PostgreSQL的组列于JSON包括阵列

Question

TABLE_NAMEË

id | name | cate | link 
---+------+------+--------- 
1 | a | A | link1 
2 | a | B | link2 
3 | a | B | link3 
4 | b | B | link4 
5 | c | A | link5 
6 | d | A | link6 
7 | e | B | link7 

欲retults：

name | A      | B 
------+------+-------------------+------------ 
a | {id: 1, link: 'link1'} | [{id: 2, link: 'link2'}, {id: 3, link: 'link3'}] 
b |       | [{id: 4, link: 'link4'}] 
c | {id: 5, link: 'link5'} | 
d | {id: 6, link: 'link6'} | 
e |       | [{id: 7, link: 'link7'}] 

的美食字段值只有A，B，C，d。但值是B必须数组

我的实验失败

select name, format('{%s}', string_agg(format('"id": "%s", "name": "%s", "link":"%s"', id, name, link), ','))::json as A from elements where cate = 'A' group by name; 
select name, string_to_array(format('[link: "%s", id: "%s", name: "%s"]', link, id, name)) as B from elements where cate = 'B' group by name; 
select name, format('{%s}', string_agg(format('"id": "%s", "name": "%s", "link":"%s"', id, name, link), ','))::json as C from elements where cate = 'C' group by name; 
select name, format('{%s}', string_agg(format('"id": "%s", "name": "%s", "link":"%s"', id, name, link), ','))::json as D from elements where cate = 'D' group by name; 

如果加入其他表P

id | e_id | role_id 
---+------+---------- 
1 | 1 | 100 
2 | 3 | 101 
3 | 4 | 102 
4 | 5 | 103 

结果：

name | checked | A          | B 
-------+---------+---------------------------------------+----------------------------------------------------------------- 
a  | true | {id: 1, link: 'link1', checked: true} | [{id: 2, link: 'link2'}, {id: 3, link: 'link3', checked: true}] 
b  | true |          | [{id: 4, link: 'link4', checked: true}] 
c  | true | {id: 5, link: 'link5', checked: true} | 
d  |   | {id: 6, link: 'link6'}    | 
e  |   |          | [{id: 7, link: 'link7'}] 

Answer 1

首先不使用格式函数来创建json字段，PostgreSQL具有解析和创建json字段所需的全部功能。你没有张贴样本DDL和数据，但您的查询应该是这样的：

select name, 
    json_agg(case cate when 'A' then json_build_object('id',id,'link',link) end) A, 
    json_agg(case cate when 'B' then json_build_object('id',id,'link',link) end) B 
from data group by name