with rollup 、with cube、grouping CUBE 和 ROLLUP 之间的区别在于: CUBE 生成的结果集显示了所选列中值的所有组合的聚合。 ROLLUP 生成的结果集显示了所选列中值的某一层次结构的聚合。 grouping: 当用 CUBE 或 ROLLUP 运算符添加行时,附加的列输出值为1,当所添加的行不是由 CUBE 或 ROLLUP 产生时,附加列值为0。
--------------------------------------------------------------------------------
首先创建测试表、添加数据。
table #t(a int,b int,c int,d int,e int)
insert into #t values(1,2,3,4,5)
insert into #t values(1,2,3,4,6)
insert into #t values(1,2,3,4,7)
insert into #t values(1,2,3,4,8)
insert into #t values(1,3,3,4,5)
insert into #t values(1,3,3,4,6)
insert into #t values(1,3,3,4,8)
insert into #t values(1,3,3,4,7)
insert into #t values(2,2,2,4,5)
insert into #t values(2,2,3,4,6)
insert into #t values(2,2,4,4,7)
insert into #t values(2,2,5,4,8)
insert into #t values(1,2,3,4,5)
insert into #t values(1,2,3,4,6)
insert into #t values(1,2,3,4,7)
insert into #t values(1,2,3,4,8)
insert into #t values(1,3,3,4,5)
insert into #t values(1,3,3,4,6)
insert into #t values(1,3,3,4,8)
insert into #t values(1,3,3,4,7)
insert into #t values(2,2,2,4,5)
insert into #t values(2,2,3,4,6)
insert into #t values(2,2,4,4,7)
insert into #t values(2,2,5,4,8)
insert into #t values(2,3,3,4,6)
insert into #t values(2,3,3,4,8)
insert into #t values(2,3,3,4,7)
情况一:只有一个分类汇总列时,只需要一个合计。只需要增加with rollup即可。
case
when grouping(a)=1 then '合计' else cast(a as varchar) end a,
sum(b),sum(c),sum(d),sum(e) from #t group by a with rollup
sum(b),sum(c),sum(d),sum(e) from #t group by a with rollup
情况二:有多个分类汇总列,只需要一个合计.增加rollup之后,需要增加判断。
case
when grouping(a)=1 then '合计' else cast(a as varchar) end a,
b,
sum(c),sum(d),sum(e) from #t
group by a,b with rollup
having grouping(b)=0 or grouping(a)=1
b,
sum(c),sum(d),sum(e) from #t
group by a,b with rollup
having grouping(b)=0 or grouping(a)=1
case
when grouping(a)=1 then '合计' else cast(a as varchar) end a,
b,
c,
sum(d),sum(e) from #t
group by a,b,c with rollup
having grouping(c)=0 or grouping(a)=1
b,
c,
sum(d),sum(e) from #t
group by a,b,c with rollup
having grouping(c)=0 or grouping(a)=1
情况三:有多个分类汇总列,需要全部的小计和合计。
case
when grouping(a)=1 then '合计' else cast(a as varchar) end a,
case when grouping(b)=1 and grouping(a)=0 then '小计' else cast(b as varchar) end b,
case when grouping(c)=1 and grouping(b)=0 then '小计' else cast(c as varchar) end c,
sum(d),sum(e) from #t
group by a,b,c with rollup
case when grouping(b)=1 and grouping(a)=0 then '小计' else cast(b as varchar) end b,
case when grouping(c)=1 and grouping(b)=0 then '小计' else cast(c as varchar) end c,
sum(d),sum(e) from #t
group by a,b,c with rollup
另外一种显示小计的方式
case
when grouping(a)=1 then '合计'
when grouping(b)=1 then cast(a as varchar)+'小计'
else cast(a as varchar) end a,
case when grouping(b)=0 and grouping(c)=1
then cast(b as varchar)+'小计' else cast(b as varchar) end b,
case when grouping(c)=1 and grouping(b)=0
then '' else cast(c as varchar) end c,
sum(d),sum(e) from #t
group by a,b,c with rollup
when grouping(b)=1 then cast(a as varchar)+'小计'
else cast(a as varchar) end a,
case when grouping(b)=0 and grouping(c)=1
then cast(b as varchar)+'小计' else cast(b as varchar) end b,
case when grouping(c)=1 and grouping(b)=0
then '' else cast(c as varchar) end c,
sum(d),sum(e) from #t
group by a,b,c with rollup
情况四:有多个分类汇总列,需要部分的小计和合计
case
when grouping(a)=1 then '合计' else cast(a as varchar) end a,
b,
case when grouping(c)=1 and grouping(b)=0 then '小计' else cast(c as varchar) end c,
sum(d),sum(e) from #t
group by a,b,c with rollup
having grouping(a)=1 or grouping(b)=0
b,
case when grouping(c)=1 and grouping(b)=0 then '小计' else cast(c as varchar) end c,
sum(d),sum(e) from #t
group by a,b,c with rollup
having grouping(a)=1 or grouping(b)=0
case
when grouping(a)=1 then '合计' else cast(a as varchar) end a,
case when grouping(b)=1 and grouping(a)=0 then '小计' else cast(b as varchar) end b,
c,
sum(d),sum(e) from #t
group by a,b,c with rollup
having grouping(a)=1 or grouping(b)=1 or grouping(c)=0
case when grouping(b)=1 and grouping(a)=0 then '小计' else cast(b as varchar) end b,
c,
sum(d),sum(e) from #t
group by a,b,c with rollup
having grouping(a)=1 or grouping(b)=1 or grouping(c)=0
引用:http://www.cnblogs.com/yiway/archive/2007/11/12/957281.html