My comments are prepended with ACM

brice_spec.rb

# This is the method I am testing.
class Log
  private
  class Parse
    @queue = :parse
    # This is a resque job. perform will ultimately be called asynchronously.
    # Therefore, must find the Log before performing the action because it
    # may be dirty or have been deleted since.
    def self.perform(log_id)
      # ACM: I would change
      # Log.find(log_id).parse if Log.exists?(conditions: {id: log_id})
      # to
      log = Log.find(log_id) and log.parse
      # because it would require one query instead of two
      nil 
    end 
  end 
end

# Here is my spec
# resque/log_parse_spec.rb
describe Log::Parse do
  context "when given a valid log id" do
    before(:each) do
      @log = create(:log)
    end 

    it "should parse a log" do
      # ACM: This is a typical pattern used
      @log.should_receive(:parse).once
      Log.stub!(:find).and_return(@log)

      Log::Parse.perform(@log.id)
    end 
  end 

  context "when given a stale log id" do
    before(:each) do
      @log = create(:log)
      @log.delete
    end 

    it "should not parse any instance of Log" do
      # ACM: I agree for this instance
      # Here is a valid usage of any_instance.
      Log.any_instance.should_not_receive(:parse)

      Log::Parse.perform(@log.id)
    end 

    it "shouldn't raise an error if the log doesn't exist" do
      expect { Log::Parse.perform(@log.id) }.to_not raise_error
    end 
  end 
end