API REST using Google Cloud Functions (Serverless)
Serverless application has gained a lot of importance over time. It allows focussing on your app code/tests without worrying about configurations, deployment process or scalability.
We're going to create a function that will be exposed via rest URL. That function will be invoked every time a HTTP(S) request is received.
During execution, an express server will be summoned exposing our REST services.
What we are going to build?
- Express API for CRUD services(create, read, update and delete) on a Firestore database.
- Use Google Cloud Function to expose our Express server
- Deploy our Google Cloud Function using Cloud CLI.
Creating our firebase project
In order to create our first project, let's do it here. Select add project, project's name must be unique, let's use prefix github-ring-{github_user}
, github-ring-levinm in my case. Be sure to select Firestore as our database.
For creating our database, click on Develop>Database and select "start in test mode".
Initializing our project locally
We need to install firebase using NPM.
npm install -g firebase-tools
Then, let's login into our firebase account.
firebase login
........... input credentials
Initialize the project
firebase init
........ select project
It will prompt an interactive console.
- Select Functions and Hosting options.
- What language would you like to use to write Cloud Functions? TypeScript
- Do you want to use TSLint to catch probable bugs and enforce style? Yes
- Do you want to install dependencies with npm now? Yes
- What do you want to use as your public directory? Press enter to select public (it is the default option)
- Configure as a single-page app (rewrite all urls to /index.html)? No
We're ready, our firebase project was initialized.
Installing Express.js and dependencies
cd functions
npm install --save express body-parser
Creating our Google Cloud Function
Open src/index.ts
, it will be the entrypoint for our Express.js server
Import main libraries
import * as functions from 'firebase-functions';
import * as admin from 'firebase-admin';
import * as express from 'express';
import * as bodyParser from "body-parser";
Initialize firebase for accesing to its services
admin.initializeApp(functions.config().firebase);
Intialize Express.js server
const app = express();
const main = express();
Configure the server.
- Let's add the path used for receiving the request.
- Select JSON as our main parser for processing requests body.
main.use('/api/v1', app);
main.use(bodyParser.json());
Export our function.
Last but not least, let's define our Google Cloud Function name, we are going to expose it using export
. Our function will receive an express server object(this case main
) which will be used for request processing. If you want more information regarding how it works, you can check this good answer on Stackoverflow
export const webApi = functions.https.onRequest(main);
Creating our first service
Let's expose a GET endpoint returning just a string.
app.get('/warm', (req, res) => {
res.send('Calentando para la pelea');
})
Our src/index.ts
file should look like this:
import * as functions from 'firebase-functions';
import * as admin from 'firebase-admin';
import * as express from 'express';
import * as bodyParser from "body-parser";
admin.initializeApp(functions.config().firebase);
const app = express();
const main = express();
main.use('/api/v1', app);
main.use(bodyParser.json());
export const webApi = functions.https.onRequest(main);
app.get('/warmup', (request, response) => {
response.send('Warming up friend.');
})
Deploying our function.
Before deploying it, we need to change our config file firebase.json
as follows:
{ "functions": { "predeploy": [ "npm --prefix \"$RESOURCE_DIR\" run lint", "npm --prefix \"$RESOURCE_DIR\" run build" ] }, "hosting": { "public": "public", "ignore": [ "firebase.json", "**/.*", "**/node_modules/**" ], "rewrites": [ { "source": "/api/v1/**", "function": "webApi" } ] } }
This rule allows "to route" all requests sent through api/v1
to be served by webApi
function (Our exported one).
Also, Google CLI installs Typescript v2 by default. So, we need to update our typescript version >=3.3.1
. You can do it in functions.package.json
.
"devDependencies": { "tslint": "~5.8.0", "typescript": "~3.3.1" },
Re-install dependencies.
cd functions
npm install
We are ready for deploying.
firebase deploy
.....
✔ Deploy complete!
Project Console: https://console.firebase.google.com/project/github-ring-levivm/overview
Hosting URL: https://github-ring-levivm.firebaseapp.com
If everything is ok, Hosting URL will be our Google Cloud Function endpoint.
Testing our function
Let's send a GET
request using CURL
$ curl -G "https://github-ring-levivm.firebaseapp.com/api/v1/warmup"
Warming up friend.
Rest API CRUD
Let's add our CRUD endpoints. We are going to manage fights
information.
Create a record
First, let's initialize our database. We open our src/index.ts
and add this after admin initialization
admin.initializeApp(functions.config().firebase);
const db = admin.firestore(); // Add this
In order to create a fight record, let's create POST /fights/
endpoint. Our fight record is going to have a winner
, loser
and title
.
app.post('/fights', async (request, response) => {
try {
const { winner, loser, title } = request.body;
const data = {
winner,
loser,
title
}
const fightRef = await db.collection('fights').add(data);
const fight = await fightRef.get();
response.json({
id: fightRef.id,
data: fight.data()
});
} catch(error){
response.status(500).send(error);
}
});
- We get our post data using
request.body
- We use
add()
method to add a new fight, if the collection doesn't exist (our case), it will create it automatically. - In order to get the actual record data, we must use
get()
over the ref. - Return a json using
response.json
.
Get a record
We create a GET /fights/:id
endpoint in order to fetch a fight by id.
app.get('/fights/:id', async (request, response) => {
try {
const fightId = request.params.id;
if (!fightId) throw new Error('Fight ID is required');
const fight = await db.collection('fights').doc(fightId).get();
if (!fight.exists){
throw new Error('Fight doesnt exist.')
}
response.json({
id: fight.id,
data: fight.data()
});
} catch(error){
response.status(500).send(error);
}
});
- We get the fight id using
request.params
. - We validate if the id is not blank.
- We get the fight and check if it exists.
- If fight doesn't exist we throw an error
- If fight exists, we return the data.
Get a record list
We create a GET /fights/
endpoint.
app.get('/fights', async (request, response) => {
try {
const fightQuerySnapshot = await db.collection('fights').get();
const fights = [];
fightQuerySnapshot.forEach(
(doc) => {
fights.push({
id: doc.id,
data: doc.data()
});
}
);
response.json(fights);
} catch(error){
response.status(500).send(error);
}
});
- We get a collection snapshot.
- We iterate over every document and push the data into an array.
- We return our fight list.
Update a record
We must create a PUT /fights/:id
endpoint in order to update a fight by id
.
app.put('/fights/:id', async (request, response) => {
try {
const fightId = request.params.id;
const title = request.body.title;
if (!fightId) throw new Error('id is blank');
if (!title) throw new Error('Title is required');
const data = {
title
};
const fightRef = await db.collection('fights')
.doc(fightId)
.set(data, { merge: true });
response.json({
id: fightId,
data
})
} catch(error){
response.status(500).send(error);
}
});
- We get request data.
- We validate the data
- We update a record using
set(data, merge: true)
. It means it is going to update only the fields passed on data parameter.
Deleting a record.
For deleting a fight, we need to add an endpoint DELETE /fights/:id
.
app.delete('/fights/:id', async (request, response) => {
try {
const fightId = request.params.id;
if (!fightId) throw new Error('id is blank');
await db.collection('fights')
.doc(fightId)
.delete();
response.json({
id: fightId,
})
} catch(error){
response.status(500).send(error);
}
});
- We get the fight id.
- We use
delete()
in order to delete a doc instance (Remember that firestore is database based on documents( "NoSQL" ))
Our src/index.ts
file should looks like this
import * as functions from 'firebase-functions';
import * as admin from 'firebase-admin';
import * as express from 'express';
import * as bodyParser from "body-parser";
admin.initializeApp(functions.config().firebase);
const db = admin.firestore(); // Add this
const app = express();
const main = express();
main.use('/api/v1', app);
main.use(bodyParser.json());
export const webApi = functions.https.onRequest(main);
app.get('/warmup', (request, response) => {
response.send('Warming up friend.');
});
app.post('/fights', async (request, response) => {
try {
const { winner, losser, title } = request.body;
const data = {
winner,
losser,
title
}
const fightRef = await db.collection('fights').add(data);
const fight = await fightRef.get();
response.json({
id: fightRef.id,
data: fight.data()
});
} catch(error){
response.status(500).send(error);
}
});
app.get('/fights/:id', async (request, response) => {
try {
const fightId = request.params.id;
if (!fightId) throw new Error('Fight ID is required');
const fight = await db.collection('fights').doc(fightId).get();
if (!fight.exists){
throw new Error('Fight doesnt exist.')
}
response.json({
id: fight.id,
data: fight.data()
});
} catch(error){
response.status(500).send(error);
}
});
app.get('/fights', async (request, response) => {
try {
const fightQuerySnapshot = await db.collection('fights').get();
const fights = [];
fightQuerySnapshot.forEach(
(doc) => {
fights.push({
id: doc.id,
data: doc.data()
});
}
);
response.json(fights);
} catch(error){
response.status(500).send(error);
}
});
app.put('/fights/:id', async (request, response) => {
try {
const fightId = request.params.id;
const title = request.body.title;
if (!fightId) throw new Error('id is blank');
if (!title) throw new Error('Title is required');
const data = {
title
};
const fightRef = await db.collection('fights')
.doc(fightId)
.set(data, { merge: true });
response.json({
id: fightId,
data
})
} catch(error){
response.status(500).send(error);
}
});
app.delete('/fights/:id', async (request, response) => {
try {
const fightId = request.params.id;
if (!fightId) throw new Error('id is blank');
await db.collection('fights')
.doc(fightId)
.delete();
response.json({
id: fightId,
})
} catch(error){
response.status(500).send(error);
}
});
Testing
We deploy our function.
firebase deploy
....
We test all our endpoints.
# Testing create fight (POST /fights)
$ curl -d '{"winner":"levi", "losser":"henry", "title": "fight1"}' -H "Content-Type: application/json" -X POST "https://github-ring-levivm.firebaseapp.com/api/v1/fights/"
> {"id":"zC9QORei07hklkKUB1Gl","data":{"title":"fight1","winner":"levi","losser":"henry"}
# Testing get a fight (GET /fight:id)
$ curl -G "https://github-ring-levivm.firebaseapp.com/api/v1/fights/zC9QORei07hklkKUB1wGl/"
>{"id":"zC9QORei07hklkKUB1Gl","data":{"winner":"levi","losser":"henry","title":"fight1"}}
# Testing get fights list (GET /fights/)
$ curl -G "https://github-ring-levivm.firebaseapp.com/api/v1/fights/"
> [{"id":"zC9QORei07hklkKUB1Gl","data":{"title":"fight1","winner":"levi","losser":"henry"}}]
# Testing update a fight (PUT /fights/:id)
$ curl -d '{"title": "new fight title"}' -H "Content-Type: application/json" -X PUT "https://github-ring-levivm.firebaseapp.com/api/v1/fights/zC9QORei07hklkKUB1Gl/"
> {"id":"zC9QORei07hklkKUB1Gl","data":{"title":"new fight title"}}
# Testing delete a fight (DELETE /fight/:id)
$ curl -X DELETE "https://github-ring-levivm.firebaseapp.com/api/v1/fights/zC9QORei07hklkKUB1Gl/"
> {"id":"zC9QORei07hklkKUB1Gl"}
And we are done, we have built our API Rest using Google Cloud Function (Serverless).
Note: You can check your database using Firestore interface within our Firebase console.
If this was helpful, share it :).
If you like my content or it was helpful, you can motivate me to write more content by buying me a coffee
Top comments (44)
Thank you Levi for this tutorial. It was very clear and easy to follow for the most part. I had an error come up from
firebase deploy
regardingfirebase-function-helper
but I see that you have since removed that import line.In my case, I did not use Typescript or TSLint, so I had to remove the
predeploy
steps infirebase.json
. Also theimports
inindex.js
had to berequire
in my case.The only thing I added was
cors
, so as to be able to call the endpoints from a front-end on a different domain, like now or Netlify, for example.Also, other (older) tutorials I followed had mentioned downloading service account settings as a json file and then importing that in the project. A bit of a hassle. Your solution is much cleaner in this regard and, as a result, more portable.
Thank you again for a very helpful article.
I'm glad it was useful to you.
Yep, I need to add
cors
so you can use the endpoints from another domain.Thanks for the feedback.
Hey there, I'm having a hard time getting this deployed. I've copied your index.ts file so I know the file is good, I just can't figure out why my code is breaking. I'm attaching a screenshot but not sure it's coming threw.
I can't see the screenshot.
Yeah.. not sure why it didn't work.. here is an url.. sidneyphillip.com/screenshot.png
This is about your TS lint configuration, it isn't nothing about the platform itself.
Check this answer: stackoverflow.com/questions/430642...
Let me know if you need anything else.
I have a dumb question :p
Why we have to build an api for firebase database and not directly use the firebase api to access our firestore database?
This is a simple tutorial, I tried to keep our endpoints simpler as possible but most cases you need to have authentication, you need to do extra processes before inserting a record, you need to create an extra task, handle permissions, roles, etc. This is a backend service, it isn’t about just accessing database.
If you have any other question, feel free to ask :)
i have a url-shortener without users logging in. I cannot write to firestore from the Vue client because the firestore rules need a user logged in. I can also not use the firebase-admin because the code runs on the client. So, now i am trying a cloud function as an API relay.
But you can authenticate using an anonymous login(it is available on firebase), so, you can login user using token generates from the client without having username or password.
HI Levi, Thanks for the tip, i will look into that!
Functions as a backend only works for the US region.
In other words, it's a no-go for people using firebase hosting, database, storage in other regions, as latency will be atrocious.
We had our entire backend rewritten as functions before we discovered this issue. Had to completely abandon Firebase functions because all our other Firebase assets are located in Europe. So we moved our backend elsewhere, unfortunately.
Just be aware guys: functions as an API for your frontend ONLY works for US region. We had 500ms latency for every request because of this.
Hey Man thanks for helping out everyone.I have read data from all documents in a collection but I cannot get into my subcollections of cloud firestore. I am trying to get result from subcollections but I always get []. I would be very helpful if you share piece of code.
Sample Collection:
Users(collection)->Id(Documents in collection)->ratedBy(Sub collection in document)-> detail(ratedby document)
Since Jan 2020, we need extra settings to make the API calls public once deployed.
gcloud functions add-iam-policy-binding webApi \
--member="allUsers" \
--role="roles/cloudfunctions.invoker"
Hi, Excellent tutorial. I using Windows 10, and I managed to get almost all the way through but after I run the firebase deploy, I am getting the following error
"debug] [2021-01-26T09:35:46.363Z] Error: EPERM: operation not permitted, scandir 'c:/Users/abc/AppData/Local/Application Data'
Any pointers please.
Great tutorial!
Should we have only one large Express app to handle all requests? Or should we divide it to multi-app by functionally?
Because of some limits
cloud.google.com/functions/quotas
Will you or have you created a tutorial that follows this for authentication and use of API Keys? I found this article EXTREMELY useful and would love a follow up!
Yep, I'm just writing it, I'm planning to release it next week.
Glad you found it useful.
Hi Levi - Did you ever get around to writing this next post on authentication? I'm looking for a tutorial on how to implement authentication via a firebase-hosted RESTful API, e.g.
The approach I'm familiar with is using JWTs with Node, but I appreciate this may be different with Firebase Cloud Functions. I'm not sure.
Hi Levi, Thanks for the tutorial. Currently, the API endpoints work fine after they are deployed.
However, when running the emulator locally, the API endpoints seem not to be recognized.
URL: mylocal_server/v1/warm or mylocal_server/crew-iq/us-central1...
Browser Output: 'Cannot GET /v1/warm'
Do you know what might cause the issue or what would be the local API endpoint?
$ firebase emulators:start --only functions
i emulators: Starting emulators: functions
✔ functions: Using node@10 from host.
✔ functions: Emulator started at 0.0.0.0:8080
i functions: Watching "/home/ubuntu/environment/Crew-IQ/functions" for Cloud Functions...
⚠ functions: The Cloud Firestore emulator is not running, so calls to Firestore will affect production.
✔ functions[webApi]: http function initialized (0.0.0.0:8080/crew-iq/us-central1/w...).
✔ All emulators started, it is now safe to connect.