Sql POSTGRES-仅在填写特定列时插入

拥有：

ID start_date  end_date     account 
1   2016-06-30  null         123
1   2017-04-19  null         111
1   2017-04-26  2017-07-30   789

insert into table (ID, start_date, account) values (1,2016-08-01,123) 

Insert into insert_only (ID, start_date, account) values (1,'2016-08-01',123) ON CONFLICT DO NOTHING;

Insert into insert_only (ID, start_date, end_date,account) values (1,'2016-08-01','2016-08-01',789) ON CONFLICT DO NOTHING;

BEFORE: 

stackoverflow=# select * from insert_only;
 id | start_date |  end_date  | account 
----+------------+------------+---------
  1 | 2016-08-01 |            |     123
  1 | 2017-08-01 |            |     111
  1 | 2017-04-26 | 2017-07-30 |     789
  1 | 2016-08-01 | 2016-08-01 |     789
(4 rows)

stackoverflow=# Insert into insert_only (ID, start_date, account) values (1,'2016-08-01',123) ON CONFLICT DO NOTHING;
INSERT 0 0

stackoverflow=# Insert into insert_only (ID, start_date, end_date,account) values (1,'2016-08-01','2016-08-01',789) ON CONFLICT DO NOTHING;
INSERT 0 1

AFTER:

stackoverflow=# select * from insert_only;
 id | start_date |  end_date  | account 
----+------------+------------+---------
  1 | 2016-08-01 |            |     123
  1 | 2017-08-01 |            |     111
  1 | 2017-04-26 | 2017-07-30 |     789
  1 | 2016-08-01 | 2016-08-01 |     789
  1 | 2016-08-01 | 2016-08-01 |     789
(5 rows)

想要实现的目标：

ID start_date  end_date     account 
1   2016-06-30  null         123
1   2017-04-19  null         111
1   2017-04-26  2017-07-30   789

insert into table (ID, start_date, account) values (1,2016-08-01,123) 

Insert into insert_only (ID, start_date, account) values (1,'2016-08-01',123) ON CONFLICT DO NOTHING;

Insert into insert_only (ID, start_date, end_date,account) values (1,'2016-08-01','2016-08-01',789) ON CONFLICT DO NOTHING;

BEFORE: 

stackoverflow=# select * from insert_only;
 id | start_date |  end_date  | account 
----+------------+------------+---------
  1 | 2016-08-01 |            |     123
  1 | 2017-08-01 |            |     111
  1 | 2017-04-26 | 2017-07-30 |     789
  1 | 2016-08-01 | 2016-08-01 |     789
(4 rows)

stackoverflow=# Insert into insert_only (ID, start_date, account) values (1,'2016-08-01',123) ON CONFLICT DO NOTHING;
INSERT 0 0

stackoverflow=# Insert into insert_only (ID, start_date, end_date,account) values (1,'2016-08-01','2016-08-01',789) ON CONFLICT DO NOTHING;
INSERT 0 1

AFTER:

stackoverflow=# select * from insert_only;
 id | start_date |  end_date  | account 
----+------------+------------+---------
  1 | 2016-08-01 |            |     123
  1 | 2017-08-01 |            |     111
  1 | 2017-04-26 | 2017-07-30 |     789
  1 | 2016-08-01 | 2016-08-01 |     789
  1 | 2016-08-01 | 2016-08-01 |     789
(5 rows)

是否有办法创建约束/索引/其他机制来检查：

如果ID和帐户存在且结束日期为空->不执行任何操作

如果ID和帐户存在且结束日期不为空->插入行

所以：

ID start_date  end_date     account 
1   2016-06-30  null         123
1   2017-04-19  null         111
1   2017-04-26  2017-07-30   789

insert into table (ID, start_date, account) values (1,2016-08-01,123) 

Insert into insert_only (ID, start_date, account) values (1,'2016-08-01',123) ON CONFLICT DO NOTHING;

Insert into insert_only (ID, start_date, end_date,account) values (1,'2016-08-01','2016-08-01',789) ON CONFLICT DO NOTHING;

BEFORE: 

stackoverflow=# select * from insert_only;
 id | start_date |  end_date  | account 
----+------------+------------+---------
  1 | 2016-08-01 |            |     123
  1 | 2017-08-01 |            |     111
  1 | 2017-04-26 | 2017-07-30 |     789
  1 | 2016-08-01 | 2016-08-01 |     789
(4 rows)

stackoverflow=# Insert into insert_only (ID, start_date, account) values (1,'2016-08-01',123) ON CONFLICT DO NOTHING;
INSERT 0 0

stackoverflow=# Insert into insert_only (ID, start_date, end_date,account) values (1,'2016-08-01','2016-08-01',789) ON CONFLICT DO NOTHING;
INSERT 0 1

AFTER:

stackoverflow=# select * from insert_only;
 id | start_date |  end_date  | account 
----+------------+------------+---------
  1 | 2016-08-01 |            |     123
  1 | 2017-08-01 |            |     111
  1 | 2017-04-26 | 2017-07-30 |     789
  1 | 2016-08-01 | 2016-08-01 |     789
  1 | 2016-08-01 | 2016-08-01 |     789
(5 rows)

->什么也没发生

insert into table (ID, start_date, account) values (1,2017-04-26, 789)

->插入行

我在想，可能是一个带有存储函数的检查约束，它将检查上述内容或类似的内容？或索引：

create unique index unique_account on table (ID, account) where end_date is null

---

这是最佳实践

警告：在高度事务数据库中使用触发器将导致性能问题

解决方案：我假设您使用PostgreSQL 9.5或更高版本

直到这一步你几乎完成

create unique index unique_account on table (ID, account) where end_date is null

您只需在PostgreSQL 9.5或更高版本中使用

UPSERT

query即可

ANS案例1:

ID start_date  end_date     account 
1   2016-06-30  null         123
1   2017-04-19  null         111
1   2017-04-26  2017-07-30   789

insert into table (ID, start_date, account) values (1,2016-08-01,123) 

Insert into insert_only (ID, start_date, account) values (1,'2016-08-01',123) ON CONFLICT DO NOTHING;

Insert into insert_only (ID, start_date, end_date,account) values (1,'2016-08-01','2016-08-01',789) ON CONFLICT DO NOTHING;

BEFORE: 

stackoverflow=# select * from insert_only;
 id | start_date |  end_date  | account 
----+------------+------------+---------
  1 | 2016-08-01 |            |     123
  1 | 2017-08-01 |            |     111
  1 | 2017-04-26 | 2017-07-30 |     789
  1 | 2016-08-01 | 2016-08-01 |     789
(4 rows)

stackoverflow=# Insert into insert_only (ID, start_date, account) values (1,'2016-08-01',123) ON CONFLICT DO NOTHING;
INSERT 0 0

stackoverflow=# Insert into insert_only (ID, start_date, end_date,account) values (1,'2016-08-01','2016-08-01',789) ON CONFLICT DO NOTHING;
INSERT 0 1

AFTER:

stackoverflow=# select * from insert_only;
 id | start_date |  end_date  | account 
----+------------+------------+---------
  1 | 2016-08-01 |            |     123
  1 | 2017-08-01 |            |     111
  1 | 2017-04-26 | 2017-07-30 |     789
  1 | 2016-08-01 | 2016-08-01 |     789
  1 | 2016-08-01 | 2016-08-01 |     789
(5 rows)

*我们需要

“冲突”

的唯一索引。此查询意味着，如果您插入相同的

id+account+date为空

，则

“冲突时”

将执行且

不执行任何操作

ANS案例2:

ID start_date  end_date     account 
1   2016-06-30  null         123
1   2017-04-19  null         111
1   2017-04-26  2017-07-30   789

insert into table (ID, start_date, account) values (1,2016-08-01,123) 

Insert into insert_only (ID, start_date, account) values (1,'2016-08-01',123) ON CONFLICT DO NOTHING;

Insert into insert_only (ID, start_date, end_date,account) values (1,'2016-08-01','2016-08-01',789) ON CONFLICT DO NOTHING;

BEFORE: 

stackoverflow=# select * from insert_only;
 id | start_date |  end_date  | account 
----+------------+------------+---------
  1 | 2016-08-01 |            |     123
  1 | 2017-08-01 |            |     111
  1 | 2017-04-26 | 2017-07-30 |     789
  1 | 2016-08-01 | 2016-08-01 |     789
(4 rows)

stackoverflow=# Insert into insert_only (ID, start_date, account) values (1,'2016-08-01',123) ON CONFLICT DO NOTHING;
INSERT 0 0

stackoverflow=# Insert into insert_only (ID, start_date, end_date,account) values (1,'2016-08-01','2016-08-01',789) ON CONFLICT DO NOTHING;
INSERT 0 1

AFTER:

stackoverflow=# select * from insert_only;
 id | start_date |  end_date  | account 
----+------------+------------+---------
  1 | 2016-08-01 |            |     123
  1 | 2017-08-01 |            |     111
  1 | 2017-04-26 | 2017-07-30 |     789
  1 | 2016-08-01 | 2016-08-01 |     789
  1 | 2016-08-01 | 2016-08-01 |     789
(5 rows)

*使用相同的查询，我们输入账户的

end\u date

值

789

插入将成功

输出：

ID start_date  end_date     account 
1   2016-06-30  null         123
1   2017-04-19  null         111
1   2017-04-26  2017-07-30   789

insert into table (ID, start_date, account) values (1,2016-08-01,123) 

Insert into insert_only (ID, start_date, account) values (1,'2016-08-01',123) ON CONFLICT DO NOTHING;

Insert into insert_only (ID, start_date, end_date,account) values (1,'2016-08-01','2016-08-01',789) ON CONFLICT DO NOTHING;

BEFORE: 

stackoverflow=# select * from insert_only;
 id | start_date |  end_date  | account 
----+------------+------------+---------
  1 | 2016-08-01 |            |     123
  1 | 2017-08-01 |            |     111
  1 | 2017-04-26 | 2017-07-30 |     789
  1 | 2016-08-01 | 2016-08-01 |     789
(4 rows)

stackoverflow=# Insert into insert_only (ID, start_date, account) values (1,'2016-08-01',123) ON CONFLICT DO NOTHING;
INSERT 0 0

stackoverflow=# Insert into insert_only (ID, start_date, end_date,account) values (1,'2016-08-01','2016-08-01',789) ON CONFLICT DO NOTHING;
INSERT 0 1

AFTER:

stackoverflow=# select * from insert_only;
 id | start_date |  end_date  | account 
----+------------+------------+---------
  1 | 2016-08-01 |            |     123
  1 | 2017-08-01 |            |     111
  1 | 2017-04-26 | 2017-07-30 |     789
  1 | 2016-08-01 | 2016-08-01 |     789
  1 | 2016-08-01 | 2016-08-01 |     789
(5 rows)

---

可以通过将insert包装到存储过程中来实现这一点。然后拒绝对表的写访问。因此，所有写入操作都必须通过应用所有必要检查的过程。哦，但你已经考虑过了。我感觉你实际上想要一个约束来强制执行非重叠的日期范围（每个{id，account}）。在这种情况下，使用

+无限

作为开放区间比空区间更实用。希望这个解决方案对你有所帮助。永远不要在insert中使用触发器进行验证，当您的数据增长到100万时，这将在将来造成麻烦。