Anyone who knows me, knows I 😍😍 Firebase. It could be considered unhealthy 😂. Despite my love for it, I have had my fair share of disagreements with it. The biggest one that comes to mind relates to Firestore.
NOTE: This article assumes you have a basic knowledge of how Firestore works. (docs)
This article will contain:
- 🤔 The Problem - Something that annoyed me
- 💪 My Solution - A brief overview
- 🤩 LET'S BUILD! - Party time! 🎉🎉
🤔 The Problem
I tend to use Firestore as my go to NoSQL store. But when we pair it with AngularFire the examples shown are at times not perfect, especially when we try to adhere to the DRY Principle.
The examples all tend to start with your call to the collection
method in your Component
to ensure you are working with the correct collection in Firestore. But all these calls to collection
add up. There must be a better way?
💪 My Solution
To me, there is! But, this is subjective. I create a Generic Firebase CRUD Service*, that accepts a Type to define the model that I want to store in my Collection on Firestore.
This is what we are going to build in this article!
* I call this a service, but it is unlike a standard Angular Service that can be injected into a constructor
, rather it is simply an instantiated class.
🤩 LET'S BUILD!
Ok, before we begin, let me take a moment to state that when I do this in codebases I work on, I tend to use the Bridge Pattern, by setting up a base Implementation for the CRUD Service, then define a Concrete Implementation of this, specific to Firetore.
My Abstractions have reference to the base Implementation but use the Firestore Concrete Implementation.
If any of this seems confusing, I highly recommend you read the Bridge Pattern article linked!
We'll break this build down into a few steps:
- Setup - Setting up the class!
- Create - The code to add the Document (henceforth called the Entity)
- Read - The code to read one or many Entities in the Collection
- Update - The code to update the Entity
- Delete - The code to delete the Entity
- Let's use it!
Let's get started!
🔧 Setup
We will assume you have an existing Angular project with AngularFire installed that you can work in.
If not, follow the instructions from the AngularFire docs.
First, we need to setup the class that will hold our logic.
import { AngularFirestore, AngularFirestoreCollection } from '@angular/fire/firestore';
// We need a function that will turn our JS Objects into an Object
// that Firestore can work with
function firebaseSerialize<T>(object: T) {
return JSON.parse(JSON.stringify(object));
}
// We need a base Entity interface that our models will extend
export interface Entity {
id?: string; // Optional for new Entities
}
export class FirestoreCrudService<T extends Entity> {
// Reference to the Collection in Firestore
private collection: AngularFirestoreCollection<T>;
/* We need to ask for the AngularFirestore Injectable
* and a Collection Name to use in Firestore
*/
constructor(private afs: AngularFirestore, collectionName: string) {
// We then create the reference to this Collection
this.collection = this.afs.collection(collectionName);
}
}
NOTE: If your collection does not exist on Firebase, don't worry, this will create it for you when you add your first Document to the Collection
Now that the setup is done, lets move on!
➕ Create - Time to Add
We now need to define our first method that will allow us to add Entities into our Collection.
/**
* We look for the Entity we want to add as well
* as an Optional Id, which will allow us to set
* the Entity into a specific Document in the Collection
*/
add(entity: T, id?: string): Promise<T> {
// We want to create a Typed Return of the added Entity
return new Promise<T>((resolve, reject) => {
if(id) {
// If there is an ID Provided, lets specifically set the Document
this.collection
.doc(id)
.set(firebaseSerialize(entity))
.then(ref => {
resolve(entity);
});
} else {
// If no ID is set, allow Firestore to Auto-Generate one
this.collection.add(firebaseSerialize(entity)).then(ref => {
// Let's make sure we return the newly added ID with Model
const newentity = {
id: ref.id,
...entity
};
resolve(newentity);
})
}
})
}
What's going on here? 🤔
We set up a reusable method that will allow us to Add an Entity to the pre-defined Collection. We want to ensure the returned Promise
is of the correct Entity Type so that our app will not break.
There is a use-case to add the Entity to a specific ID for scenarios such as adding a User
to a Users
Collection where the ID of the User comes from an external system.
📚 Read - Let's get Entities
Reading from the Collection comes in two forms. Get one specific Entity, or all the Entities in the Collection. We will define both below.
They will open an Observable
Stream which will allow our App to remain up to date with the Hosted Collection, wherein any change to the Hosted Collection will be piped down into your App via this Stream. (REAL-TIME BABY 🚀🚀)
// Our get method will fetch a single Entity by it's Document ID
get(id: string): Observable<T> {
return this.collection
.doc<T>(id)
.snapshotChanges()
.pipe(
// We want to map the document into a Typed JS Object
map(doc => {
// Only if the entity exists should we build an object out of it
if (doc.payload.exists) {
const data = doc.payload.data() as T;
const payloadId = doc.payload.id;
return { id: payloadId, ...data };
}
})
);
}
// Our list method will get all the Entities in the Collection
list(): Observable<T[]> {
return this.collection.snapshotChanges().pipe(
// Again we want to build a Typed JS Object from the Document
map(changes => {
return changes.map(a => {
const data = a.payload.doc.data() as T;
data.id = a.payload.doc.id;
return data;
});
})
);
}
I feel like the code above is pretty self-explanatory. We will discuss usage of these methods after we complete this class.
☁️ Update - We modified some data, let's save it
We also need the ability to modify existing Entities in our Collection, so this little method will handle that for us!
// Our Update method takes the full updated Entity
// Including it's ID property which it will use to find the
// Document. This is a Hard Update.
update(entity: T): Promise<T> {
return new Promise<T>((resolve, reject) => {
this.collection
.doc<T>(entity.id as string)
.set(firebaseSerialize(entity))
.then(() => {
resolve({
...entity
});
});
});
}
Pretty straightfoward, right? One method left, then we will show the full class!
🗑️ Delete - We don't like this Entity, let's dump it!
Finally, our Delete method will remove the Entity at a specific ID:
delete(id: string): Promise<void> {
return new Promise<void>((resolve, reject) => {
this.collection
.doc<T>(id)
.delete()
.then(() => {
resolve();
});
});
}
Ok, here is the completed class:
import { AngularFirestore, AngularFirestoreCollection } from '@angular/fire/firestore';
import { Observable } from 'rxjs';
import { take, map } from 'rxjs/operators';
// We need a function that will turn our JS Objects into an Object
// that Firestore can work with
function firebaseSerialize<T>(object: T) {
return JSON.parse(JSON.stringify(object));
}
// We need a base Entity interface that our models will extend
export interface Entity {
id?: string; // Optional for new entities
}
export class FirestoreCrudService<T extends Entity> {
// Reference to the Collection in Firestore
private collection: AngularFirestoreCollection<T>;
/* We need to ask for the AngularFirestore Injectable
* and a Collection Name to use in Firestore
*/
constructor(private afs: AngularFirestore, collectionName: string) {
// We then create the reference to this Collection
this.collection = this.afs.collection(collectionName);
}
/**
* We look for the Entity we want to add as well
* as an Optional Id, which will allow us to set
* the Entity into a specific Document in the Collection
*/
add(entity: T, id?: string): Promise<T> {
// We want to create a Typed Return of the added Entity
return new Promise<T>((resolve, reject) => {
if (id) {
// If there is an ID Provided, lets specifically set the Document
this.collection
.doc(id)
.set(firebaseSerialize(entity))
.then(ref => {
resolve(entity);
});
} else {
// If no ID is set, allow Firestore to Auto-Generate one
this.collection.add(firebaseSerialize(entity)).then(ref => {
// Let's make sure we return the newly added ID with Model
const newentity = {
id: ref.id,
...entity,
};
resolve(newentity);
});
}
});
}
/**
* Our get method will fetch a single Entity by it's Document ID
*/
get(id: string): Observable<T> {
return this.collection
.doc<T>(id)
.snapshotChanges()
.pipe(
// We want to map the document into a Typed JS Object
map(doc => {
// Only if the entity exists should we build an object out of it
if (doc.payload.exists) {
const data = doc.payload.data() as T;
const payloadId = doc.payload.id;
return { id: payloadId, ...data };
}
})
);
}
/*
* Our list method will get all the Entities in the Collection
*/
list(): Observable<T[]> {
return this.collection.snapshotChanges().pipe(
// Again we want to build a Typed JS Object from the Document
map(changes => {
return changes.map(a => {
const data = a.payload.doc.data() as T;
data.id = a.payload.doc.id;
return data;
});
})
);
}
/* Our Update method takes the full updated Entity
* Including it's ID property which it will use to find the
* Document. This is a Hard Update.
*/
update(entity: T): Promise<T> {
return new Promise<T>((resolve, reject) => {
this.collection
.doc<T>(entity.id as string)
.set(firebaseSerialize(entity))
.then(() => {
resolve({
...entity,
});
});
});
}
delete(id: string): Promise<void> {
return new Promise<void>((resolve, reject) => {
this.collection
.doc<T>(id)
.delete()
.then(() => {
resolve();
});
});
}
}
That's it, thats our Generic Class!
🔥 Let's use it!
Ok, now that we have created our generic class, let's take the Traditional Todo List example and recreate it with our new class.
Let's start with our Todo Model:
export interface Todo extends Entity {
todo: string;
category: string;
}
When we typically work with Entities in our code, we usually have a service that will handle specific logic relating to that Entity. We will also want this service to talk to our Firestore. We will use our newly created Crud Class for this.
So let's create a service:
@Injectable({
providedIn: 'root'
})
export class TodoService {
private crudService: FirestoreCrudService;
// AngularFirestore should be found by Angular DI System
constructor(private afs: AngularFirestore) {
// Let's create our CrusService and use the a Collection with the name 'todos'
this.crudService = new FirestoreCrudService<Todo>(afs, 'todos');
}
addTodo(todo: string, category: string) {
return this.crudService.add({todo, category});
}
updateTodoCategory(todo: Todo, category: string) {
return this.crudService.update({..todo, category});
}
deleteTodo(todo: Todo) {
return this.crudService.delete(todo.id);
}
getAllTodos() {
return this.crudService.list();
}
}
Hopefully, you can see from this service above how easy it will now be to create custom logic, but reuse one class to talk to our Firestore, for multiple different models!
Isn't that awesome! 🚀🚀🚀
Hopefully this has been educational in some form or other!
If you have any questions, feel free to ask below or reach out to me on Twitter: @FerryColum.
Top comments (9)
I'm not sure this is using the Bridge Pattern (as I understand it) - the
TodoService
is tightly coupled to theFirestoreCrudService
using a class reference not an interface, which couldn't really be substituted without changing the implementation of bothI think the use of JSON.parse(JSON.stringify(object)); might be very slow.
Kind regards,
Tailor
I've yet to see a performance hit with it in production environments running on older mobile devices :)
When it comes to Firebase apps, this will probably slow down your data fetching by hundreds of milliseconds, which, in many cases, is a lot for synchronized experiences developed with this type of database.
Perhaps it would be faster to spread the typed object into a new object:
{...entity}
What problem are you trying to solve with that? Why not just put the object directly into firebase?
You can't access data.id if you're using a generic type, isn't better to use a abstract class?
I missed a part of my total usage of this.
I create a base interface called Entity:
Then the Crud Type Generic should be
Then finally, all our models will extend the Entity interface
Thanks! This is a great idea!
Is there any reason why you manually create Promises though? I normally just return the promises returned by firebase's set/add/delete methods.