I'm not really sure why we're looking at RAILS_ENV and the current ActiveRecord connection in init.rb. Why not just patch both PostgreSQL and MySQL all the time? This would give the ability to use ActiveRecord and the spatial_adapter outside of Rails.

It looks like get_rails2_conditions (called here)

  #For Rails >= 2
  def self.sanitize_sql_hash_for_conditions(attrs)
    conditions = get_rails2_conditions(attrs)
    replace_bind_variables(conditions, expand_range_bind_variables(attrs.values))
  end

Doesn't fully replicate the 2.3.2 base.rb code in the case where value.is_a?(Hash)

Unfortunately I'm new to rails to haven't yet figured out what aspects are missing, but I get a bind error when doing a join even if the table has no spatial columns (and the error goes away if I comment out overriding sanitize_sql_hash_for_conditions in the spatial_adapter code).

Any help is appreciated.

I'm trying to get my code using PostGIS and this plugin under unit test using rspec

I'm having trouble getting tables with spatial columns loaded in the test db with the rake db:test:load rake task.

rake db:schema:dump is giving me this

create_table "room_geometries", :force => true do |t|
  t.column "room_id", :integer
  t.column "created_at", :timestamp
  t.column "updated_at", :timestamp
  t.column "the_geom", :multi_polygon
end

DB Test load is giving me all the columns except for 'the_geom'.

Is there instructions out there on what I need to do to get this testable? (I've not seen much with the google searches I've done)

The PostGIS adapter is currently initialising the column with super(base, name, type, limit, default, null). It should actually be super(base, name, type, limit, nil, nil, default, null) because the fifth and sixth arguments are actually for :precision and :scale. Consequently the :null option is not getting handled properly. I'm guessing this was due to a change in Rails? I'm using the 2-3-stable branch.