SQL

Question

我不确定如何给这个问题起标题...希望没问题。

我有一个显示速率值的相当基本的数据集：

从这个逻辑：

SELECT  HotelId, CompetitorId,
        RateShopDate AS ShopDate,
        Rate, RateRemark,
        RowNumber
    FROM
    (
        SELECT  HotelId, CompetitorId, RateShopDate, ChannelId, RequestedRateType, Rate, RateRemark,
                PreviousRate, PreviousRateRemark, NextRate, NextRateRemark,
                ROW_NUMBER() OVER ( ORDER BY HotelId, CompetitorId, ChannelId, RequestedRateType, RateShopDate ) AS RowNumber
            FROM
            (
                SELECT  HotelId, CompetitorId, RateShopDate, ChannelId, RequestedRateType, Rate, RateRemark, RowNumber,
                        LAG( Rate ) OVER
                        (
                            PARTITION BY HotelId, CompetitorId, ChannelId, RequestedRateType
                            ORDER BY RowNumber
                        ) AS PreviousRate,
                        LAG( RateRemark ) OVER
                        (
                            PARTITION BY HotelId, CompetitorId, ChannelId, RequestedRateType
                            ORDER BY RowNumber
                        ) AS PreviousRateRemark,
                        LEAD( Rate ) OVER
                        (
                            PARTITION BY HotelId, CompetitorId, ChannelId, RequestedRateType
                            ORDER BY RowNumber
                        ) AS NextRate,
                        LEAD( RateRemark ) OVER
                        (
                            PARTITION BY HotelId, CompetitorId, ChannelId, RequestedRateType
                            ORDER BY RowNumber
                        ) AS NextRateRemark
                    FROM
                    (
                        SELECT  HotelId, CompetitorId, RateShopDate, ChannelId, RequestedRateType, Rate, RateRemark,
                                ROW_NUMBER() OVER ( ORDER BY HotelId, CompetitorId, ChannelId, RequestedRateType, RateShopDate ) AS RowNumber
                            FROM #TempRateShop
                    ) Rates
            ) PrevNextRates
                WHERE COALESCE(Rate, 0) <> COALESCE(PreviousRate, 0)
                OR COALESCE(RateRemark, '') <> COALESCE(PreviousRateRemark, '')
                OR COALESCE(Rate, 0) <> COALESCE(NextRate, 0)
                OR COALESCE(RateRemark, '') <> COALESCE(NextRateRemark, '')
    ) FilteredRates
        ORDER BY RateShopDate

我正在尝试获取每次汇率变化的开始和结束 ShopDate。但是由于利率有可能从一个变为另一个并回到原来的状态，我遇到了一点困难；例如价格在一段时间内从 90 美元涨到 95 美元，然后又回落到 90 美元。

我想得到的是：

我想出的唯一解决方案是使用 LEAD/LAG 并按 RowNumber 列排序，然后按 odd/even 数字过滤...但我真的不认为是最好的解决方案。

    LEAD ( RateShopDate ) OVER
    (
        ORDER BY RowNumber
    ) AS ShopEndDate

Answer 1

这是一种空隙和孤岛问题。

解决方法有很多，这里是一个：

使用 LAG 来识别每组开头的行
一个运行条件 COUNT 给我们每个组的 ID
然后简单地按那个 ID 分组

WITH PrevValues AS (
    SELECT *,
      IsStart = CASE WHEN LAG(rs.Rate, 1, -1) OVER (PARTITION BY rs.HotelId, rs.CompetitorId
                    ORDER BY rs.RateShopDate) <> rs.Rate THEN 1 END
    FROM #RateShop rs
),
Groupings AS (
    SELECT *,
      GroupId = COUNT(IsStart) OVER (PARTITION BY rs.HotelId, rs.CompetitorId
                    ORDER BY rs.RateShopDate ROWS UNBOUNDED PRECEDING)
    FROM PrevValues rs
)
SELECT
  rs.HotelId,
  rs.CompetitorId,
  StartDate  = MIN(rs.RateShopDate),
  EndDate    = MAX(rs.RateShopDate),
  Rate       = MIN(rs.Rate),
  RateRemark = STRING_AGG(rs.RateRemark, '; ')
FROM Groupings rs
GROUP BY
  rs.HotelId,
  rs.CompetitorId,
  rs.GroupId;

db<>fiddle

SQL - LEAD 和 LAG 查询

SQL - LEAD and LAG Query

tsql

sql-server

sql-server-2019