hair trigger should drop raw triggers before creating in schema.rb

[jenseng]

Spawned from discussion #26 ... in schema.rb, although create_trigger will drop the existing trigger first (if it exists), that is not the case for raw execute "CREATE TRIGGER..." calls that it does. Because of this, if you have any triggers that don't correspond to a create_trigger in a migration, schema.rb cannot be used to update an existing test database (à la rake db:test:prepare), only an empty one.

For any adapter-specific execute "CREATE TRIGGER..." calls, HairTrigger should first do a DROP TRIGGER IF EXISTS call

[jenseng]

@sshaw can you provide a little more info on how you're getting duplicate trigger errors from the execute "CREATE TRIGGER..." calls in schema.rb (per your comments on #26)? I have a fix, but am unable to actually reproduce the problem on the current code. I'd like to truly verify it fixes the problem you're seeing.

I've got a vanilla Rails app and have been testing it under MySQL, Postgres and SQLite. Some findings:

1. db:test:prepare drops the db before doing a db:schema:load, so any triggers are wiped out long before schema.rb adds any
2. db:schema:load doesn't drop the db, but it does drop every table (due to force: true). When a table is dropped, any triggers on it are dropped automatically as well. I've verified this under all three adapters.

Any info you can provide would be helpful (which rake command you run, other gems that might be affecting things, database you're using, etc.)

[sshaw]

Yes, I had trouble reproducing it too... Let me have a look at the app that was causing the problem, I'll get back to you.

[sshaw]

It turns out that some code I had added to deal with triggers before I found your gem was causing the problem. Sorry for the runaround... 🙊

[jenseng]

haha, no worries. thanks for all the other good info and pointers 👍