#374 - allow concurrent consumption of events

[nachogiljaldo]

Closes ThreeDotsLabs/watermill#374

Why?

It would be nice to have a simple way to increase the concurrency of message processing. With the current options, it is only possible to consume one message at the time (i.e. you need to get > process > ACK > get > process > ...).

That is not actually required in many use cases where events with the same key are sent to the same partition and, therefore, a partial order exists within the partition.

What?

This PR:

• documents the previous and new consumption models
• extracts the messageHandler struct to an interface, so multiple implementations can be provided while preserving the subscriber's structure
• creates 2 new implementations of the interface:
  • batchedMessageHandler works by collecting batches of messages up to a certain time or duration. Those messages are then sent to the output channel. NACKed messages are managed by resending every message following the NACKed one in that topic / partition
  • partitionConcurrentMessageHandler works by multiplexing each claim into the output channel, that allows having up to N messages in-flight, where N is the number of topics/partitions in the subscriber. NACKs / ACKs are handled by each multiplexed
• updates the tests so they run with both models (which makes them slower :/ )

IMPORTANT: I tried to preserve the current default behavior. Therefore, there is a Default consumption model which behaves the same way it does now.

Potential follow-up

A potential follow-up would be a consumer that allows for up to N messages being in-flight. That would allow for cases in which the developers want to maximize parallelism at the expense of complexity on their code.

[nachogiljaldo]

Hey @m110 sorry for the direct ping, but I wonder if you could, as member of the org, review the PR or ask somebody who could do it.

[nachogiljaldo]

@aldinugroho24 @yangyangrusli @m110 are you the maintainers of this repo? If yes, would you mind having a look at this PR? I think it would be a nice improvement. Also I wonder if you could use any help maintaining this repo. I see there's a bunch of piled PRs and I'm affected by some of those bugs, so I'd be happy to help if it works.

[m110]

I'm sorry about the long wait @nachogiljaldo, I'm on it now. Thanks for contributing!