Castore π¦«
Making Event Sourcing easy π
Event Sourcing is a data storage paradigm that saves changes in your application state rather than the state itself. It is powerful but tricky to implement.
After years of using it at Kumo, we have grown to love it, but also experienced first-hand the lack of consensus and tooling around it. That's where Castore comes from!
Castore is a TypeScript library that makes Event Sourcing easy π
With Castore, you'll be able to:
All that with first-class developer experience and minimal boilerplate β¨
π« Core Design
Some important decisions that we've made early on:
π Abstractions first
Castore has been designed with flexibility in mind. It gives you abstractions that are meant to be used anywhere: React apps, containers, Lambdas... you name it!
For instance, EventStore
classes are stack agnostic: They need an EventStorageAdapter
class to interact with actual data. You can code your own EventStorageAdapter
(simply implement the interface), but it's much simpler to use an off-the-shelf adapter like DynamoDBEventStorageAdapter
.
π
ββοΈ We do NOT deploy resources
While some packages like DynamoDBEventStorageAdapter
require compatible infrastructure, Castore is not responsible for deploying it.
Though that is not something we exclude in the future, we are a small team and decided to focus on DevX first.
β Full type safety
Speaking of DevX, we absolutely love TypeScript! If you do too, you're in the right place: We push type-safety to the limit in everything we do!
If you don't, that's fine π Castore is still available in Node/JS. And you can still profit from some nice JSDocs!
π Best practices
The Event Sourcing journey has many hidden pitfalls. We ran into them for you!
Castore is opiniated. It comes with a collection of best practices and documented anti-patterns that we hope will help you out!
Table of content
Getting Started
π₯ Installation
npm install @castore/core
yarn add @castore/core
π¦ Packages structure
Castore is not a single package, but a collection of packages revolving around a core
package. This is made so every line of code added to your project is opt-in, wether you use tree-shaking or not.
Castore packages are released together. Though different versions may be compatible, you are guaranteed to have working code as long as you use matching versions.
Here is an example of working package.json
:
{
...
"dependencies": {
"@castore/core": "1.3.1",
"@castore/dynamodb-event-storage-adapter": "1.3.1"
...
},
"devDependencies": {
"@castore/test-tools": "1.3.1"
...
}
}
The Basics
π Events
Event Sourcing is all about saving changes in your application state. Such changes are represented by events, and needless to say, they are quite important π
Events that concern the same business entity (like a User
) are aggregated through a common id called aggregateId
(and vice versa, events that have the same aggregateId
represent changes of the same business entity). The index of an event in such a serie of events is called its version
.
In Castore, stored events (also called event details) always have exactly the following attributes:
aggregateId (string)
version (integer β₯ 1)
timestamp (string)
: A date in ISO 8601 formattype (string)
: A string identifying the business meaning of the eventpayload (?any = never)
: A payload of any typemetadata (?any = never)
: Some metadata of any type
import type { EventDetail } from '@castore/core';
type UserCreatedEventDetail = EventDetail<
'USER_CREATED',
{ name: string; age: number },
{ invitedBy?: string }
>;
type UserCreatedEventDetail = {
aggregateId: string;
version: number;
timestamp: string;
type: 'USER_CREATED';
payload: { name: string; age: number };
metadata: { invitedBy?: string };
};
π· EventType
Events are generally classified in events types (not to confuse with TS types). Castore lets you declare them via the EventType
class:
import { EventType } from '@castore/core';
export const userCreatedEventType = new EventType<
'USER_CREATED',
{ name: string; age: number },
{ invitedBy?: string }
>({ type: 'USER_CREATED' });
Note that we only provide TS types for payload
and metadata
properties. That is because, as stated in the core design, Castore is meant to be as flexible as possible, and that includes the validation library you want to use: The EventType
class is not meant to be used directly, but rather extended by other classes which will add run-time validation methods to it π
See the following packages for examples:
Constructor:
type (string)
: The event type
import { EventType } from '@castore/core';
export const userCreatedEventType = new EventType({ type: 'USER_CREATED' });
Properties:
type (string):
The event type
const eventType = userCreatedEventType.type;
Type Helpers:
EventTypeDetail
: Returns the event detail TS type of an EventType
import type { EventTypeDetail } from '@castore/core';
type UserCreatedEventTypeDetail = EventTypeDetail<typeof userCreatedEventType>;
type UserCreatedEventTypeDetail = {
aggregateId: string;
version: number;
timestamp: string;
type: 'USER_CREATED';
payload: { name: string; age: number };
metadata: { invitedBy?: string };
};
EventTypesDetails
: Return the events details of a list of EventType
import type { EventTypesDetails } from '@castore/core';
type UserEventTypesDetails = EventTypesDetails<
[typeof userCreatedEventType, typeof userRemovedEventType]
>;
π Aggregate
Eventhough entities are stored as series of events, we still want to use a stable interface to represent their states at a point in time rather than directly using events. In Castore, it is implemented by a TS type called Aggregate
.
βοΈ Think of aggregates as "what the data would look like in CRUD"
In Castore, aggregates necessarily contain an aggregateId
and version
attributes (the version
of the latest event
). But for the rest, it's up to you π€·ββοΈ
For instance, we can include a name
, age
and status
properties to our UserAggregate
:
import type { Aggregate } from '@castore/core';
interface UserAggregate extends Aggregate {
name: string;
age: number;
status: 'CREATED' | 'REMOVED';
}
interface UserAggregate {
aggregateId: string;
version: number;
name: string;
age: number;
status: 'CREATED' | 'REMOVED';
}
βοΈ Reducer
Aggregates are derived from their events by reducing them through a reducer
function. It defines how to update the aggregate when a new event is pushed:
import type { Reducer } from '@castore/core';
export const usersReducer: Reducer<UserAggregate, UserEventsDetails> = (
userAggregate,
newEvent,
) => {
const { version, aggregateId } = newEvent;
switch (newEvent.type) {
case 'USER_CREATED': {
const { name, age } = newEvent.payload;
return {
aggregateId,
version,
name,
age,
status: 'CREATED',
};
}
case 'USER_REMOVED':
return { ...userAggregate, version, status: 'REMOVED' };
}
};
const johnDowAggregate: UserAggregate = johnDowEvents.reduce(usersReducer);
βοΈ Note that aggregates are always computed on the fly, and NOT stored. Changing them does not require any data migration whatsoever (except if you use snapshots, an invalidation is needed first).
π EventStore
Once you've defined your event types and how to aggregate them, you can bundle them together in an EventStore
class.
Each event store in your application represents a business entity. Think of event stores as "what tables would be in CRUD", except that instead of directly updating data, you just append new events to it!
In Castore, EventStore
classes are NOT responsible for actually storing data (this will come with event storage adapters). But rather to provide a boilerplate-free and type-safe interface to perform many actions such as:
- Listing aggregate ids
- Accessing events of an aggregate
- Building an aggregate with the reducer
- Pushing new events etc...
import { EventStore } from '@castore/core';
const userEventStore = new EventStore({
eventStoreId: 'USERS',
eventTypes: [
userCreatedEventType,
userRemovedEventType,
...
],
reducer: usersReducer,
});
βοΈ The EventStore
class is the heart of Castore, it even gave it its name!
Constructor:
...coming soon
Properties:
...coming soon
Type Helpers:
...coming soon
πΎ EventStorageAdapter
...coming soon
π¨ Command
...coming soon
πΈ Snapshots
...coming soon
Resources
π― Test Tools
...coming soon
π Packages List
...coming soon
π Common Patterns
...coming soon