ES Java High Level REST Client 聚合返回值解析总结

ES Java High Level REST Client 聚合返回值解析总结

在使用ES Java Rest High Client的时候，因为初学ES，所以对解析返回值一脸懵逼。
现针对返回值解析写下此文。

原始Rest请求

GET log/orderLog/_search?size=0
{
  "query":{
      "bool": {
        "must": [
          {"match_phrase": {
            "operator": "15061106601"
          }},
          {"range": {
            "time": {
            "gte": 1536019200000,
            "lte": 1536205600000
            }
          }}
        ]
      }
    },
    "aggs":{
      "day_order":{
        "date_histogram": {
        "field": "time",
        "interval": "day"
        }
        , "aggs": {
          "grap_order_success": {
            "filter": {
              "terms": {
                "operating": [
                  "抢",
                  "单",
                  "成",
                  "功"
                ]
              }
            },
            "aggs": {
              "order_success_count": {
                "value_count": {
                  "field": "time"
                }
              }
            }
          },
          "order_ok":{
            "filter": {
              "terms": {
                "operating": [
                  "确",
                  "认",
                  "方",
                  "案"
                ]
              }
            },
            "aggs": {
              "order_ok_count": {
                "value_count": {
                  "field": "time"
                }
              }
            }
          },
          "percent_customizer":{
            "bucket_script": {
              "buckets_path": {
                "orderGrap":"grap_order_success>order_success_count",
                "orderOk":"order_ok>order_ok_count"
              },
              "script": "params.orderOk/params.orderGrap*100"
            }
          }
        }
      }
    }
}

此请求的内容为：计算一段时间内，某一操作人确认方案和抢单成功的比，我们称之为确认方案率。计算方式为：

每日某人的抢单数量/每日某人的确认方案数量。

1. 先使用起止时间和操作人进行筛选；
2. 主要使用的Date Histogram Aggregation，对日期以天为间隔分桶；
3. 对每天的数据，进行以下过滤：
   
   1. 选出每天进行操作为“抢单成功”的文档（等一下说以下Term），使用fliter的terms，并统计数量，使用value count。
   2. 选出每天进行操作为“确认方案”的文档，并统计数量
4. 使用Bucket Script Aggregation（桶脚本聚合），计算确认方案率。

Java 代码

 public SearchResponse orderOkPercent(GrabOrderDTO grabOrderDTO) throws IOException {
        SearchRequest request = new SearchRequest(ES_INDEX);
        request.types(TYPE);

        SearchSourceBuilder searchSourceBuilder = new SearchSourceBuilder().size(0);
        BoolQueryBuilder boolQueryBuilder = QueryBuilders.boolQuery();

        //如果指定了定制师
        if (!Strings.isNullOrEmpty(grabOrderDTO.getOperator())){
            MatchPhraseQueryBuilder matchPhraseQueryBuilder = QueryBuilders.matchPhraseQuery("operator", grabOrderDTO.getOperator());
            boolQueryBuilder.must(matchPhraseQueryBuilder);
        }

        //如果指定了起始时间则开启范围查询
        if (!Strings.isNullOrEmpty(grabOrderDTO.getStartTime())){
            RangeQueryBuilder rangeQueryBuilder = QueryBuilders.rangeQuery("time");
            rangeQueryBuilder.gte(grabOrderDTO.getStartTime());
            rangeQueryBuilder.lte(grabOrderDTO.getEndTime());
            boolQueryBuilder.must(rangeQueryBuilder);
        }

        //日期直方图聚合
        DateHistogramAggregationBuilder dateHistogramAggregationBuilder = AggregationBuilders.dateHistogram("day_order");
        dateHistogramAggregationBuilder.field("time");

        //抢单成功聚合
        dateHistogramAggregationBuilder.dateHistogramInterval(setInterval(grabOrderDTO.getInterval()));
        AggregationBuilder grabAggregationBuilders =
                AggregationBuilders.filter("grab_order_success",
                        QueryBuilders.termsQuery("operating", "抢", "单","成","功"));
        ValueCountAggregationBuilder grabValueCountAggregationBuilder = AggregationBuilders.count("order_success_count").field("time");
        grabAggregationBuilders.subAggregation(grabValueCountAggregationBuilder);

        //确认方案聚合
        AggregationBuilder okAggregationBuilder =
                AggregationBuilders.filter("ok_order",
                    QueryBuilders.termsQuery("operating", "确", "认", "方", "案"));
        ValueCountAggregationBuilder okValueCountAggregationBuilder = AggregationBuilders.count("order_ok_count").field("time");
        okAggregationBuilder.subAggregation(okValueCountAggregationBuilder);

        // 脚本
        HashMap script = Maps.newHashMap();
        script.put("orderGrab", "grab_order_success>order_success_count");
        script.put("orderOk", "ok_order>order_ok_count");
        Script okOrderPercent = new Script("params.orderOk/params.orderGrab*100");
        BucketScriptPipelineAggregationBuilder bucketScriptPipelineAggregationBuilder = new BucketScriptPipelineAggregationBuilder("percent_customizer", script, okOrderPercent);

        dateHistogramAggregationBuilder.subAggregation(grabAggregationBuilders);
        dateHistogramAggregationBuilder.subAggregation(okAggregationBuilder);
        dateHistogramAggregationBuilder.subAggregation(bucketScriptPipelineAggregationBuilder);

        searchSourceBuilder.query(boolQueryBuilder).aggregation(dateHistogramAggregationBuilder);
        request.source(searchSourceBuilder);

        return client.search(request);
    }

不得不说，Java Rest High client 有一种让人望文生意的感觉，用起来还是很爽的。
直接添加查询条件，添加聚合、子聚合，添加source，然后查询就可以了，此处不再赘述。
官方api

返回结果

以下是返回的结果(保留java 返回的原格式，再名称前有聚合名称标注)：

{
    "took":3,
    "timed_out":false,
    "_shards":{
        "total":10,
        "successful":10,
        "skipped":0,
        "failed":0
    },
    "hits":{
        "total":10,
        "max_score":0,
        "hits":[
        ]
    },
    "aggregations":{
        "date_histogram#day_order":{
            "buckets":[
                {
                    "key_as_string":"1536019200000",
                    "key":1536019200000,
                    "doc_count":1,
                    "filter#ok_order":{
                        "doc_count":0,
                        "value_count#order_ok_count":{ "value":0 } },
                    "filter#grab_order_success":{
                        "doc_count":1,
                        "value_count#order_success_count":{ "value":1 } },
                    "simple_value#percent_customizer":{
                        "value":0 }
                },
                {
                    "key_as_string":"1536105600000",
                    "key":1536105600000,
                    "doc_count":9,
                    "filter#ok_order":{
                        "doc_count":3,
                        "value_count#order_ok_count":{ "value":3 } },
                    "filter#grab_order_success":{
                        "doc_count":6,
                        "value_count#order_success_count":{ "value":6 } },
                    "simple_value#percent_customizer":{
                        "value":50 }
                }
            ]
        }
    }
}

此处返回了两条记录。
再aggregations中，时一个名为day_order的date_histogram聚合，包含两个桶（bucket），获取桶的Java代码如下：

//此处其实返回的是一个data_date_histogram类型，'#'前面即表示类型
Aggregation aggregation = response.getAggregations().get("day_order");
ListHistogram.Bucket> buckets = ((Histogram)aggregation).getBuckets();
// 遍历返回的桶
for (Histogram.Bucket bucket : buckets){ 
    // 做你的处理
}

针对aggregations的第一条：

{
    "key_as_string":"1536019200000",
    "key":1536019200000,
    "doc_count":1,
    "filter#ok_order":{
        "doc_count":0,
        "value_count#order_ok_count":{
            "value":0
        }
    },
    "filter#grab_order_success":{
        "doc_count":1,
        "value_count#order_success_count":{
            "value":1
        }
    },
    "simple_value#percent_customizer":{
        "value":0
    }
}

结果中分别有：
1. ok_order，订单确认数，类型为filter，里面包含一个value_count 类型的数值计数order_ok_count
2. grab_order_success，订单抢单成功数量，结构同上
3. percent_customizer，类型为simple_value，为桶脚本聚合产生的结果的值。
话不多说，上代码：

// 获得Filter过滤,使用okFilter.getValue()，获取对应的值
Filter okFilter = bucket.getAggregations().get("ok_order");
// 获得ValueCount结果,使用okCount.getValue()
ValueCount okCount = okFilter.getAggregations().get("order_ok_count");
Filter grabFilter = bucket.getAggregations().get("grab_order_success");
ValueCount grabCount = grabFilter.getAggregations().get("order_success_count");
// 获得百分比，使用value.getValueAsString()获取对应的值
SimpleValue value = bucket.getAggregations().get("percent_customizer");

解析完整代码：

public void aggreation2Percent(SearchResponse response){
        Aggregation aggregation = response.getAggregations().get("day_order");
        List.Bucket> buckets = ((Histogram)aggregation).getBuckets();

        // 遍历返回的桶
        for (Histogram.Bucket bucket : buckets){
            PercentCustomizer percentCustomizer = new PercentCustomizer();
            percentCustomizer.setKey_as_string(bucket.getKeyAsString());
            // 获得Filter过滤
            Filter okFilter = bucket.getAggregations().get("ok_order");
            // 获得ValueCount结果
            ValueCount okCount = okFilter.getAggregations().get("order_ok_count");
            Filter grabFilter = bucket.getAggregations().get("grab_order_success");
            ValueCount grabCount = grabFilter.getAggregations().get("order_success_count");
        }
    }

总结

知道返回结果各个字段代表的意义后，就很简单了。可是之前一直在傻傻的想get后的值到底是什么意思…
知道是什么意思之后，敲起代码来就舒服多了。