Vercel/NextJS: How to access serverless functions from frontend during local development? - aws-lambda

My React/NextJS front end has a Button component that fetches data via a serverless function when the button is clicked. I want to test this functionality during local development with the Vercel dev/CLI tools. I am getting a 404 result when attempting to access my lambda functions. Here are the approximate steps that I've gone through so far:
Create package.json with a dev script:
...
"scripts": {
"dev": "yarn codegen && next --hostname=0.0.0.0 --port=3001",
}
...
Link to deployed vercel project
Create vercel.json to specify builds and routes:
...
"builds": [
{ "src": "*.html", "use": "#now/static" },
{ "src": "pages/api/*.py", "use": "#now/python" },
],
"routes": [
{ "src": "/api/validate", "dest": "/pages/api/validate.py" }
]
...
Create my test Lambda function (in python):
from http.server import BaseHTTPRequestHandler
from datetime import datetime
class handler(BaseHTTPRequestHandler):
def do_GET(self):
self.send_response(200)
self.send_header('Content-type', 'text/plain')
self.end_headers()
self.wfile.write(str(datetime.now().strftime('%Y-%m-%d %H:%M:%S')).encode())
return
Create my Button component:
...
<Button
variant="contained"
color="secondary"
onClick={() => {
fetch('api/validate')
.then(response => { console.log(response)
response.json() })
.then(data => console.log(data))
}}
>
Generate sample dataset
</Button>
...
Run vercel dev
Access website at localhost:3001 (next dev server address)
Click button
Result:
I'm receiving a 404 response
Note: I can access the lambda function from localhost:3000/pages/api/validate.py (vercel dev server address). This appears to manually kickstart the lambda function build and serve process. I thought that it should have been built and served already from the vercel.json specification and be available at localhost:3001/api/validate. This seems to agree with the Vercel documentation.
Note 2: Next dev/CLI tools build and serve javascript/typescript files just fine. I'm using python and Go functions as well, which are supported by Vercel dev/CLI but not Next

My solution was to use vercel dev instead of next dev or yarn dev, and to use an environment variable in a .env file that points to the function url. This env variable should be prepended with NEXT_PUBLIC_ so that it is registered by next.js and passed to process.env during the build process.
# .env
NEXT_PUBLIC_FUNCTIONS_BASE_URL="http://localhost:3000" # 3000 is vercel port
# component.js
...
fetch(process.env.NEXT_PUBLIC_FUNCTIONS_BASE_URL + '/api/function-name')
...

You need to pass the port from vercel dev to the upstream CLI, in this case next dev.
{
"scripts": {
"dev": "yarn codegen && next dev --port=$PORT",
}
}
Now when you run vercel dev, the ephemeral port will be proxied from the dev server.
You can also remove vercel.json if you rename /pages/api to /api.

Related

Laravel Vite: Assets blocked/Mixed Content issues in production environment

I'm hosting my App on an EC2-instance behind an Elastic Load Balancer which manages my SSL-Certificate. On this EC2-Instance my nginx-configuration is redirecting all http-Requests to https.
I recently switched to Vite which caused me a lot of trouble. When I push my app to the server after calling npm run build my assets are blocked. In the browser console I get:
Mixed Content: The page at 'example.com' was loaded over HTTPS, but requested an insecure ...
My Setup:
vite.config.js
export default defineConfig({
server: {
host: 'localhost',
},
plugins: [
laravel([
'resources/assets/sass/app.sass',
// etc...
]),
vue({
template: {
transformAssetUrls: {
base: null,
includeAbsolute: false,
},
},
}),
],
});
Setting "https: true" in the server-block didn't help me.
.env
APP_ENV=production
APP_URL=https://example.com
ASSET_URL=https://example.com
In my blade template I'm using the Vite-directive:
#vite('resources/assets/sass/app.sass')
I tried the following solutions:
Setting $proxies = '*' in TrustProxies.php, which doesn't have any effect.
Setting URL::forceScheme('https'); in AppServiceProvider.php, which will load the assets but lead to a lot of other issues.
Somehow the #vite-directive is not resolving my assets as secure assets. With Laravel Mix I could just call secure_asset.
How can I fix this?
In the end I used the TrustedProxies-middleware. However back then I forgot to register it as global middleware.
import fs from 'fs';
const host = 'example.com';
server: {
host,
hmr: {host},
https: {
key: fs.readFileSync(`ssl-path`),
cert: fs.readFileSync(`ssl-cert-path`),
},
},
you should add this to vite.config.js file along with ASSET_URL to your .env file

Upgrade Prisma 1 to Prisma 2 with Apollo + GraphQL

I have a problem upgrading from Prisma 1 to Prisma 2.
The documentation is quite complicated for me.
I currently have a small project using :
"dependencies": {
"bcryptjs": "2.4.3",
"graphql-yoga": "1.18.3",
"jsonwebtoken": "8.5.1",
"prisma-binding": "1.5.19"
},
"devDependencies": {
"dotenv": "5.0.1",
"graphql-cli": "2.17.0",
"nodemon": "1.19.4",
"npm-run-all": "4.1.5",
"prisma": "^1.34.10"
}
My prisma.yml :
endpoint: ${env:PRISMA_ENDPOINT}
secret: ${env:PRISMA_SECRET}
datamodel: datamodel.graphql
hooks:
  post-deploy:
    - prisma generate
generate:
  - generator: graphql-schema
    output: ../src/generated/prisma.graphql
I used scripts :
"scripts": {
"start:dev": "nodemon -e js,graphql -x node -r dotenv/config src/index.js",
"start": "node src/index.js",
"debug": "nodemon -e js,graphql -x node --inspect -r dotenv/config src/index.js",
"playground": "graphql playground",
"dev": "npm-run-all --parallel start playground",
"deploy": "prisma1 deploy --env-file .env"
},
And this graphqlconfig
projects:
app:
schemaPath: "src/schema.graphql"
extensions:
endpoints:
default: "http://localhost:4000"
prisma:
schemaPath: "src/generated/prisma.graphql"
extensions:
prisma: database/prisma.yml
How can I update prisma?
Knowing that my frontend is based on VueJS with ApolloClient, graphlq, graphql-tag
Thanks a lot to you, the backend part is not something simple for me
And here is my tree structure
E D I T
Thanks for your anwser. Nice, #nburk
But I have a problem on the 3rd step : https://www.prisma.io/docs/guides/upgrade-guides/upgrade-from-prisma-1/upgrading-the-prisma-layer-postgres
(Connection URL)
Previously I don't used "docker-compose". I deployed my front + back + DB on Heroku with Prisma
const { Prisma } = require("prisma-binding");
const resolvers = require("./resolvers");
// GraphQL Yoga Server
const server = new GraphQLServer({
typeDefs: "src/schema.graphql",
resolvers,
context: (req) => ({
...req,
db: new Prisma({
typeDefs: "src/generated/prisma.graphql", // DB Prisma Schema
endpoint: process.env.PRISMA_ENDPOINT, // Prisma Service
secret: process.env.PRISMA_SECRET, // Prisma Secret
debug: true,
}),
}),
});
server.start(() =>
console.log(`Server is running on ${process.env.PRISMA_ENDPOINT}`)
);
With a DB hosted on Heroku (.env file)
PRISMA_ENDPOINT="https://lprojet-name-db.herokuapp.com/database/prod"
And when I used npx prisma introspect I have this error
Introspecting based on datasource defined in prisma/schema.prisma …
Error: P1001
Can't reach database server at `'localhosh':'5432'
Please make sure your database server is running at 'localhost':'5432'
I think the problem comes from the schema.prisma which requires to have a url starting with postgresql:// but with Prisma1 I didn't need to go through that.
How can I transform my old DB URL (currently hosted on Heroku)?
Thanks
EDIT 2
I used
DATABASE_URL=postgres://..........eu-west-1.compute.amazonaws.com:5432/d9ptc61fera9g1
And I have a "database empty" error, but my database isn't empty. This UR come from Heroku Database Config
Nikolas from Prisma here!
We have written extensive upgrade documentation that walks you through the upgrade process. Here's some guides you can follow:
How to upgrade: Provides a general overview and explains different upgrade strategies
Upgrading the Prisma layer: Explains how to adjust your database schema using the Prisma Upgrade CLI
prisma-binding to SDL-first: Explains how to upgrade your GraphQL schema and resolvers
Feel free to follow up in case you have any questions along the way, always happy to help :)

Getting svelte-material-ui working with snowpack and sass

I'm trying to get svelte material UI working with snowpack.
I have installed Snowpack and Snowpacks svelte template like so:
npm install --save-dev snowpack#next
npx create-snowpack-app xpapp --template #snowpack/app-template-svelte
This works, the sample svelte page shows up. Next I followed the Svelte Material UI instructions to "bundle this in your own code" as cited on the Usage chapter in the instructions here: https://github.com/hperrin/svelte-material-ui#usage
So I installed Sass and configured it in my snowpack.config.json file like this:
{
"extends": "#snowpack/app-scripts-svelte",
"scripts": {
"build:scss": "sass"
},
"devOptions": {},
"installOptions": {}
}
I followed the (very concise) instructions here: https://www.snowpack.dev/#sass
I've also added an empty src/theme/_smui-theme.scss file to my source files as the instructions say, and I installed the nessecary #smui components.
The problem is that I'm currently getting this error when starting the snowpack dev server:
> snowpack dev
Snowpack Dev Server (Beta)
NOTE: Still experimental, default behavior may change.
Starting up...
⠙ snowpack installing... #smui/icon-button, #smui/top-app-bar, svelte/internal
✘ /home/erik/Projects/svelte-xpapp/xpapp/node_modules/#smui/icon-button/_index.scss
Error: Unexpected character '#' (Note that you need plugins to import files that are not JavaScript)
at error (/home/erik/Projects/svelte-xpapp/xpapp/node_modules/snowpack/node_modules/rollup/dist/shared/rollup.js:161:30)
at Module.error (/home/erik/Projects/svelte-xpapp/xpapp/node_modules/snowpack/node_modules/rollup/dist/shared/rollup.js:15120:16)
at tryParse (/home/erik/Projects/svelte-xpapp/xpapp/node_modules/snowpack/node_modules/rollup/dist/shared/rollup.js:15009:23)
at Module.setSource (/home/erik/Projects/svelte-xpapp/xpapp/node_modules/snowpack/node_modules/rollup/dist/shared/rollup.js:15410:30)
at ModuleLoader.addModuleSource (/home/erik/Projects/svelte-xpapp/xpapp/node_modules/snowpack/node_modules/rollup/dist/shared/rollup.js:17460:20)
at async ModuleLoader.fetchModule (/home/erik/Projects/svelte-xpapp/xpapp/node_modules/snowpack/node_modules/rollup/dist/shared/rollup.js:17521:9)
at async /home/erik/Projects/svelte-xpapp/xpapp/node_modules/snowpack/node_modules/rollup/dist/shared/rollup.js:17491:36
at async Promise.all (index 0)
at async ModuleLoader.fetchModule (/home/erik/Projects/svelte-xpapp/xpapp/node_modules/snowpack/node_modules/rollup/dist/shared/rollup.js:17522:9)
at async Promise.all (index 0)
It seems that the #import statements in Material UI's _index.scss aren't recognized. I figure Snowpack should interpret/transpile .scss files, but it doesn't seem to be doing that.
So I came across the same problem using Svite as well as Snowpack. I was able to use the bare implementation:
// component.svelte <script>
import Button, { Label } from '#smui/button/bare'
import '#smui/button/bare.css'
That's all that's required with Svite.
With Snowpack, I needed to add rollup-plugin-svelte and update snowpack.config.js
// snowpack.config.js
module.exports = {
// ...
installOptions: {
rollup: { plugins: [require('rollup-plugin-svelte')()] }
},
// ...
}
I got it working with these install options:
installOptions: {
rollup: {
plugins: [
require("rollup-plugin-svelte")({
include: ["./node_modules"],
}),
require("rollup-plugin-postcss")({
use: [
[
"sass",
{
includePaths: ["./src/theme", "./node_modules"],
},
],
],
}),
],
},
},
Unfortunately, you'll have to run npx snowpack dev --reload for changes to the theme to take effect.
This won't extract css into .css files.
I also got an error message with the Dialog component during a production build.
Here is a full example: https://github.com/LeanderG/svelte-smui

Heroku Deployment Error: displaying backend, not frontend

I am trying to deploy my local MERN application to Heroku. Application works offline. After deployment, when click "Open app", all I see is the data from the backend. Not the front end. Deployment here: https://whispering-falls-45660.herokuapp.com/.
Set up new Heroku project, and after successful "git push Heroku master", application only shows backend data. Heroku CLI version: heroku/7.24.1, Node version: v10.13.0.
Github repo: https://github.com/neilhsieh/whereToEat
Package.json file has the proper scripts in place as per Brad Traversy:
"scripts": {
"start": "node server.js",
"server": "nodemon server.js",
"client": "npm start --prefix client",
"dev": "concurrently \"npm run server\" \"npm run client\"",
"heroku-postbuild": "NPM_CONFIG_PRODUCTION=false npm install --prefix client && npm run build --prefix client"
},
server.js code has appropriate code to point to client build file:
if (process.env.NODE_ENV === 'production') {
app.use(express.static('client/build'))
app.get('*', (req, res) => {
res.sendFile(path.resolve(__dirname, 'client', 'build', 'index.html')) // relative path
})
}
Expecting front end deployed on Heroku, ended up only having backend.
UPDATE: Moved process.env.NODE_ENV test to before all API calls, this fixed my issue.
First if you have a route like this in your server/app.js file
app.get("/", (req, res) => res.send("Api Running"));
Definitely get rid of that
Second add these lines in your server/app.js file
const path = require("path");
if (process.env.NODE_ENV === "production") {
app.use(express.static("client/build"));
app.get("*", (req, res) => {
res.sendFile(path.resolve(__dirname, "client", "build", "index.html"));
});
}

Laravel environment variable with React: is this a good practice?

I have a React app which makes API requests to a Laravel backend.
My app is hosted on Heroku (I do not know if it changes something for my question).
I would like to differentiate a production and a local environment for these requests. I do it as follows.
In my "welcome.blade.php", I add this meta tag:
<meta name="app_env" content="<?php echo env("APP_ENV") ?>" />
The APP_ENV contains either "production" or "local".
In my React app, I have this script:
export let urlApi = (document.querySelector("[name=app_env]").content === "production" ) ?
"https://laravel-react.herokuapp.com/api"
:
"http://localhost/laravel_react/public/api"
;
And I import this function in each component which needs it:
import { urlApi } from './../findUrlApi';
// .....
return fetch(`${urlApi}/products`,{
headers: {
"Content-Type": "application/json",
"Authorization": "Bearer "+localStorage.getItem("token")
}
})
It works fine.
But my question is, is it a good practice? (I am a beginner in React).
I don't think that it is a good practice. Environment variables (like APP_ENV and API URLs) should not reside in the source code.
However, you could store them like usual in Laravel .env file but by prefixing the key with MIX_. For example: MIX_API_URL=http://localhost. Every MIX_* variables in .env file will be exposed to your React application. Then, you could get MIX_API_URL value from your React application by calling process.env.MIX_API_URL.
Updated Laravel .env file
...
MIX_APP_ENV=production (or local) # Should be same as APP_ENV
MIX_API_LOCAL_URL=http://localhost/laravel_react/public/api
MIX_API_PRODUCTION_URL=https://laravel-react.herokuapp.com/api
In React components that need it
const { MIX_APP_ENV, MIX_API_LOCAL_URL, MIX_API_PRODUCTION_URL } = process.env;
const apiUrl = MIX_APP_ENV === 'local'? MIX_API_LOCAL_URL: MIX_API_PRODUCTION_URL;
return fetch(apiUrl + '/products', { ... });
If calling process.env.MIX_API_URL does not work and you are running npm run watch, try restarting npm run watch and hard reload your browser.
Reference
Laravel Documentation - Compiling Assets (Mix) - Environment
Variables

Resources