Prevent Duplicates with Delayed Jobs

add_signature_fields_to_delayed_jobs.rb

class AddFieldsToDelayedJobs < ActiveRecord::Migration
  def change
    add_column :delayed_jobs, :signature, :string
    add_column :delayed_jobs, :args, :text
  end
end

delayed_duplicate_prevention_plugin.rb

# /lib/delayed_duplicate_prevention_plugin.rb
require 'delayed_job'
class DelayedDuplicatePreventionPlugin < Delayed::Plugin
 
  module SignatureConcern
    extend ActiveSupport::Concern

    included do
      before_validation :add_signature
      validate :prevent_duplicate
    end

    private
    def add_signature
      self.signature = generate_signature
      self.args = self.payload_object.args
    end

    def generate_signature
      pobj = payload_object
      if pobj.object.respond_to?(:id) and pobj.object.id.present?
        sig = "#{pobj.object.class}"
        sig += ":#{pobj.object.id}" 
      else
        sig = "#{pobj.object}"
      end
      
      sig += "##{pobj.method_name}"
      return sig
    end    

    def prevent_duplicate
      if DuplicateChecker.duplicate?(self)
        Rails.logger.warn "Found duplicate job(#{self.signature}), ignoring..."
        errors.add(:base, "This is a duplicate") 
      end
    end
  end

  class DuplicateChecker
    attr_reader :job

    def self.duplicate?(job)
      new(job).duplicate?
    end

    def initialize(job)
      @job = job
    end

    def duplicate?
      possible_dupes = Delayed::Job.where(signature: job.signature)
      possible_dupes = possible_dupes.where.not(id: job.id) if job.id.present?
      result = possible_dupes.any?{|possible_dupe| args_match?(possible_dupe, job)}
      result
    end

    private

    def args_match?(job1, job2)
      # TODO: make this logic robust
      normalize_args(job1.args) == normalize_args(job2.args)
    end

    def normalize_args(args)
      args.kind_of?(String) ? YAML.load(args) : args
    end
  end
end

delayed_job.rb

# config/initializers/delayed_job.rb
require 'delayed_duplicate_prevention_plugin'

Delayed::Backend::ActiveRecord::Job.send(:include, DelayedDuplicatePreventionPlugin::SignatureConcern)
Delayed::Worker.plugins << DelayedDuplicatePreventionPlugin

Why create a new column instead of querying the handler?

https://stackoverflow.com/a/70041500/1536309

Because querying a long serialized string on a large table won't scale well when the table is large. I think a hybrid approach would work well though where you debounce not against the handler but against a key or signature that identifies the job.

Hey @synth thanks a lot!
We did a gem based on that :)
https://github.com/noesya/delayed_job_prevent_duplicate

Very cool! gem > gist :)

@synth thanks.

I like interacting with active job API rather than the delayed_job syntax. Taking some concept from here I implemented a unique job using an active job callback https://gist.github.com/channainfo/b920eeda6b20576310c1fae9780dbedc

@SebouChu @pabois should we integrate in the gem?