A event-driven framework for writing reliable and scalable system.
At a high level, it provides a few major components:
Event-Driven-Library consists of a number of modules that provide a range of functionality essential for implementing messagebus-like applications in Rust. In this section, we will take a brief tour, summarizing the major APIs and their uses.
You can register any general struct with [Command] Derive Macro as follows: ```rust
pub struct MakeOrder {
pub user_id: i64,
pub items: Vec
Likewise, you can do the same thing for Event: ```rust
pub struct OrderFailed { #[identifier] pub user_id: i64, }
pub struct OrderSucceeded{
#[identifier]
pub userid: i64,
pub items: Vec
Note that use of
internally(or
externally_notifiable) and
identifier` is MUST.
internally_notifiable
is marker to let the system know that the event should be handled
within the applicationexternally_notifiable
is to leave OutBox
.identifier
is to record aggregate id.Command handlers are responsible for handling commands in an application, the response of which is sent directly to clients. Commands are imperative in nature, meaning they specify what should be done.
```rust use eventdrivenlibrary::prelude::{initcommandhandler, initeventhandler};
initcommandhandler!(
{
MakeOrder: OrderHandler::makeorder,
CancelOrder: OrderHandler::cancelorder
}
);
``
In the example above, you see
MakeOrderis mapped to
OrderHandler::make_order`, handler in application layer.
At this point, imagine you want to handle both success/failure case of the MakeOrder
command processing.
Then you have to think about using event handlers.
Event
is a side effect of [Command] or yet another [Event] processing.
You can register as many handlers as possible as long as they all consume same type of Event as follows:
rust
init_event_handler!(
{
OrderFaild: [
NotificationHandler::send_mail,
],
OrderSucceeded: [
DeliveryHandler::checkout_delivery_items,
InventoryHandler::change_inventory_count
]
}
);
In the MakeOrder
Command Handling, we have either OrderFailed
or OrderSucceeded
event with their own processing handlers.
Events are raised in the handlers that are thrown to [MessageBus] by [Context].
[MessageBus] then loops through the handlers UNLESS StopSentinel
is received.
Handlers can be located anywhere as long as they accept two argument:
```rust
pub async fn make_order(
cmd: MakeOrder,
context: AtomicContextManager,
) -> Result
let mut order_aggregate = OrderAggregate::new(cmd);
uow.repository().add(&mut task_aggregate).await?;
uow.commit::<ServiceOutBox>().await?;
Ok(().into())
}
``` But sometimes, you may want to add yet another dependencies. For that, Dependency Injection mechanism has been implemented. So, you can also do something along the lines of:
```rust
pub async fn makeorder(
cmd: MakeOrder,
context: AtomicContextManager,
paymentgateway_caller: Box
let mut order_aggregate = OrderAggregate::new(cmd,payment_gateway_caller);
uow.repository().add(&mut task_aggregate).await?;
uow.commit::<ServiceOutBox>().await?;
Ok(().into())
} ```
How is this possible? because we preprocess handlers so it can allow for DI container
.
You can simply register dependencies by putting attribute on top of free function.
```rust
pub fn paymentgatewaycaller() -> Box
```
This is great as you can take your mind off static nature of the language.
At the core is event driven library is [MessageBus], which gets command and gets raised event from
UnitOfWork
and dispatch the event to the right handlers.
As this is done only in framework side, the only way you can 'feel' the presence of messagebus is
when you invoke it. Everything else is done magically.
```rust
pub struct MakeOrder { // Test Command
pub user_id: i64,
pub items: Vec
async fn testfunc(){
let bus = MessageBus::new(commandhandler(), eventhandler())
let command = MakeOrder{userid:1, items:vec!["shirts","jeans"]}
match bus.handle(command).await{
Err(err)=> { // test for error case }
Ok(val)=> { // test for happy case }
}
}
}
}
```
When command has not yet been regitered, it returns an error - BaseError::CommandNotFound
Be mindful that bus does NOT return the result of event processing as in distributed event processing.