sql server多重行列转置的优化

将表1转化成表2：

表1

表2

得到表2的结果，需要经过多次pivot转换，再经union连接到一起，代码如下：

 select id, type,sum([]) [],sum([]) [],sum([]) [],sum([]) [] from
 (
     select 'a' as type, * from   Table_1
     pivot(sum(a)  for p in([],[],[],[])) as a
 union all
     select 'b' as type,* from   Table_1
     pivot(sum(b)  for p in([],[],[],[])) as b
 union all
     select 'c'  as type,* from   Table_1
     pivot(sum(c)  for p in([],[],[],[])) as c
 union all
     select 'd'  as type,* from   Table_1
     pivot(sum(d)  for p in([],[],[],[])) as d
 ) t1
 group by id,type
 order by id,type

此时代码看起来比较多，如果需要n多次pivot转换，代码过于繁多。

此时，可通过定义一个变量，以拼字符串的形式，来代替繁多的代码：

 declare @str varchar(8000)
 set @str = ''
 select  @str = @str + ' SELECT '''+ NAME + ''' AS TYPE,* FROM Table_1 pivot(SUM('+ NAME +')
         for p in ([1],[2],[3],[4])) as '+ NAME +' union ALL '
         from syscolumns
         where object_id('Table_1') = id AND NAME <> 'P' AND NAME <> 'ID'
 select @str = left(@str,len(@str)-len('union ALL'))
 select @str ='select id, type,sum([1]) [1],sum([2]) [2],sum([3]) [3],sum([4]) [4] from ('+ @str +') t1 group by id,type order by id,type'
 exec (@str)

两种方法得出的结果是一样的，只是后者代码更为简洁。