Questions about how this library works

[samijnih]

Hey,

First of all, I'd like to thank you for that great work on this library. It seems very promising. I'm currently testing it because I'm fed up of Symfony Messenger. I've been using messaging libraries for a while now:

• tactician
• simplebus
• prooph
• event sauce
• messenger
• php-amqplib/RabbitMqBundle
• ...

I really want to use your lib for my next professional projects.

To be honest, it's pretty hard to understand how it works out of the box by reading the documentation :/

I'll list some feedbacks about my misunderstanding! :)

1/ For Symfony, what is the difference between an enqueue.processor and enqueue.transport.processor? In which case should I use one instead of the other?

• If I tag a service with enqueue.processor and run enqueue:transport:consume then, I'll get an error because it's incompatible
• If I tag a service with enqueue.transport.processor and run enqueue:consume then, I'll get an error because it's incompatible
• https://php-enqueue.github.io/bundle/message_processor/ does not explain cases when we should use a transport or not

My thought is that enqueue:transport:consume enables to handle a message from a given queue with a given processor even if the processor is not configured on that queue. enqueue:consume will select the appropriate queue binded processor to handle that message

2/ I want to dispatch commands or events using async. For commands, while reading the documentation, I thought the only way to do it, was to use enqueue/async-command library. But it is not. I'm able to async send commands and consume those later using enqueue:consume with an SQS client.

I think, you should make a clear distinction between Producer sending commands/events and cases when we need to use async-command

3/ Configuration

I'm using the symfony bridge. I want to declare a queue named "place-order" for my transport named "sqs"

enqueue:
    sqs:
        transport: '%env(ENQUEUE_SQS_DSN)%'
        client:
            traceable_producer: '%kernel.debug%'
            prefix: 'dev'
            app_name: ''
            separator: '-'
            router_queue: 'place-order'
            redelivered_delay_time: 0

When I run enqueue:setup-broker then it perfectly creates a queue named "dev-place-order" into SQS.

Now, let's say I want to manually register a processor on that queue by tagging it. If I rerun enqueue:setup-broker, then it will create another queue named dev-default:

Configuration:

    Domain\Order\Saga\PlaceOrderSaga:
        tags:
            - {
                name: 'enqueue.processor',
                command: Domain\Order\Features\Place\PlaceOrder
            }

Command output:

/app # bin/console enqueue:setup-broker -vvv
[debug] [SqsDriver] Declare router queue: dev-place-order
[debug] [SqsDriver] Declare processor queue: dev-default
Broker set up

As your you can see, It creates a queue named dev-default only because I've tagged my processor. I don't want that. So following the documentation, I have to specify a queue name tag, e.g:

    Domain\Order\Saga\PlaceOrderSaga:
        tags:
            - {
                name: 'enqueue.processor',
                command: Domain\Order\Features\Place\PlaceOrder,
                queue: place-order
            }

That will work. But I've debugged the source code and it will actually make a HTTP request to AWS SQS for creating this queue which already exist:

/app # bin/console enqueue:setup-broker -vvv
[debug] [SqsDriver] Declare router queue: dev-place-order
[debug] [SqsDriver] Declare processor queue: dev-place-order
Broker set up

So... I must define the client key in a transport, otherwise an exception will be raised. If I don't want any default queue to be created, I have to define router_queue in client with my queue name.
If I wanna manually register a processor and if I don't want a default queue to be created, I have to define the queue tag.

How can I avoid that double queue creation?

4/ Multiple queues on a same transport

Let's say I want only one transport (SQS) with N queues (queue1, queue2, queue3). Should I do that?

enqueue:
    queue1:
        transport: '%env(ENQUEUE_SQS_DSN)%'
        client:
            router_queue: 'queue1'

    queue2:
        transport: '%env(ENQUEUE_SQS_DSN)%'
        client:
            router_queue: 'queue2'

    queue3:
        transport: '%env(ENQUEUE_SQS_DSN)%'
        client:
            router_queue: 'queue3'

If so, I found it pretty confusing to get 3 different clients for the same transport to achieve this.

I'm more expecting something like:

enqueue:
    transports:
        monolith:
            dsn: '%env(ENQUEUE_SQS_DSN)%'
            default_queue: foo
            queues:
                one:
                    name: queue1
                two:
                    name: queue2
                three:
                    name: queue3

        order:
            dsn: '%env(ENQUEUE_SQS_DSN)%'
            default_queue: foo
            queues:
                one:
                    name: place-order
                two:
                    name: order-placed
                three:
                    name: order-payment

        monolith-topic:
            dsn: '%env(ENQUEUE_SNS_DSN)%'
            default_topic: bar
            topics:
                one:
                    name: topic1

Thanks!

[samijnih]

There is something weird I've noticed.

If you have a processor queue and a queue with the same name

• when you send a message, 2 SQS calls will be made
• when you run enqueue:consume, 3 calls will be made but the message will be consumed only once