执行数据透视但不聚合的T-SQL查询

Question

Hi I have the following table of data

Class | Member | Value
----------------------
c1    | m1     | 10
c1    | m2     | 20
c1    | list   | 30
c1    | list   | 40
c1    | list   | 50
c2    | m1     | 60
c2    | m2     | 70
c2    | list   | 80

and would like to pivot the data into this form

Class | Member 1 | Member 2 | List
----------------------------------
c1    | 10       | 20       | 30
c1    | 10       | 20       | 40
c1    | 10       | 20       | 50
c2    | 60       | 70       | 80

a normal pivot using max as the aggregate function would have given me

Class | Member 1 | Member 2 | List
----------------------------------
c1    | 10       | 20       | 50
c2    | 60       | 70       | 80

but I want each of the List values of any Class to be listed out.

Instead of finding alternatives by writing SQL queries full of CASEes eg CASE Member WHEN 'm1' then Value, CASE Member WHEN 'm2' then Value, ... to achieve what I want, I want to know if there is any chance to use pivot with some tweaks to make it work for my task?

The database is SQL 2008 R2

Thank you

Answer 1

No. PIVOT is effectively syntactic sugar for doing aggregates around CASE expressions. If the sugar doesn't work, you need to go back to the longer form.

The syntax for PIVOT is fully described in the FROM clause:

<pivoted_table> ::=
    table_source PIVOT <pivot_clause> [ AS ] table_alias

<pivot_clause> ::=
        ( aggregate_function ( value_column [ [ , ]...n ]) 
        FOR pivot_column 
        IN ( <column_list> ) 
    ) 

And note that aggregate_function must be provided. All aggregate functions in SQL Server operate on any number of input values and produce a single output value. There's no aggregate which can produce multiple output values, as you would require here.

---

This gives the result you've asked for, but does rely on each Class value only having one row for each of m1 and m2 :

declare @t table (Class char(2) not null,Member varchar(4) not null,Value int not null)
insert into @t(Class,Member,Value) values
('c1','m1',10),
('c1','m2',20),
('c1','list',30),
('c1','list',40),
('c1','list',50),
('c2','m1',60),
('c2','m2',70),
('c2','list',80)

select l.Class,m1.Value as m1,m2.Value as m2,l.Value as list
from
    @t l
        inner join
    @t m1
        on
            l.Class = m1.Class and
            m1.Member = 'm1'
        inner join
    @t m2
        on
            l.Class = m2.Class and
            m2.Member = 'm2'where
l.Member='list'

Result:

Class m1          m2          list
----- ----------- ----------- -----------
c1    10          20          30
c1    10          20          40
c1    10          20          50
c2    60          70          80

If there are multiple rows for m1 and m2 , and you just want, say, the MAX values for them, then you'd make m1 and m2 in my above query subqueries:

...
    inner join
(select Class,MAX(Value) from @t where Member='m1' group by Class) m1
    on
        l.Class = m1.Class
...

Answer 2

SELECT  a.*, 
        b.Value
FROM    
        (
            SELECT  Class,
                    MAX(CASE WHEN Member = 'm1' THEN Value ELSE NULL END) [Member 1],
                    MAX(CASE WHEN Member = 'm2' THEN Value ELSE NULL END) [Member 2]
            FROM  tableName
            GROUP BY Class
        ) a 
        INNER JOIN
        (
            SELECT  Class, Value
            FROM    tableName 
            WHERE   Member = 'List'
        ) b ON a.Class = b.Class

• SQLFiddle Demo

RESULT

╔═══════╦══════════╦══════════╦═══════╗
║ CLASS ║ MEMBER 1 ║ MEMBER 2 ║ VALUE ║
╠═══════╬══════════╬══════════╬═══════╣
║ c1    ║       10 ║       20 ║    30 ║
║ c1    ║       10 ║       20 ║    40 ║
║ c1    ║       10 ║       20 ║    50 ║
║ c2    ║       60 ║       70 ║    80 ║
╚═══════╩══════════╩══════════╩═══════╝