Oh application state. The source of so many bugs. React has Redux, Vue has Pinia, and Angular has NgRx as their most used state management libraries.
NgRx is a great solution for folks well versed in Angular-ese and can make tackling problems with enterprise-sized applications way easier. There's a very big application I work on where we as a team started using NgRx to help us out for this very reason.
I work on a team where most of the developers are hardened Java veterans and don't spend too much time on the frontend. NgRx helped a lot but we still had to deal with reducers, actions, effects, registries, and getting it all wired up properly. As a team we had to understand new concepts and basically adopt an additional framework on top of the framework many of us were already still struggling to learn.
To their credit, the folks behind NgRx have simplified it and have removed a lot of the boilerplate code (there used to be A LOT of boilerplate).
It's much better than trying to manage state without a library by passing around properties from child to parent/parent to child or by using services. I can say from experience that doing so gets very messy very quickly if you have even a moderately complicated application.
In 2020, Dan Wahlin (a big contributor to the Angular community), did a talk about a new library he came up with called Observable Store.
This was game changing and something I don't think got enough attention. Why? Because it's just so dang simple to use and easy to understand.
We just need a basic understanding of RxJS.
That's it!
No, really. Read on and I'll prove it.
To be clear, even though this post is tailored to Angular developers, this is not an Angular-only thing. It can be used with any JS framework - or without, as long as you have RxJS installed.
Even though the documentation is excellent I thought I'd share a starter project and a quick rundown of how it works.
The Rundown
Assuming you have an Angular app set up (and therefore RxJS installed), run the following command:
npm i @codewithdan/observable-store
Create a new class. It'll be treated as a service so it'll have that @Injectable decorator. Extend the ObservableStore class and type it with a model:
customer-store.ts
export interface StoreStateModel {
customers: CustomerModel[],
addMode: boolean,
}
@Injectable()
export class CustomersStore extends ObservableStore<StateStoreModel> {
initialState = {
customers: [],
addMode: false,
};
constructor() {
super({
logStateChanges: true,
})
this.setState(this.initialState, 'INIT_STATE');
}
get(): Observable<StoreStateModel> {
const state = this.getState();
return of(state);
}
}
Let's pump the breaks for a sec and see where we're at.
We have a class property initialState that gets passed in the constructor to setState() along with the INIT_STATE action. Actions are more for the developers' convenience here and can be any arbitrary string.
We're calling super() since we're extending the ObservableStore class and passing it logStateChanges: true. This prints out the state in the console like so:
That's right. No additional libraries needed. No browser extensions needed. Just a property passed to super(). That's pretty darn super! (Sorry)
Next, we have the get() method that gets that state and wraps it up nice in an Observable.
Now we can inject the CustomersStore into our component and capture the state changes in an Observable class property:
app.component.ts
export class AppComponent {
state$ = this.customersStore.stateChanged;
constructor(private customersStore: CustomersStore) {}
So far, so good. But how do we handle getting data? There's a couple different approaches we can take. One would be to add our data service to our CustomersStore, keeping it as the single source of truth and our app nice and clean.
customer-store.ts
export interface StoreStateModel {
customers: CustomerModel[],
addMode: boolean,
}
@Injectable()
export class CustomersStore extends ObservableStore<StateStoreModel> {
initialState = {
customers: [],
addMode: false,
};
constructor(private dataService: DataService) {
super({
logStateChanges: true,
})
this.setState(this.initialState, 'INIT_STATE');
}
get(): Observable<StoreStateModel> {
const state = this.getState();
return of(state);
}
fetchData(): void {
this.fetchSub = this.dataService.getData()
.subscribe({
next: (customers: CustomerModel[]) => {
const updatedState = {
...this.initialState,
customers,
}
this.setState(updatedState, 'FETCHED_DATA');
},
error: response => // do stuff to handle error
});
}
}
Back in our AppComponent we can fetch it in the ngOnInit hook.
app.component.ts
export class AppComponent implements OnInit {
state$ = this.customersStore.stateChanged;
constructor(private customersStore: CustomersStore) {}
ngOnInit(): void {
this.getCustomerData();
}
getCustomerData(): void {
this.customersStore.fetchData();
}
I try to avoid putting logic directly in lifecycle hooks. I think it makes unit tests easier to write and the code easier to read.
Now our console output looks like this:
Great, but what about updating the state with new customers?
Let's take a look!
We should probably display our customers in its own component to keep our components nice and single-minded.
Let's update the AppComponent's html to pass the state to our new component:
app.component.html
<div *ngIf="state$ | async as state">
<app-customers [state]="state"></app-customers>
</div>
customers.component.ts
export class CustomersComponent implements OnChanges {
@Input() state: StoreStateModel = { customers: [], addMode: false };
customersData: CustomerModel[] = [];
customersStore: CustomersStore;
displayedColumns: string[] = ['id', 'firstName', 'lastName', 'phone', 'email', 'memberSince', 'delete'];
constructor(customersStore: CustomersStore) {
this.customersStore = customersStore;
}
ngOnChanges(changes: SimpleChanges): void {
const stateChanges = changes['state'];
if (stateChanges && stateChanges.currentValue) {
this.customersData = stateChanges.currentValue.customers;
}
}
Our customersStore was injected as a public property so we can directly call initAddMode() in the html.
customers.component.html
<!-- a table or something to display the data -->
<button
class="addBtn"
(click)="customersStore.initAddMode()"
[disabled]="state.addMode">
Add
</button>
customers-store.ts
initAddMode(): void {
const state = this.getState();
const updatedState = {
...state,
addMode: true
};
this.setState(updatedState, 'INIT_ADD_MODE');
}
Let's assume you have another component AddCustomerComponent and you have its selector in the CustomersComponent's html:
<div *ngIf="state.addMode">
<app-add-customer></app-add-customer>
</div>
AddCustomerComponent would look something like this:
add-customer.component.ts
export class AddCustomerComponent {
customersStore: CustomersStore;
addCustomerForm = this.fb.group({
firstName: ['', Validators.required],
lastName: ['', Validators.required],
phone: ['', Validators.required],
email: ['', Validators.required],
})
constructor(
private fb: FormBuilder,
customersStore: CustomersStore,
) {
this.customersStore = customersStore;
}
}
Like with the CustomersComponent we injected customersStore as a public property so we can directly call addCustomer() and resetView() in the html.
<!-- form stuff -->
<div class="button-group">
<button
type="submit"
[disabled]="!addCustomerForm.valid"
(click)="customersStore.addCustomer(addCustomerForm.value)">
Save
</button>
<button
type="button"
(click)="customersStore.resetView()">
Cancel
</button>
customers-store.ts
addCustomer(customer: CustomerModel): void {
const state = this.getState();
const ids = state.customers.map(c => c.id)
const newId = ids.length ? Math.max(...ids) + 1 : 1;
customer.id = newId;
customer.memberSince = new Date();
this.addSub = this.dataService.addCustomer(customer)
.subscribe({
next: (customers: CustomerModel[]) => {
const updatedState = {
customers: [ ...state.customers, customer ],
addMode: false,
};
this.setState(updatedState, 'ADDED_CUSTOMER');
},
error: response => // do stuff to handle error
});
}
resetView(): void {
const state = this.getState();
const updatedState = {
...state,
addMode: false,
};
this.setState(updatedState, 'RESET_VIEW');
}
By now you should be getting the idea. For a more in-depth example feel free to take a peek at that example project I mentioned earlier.
Bonus! Unit Tests
Unit testing Observable Store is pretty straight forward. After all, we're just dealing with RxJS and class methods.
Yes, I write my tests after I write my application code. Yes, I know it's better to write tests first and then write application code. I fully embrace unit tests (as should you as a professional) but I haven't quite made the mental switch to TDD. I'll get there, promise.
Let's start off with creating some mock data and a stub for our DataService:
customers-store.spec.ts
const mockData = require('../../test-data/customers.json');
const dataServiceStub = {
getData: () => of([]),
}
Next, the setup. Just like with any other service we'll want to inject this class into the TestBed:
customers-store.spec.ts
describe('CustomersStore', () => {
let customersStore: CustomersStore;
beforeEach(() => {
TestBed.configureTestingModule({
providers: [
CustomersStore,
{ provide: DataService, useValue: dataServiceStub }
],
});
customersStore = TestBed.inject(CustomersStore);
});
Now we test the class like we would any other component:
describe('addCustomer()', () => {
it('should set the ID to 1 if no customers are present', fakeAsync(() => {
customersStore.addCustomer(mockData[0]);
customersStore.get()
.subscribe({
next: data => {
expect(data.customers[0].id).toBe(1);
}
});
}));
it('should set the ID to 1 more than the last one', fakeAsync(() => {
customersStore.addCustomer(mockData[0]);
customersStore.addCustomer(mockData[1]);
customersStore.get()
.subscribe({
next: data => {
expect(data.customers[1].id).toBe(2);
}
});
}));
it('should set the "memberSince" property to today\'s date', fakeAsync(() => {
customersStore.addCustomer(mockData[0]);
customersStore.get()
.subscribe({
next: data => {
expect(data.customers[0].memberSince).toEqual(new Date());
}
});
}));
});
That's all folks! I'm curious to hear your thoughts or experiences with Observable Store. Let me know!
Top comments (4)
excellent explanation, this package is great. thanks.
Many thanks for this great article!
ng-specific
The article uses an Angular app as an example (which is why the post has an Angular tag) but the library featured is framework agnostic. It just relies on RxJS