Crates.io | streameroo |
lib.rs | streameroo |
version | 0.1.2 |
source | src |
created_at | 2024-12-01 22:35:23.69879 |
updated_at | 2024-12-05 13:40:30.281099 |
description | Compilation of mini-frameworks & utilities for data-streaming applications |
homepage | |
repository | https://github.com/lur1an/streameroo |
max_upload_size | |
id | 1467928 |
size | 49,921 |
A collection of mini-frameworks & tools for building asynchronous applications in rust.
streameroo
is a collection of mini-frameworks & tools for building asynchronous applications in rust.
I built a small version of my ideas for working with AMQP using lapin
and tokio
, I have a few plans for working with fluvio.
Through traits & wrapper types you can instrument the framework to take care of a lot of boilerplate and verbose pain when working with lapin
.
The inner workings of streameroo
might change, however I am satisfied with the API, so that will very very likely NOT change in a breaking manner.
This mini-framework takes care of:
ack/nack
operations on various types of errorsreply-to
, standard)ChannelExt
for slightly more convenient publishing of eventsBatch consuming/acknowledging to maximize throughput. streameroo
currently runs on a single Channel
, (You can initialize multiple instances of Streameroo
with different channels) and also doesn't support message batching & batch acks
Create a struct for the event you want to handle. If you are using common formats like JSON, MsgPack or other support formats by streameroo
all you need is a DeserializeOwned
implementing struct
.
#[derive(Debug, Deserialize)]
struct MyEvent {
hello: String
}
Now create a handler. A handler is anything that implements the AMQPHandler
trait, you're not supposed to implement this trait yourself, its implemented automatically for matching async functions.
async fn test_event_handler(
pgpool: State<PgPool>,
redelivered: Redelivered,
event: Json<MyEvent>,
)
Lets dissect this a bit:
pgpool
: This is a global state that you can access in your handler, if you don't provide it in the context it'll panic. You can extract any piece of global data, as long as you have initialized it in the context.exchange
: This uses the Redelivered
extractor to access the redelivered
flag of the delivery.event
: This is the event you want to handle. The Json
extractor instructs streameroo
to deserialize MyEvent
using serde_json
. For custom protocols that are not covered by the framework you can either just use Vec<u8>
or implement Decode
yourself.Now you can create a Streameroo
instance and consume events from a queue.
The instance of the application runs on top of a Channel
, if you want to use multiple channels you need to re-initialize the context, plans to improve on this will come later. However most of the time only a single Channel
is used for consumption.
// Connect to the broker and create a channel
// This could be shortened and implemented on top of `Streameroo` in the future if this feels clunky
let url = "amqp://user:password@localhost:5672";
let connection = lapin::Connection::connect(url, Default::default()).await?;
let channel = connection.create_channel().await?;
// Initialize your application's context (global state + channel)
let mut context = Context::new(ctx.channel.clone());
// Add any global state you want to access in your handler, for example a database connection
context.data(pgpool);
// Give your app a consumer tag and the context
let mut app = Streameroo::new(context, "test-consumer");
// Now you can spawn as many consumers as you want
app.consume(test_event_handler, "test").await?;
AMQPHandler
The AMQPHandler
trait is used to define a handler for a specific event. Its implemented for all async functions that return a Result<T, E>
where T
is AMQPResult
and E
can be turned into a boxed error type.
async fn test_event_handler(
counter: StateOwned<Arc<AtomicU8>>,
exchange: Exchange,
redelivered: Redelivered,
event: Json<TestEvent>,
) -> anyhow::Result<()> {
let event = event.into_inner();
assert_eq!(exchange.into_inner(), "");
assert_eq!(event.0, "hello");
let count = counter.load(Ordering::Relaxed);
tracing::info!(?count);
if count == 0 {
assert!(!redelivered.into_inner());
} else {
assert!(redelivered.into_inner());
}
if count < 3 {
counter.fetch_add(1, Ordering::Relaxed);
anyhow::bail!("Go again");
}
Ok(())
}
This is a valid handler. All required values are extracted and injected into it at runtime.
AMQPResult
The AMQPResult
trait is used to control what happens to a delivery after it has been handled by your handler. The body of the AMQPResult
implementation is executed before attempting to acknowledge the delivery. If the impl fails the delivery is nack
'ed
Some common impls are:
()
Nothing happens, the delivery is acknowledged as is.Publish<E>
Publishes the event to the exchange specified in the Publish
struct.
E
implements Encode
. Encode
is implemented for most common formats like JSON, MsgPack and others through the wrapper types.DeliveryAction
is for fine-grained control, the best way to use this is set your error type to Infallible
and handle everything yourself by returning the correct Deliveryaction
for your usecase. When returning this value the automatic ack
of the framework is disabled as its taken care of by the DeliveryAction
impl.PublishReply<E>
follows the RPC pattern for rabbitmq. It publishes the event E
t the queue specified in the reply-to
header.ChannelExt
ChannelExt is an extension trait for lapin's Channel
type that allows working with events with slightly less boilerplate and reuse the wrapper types and Decode/Encode
traits.
The most useful one is the direct_rpc
method which implements the direct reply to RPC pattern from the RabbitMQ docs.
let result: Json<TestEvent> = channel
.direct_rpc(
"",
&queue,
Duration::from_secs(5),
Json(TestEvent("hello".into())),
)
.await?;