ES對應mysql的group by分組查詢javaApi，多對多關係的分組查詢

比如我這邊有個下列訂單索引資料，現在的需求是按使用者(fmerchantId)和支付方式(fchannelId)進行分組統計訂單總金額(famt)和總筆數，其中使用者和支付方式是多對多的關係，就是一個使用者會對應多個支付方式，一個支付方式會對應多個使用者

{
          "famt": "2",
          "fbankCode": "0000_0002",
          "fbankName": "支付寶",
          "fchannelCode": "ALIPAY",
          "fchannelId": "993",
          "fchannelName": "支付寶",
          "fchannelTradeNo": "2020072222001439181419030679",
          "fchgAgenCode": "111111",
          "fcreateDate": "2020-07-22",
          "fcreateDay": "22",
          "fcreateMonth": "07",
          "fcreateTime": "2020-07-22 14:46:57",
          "fcreateYear": "2020",
          "fdeviceType": "phone",
          "fmerchantId": "5200002020072001",
          "fmerchantName": "測試使用者",
          "forderNo": "483325941654679552",
          "fpayCode": "36000019115000000287",
          "fthirdpayTradeNo": "2007221446570296",
          "ftradeStatus": "1"
        }
 

1.建立查詢條件，相當於mysql的where條件

其中SearchSourceBuilder相當於mysql中的一條完整sql語句，BoolQueryBuilder相當於where條件，根據需求自行新增where條件，最後把boolQueryBuilder的where條件新增到SearchSourceBuilder的sql中

//查詢條件
SearchSourceBuilder searchSourceBuilder = new SearchSourceBuilder();
BoolQueryBuilder boolQueryBuilder =  QueryBuilders.boolQuery();
//成功狀態的支付訂單
boolQueryBuilder.must(QueryBuilders.termsQuery("ftradeStatus", Arrays.asList("1", "2", "3", "4")));
//使用者，支付方式不能為空
boolQueryBuilder.must(QueryBuilders.existsQuery("fmerchantId"));
boolQueryBuilder.must(QueryBuilders.existsQuery("fchannelId"));
boolQueryBuilder.must(QueryBuilders.existsQuery("fchannelName"));

String startDate = queryMap.get("startDate") == null ? null : queryMap.get("startDate").toString();
String endDate = queryMap.get("endDate") == null ? null : queryMap.get("endDate").toString();
//大於等於
if (!StringUtils.isEmpty(startDate)) {
    boolQueryBuilder.must(QueryBuilders.rangeQuery("fcreateDate").gte(startDate));
}
//小於
if (!StringUtils.isEmpty(endDate)) {
    boolQueryBuilder.must(QueryBuilders.rangeQuery("fcreateDate").lte(endDate));
}
//新增查詢條件
searchSourceBuilder.query(boolQueryBuilder);
 

2.新增分組條件，相當於group by條件

TermsAggregationBuilder相當於mysql中的group by分組查詢條件欄位，建立要分組的各個欄位TermsAggregationBuilder，AggregationBuilders.terms("fchannelTypeId").field(
"fchannelId").size(searchSize).order(BucketOrder.key(true))中terms是建立的別名欄位（類似mysql select a as "A"）,field是索引中的欄位，size可設定查詢數量大小，order進行排序。

然後進行group by欄位的拼接，用termsAggregationBuilder.subAggregation(termsAggregationBuilder3)，相當於group by a,b欄位，a和b都是欄位.

注意ES中是可以拼接物件的，比如我先執行termsAggregationBuilder2.subAggregation(AggregationBuilders.sum("money").field("famt"))，這是根據使用者ID和使用者總金額分組了；再執行termsAggregationBuilder.subAggregation(termsAggregationBuilder2)，相當於使用者ID和使用者總金額分組當作一個物件b和支付方式a欄位一起分組了（group by a,b），其中a是一個欄位，b是一個物件（b中包含使用者和總金額的分組），這就是ES的分組內再分組

//按聚合名稱標識對桶進行升序排序
TermsAggregationBuilder termsAggregationBuilder = AggregationBuilders.terms("fchannelTypeId").field(
    "fchannelId").size(searchSize).order(BucketOrder.key(true));//支付方式ID
TermsAggregationBuilder termsAggregationBuilder2 = AggregationBuilders.terms("fmerchantTypeId").field(
    "fmerchantId");//使用者ID編號
TermsAggregationBuilder termsAggregationBuilder3 = AggregationBuilders.terms("fbankTypeCode").field(
    "fbankCode");//渠道商編號
TermsAggregationBuilder termsAggregationBuilder4 = AggregationBuilders.terms("fbankTypeName").field(
    "fbankName.keyword");//渠道商名稱
TermsAggregationBuilder termsAggregationBuilder5 = AggregationBuilders.terms("fchannelTypeName").field(
    "fchannelName.keyword");//支付渠道名稱

//1.先按支付渠道，渠道商編號，渠道商名稱，支付渠道名稱進行分組
termsAggregationBuilder.subAggregation(termsAggregationBuilder3).subAggregation(termsAggregationBuilder4).subAggregation(termsAggregationBuilder5);
//2.再在商戶編號裡統計金額分組
termsAggregationBuilder2.subAggregation(AggregationBuilders.sum("money").field("famt"));
//2.1按金額倒序排列
List<FieldSortBuilder> fieldSorts=new ArrayList<>();
fieldSorts.add(new FieldSortBuilder("money").order(SortOrder.DESC));
termsAggregationBuilder2.subAggregation(new BucketSortPipelineAggregationBuilder("bucket_field", fieldSorts).size(searchSize));
//3.拼接分組
termsAggregationBuilder.subAggregation(termsAggregationBuilder2);

3.執行查詢語句

這個總的語句相當與 select（支付方式，其他欄位，(使用者，sum(amt) ) as b ）as a,sum(amt) from 表 group by a ,其中a是以支付方式為主鍵的一個分組物件，a物件中包含了支付方式，其他欄位和使用者物件b的分組。b物件是以使用者為主鍵的使用者，使用者總金額分組。

//總的分組，把第二步建立的分組看作一個物件，在進行總分組
searchSourceBuilder.aggregation(termsAggregationBuilder);
searchSourceBuilder.aggregation(AggregationBuilders.sum("totalAmt").field("famt"));
//執行ES的查詢
SearchResponse response = ESUtils.findAll(payTradeIndex, payTradeType, searchSourceBuilder, null);

4.取值物件

取值總金額，上面最外層的sum(amt)就是所用訂單的總金額

取值a物件的分組，可以獲取分組欄位的值和b物件

再取值b物件裡面的使用者和金額（這個金額就是對應的每個使用者和支付方式分組統計的總金額了）

Aggregations aggregations = response.getAggregations();
Sum totalAmtSum = aggregations.get("totalAmt");
//總金額
double totalAmt = totalAmtSum.getValue();
DecimalFormat df = new DecimalFormat("0.00");
String totalMoney = df.format(totalAmt / 100);

Map<String, Aggregation> aggMap = response.getAggregations().asMap();
ParsedStringTerms codeTerms = (ParsedStringTerms) aggMap.get("fchannelTypeId");
Iterator<Terms.Bucket> codeBucketIt = (Iterator<Terms.Bucket>) codeTerms.getBuckets().iterator();
while (codeBucketIt.hasNext()) {
    Terms.Bucket codeBucket = codeBucketIt.next();

    //使用者編號的分組terms物件
    ParsedStringTerms nameTerms = (ParsedStringTerms) codeBucket.getAggregations().asMap().get("fmerchantTypeId");
    //渠道商名稱
    ParsedStringTerms nameTerms1 = (ParsedStringTerms) codeBucket.getAggregations().asMap().get("fbankTypeName");
    String fbankTypeName = nameTerms1.getBuckets().get(0).getKey().toString();
    //渠道商編號
    ParsedStringTerms nameTerms2 = (ParsedStringTerms) codeBucket.getAggregations().asMap().get("fbankTypeCode");
    String fbankTypeCode = nameTerms2.getBuckets().get(0).getKey().toString();
    //支付渠道名稱
    ParsedStringTerms nameTerms3 = (ParsedStringTerms) codeBucket.getAggregations().asMap().get("fchannelTypeName");
    String fchannelTypeName = nameTerms3.getBuckets().get(0).getKey().toString();


    Iterator<Terms.Bucket> nameBucketIt = (Iterator<Terms.Bucket>) nameTerms.getBuckets().iterator();
    while (nameBucketIt.hasNext()){
        Terms.Bucket nameBucket = nameBucketIt.next();
        //金額
        Sum term = nameBucket.getAggregations().get("money");
        String money = df.format(term.getValue() / 100);
        //使用者編號
        String fmerchantTypeId = nameBucket.getKey().toString();
        //統計筆數
        Long count = nameBucket.getDocCount();

        esDataList.add(new MerchantChannelCountModel(fmerchantTypeId,null,fbankTypeCode,fbankTypeName,
                                                     String.valueOf(codeBucket.getKey()),fchannelTypeName,money,
                                                     count.intValue()));
    }

}