Tests can be slow. This is how to find out why they’re slow.
Toss this bad boy into spec/support/profile.rb and tag any example with :profile => true and it’ll spit out callgrind dumps for your consumption in KCachegrind or similar.
If you specify PROFILE=all on your command line, it’ll profile all examples, regardless of tagging. If you pass PROFILE=true (or any other non-nil, non-ALL value) then it’ll profile tagged examples.
RSpec.configure do |c|
def profile
result = RubyProf.profile { yield }
name = example.metadata[:full_description].downcase.gsub(/[^a-z0-9_-]/, "-").gsub(/-+/, "-")
printer = RubyProf::CallTreePrinter.new(result)
open("tmp/performance/callgrind.#{name}.#{Time.now.to_i}.trace", "w") do |f|
printer.print(f)
end
end
c.around(:each) do |example|
if ENV['PROFILE'] == 'all' or (example.metadata[:profile] and ENV['PROFILE'])
profile { example.run }
else
example.run
end
end
end
Then add the tag to your tests:
describe "A test" do
it "can be profiled", :profile => true do
expect { some_complex_operation }.to_not raise_error
end
end
Bam.