Return a custom wrapped JobError with clean backtrace.

Author: metaskills

TODO.md

@@ -24,11 +24,9 @@ def _enqueue(job, send_message_opts = {})
 end
 ```
 
-* Use job's `attr_accessor :executions` vs `ApproximateReceiveCount`
 * Error handlers. Ensure we easily hook into Rollbar, etc.
-* Is `delete_message` message needed? Is 200 from consumer implied delete?
 * Can I set Rails tempalte `VisibilityTimeout` to just +1 of function timeout or full 43200?
-* Can I get rid of the Job re-raising and rely on change message visibility alone?
+* Do this in our gem. `ActiveJob::Base.logger = Logger.new(IO::NULL)`
 
 ## Doc Points
 
@@ -60,7 +58,6 @@ DO I MIRROR or MIGRATE
 ## Max Retries
 
 * Max is twelve.
-*
 
 ## Migrating from Sidekiq
 

lib/lambdakiq.rb

@@ -4,6 +4,7 @@
 require 'active_job/queue_adapters'
 require 'active_support/all'
 require 'lambdakiq/version'
+require 'lambdakiq/error'
 require 'lambdakiq/adapter'
 require 'lambdakiq/client'
 require 'lambdakiq/queue'

lib/lambdakiq/error.rb

@@ -0,0 +1,13 @@
+module Lambdakiq
+  class Error < StandardError
+    attr_reader :original_exception, :job
+
+    def initialize(error)
+      @original_exception = error
+      super(error.message)
+      set_backtrace Rails.backtrace_cleaner.clean(error.backtrace)
+    end
+  end
+
+  class JobError < Error ; end
+end

lib/lambdakiq/job.rb

@@ -10,7 +10,8 @@ def handler(event)
         jobs = records.map { |record| new(record) }
         jobs.each(&:perform)
         jwerror = jobs.detect{ |j| j.error }
-        jwerror ? raise(jwerror.error) : true
+        return unless jwerror
+        raise JobError.new(jwerror.error)
       end
 
     end