ES6分页from+size、search_after两种查询

1、from+size

//分页查询
@RequestMapping(value = "/get", method = RequestMethod.GET)
public BaseResponse<List<Object>> get(@RequestParam int from ,@RequestParam int size) {
    SearchSourceBuilder searchSourceBuilder = new SearchSourceBuilder();
    //排序
    searchSourceBuilder.sort("timestamp", SortOrder.ASC);
    //筛选返回的字段，只要referer和clientip
    searchSourceBuilder.fetchSource(new String[]{"referer", "clientip"},null);
    searchSourceBuilder.from(from*size);
    searchSourceBuilder.size(size);
    SearchResponse response = transportClient.prepareSearch("kibana_sample_data_logs").setTypes("_doc")
            //设置查询类型java
            .setSearchType(SearchType.DFS_QUERY_THEN_FETCH)
            .setSource(searchSourceBuilder)
            .get();
    SearchHits hits = response.getHits();
    Map<String, String> map = new HashMap<>();
    for(int i=0;i<hits.getHits().length;i++) {
        Map<String, Object> sourceAsMap = hits.getHits()[i].getSourceAsMap();
        System.out.println(JSONObject.toJSONString(sourceAsMap));
        map.put(hits.getHits()[i].getId(),hits.getHits()[i].getIndex());
    }
    return BaseResponse.ok(map);
}

建议：避免过度使用 from 和 size 来分页或一次请求太多结果，深度分页问题。

注意：max_result_window默认10000，分页查询超过10000会报错，可以修改这个值（不建议）

2、search_after

@RequestMapping(value = "/searchAfter", method = RequestMethod.GET)
    public BaseResponse<List<Object>> searchAfter(@RequestParam int size,
                                                  @RequestParam Long searchAfter) {
        SearchSourceBuilder searchSourceBuilder = new SearchSourceBuilder();
        //排序字段
        searchSourceBuilder.sort("timestamp", SortOrder.ASC);
        //筛选返回的字段，只要referer和clientip
        searchSourceBuilder.fetchSource(new String[]{"referer", "clientip"}, null);
        //上一页最后数据sort值，这个参数应该是个数组，数组的个数与排序字段个数一致，必填字段，首页时为-1，
        searchSourceBuilder.searchAfter(new Object[]{searchAfter});
        //不用传，默认0
        searchSourceBuilder.from(0);
        searchSourceBuilder.size(size);
        SearchResponse response = transportClient.prepareSearch("kibana_sample_data_logs").setTypes("_doc")
                //设置查询类型java
                .setSearchType(SearchType.DFS_QUERY_THEN_FETCH)
                .setSource(searchSourceBuilder)
                .get();
        SearchHits hits = response.getHits();
        Map<String, String> map = new HashMap<>();
        for (int i = 0; i < hits.getHits().length; i++) {
            Map<String, Object> sourceAsMap = hits.getHits()[i].getSourceAsMap();
            System.out.println(JSONObject.toJSONString(sourceAsMap));
            map.put(hits.getHits()[i].getId(), hits.getHits()[i].getIndex());
        }
        //这个值就是排序字段sort值，需要返回，下一页请求时带过来
        map.put("searchAfter",String.valueOf(Arrays.stream(hits.getHits()[hits.getHits().length-1].getRawSortValues()).findFirst().get()));
        return BaseResponse.ok(map);
    }

对比：

from size，深度分页或者size特别大的情况，会出深度分页问题，max_result_window也会阻预设的查询。

search_after: 性能优秀，是分页的优化。

深度分页问题可以从业务上想办法避免是最好的