By行处理数据,推荐2种方式:
1、游标
2、While循环
我们来了解下这两种方案处理1w行数据分别需要多长时间。
一、游标。
首先我们填充一个表,用优雅的递归方式填充。
create table Orders(OrderID int,CostValue decimal(18,2) ) ;with cte_temp as ( select 1 as OrderID union all select OrderID+1 from cte_temp where OrderID<10000 ) insert into Orders(OrderID) select OrderID from cte_temp option (maxrecursion 32767);
现在我们的订单表Orders有了一万条订单,但是CostValue还是NULL值。
我们用游标的方式给每一条订单添加一个CostValue,耗时44s。
--游标 DECLARE @OrderID int DECLARE cursor_CostValue CURSOR FOR SELECT OrderID FROM Orders OPEN cursor_CostValue FETCH NEXT FROM cursor_CostValue INTO @OrderID WHILE @@FETCH_STATUS = 0 BEGIN UPDATE Orders SET CostValue = OrderID+100 WHERE OrderID = @OrderID FETCH NEXT FROM cursor_CostValue INTO @OrderID END CLOSE cursor_CostValue DEALLOCATE cursor_CostValue
二、While循环
将数据放在临时表中,然后操作临时表,最后更新回总表。耗时16s。
DECLARE @RowID int -- 获取待处理的数据记录到临时表 -- 字段说明:RowID:记录行号 / DealFlg:行处理标识 SELECT RowID = IDENTITY(INT , 1, 1),DealFlg=0,OrderID,CostValue = 0 INTO #Tmp FROM Orders SELECT @RowID = MIN(RowID) FROM #Tmp WHERE DealFlg = 0 -- 若最小行号不为空(有需要处理的数据) WHILE @RowID IS NOT NULL BEGIN UPDATE #Tmp SET DealFlg = 1,CostValue=OrderID+100 WHERE RowID = @RowID SELECT @RowID = MIN(RowID) FROM #Tmp WHERE DealFlg = 0 END update O set O.CostValue=T.CostValue from Orders O inner join #Tmp T on O.OrderID=T.OrderID
还有一种错误的While循环,即不把数据放在临时表中,直接操作本表,会大大增加耗时。
因为多次调用本表,如果在生产环境,将是一个灾难。
DECLARE @OrderID INT --表中OrderID最小的值 SELECT @OrderID = MIN(OrderID) FROM Orders where CostValue is null WHILE @OrderID IS NOT NULL BEGIN UPDATE Orders SET CostValue = OrderID+100 WHERE OrderID = @OrderID SELECT @OrderID = MIN(OrderID) FROM Orders where CostValue is null END