fix table_exists? in postgresql adapter to always use current search_path

[bradrobertson]

modified table_exists? in the postgresql adapter to always use the current search_path unless a explicit schema is specified.

Currently one cannot load schema.rb into a new (empty) postgresql schema. create_table forces a table drop which checks table_exists? first. This returns true even though the table doesn't actually exist in the current schema. Thus a PG error is thrown when it tries to drop the nonexistent table.

[tardate]

fyi, I was working on something similar based off master, see: #1632
+1 yes I'd like to see this fixed too, as it also enables migration into independent schemas and compatibility with schema-partitioning approaches for multi-tenancy

[bradrobertson]

ah good one! Ya I planned on adding a patch for master also, just haven't gotten around to it as we don't use 3.1 in production ATM.

[bradrobertson]

Can someone give me some feedback on this pull request? It's been sitting here for a month without any comments (save 1) I'm interested to know how I can improve it

[mitchlloyd]

Yes! If this were in Rails it would have saved me many hours. I had monkey patched table_exists? so that I could use Postgresql schemas and I was running into problems when I used set_table_name on a model with a schema (i.e. set_table_name 'meta.accounts'). I applied your patch and it fixed my issues with table_exists?

Would love to see this added into Rails as I have an app that makes heavy use of Postgresql schemas, a feature which is mostly unusable with an unhacked Rails install.

[bradrobertson]

fyi this is what we're using in production if you need to use this patch:

gem 'rails', '3.0.8', :git => 'git://github.com/bradrobertson/rails', :branch => '3-0-stable'

[tardate]

Brad, +1 (again;-)
Note that I just realised my similar patch on 3.1 has been merged to master (covering view and schema-related improvements for the PostgreSQL adapter), suggest you see if there's anything you can scavenge from this to converge/align what's to go on 3.0 #1632

[bradrobertson]

cool, thx @tardate, will take a look.
If anyone on the Rails core team can comment as to what extra work from @tardate's might make this pull request more enticing I'm all ears. I was pretty happy with the simplicity of my patch, but am definitely happy to provide more functionality if that's what people are looking for.

[spastorino]

/cc @tenderlove

[spastorino]

Was this merged in master and 3-1-stable?

[bradrobertson]

3-1-stable has a different fix from @tardate. This was just for 3-0-stable

[spastorino]

cool

[holodigm]

This issue still exists in 3-1-stable

[bradrobertson]

really @tardate can you comment? I'm not on 3.1 yet

[holodigm]

I hit this on a 3.0.10 to 3.1 upgrade yesterday on a relative new multi-tenant build. Big thanks for the original fix.
#2909

[tardate]

#1632 addressed this for 3.1 but it looks like that while it's in rails/master it didn't get merged to an RC or 3.1 stable.