PostgreSQL 9.1:如何连接数组中没有重复的行,加入另一个表

Question

我正在使用PostgreSQL 9.1并需要帮助将多行连接在一起.我需要在2个表中这样做.当我使用两次array_agg()函数时,我得到结果中的重复值.

表:

CREATE TABLE rnp (id int, grp_id int, cabinets varchar(15) );

INSERT INTO rnp VALUES
 (1,'11','cabs1')
,(2,'11','cabs2')
,(3,'11','cabs3')
,(4,'11','cabs4')
,(5,'22','c1')
,(6,'22','c2');

CREATE TABLE ips (id int, grp_id int, address varchar(15));

INSERT INTO ips VALUES
 (1,'11','NY')
,(2,'11','CA')
,(3,'22','DC')
,(4,'22','LA');

SQL:

SELECT DISTINCT

  rnp.grp_id,
  array_to_string(array_agg(rnp.cabinets)OVER (PARTITION BY rnp.grp_id), ',') AS cabinets,
  array_to_string(array_agg(ips.address) OVER (PARTITION BY ips.grp_id), ',') AS addresses


FROM rnp JOIN ips ON rnp.grp_id=ips.grp_id

结果:

GRP_ID  CABINETS                                             ADDRESSES
11  cabs1,cabs1,cabs2,cabs2,cabs3,cabs3,cabs4,cabs4     NY,CA,NY,CA,NY,CA,NY,CA
22  c1,c1,c2,c2                                             DC,LA,DC,LA

而我需要的是:

 GRP_ID     CABINETS                 ADDRESSES
    11  cabs1,cabs2,cabs3,cabs4       NY,CA,
    22  c1,c2                         DC,LA

这个例子在SQLFiddle中:http://sqlfiddle.com/#!1/4815e/19

如果使用一个表没有问题 - SQLFiddle:http://sqlfiddle.com/#!1/4815e/20

我错过了什么？是否可以这样做,因为加入？

Answer 1

而不是使用窗口函数和patitioning,使用查询级GROUP BY并与DISTINCT子句聚合:

SELECT         
  rnp.grp_id,
  array_to_string(array_agg(distinct rnp.cabinets),',') AS cabinets,
  array_to_string(array_agg(distinct ips.address),',')  AS addresses
FROM rnp JOIN ips ON rnp.grp_id=ips.grp_id GROUP BY rnp.grp_id, ips.grp_id;

结果:

 grp_id |        cabinets         | addresses 
--------+-------------------------+-----------
     11 | cabs1,cabs2,cabs3,cabs4 | CA,NY
     22 | c1,c2                   | DC,LA
(2 rows)

这里的关键是,不使用窗口函数和修改,而是使用查询级别GROUP BY和聚合与DISTINCT子句.

这也适用于窗口函数方法,除了PostgreSQL(至少9.1)不支持DISTINCT窗口函数:

regress=# SELECT DISTINCT
  rnp.grp_id,
  array_to_string(array_agg(distinct rnp.cabinets)OVER (PARTITION BY rnp.grp_id), ',') AS cabinets,                    
  array_to_string(array_agg(distinct ips.address) OVER (PARTITION BY ips.grp_id), ',') AS addresses
FROM rnp JOIN ips ON rnp.grp_id=ips.grp_id;
ERROR:  DISTINCT is not implemented for window functions
LINE 3:   array_to_string(array_agg(distinct rnp.cabinets)OVER (PART...