Spring Data Elasticsearch是否支持Elasticsearch 6.2的Multi Search API？

我在使用Spring Data Elasticsearch配合ES6.2的时候也遇到过这个问题——确实当时的Spring Data Elasticsearch没有直接提供Multi Search的高层封装API，不过我们可以借助它底层的Elasticsearch客户端来实现，而且完全可以复用你已经用NativeSearchQueryBuilder写好的查询逻辑。

实现步骤 #

1. 构建多个NativeSearchQuery

首先你可以继续用NativeSearchQueryBuilder创建需要的搜索查询，比如：

// 第一个查询：匹配title包含"spring"的书籍
NativeSearchQuery bookTitleQuery = new NativeSearchQueryBuilder()
    .withQuery(QueryBuilders.matchQuery("title", "spring"))
    .withIndices("books")
    .withTypes("book")
    .build();

// 第二个查询：价格低于50的书籍
NativeSearchQuery bookPriceQuery = new NativeSearchQueryBuilder()
    .withQuery(QueryBuilders.rangeQuery("price").lt(50))
    .withIndices("books")
    .withTypes("book")
    .build();

2. 获取底层Elasticsearch客户端并构建MultiSearch请求

从ElasticsearchTemplate里拿到底层客户端（ES6.2支持RestHighLevelClient或TransportClient，这里分别给出两种示例）：

如果你用的是RestHighLevelClient

// 从ElasticsearchTemplate获取底层客户端
RestHighLevelClient client = elasticsearchTemplate.getClient();

// 创建MultiSearch请求实例
MultiSearchRequest multiSearchRequest = new MultiSearchRequest();

// 把NativeSearchQuery转换成原生SearchRequest
SearchRequest titleSearchRequest = new SearchRequest(bookTitleQuery.getIndices());
titleSearchRequest.types(bookTitleQuery.getTypes());
titleSearchRequest.source(bookTitleQuery.getSearchSourceBuilder());

SearchRequest priceSearchRequest = new SearchRequest(bookPriceQuery.getIndices());
priceSearchRequest.types(bookPriceQuery.getTypes());
priceSearchRequest.source(bookPriceQuery.getSearchSourceBuilder());

// 将两个搜索请求加入MultiSearch
multiSearchRequest.add(titleSearchRequest);
multiSearchRequest.add(priceSearchRequest);

如果你用的是TransportClient

// 从ElasticsearchTemplate获取底层客户端
TransportClient client = elasticsearchTemplate.getClient();

// 创建MultiSearch请求构建器
MultiSearchRequestBuilder multiSearchBuilder = client.prepareMultiSearch();

// 把NativeSearchQuery转换成原生SearchRequestBuilder
SearchRequestBuilder titleSearchBuilder = client.prepareSearch(bookTitleQuery.getIndices())
    .setTypes(bookTitleQuery.getTypes())
    .setSource(bookTitleQuery.getSearchSourceBuilder());

SearchRequestBuilder priceSearchBuilder = client.prepareSearch(bookPriceQuery.getIndices())
    .setTypes(bookPriceQuery.getTypes())
    .setSource(bookPriceQuery.getSearchSourceBuilder());

// 将两个搜索请求加入MultiSearch
multiSearchBuilder.add(titleSearchBuilder);
multiSearchBuilder.add(priceSearchBuilder);

3. 执行MultiSearch请求并处理结果

提交请求后，遍历结果并处理，还可以利用ElasticsearchTemplate把返回的文档转换成你定义的实体类：

// RestHighLevelClient的执行方式
MultiSearchResponse response = client.msearch(multiSearchRequest, RequestOptions.DEFAULT);

// （如果是TransportClient，执行方式是：MultiSearchResponse response = multiSearchBuilder.get();）

// 遍历处理每个请求的结果
for (MultiSearchResponse.Item item : response.getResponses()) {
    if (!item.isFailure()) {
        // 获取单个搜索的响应
        SearchResponse searchResponse = item.getResponse();
        // 利用ElasticsearchTemplate将结果映射为实体类
        List<Book> bookList = elasticsearchTemplate.mapResults(searchResponse, Book.class);
        // 这里可以根据业务需求处理bookList
        System.out.println("查询到书籍数量：" + bookList.size());
    } else {
        // 处理请求失败的情况
        String errorMsg = item.getFailureMessage();
        System.err.println("搜索请求失败：" + errorMsg);
    }
}

关键说明 #

• 之所以需要这么做，是因为Spring Data Elasticsearch在对应ES6.2的版本中，高层API（比如ElasticsearchTemplate的query方法）没有封装Multi Search的能力，所以必须借助底层的Elasticsearch原生客户端来实现。
• 这种方式的好处是完全复用你已经用NativeSearchQueryBuilder构建的查询逻辑，不需要重新写原生的查询DSL。

内容的提问来源于stack exchange，提问作者nalapoke