Skip to main content

Randomization can be reproduced across test runs

In Ruby, randomness is seeded by calling srand and passing it the seed that you want to use. By doing this, subsequent calls to rand, shuffle, sample, etc. will all be randomized the same way given the same seed is passed to srand.

RSpec takes care not to seed randomization directly when taking action that involves randomness (such as random ordering of examples).

Since RSpec does not ever invoke srand, this means that you are free to choose which, if any, mechanism is used to seed randomization.

There is an example below of how to use RSpec's seed for this purpose if you wish to do so.

If you would like to manage seeding randomization without any help from RSpec, please keep the following things in mind:

  • The seed should never be hard-coded.

    The first example below only does this to show that seeding randomization with a seed other than the one used by RSpec will correctly seed randomization.

  • Report the seed that was chosen.

    The randomization that was used for a given test run can not be reproduced if no one knows what seed was used to begin with.

  • Provide a mechanism to feed the seed into the tests.

    Without this, the call to srand will have to be hard-coded any time it is necessary to replicate a given test run's randomness.

Background​

Given a file named ".rspec" with:

--require spec_helper

Given a file named "spec/random_spec.rb" with:

RSpec.describe 'randomized example' do
  it 'prints random numbers' do
    puts 5.times.map { rand(99) }.join("-")
  end
end

Specifying a seed using srand provides predictable randomization​

Given a file named "spec/spec_helper.rb" with:

srand 123

When I run rspec

Then the output should contain "66-92-98-17-83".

Passing the RSpec seed to srand provides predictable randomization​

Given a file named "spec/spec_helper.rb" with:

srand RSpec.configuration.seed

When I run rspec --seed 123

Then the output should contain "66-92-98-17-83".