PostgreSQL - 使用子查询更新多个列值

Question

我需要能够使用子查询的结果更新表中的多个列。一个简单的示例如下所示 -

UPDATE table1
SET (col1, col2) =
  ((SELECT MIN (ship_charge), MAX (ship_charge) FROM orders))
WHERE col4 = 1001; 

如何在 PostgreSQL 中做到这一点？

感谢您的任何提示！

更新：对于让示例对于我的实际用例而言过于简单，我深表歉意。下面的查询更准确 -

UPDATE table1
SET    (TOTAL_MIN_RATE, TOTAL_MAX_RATE) = (SELECT AVG(o.MIN_RATE), AVG(o.MAX_RATE)
                           FROM   ORDR o INNER JOIN table2 ba ON (o.PAY_ACCT_ID = ba.ACCT_ID) 
                                         INNER JOIN table3 mb ON (ba.BANK_ID = mb.BANK_ID)
                               WHERE  ba.CNTRY_ID = table1.CNTRY_ID AND 
                                              o.STUS_CD IN ('01','02','03','04','05','06') AND
                                  ((o.FRO_CRNCY_ID = table1.TO_CRNCY_ID AND o.TO_CRNCY_ID = table1.FRO_CRNCY_ID) OR
                                   (o.TO_CRNCY_ID = table1.TO_CRNCY_ID AND o.FRO_CRNCY_ID = table1.FRO_CRNCY_ID))   
                               GROUP BY ba.CNTRY_ID)

Answer 1

一个选项（但不是唯一一个）是使用两个单独的子查询：

update table1
set col1 = (select min(ship_charge) from orders),
    col2 = (select max(ship_charge) from orders)
where col4 = 1001;

来自fine manual for PostgreSQL 9.0's UPDATE：

根据标准，column-list 语法应该允许从单个行值表达式分配列列表，例如子选择：

UPDATE accounts SET (contact_last_name, contact_first_name) =
(SELECT last_name, first_name FROM salesmen
 WHERE salesmen.id = accounts.sales_id);

这目前没有实现——源必须是一个独立表达式的列表。

Answer 2

这不是最有效的方法，但很简单：

UPDATE table1 SET
col1 = (SELECT MIN (ship_charge) FROM orders),
col2 = (SELECT MAX (ship_charge) FROM orders)
WHERE col4 = 1001; 

Answer 3

如果你想避免两个子选择，查询可以这样重写：

UPDATE table1
  SET col1 = o_min, col2 = o_max
FROM ( 
    SELECT min(ship_charge) as o_min, 
           max(ship_charge) as o_max
    FROM orders
) t 
WHERE col4 = 1001

如果 ship_charge 没有被索引，这应该比两个子选择更快。如果 ship_charge 被索引，它可能不会有很大的不同

---

编辑

从 Postgres 9.5 开始，这也可以写成：

UPDATE table1
  SET (col1, col2) = (SELECT min(ship_charge), max(ship_charge) FROM orders)
WHERE col4 = 1001

Answer 4

UPDATE table1
SET
    col1 = subquery.min_value,
    col2 = subquery.max_value
FROM
(

    SELECT
        1001 AS col4,
        MIN (ship_charge) AS min_value,
        MAX (ship_charge) AS max_value
    FROM orders
) AS subquery
WHERE table1.col4 = subquery.col4

如果要一次更新table1中的多行，也可以在子查询中返回多行。

Answer 5

当您没有简单的子选择时，使用UPDATE FROM 是一个很好的解决方案。在这个UPDATE 中，我想将photos 表的event_profile_id 设置为照片所属照片集的所有者（事件配置文件也是所有者）。

UPDATE photos
SET event_profile_id=photos_and_events.event_profile_id
FROM (
  SELECT
    ph.id photo_id,
    pr.id event_profile_id
  FROM photos ph, profiles pr, photo_sets ps
  WHERE ph.main_photo_set_id=ps.id AND ps.owner_profile_id=pr.id
) AS photos_and_events
WHERE photos.id=photos_and_events.photo_id;

Answer 6

正如official document所说：你可以使用PostgreSQL更新的标准更新概要
UPDATE table SET { column = { expression | DEFAULT } | ( column [, ...] ) = ( { expression | DEFAULT } [, ...] ) } [, ...] [ FROM from_list ] [ WHERE condition ]

所以你可以这样写：

UPDATE table1 SET TOTAL_MIN_RATE = subQuery."minRate", TOTAL_MAX_RATE = subQuery.maxRate FROM ( SELECT AVG (o.MIN_RATE) AS minRate, AVG (o.MAX_RATE) AS maxRate FROM ORDR o INNER JOIN table2 ba ON (o.PAY_ACCT_ID = ba.ACCT_ID) INNER JOIN table3 mb ON (ba.BANK_ID = mb.BANK_ID) WHERE ba.CNTRY_ID = table1.CNTRY_ID AND o.STUS_CD IN ( '01', '02', '03', '04', '05', '06' ) AND ( ( o.FRO_CRNCY_ID = table1.TO_CRNCY_ID AND o.TO_CRNCY_ID = table1.FRO_CRNCY_ID ) OR ( o.TO_CRNCY_ID = table1.TO_CRNCY_ID AND o.FRO_CRNCY_ID = table1.FRO_CRNCY_ID ) ) GROUP BY ba.CNTRY_ID ) subQuery;

或者更简单的方法：

UPDATE table1 SET ( TOTAL_MIN_RATE, TOTAL_MAX_RATE ) = ( SELECT AVG (o.MIN_RATE) AS minRate, AVG (o.MAX_RATE) AS maxRate FROM ORDR o INNER JOIN table2 ba ON (o.PAY_ACCT_ID = ba.ACCT_ID) INNER JOIN table3 mb ON (ba.BANK_ID = mb.BANK_ID) WHERE ba.CNTRY_ID = table1.CNTRY_ID AND o.STUS_CD IN ( '01', '02', '03', '04', '05', '06' ) AND ( ( o.FRO_CRNCY_ID = table1.TO_CRNCY_ID AND o.TO_CRNCY_ID = table1.FRO_CRNCY_ID ) OR ( o.TO_CRNCY_ID = table1.TO_CRNCY_ID AND o.FRO_CRNCY_ID = table1.FRO_CRNCY_ID ) ) GROUP BY ba.CNTRY_ID );

Answer 7

我需要在一个表上进行多次插入，从两个表中获取数据，它们之间没有公共列，并忽略已经存在的记录。

以下 sql 在 Postgresql 11 上进行了测试，尽管它在 v9+ 上应该可以正常工作：

WITH permission_info AS (
    SELECT id
    FROM permission
    WHERE permission."key" LIKE 'prefix_for_admin_%'
), role_info AS (
    SELECT id 
    FROM role
    WHERE role."name" = 'Admin'
)
INSERT INTO role_permission_table
(
    role_id,
    permission_id
)
    SELECT role_info.id, permission_info.id FROM role_info, permission_info

ON CONFLICT DO NOTHING
;