Answer a question

I have a fairly large postgreql database that I've inherited. We have a job that runs ~monthly that backs up the existing database and creates a new database with updated vendor data that we receive.

Currently there is a small issue with it. Without going into details of the table setup, what the data is modeling, etc, I believe it can be fixed with a simple delete query, as the tables are set-up to use cascading deletes.

However, it takes about 9 hours to generate this database from the vendor's provided source, so I'm always hesitant to introduce new changes. I currently have a copy of this database that I intend to run the query on first, to make user it 'CAN' run successfully. However, one of the downsides of sql in general is that when deletes are made, the output is typically something alone the lines of:

Query returned successfully: x rows affected, y ms execution time.

Is there a way in postgres to determine what rows were removed from which tables when using cascading deletes? I'd like to run the query on my copy and see if what is being dropped is what I expect, at least in what tables are being hit. Is this possible?

Answers

Is there a way in postgres to determine what rows were removed from which tables when using cascading deletes?

Unfortunately no simple built-in way, but it's a great idea. Since cascading deletes are implemented under the hood by triggers, all you'd have to do is modify the referential integrity triggers so that cascade deletes raise a LOG level message with with the row information before deleting it. This would require changes to the PostgreSQL source code and a recompile.

Alternately, you can log all deletes by creating a new AFTER DELETE ... FOR EACH ROW trigger on each table referred to by a cascade constraint. It can be as simple as:

CREATE OR REPLACE FUNCTION log_delete() RETURNS trigger AS $$
BEGIN
    RAISE LOG 'Deleting row % (statement is %)', OLD, current_query();
    RETURN NULL;
END;
$$ LANGUAGE plpgsql;

That way you don't need to modify the DB, though you do need to install the trigger on every table that's subject to cascade deletes.

For bonus points you can even automate its creation by querying the information_schema for foreign key relationships, but it's probably more hassle than its worth.

# postgresql # cascadeur
Logo
PostgreSQL

PostgreSQL社区为您提供最前沿的新闻资讯和知识内容

更多推荐

PostgreSQL 计数查询效率,物化视图 [重复]

问题:PostgreSQL 计数查询效率,物化视图 [重复] 可能重复: PostgreSQL 计数查询优化 使用 PostgreSQL 9.2,我们试图弄清楚是否有一种方法可以跟踪查询的结果数量,并以有效的方式返回该数字。这个查询应该每秒执行几次(可能几十到几百甚至几千次)。我们现在的查询看起来像这样,但我们想知道这是否效率低下: -- Get # of rows that do not hav

avatar PostgreSQL

多对多中的唯一性

问题:多对多中的唯一性 我无法弄清楚谷歌的哪些术语,所以帮助标记这个问题或只是以相关问题的方式向我指出会有所帮助。 我相信我有一个典型的多对多关系: CREATE TABLE groups ( id integer PRIMARY KEY); CREATE TABLE elements ( id integer PRIMARY KEY); CREATE TABLE groups_elements

avatar PostgreSQL

Django 与 postgresql - manage.py syncdb 返回错误

问题:Django 与 postgresql - manage.py syncdb 返回错误 我从 Django 开始。我设置了一些使用 SQLite 工作的站点,但是在将 DB 引擎更改为 postgresql manage.py syncdb 后返回错误。我已经用谷歌搜索了 2 天,但对我仍然没有任何作用。Postgres 用户 'joe' 具有超级用户权限和本地 'joe ' 数据库存在。

avatar PostgreSQL