我们正在重构框架Postgresql驱动程序以允许事务。在此过程中,我们引入了一些问题,导致以下错误
pg_query(): duplicate key value violates unique constraint DETAIL: Key (id)=(1) already exists
链接到特拉维斯测试与一些更多的细节https://travis-ci.org/photodude/database/jobs/175596877
有问题的驱动程序的相关部分在这个链接
https://github.com/joomla-framework/database/blob/master/src/Postgresql/PostgresqlDriver.php L711-L819
与该问题相关的测试为
https://github.com/joomla-framework/database/blob/master/Tests/DriverPostgresqlTest.php L1116-L1163
我得到的表序列是搞砸了,但我在为什么表序列搞砸了,甚至只是如何修复代码,使测试功能正确的损失。
注意:我认为这个失败与准备好的和未准备好的语句有关
在第519行重新启动序列和截断表,这看起来不错,但如果在回滚事务中运行,将不会发生截断,但序列将重新启动
Important: Because sequences are non-transactional, changes made by setval are not undone if the transaction rolls back.
:
s1=> create table test1 ( id serial primary key, a text not null);
CREATE TABLE
s1=> d
List of relations
Schema | Name | Type | Owner
--------+--------------+----------+--------
public | test1 | table | albert
public | test1_id_seq | sequence | albert
(2 rows)
s1=> insert into test1(a) values ('apple');
INSERT 0 1
s1=> select * from test1;
id | a
----+-------
1 | apple
(1 row)
s1=> select * from test1_id_seq;
sequence_name | last_value | start_value | increment_by | max_value | min_value | cache_value | log_cnt | is_cycled | is_called
---------------+------------+-------------+--------------+---------------------+-----------+-------------+---------+-----------+-----------
test1_id_seq | 1 | 1 | 1 | 9223372036854775807 | 1 | 1 | 32 | f | t
(1 row)
s1=> insert into test1(a) values ('bannana');
INSERT 0 1
s1=> select * from test1;
id | a
----+---------
1 | apple
2 | bannana
(2 rows)
s1=> insert into test1(a) values ('bannana');
INSERT 0 1
s1=> select * from test1;
id | a
----+---------
1 | apple
2 | bannana
3 | bannana
(3 rows)
s1=> select * from test1_id_seq;
sequence_name | last_value | start_value | increment_by | max_value | min_value | cache_value | log_cnt | is_cycled | is_called
---------------+------------+-------------+--------------+---------------------+-----------+-------------+---------+-----------+-----------
test1_id_seq | 3 | 1 | 1 | 9223372036854775807 | 1 | 1 | 30 | f | t
(1 row)
s1=> begin;
BEGIN
s1=> alter sequence test1_id_seq RESTART WITH 1;
ALTER SEQUENCE
s1=> truncate table test1;
TRUNCATE TABLE
s1=> rollback;
ROLLBACK
s1=> select * from test1_id_seq;
sequence_name | last_value | start_value | increment_by | max_value | min_value | cache_value | log_cnt | is_cycled | is_called
---------------+------------+-------------+--------------+---------------------+-----------+-------------+---------+-----------+-----------
test1_id_seq | 1 | 1 | 1 | 9223372036854775807 | 1 | 1 | 0 | f | f
(1 row)
s1=> select * from test1;
id | a
----+---------
1 | apple
2 | bannana
3 | bannana
(3 rows)
s1=>
我们仍然不确定根本原因,但整个问题都与单元测试有关。某些东西正在/正在全局重新启动表序列,即使只检查了一个失败的测试。我们找到了解决问题的方法,但仍在寻找根本原因。
我们还发现需要改进所有驱动程序测试的tearDown()方法。