Postgres 连接查询有时采用笛卡尔积

Question

我正在尝试为一个查询连接多个表，但从数据库中得到的结果不一致，我相信我的查询采用的是所有用户的笛卡尔积，而我只需要 DirectConversation 中的用户.

供参考的架构：

查询为（其中$id代表变量User.id）：

SELECT c.*, count(dm.id), 
u1.first_name, u1.last_name, u1.company, u1.picture, 
u2.first_name, u2.last_name, u2.company, u2.picture
FROM "DirectConversation" as c, "DirectMessage" as dm, "Profile" as u1, "Profile" as u2
WHERE u1."id_User" = c."id_User1"
AND u2."id_User" = c."id_User2"
AND c.id = dm."id_DirectConversation" 
AND dm.viewed = 'f' AND dm.deleted = 'f'
AND c."id_User1" = $id OR c."id_User2" = $id
GROUP BY c.id, u1.id, u2.id;

预期结果（用户id=1时的结果）：

id | id_User1 | id_User2 | count | first_name | last_name |      company       |                                   picture                                   | first_name | last_name |    company     |                                 picture                                  
----+----------+----------+-------+------------+-----------+--------------------+-----------------------------------------------------------------------------+------------+-----------+----------------+--------------------------------------------------------------------------
  1 |        1 |        2 |     3 | Albert     | Einstein  | alberts inventions | http://upload.wikimedia.org/wikipedia/commons/d/d3/Albert_Einstein_Head.jpg | Nikola     | Tesla     | Teslas Widgets | http://upload.wikimedia.org/wikipedia/commons/7/79/Tesla_circa_1890.jpeg
(1 row)

(END)

错误结果（用户id=2时的结果）：

id | id_User1 | id_User2 | count | first_name | last_name |      company       |                                                    picture                                                     | first_name | last_name |      company       |                                                    picture                                                     
----+----------+----------+-------+------------+-----------+--------------------+----------------------------------------------------------------------------------------------------------------+------------+-----------+--------------------+----------------------------------------------------------------------------------------------------------------
  1 |        1 |        2 |     4 | Albert     | Einstein  | alberts inventions | http://upload.wikimedia.org/wikipedia/commons/d/d3/Albert_Einstein_Head.jpg                                    | Albert     | Einstein  | alberts inventions | http://upload.wikimedia.org/wikipedia/commons/d/d3/Albert_Einstein_Head.jpg
  1 |        1 |        2 |     4 | Albert     | Einstein  | alberts inventions | http://upload.wikimedia.org/wikipedia/commons/d/d3/Albert_Einstein_Head.jpg                                    | Nikola     | Tesla     | Teslas Widgets     | http://upload.wikimedia.org/wikipedia/commons/7/79/Tesla_circa_1890.jpeg
  1 |        1 |        2 |     4 | Albert     | Einstein  | alberts inventions | http://upload.wikimedia.org/wikipedia/commons/d/d3/Albert_Einstein_Head.jpg                                    | Rosalind   | Franklin  | DNA R US           | http://upload.wikimedia.org/wikipedia/en/9/97/Rosalind_Franklin.jpg
  1 |        1 |        2 |     4 | Albert     | Einstein  | alberts inventions | http://upload.wikimedia.org/wikipedia/commons/d/d3/Albert_Einstein_Head.jpg                                    | Charles    | Babbage   | Babbages Cabbages  | http://upload.wikimedia.org/wikipedia/commons/6/6b/Charles_Babbage_-_1860.jpg

...请注意，为简洁起见，这被截断了。我相信这是所有用户的笛卡尔积，但我不知道为什么

我使用的 postgres 版本：

版本

-----------------------------------------------------------------------------------------------------
 PostgreSQL 9.3.6 on x86_64-unknown-linux-gnu, compiled by gcc (Ubuntu 4.8.2-19ubuntu1) 4.8.2, 64-bit

Answer 1

我刚刚将你的连接的 where 子句移动到 ON 语句并进行了正确的连接，如果这不起作用我会设置一个 sqlfiddle 看看这个 sql 是

SELECT c.*, count(dm.id), 
    u1.first_name, u1.last_name, u1.company, u1.picture, 
    u2.first_name, u2.last_name, u2.company, u2.picture
    FROM "DirectConversation" c
    JOIN "DirectMessage" dm ON c.id = dm."id_DirectConversation" 
    JOIN "Profile" u1 ON u1."id_User" = c."id_User1"
    JOIN "Profile" u2 ON u2."id_User" = c."id_User2"
    WHERE 
    dm.viewed = 'f' AND dm.deleted = 'f'
    AND (c."id_User1" = $id OR c."id_User2" = $id)
    GROUP BY c.id, u1.id, u2.id;

编辑：为了安全起见将 OR 子句分组

Postgres 连接查询有时采用笛卡尔积

Postgres Join Query is SOMETIMES taking the cartesian product

sql

postgresql

postgresql-9.3

版本