler function, create an index (if it doesn’t exist):
BooleanResponse exists = client.indices().exists(new ExistsRequest.Builder().index("crudsample").build());
if (!exists.value()) {
client.indices().create(new CreateIndexRequest.Builder().index("crudsample").build());
}
Next, you add a document to it:
CrudData crudData = new CrudData();
crudData.setValue("value");
IndexResponse response = client.index(new IndexRequest.Builder<CrudData>().index("crudsample").document(crudData).build());
The BaseExtensionRestHandler
provides an exceptionalRequest()
method to handle exceptions:
return exceptionalRequest(request, e);
The user needs the ID of the created document (response.id()
) for further handling. The BaseExtensionRestHandler
provides a createJsonResponse()
method for this:
return createJsonResponse(request, RestStatus.OK, "_id", response.id());
Finally, you have the following code for the create handler method:
Function<RestRequest, RestResponse> createHandler = (request) -> {
IndexResponse response;
try {
// Create index if it doesn't exist
BooleanResponse exists = client.indices().exists(new ExistsRequest.Builder().index("crudsample").build());
if (!exists.value()) {
client.indices().create(new CreateIndexRequest.Builder().index("crudsample").build());
}
// Now add our document
CrudData crudData = new CrudData();
crudData.setValue("value");
response = client.index(new IndexRequest.Builder<CrudData>().index("crudsample").document(crudData).build());
} catch (OpenSearchException | IOException e) {
return exceptionalRequest(request, e);
}
if (response.result() == Result.Created) {
return createJsonResponse(request, RestStatus.OK, "_id", response.id());
}
return createJsonResponse(request, RestStatus.INTERNAL_SERVER_ERROR, "failed", response.result().toString());
};
You can now use the read handler function to get the document you just created, using its ID, which you will pass as a named parameter in the path. You can then get the document by ID.
String id = request.param("id");
GetResponse<CrudData> response = client.get(new GetRequest.Builder().index("crudsample").id(id).build(), CrudData.class);
Adding exception handling, the following is the full handler method:
Function<RestRequest, RestResponse> readHandler = (request) -> {
GetResponse<CrudData> response;
// Parse ID from request
String id = request.param("id");
try {
response = client.get(new GetRequest.Builder().index("crudsample").id(id).build(), CrudData.class);
} catch (OpenSearchException | IOException e) {
return exceptionalRequest(request, e);
}
if (response.found()) {
return createJsonResponse(request, RestStatus.OK, "value", response.source().getValue());
}
return createJsonResponse(request, RestStatus.NOT_FOUND, "error", "not_found");
};
You will create a new document similar to the one you created in the create handler, parse the ID as you did in the read handler, and then update that document. With exception handling, the following is the update handler method:
Function<RestRequest, RestResponse> updateHandler = (request) -> {
UpdateResponse<CrudData> response;
// Parse ID from request
String id = request.param("id");
// Now create the new document to update with
CrudData crudData = new CrudData();
crudData.setValue("new value");
try {
response = client.update(
new UpdateRequest.Builder<CrudData, CrudData>().index("crudsample").id(id).doc(crudData).build(),
CrudData.class
);
} catch (OpenSearchException | IOException e) {
return exceptionalRequest(request, e);
}
if (response.result() == Result.Updated) {
return createEmptyJsonResponse(request, RestStatus.OK);
}
return createJsonResponse(request, RestStatus.INTERNAL_SERVER_ERROR, "failed", response.result().toString());
};
You only need the ID to delete a document, so the delete handler method is implemented as follows:
Function<RestRequest, RestResponse> deleteHandler = (request) -> {
DeleteResponse response;
// Parse ID from request
String id = request.param("id");
try {
response = client.delete(new DeleteRequest.Builder().index("crudsample").id(id).build());
} catch (OpenSearchException | IOException e) {
return exceptionalRequest(request, e);
}
if (response.result() == Result.Deleted) {
return createEmptyJsonResponse(request, RestStatus.OK);
}
return createJsonResponse(request, RestStatus.INTERNAL_SERVER_ERROR, "failed", response.result().toString());
};