Question

我得到了一个简单的任务：为表中的每个唯一ID选择“id”，min和max值。所以我写了简单的group by，但查询需要花费很长时间才能执行（30-60秒）

SELECT     CHPDataElement.DataElementID, MIN(CHPDataElementData.UTCDataTime) AS MinDataTime, MAX(CHPDataElementData.UTCDataTime) AS MaxDataTime
FROM         CHPDataElement INNER JOIN
                      CHPDataElementData ON CHPDataElement.DataElementID = CHPDataElementData.DataElementID
GROUP BY CHPDataElement.DataElementID

order by 
CHPDataElement.DataElementID

所以我开始致力于改进。并提出了简单的迭代，它在0.3-0.5秒内返回相同的数据。

declare @result table
(
    DataElementID int,
    MinDataTime datetime NULL,
    MaxDataTime datetime null
)

declare @currentID int
declare @nextID int
declare @time datetime

insert into @result (DataElementID, MinDataTime, MaxDataTime)
select DataElementID,null,null from CHPDataElement
order by DataElementID

select top 1 @currentID=DataElementID from @result 

while @currentID is not null
begin
    print @currentID


    select top 1 @time=UTCDataTime  from CHPDataElementData
    where DataElementID = @currentID 
    order by UTCDataTime asc

    update @result set MinDataTime = @time
    where DataElementID = @currentID

    select top 1 @time=UTCDataTime  from CHPDataElementData
    where DataElementID = @currentID 
    order by UTCDataTime desc

    update @result set MaxDataTime = @time
    where DataElementID = @currentID



    set @nextID = null
    select top 1 @nextID=DataElementID from @result where DataElementID > @currentID
    set @currentID = @nextID
end

select * from @result

有没有人可以解释为什么'group by'与第二个查询相比效率低？

Answer 1

为CHPDataElementData提供DataElementID的索引。

Answer 2

在CHPDataElementData上为DataElementID, UTCDateTime添加索引，

CREATE NONCLUSTERED INDEX IX_CHPDataElementData_DataElementID_UTCDataTime
    ON CHPDataElementData
(
    DataElementID ASC,
    UTCDataTime ASC
)

然后使用此声明，

SELECT
            CHPDataElementData.DataElementID,
            MIN(CHPDataElementData.UTCDataTime) AS MinDataTime,
            MAX(CHPDataElementData.UTCDataTime) AS MaxDataTime
    FROM
            CHPDataElementData
    GROUP BY
            CHPDataElementData.DataElementID;

sql迭代vs group by，巨大的性能差异

2 个答案: