[Nitro] Og newbie problems

[Kristof Jozsa]

Hi,

admittedly I started to play with Og today so please feel free to correct me 
with anything (or just send over /dev/null :), but I'd like to share my 
experiences and ask for suggestions. All my experiments happened on my gentoo 
linux box, Og installed by gem, version 0.22.0 on ruby 1.8.2.

The first problem I bumped into can be demonstrated with the following piece of 
code:

----
#!/usr/bin/ruby -rubygems

require 'test/unit'

require 'og'
require 'og/relation'

class Order
   property :name, String

   has_many :items, LineItem
end

class LineItem
   belongs_to Order
end

class TestOg < Test::Unit::TestCase
   include Og

   def test_basic
   end

   def setup
     og = Og.setup(:store => :memory, :name => 'test')
   end
end
----

running this script dumps this error:
/test_og.rb:15: undefined method `belongs_to' for LineItem:Class (NoMethodError)

I kept on fighting and after some time I realized that adding any property to 
LineItem before the belongs_to declaration fixes the problem. Is there any other 
way to trigger initializing Og (I think that's what fails to happen in this 
case) than this?


Next I had set up an unidirectional many-to-many relationship. I tried very hard 
to set it up correctly and get the join table generated, right until I realized 
that it IS generated just not in the logs.. (the two other create table sql 
statements were displayed as INFO messages but not the join table :) Do you 
think it would make sense displaying that info as well?

Still at the many-to-many relationship (example was many SoftwarePackage can 
contain many Software), I tried to create a constructor for SoftwarePackage 
which accepted an array of Software (simplifying the example here):

----
class Software
   property :name, String
end

class SoftwarePackage < Software
   many_to_many :software, Software

   def initialize(software)
     software.each { |x| add_software x }
   end
end
----

and well, it did not really work out. I find out that I have to call save first 
in the SoftwarePackage constructor, else an invalid sql will be generated like:

INSERT INTO j_software_softwarepackage (software_oid,softwarepackage_oid) VALUES 
(1, )

with a missing id. Again, I've got a workaround here so I'm not sure if that's 
the way it's meant to work or that's clarified as a rough edge and you plan to 
smooth it a bit :) Still at this topic slightly against following the least 
surprise principle first I assumed the following line to work in the constructor:

software.each { |x| @software << x }

Maybe it would be nice to have that too if possible.


As a last question, is there a way to change the oid naming strategy followed by 
Og? A similar approach as setting Og.table_prefix = nil would be really useful.

Wrapping up, I'm absolutely amazed with the power of Og. I'm actually sketching 
a new java project's software design from scratch with ruby and Og (first time 
attempting this), and I'm blown away with the speed I could build/test and play 
around with my domain objects using Og.

Thanks much, and keep up the excellent work,

Kristof Jozsa