1.grouping sets
记得前几天第一次接触grouping sets时，笔者的感觉是一脸懵逼。
后来一不小心看到msdn上对grouping sets的说明，顿时豁然开朗，其实grouping sets就是由多个group by联合起来，关系如下。
select A , B from table group by grouping sets(A, B) 等价于
select A , null as B from table group by A
union all
select null as A , B from table group by B
为了更好的理解我创建了teacher表，表数据如下，查询结果集中左边的为使用union all的group by字句，右边的为使用grouping sets的结果集。
select null as teacherAddress,MAX(teacherSalary),ascriptionInstitute from teacher group by ascriptionInstitute
union all
select teacherAddress,MAX(teacherSalary),NULL as ascriptionInstitute from teacher group by teacherAddress select teacherAddress,MAX(teacherSalary),ascriptionInstitute from teacher group by GROUPING SETS (ascriptionInstitute,teacherAddress)
上面提到grouping sets是等价于带union all的group by子句，之所以是等价而不是等于，从两者结果集中的对比就可以一目了之，那就是它们的顺序不一样。这说明grouping sets并不只是group by的语法糖，这两者内部的执行过程应该是全然不同的，在百度过程中发现大多数答案都是这句话：“聚合是一次性从数据库中取出所有需要操作的数据，在内存中对数据库进行聚合操作并生成结果。而UNION ALL是多次扫描表，将返回的结果进行UNION操作。性能方面grouping sets能减少IO操作但会增加CPU占用时间”。我不理解的地方是一次性取出数据后，是如何在内存中进行聚合操作的？结果集虽然顺序不一样但数据是相同的，这说明依旧进行了联合操作而这个联合操作并不是多次扫描表，关键内部多次是如何扫描的我很好奇？对于性能我想知道为什么会这样子而不是看到现象。另外在grouping sets中如果将括号中的参数换个位置那么结果也将改变，这说明结果集中的顺序与参数的位置也有关，这让我更加好奇grouping sets的内部执行过程了。
select MAX(teacherSalary),ascriptionInstitute ,teacherAddress from teacher group by GROUPING SETS (ascriptionInstitute,teacherAddress)
select MAX(teacherSalary),ascriptionInstitute ,teacherAddress from teacher group by GROUPING SETS (teacherAddress,ascriptionInstitute)
2.grouping( )
grouping函数用来区分NULL值，这里NULL值有2种情况，一是原本表中的数据就为NULL，二是由rollup、cube、grouping sets生成的NULL值。
当为第一种情况中的空值时，grouping(NULL)返回0；当为第二种情况中的空值时，grouping(NULL)返回1。实例如下，从结果中可以看到第二个结果集中原本为null的数据由于grouping函数为1，故显示ROLLUP-NULL字符串。
select teacherAddress,ascriptionInstitute,COUNT(teacherId ) from teacher group by teacherAddress,ascriptionInstitute
select teacherAddress,ascriptionInstitute,COUNT(teacherId ) from teacher group by rollup(teacherAddress,ascriptionInstitute) select ISNULL(teacherAddress,case when GROUPING(teacherAddress)=1 then 'ROLLUP-NULL' end) as teacherAddress,
ISNULL(ascriptionInstitute,case when GROUPING(ascriptionInstitute)=1 then 'ROLLUP-NULL' end) as ascriptionInstitute,
COUNT(teacherId )
from teacher group by rollup(teacherAddress,ascriptionInstitute)
3.grouping_id( )
grouping_id函数也是计算分组级别的函数，注意如果要使用grouping_id函数那必须得有group by字句，而且group by字句的中的列与grouping_id函数的参数必须相等。比如group by A,B，那么必须使用grouping_id（A,B）。下面用一个等效关系来说明grouping_id()与grouping()的联系，grouping_id(A, B)等效于grouping(A) + grouping(B)，但要注意这里的+号不是算术相加，它表示的是二进制数据组合在一起，比如grouping（A）=1，grouping（B）=1，那么grouping_id(A, B)=11B，也就是十进制数3。原来的表数据执行下面的sql语句结果太多效果不明显，所以我改了下表数据，不过对比两个结果集效果很明显。
select ISNULL(teacherAddress,case when GROUPING(teacherAddress)=1 then 'ROLLUP-NULL' end) as teacherAddress,
ISNULL(ascriptionInstitute,case when GROUPING(ascriptionInstitute)=1 then 'ROLLUP-NULL' end) as ascriptionInstitute,
ISNULL(teacherSex,case when GROUPING(teacherSex)=1 then 'ROLLUP-NULL' end) as teacherSex,
COUNT(teacherId )
from teacher group by rollup(teacherAddress,ascriptionInstitute,teacherSex) select ISNULL(teacherAddress,case when GROUPING(teacherAddress)=1 then 'ROLLUP-NULL' end) as teacherAddress,
ISNULL(ascriptionInstitute,case when GROUPING(ascriptionInstitute)=1 then 'ROLLUP-NULL' end) as ascriptionInstitute,
ISNULL(teacherSex,case when GROUPING(teacherSex)=1 then 'ROLLUP-NULL' end) as teacherSex,
COUNT(teacherId ) as '数量' ,
GROUPING_ID(teacherAddress,ascriptionInstitute,teacherSex)
from teacher group by rollup(teacherAddress,ascriptionInstitute,teacherSex)