Same-seeded rspec test runs no longer generate deterministic data

[jasonkarns]

Describe the bug

Given an identical Kernel seed value, Faker should generate the same deterministic data, but it does not.

Seems to have been caused by #2226

To Reproduce

Given a conventional rspec test suite, with randomization enabled: the conventional rspec configuration sets the Kernel seed to the rspec seed:

  # conventional spec_helper.rb
  config.order = :random
  Kernel.srand config.seed

Then rspec --seed 1234 generates non-deterministic faker data.

Expected behavior

I would expect Faker to be based on the Kernel.srand seed, as configured with rspec. This way two same-seeded rspec runs would generate identical test data.

Additional context

We're currently working around this by adding the following to our rspec configuration:

   config.order = :random
   Kernel.srand config.seed
+  Faker::Config.random = Random.new(config.seed)

[connorshea]

oooh 😬 I'm not sure how to fix this in the Faker gem itself other than just telling people to use the workaround you've suggested.

@vbrazo @koic any ideas here?

[havenwood]

Should Random.new here be Random.new(Random.seed) instead? The default seed argument is currently resulting in Random.new(Random.new_seed).

[hosamaly]

How about using RSpec's seed if it's loaded?

module Faker
  module Config
    def random
        @random || (defined?(RSpec) ? Random.new(RSpec.configuration.seed) : Random.new)
      end
  end
end

[connorshea]

This should be fixed now by #2488, although it's not in the 2.21.0 release that just went out. Regardless, I'm going to close this.