ES实现类似sql的group by后如何分页？

{
    "query": {
        ...... //搜索条件
    },
    "aggs": {
        "count": {   // COUNT(*)，统计GROUP BY后的总数
            "cardinality": {
                "field": "goods_id"    // 因为我这里GROUP BY的字段是goods_id，所以就用goods_id来计数了
            }
        },
        "goods_id": {
            "terms": {
                "field": "goods_id",    // 选择GROUP BY的字段
                "size": 20    // 取出20条GROUP BY的数据。数量应设置为sql中offset+limit的数量。注：其实es聚合操作不是很支持分页，于是只能先将数据取出，再对其做分页操作，可想而知页数越往后效率越低
            },
            "aggs": {
                "group": {
                    "top_hits": {
                        "sort": [
                            {
                                "stock_num": {
                                    "order": "desc"    // GROUP BY的数据如何排序，这里是根据stock_num 降序排列
                                }
                            }
                        ],
                        "_source": {    // 对应SQL的SELECT 
                            "includes": [
                                "goods_no"    // SELECT的列
                            ]
                        },
                        "size": 1    // es聚合时需要指定返回几条数据（即返回几条同一个goods_id的数据）我们做GROUP BY操作就只要写1就完事了
                    }
                },
                "r_bucket_sort": {    // 分页操作
                    "bucket_sort": {
                        "sort": [],
                        "from": 0,   // 对上面取出的20条数据分页，等价于SQL的OFFSET
                        "size": 10   // SQL的LIMIT
                    }
                }
            }
        }
    },
    "size": 0,   // 因为是做聚合操作，所以直接无视query筛选出的数据
    "from": 0
}

案例：统计业务应用流量数据

curl -XGET /action*/_search
{
  "from": 0,
  "size": 0,
  "query": {
    "bool": {
      "filter": [
        {
          "range": {
            "occurredAt": {
              "gte": "1659316790000",
              "lt": "1659323990000"
            }
          }
        }
      ],
      "must": [
        {
          "term": {
            "subtype.keyword": {
              "value": "/datatrans/traffic"
            }
          }
        }
      ]
    }
  },
  "aggs": {
    "count": {
      "cardinality": {
        "field": "applicationType.keyword"
      }
    },
    "group_by_app_type": {
      "terms": {
        "field": "applicationType.keyword",
        "min_doc_count": 1,
        "shard_min_doc_count": 0,
        "show_term_doc_count_error": false,
        "size": 40,
        "order": [
          {
            "_count": "desc"
          }
        ]
      },
      "aggs": {
        "sum_bytes_sent": {
          "sum": {
            "field": "bytesSent"
          }
        },
        "sum_bytes_received": {
          "sum": {
            "field": "bytesReceived"
          }
        },
        "sum_total_flow": {
          "sum": {
            "script": {
              "source": "(doc[\"bytesSent\"].value + doc[\"bytesReceived\"].value)"
            }
          }
        },
        "max_date": {
          "max": {
            "field": "occurredAt"
          }
        },
        "min_date": {
          "min": {
            "field": "occurredAt"
          }
        },
        "bucket_filed": {
          "bucket_sort": {
            "sort": [
              {
                "sum_bytes_sent": {
                  "order": "asc"
                }
              }
            ],
            "from": 0,
            "size": 40
          }
        }
      }
    }
  }
}'