Question

I’ve been using MySQL for quite some time, and I’ve decided to learn Postgres. The transition hasn’t been terrible, but I ran into a snag today:

I have the admin account, postgres, and a user specifically for this application, tv. For the sake of convenience, I was modifying some rows in a table under the admin account. The website would not reflect any of the changes I was making to the database.

After blaming my various caching strategies for about an hour I finally ran psql as the tv user and noticed that none of the rows in there reflected the alterations made while logged in as postgres. Coming from a MySQL background, this behavior was completely baffling to me.

Long story short: is this a feature, or did I misconfigure something somewhere? And is there any way to make the database not act like this?

Thanks for any help.

---

Update: Here are a sampling of commands:

[12:23:04] blake$ sudo -u postgres psql -d teevee
psql (9.1.3)
Type "help" for help.

teevee=# SELECT COUNT(*) FROM episode;
 count
-------
     1
(1 row)

[12:23:25] blake$ psql -U tv -d teevee -h localhost -W
Password for user tv:
psql (9.1.3)
SSL connection (cipher: DHE-RSA-AES256-SHA, bits: 256)
Type "help" for help.

teevee=> SELECT COUNT(*) FROM episode;
 count
-------
   176
(1 row)

Answer 1

You can list schemas using \dn or \dn+ under psql, but I would suggest using a tool like pgAdmin to have a better visual representation. You can find whether you have conflicting tables in those potential tv and public schemas using \dt public.* and \dt tv.*.

Using multiple schemas can be useful, but is often not required.

The default search path tends to be "$user",public, which means that it will first try to use the tables in the schema named after the user and fall back onto the public schema (the most common).

I would suggest using the public schema in general, unless you have a good reason not to. This should generally be the default. (I’m not sure why you had another schema.)

You can change a table schema using ALTER TABLE xxxxxx SET SCHEMA yyyyyyy, although you may need to drop the other one with a conflicting name beforehand. You would certainly want to copy the data out first.

Something like INSERT INTO public.episode SELECT * FROM tv.episode should work in your case, even better with the column names specified. Whether this works may depend on other constraints.

If you have nothing useful in your public schema at the moment, you might as well drop the schema entirely (DROP SCHEMA public) and rename the other one (ALTER SCHEMA tv RENAME to public). Depending on what’s already granted, you may then also need something like this:

GRANT ALL ON SCHEMA public TO postgres;
GRANT ALL ON SCHEMA public TO public;