Implement GoodJob::Batch

[bensheldon]

Connects to #417, #691.

• Add description field
• Add Dashboard view
• Add cleanup for batches
• Rename batch params to properties
• Make migration-less patch-release compatible

[mollerhoj]

I find this instruction unclear. To me, it looks like the code below contains a race condition. What happens if all the jobs finishes before enqueue is called?.

On another note: What happens if you call enqueue twice on the same batch?
What happens if you call enqueue on a batch without any jobs?

[bensheldon]

Great callout! I will cover that in the documentation. Briefly:

• the callback job won't be enabled until #enqueue is called on the batch.
• calling #enqueue on a batch without any jobs will immediately enqueue the callback job
• #enqueue can be called multiple times, which will re-enable the callback job to be enqueued if it had previously been triggered because all jobs had finished.
• #add will add jobs to the batch without triggering an enqueue

Do you think this kind of example makes sense for creating a multistage batch?

class MyBatchJob < ApplicationJob
  def perform(batch, options = {})
    if batch.properties[:stage] == 1
      batch.properties[:stage] = 2
      batch.enqueue do
        10.times { OtherJob.perform_later}
      end
    elsif batch.properties[:stage] == 2
      batch.properties[:stage] = 3
      batch.enqueue do
        OtherJob.perform_later
      end
    else
      # all done
    end 
  end
end

GoodJob::Batch.enqueue(MyBatchJob, stage: 1)

[mollerhoj]

I have not have time yet to go over the code in detail, but I've written some questions that popped up when reading the documentation.

All in all, this looks very, very good to me. Nice API, seems like a pretty clean implementation, and an absolute KILLER feature.

I think the documentation would be even better with an example of a complex workflow (using batches to run both serial and parallel jobs in succession.) E.g, show how one would implement this workflow: