1. 技术文章
  2. SQL Server Pivot Table with multiple column aggregates

SQL Server Pivot Table with multiple column aggregates

由 陌路散爱 提交于 2019-11-26 11:48:00

I would do this slightly different by applying both the UNPIVOT and the PIVOT functions to get the final result. The unpivot takes the values from both the totalcount and totalamount columns and places them into one column with multiple rows. You can then pivot on those results.:

select chardate,
  Australia_totalcount as [Australia # of Transactions], 
  Australia_totalamount as [Australia Total $ Amount],
  Austria_totalcount as [Austria # of Transactions], 
  Austria_totalamount as [Austria Total $ Amount]
from
(
  select 
    numericmonth, 
    chardate,
    country +'_'+col col, 
    value
  from
  (
    select numericmonth, 
      country, 
      chardate,
      cast(totalcount as numeric(10, 2)) totalcount,
      cast(totalamount as numeric(10, 2)) totalamount
    from mytransactions
  ) src
  unpivot
  (
    value
    for col in (totalcount, totalamount)
  ) unpiv
) s
pivot
(
  sum(value)
  for col in (Australia_totalcount, Australia_totalamount,
              Austria_totalcount, Austria_totalamount)
) piv
order by numericmonth

See SQL Fiddle with Demo.

If you have an unknown number of country names, then you can use dynamic SQL:

DECLARE @cols AS NVARCHAR(MAX),
    @colsName AS NVARCHAR(MAX),
    @query  AS NVARCHAR(MAX)

select @cols = STUFF((SELECT distinct ',' + QUOTENAME(country +'_'+c.col) 
                      from mytransactions
                      cross apply 
                      (
                        select 'TotalCount' col
                        union all
                        select 'TotalAmount'
                      ) c
            FOR XML PATH(''), TYPE
            ).value('.', 'NVARCHAR(MAX)') 
        ,1,1,'')

select @colsName 
    = STUFF((SELECT distinct ', ' + QUOTENAME(country +'_'+c.col) 
               +' as ['
               + country + case when c.col = 'TotalCount' then ' # of Transactions]' else 'Total $ Amount]' end
             from mytransactions
             cross apply 
             (
                select 'TotalCount' col
                union all
                select 'TotalAmount'
             ) c
            FOR XML PATH(''), TYPE
            ).value('.', 'NVARCHAR(MAX)') 
        ,1,1,'')

set @query 
  = 'SELECT chardate, ' + @colsName + ' 
     from 
     (
      select 
        numericmonth, 
        chardate,
        country +''_''+col col, 
        value
      from
      (
        select numericmonth, 
          country, 
          chardate,
          cast(totalcount as numeric(10, 2)) totalcount,
          cast(totalamount as numeric(10, 2)) totalamount
        from mytransactions
      ) src
      unpivot
      (
        value
        for col in (totalcount, totalamount)
      ) unpiv
     ) s
     pivot 
     (
       sum(value)
       for col in (' + @cols + ')
     ) p 
     order by numericmonth'

execute(@query)

See SQL Fiddle with Demo

Both give the result:

|             CHARDATE | AUSTRALIA # OF TRANSACTIONS | AUSTRALIA TOTAL $ AMOUNT | AUSTRIA # OF TRANSACTIONS | AUSTRIA TOTAL $ AMOUNT |
--------------------------------------------------------------------------------------------------------------------------------------
| Jul-12               |                          36 |                   699.96 |                        11 |                 257.82 |
| Aug-12               |                          44 |                  1368.71 |                         5 |                 126.55 |
| Sep-12               |                          52 |                  1161.33 |                         7 |                  92.11 |
| Oct-12               |                          50 |                  1099.84 |                        12 |                 103.56 |
| Nov-12               |                          38 |                  1078.94 |                        21 |                 377.68 |
| Dec-12               |                          63 |                  1668.23 |                         3 |                  14.35 |

I have added the dynamic query/solution.

Static

SELECT  t.chardate,
        SUM(CASE WHEN t.country='Australia' THEN t.totalcount ELSE 0 END) AS "Australia # of Transactions",
        SUM(CASE WHEN t.country='Australia' THEN t.totalamount ELSE 0 END) AS "Australia Total $ amount",
        SUM(CASE WHEN t.country='Austria' THEN t.totalcount ELSE 0 END) AS "Austria # of Transactions",
        SUM(CASE WHEN t.country='Austria' THEN t.totalamount ELSE 0 END) AS "Austria Total $ amount"
FROM    mytransactions t
GROUP BY t.chardate;

Note:

1) ORDER BY t.chardate will not work because the values from chardate column are chars.

2) My advice is to split chardate in two columns numericmonth and numericyear. In this last case, you could use this solution:

SELECT  t.numericyear, t.numericmonth,
        SUM(CASE WHEN t.country='Australia' THEN t.totalcount ELSE 0 END) AS "Australia # of Transactions",
        SUM(CASE WHEN t.country='Australia' THEN t.totalamount ELSE 0 END) AS "Australia Total $ amount",
        SUM(CASE WHEN t.country='Austria' THEN t.totalcount ELSE 0 END) AS "Austria # of Transactions",
        SUM(CASE WHEN t.country='Austria' THEN t.totalamount ELSE 0 END) AS "Austria Total $ amount"
FROM    mytransactions t
GROUP BY t.numericyear, t.numericmonth
ORDER BY BY t.numericyear, t.numericmonth;

Dynamic

DECLARE @Sql NVARCHAR(MAX)='SELECT t.chardate';
DECLARE @ColumnTemplate NVARCHAR(MAX)='SUM(CASE WHEN t.country=''{country}'' THEN t.totalcount ELSE 0 END) AS "{country} # of Transactions"
,SUM(CASE WHEN t.country=''{country}'' THEN t.totalamount ELSE 0 END) AS "{country} Total $ amount"'

SELECT @Sql=@Sql+CHAR(13)+','+REPLACE(@ColumnTemplate, '{country}', REPLACE(c.name,'''','''''')e)
FROM (
    SELECT  DISTINCT t.country AS name
    FROM    mytransactions t
) c

SELECT @Sql=@Sql+'
FROM mytransactions t
GROUP BY t.chardate;'

PRINT @Sql;


EXEC(@Sql);

Results:

SELECT t.chardate
,SUM(CASE WHEN t.country='Australia' THEN t.totalcount ELSE 0 END) AS "Australia # of Transactions"
,SUM(CASE WHEN t.country='Australia' THEN t.totalamount ELSE 0 END) AS "Australia Total $ amount"
,SUM(CASE WHEN t.country='Austria' THEN t.totalcount ELSE 0 END) AS "Austria # of Transactions"
,SUM(CASE WHEN t.country='Austria' THEN t.totalamount ELSE 0 END) AS "Austria Total $ amount"
FROM mytransactions t
GROUP BY t.chardate;

Note: The REPLACE function from SELECT @Sql=@Sql+CHAR(13)+ ... REPLACE(c.name,'''','''''')) is used to prevent SQL injections.

I used your own pivot as a nested query and came to this result:

SELECT
  [sub].[chardate],
  SUM(ISNULL([Australia], 0)) AS [Transactions Australia],
  SUM(CASE WHEN [Australia] IS NOT NULL THEN [TotalAmount] ELSE 0 END) AS [Amount Australia],
  SUM(ISNULL([Austria], 0)) AS [Transactions Austria],
  SUM(CASE WHEN [Austria] IS NOT NULL THEN [TotalAmount] ELSE 0 END) AS [Amount Austria]
FROM
(
  select * 
  from  mytransactions
  pivot (sum (totalcount) for country in ([Australia], [Austria])) as pvt
) AS [sub]
GROUP BY
  [sub].[chardate],
  [sub].[numericmonth]
ORDER BY 
  [sub].[numericmonth] ASC

Here is the Fiddle.

The least complicated, most straight-forward way of doing this is by simply wrapping your main query with the pivot in a common table expression, then grouping/aggregating.

WITH PivotCTE AS
(
    select * from  mytransactions
    pivot (sum (totalcount) for country in ([Australia], [Austria])) as pvt
)
SELECT
    numericmonth,
    chardate,
    SUM(totalamount) AS totalamount,
    SUM(ISNULL(Australia, 0)) AS Australia,
    SUM(ISNULL(Austria, 0)) Austria
FROM PivotCTE
GROUP BY numericmonth, chardate

The ISNULL is to stop a NULL value from nullifying the sum (because NULL + any value = NULL)

标签
sql-server
sql-server-2005
pivot
unpivot
易学教程内所有资源均来自网络或用户发布的内容，如有违反法律规定的内容欢迎反馈！
该文章没有解决你所遇到的问题?点击提问,说说你的问题,让更多的人一起探讨吧!