系列文章目录

文章目录

系列文章目录
前言
一、用"行转列"来得到隐含信息
二、用隐藏数据进行行转列
总结

前言

本篇文章讲解的主要内容是：需求一：T表内始终只有近两年的数据，要求时刻返回两列并且分别显示其中一年的数据。需求二：以Emp表中为例，要求每个job显示为一行，同类job的姓名分列显示，不够的显示为空
【SQL开发实战技巧】这一系列博主当作复习旧知识来进行写作，毕竟SQL开发在数据分析场景非常重要且基础，面试也会经常问SQL开发和调优经验，相信当我写完这一系列文章，也能再有所收获，未来面对SQL面试也能游刃有余~。

一、用"行转列"来得到隐含信息

示例数据如下：

with t as (
select 'A' as shop,'2013' as nyear, 123 as amount from dual 
union all
select 'A' as shop,'2012' as nyear, 200 as amount from dual 
)
select * from t;
SHOP NYEAR     AMOUNT
---- ----- ----------
A    2013         123
A    2012         200

T表内始终只有两年的数据，要求返回两列分别显示其中一年的数据，原始写法如下：

with t as (
select 'A' as shop,'2013' as nyear, 123 as amount from dual 
union all
select 'A' as shop,'2012' as nyear, 200 as amount from dual 
)
select 
max(decode(nyear,'2012',amount)) as s2012,
max(decode(nyear,'2013',amount)) as s2013
from t;
     S2012      S2013
---------- ----------
       200        123

现在的语句中，2012年与2013年是固定的，而数据库中每一年的数据都在变（上一年与本年），现要求不再固定为2012年与2013年。怎么办？
其实这就是把上一年与本年数据各写为两列，本例的max用错了，应该为sum。
首先可以用分析函数取出上一年的年份(min)和本年的年份(max),分别放在两列
里。

with t as
 (select 'A' as shop, '2013' as nyear, 123 as amount
    from dual
  union all
  select 'A' as shop, '2012' as nyear, 200 as amount
    from dual)
select shop,
       nyear,
       max(nyear) over() as max_year,
       min(nyear) over() as min_year,
       sum(amount) as amount
  from t
 group by shop, nyear
SHOP NYEAR MAX_YEAR MIN_YEAR     AMOUNT
---- ----- -------- -------- ----------
A    2012  2013     2012            200
A    2013  2013     2012            123

SQL>

可以看到，我们把年份信息取出来，分别放在了两列中，这样就有了一个参照：

with t as
 (select 'A' as shop, '2013' as nyear, 123 as amount
    from dual
  union all
  select 'A' as shop, '2012' as nyear, 200 as amount
    from dual)
select shop,
       max(decode(nyear, max_year, amount)) as 今年,
       max(decode(nyear, min_year, amount)) as 去年
  from (select shop,
               nyear,
               max(nyear) over() as max_year,
               min(nyear) over() as min_year,
               sum(amount) as amount
          from t
         group by shop, nyear
        )
 group by shop;
 
SHOP         今年         去年
---- ---------- ----------
A           123        200

二、用隐藏数据进行行转列

有网友提出如下需求，用现有表scott.emp模拟，表中每个job显示为一行，同类job的姓名分列显示，不够的显示为空，结果如下：

JOB       S1         S2         S3         S4
--------- ---------- ---------- ---------- ----------
ANALYST   FORD       SCOTT                 
CLERK     ADAMS      JAMES      MILLER     SMITH
MANAGER   BLAKE      CLARK      JONES      
PRESIDENT KING                             
SALESMAN  ALLEN      MARTIN     TURNER     WARD

乍一看，这个语句不好写，因为没有可供case when选用的条件。其实这里的条件是隐藏的，就是各员工姓名的顺序。

with t as (
select ename, job, row_number() over(partition by job order by ename) as rn
  from emp
 order by job)
 select * from t;
ENAME      JOB               RN
---------- --------- ----------
FORD       ANALYST            1
SCOTT      ANALYST            2
ADAMS      CLERK              1
JAMES      CLERK              2
MILLER     CLERK              3
SMITH      CLERK              4
BLAKE      MANAGER            1
CLARK      MANAGER            2
JONES      MANAGER            3
KING       PRESIDENT          1
ALLEN      SALESMAN           1
MARTIN     SALESMAN           2
TURNER     SALESMAN           3
WARD       SALESMAN           4

14 rows selected

有这个序号后，我们可以把sn=l的放第一列，sn=2的放第二列…

with t as (
select ename, job, row_number() over(partition by job order by ename) as rn
  from emp
 order by job)
 select job, 
 max(case when rn=1 then ename end) as s1,
 max(case when rn=2 then ename end) as s2,
 max(case when rn=3 then ename end) as s3,
 max(case when rn=4 then ename end) as s4
 from t
 group by job;
 JOB       S1         S2         S3         S4
--------- ---------- ---------- ---------- ----------
ANALYST   FORD       SCOTT                 
CLERK     ADAMS      JAMES      MILLER     SMITH
MANAGER   BLAKE      CLARK      JONES      
PRESIDENT KING                             
SALESMAN  ALLEN      MARTIN     TURNER     WARD

如果用行转列函数，则为

 select job, s1, s2, s3, s4
   from (select ename,
                job,
                row_number() over(partition by job order by ename) as rn
           from emp
          order by job)
 pivot(max(ename)
    for rn in(1 as s1, 2 as s2, 3 as s3, 4 as s4));
JOB       S1         S2         S3         S4
--------- ---------- ---------- ---------- ----------
ANALYST   FORD       SCOTT                 
CLERK     ADAMS      JAMES      MILLER     SMITH
MANAGER   BLAKE      CLARK      JONES      
PRESIDENT KING                             
SALESMAN  ALLEN      MARTIN     TURNER     WARD

如果不知道有多少列怎么办？那就需要先查询，然后用循环语句自动拼装成需要的SQL然后查询即可：

declare
  v_max number;
  v_sql varchar2(4000);
begin
  select max(count(*)) into v_max from emp group by job;
  v_sql := 'select job, ' || chr(10);
  for i in 1 .. v_max loop
    if (i <> v_max) then
      v_sql := v_sql || ' max(case when rn=' || i ||
               ' then ename end) as s' || i || ',' || chr(10);
    else
      v_sql := v_sql || ' max(case when rn=' || i ||
               ' then ename end) as s' || i || chr(10);
    end if;
  end loop;
  v_sql := v_sql || ' from (
select ename, job, row_number() over(partition by job order by ename) as rn
  from emp
 order by job)  group by job';
  dbms_output.put_line(v_sql);
end;
 
 
select job, 
 max(case when rn=1 then ename end) as s1,
 max(case when rn=2 then ename end) as s2,
 max(case when rn=3 then ename end) as s3,
 max(case when rn=4 then ename end) as s4
 from (
select ename, job, row_number() over(partition by job order by ename) as rn
  from emp
 order by job)  group by job

总结

本篇文章讲解的主要内容是：需求一：T表内始终只有近两年的数据，要求时刻返回两列并且分别显示其中一年的数据。需求二：以Emp表中为例，要求每个job显示为一行，同类job的姓名分列显示，不够的显示为空

【SQL开发实战技巧】系列（三十七）：数仓报表场景☞从表内始终只有近两年的数据，要求用两列分别显示其中一年的数据聊行转列隐含信息的重要性

系列文章目录

文章目录

前言

一、用"行转列"来得到隐含信息

二、用隐藏数据进行行转列

总结

猜你喜欢