automatthew-stevedore

Stevedore is a tool for performing and comparing sets of benchmarks based on similar setup and teardown needs. Instances use their class's setup/teardown where available, and you can set or override these methods per instance.

Steve (and his subclasses, naturally) track any instances created and can run a comparison using Steve.compare_instances. If the R language is available, you can do a power analysis using the data from a trial run to suggest sample sizes and number of runs needed to meet certain figures of statistical reliability.

Steve uses the rsruby gem, if available, for basic stats and power analysis. As the rsruby gem uses a C extension, this could be problematic for non-MRI Rubys. In the absence of rsruby, stevedore uses bmarini-mathstats for the basic stats and attempts to shell out to R for power analysis (i.e. power.t.test)

class StringConcat < Stevedore before_sample do ... end after_sample do ... end end

plus = StringConcat.new "Combining strings with +" do measure do 10000.times do "supercalifragilisticexpiali" + "docious" end end end

append = StringConcat.new "Combining strings with <<" do measure do 10000.times do "supercalifragilisticexpiali" << "docious" end end end

interpolate = StringConcat.new "Combining strings with interpolation" do measure do 10000.times do "supercalifragilisticexpiali#{'docious'}" end end end

run_count, sample_size = 3, 8

puts "Measuring #{run_count} runs of #{sample_size} for each test." StringConcat.run([append, interpolate], run_count, sample_size) puts StringConcat.report([append, interpolate]) => putses the following:

Combining strings

                            Mean       Stddev      Minimum       Median          Max

---

interpolating strings 0.003064 0.002386 0.001356 0.001682 0.008544 appending strings 0.009267 0.002933 0.005337 0.010756 0.013299 Ratio of means: 3.02487110476993