用上一行更新行答案

【问题标题】：Update Row with Previous Row用上一行更新行
【发布时间】：2021-06-08 17:33:11
【问题描述】：

我有下表：

我正在尝试编写一个更新，用大于零的先前值填充零

我尝试了以下查询但没有成功

With cte As
(
    SELECT [ID], [Q_G_R_BUOM], ROW_NUMBER() OVER (ORDER BY [ID] Asc) AS RN
    FROM [Testing_Lag]
)
update cte set [Q_G_R_BUOM]=iif([Q_G_R_BUOM]>0,[Q_G_R_BUOM],(SELECT [Q_G_R_BUOM] FROM cte WHERE [ID]=[ID]-1))

With cte As
(
    SELECT [ID], [Q_G_R_BUOM], ROW_NUMBER() OVER (ORDER BY [ID] Asc) AS RN
    FROM [Testing_Lag]
)
update cte set [Q_G_R_BUOM]=iif([Q_G_R_BUOM]>0,[Q_G_R_BUOM],(Lag([Q_G_R_BUOM], 1) OVER(ORDER BY [ID] ASC)))

如果有人能帮助我，我将不胜感激谢谢

【问题讨论】：

标签： sql asp.net sql-server tsql azure-sql-database

【解决方案1】：

SQL Server 中的一种方法使用可更新的 CTE 和窗口函数：

with toupdate as (
      select tl.*,
             max(Q_G_R_BUOM) over (partition by grp) as imputed_Q_G_R_BUOM
      from (select tl.*,
                   sum(case when Q_G_R_BUOM <> 0 then 1 else 0 end) over (order by id) as grp
            from testing_lag tl
           ) tl
     ) 
update toupdate
    set Q_G_R_BUOM = imputed_Q_G_R_BUOM
    where Q_G_R_BUOM = 0;

这会根据直到该行的非零值的数量为每一行分配一个分组。然后它将该值“传播”到整个组。

另一种方法是使用apply：

update tl
    set q_g_r_buom = tl2.q_g_r_buom
    from test_lag tl cross apply
         (select top (1) tl2.*
          from test_lag tl2
          where tl2.q_g_r_buom <> 0 and
                tl2.id < tl.id
          order by tl2.id
         ) tl2
    where tl.q_g_r_buom = 0

【讨论】：

感谢回答，评论说更新一定要根据Asc ID，请问这个查询是这样工作的吗？
@Laureanorubioaguilar 。 . .这两个答案都依赖于id 来定义表格的顺序。两者都基于 id 采用“前面的”非0 值。

【解决方案2】：

这个怎么样？

--DROP TABLE #C
CREATE TABLE #C (ID INT, Q_G_R_BUOM INT)
INSERT INTO #C values(1, 1805)
INSERT INTO #C values(2, 0)
INSERT INTO #C values(3, 0)
INSERT INTO #C values(4, 4732)
INSERT INTO #C values(5, 0)
INSERT INTO #C values(6, 0)

select * from #c

如果您的表中有空值，则解决方案会有所不同。

SELECT ID,CASE WHEN Q_G_R_BUOM >0
            THEN Q_G_R_BUOM
            ELSE (SELECT max(Q_G_R_BUOM)
                  FROM #C
                  WHERE ID <= t.ID)
       END AS X
FROM #C t

DROP TABLE #C
CREATE TABLE #C (ID INT, Q_G_R_BUOM INT)
INSERT INTO #C values(1, 1805)
INSERT INTO #C values(2, null)
INSERT INTO #C values(3, null)
INSERT INTO #C values(4, 4732)
INSERT INTO #C values(5, null)
INSERT INTO #C values(6, null)

select * from #c

SELECT ID,
COALESCE(Q_G_R_BUOM,
MAX(COALESCE(Q_G_R_BUOM,'')) OVER (ORDER BY ID  ROWS BETWEEN UNBOUNDED PRECEDING AND  1 PRECEDING)) AS Name
FROM #c

【讨论】：