Class: InstagramConnect::SendMessageJob

Inherits:
ApplicationJob
  • Object
show all
Defined in:
app/jobs/instagram_connect/send_message_job.rb

Overview

Sends one pending outbound Message via the Graph API, exactly once.

The claim is an atomic UPDATE from pending to sending, so a duplicate enqueue cannot double-send. Everything after it keeps an at-most-once bias: on any doubt the message is marked failed for an operator to retry deliberately, never resent automatically. A duplicate message reaching a customer is worse than a visible failure.

Instance Method Summary collapse

Instance Method Details

#perform(message_id) ⇒ Object 



12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
 
# File 'app/jobs/instagram_connect/send_message_job.rb', line 12

def perform(message_id)
  claimed = Message.where(id: message_id, status: "pending")
                   .update_all(status: "sending", updated_at: Time.current)
  return unless claimed == 1

  message = Message.find(message_id)
  # The claim was callback-free, so a host rendering this thread live needs
  # telling before the send starts rather than after it finishes.
  message.broadcast_refresh

  conversation = message.conversation

  # Meta rejects a send on a thread another app owns, and the error it
  # returns says nothing useful about why. Refusing here names the reason.
  if conversation.controlled_elsewhere?
    return fail_message(message, "thread_controlled_elsewhere",
                        "Another app currently has control of this conversation.")
  end

  tag = MessagingWindow.new(last_inbound_at: conversation.last_inbound_at).send_tag
  if tag == :blocked
    return fail_message(message, "outside_messaging_window",
                        "The reply window has closed; the customer must message again.")
  end

  deliver(message, conversation, tag)
rescue InstagramConnect::TokenUnreadableError => e
  # The generic ApplicationJob rescue logs and stops — right for a sync
  # job, wrong here: this job has already claimed the message into
  # "sending", and swallowing the error leaves that bubble spinning
  # forever with nothing to retry. The operator must see it fail, with
  # the real reason on it.
  fail_message(message, "token_unreadable", e.message)
end