Optionally fork before running each job

[bkeepers]

This should be an option that is easy to turn on or off.

[johnae]

Or threads for jruby?

[bkeepers]

Sure, I'm not opposed to that.

[DeTeam]

Possible configuration option:

Qu.process_with = lambda do |job_closure|
  Thread.new { job_closure.call }
end

Qu.process_with = lambda do |job_closure|
  job_closure.call if fork
end

By default it's just .call on arg. What do you think?

[bkeepers]

I think I would rather see it wrapped up in a class so it can be tested and extended. I have a hunch the different strategies will be more complicated than you'd want to handle in a lambda.

class ThreadedProcessor
  def initializer(max_threads = 5)
    @max_threads = max_threads
  end

  def run(job)
    @threads << Thread.new { job.process }
    wait_until_a_thread_dies if at_max_threads?
  end

  # …
end

Qu.job_processor = ThreadedProcessor.new(10)

[DeTeam]

Agree, strategy classes seem fine.

There's also an idea to move backend interface to a separate class with "implement me" warning.

[jnunemaker]

This is done. There is a runner that can be customized in #86