ES source code API call link source code analysis

Regarding the API call link, let's start with the familiar three-tier MVC architecture.Those who are doing development are more familiar with spring springMVC. The routine we are familiar with is that the request goes to the controller layer first, then the controller calls the service layer, and then the result is returned.Usually, we also have tomcat as a server to receive requests, and the received requests call the controller layer.A user's request, such as an http request, usually goes to the tomcat service first, and then calls the program's controller.
Then look at how es play, es does not use spring spingMVC, only Guice is used for container management, es does not choose to use a server similar to tomcat to receive requests, but chooses netty to implement the server,Complete the receiving and sending of requests.This corresponds to the RestController in the figure below. This layer has the same function as tomcat.The BaseRestHandler in the figure below can actually be understood as selecting a controller router, which selects a controller according to the url path in the request.The RestSearchAction in the third step in the figure below is the controller layer in the MVC that we understood before. Different functions are placed in different RestActions. In fact, some logic is also placed here.The NodeClient layer actually has a taste of service, and it is where the logic is really executed.Unlike traditional services, es is a natural distributed application.

1. Netty4HttpServerTransport.dispatchRequest()

2. RestController.dispatchRequest().tryAllHandlers()

3. RestController.dispatchRequest() calls wrappedHandler.handleRequest() in the method

4. BaseRestHandler.handleRequest() Then call the prepareRequest() method in the method. At this point, a specific action will actually be selected according to the request.prepareRequest() is actually an interface. You can see the implementation class below. The red box is just a case (this is an Action corresponding to deleting an index). In prepareRequest, the rest request is converted into a request recognized by es.

5. Analyze with a case (RestDeleteIndexAction)

   Then see:

@Overridepublic RestChannelConsumer prepareRequest(final RestRequest request, final NodeClient client) throws IOException { DeleteIndexRequest deleteIndexRequest = new DeleteIndexRequest(Strings.splitStringByCommaToArray(request.param("index"))); deleteIndexRequest.timeout(request.paramAsTime("timeout", deleteIndexRequest.timeout())); deleteIndexRequest.masterNodeTimeout(request.paramAsTime("master_timeout", deleteIndexRequest.masterNodeTimeout())); deleteIndexRequest.indicesOptions(IndicesOptions.fromRequest(request, deleteIndexRequest.indicesOptions())); return channel -> client.admin().indices().delete(deleteIndexRequest, new RestToXContentListener<>(channel));}

6. Follow the delete in the fifth step and go down to ElasticsearchClient.execute(), which is another interface, and then follow its implementation class, come to AbstractClient.execute(), and call doExecute(), this is an abstract method, its implementation class NodeClient.doExecute(), calls executeLocally() in the method, this method will call transportAction(action).execute(), you can see that the action is executed!

7. Continue to step 6, click in transportAction(action).execute() and come to TransportAction.execute() In this class, continue to click execute(), which calls requestFilterChain.proceed(), inGo down to proceed() and call this.action.doExecute(task, request, listener), this doExecute() is a method in the TransportAction class, and this class is an abstract class, so look at it according to doExecuteImplementation class, different functions correspond to different actions, or take the example of deleting the index,