What does inverse_of do? What SQL does it generate?

From the documentation, it seems like the :inverse_of option is a method for avoiding SQL queries, not generating them. It's a hint to ActiveRecord to use already loaded data instead of fetching it again through a relationship.

Their example:

class Dungeon < ActiveRecord::Base
  has_many :traps, :inverse_of => :dungeon
  has_one :evil_wizard, :inverse_of => :dungeon
end

class Trap < ActiveRecord::Base
  belongs_to :dungeon, :inverse_of => :traps
end

class EvilWizard < ActiveRecord::Base
  belongs_to :dungeon, :inverse_of => :evil_wizard
end

In this case, calling dungeon.traps.first.dungeon should return the original dungeon object instead of loading a new one as would be the case by default.

I think :inverse_of is most useful when you are working with associations that have not yet been persisted. E.g.:

class Project < ActiveRecord::Base
  has_many :tasks, :inverse_of=>:project
end

class Task < ActiveRecord::Base
  belongs_to :project, :inverse_of=>:tasks
end

Now, in the console:

irb> p = Project.new
=> #<Project id: nil, name: nil, ...>
irb> t = p.tasks.build
=> #<Task id: nil, project_id: nil, ...>
irb> t.project
=> #<Project id: nil, name: nil, ...>

Without the :inverse_of arguments, t.project would return nil, because it triggers an sql query and the data isn't stored yet. With the :inverse_of arguments, the data is retrieved from memory.