PostgreSQL函数如何返回数据集

背景： PostgreSQL里面没有存储过程，只有函数，其他数据库里的这两个对象在PG里都叫函数。 函数由函数头，体和语言所组成，函数头主要是函数的定义，变量的定义等，函数体主要是函数的实现，函数的语言是指该函数实现的方式，目前内置的有c,plpgsql,sql和internal,可以通过pg_language来查看当前DB支持的语言，也可以通过扩展来支持python等
函数返回值一般是类型，比如return int,varchar，返回结果集时就需要setof来表示。
一、数据准备

create table department(id int primary key, name text);
create table employee(id int primary key, name text, salary int, departmentid int references department);

insert into department values (1, 'Management'),(2, 'IT'),(3, 'BOSS');

insert into employee values (1, 'kenyon', 30000, 1);
insert into employee values (2, 'francs', 50000, 1);
insert into employee values (3, 'digoal', 60000, 2);
insert into employee values (4, 'narutu', 120000, 3);

二、例子
1.sql一例

create or replace function f_get_employee()
returns setof employee
as
$$
select * from employee;
$$
language 'sql';

等同的另一个效果(Query)

create or replace function f_get_employee_query()
returns setof employee
as
$$
begin
return query select * from employee;
end;
$$
language plpgsql;

查询图解如下

postgres=# select * from f_get_employee();
 id |  name  | salary | departmentid
----+--------+--------+--------------
  1 | kenyon |  30000 |            1
  2 | francs |  50000 |            1
  3 | digoal |  60000 |            2
  4 | narutu | 120000 |            3
(4 rows)

查询出来的函数还可以像普通的表一样按条件查询 ,但如果查询的方式不一样，则结果也不一样，以下查询方式将会得到类似数组的效果

postgres=# select f_get_employee();
   f_get_employee
---------------------
 (1,kenyon,30000,1)
 (2,francs,50000,1)
 (3,digoal,60000,2)
 (4,narutu,120000,3)
(4 rows)

因为返回的结果集类似一个表的数据集，PostgreSQL还支持对该函数执行结果进行条件判断并过滤

postgres=# select * from f_get_employee() where id >3;
 id |  name  | salary | departmentid
----+--------+--------+--------------
  4 | narutu | 120000 |            3
(1 row)

上面的例子相对简单，如果要返回不是表结构的数据集该怎么办呢？看下面

2.返回指定结果集
a.用新建type来构造返回的结果集

--新建的type在有些图形化工具界面中可能看不到，
要查找的话可以通过select * from pg_class where relkind='c'去查，c表示composite type

create type dept_salary as (departmentid int, totalsalary int);

create or replace function f_dept_salary()
returns setof dept_salary
as
$$
declare
rec dept_salary%rowtype;
begin
for rec in select departmentid, sum(salary) as totalsalary from f_get_employee() group by departmentid loop
  return next rec;
  end loop;
return;
end;
$$
language 'plpgsql';

b.用Out传出的方式

create or replace function f_dept_salary_out(out o_dept text,out o_salary text)
returns setof record as
$$
declare
    v_rec record;
begin
    for v_rec in select departmentid as dept_id, sum(salary) as total_salary from f_get_employee() group by departmentid loop
        o_dept:=v_rec.dept_id;
        o_salary:=v_rec.total_salary;
        return next;
    end loop;
end;
$$
language plpgsql;

执行结果:

postgres=# select * from f_dept_salary();
 departmentid | totalsalary
--------------+-------------
            1 |       80000
            3 |      120000
            2 |       60000
(3 rows)

postgres=# select * from f_dept_salary_out();
 o_dept | o_salary
--------+----------
 1      | 80000
 3      | 120000
 2      | 60000
(3 rows)

c.根据执行函数变量不同返回不同数据集

create or replace function f_get_rows(text) returns setof record as
$$
declare
rec record;
begin
for rec in EXECUTE 'select * from ' || $1 loop
return next rec;
end loop;
return;
end
$$
language 'plpgsql';

执行结果：

postgres=# select * from f_get_rows('department') as dept(deptid int, deptname text);
 deptid |  deptname
--------+------------
      1 | Management
      2 | IT
      3 | BOSS
(3 rows)

postgres=# select * from f_get_rows('employee') as employee(employee_id int, employee_name text,employee_salary int,dept_id int);
 employee_id | employee_name | employee_salary | dept_id
-------------+---------------+-----------------+---------
           1 | kenyon        |           30000 |       1
           2 | francs        |           50000 |       1
           3 | digoal        |           60000 |       2
           4 | narutu        |          120000 |       3
(4 rows)

这样同一个函数就可以返回不同的结果集了，很灵活。