gistfile1.txt

The problem with the above is that the two rspec "worlds" get mixed together. Despite passing a separate Options object to
my rpsec builder, it ends up using the same Reporter object as the spec itself! So at the end of the run rspec reports that 
it ran 2 specs - the one run by RspecBuilder and the one run by the spec itself.

rspec_builder.rb

class RspecBuilder

  require 'spec'


  # 'suites' is data structure that can be converted to a nested rspec example group
  def initialize(suites, spec_options)
    @suites = suites
    @spec_options = spec_options
  end

  def run
    declare_suites
    @spec_options.run_examples
  end

  def declare_suites
    @suites.each do |suite|
      declare_suite(self, suite)
    end
  end

  def declare_suite(parent, suite)
    me = self
    parent.describe suite["name"] do
      suite["children"].each do |suite_or_spec|
        type = suite_or_spec["type"]
        if type == "suite"
          me.declare_suite(self, suite_or_spec)
        elsif type == "spec"
          me.declare_spec(self, suite_or_spec)
        else
          raise "unknown type #{type} for #{suite_or_spec.inspect}"
        end
      end
    end
  end

  def declare_spec(parent, spec)
    example_name = spec["name"]
    parent.it example_name do
      # run spec here
    end
  end

end  

rspec_builder_spec.rb

require 'rubygems'
require 'spec'
require 'spec/autorun'

describe RspecBuilder do
  it "runs the given specs and reports results like an rspec" do

    error_stream = StringIO.new
    out_stream = StringIO.new
    options = Spec::Runner::Options.new(error_stream, out_stream)

    suites = [] # this would build the representation
    builder = RspecBuilder.new(suites)
    
    builder.spec_options = options
    builder.run
    
    # assertions here
    
  end
end