MYSQL 分组排名

今天遇到一个MYSQL排序的问题,要求按某列进行分组,组内进行排序.

百度一下发现MYSQL不支持row_number(),rank()等函数.

采用的办法如下,我们首先创建一个测试表:

--创建表

create table Rank_test(ID int,SCORE int,grp int)



--插入数据

insert into Rank_test values(1 ,     28,1);

insert into Rank_test values(2 ,     33,1);

insert into Rank_test values(3 ,     33,1);

insert into Rank_test values(4 ,     89,1);

insert into Rank_test values(5 ,     99,1);

insert into Rank_test values(6 ,     68,1);

insert into Rank_test values(7 ,     68,1);

insert into Rank_test values(8 ,     78,1);

insert into Rank_test values(9 ,     88,1);

insert into Rank_test values(10,    90 ,1);

insert into Rank_test values(11,     28,2);           

insert into Rank_test values(12,     33,2);           

insert into Rank_test values(13,     33,2);    

insert into Rank_test values(14,     89,2);           

insert into Rank_test values(15,     99,2);           

insert into Rank_test values(16,     68,2);           

insert into Rank_test values(17,     68,2);           

insert into Rank_test values(18,     78,2);           

insert into Rank_test values(19,     88,2);           

insert into Rank_test values(20,    90 ,2);         

分组排序的SQL代码:

SELECT id,  

       score,  

       rank  

  FROM (SELECT tmp.id,  

               tmp.score,                              

               @rank := (case when @Grp = grp then @rank + 1 else 1 end) AS rank  ,

               @Grp:=grp as grp2

          FROM (SELECT id,  

                       score,

                       grp

                  FROM rank_test) tmp,  

               (SELECT @rank   := 0,@grp:=0) a

            order by grp,score desc) RESULT

;  

结果如下:

 

测试结果OK,但重点是套用到我们的SQL中后,排序结果完全不是我们预期中的,而且没有找到规律.

经过几次尝试,最后重新改写了SQL代码,终于成功了.

原因是套用下面这段代码时,我们是用了几张表进行关联(inner join ),最后将inner join 的表全部改为where 语句中的子句,结果满足我们的预期.

(SELECT id,  

                       score,

                       grp

                  FROM rank_test) tmp,

所以建议在这里的代码尽量使用单表,不要多张表关联

附上修改前后的代码:(红色部分是修改前,绿色部分是修改后)

select * from (

select evtdate

        ,stkabb,chng,PCTCSHG,cshg,SHHNAME

        ,@rank := (case when @SHHNAME=SHHNAME then @rank + 1 

                                        else 1 end) AS rank  

        ,@SHHNAME:=SHHNAME

from 

(

select tt.SHHNAME -- 股东名称

      ,s.evtdate  -- 变动日期

      ,r.stkabb   -- 变动股票名称

      ,ifnull(s.RLDSHG,s.cshg-s.SHGBCH)/10000 as chng -- 变动股数（万股）

      ,s.cshg/10000 as cshg-- 变动后持股（万股）

      ,s.PCTCSHG -- 变动后持股比例%

  from pgenius.hk_stkcode r inner join pgenius.HK_SAKCHMJSHH s on r.comunic = s.comunic inner join (select distinct b.SHHNAME from pgenius.hk_stkcode a inner join pgenius.HK_SAKCHMJSHH b on a.comunic = b.comunic and b.evtdate = (select max(evtdate) from pgenius.HK_SAKCHMJSHH where comunic = b.comunic and SHHNAME = b.SHHNAME and NTUREFCINTRTS = b.NTUREFCINTRTS) where a.lssturefc = 1

          and b.cshg > 0

          and a.stkcode = '00001' ) tt on s.SHHNAME = tt.SHHNAME

) t1,

(SELECT @rank := 0,@SHHNAME:='') t2

order by SHHNAME,evtdate desc,stkabb ) t

 

select t.*

        , o.stkabb 

from 

(

    select t1.*

            ,@rank := (case when @SHHNAME=SHHNAME then @rank + 1 

                                            else 1 end) AS rank  

            ,@SHHNAME:=SHHNAME

    from 

    (

        select s.evtdate ,ifnull(s.RLDSHG,s.cshg-s.SHGBCH)/10000 as chng -- 变动股数（万股）

                    ,s.cshg/10000 as cshg-- 变动后持股（万股）

                    ,s.PCTCSHG -- 变动后持股比例%

 ,SHHNAME ,s.COMUNIC from pgenius.HK_SAKCHMJSHH s where SHHNAME in ( select distinct shhname from pgenius.HK_SAKCHMJSHH a where comunic=(select COMUNIC from pgenius.hk_stkcode where stkcode='00001') and ISVALID=1 

            and not exists(select 1 from pgenius.HK_SAKCHMJSHH where COMUNIC=a.COMUNIC and SHHNAME=a.SHHNAME and evtdate>a.evtdate)

            and cshg > 0)

    ) t1,

    (SELECT @rank := 0,@SHHNAME:='') t2

    order by SHHNAME,evtdate desc 

) t

    inner join pgenius.hk_stkcode o on t.COMUNIC= o.COMUNIC

where t.rank<=10

order by shhname,rank