PostgreSQL 实现列转行问题

1 测试表数据

SELECT
  relative_label_content
FROM
  frk_s.label_cor_gene
relative_label_content
------
AA
BB
CC

2 列转行写法

写法1：

string_agg

SELECT
  frwybs,
  string_agg (relative_label_content, ',') as relative_label_content
FROM
  frk_s.label_cor_gene
GROUP BY
  frwybs
relative_label_content
------------
AA,BB,CC

写法2：

array_to_string(ARRAY_AGG (text),',')

SELECT
  frwybs,
  array_to_string(
    ARRAY_AGG (DISTINCT relative_label_content),
    ','
  ) as labels_content
FROM
  frk_s.label_cor_gene
GROUP BY
  frwybs
labels_content
------------
AA,BB,CC

补充：PostgreSQL行列转换(兼容oracle pivot unpivot)

oracle11g开始内置了数据透视表pivot table这一功能，可以用来实现行列转换的功能，但是在数据量较大的时候使用性能就会较差。

pivot语法为：

SELECT ...
FROM  ...
PIVOT [XML]
  (pivot_clause
  pivot_for_clause
  pivot_in_clause )
WHERE ... 

oracle pivot使用例子：

–创建测试表并插入数据

create table usr
(name varchar2(20),
 score int,
 class varchar2(20)
);
insert into usr values('a',20,'math');
insert into usr values('a',22,'phy');
insert into usr values('b',23,'phy');
insert into usr values('b',21,'math');
insert into usr values('c',22,'phy');
insert into usr values('c',24,'math');
insert into usr values('d',25,'math');
insert into usr values('d',23,'phy');

–使用pivot进行行列转换

SQL> select * from usr
 2 pivot(
sum(score)
for class in ('math','phy')
 3  4  5 );
NAME           'math'   'phy'
-------------------- ---------- ----------
d              25     23
a              20     22
b              21     23
c              24     22

我们还可以使用unpivot来实现列转行。

unpivot语法为：

SELECT ...
FROM ...
UNPIVOT [INCLUDE|EXCLUDE NULLS]
  (unpivot_clause
  unpivot_for_clause
  unpivot_in_clause )
WHERE ... 

oracle unpivot使用例子：

–创建测试表并插入数据

CREATE TABLE t1
(
  VendorID int,
  Emp1 int,
  Emp2 int,
  Emp3 int,
  Emp4 int,
  Emp5 int
);
INSERT INTO t1 VALUES (1,4,3,5,4,4);
INSERT INTO t1 VALUES (2,4,1,5,5,5);
INSERT INTO t1 VALUES (3,4,3,5,4,4);
INSERT INTO t1 VALUES (4,4,2,5,5,4);
INSERT INTO t1 VALUES (5,5,1,5,5,5);

–使用unpivot进行列转行

SQL> select * from t1
 2 UNPIVOT(
orders for Employee in(emp1,emp2,emp3,emp4,emp5)
); 3  4
 VENDORID EMPL   ORDERS
---------- ---- ----------
     1 EMP1     4
     1 EMP2     3
     1 EMP3     5
     1 EMP4     4
     1 EMP5     4
     2 EMP1     4
     2 EMP2     1
     2 EMP3     5
     2 EMP4     5
     2 EMP5     5
     3 EMP1     4
 VENDORID EMPL   ORDERS
---------- ---- ----------
     3 EMP2     3
     3 EMP3     5
     3 EMP4     4
     3 EMP5     4
     4 EMP1     4
     4 EMP2     2
     4 EMP3     5
     4 EMP4     5
     4 EMP5     4
     5 EMP1     5
     5 EMP2     1
 VENDORID EMPL   ORDERS
---------- ---- ----------
     5 EMP3     5
     5 EMP4     5
     5 EMP5     5
25 rows selected.

那么在pg中该如何实现oracle的pivot/unpivot的行列转行功能呢？pg中自带的tablefunc插件可以实现，我们可以使用该插件中的crosstab函数接口进行行列转换。

pg行转列例子：

–建表插入测试数据

create table tbl (seller text,se_year int,se_month int,se_amount int);
insert into tbl values ('test1',2020,01,123456);
insert into tbl values ('test1',2020,02,234567);
insert into tbl values ('test1',2020,03,345678);
insert into tbl values ('test1',2020,04,345678);
insert into tbl values ('test1',2020,05,567890);
insert into tbl values ('test2',2020,01,12);
insert into tbl values ('test2',2020,02,23);
insert into tbl values ('test2',2020,03,34);
insert into tbl values ('test2',2020,04,45);
insert into tbl values ('test2',2020,05,56);
insert into tbl values ('test3',2020,03,12);
insert into tbl values ('test3',2020,04,45);
insert into tbl values ('test3',2020,05,56);
insert into tbl values ('test4',2020,02,20);
insert into tbl values ('test4',2020,03,30);
insert into tbl values ('test4',2020,04,40);
insert into tbl values ('test4',2020,05,50);
insert into tbl values ('test1',2019,01,123456);
insert into tbl values ('test1',2019,02,234567);
insert into tbl values ('test1',2019,03,345678);
insert into tbl values ('test1',2019,04,345678);
insert into tbl values ('test1',2019,05,567890);
insert into tbl values ('test1',2019,06,123456);
insert into tbl values ('test1',2019,07,234567);
insert into tbl values ('test1',2019,08,345678);
insert into tbl values ('test1',2019,09,345678);
insert into tbl values ('test1',2019,10,567890);
insert into tbl values ('test1',2019,11,123456);
insert into tbl values ('test1',2019,12,234567);
insert into tbl values ('test2',2019,11,12);
insert into tbl values ('test2',2019,12,23);
insert into tbl select * from tbl;

–行转列

bill=# select
bill-#  js->>'seller' as seller,
bill-#  js->>'se_year' as se_year,
bill-#  jan ,
bill-#  feb ,
bill-#  mar ,
bill-#  apr ,
bill-#  may ,
bill-#  jun ,
bill-#  jul ,
bill-#  aug ,
bill-#  sep ,
bill-#  oct ,
bill-#  nov ,
bill-#  dec
bill-# from crosstab(
bill(#  -- 这个是需要进行行列变换的源SQL ， 数据源。
bill(#  -- 排序字段为group by字段 ，最后一个字段为转换后的内容字段，导数第二个字段为行列变换的字段（内容为枚举，比如月份）
bill(#  -- （必须在下一个参数中提取出对应的所有枚举值）
bill(#  $$select jsonb_build_object('seller', seller, 'se_year', se_year) as js, se_month, sum(se_amount) from tbl group by 1,2 order by 1$$,
bill(#  -- 行列转换的行，有哪些值被提取出来作为列。 这个在这里代表的是月份，也就是se_month的值
bill(#  -- 或(select * from (values('jan'),...('dec')) t(se_month))
bill(#  'select distinct se_month from tbl order by 1'
bill(# )
bill-# as  -- crosstab 输出格式
bill-# ( js jsonb, -- 第一个参数SQL内对应的order by对应的字段（1个或多个）
bill(#  Jan numeric, -- 第一个参数SQL内对应导数第二个字段的枚举值，（行转列）
bill(#  feb numeric, -- ...同上
bill(#  mar numeric,
bill(#  apr numeric,
bill(#  may numeric,
bill(#  jun numeric,
bill(#  jul numeric,
bill(#  aug numeric,
bill(#  sep numeric,
bill(#  oct numeric,
bill(#  nov numeric,
bill(#  dec numeric
bill(# )
bill-# order by 1,2;
 seller | se_year | jan  | feb  | mar  | apr  |  may  | jun  | jul  | aug  | sep  |  oct  | nov  | dec
--------+---------+--------+--------+--------+--------+---------+--------+--------+--------+--------+---------+--------+--------
 test1 | 2019  | 246912 | 469134 | 691356 | 691356 | 1135780 | 246912 | 469134 | 691356 | 691356 | 1135780 | 246912 | 469134
 test1 | 2020  | 246912 | 469134 | 691356 | 691356 | 1135780 |    |    |    |    |     |    |
 test2 | 2019  |    |    |    |    |     |    |    |    |    |     |   24 |   46
 test2 | 2020  |   24 |   46 |   68 |   90 |   112 |    |    |    |    |     |    |
 test3 | 2020  |    |    |   24 |   90 |   112 |    |    |    |    |     |    |
 test4 | 2020  |    |   40 |   60 |   80 |   100 |    |    |    |    |     |    |
(6 rows)

–列转行

bill=# with a as ( -- A对应原始数据（即需要列转行的数据）
bill(# select
bill(#  js->>'seller' as seller,
bill(#  js->>'se_year' as se_year,
bill(#  jan ,
bill(#  feb ,
bill(#  mar ,
bill(#  apr ,
bill(#  may ,
bill(#  jun ,
bill(#  jul ,
bill(#  aug ,
bill(#  sep ,
bill(#  oct ,
bill(#  nov ,
bill(#  dec
bill(# from crosstab(
bill(#  -- 这个是需要进行行列变换的源SQL ， 数据源。
bill(#  -- 排序字段为group by字段 ，最后一个字段为转换后的内容字段，导数第二个字段为行列变换的字段（内容为枚举，比如月份）
bill(#  -- （必须在下一个参数中提取出对应的所有枚举值）
bill(#  $$select jsonb_build_object('seller', seller, 'se_year', se_year) as js, se_month, sum(se_amount) from tbl group by 1,2 order by 1$$,
bill(#  -- 行列转换的行，有哪些值被提取出来作为列。 这个在这里代表的是月份，也就是se_month的值
bill(#  -- 或(select * from (values('jan'),...('dec')) t(se_month))
bill(#  'select distinct se_month from tbl order by 1'
bill(# )
bill(# as  -- crosstab 输出格式
bill(# ( js jsonb, -- 第一个参数SQL内对应的order by对应的字段（1个或多个）
bill(#  Jan numeric, -- 第一个参数SQL内对应导数第二个字段的枚举值，（行转列）
bill(#  feb numeric, -- ...同上
bill(#  mar numeric,
bill(#  apr numeric,
bill(#  may numeric,
bill(#  jun numeric,
bill(#  jul numeric,
bill(#  aug numeric,
bill(#  sep numeric,
bill(#  oct numeric,
bill(#  nov numeric,
bill(#  dec numeric
bill(# )
bill(# order by 1,2
bill(# )
bill-# ,
bill-# -- b ， 用jsonb把多列合并为一列，并使用jsonb_each展开。
bill-# b as (select seller, se_year, jsonb_each(row_to_json(a)::jsonb-'seller'::text-'se_year'::text) as rec from a)
bill-# select seller, se_year, (b.rec).key as month, (b.rec).value as sum from b;
 seller | se_year | month |  sum
--------+---------+-------+---------
 test1 | 2019  | apr  | 691356
 test1 | 2019  | aug  | 691356
 test1 | 2019  | dec  | 469134
 test1 | 2019  | feb  | 469134
 test1 | 2019  | jan  | 246912
 test1 | 2019  | jul  | 469134
 test1 | 2019  | jun  | 246912
 test1 | 2019  | mar  | 691356
 test1 | 2019  | may  | 1135780
 test1 | 2019  | nov  | 246912
 test1 | 2019  | oct  | 1135780
 test1 | 2019  | sep  | 691356
 test1 | 2020  | apr  | 691356
 test1 | 2020  | aug  | null
 test1 | 2020  | dec  | null
 test1 | 2020  | feb  | 469134
 test1 | 2020  | jan  | 246912
 test1 | 2020  | jul  | null
 test1 | 2020  | jun  | null
 test1 | 2020  | mar  | 691356
 test1 | 2020  | may  | 1135780
 test1 | 2020  | nov  | null
 test1 | 2020  | oct  | null
 test1 | 2020  | sep  | null
 test2 | 2019  | apr  | null
 test2 | 2019  | aug  | null
 test2 | 2019  | dec  | 46
 test2 | 2019  | feb  | null
 test2 | 2019  | jan  | null
 test2 | 2019  | jul  | null
 test2 | 2019  | jun  | null
 test2 | 2019  | mar  | null
 test2 | 2019  | may  | null
 test2 | 2019  | nov  | 24
 test2 | 2019  | oct  | null
 test2 | 2019  | sep  | null
 test2 | 2020  | apr  | 90
 test2 | 2020  | aug  | null
 test2 | 2020  | dec  | null
 test2 | 2020  | feb  | 46
 test2 | 2020  | jan  | 24
 test2 | 2020  | jul  | null
 test2 | 2020  | jun  | null
 test2 | 2020  | mar  | 68
 test2 | 2020  | may  | 112
 test2 | 2020  | nov  | null
 test2 | 2020  | oct  | null
 test2 | 2020  | sep  | null
 test3 | 2020  | apr  | 90
 test3 | 2020  | aug  | null
 test3 | 2020  | dec  | null
 test3 | 2020  | feb  | null
 test3 | 2020  | jan  | null
 test3 | 2020  | jul  | null
 test3 | 2020  | jun  | null
 test3 | 2020  | mar  | 24
 test3 | 2020  | may  | 112
 test3 | 2020  | nov  | null
 test3 | 2020  | oct  | null
 test3 | 2020  | sep  | null
 test4 | 2020  | apr  | 80
 test4 | 2020  | aug  | null
 test4 | 2020  | dec  | null
 test4 | 2020  | feb  | 40
 test4 | 2020  | jan  | null
 test4 | 2020  | jul  | null
 test4 | 2020  | jun  | null
 test4 | 2020  | mar  | 60
 test4 | 2020  | may  | 100
 test4 | 2020  | nov  | null
 test4 | 2020  | oct  | null
 test4 | 2020  | sep  | null
(72 rows)

以上为个人经验，希望能给大家一个参考，也希望大家多多支持我们。如有错误或未考虑完全的地方，望不吝赐教。