Elasticsearch客户端源码剖析

注：本文出自博主 Chloneda：个人博客 | 博客园 | Github | Gitee | 知乎

注：本文源链接：https://www.cnblogs.com/chloneda/p/es-clients.html

前言

今天我们来聊聊Elasticsearch客户端的类型。我们知道Elasticsearch是一种分布式的海量数据搜索与分析的技术，可以用于电商网站、门户网站、企业IT系统等各种场景下的搜索引擎，也可以用于对海量的数据进行近实时的数据分析。

但Elasticsearch版本迭代更新太快，这就意味着在Elasticsearch升级过程中容易出现兼容性问题。也引出了今天对Elasticsearch客户端种类及使用的问题讨论！

ES客户端种类

ES官方客户端有TransportClient、Java Low Level REST Client和Java High Level REST Client三种。官方文档对他们的说明是：

TransportClient

We plan on deprecating the TransportClient in Elasticsearch 7.0 and removing it completely in 8.0.

Java Low Level REST Client

the official low-level client for Elasticsearch. It allows to communicate with an Elasticsearch cluster through http. Leaves requests marshalling and responses un-marshalling to users. It is compatible with all Elasticsearch versions.

Java High Level REST Client

the official high-level client for Elasticsearch. Based on the low-level client, it exposes API specific methods and takes care of requests marshalling and responses un-marshalling.

意思就是说，TransportClient将会在将来版本进行废弃移除，官方建议使用Java High Level REST Client。

为什么会这样呢？这里涉及到两个问题：

• 未来版本为什么会淘汰TransportClient客户端？
• Java Low/High Level REST Client客户端优点在哪里？

先别急，我们来看看这两个问题！

客户端的使用

各客户端使用需要引入相关依赖，这里统一引入相关依赖，后面就不多赘述了！

<!-- elasticsearch core -->
<dependency>
    <groupId>org.elasticsearch</groupId>
    <artifactId>elasticsearch</artifactId>
    <version>${elasticsearch.version}</version>
</dependency>
<!-- low level rest client -->
<dependency>
    <groupId>org.elasticsearch.client</groupId>
    <artifactId>elasticsearch-rest-client</artifactId>
    <version>${elasticsearch.version}</version>
</dependency>
<!-- high level rest client -->
<dependency>
    <groupId>org.elasticsearch.client</groupId>
    <artifactId>elasticsearch-rest-high-level-client</artifactId>
    <version>${elasticsearch.version}</version>
</dependency>

注：TransportClient将会在8.x版本后完全移除！

TransportClient

初始化TransportClient客户端代码示例：

public TransportClient initTransportClient(String esClusterName,String host,String port) throws UnknownHostException {
	Settings settings = Settings.builder()
			.put("cluster.name", esClusterName)
			.put("client.transport.sniff", true)
			.build();
	TransportClient client = new PreBuiltTransportClient(settings)
			.addTransportAddress(new TransportAddress(InetAddress.getByName(host),port);

	return client;
}

Java Low Level REST Client

初始化 RestClient 客户端代码示例：

public RestClient initRestClient(String host, int port) {
	RestClientBuilder builder = RestClient.builder(new HttpHost(host,
			port, "http"));
	Header[] defaultHeaders = new Header[]{new BasicHeader("header", "value")};
	builder.setDefaultHeaders(defaultHeaders);
	RestClient restClient = builder.build();
	return restClient;
}

Java High Level REST Client

初始化 RestHighLevelClient 客户端代码示例：

public RestHighLevelClient restHighLevelClient(List<String> hostArray) {
	//创建HttpHost数组，其中存放es主机和端口的配置信息
	HttpHost[] httpHostArray = new HttpHost[hostArray.size()];
	for (int i = 0; i < hostArray.size(); i++) {
		String item = hostArray.get(i);
		httpHostArray[i] = new HttpHost(item.split(":")[0],
				Integer.parseInt(item.split(":")[1]),
				"http");
	}
	//创建RestHighLevelClient客户端
	return new RestHighLevelClient(RestClient.builder(httpHostArray));
}

以上就是初始化三种不同客户端的示例代码！下面我们深入客户端代码底层，看看他们之间有什么不一样？

深入客户端的底层

TransportClient

TransportClient客户端自从Elasticsearch诞生以来，一直是Elasticsearch的一部分。 它是一个特殊的客户端，因为它使用传输层协议（TCP）与Elasticsearch进行通信，如果该客户端与其所使用的Elasticsearch不在同一版本上，则会导致兼容性问题。基于这个原因，官方会在8.x后完全移除！

因此，在这里就不对 TransportClient 客户端底层进行深究了！

Java Low Level REST Client

2016年，Elasticsearch官方发布了一个低级REST客户端，该客户端基于众所周知的Apache HTTP客户端，它允许使用 HTTP 与任何版本的Elasticsearch集群进行通信。

我们来看看RestClient客户端的代码：

package org.elasticsearch.client;

public class RestClient implements Closeable {

    //已省略其他非必要属性代码。。。

    // RestClient 类构造器的第一个参数是 CloseableHttpAsyncClient，是Apache HTTP client 中的类，相关请求也是通过该参数

    RestClient(CloseableHttpAsyncClient client, long maxRetryTimeoutMillis, Header[] defaultHeaders, HttpHost[] hosts, String pathPrefix, RestClient.FailureListener failureListener) {
        this.client = client;
        this.maxRetryTimeoutMillis = maxRetryTimeoutMillis;
        this.defaultHeaders = Collections.unmodifiableList(Arrays.asList(defaultHeaders));
        this.failureListener = failureListener;
        this.pathPrefix = pathPrefix;
        this.setHosts(hosts);
    }

    //已省略其他非必要代码。。。

    public void performRequestAsync(String method, String endpoint, Map<String, String> params, HttpEntity entity, HttpAsyncResponseConsumerFactory httpAsyncResponseConsumerFactory, ResponseListener responseListener, Header... headers) {
        try {
            Objects.requireNonNull(params, "params must not be null");
            Map<String, String> requestParams = new HashMap(params);
            String ignoreString = (String)requestParams.remove("ignore");
            Object ignoreErrorCodes;
            if (ignoreString == null) {
                if ("HEAD".equals(method)) {
                    ignoreErrorCodes = Collections.singleton(404);
                } else {
                    ignoreErrorCodes = Collections.emptySet();
                }
            } else {
                String[] ignoresArray = ignoreString.split(",");
                ignoreErrorCodes = new HashSet();
                if ("HEAD".equals(method)) {
                    ((Set)ignoreErrorCodes).add(404);
                }

                String[] var12 = ignoresArray;
                int var13 = ignoresArray.length;

                for(int var14 = 0; var14 < var13; ++var14) {
                    String ignoreCode = var12[var14];

                    try {
                        ((Set)ignoreErrorCodes).add(Integer.valueOf(ignoreCode));
                    } catch (NumberFormatException var17) {
                        throw new IllegalArgumentException("ignore value should be a number, found [" + ignoreString + "] instead", var17);
                    }
                }
            }

            URI uri = buildUri(this.pathPrefix, endpoint, requestParams);
            HttpRequestBase request = createHttpRequest(method, uri, entity);
            this.setHeaders(request, headers);
            RestClient.FailureTrackingResponseListener failureTrackingResponseListener = new RestClient.FailureTrackingResponseListener(responseListener);
            long startTime = System.nanoTime();
            this.performRequestAsync(startTime, this.nextHost(), request, (Set)ignoreErrorCodes, httpAsyncResponseConsumerFactory, failureTrackingResponseListener);
        } catch (Exception var18) {
            responseListener.onFailure(var18);
        }

    }

    //已省略其他非必要代码。。。

    private static HttpRequestBase createHttpRequest(String method, URI uri, HttpEntity entity) {
        String var3 = method.toUpperCase(Locale.ROOT);
        byte var4 = -1;
        switch(var3.hashCode()) {
        case -531492226:
            if (var3.equals("OPTIONS")) {
                var4 = 3;
            }
            break;
        case 70454:
            if (var3.equals("GET")) {
                var4 = 1;
            }
            break;
        case 79599:
            if (var3.equals("PUT")) {
                var4 = 6;
            }
            break;
        case 2213344:
            if (var3.equals("HEAD")) {
                var4 = 2;
            }
            break;
        case 2461856:
            if (var3.equals("POST")) {
                var4 = 5;
            }
            break;
        case 75900968:
            if (var3.equals("PATCH")) {
                var4 = 4;
            }
            break;
        case 80083237:
            if (var3.equals("TRACE")) {
                var4 = 7;
            }
            break;
        case 2012838315:
            if (var3.equals("DELETE")) {
                var4 = 0;
            }
        }

        switch(var4) {
        case 0:
            return addRequestBody(new HttpDeleteWithEntity(uri), entity);
        case 1:
            return addRequestBody(new HttpGetWithEntity(uri), entity);
        case 2:
            return addRequestBody(new HttpHead(uri), entity);
        case 3:
            return addRequestBody(new HttpOptions(uri), entity);
        case 4:
            return addRequestBody(new HttpPatch(uri), entity);
        case 5:
            HttpPost httpPost = new HttpPost(uri);
            addRequestBody(httpPost, entity);
            return httpPost;
        case 6:
            return addRequestBody(new HttpPut(uri), entity);
        case 7:
            return addRequestBody(new HttpTrace(uri), entity);
        default:
            throw new UnsupportedOperationException("http method not supported: " + method);
        }
    }

}

看到上面的代码，RestClient 类构造器的第一个参数是 CloseableHttpAsyncClient，是 Apache HTTP client 中的类，也就是说 RestClient 是基于 Apache HTTP 实现的，这里是 Apache HTTP 的依赖！

<dependency>
    <groupId>org.apache.httpcomponents</groupId>
    <artifactId>httpclient</artifactId>
    <version>${http.version}</version>
</dependency>

Java High Level REST Client

最重要的是，我们发布了基于低级客户端的高级REST客户端，它负责请求编组和响应解组。

我们来看看 RestHighLevelClient 的底层代码！

package org.elasticsearch.client;

public class RestHighLevelClient {
    private final RestClient client;
    private final NamedXContentRegistry registry;

    public RestHighLevelClient(RestClient restClient) {
        this(restClient, Collections.emptyList());
    }

    // 此处省略多处代码！

    // 该类大部分方法最终会调用以下 performRequestAndParseEntity 方法，我们主要看该方法的调用关系

    protected <Req extends ActionRequest, Resp> Resp performRequestAndParseEntity(Req request, CheckedFunction<Req, Request, IOException> requestConverter, CheckedFunction<XContentParser, Resp, IOException> entityParser, Set<Integer> ignores, Header... headers) throws IOException {
        return this.performRequest(request, requestConverter, (response) -> {
            return this.parseEntity(response.getEntity(), entityParser);
        }, ignores, headers);
    }

    protected <Req extends ActionRequest, Resp> Resp performRequest(Req request, CheckedFunction<Req, Request, IOException> requestConverter, CheckedFunction<Response, Resp, IOException> responseConverter, Set<Integer> ignores, Header... headers) throws IOException {
        ActionRequestValidationException validationException = request.validate();
        if (validationException != null) {
            throw validationException;
        } else {
            Request req = (Request)requestConverter.apply(request);

            Response response;
            try {

            	// 这里的 client 就是RestClient，最终还是调用 RestClient 的方法，也就是说 RestHighLevelClient 是基于 RestClient 的

                response = this.client.performRequest(req.getMethod(), req.getEndpoint(), req.getParameters(), req.getEntity(), headers);
            } catch (ResponseException var13) {
                ResponseException e = var13;
                if (ignores.contains(var13.getResponse().getStatusLine().getStatusCode())) {
                    try {
                        return responseConverter.apply(e.getResponse());
                    } catch (Exception var11) {
                        throw this.parseResponseException(var13);
                    }
                }

                throw this.parseResponseException(var13);
            }

            try {
                return responseConverter.apply(response);
            } catch (Exception var12) {
                throw new IOException("Unable to parse response body for " + response, var12);
            }
        }
    }

}

看上面的代码及注解，我相信你很快就豁然开朗了！

其实上面的问题现在就有答案了！TransportClient废弃的主要原因就是考虑到兼容性的问题，而后续两个客户端在兼容性方面就做的很好！

小结

关于Elasticsearch的客户端问题，其实 ES官网 已经说得很明确了，这里也通过代码剖析的方式去认识一下底层的代码，加深理解！

由此可见，HighLevelClient 是基于 RestClient,而 RestClient 又是基于 Apache HTTP 客户端， 这样一来, 在客户端方面, Elasticsearch 将 Java, Python, Php, Javascript 等各种语言的底层接口就都统一起来了; 与此同时, 使用 rest api, 还可以屏蔽各版本之前的差异。

这也提醒我们，在代码的升级过渡期, 处理好新 client 和旧 client 的关系，可以减少代码后期维护的工作量！

End！