A lot of effort has been put into Faker to create a useful and handy library.
There are still a lot of things to be done, so all contributions are welcome!
If you want to make Faker a better, please read the following contribution guide.
Please make sure that you run pnpm run preflight before making a PR to ensure that everything is working from the start.
This is a shorthand for running the following scripts in order:
pnpm install- installs npm packages defined in package.jsonpnpm run generate:locales- generates locale filespnpm run generate:api-docs- generates API documentationpnpm run format- runs prettify to format codepnpm run lint- runs ESLint to enforce project code standardspnpm run build:clean- removes artifacts from previous buildspnpm run build:code- builds the code, both CommonJS and ESM versionspnpm run build:types- builds the TypeScript type definitionspnpm run test:update-snapshots- runs all tests, and updates any snapshots if neededpnpm run ts-check- checks that there are no TypeScript errors in any files
- The project is being built by esbuild (see bundle.ts)
- The documentation is running via VitePress.
Make sure you build the project before running the docs, cause some files depend on
dist. Usepnpm run docs:devto edit them in live mode. - The tests are executing
vitestagainsttest/**/*.spec.ts - If you update the locales, make sure to run
pnpm run generate:localesto generate/update the related files.
The sources are located in the src directory. All fake data generators are divided into namespaces (each namespace being a separate module). Most of the generators use the definitions, which are just plain JavaScript objects/arrays/strings that are separate for each locale.
If adding new data definitions to Faker, you'll often need to find source data. Note that:
- Faker must not contain copyrighted materials.
- Facts cannot be copyrighted, so if you are adding or translating a finite, known, list of things such as the names of chemical elements into another language, that's OK.
- But if you are compiling a list of, for example, popular personal names or cities, don't copy directly from a single source (Wikipedia, 'most popular' articles, government data sites etc). A compilation of facts can be copyrighted.
- It's best to refer to multiple sources and use your own judgement/knowledge to make a sample list of data.
After adding new or updating existing locale data, you need to run pnpm run generate:locales to generate/update the related files.
Please only change files related to one module (e.g. person, location) whenever possible. This can simplify/speed up the review process. Additionally, it allows the maintainers to track PRs in a meaningful way by adding related labels.
The project is being built by esbuild (see bundle.ts)
pnpm install
pnpm run buildBefore you can run the tests, you need to install all dependencies and build the project, because some tests depend on the bundled content.
pnpm install
pnpm run build
pnpm run test
# or
pnpm run coverageYou can view a generated code coverage report at coverage/index.html.
All methods should have tests for all their parameters.
Usually, there will be a test case for each of the following scenarios:
- No arguments/Only required parameters
- One parameter/option at a time
- All parameters at once
- Special cases
We won't test for arguments that don't match the expected types.
Our tests are separated into two parts:
- Fixed Seeded Tests
- Random Seeded Tests
The fixed seeded tests are used to check that the returned results are matching the users expectations and are deterministic. Each iteration will return in the same results as the previous. Here, the automatically generated test snapshots should be reviewed in depth. This is especially important if you refactor a method to ensure no unexpected behavior occurs.
There are two ways to write these tests.
Methods without arguments can be tested like this:
import { faker } from '../src';
import { seededTests } from './support/seededRuns';
seededTests(faker, 'someModule', (t) => {
t.it('someMethod');
// Or if multiple similar methods exist:
t.itEach('someMethod1', 'someMethod2', 'someMethod3');
});Methods with arguments can be tested like this:
import { faker } from '../src';
import { seededTests } from './support/seededRuns';
seededTests(faker, 'someModule', (t) => {
t.describe('someMethod', (t) => {
t.it('noArgs')
.it('with param1', true)
.it('with param1 and param2', false, 1337);
});
// Or if multiple similar methods exist:
t.describeEach(
'someMethod1',
'someMethod2',
'someMethod3'
)((t) => {
t.it('noArgs')
.it('with param1', true)
.it('with param1 and param2', false, 1337);
});
});You can update the snapshot files by running pnpm run test -u.
The random seeded tests return a random result in each iteration. They are intended to check for edge cases and function as general result checks. The tests will usually use regex or preferably validator.js to ensure the method returns valid results. We repeat these tests a few times to reduce the likelihood of flaky tests caused by the various corner cases that the implementation or the relevant locale data might have. The loop can also be used to steeply increase the test count to trigger rare issues.
import { describe, expect, it } from 'vitest';
import { faker } from '../src';
describe('someModule', () => {
describe(`random seeded tests for seed ${faker.seed()}`, () => {
for (let i = 1; i <= NON_SEEDED_BASED_RUN; i++) {
describe('someMethod', () => {
it('Should return a valid result', () => {
const actual = faker.someModule.someMethod();
expect(actual).toBeTypeOf('string');
expect(actual).toSatisfy(validatorjs.isAlphanumeric);
// ...
});
// ...
});
}
});
});If you ever find yourself deprecating something in the source code, you can follow these steps to save yourself (and the reviewers) some trouble.
If the code you want to deprecate is a property, convert it to a getter first. Now that you have a function, the first thing you want to do is call the internal deprecated function. Afterwards, add a @deprecated parameter to the end of the JSDoc with a human readable description message with a suitable replacement for the deprecated function. Lastly, add a @see parameter to the JSDoc with a link to the replacement in the faker library (if it exists). The syntax for the link is faker.[module].[function].
Example:
/**
* @see faker.cat.random()
*
* @deprecated Use `faker.cat.random()` instead.
*/
get cat() {
deprecated({
deprecated: 'faker.animal.cat',
});
return 'cat';
}Each major version has an upgrading guide, e.g. next.fakerjs.dev/guide/upgrading.
While developing new features and fixing bugs for a new release, changes are added to the migration guide to aid developers when the version is released.
The general principle is to document anything which requires a normal user of the library to change their code which uses Faker when upgrading to the new major version.
There are two sections:
- Breaking changes (user MUST change their code)
- Deprecations and other changes (user SHOULD change their code but it will still work for this major version even if they don't)
Not every change needs to be in the migration guide. If it is too long, it becomes hard for users to spot the important changes.
- Breaking changes, e.g. removal of methods
- Behavior changes, e.g. a different default for a parameter, or a parameter becoming required
- Whole modules renaming (e.g. faker.name to faker.person)
- Locale renames
- Changes to minimum versions e.g. requiring a new version of Node
- Changes to how Faker is imported
- New locales
- Changes to locale data in existing locales
- Bugfixes where it's unlikely anyone was relying on the old behavior (eg broken values in locale files)
- New methods and parameters
- Straightforward method aliases, e.g. where a method or parameter is renamed but the old name still works identically. (Runtime warnings will already guide the user in this case)
- Changes to locale definition files which only affect usage via
faker.helpers.fake, e.g. if a definition file is renamed, but the public API for the method stays the same
JSDoc are comments above any code structure (variable, function, class, etc.) that begin with /** and end with */. Multiline comments start (if not being the start or end line) with a *.
For more info checkout jsdoc.app.
JSDoc will be read and automatically processed by generate:api-docs and therefore need to follow some project conventions. Other standards are in place because we think they increase the code quality.
We have a small set of JSDoc tags that all methods should have.
- Description
@param- If the method has parameters@see- If there are other important methods@example- Example calls without and with parameters, including a sample result of each call@since- The version this method was added (or is likely to be added)@deprecated- If the method is deprecated, with additional information about replacements
| Do | Dont |
|---|---|
/**
* This is a good JSDoc description for a method that generates foos.
*
* @param options The optional options to use.
* @param options.test The parameter to configure test. Defaults to `'bar'`.
*
* @see faker.helper.fake
*
* @example
* faker.bar.foo() // 'foo'
* faker.bar.foo({ test: 'oof' }) // 'of'
*
* @since 7.5.0
*
* @deprecated Use `faker.cat.random()` instead.
*/
function foo(options: { test: string } = {}): string {
// implementation
} |
/**
* This is a bad JSDoc description.
*
* @return foo
*/
function foo(options: { test: string }) {
// implementation
} |
We use eslint-plugin-jsdoc to test for basic styling and sorting of doc-tags.
This is in place so all JSDoc tags will get sorted automatically, so you don't have to bother with it. This also means that most rules in this section can get auto fixed by the eslint formatter.
JSDocs should always be multiline
While single line JSDoc are technically valid, we decided to follow this rule since it makes changes in the git diff much more clear and easier to understand.
| Do | Dont |
|---|---|
/**
* This is a good JSDoc description.
*/
function foo() {
// implementation
} |
/** This is a bad JSDoc description. */
function foo() {
// implementation
} |
Everything that can be accessed directly by a user should have JSDoc.
This rule is aimed to target anything that is exported from the faker library. This includes types, interfaces, functions, classes and variables. So if you introduce anything new that is not internal, write JSDoc for it.
If a
@paramhas a default value, it needs to be mentioned at the end of the sentence.
/**
* This is a good JSDoc description.
*
* @param bar this is a parameter description. Defaults to `0`.
*/
function foo(bar: number = 0) {
// implementation
}If a function can throw an error (FakerError) you have to include the
@throwstag with an explanation when an error could be thrown
/**
* This is a good JSDoc description.
*
* @param bar this is a parameter description. Defaults to `0`.
*
* @throws If bar is negative.
*/
function foo(bar: number = 0) {
// implementation
}Sentences should always end with a period.
This rule ensures minimal grammatical correctness in the comments and ensures that all comments look the same.
Different tags have to be separated by an empty line.
This rule improves the comments readability by grouping equivalent tags and making them more distinguishable from others.
| Do | Dont |
|---|---|
/**
* This is a good JSDoc block, because it follows the Faker preferences.
*
* @param bar The first argument.
* @param baz The second argument.
*
* @example foo(1, 1) // [1, 1]
* @example foo(13, 56) // [13, 56]
*/
function foo(bar: number, baz: number): [number, number] {
// implementation
} |
/**
* This is a bad JSDoc block, because it has no linebreaks between sections.
* @param bar The first argument.
* @param baz The second argument.
* @example foo(1, 1) // [1, 1]
* @example foo(13, 56) // [13, 56]
*/
function foo(bar: number, baz: number): [number, number] {
// implementation
} |
Before running the docs, build the Faker dist, it's used inside of certain routes.
pnpm run build
pnpm run docs:devIf you changed something heavily in the docs, like auto-generating content, you should check the docs statically, because it could differ from the dev version. Before running the docs, build the Faker dist, it's used inside of certain routes.
pnpm run build
pnpm run docs:build # Output docs to /dist
pnpm run docs:serve # Serve docs from /distSee the netlify.toml for configuration.
Pull Request titles need to follow our semantic convention.
PR titles are written in following convention: type(scope): subject
type is required and indicates the intent of the PR
The types
featandfixwill be shown in the changelog as### Featuresor### Bug Fixes
All other types wont show up except for breaking changes marked with the!in front of:
Allowed types are:
| type | description |
|---|---|
| feat | A new feature is introduced |
| fix | A bug was fixed |
| chore | No user affected code changes were made |
| refactor | A refactoring that affected also user (e.g. log a deprecation warning) |
| docs | Docs were changed |
| test | Test were changed |
| ci | CI were changed |
| build | Build scripts were changed |
| infra | Infrastructure related things were made (e.g. issue-template was updated) |
| revert | A revert was triggered via git |
scope is optional and indicates the scope of the PR
The scope will be shown in the changelog in front of the subject in bold text
Also as the commits are sorted alphabetically, the scope will group the commits indirectly into categories
Allowed scopes are:
| scope | description |
|---|---|
| <module-name> | The specific module name that was affected by the PR |
| locale | When only locale(s) are added/updated/removed |
| module | When some modules where updates or something related to modules were changed |
| revert | When a revert was made via git |
| deps | Will mostly be used by Renovate |
| release | Will be set by release process |
The scope is not checkable via
Semantic Pull Requestaction as this would limit the scopes to only existing modules,
but if we add a new module likecolor, then the PR author couldn't use the new module name as scope.
As such, we (the Faker team) must be mindful of valid scopes and we reserve the right to edit titles as we see fit.
subject is required and describes what the PR does
Please note that the PR title should not include a suffix of e.g.
(#123)as this will be done automatically by GitHub while merging
Some examples of valid pull request titles:
feat: add casing option
feat(locale): extend Hebrew (he)
fix: lower target to support Webpack 4
chore: add naming convention rule
refactor(location): deprecate streetPrefix and streetSuffix
docs: remove unused playground
test: validate @see contents
ci: allow breaking change commits
build: add node v18 support
infra: rework bug-report template
revert: add more arabic names dataset (#362)
# Breaking changes
refactor!: remove faker default export
build!: remove node v12 support
# A release PR will look like this
chore(release): 7.4.0
# Renovate automatically generates these
chore(deps): update devdependencies
chore(deps): update typescript-eslint to ~5.33.0Previous pull request titles that could have been written in a better way:
- feat: `datatype.hexadecimal` signature change
+ feat(datatype): hexadecimal signature change
datatype is one of our modules and can be used as scope
- feat(image): add image via.placeholder provider
+ feat(image): add via.placeholder provider
image was redundant in the subject
- feat(system.networkInterface): add networkInterface faker
+ feat(system): add networkInterface method
networkInterface was redundant in the scope and made the whole commit message long
also method in the subject explains a bit better what it is
- chore(bug-report-template): new design
+ infra: rework bug-report template
the type infra tells that no actual code-changes were made
the subject contains what the PR does
- chore: rename Gender to Sex
+ refactor(name): rename Gender to Sex
this was not a chore as it touched runtime code that affected the end-user
scope name can be used here to tell that the change affects only the name module