Comments on End Point Blog: pgcrypto pg_cipher_exists errors on upgrade from P...

Robert: Yes, that would be nice, but I'm not h...

2010-06-14T10:33:49.369-04:00

Robert: Yes, that would be nice, but I'm not holding my breath on that one. For the final iteration, I will probably export the schema only, run uninstall plus install for each contrib module, and then install the data. What's annoying is some contrib modules do not have uninstall SQL in older versions!

This is why I have been advocating putting contrib...

2010-06-12T17:46:09.592-04:00

This is why I have been advocating putting contrib modules in their own schemas for years. In most cases, it's much simpler to exclude the contrib modules on the dump, and install the new version of contrib module before upgrading.

Anon: In addition to Jon's reasoning (which is...

2010-06-11T13:02:17.461-04:00

Anon: In addition to Jon's reasoning (which is quite correct: I hate the fact that the functions simply disappeared), I prefer my way because I was running the dump often and on different databases, so that would have been a lot of extra steps to do each time compared with pg_dump --clean mydb | psql

Certainly you could do that. That's what I wou...

2010-06-11T12:48:36.253-04:00

Certainly you could do that. That's what I would have done myself.

But I think Greg makes a good case that it's nice to have the old functions defined but throwing a meaningful error, rather than just not working and making it appear that pgcrypto isn't installed, or isn't installed correctly.

In other words, I think Greg's solution here is about helping users of the system after the import, not just fixing the import itself.

Can't you just dump in custom format, generate...

2010-06-11T11:00:13.589-04:00

Can't you just dump in custom format, generate the dump list, remove the lines you don't want and import that?