Simple es6-friendly mocking library inspired by the best libraries:
- mockery - the library I like the most.
- mock-require - the simplest thing ever.
- proxyquire - the one I know better than others.
By its nature rewiremock has same behavior as Mockery. But it can behave like others too. It covers any cases. It is the right way to mock your dependencies or perform dependency injection.
Rewiremock is an evolution of lessons I learned from: the better proxyquire, the way of resolveQuire, and magic of proxyquire-webpack-alias.
- be right, and enable
true
testing experience. - be simple, and ease to use.
- be modular, to cover all cases.
- be secure, and isolate target under test.
- be fast, to be faster.
- give ability to mock everything - CommonJS, ES6, Webpack, anything.
- give ability to do correctly - isolation, typechecks, powerfull API
I have wrote some articles about these ideas - https://medium.com/tag/rewiremock/latest
- rewiremock.enable() - wipes cache and enables interceptor.
- rewiremock.disable() - wipes cache and disables interceptor.
- rewiremock.around(loader, creator):Promise< T > - loads a module in an asynchronous sandbox.
- rewiremock.proxy(file, stubs):T - proxyquire like mocking api, where file is file name, and stubs are an object or a function.
- rewiremock.module(loader, stubs):Promise< T > - async version of proxy, where loader is a function.
- rewiremock(moduleName: string) - fabric for a moduleNames's mock
- rewiremock(moduleImport: loader) - async fabric for a module import function.
- .enable/disable() - to enable or disable mock (enabled by default).
- .with(stubs: function | Object) - overloads module with a value
- .withDefault(stub: function | Object) - overload
default
es6 export - .by(otherModule: string| function) - overload by another module(if string provider) or by result of a function call.
- .callThrough() - first load original module, and next extend it by provided stub.
- .toBeUsed() - enables usage checking.
- .directChildOnly - will do mock only direct dependencies.
- .calledFromMock - will do mock only dependencies of mocked dependencies.
- rewiremock(moduleName: string|loader) - returns existing mock
- rewiremock.isolation() - enables isolation
- rewiremock.withoutIsolation() - disables isolation
- rewiremock.passBy(pattern or function) - enables some modules to pass thought isolation.
- rewiremock.inScope(callback) - place synchronous callback inside a sandbox.
Yep - there is 4 top level ways to activate a mock - inScope, around, proxy or just enable.
Which one to choose? Any! It just depends:
- If everything is simply - use rewiremock.proxy.
- If you have issues with name resolve - use rewiremock.module and resolve names by yourself.
- If you need scope isolation - use rewiremock.around, or inScope.
- If you advanced syntax and type checking - use rewiremock.around.
- If you need full control - you will always have it.
- You always can just use .enable/.disable.
#Usage
// 1. proxy will load a file by it's own ( keep in mind - name resolution is a complex thing)
const mock = rewiremock.proxy('somemodule', (r) => ({
'dep1': { name: 'override' },
'dep2': r.with({name: 'override' }).toBeUsed().directChildOnly() // use all `mocking API`
}));
// 2. you can require a file by yourself. ( yep, proxy is a god function)
const mock = rewiremock.proxy(() => require('somemodule'), {
'dep1': { name: 'override' },
'dep2': { onlyDump: 'stubs' }
}));
// 3. or use es6 import (not for node.js mjs `real` es6 modules)
// PS: module is an async version of proxy, so you can use imports
const mock = await rewiremock.module(() => import('somemodule'), {
'dep1': { name: 'override' },
'dep2': { onlyDump: 'stubs' }
}));
// 3. another version of .module, where you can do just ~anything~.
const mock = await rewiremock.around(() => import('somemodule'), () => {
rewiremock('dep1').with('something');
callMom();
}));
// 4. Low level API
rewiremock('someThing').with('someThingElse')
rewiremock.enable();
rewiremock.disable();
Rewiremock is able to provide a type-safe mocks. To enable type-safety follow these steps:
- Use TypeScript or Flow.
- Use dynamic import syntax.
- Use rewiremock.around or rewiremock.module to perform a mock.
- Use async form of rewiremock mock declaration.
// @flow
import rewiremock from 'rewiremock';
rewiremock.around(
() => import('./a.js'),
mock => {
mock(() => import('./b.js'))
.withDefault(() => "4")
.with({testB: () => 10})
.nonStrict() // turn off type system
.with({ absolutely: "anything" })
}
);
If default export is not exists on module 'b', or there is no named export testB, or types do not match - type system will throw.
If you will declare an async mock, you it will not be resolved by the time of execution - Rewiremock will throw on Error.
If you have async imports inside mocked file, follow this syntax
rewiremock.around(
() => import('./a.js'),
mock => {
// just before loader function rewiremock enabled itself
mock(() => import('./b.js').then(mock=>mock)) // mocks `live` one `tick` more
// just after loader function resolved rewiremock disables itself
mock => {
....
}
}
);
Rewiremock can check mock against the real implementation. This does not perform type checking
, but
could check exported names and exported types (function vs number, for example).
Rewiremock expects that mock will be less or equal than original file.
rewiremock: mocked export "somethingMissing" does not exists in ./b.js
rewiremock: exported type mismatch: ./b.js:default. Expected function, got number
To activate exports comparison
rewiremock('somemoduname')
.toMatchOrigin(); // to activate
// or
import rewiremock, { addPlugin, removePlugin, plugins } from 'rewiremock';
addPlugin(plugins.alwaysMatchOrigin);
Just use it. You can also activate nodejs, which will double check all modules names on a real FS, but.. everything might work out of the box.
PS: Just use usedByDefault to ensure module names are resolved correctly.
Rewiremock can emulate
few webpack features(like aliases) in node.js environment, but it also can be run inside webpack.
Actually rewiremock is the first client side mocking library
But not so fast, hanny. First you have to have 3(!) Plugins enabled.
- webpack.NamedModulesPlugin(). To enlight the real names of modules. Not "numbers".
- webpack.HotModuleReplacementPlugin(). To provide some information about connections between modules.
- rewiremock.webpackPlugin. To add some magic and make gears rolling.
plugins: [
new webpack.NamedModulesPlugin(),
new webpack.HotModuleReplacementPlugin(),
new (require("rewiremock/webpack/plugin"))()
]
That's all. Now all magic will happens at client side.
It is better to use .proxy/module command with direct require/import and leave all names conversion to webpack.
First - define your mocks. You can do it in any place, this is just a setup.
import rewiremock from 'rewiremock';
...
// totaly mock `fs` with your stub
rewiremock('fs')
.with({
readFile: yourFunction
});
// replace path, by other module
rewiremock('path')
.by('path-mock');
// replace enzyme by preconfigured one (from https://medium.com/airbnb-engineering/unlocking-test-performance-migrating-from-mocha-to-jest-2796c508ec50)
rewiremock('enzyme')
.by(({requireActual}) => {
// see rest of possible params in d.ts file
const enzyme = requireActual('enzyme');
if (!mockSetup) {
const chai = requireActual('chai');
const chaiEnzyme = requireActual('chai-enzyme');
chai.use(chaiEnzyme());
}
return enzyme;
});
// replace default export of ES6 module
rewiremock('reactComponent')
.withDefault(MockedComponent)
// replace only part of some library and keep the rest
rewiremock('someLibrary')
.callThrough()
.with({
onlyOneMethod
})
// secure yourself and from 'unexpected' mocks
rewiremock('myDep')
.with(mockedDep)
.calledFromMock()
There is a simply way to do it: Just enable it, and dont forget to disable it later.
//in mocha tests
beforeEach( () => rewiremock.enable() );
//...
// here you will get some advantage in type casting and autocompleting.
// it will actually works...
const someModule = require('someModule');
//...
afterEach( () => rewiremock.disable() );
Once enabled, rewiremock will wipe all mocked modules from cache, and all modules which require them.
Including your test.
Once disabled it will restore everything.
All unrelated to test dependencies will be kept. Node modules, react, common files - everything.
As result - it will run faster.
Sometimes you will have independent tests in a single file, and you might need separate mocks for each one.
inScope
execute callback inside sandbox, and all mocks or plugins or anything else you have added will not leaks away.
rewiremock.inScope( () => {
rewiremock('something').with(something);
rewiremock.enable();
// is 'something' mocked? Yes
....
rewiremock.disable();
// is 'something' mocked? No
// is it still listed as mock? Yes
});
// is 'something' mocked or listed? No
And there is a bit harder way to do it - scope. inScope will create new internal scope, next you can add something new to it, and then it will be destroyed. It will also enable/disable rewiremock just in time.
This helps keep tests in isolation.
PS: scopes are nesting each other like javascript prototypes do.
rewiremock.around(
() => import('somemodule'), // load a module. Using import or require.
// just before it you can specify mocks or anything else
(mock) => {
addPlugin(nodePlugin);
mock('./lib/a/foo').with(() => 'aa');
mock('./lib/a/../b/bar').with(() => 'bb');
mock('./lib/a/../b/baz').with(() => 'cc');
}
) // at this point scope is dead
.then((mockedBaz) => {
expect(mockedBaz()).to.be.equal('aabbcc');
});
or just
rewiremock.around(() => import('somemodule')).then(mockedModule => doSomething)
or
rewiremock.around(
() => import('somemodule').then( mockedModule => doSomething),
(mock) => aPromise
);
Currently .inScope is the only API capable to handle es6(not node [m]js!) dynamic imports.
Sometimes it is much easier to combine all the things together.
// preferred way - crete stubs using a function, where R is mock creator
rewiremock.proxy('somemodule', (r) => ({
'dep1': { name: 'override' },
'dep2': r.with({name: 'override' }).toBeUsed().directChildOnly() // same powerfull syntax
}));
// straight way - just provide stubs.
rewiremock.proxy('somemodule', {
'dep1': { name: 'override' },
'dep2': { name: 'override' }
}));
By default - rewiremock has limited features. You can extend its behavior via plugins.
- relative. A bit simply, proxyquire-like behavior. Will override only first level deps, and will wipe a lot of modules from a cache.
- nodejs. Common support to
usual
node.js application. Will absolutize all paths. Will wipe cache very accurately. - webpack-alias. Enabled you to use webpack aliases as module names.
- childOnly. Only first level dependencies will be mocked.
- protectNodeModules. Ensures that any module from node_modules will not be wiped from a cache.
- toBeUsed. Adds feature. The only plugin enabled by default.
- disabledByDefault. All mocks will be disabled on create and at the end of each cycle.
- usedByDefault. All mocks to be used by fact (reverse isolation)
import rewiremock, { addPlugin, removePlugin, plugins } from 'rewiremock';
addPlugin(plugins.webpackAlias);
removePlugin(plugins.webpackAlias);
If you import rewiremock from other place, for example to add some defaults mocks - it will not gonna work. Each instance of rewiremock in independent. You have to pass your instance of rewiremock to build a library. PS: note, rewiremock did have nested API, but it were removed.
Unit testing requires all decencies to be mocked. All! To enable it, run
rewiremock.isolation();
//or
rewiremock.withoutIsolation();
Then active - rewiremock will throw error on require of any unknown module.
Unknown is module which is nor mocked, nor marked as pass-through.
To make few modules to be invisible
to rewiremock, run
rewiremock.passBy(/*pattern or function*/);
rewiremock.passBy(/common/);
rewiremock.passBy(/React/);
rewiremock.passBy(/node_modules/);
rewiremock.passBy((name) => name.indexOf('.node')>=0 )
Sometimes you have to be sure, that you mock is actually was called.
Isolation will protect you then you add new dependencies, .toBeUsed
protect you from removal.
Jest is a very popular testing framework, but it has one issue - is already contain mocking support.
To use rewiremock with Jest add to the beginning of your file
// better to disable automock
jest.disableAutomock();
// Jest breaks the rules, and you have to restore nesting of a modules.
rewiremock.overrideEntryPoint(module);
// There is no way to use overload by Jest require or requireActual.
// use the version provided by rewiremock.
require = rewiremock.requireActual;
!!! the last line here may disable Jest sandboxing. !!!
Also it will disable Jest transformation.
To be able continue use ES6/imports - you have to enforce Babel to be applied in the common
way.
describe('block of tests', () => {
// require babel-register in describe or it block.
// NOT! On top level. Jest sandboxing and isolation are still in action,
// and will reset all settings to default
require("babel-register");
})
PS: Jest will set BABEL_ENV to test
.
It is better just to use rewiremock.requireActual
, without overriding global require.
In most cases you have to:
- add plugin
- setup default passBy rules
- add some common mocks
- do something else.
And it is not a good idea to do it in every test you have.
It is better to have one setup file, which will do everything for you
- Part 1 - man in the middle
// this is your test file
// instead of importing original file - import your own one
// import rewiremock from 'rewiremock';
import rewiremock from 'test/rewiremock';
- Part 2 - create your own one
// this tests/rewiremock.js
import rewiremock, { addPlugin, overrideEntryPoint} from 'rewiremock';
// do anything you need
addPlugin(something);
rewiremock('somemodule').with(/*....*/);
// but don't forget to add some magic
overrideEntryPoint(module); // <-- set yourself as top module
// PS: rewiremock will wipe this module from cache to keep magic alive.
export default rewiremock;
- Part 3 - enjoy. You extract some common code into helper. And things become a lot easier.
Default cache policy follow these steps:
- Preparation:
- all files required from original test, while interceptor is active, will bypass cache.
- all files you indicate as mocks will be removed from the cache.
- all "soiled" files which rely on mocks - will also be removed from the cache.
- repeat .
- Finalization
- repeat all mocks, and possible "soiled" by mocks files.
- copy over the old cache.
- or restore the old cache complitely if
forceCacheClear
mode is set.
The last variant is default for proxyquire and mockery, also it is more sequre
from different side effects.
Regardless, default is the first variant - as a way faster, and secure enough.
As result of this mocking strategy you can mock any file at any level, while keeping another files cached.
If you don't want this - just add relative
plugin. It will allow mocking only for modules
required from module with parent equals entryPoint.
PS: module with parent equals entryPoint - any module you require from test(it is an entry point). required from that module - first level require. Simple.
Don't forget - you can write your own plugins. plugin is an object with fields:
{
// to transform name. Used by alias or node.js module
fileNameTransformer: (fileName, parentModule) => fileName;
// check should you wipe module or not. Never used :)
wipeCheck: (stubs, moduleName) => boolean,
// check is mocking allowed for a module. User in relative plugin
shouldMock: (mock, requestFilename, parentModule, entryPoint) => boolean
}
If something is not working - just check that you:
- added a plugin to transform names (nodejs, webpackAlias or relative)
- use .toBeUsed for each mocks And they actually were mocked. If not - rewiremock will throw an Error.
MIT
Happy mocking!