So, I think there is probably a simpler way to test what you're trying to test, but without more context I don't know what to recommend. However, here is some proof-of-concept code to show that what you want to do can be done:
describe "test" do
class TestClass
attr_accessor :opts
def initialize(opts={})
@opts = opts
end
def bar
[]
end
end
let!(:stubbed) do
TestClass.new(args).tap{|obj| obj.stub(:bar).and_return("bar")}
end
let!(:unstubbed) { TestClass.new(args) }
before :each do
TestClass.stub(:new) do |args|
case args
when { :foo => "foo" }
stubbed
else
unstubbed
end
end
end
subject { TestClass.new(args) }
context "special arguments" do
let(:args) { { :foo => "foo" } }
its(:bar) { should eq "bar" }
its(:opts) { should eq({ :foo => "foo" }) }
end
context "no special arguments" do
let(:args) { { :baz => "baz" } }
its(:bar) { should eq [] }
its(:opts) { should eq({ :baz => "baz" }) }
end
end
test
special arguments
bar
should == bar
opts
should == {:foo=>"foo"}
no special arguments
bar
should == []
opts
should == {:baz=>"baz"}
Finished in 0.01117 seconds
4 examples, 0 failures
However I'm making a lot of use of special subject/let context blocks here. See http://benscheirman.com/2011/05/dry-up-your-rspec-files-with-subject-let-blocks/ for more on that subject.
result
versusresults
in the question. Probably not your answer, but worth fixing? – Bridget