0

Not sure if the title makes sense but I'll give this a shot.

Basically I have a report where I throw a bunch of data into a temp table and then display it on a webpage. The totals for this report need to be a weighted average but some data in the table throws this average off.

Here's an example of the data set in the temp table

CustomerNo  CountryCode qJAN    rJAN    gJAN    qFEB    rFEB    gFEB
IDS1        ID          1500    0       0       1500    0       0       
ID18J       ID          8000    15      120000  8000    15      120000      
ID21        ID          3000    0       0       0       0       0       
ID22J       ID          9000    0       0       12000   0       0       
ID22S       ID          9500    0       0       0       0       0       
ID22S       ID          9500    0       0       0       0       0       
ID31S       ID          3000    0       0       0       0       0

Where qCOL is quantity, rCOL is rate and gCOL is gross (quantity * rate).

Now values are only to count towards the country's total if there is a quantity and rate. The query I have been using is as follows:

SELECT CASE SUM(qJAN) WHEN 0 THEN 0 ELSE SUM(gJAN) / SUM(qJAN) END AS JAN,
       CASE SUM(qFEB) WHEN 0 THEN 0 ELSE SUM(gFEB) / SUM(qFEB) END AS FEB,
       ...
       CASE(SUM(qJAN) + SUM(qFEB) + ... )
       WHEN 0 THEN 0
       ELSE
       (SUM(gJAN) + SUM(gFEB) + ... ) / (SUM(qJAN) + SUM(qFEB) + ...) END AS TOT
FROM table
WHERE CountryCode = N'ID'

The above query returns the following weighted averages for JAN and FEB respectively: 3.24 and 4.90 both of which are wrong. The correct result would be 15.00 for both.

The problem is that the above takes the quantities that have no rates into account.

I can get around this by using subqueries but what I came up with started to get out of control and will really bog things down:

SELECT CASE 
(SELECT SUM(qJAN) FROM table WHERE CountryCode = N'ID' AND rJAN <> 0)
WHEN 0 THEN 0 ELSE 
SUM(gJAN) / 
(SELECT SUM(qJAN) FROM table WHERE CountryCode = N'ID' AND rJAN <> 0) 
END AS JAN,
CASE 
(SELECT SUM(qFEB) FROM table WHERE CountryCode = N'ID' AND rFEB <> 0)
WHEN 0 THEN 0 ELSE 
SUM(gFEB) / 
(SELECT SUM(qFEB) FROM table WHERE CountryCode = N'ID' AND rFEB <> 0) 
END AS FEB,
...
CASE 
(
(SELECT SUM(qJAN) FROM table WHERE CountryCode = N'ID' AND rJAN <> 0) + 
(SELECT SUM(qFEB) FROM table WHERE CountryCode = N'ID' AND rFEB <> 0) + 
...
) 
WHEN 0 THEN 0 ELSE 
(SUM(gJAN) + SUM(gFEB) + ...) / 
(
(SELECT SUM(qJAN) FROM table WHERE CountryCode = N'ID' AND rJAN <> 0) + 
(SELECT SUM(qFEB) FROM table WHERE CountryCode = N'ID' AND rFEB <> 0) + ...
) END AS TOT

This returns the correct totals but as you can see it gets ugly fast. All these subqueries are going to kill performance.

Is there a simpler way to get the result I'm looking for? I feel like there must be but I'm completely brain-locking on this.

Thanks.

4

1 回答 1

0

为什么不使用 PIVOT？

多个枢轴？每月需要按小时计算

并且有几个关于 sqlfiddle 的例子。

于 2013-10-11T17:50:17.973 回答