Elasticsearch — 聚合查詢，先分組後排序

需求

      對明細數據先按waybillId分組，再按eventTime降序，取最新一條數據。

桶聚合（bucket）

   Elasticsearch桶聚合，目的就是數據分組，先將數據按指定的條件分成多個組，然後對每一個組進行統計。

1. ES SQL寫法

{
  "query": {
    "bool": {
      "must": [
        {
          "match_all": {}
        }
      ]
    }
  },
  "aggs": {
    "waybillIdAgg": {
      "terms": {
        "field": "waybillId",
        "size": 1000,
        "min_doc_count": 1
      },
      "aggs": {
        "top1": {
          "top_hits": {
            "size": 1,
            "sort": [
              {
                "eventTime": {
                  "order": "desc"
                }
              }
            ]
          }
        }
      }
    }
  }
}

返回結果如下：

    

2. Java Elasticsearch寫法及結果解析

    // 查詢條件
    BoolQueryBuilder queryBool = QueryBuilders.boolQuery();

    BoolQueryBuilder inFilter = new BoolQueryBuilder();
    waybillIds.forEach(
        waybillId -> inFilter.should(QueryBuilders.termQuery("waybillId", waybillId)));
    queryBool.must(inFilter);

    // 桶聚合(bucket),按waybillId分組
    TermsAggregationBuilder termsAggregationBuilder =
        AggregationBuilders.terms("waybillIdAgg").field("waybillId").size(1000).minDocCount(1);
    // 嵌套桶，再按時間倒序取第一條數據
    TopHitsAggregationBuilder sort =
        AggregationBuilders.topHits("top1").size(1).sort("eventTime", SortOrder.DESC);
    termsAggregationBuilder.subAggregation(sort);

    SearchSourceBuilder searchSourceBuilder =
        SearchSourceBuilder.searchSource().query(queryBool).aggregation(termsAggregationBuilder);

    // 查詢請求
    SearchRequest searchRequest = new SearchRequest(esIndexConfig.getIndexNameTrackingDetail());
    searchRequest.source(searchSourceBuilder);

    // 執行查詢
    SearchResponse searchResponse =
        restHighLevelClient.search(searchRequest, RequestOptions.DEFAULT);
    // 處理聚合查詢結果
    Aggregations aggregations = searchResponse.getAggregations();
    Terms terms = aggregations.get("waybillIdAgg");
    List<TrackingDetail> trackingDetails =
        terms.getBuckets().stream()
            .map(
                t -> {
                  Aggregation top1 = t.getAggregations().get("top1");
                  Optional<SearchHit> first =
                      Arrays.stream(((ParsedTopHits) top1).getHits().getHits()).findFirst();
                  if (first.isPresent()) {
                    Map<String, Object> trackingDetailMap = first.get().getSourceAsMap();
                    return BeanUtil.fillBeanWithMap(trackingDetailMap, new TrackingDetail(), false);
                  }
                  return null;
                })
            .collect(Collectors.toList());