SQL：向后获取时间序列的结果

Question

给出的是每天保存的实体的库存值表。我想得到：

给出今天缺货（数量 = 0）且前一天没有缺货的所有实体。

这是前天与前天相比的前天。以此类推。

我的方法有效，但我必须手动在 WHERE 条件下创建部件。

2019年全年查询怎么样？

#standardSQL
WITH
  TableStockDaily AS (
  SELECT TIMESTAMP('2019-10-10 00:00:00 UTC') AS ExportDate, 1001 AS EntityId, 0 as Qty UNION ALL
  SELECT TIMESTAMP('2019-10-10 00:00:00 UTC') AS ExportDate, 1002 AS EntityId, 0 as Qty UNION ALL
  SELECT TIMESTAMP('2019-10-10 00:00:00 UTC') AS ExportDate, 1003 AS EntityId, 1 as Qty UNION ALL
  SELECT TIMESTAMP('2019-10-11 00:00:00 UTC') AS ExportDate, 1001 AS EntityId, 1 as Qty UNION ALL
  SELECT TIMESTAMP('2019-10-11 00:00:00 UTC') AS ExportDate, 1002 AS EntityId, 1 as Qty UNION ALL
  SELECT TIMESTAMP('2019-10-11 00:00:00 UTC') AS ExportDate, 1003 AS EntityId, 0 as Qty UNION ALL
  SELECT TIMESTAMP('2019-10-12 00:00:00 UTC') AS ExportDate, 1001 AS EntityId, 0 as Qty UNION ALL
  SELECT TIMESTAMP('2019-10-12 00:00:00 UTC') AS ExportDate, 1002 AS EntityId, 1 as Qty UNION ALL
  SELECT TIMESTAMP('2019-10-12 00:00:00 UTC') AS ExportDate, 1003 AS EntityId, 1 as Qty UNION ALL
  SELECT TIMESTAMP('2019-10-13 00:00:00 UTC') AS ExportDate, 1001 AS EntityId, 1 as Qty UNION ALL
  SELECT TIMESTAMP('2019-10-13 00:00:00 UTC') AS ExportDate, 1002 AS EntityId, 0 as Qty UNION ALL
  SELECT TIMESTAMP('2019-10-13 00:00:00 UTC') AS ExportDate, 1003 AS EntityId, 0 as Qty UNION ALL
  SELECT TIMESTAMP('2019-10-14 00:00:00 UTC') AS ExportDate, 1001 AS EntityId, 0 as Qty UNION ALL
  SELECT TIMESTAMP('2019-10-14 00:00:00 UTC') AS ExportDate, 1002 AS EntityId, 0 as Qty UNION ALL
  SELECT TIMESTAMP('2019-10-14 00:00:00 UTC') AS ExportDate, 1003 AS EntityId, 1 as Qty UNION ALL
  SELECT TIMESTAMP('2019-10-15 00:00:00 UTC') AS ExportDate, 1001 AS EntityId, 1 as Qty UNION ALL
  SELECT TIMESTAMP('2019-10-15 00:00:00 UTC') AS ExportDate, 1002 AS EntityId, 1 as Qty UNION ALL
  SELECT TIMESTAMP('2019-10-15 00:00:00 UTC') AS ExportDate, 1003 AS EntityId, 0 as Qty UNION ALL
  SELECT TIMESTAMP('2019-10-16 00:00:00 UTC') AS ExportDate, 1001 AS EntityId, 1 as Qty UNION ALL
  SELECT TIMESTAMP('2019-10-16 00:00:00 UTC') AS ExportDate, 1002 AS EntityId, 1 as Qty UNION ALL
  SELECT TIMESTAMP('2019-10-16 00:00:00 UTC') AS ExportDate, 1003 AS EntityId, 1 as Qty
  )
SELECT
  sd1.ExportDate AS DateOutOfStock,
  sd2.ExportDate AS DateNotOutOfStock,
  sd1.EntityId AS EntityId,
  sd1.Qty AS Qty1,
  sd2.Qty AS Qty2
FROM
  TableStockDaily sd1
LEFT JOIN
  TableStockDaily sd2
ON
  sd1.EntityId = sd2.EntityId
WHERE
  sd1.Qty = 0
  AND sd2.Qty > 0
  AND sd1.ExportDate > sd2.ExportDate
  AND
  (
    ( 
      DATE(sd1.ExportDate) <= DATE_ADD(DATE(CURRENT_TIMESTAMP()), INTERVAL -1 DAY)
      AND DATE(sd1.ExportDate) >= DATE_ADD(DATE(CURRENT_TIMESTAMP()), INTERVAL -2 DAY)
      AND DATE(sd2.ExportDate) <= DATE_ADD(DATE(CURRENT_TIMESTAMP()), INTERVAL -2 DAY)
      AND DATE(sd2.ExportDate) > DATE_ADD(DATE(CURRENT_TIMESTAMP()), INTERVAL -3 DAY)
    )
    OR
    ( 
      DATE(sd1.ExportDate) <= DATE_ADD(DATE(CURRENT_TIMESTAMP()), INTERVAL -2 DAY)
      AND DATE(sd1.ExportDate) >= DATE_ADD(DATE(CURRENT_TIMESTAMP()), INTERVAL -3 DAY)
      AND DATE(sd2.ExportDate) <= DATE_ADD(DATE(CURRENT_TIMESTAMP()), INTERVAL -3 DAY)
      AND DATE(sd2.ExportDate) > DATE_ADD(DATE(CURRENT_TIMESTAMP()), INTERVAL -4 DAY)
    )
    OR
    ( 
      DATE(sd1.ExportDate) <= DATE_ADD(DATE(CURRENT_TIMESTAMP()), INTERVAL -3 DAY)
      AND DATE(sd1.ExportDate) >= DATE_ADD(DATE(CURRENT_TIMESTAMP()), INTERVAL -4 DAY)
      AND DATE(sd2.ExportDate) <= DATE_ADD(DATE(CURRENT_TIMESTAMP()), INTERVAL -4 DAY)
      AND DATE(sd2.ExportDate) > DATE_ADD(DATE(CURRENT_TIMESTAMP()), INTERVAL -5 DAY)
    )
    OR
    ( 
      DATE(sd1.ExportDate) <= DATE_ADD(DATE(CURRENT_TIMESTAMP()), INTERVAL -4 DAY)
      AND DATE(sd1.ExportDate) >= DATE_ADD(DATE(CURRENT_TIMESTAMP()), INTERVAL -5 DAY)
      AND DATE(sd2.ExportDate) <= DATE_ADD(DATE(CURRENT_TIMESTAMP()), INTERVAL -5 DAY)
      AND DATE(sd2.ExportDate) > DATE_ADD(DATE(CURRENT_TIMESTAMP()), INTERVAL -6 DAY)
    )
    OR
    ( 
      DATE(sd1.ExportDate) <= DATE_ADD(DATE(CURRENT_TIMESTAMP()), INTERVAL -5 DAY)
      AND DATE(sd1.ExportDate) >= DATE_ADD(DATE(CURRENT_TIMESTAMP()), INTERVAL -6 DAY)
      AND DATE(sd2.ExportDate) <= DATE_ADD(DATE(CURRENT_TIMESTAMP()), INTERVAL -6 DAY)
      AND DATE(sd2.ExportDate) > DATE_ADD(DATE(CURRENT_TIMESTAMP()), INTERVAL -7 DAY)
    )
  )

Answer 1

我认为您正在寻找LAG()。此窗口函数可用于根据排序标准（此处为 ExportDate）恢复一组记录（此处为共享相同 EntityID 的记录）中给定列的值：

SELECT *
FROM (
    SELECT
        EntityId,
        ExportDate AS DateOutOfStock,
        Qty AS QtyOutOfStock,
        LAG(ExportDate) OVER(PARTITION BY EntityId ORDER BY ExportDate) AS DateNotOutOfStock,
        LAG(Qty) OVER(PARTITION BY EntityId ORDER BY ExportDate) AS QtyNotOutOfStock
    FROM
      TableStockDaily sd1
) x
WHERE QtyOutOfStock = 0 AND QtyNotOutOfStock > 0