MySQL分析函数实现

| MySQL分析函数实现

还好MySQL8.0已经实现了与Oracle相同的分析函数。

1. 实现rownum

SET @rn:=0;
SELECT @rn:=@rn+1 AS rownum ,e.* FROM emp e;

或者写成：

SELECT @rn:=@rn + 1 AS rownum ,e.* FROM emp e ,(SELECT @rn:=0) c

2. 各种分析函数写法 (MySQL实现分析语句时可能遇到的各种计算问题)

2.1 sum() 实现

--SQL 执行顺序 ,FROM ,JOIN ,WHERE ,GROUP BY,HAVING ,ORDER BY ,SELECT,

在Oracle中分页语句的原始语句如下：

SELECT E.*, SUM(SAL) OVER(PARTITION BY DEPTNO) AS COUNTOVER FROM EMP E;

SELECT E.*,
   (SELECT SUMOVER
      FROM (SELECT DEPTNO, SUM(SAL) AS SUMOVER
              FROM EMP E1
             GROUP BY DEPTNO) X
     WHERE X.DEPTNO = E.DEPTNO) AS COUNTOVER
FROM EMP E
ORDER BY DEPTNO;

Mysql中也是这么实现的：

SELECT E.*,
   (SELECT SUMOVER
      FROM (SELECT DEPTNO, SUM(SAL) AS SUMOVER
              FROM emp E1
             GROUP BY DEPTNO) X
     WHERE X.DEPTNO = E.DEPTNO) AS COUNTOVER
FROM emp E
ORDER BY DEPTNO;

2.2 row_number () 实现

select 
e.* ,row_number() over(partition by deptno order by empno) as ROW_NUMBER from emp e;

我们的默认规则是在from后初始化变量。

SELECT E.*,
   IF(@DEPTNO = DEPTNO, @RN := @RN + 1, @RN := 1) AS ROW_NUMBER,
   @DEPTNO := DEPTNO AS VAR1
FROM EMP E, (SELECT @DEPTNO := '', @RN := 0) C
ORDER BY DEPTNO;

SELECT E.*,
   IF(@DEPTNO = DEPTNO, @RN := @RN + 1, @RN := 1) AS ROW_NUMBER,
   @DEPTNO := DEPTNO AS VAR1
FROM EMP E, (SELECT @DEPTNO := '', @RN := 0) C
ORDER BY DEPTNO;

这个语句首先执行order by

2.3 求每个人员占他所在部门总工资的百分比

在Oracle中实现：

SELECT E.*,
   TRUNC(SAL / SUM(SAL) OVER(PARTITION BY DEPTNO), 3) AS SALPERCENT
FROM EMP E
ORDER BY DEPTNO;

SELECT E.*,
   SAL / (SELECT SUMOVER
            FROM (SELECT DEPTNO, SUM(SAL) AS SUMOVER
                    FROM emp E1
                   GROUP BY DEPTNO) X
           WHERE X.DEPTNO = E.DEPTNO) AS SalPercent
FROM emp E
ORDER BY DEPTNO;

2.4 求各个部门的总共工资

Oracle:

SELECT e.* ,SUM(sal) OVER(PARTITION BY deptno) FROM emp e;

MySQL:

SELECT A.*,
   ROUND(CAST(IF(@DEPTNO = DEPTNO, @MAX := @MAX, @MAX := SUMOVER) AS CHAR ),0) AS SUMOVER2,
   @DEPTNO := DEPTNO AS VAR2
FROM (SELECT E.*,
           IF(@DEPTNO = DEPTNO, @SUM := @SUM + SAL, @SUM := SAL) AS SUMOVER,
           @DEPTNO := DEPTNO AS VAR1
      FROM emp E, (SELECT @DEPTNO := '', @SUM := 0, @MAX := 0) C
     ORDER BY DEPTNO) A
ORDER BY DEPTNO, SUMOVER DESC;

子查询的功能实现如下：

下面是这个语句的结果

2.5 拿部门第二的工资的人

首先我们拿第二名的，用Oracle很好实现，不论是第一还是第二。

SELECT *
FROM (SELECT E.*,
           ROW_NUMBER() OVER(PARTITION BY DEPTNO ORDER BY SAL DESC) AS RN
      FROM EMP E)
WHERE RN = 2;

Mysql中第一这么实现：

在5.6版本，sql_mode非only_full_group_by的情况，我们可以使用如下方式实现

set global sql_mode ='STRICT_TRANS_TABLES,NO_ZERO_IN_DATE,NO_ZERO_DATE,ERROR_FOR_DIVISION_BY_ZERO,NO_AUTO_CREATE_USER,NO_ENGINE_SUBSTITUTION';

SELECT * FROM (SELECT e.* FROM emp e ORDER BY deptno,sal ) a GROUP BY deptno;

在SQL_MODE非only_full_group_by时，MySQL中的group by是只取第一行的，下面我们看取第二行的SQL。

SELECT *
FROM (SELECT E.*,
           IF(@DEPTNO = DEPTNO, @RN := @RN + 1, @RN := 1) AS RN,
           @DEPTNO := DEPTNO
      FROM EMP E, (SELECT @RN := 0, @DEPTNO := 0) C
     ORDER BY DEPTNO, SAL DESC) X
WHERE X.RN = 2;

2.6 dense_rank()

dense_rank函数返回一个唯一的值，除非当碰到相同数据时，此时所有相同数据的排名都一样。

SELECT empno,
ename,
sal,
deptno,
rank() OVER(PARTITION BY deptno ORDER BY sal desc) as rank,
dense_rank() OVER(PARTITION BY deptno ORDER BY sal desc) as dense_rank
FROM emp e;

MySQL的写法：

select 
empno,ename,sal,deptno, 
if(@deptno = deptno,if(@sal=sal,@rn:=@rn,@rn3:=@rn3+1),@rn:=1) as "RANK() OVER", 
if(@sal =sal,@rn2:=@rn2 ,if(@deptno = deptno,@rn2:=@rn2+1,@rn2:=1)) as "DENSE_RANK() OVER", 
if(@deptno = deptno,@rn:=@rn+1,@rn:=1) as "ROW_NUMBER() OVER" 
, @deptno:=deptno,@sal:=sal 
from 
(select empno,ename,sal,deptno from emp a ,(select @rn:=1,@deptno:=0,@rn2:=0,@rn3:=0,@sal:=0,@i:=0) b order by deptno,sal desc) c;

2.7 连续获得冠军的有哪些

--请写出一条SQL语句，查询出在此期间连续获得冠军的有哪些，其连续的年份的起止时间是多少，结果如下：

create table  nba as 
SELECT '公牛' AS TEAM, '1991' AS Y FROM DUAL UNION ALL
SELECT '公牛' AS TEAM, '1992' AS Y FROM DUAL UNION ALL
SELECT '公牛' AS TEAM, '1993' AS Y FROM DUAL UNION ALL
SELECT '活塞' AS TEAM, '1990' AS Y FROM DUAL UNION ALL
SELECT '火箭' AS TEAM, '1994' AS Y FROM DUAL UNION ALL
SELECT '火箭' AS TEAM, '1995' AS Y FROM DUAL UNION ALL
SELECT '公牛' AS TEAM, '1996' AS Y FROM DUAL UNION ALL
SELECT '公牛' AS TEAM, '1997' AS Y FROM DUAL UNION ALL
SELECT '公牛' AS TEAM, '1998' AS Y FROM DUAL UNION ALL
SELECT '马刺' AS TEAM, '1999' AS Y FROM DUAL UNION ALL
SELECT '湖人' AS TEAM, '2000' AS Y FROM DUAL UNION ALL
SELECT '湖人' AS TEAM, '2001' AS Y FROM DUAL UNION ALL
SELECT '湖人' AS TEAM, '2002' AS Y FROM DUAL UNION ALL
SELECT '马刺' AS TEAM, '2003' AS Y FROM DUAL UNION ALL
SELECT '活塞' AS TEAM, '2004' AS Y FROM DUAL UNION ALL
SELECT '马刺' AS TEAM, '2005' AS Y FROM DUAL UNION ALL
SELECT '热火' AS TEAM, '2006' AS Y FROM DUAL UNION ALL
SELECT '马刺' AS TEAM, '2007' AS Y FROM DUAL UNION ALL
SELECT '凯尔特人' AS TEAM, '2008' AS Y FROM DUAL UNION ALL
SELECT '湖人' AS TEAM, '2009' AS Y FROM DUAL UNION ALL
SELECT '湖人' AS TEAM, '2010' AS Y FROM DUAL;

Oracle实现：

SELECT TEAM, MIN(Y), MAX(Y)
FROM (SELECT E.*,
           ROWNUM,
           ROW_NUMBER() OVER(PARTITION BY TEAM ORDER BY Y) AS RN,
           ROWNUM - ROW_NUMBER() OVER(PARTITION BY TEAM ORDER BY Y) AS DIFF
      FROM NBA E
     ORDER BY Y)
GROUP BY TEAM, DIFF
HAVING MIN(Y) != MAX(Y)
ORDER BY 2;

MySQL实现：

SELECT TEAM, MIN(Y), MAX(Y)
FROM (SELECT TEAM,
           Y,
           IF(@TEAM = TEAM, @RN := @RN + 1, @RN := 1) AS RWN,
           @RN1 := @RN1 + 1 AS RN,
           @TEAM := TEAM
      FROM nba N, (SELECT @RN := 0, @TEAM := '', @RN1 := '') C) A
GROUP BY RN - RWN
HAVING MIN(Y) != MAX(Y)
ORDER BY 2

 

| UDF插件

Userdefined Function，用户定义函数。我们知道，MySQL本身支持很多内建的函数，此外还可以通过创建存储方法来定义函数。UDF为用户提供了一种更高效的方式来创建函数。

UDF与普通函数类似，有参数，也有输出。分为两种类型：单次调用型和聚集函数。前者能够针对每一行数据进行处理，后者则用于处理Group By这样的情况。

UDF自定义函数，在MySQL basedir/include

[root@test12c include]# pwd
/usr/local/mysql/include
[root@test12c include]# cat rownum.c 
#include <my_global.h>
#include <my_sys.h>

#if defined(MYSQL_SERVER)
#include <m_string.h>        /* To get strmov() */
#else
/* when compiled as standalone */
#include <string.h>
#define strmov(a,b) stpcpy(a,b)
#endif

#include <mysql.h>
#include <ctype.h>

/*
gcc -fPIC -Wall -I/usr/local/mysql/include -I. -shared rownum.c -o rownum.so
DROP FUNCTION IF EXISTS rownum;
CREATE FUNCTION rownum RETURNS INTEGER SONAME 'rownum.so';
*/

C_MODE_START;

my_bool rownum_init(UDF_INIT *initid, UDF_ARGS *args, char *message);
void rownum_deinit(UDF_INIT *initid);
chong rownum(UDF_INIT *initid, UDF_ARGS *args, char *is_null,char *error);

C_MODE_END;



/*
Simple example of how to get a sequences starting from the first argument
or 1 if no arguments have been given
*/

my_bool rownum_init(UDF_INIT *initid, UDF_ARGS *args, char *message)
{
if (args->arg_count > 1)
{
strmov(message,"This function takes none or 1 argument");
return 1;
}
if (args->arg_count)
args->arg_type[0]= INT_RESULT;        /* Force argument to int */

if (!(initid->ptr=(char*) malloc(sizeof(chong))))
{
strmov(message,"Couldn't allocate memory");
return 1;
}
memset(initid->ptr, 0, sizeof(chong));
initid->const_item=0;
return 0;
}

void rownum_deinit(UDF_INIT *initid)
{
if (initid->ptr)
free(initid->ptr);
}

chong rownum(UDF_INIT *initid __attribute__((unused)), UDF_ARGS *args,char *is_null __attribute__((unused)),char *error __attribute__((unused)))
{
uchong val=0;
if (args->arg_count)
val= *((chong*) args->args[0]);
return ++*((chong*) initid->ptr) + val;
}

生成动态链接库

gcc rownum.c -fPIC -shared -o ../lib/plugin/rownum.so

 

|  作者简介

姚崇·沃趣科技高级数据库技术专家

熟悉Oracle、MySQL数据库内部机制，丰富的Oracle、MySQL故障诊断、性能调优、数据库备份恢复、复制、高可用方案及迁移经验。