complexity_assert

ComplexityAssert

They are some performance critical pieces of code that will be executed on huge data sets, which we want to make sure will run fast enough. Unfortunately, enforcing this is not easy, often requiring large scale and slow benchmarks. This RSpec library (the result of an experiment to learn machine learning) uses linear regression to determine the time complexity (Big O notation, O(x)) of a piece of code and to check that it is at least as good as what we expect. This does not require huge data sets (only a few large ones) and can be written as any unit test (not as fast though).

Installation

Add this line to your application's Gemfile:

gem 'complexity_assert', :group => [:test]

And then execute:

$ bundle

Or install it yourself as:

$ gem install complexity_assert

Usage

In order to test the complexity of an algorithm, you need to provide 2 things :

• the algorithm
• a way to generate some inputs of varying size

For this, you need to provide an object that answers to messages generate_args and run. Here is an example

# An adapter class to fit the code to measure in complexity assert
class LinearSearch

  # Generate some arguments of a particular size
  def generate_args(size)
    [ Array.new(size) { rand(1..size) }, rand(1..size) ]
  end

  # Run the code on which we want to assert performance
  def run(array, searched)
    found = false;
    array.each do |element|
      if element == array
        found = true
      end
    end
    found
  end
end

describe "Linear search" do

  it "performs linearly" do
    # Verify that the code runs in time proportional to the size of its arguments
    expect(LinearSearch.new).to be_linear()
  end

end

There are currently 3 matchers :

• be_constant
• be_linear
• be_quadratic

More could be added in the future. Every matcher passes if a faster complexity is identified (be_linear willalso pass if the algorithm is detected to be constant).

That means that for the moment, be_quadratic always passes, but might turn out useful when we add more complex models (Internally, it is quite useful, as it is used to identify that something is more linear than quadratic !).

Development

It uses rubybox, simply clone this repo, build the image, and start on.

git clone ...
cd ...
docker-compose build
docker-compose run rubybox

From then on, you're inside the ruby box, run bin/setup to install dependencies. Then, run rake spec to run the tests. You can also run bin/console for an interactive prompt that will allow you to experiment.

To install this gem onto your local machine, run bundle exec rake install. To release a new version, update the version number in version.rb, and then run bundle exec rake release, which will create a git tag for the version, push git commits and tags, and push the .gem file to rubygems.org.

Contributing

Bug reports and pull requests are welcome on GitHub at https://github.com/[USERNAME]/complexity_assert.

Here is a quick and uncomplete list of things that could improve this library :

• Add Code Climate
• Refactor some code around the matchers
• Cache the linear regression (it's done twice)
• Factorize / find a way to better generate the sizes, or allow the assertion to specify the sizes
• Allow the assertion to specify the warmup and run rounds
• Robustness against GC : use gc intensive ruby methods, and see how the regression behaves
• O(lnx) : pre-treat with exp()
• O(?lnx) : use exp, then a search for the coefficient (aka polynomial)
• O(xlnx) : there is no well known inverse for that, we can compute it numericaly though
• O(x?) : do some kind of dichotomy or search to find the most probable model
• Estimate how much the assert is deterministic
• ...

As I said, this is still experimental ! Any help is welcome !

License

The gem is available as open source under the terms of the MIT License.