Having this node.js app which is going to be quite huge.
First I created a file named
user.account.test.js
In that I begun putting all the possible tests (positive and negative tests) for the usuale flow: signup, singin, activation, restore password etc.
At the end I have this file that is over 600 rows. Now, Im going to create a lot of more tests. And having everything in the same file sounds silly to me.
I could not really find resources that explain how to split the test in severals test files.
Im having a nightmare when I created a new test file where to put other tests. I mostly got timeout issues.
And a lot of things look strange. For example:
In the user.account.test.js I had this line:
beforeAll(async () => {
await mongoose.connect(process.env.MONGODB_TEST_URI);
});
In the second test file, named user.step2.test.js, I was unsure if I had to also put the same function.
At the end I did it, and incredibly that file did not know anything about "process.env.MONGODB_TEST_URI".
What is the best practice when you want to split tests into multiple files?
Ok, solution seems to be adding the flag --runInBand. Then they will run sequentially.
For me each routes I write one test file.
For example:
Each test file have sth like this:
import request from 'supertest'
import {app} from '../../app'
it('create new user', async() => {
return request(app).post('/api/users')
.send({account: "123", password:"123"})
.expect(201)
})
and I create a test setup file:
beforeAll(async () => {
// init your database connection here
})
beforeEach(() => {
// delete all data in your database
})
afterAll(async() => {
//close your db connection
})
and in package.json:
"jest": {
"preset": "ts-jest",
"testEnvironment": "node",
"setupFilesAfterEnv": [
"./src/test/setup.ts"
]
},
I am trying to set up a testbench for a more complex app. We have the issue that our app wants to call backend code which needs to be mocked in order to keep the testbench fast. Therefor we have two files per module:
connector.ts and connector.mocked.ts the first contains the "live" code, the second one contains some mocked/ dummy implementation. In our Modules we then import them by just doing
import {...} from './connector.ts'
which will be executed directly in our app, or in storybook will be rewritten to the connector.mocked.ts. In Storybook this happens in the global config:
module.exports = {
...
webpackFinal: (config) => {
...
config.resolve.alias = {
...config.resolve.alias,
'./connector': './connector.mocked',
'#': path.resolve(__dirname, '..'),
}
}
}
whats the equivalent in Jest for this? I don't want to write this for every test as the unit under test might not use any connector directly but might rely on some other module that uses their own connector.s (e.g. a Modal-View, which uses a Menu, which uses a function to check if the menu entry should be shown which calls the backend and is mocked)
i've found something called moduleNameMapper in jest, but I have no clue how to use it. is this the right way to go? how do I debug it?
My current Jest-Config:
module.exports = {
preset: 'ts-jest',
testEnvironment: 'jsdom',
testPathIgnorePatterns: ['jest-setup.spec.ts'],
moduleFileExtensions: ['ts', 'tsx', 'js', 'jsx'],
moduleNameMapper: {
'#/(.*)$': '<rootDir>/$1',
// '.\\/connector': './connector.mocked.ts', // does not work
},
setupFilesAfterEnv: ['<rootDir>/jest-setup.spec.ts'],
};
You need to swap the order of your keys in moduleNameMapper. So it becomes something like this:
moduleNameMapper: {
'#/(.*)connector(.ts)?$': '<rootDir>/$1connector.mock',
'#/(.*)$': '<rootDir>/$1'
}
Then in your screen you do
import { something } from '#/connector'
something() // live connector
but your spec files will import #/connector.mock and something() will do the mocked stuff.
Unable to identify what's happening in my next.js app. As fs is a default file system module of nodejs. It is giving the error of module not found.
If you use fs, be sure it's only within getInitialProps or getServerSideProps. (anything includes server-side rendering).
You may also need to create a next.config.js file with the following content to get the client bundle to build:
For webpack4
module.exports = {
webpack: (config, { isServer }) => {
// Fixes npm packages that depend on `fs` module
if (!isServer) {
config.node = {
fs: 'empty'
}
}
return config
}
}
For webpack5
module.exports = {
webpack5: true,
webpack: (config) => {
config.resolve.fallback = { fs: false };
return config;
},
};
Note: for other modules such as path, you can add multiple arguments such as
{
fs: false,
path: false
}
I spent hours on this and the solution is also here on Stackoverflow but on different issue -> https://stackoverflow.com/a/67478653/17562602
Hereby I asked for MOD permission to reshare this, since this issue is the first one to show up on Google and probably more and more people stumble would upon the same problem as I am, so I'll try to saved them some sweats
Soo, You need to add this in your next.config.js
module.exports = {
future: {
webpack5: true, // by default, if you customize webpack config, they switch back to version 4.
// Looks like backward compatibility approach.
},
webpack(config) {
config.resolve.fallback = {
...config.resolve.fallback, // if you miss it, all the other options in fallback, specified
// by next.js will be dropped. Doesn't make much sense, but how it is
fs: false, // the solution
};
return config;
},
};
It works for like a charm for me
Minimal reproducible example
A clean minimal example will be beneficial to Webpack beginners since auto splitting based on usage is so mind-blowingly magic.
Working hello world baseline:
pages/index.js
// Client + server code.
export default function IndexPage(props) {
return <div>{props.msg}</div>
}
// Server-only code.
export function getStaticProps() {
return { props: { msg: 'hello world' } }
}
package.json
{
"name": "test",
"version": "1.0.0",
"scripts": {
"dev": "next",
"build": "next build",
"start": "next start"
},
"dependencies": {
"next": "12.0.7",
"react": "17.0.2",
"react-dom": "17.0.2"
}
}
Run with:
npm install
npm run dev
Now let's add a dummy require('fs') to blow things up:
// Client + server code.
export default function IndexPage(props) {
return <div>{props.msg}</div>
}
// Server-only code.
const fs = require('fs')
export function getStaticProps() {
return { props: { msg: 'hello world' } }
}
fails with:
Module not found: Can't resolve 'fs'
which is not too surprising, since there was no way for Next.js to know that that fs was server only, and we wouldn't want it to just ignore random require errors, right? Next.js only knows that for getStaticProps because that's a hardcoded Next.js function name.
OK, so let's inform Next.js by using fs inside getStaticProps, the following works again:
// Client + server code.
export default function IndexPage(props) {
return <div>{props.msg}</div>
}
// Server-only code.
const fs = require('fs')
export function getStaticProps() {
fs
return { props: { msg: 'hello world' } }
}
Mind equals blown. So we understand that any mention of fs inside of the body of getStaticProps, even an useless one like the above, makes Next.js/Webpack understand that it is going to be server-only.
Things would work the same for getServerSideProps and getStaticPaths.
Higher order components (HOCs) have to be in their own files
Now, the way that we factor out IndexPage and getStaticProps across different but similar pages is to use HOCs, which are just functions that return other functions.
HOCs will normally be put outside of pages/ and then required from multiple locations, but when you are about to factor things out to generalize, you might be tempted to put them directly in the pages/ file temporarily, something like:
// Client + server code.
import Link from 'next/link'
export function makeIndexPage(isIndex) {
return (props) => {
return <>
<Link href={isIndex ? '/index' : '/notindex'}>
<a>{isIndex ? 'index' : 'notindex'}</a>
</Link>
<div>{props.fs}</div>
<div>{props.isBlue}</div>
</>
}
}
export default makeIndexPage(true)
// Server-only code.
const fs = require('fs')
export function makeGetStaticProps(isBlue) {
return () => {
return { props: {
fs: Object.keys(fs).join(' '),
isBlue,
} }
}
}
export const getStaticProps = makeGetStaticProps(true)
but if you do this you will be saddened to see:
Module not found: Can't resolve 'fs'
So we understand another thing: the fs usage has to be directly inside the getStaticProps function body, Webpack can't catch it in subfunctions.
The only way to solve this is to have a separate file for the backend-only stuff as in:
pages/index.js
// Client + server code.
import { makeIndexPage } from "../front"
export default makeIndexPage(true)
// Server-only code.
import { makeGetStaticProps } from "../back"
export const getStaticProps = makeGetStaticProps(true)
pages/notindex.js
// Client + server code.
import { makeIndexPage } from "../front"
export default makeIndexPage(false)
// Server-only code.
import { makeGetStaticProps } from "../back"
export const getStaticProps = makeGetStaticProps(false)
front.js
// Client + server code.
import Link from 'next/link'
export function makeIndexPage(isIndex) {
return (props) => {
console.error('page');
return <>
<Link href={isIndex ? '/notindex' : '/'}>
<a>{isIndex ? 'notindex' : 'index'}</a>
</Link>
<div>{props.fs}</div>
<div>{props.isBlue}</div>
</>
}
}
back.js
// Server-only code.
const fs = require('fs')
export function makeGetStaticProps(isBlue) {
return () => {
return { props: {
fs: Object.keys(fs).join(' '),
isBlue,
} }
}
}
Webpack must see that name makeGetStaticProps getting assigned to getStaticProps, so it decides that the entire back file is server-only.
Note that it does not work if you try to merge back.js and front.js into a single file, probably because when you do export default makeIndexPage(true) webpack necessarily tries to pull the entire front.js file into the frontend, which includes the fs, so it fails.
This leads to a natural (and basically almost mandatory) split of library files between:
front.js and front/*: front-end + backend files. These are safe for the frontend. And the backend can do whatever the frontend can do (we are doing SSR right?) so those are also usable from the backend.
Perhaps this is the idea behind the conventional "components" folder in many official examples. But that is a bad name, because that folder should not only contain components, but also any library non-component helpers/constants that will be used from the frontend.
back.js and back/* (or alternatively anything outside of front/*): backend only files. These can only be used by the backend, importing them on frontend will lead to the error
fs,path or other node native modules can be used only inside server-side code, like "getServerSide" functions. If you try to use it in client you get error even you just console.log it.. That console.log should run inside server-side functions as well.
When you import "fs" and use it in server-side, next.js is clever enough to see that you use it in server-side so it wont add that import into the client bundle
One of the packages that I used was giving me this error, I fixed this with
module.exports = {
webpack: (config, { isServer }) => {
if (!isServer) {
config.resolve.fallback.fs = false
}
return config
},
}
but this was throwing warning on terminal:
"Critical dependency: require function is used in a way in which
dependencies cannot be statically extracted"
Then I tried to load the node module on the browser. I copied the "min.js" of the node module from the node_modules and placed in "public/js/myPackage.js" and load it with Script
export default function BaseLayout({children}) {
return (
<>
<Script
// this in public folder
src="/js/myPackage.js"
// this means this script will be loaded first
strategy="beforeInteractive"
/>
</>
)
}
This package was attached to window object and in node_modules source code's index.js:
if (typeof window !== "undefined") {
window.TruffleContract = contract;
}
So I could access to this script as window.TruffleContract. BUt this was not an efficient way.
While this error requires a bit more reasoning than most errors you'll encounter, it happens for a straightforward reason.
Why this happens
Next.js, unlike many frameworks allows you to import server-only (Node.js APIs that don't work in a browser) code into your page files. When Next.js builds your project, it removes server only code from your client-side bundle by checking which code exists inside one any of the following built-in methods (code splitting):
getServerSideProps
getStaticProps
getStaticPaths
Side note: there is a demo app that visualizes how this works.
The Module not found: can't resolve 'xyz' error happens when you try to use server only code outside of these methods.
Error example 1 - basic
To reproduce this error, let's start with a working simple Next.js page file.
WORKING file
/** THIS FILE WORKS FINE! */
import type { GetServerSideProps } from "next";
import fs from "fs"; // our server-only import
type Props = {
doesFileExist: boolean;
};
export const getServerSideProps: GetServerSideProps = async () => {
const fileExists = fs.existsSync("/some-file");
return {
props: {
doesFileExist: fileExists,
},
};
};
const ExamplePage = ({ doesFileExist }: Props) => {
return <div>File exists?: {doesFileExist ? "Yes" : "No"}</div>;
};
export default ExamplePage;
Now, let's reproduce the error by moving our fs.existsSync method outside of getServerSideProps. The difference is subtle, but the code below will throw our dreaded Module not found error.
ERROR file
import type { GetServerSideProps } from "next";
import fs from "fs";
type Props = {
doesFileExist: boolean;
};
/** ERROR!! - Module not found: can't resolve 'fs' */
const fileExists = fs.existsSync("/some-file");
export const getServerSideProps: GetServerSideProps = async () => {
return {
props: {
doesFileExist: fileExists,
},
};
};
const ExamplePage = ({ doesFileExist }: Props) => {
return <div>File exists?: {doesFileExist ? "Yes" : "No"}</div>;
};
export default ExamplePage;
Error example 2 - realistic
The most common (and confusing) occurrence of this error happens when you are using modules that contain multiple types of code (client-side + server-side).
Let's say I have the following module called file-utils.ts:
import fs from 'fs'
// This code only works server-side
export function getFileExistence(filepath: string) {
return fs.existsSync(filepath)
}
// This code works fine on both the server AND the client
export function formatResult(fileExistsResult: boolean) {
return fileExistsResult ? 'Yes, file exists' : 'No, file does not exist'
}
In this module, we have one server-only method and one "shared" method that in theory should work client-side (but as we'll see, theory isn't perfect).
Now, let's try incorporating this into our Next.js page file.
/** ERROR!! */
import type { GetServerSideProps } from "next";
import { getFileExistence, formatResult } from './file-utils.ts'
type Props = {
doesFileExist: boolean;
};
export const getServerSideProps: GetServerSideProps = async () => {
return {
props: {
doesFileExist: getFileExistence('/some-file')
},
};
};
const ExamplePage = ({ doesFileExist }: Props) => {
// ERROR!!!
return <div>File exists?: {formatResult(doesFileExist)}</div>;
};
export default ExamplePage;
As you can see, we get an error here because when we attempt to use formatResult client-side, our module still has to import the server-side code.
To fix this, we need to split our modules up into two categories:
Server only
Shared code (client or server)
// file-utils.ts
import fs from 'fs'
// This code (and entire file) only works server-side
export function getFileExistence(filepath: string) {
return fs.existsSync(filepath)
}
// file-format-utils.ts
// This code works fine on both the server AND the client
export function formatResult(fileExistsResult: boolean) {
return fileExistsResult ? 'Yes, file exists' : 'No, file does not exist'
}
Now, we can create a WORKING page file:
/** WORKING! */
import type { GetServerSideProps } from "next";
import { getFileExistence } from './file-utils.ts' // server only
import { formatResult } from './file-format-utils.ts' // shared
type Props = {
doesFileExist: boolean;
};
export const getServerSideProps: GetServerSideProps = async () => {
return {
props: {
doesFileExist: getFileExistence('/some-file')
},
};
};
const ExamplePage = ({ doesFileExist }: Props) => {
return <div>File exists?: {formatResult(doesFileExist)}</div>;
};
export default ExamplePage;
Solutions
There are 2 ways to solve this:
The "correct" way
The "just get it working" way
The "Correct" way
The best way to solve this error is to make sure that you understand why it is happening (above) and make sure you are only using server-side code inside getStaticPaths, getStaticProps, or getServerSideProps and NOWHERE else.
And remember, if you import a module that contains both server-side and client-side code, you cannot use any of the imports from that module client-side (revisit example #2 above).
The "Just get it working" way
As others have suggested, you can alter your next.config.js to ignore certain modules at build-time. This means that when Next.js attempts to split your page file between server only and shared code, it will not try to polyfill Node.js APIs that fail to build client-side.
In this case, you just need:
/** next.config.js - with Webpack v5.x */
module.exports = {
... other settings ...
webpack: (config, { isServer }) => {
// If client-side, don't polyfill `fs`
if (!isServer) {
config.resolve.fallback = {
fs: false,
};
}
return config;
},
};
Drawbacks of this approach
As shown in the resolve.fallback section of the Webpack documentation, the primary reason for this config option is because as-of Webpack v5.x, core Node.js modules are no longer polyfilled by default. Therefore, the main purpose for this option is to provide a way for you to define which polyfill you want to use.
When you pass false as an option, this means, "do not include a polyfill".
While this works, it can be fragile and require ongoing maintenance to include any new modules that you introduce to your project. Unless you are converting an existing project / supporting legacy code, it is best to go for option #1 above as it promotes better module organization according to how Next.js actually splits the code under the hood.
If trying to use fs-extra in Next.js, this worked for me
module.exports = {
webpack: (config) => {
config.resolve.fallback = { fs: false, path: false, stream: false, constants: false };
return config;
}
}
I got this error in my NextJS app because I was missing export in
export function getStaticProps()
/** #type {import('next').NextConfig} */
module.exports = {
reactStrictMode: false,
webpack5: true,
webpack: (config) => {
config.resolve.fallback = {
fs: false,
net: false,
dns: false,
child_process: false,
tls: false,
};
return config;
},
};
This code fixed my problem and I want to share.Add this code to your next.config file.i'm using
webpack5
For me clearing the cache
npm cache clean -f
and then updating the node version to the latest stable release(14.17.0) worked
It might be that the module you are trying to implement is not supposed to run in a browser. I.e. it's server-side only.
For me, the problem was the old version of the node.js installed. It requires node.js version 14 and higher. The solution was to go to the node.js web page, download the latest version and just install it. And then re-run the project. All worked!
I had the same issue when I was trying to use babel.
For me this worked:
#add a .babelrc file to the root of the project and define presets and plugins
(in my case, I had some issues with the macros of babel, so I defined them)
{
"presets": ["next/babel"],
"plugins": ["macros"]
}
after that shut down your server and run it again
I had this exact issue. My problem was that I was importing types that I had declared in a types.d.ts file.
I was importing it like this, thanks to the autofill provided by VSCode.
import {CUSTOM_TYPE} from './types'
It should have been like this:
import {CUSTOM_TYPE} from './types.d'
In my case, I think the .d was unnecessary so I ended up removing it entirely and renamed my file to types.ts.
Weird enough, it was being imported directly into index.tsx without issues, but any helper files/functions inside the src directory would give me errors.
I ran into this in a NextJS application because I had defined a new helper function directly below getServerSideProps(), but had not yet called that function inside getServerSideProps().
I'm not sure why this created a problem, but it did. I could only get it to work by either calling that function, removing it, or commenting it out.
Don't use fs in the pages directory, since next.js suppose that files in pages directory are running in browser environment.
You could put the util file which uses fs to other directory such as /core
Then require the util in getStaticProps which runs in node.js environment.
// /pages/myPage/index.tsx
import View from './view';
export default View;
export async function getStaticProps() {
const util = require('core/some-util-uses-fs').default; // getStaticProps runs in nodes
const data = await util.getDataFromDisk();
return {
props: {
data,
},
};
}
In my case, this error appeared while refactoring the auth flow of a Next.js page. The cause was some an unused imports that I had not yet removed.
Previously I made the page a protected route like so:
export async function getServerSideProps ({ query, req, res }) {
const session = await unstable_getServerSession(req, res, authOptions)
if (!session) {
return {
redirect: {
destination: '/signin',
permanent: false,
},
}
}
//... rest of server-side logic
}
Whilst refactoring, I read up on NextAuth useSession. Based on what I read there, I was able to change the implementation such that I simply needed to add
MyComponent.auth = true to make a page protected. I then deleted the aforementioned code block inside of getServerSideProps. However, I had not yet deleted the two imports used by said code block:
import { unstable_getServerSession } from 'next-auth/next'
import { authOptions } from 'pages/api/auth/[...nextauth]'
I believe the second of those two imports was causing the problem. So the summary is that in addition to all of the great answers above, it could also be an unused import.
Sometimes this error can be because you have imported something but not mastered it anywhere. This worked for me. I reviewed my code and removed the unused dependencies.
I thought I'd post this as I stumbled around for a while before noticing what's going on. I have a test suite that uses CouchDB as its logging / recording database. I discovered you can write custom reporters in intern, so thought I could move a lot of my manual 'recordSuccess()'/'recordFailure()' calls out of my test script, and into a custom reporter responding to test pass and fail events.
My main test script still wants to do a little couchdb interaction, so I factored out the couchdb connection and reporting functions into a module, then tried to use that module from both the main test script, and the custom reporter module.
I find that the couchdb helper module is instantiated twice. This goes against the expectation that AMD/RequireJS require() will only execute a module once, and cache the result for use the next time the module is required. If I put a 'debugger' statement in its main body of code, it is clearly executed twice. The upshot, for me, is that the couchdb reference is undefined when called from the reporter.
Directory structure:
runTest.js # helper script to run intern test from this dir
src/MainTest.js
src/CouchHelper.js
src/CouchDBReporter.js
src/intern.js # intern config
runTest.js
node node_modules/.bin/intern-client config=src/intern suites=mypackage/WINTest --envConfig=src/test/dev.json
i.e. MainTest.js:
define([ 'CouchHelper' ], function (CouchHelper) {
.. test startup ..
CouchHelper.connect(username, password, etc);
CouchDBReporter.js:
define([ 'CouchHelper' ], function (CouchHelper) {
return {
'/test/fail': function (test) {
// Presume the couchdb is connected at this point
CouchHelper.recordFailure(test);
}
}
intern.js:
... blah blah ..
loader: {
// Packages that should be registered with the loader in each testing environment
packages: [
'node',
'nedb',
'nodemailer',
{ 'mypackage', 'src' }
],
reporters: [ 'console', 'src/CouchDBReporter' ]
CouchHelper.js:
define([
'intern/dojo/node!node-couchdb'
], function (Couchdb) {
debugger; // this is hit twice
var instance = 0;
function CouchHelper() {
this.couchdb = undefined;
this.instance = instance++;
console.log('Created instance ' + this.instance);
}
CouchHelper.prototype = {
connect: function () { this.couchdb = Couchdb.connect(blah); },
recordFailure: function (test) { this.couchdb.insert(blah); }
}
}
On startup, the console logs:
Created instance 0
Created instance 0
When the reporter calls recordFailure, it calls into a different instance of CouchHelper than the MainTest.js file called connect() on .. so this.couchdb is undefined, and the script crashes. I can call recordSuccess/recordFailure from in MainTest.js just fine, and this.couchdb is valid in CouchHelper, but from the CouchDBReporter the CouchHelper instance is clearly different.
Is this behaviour expected, and if so, what's the recommended way to share data and resources between the main test code, and code in a custom reporter? I see that in 3.0 the reporters config can take an object which might help mitigate this problem, but it feels like one would have to instantiate the reporter programatically rather than define it in config.
Nick
As suggested by Colin, the path to the answer lay in my loader map configuration. This means that my intern.js file, referenced as config on the command line, has a loader section where one can define the mappings of paths to AMD module (see https://theintern.github.io/intern/#option-loader). Typically I just define a list of package names, for example I know my test requires nedb, nodemailer, and my own src package:
loader: {
packages: [ 'node', 'nedb', 'nodemailer', 'src' ]
}
For some reason, I had defined my src package as being available by the name mypackage:
loader: {
packages: [ 'node', 'nedb', 'nodemailer',
{ name: 'mypackage', location: 'src' }
]
}
I had no good reason to do this. I then specified my custom reporter be loaded by intern using the 'src' package name:
intern.js:
reporters: [ 'console', 'src/CouchDBReporter' ]
And, here's the tricky bit, I referenced my helper module, CouchHelper, in two different ways, but both times by using a relative module path ./CouchHelper:
MainTest.js:
require([
'./CouchHelper',
...
], ...
CouchDBReporter.js:
require([
'./CouchHelper',
...
], ...
And on the command line, you guessed it, specified the test to be run as mypackage/MainTest.js. This conflicts with my specification of src/CouchDBReporter in intern.js's reporter section.
The result was that mypackage/MainTest.js required ./CouchHelper which resolved as mypackage/CouchHelper, and src/CouchDBReporter required ./CouchHelper, which resolved as src/CouchHelper. This loaded the CouchHelper module code twice, working around the usual guarantee with an AMD style loader that a module is only ever loaded once.
It has certainly been a good lesson in AMD module paths, and one implication of using relative paths.