Note
This documentation is forbabel-plugin-tester@β₯11
. Documentation for earlier versions can be found here.
Utilities for testing babel plugins and presets.
You are writing a babel plugin or preset and want to write tests for it too.
This is a fairly simple abstraction to help you write tests for your babel
plugin or preset. It was built to work with Jest, but most of the
functionality will work with Mocha, Jasmine, node:test
,
Vitest, and any other test runner that defines standard describe
and it
globals with async support (see appendix).
This package is tested on both Windows and nix (Ubuntu) environments.
This module is distributed via npm which is bundled with node and
should be installed as one of your project's devDependencies
:
npm install --save-dev babel-plugin-tester
ESM:
import { pluginTester } from 'babel-plugin-tester';
CJS:
const { pluginTester } = require('babel-plugin-tester');
/* file: test/unit.test.js */
import { pluginTester } from 'babel-plugin-tester';
import yourPlugin from '../src/your-plugin';
pluginTester({
plugin: yourPlugin,
tests: {
/* Your test objects */
}
});
Note how
pluginTester
does not appear inside anytest
/it
block nor within any hook functions. For advanced use cases,pluginTester
may appear within one or moredescribe
blocks, though this is discouraged.
In your terminal of choice:
# Prettier@3 requires --experimental-vm-modules for older Node versions
NODE_OPTIONS='--no-warnings --experimental-vm-modules' npx jest
This section lists the options you can pass to babel-plugin-tester. They are all optional with respect to the following:
- When testing a preset, the
preset
option is required. - When testing a plugin, the
plugin
option is required. - You must test either a preset or a plugin.
- You cannot use preset-specific options (
preset
,presetName
,presetOptions
) and plugin-specific options (plugin
,pluginName
,pluginOptions
) at the same time.
This is used to provide the babel plugin under test. For example:
/* file: test/unit.test.js */
import { pluginTester } from 'babel-plugin-tester';
import identifierReversePlugin from '../src/identifier-reverse-plugin';
pluginTester({
plugin: identifierReversePlugin,
tests: {
/* Your test objects */
}
});
/* file: src/identifier-reverse-plugin.js */
// Normally you would import this from your plugin module
function identifierReversePlugin() {
return {
name: 'identifier reverse',
visitor: {
Identifier(idPath) {
idPath.node.name = idPath.node.name.split('').reverse().join('');
}
}
};
}
This is used as the describe block name and in your tests' names. If
pluginName
can be inferred from the plugin
's name, then it will
be and you do not need to provide this option. If it cannot be inferred for
whatever reason, pluginName
defaults to "unknown plugin"
.
Note that there is a small caveat when relying on pluginName
inference.
This is used to pass options into your plugin at transform time. If provided,
the object will be lodash.mergeWith
'd with each test
object's pluginOptions
/fixture's pluginOptions
, with the latter
taking precedence. Note that arrays will be concatenated and explicitly
undefined values will unset previously defined values during merging.
This is used to provide the babel preset under test. For example:
/* file: cool-new-babel-preset.test.js */
import path from 'node:path';
import { pluginTester } from 'babel-plugin-tester';
import coolNewBabelPreset from './cool-new-babel-preset.js';
pluginTester({
preset: coolNewBabelPreset,
// A path to a directory containing your test fixtures
fixtures: path.join(__dirname, 'fixtures')
});
/* file: cool-new-babel-preset.js */
function identifierReversePlugin() {
return {
name: 'identifier reverse',
visitor: {
Identifier(idPath) {
idPath.node.name = idPath.node.name.split('').reverse().join('');
}
}
};
}
function identifierAppendPlugin() {
return {
name: 'identifier append',
visitor: {
Identifier(idPath) {
idPath.node.name = `${idPath.node.name}_appended`;
}
}
};
}
export function coolNewBabelPreset() {
return { plugins: [identifierReversePlugin, identifierAppendPlugin] };
}
This is used as the describe block name and in your tests' names.
Defaults to "unknown preset"
.
This is used to pass options into your preset at transform time. If provided,
the object will be lodash.mergeWith
'd with each test
object's presetOptions
/fixture's presetOptions
, with the latter
taking precedence. Note that arrays will be concatenated and explicitly
undefined values will unset previously defined values during merging.
This is used to provide your own implementation of babel. This is particularly useful if you want to use a different version of babel than what's required by this package.
This is used to configure babel. If provided, the object will be
lodash.mergeWith
'd with the defaults and each test
object's babelOptions
/fixture's babelOptions
, with the latter
taking precedence. Note that arrays will be concatenated and explicitly
undefined values will unset previously defined values during merging.
Also note that babelOptions.babelrc
and babelOptions.configFile
are set to false
by default, which disables automatic babel configuration
loading. This can be re-enabled if desired.
To simply reuse your project's babel.config.js
or some other
configuration file, set babelOptions
like so:
import path from 'node:path';
import { pluginTester } from 'babel-plugin-tester';
pluginTester({
plugin: yourPlugin,
// ...
babelOptions: require(path.join('..', 'babel.config.js')),
// ...
tests: {
/* Your test objects */
}
});
By default, when you include a custom list of plugins or presets in
babelOptions
, the plugin or preset under test will always be the final plugin
or preset to run.
For example, consider the myPlugin
plugin:
import { pluginTester } from 'babel-plugin-tester';
pluginTester({
plugin: myPlugin,
pluginName: 'my-plugin',
babelOptions: {
plugins: [
['@babel/plugin-syntax-decorators', { legacy: true }],
['@babel/plugin-proposal-class-properties', { loose: true }]
]
}
});
By default, myPlugin
will be invoked after @babel/plugin-syntax-decorators
and @babel/plugin-proposal-class-properties (i.e. myPlugin
is appended by
default).
It is possible to specify a custom ordering using the exported
runPluginUnderTestHere
symbol. For instance, to run myPlugin
after
@babel/plugin-syntax-decorators but before
@babel/plugin-proposal-class-properties:
import { pluginTester, runPluginUnderTestHere } from 'babel-plugin-tester';
pluginTester({
plugin: myPlugin,
pluginName: 'my-plugin',
babelOptions: {
plugins: [
['@babel/plugin-syntax-decorators', { legacy: true }],
runPluginUnderTestHere,
['@babel/plugin-proposal-class-properties', { loose: true }]
]
}
});
Or to run myPlugin
before both @babel/plugin-syntax-decorators and
@babel/plugin-proposal-class-properties:
import { pluginTester, runPluginUnderTestHere } from 'babel-plugin-tester';
pluginTester({
plugin: myPlugin,
pluginName: 'my-plugin',
babelOptions: {
plugins: [
runPluginUnderTestHere,
['@babel/plugin-syntax-decorators', { legacy: true }],
['@babel/plugin-proposal-class-properties', { loose: true }]
]
}
});
The same can be done when testing presets. Note that myPreset
is normally
prepended by default since, unlike plugins, presets are run in reverse
order:
import { pluginTester, runPresetUnderTestHere } from 'babel-plugin-tester';
pluginTester({
preset: myPreset,
presetName: 'my-preset',
babelOptions: {
presets: [
'@babel/preset-typescript',
['@babel/preset-react', { pragma: 'dom' }],
runPresetUnderTestHere
]
}
});
In this example, myPreset
will run first instead of last.
This is used to specify a custom title for the two top-level describe
blocks, the first enclosing all tests (i.e. describe(title, ...)
)
and the second enclosing all fixtures (i.e.
describe(`${title} fixtures`, ...)
).
Explicitly setting this option will override any defaults or inferred values.
Set to false
to prevent the creation of these enclosing describe blocks.
Otherwise, the title defaults to using pluginName
/presetName
.
This is used to resolve relative paths provided by the fixtures
option;
the test object properties codeFixture
, outputFixture
, and
execFixture
; and during configuration resolution for prettier.
That is: if the aforesaid properties are not absolute paths, they will be
path.join
'd with the directory name of filepath
.
filepath
is also passed to formatResult
if a more specific path is not
available, and it is used as the default value for babelOptions.filename
in
test objects.
This option defaults to the absolute path of the file that invoked the
pluginTester
function.
For backwards compatibility reasons,
filepath
is synonymous withfilename
. They can be used interchangeably, though care must be taken not to confuse the babel-plugin-tester optionfilename
withbabelOptions.filename
. They are NOT the same!
This is used to control which line endings both the actual output from babel and
the expected output will be converted to. Defaults to "lf"
.
Options | Description |
---|---|
"lf" |
Use Unix-style line endings |
"crlf" |
Use Windows-style line endings |
"auto" |
Use the system default line endings |
"preserve" |
Use the line endings from the input |
false |
Disable line ending conversion entirely |
When disabling line ending conversion, note that Babel will always output LF even if the input is CRLF.
This function will be run before every test runs, including fixtures. It can
return a function which will be treated as a teardown
function. It can
also return a promise. If that promise resolves to a function, that will be
treated as a teardown
function.
See here for the complete run order.
This function will be run after every test runs, including fixtures. You can
define this via teardown
or you can return it from the setup
function.
This can likewise return a promise if it is asynchronous.
This function, if provided, will be run after any teardown function returned
by setup
. See here for the complete run order.
This function is used to format all babel outputs, and defaults to a function that invokes prettier. If a prettier configuration file is found, then that will be used. Otherwise, prettier will use its own default configuration.
You can also override or entirely disable formatting.
Equivalent to snapshot
but applied globally across all test
objects.
Equivalent to fixtureOutputName
but applied globally across all
fixtures.
Equivalent to fixtureOutputExt
but applied globally across all
fixtures.
Determines which test titles are prefixed with a number when registering test
blocks (e.g. `1. ${title}`
, `2. ${title}`
, etc). Defaults to
"all"
.
Options | Description |
---|---|
"all" |
All test object and fixtures tests will be numbered |
"tests-only" |
Only test object tests will be numbered |
"fixtures-only" |
Only fixtures tests will be numbered |
false |
Disable automatic numbering in titles entirely |
Normally, multiple invocations of babel-plugin-tester in the same test file will share the same test title numbering. For example:
/* file: test/unit.test.js */
import { pluginTester } from 'babel-plugin-tester';
import yourPlugin from '../src/your-plugin';
pluginTester({
plugin: yourPlugin,
tests: { 'test one': testOne, 'test two': testTwo }
});
pluginTester({
plugin: yourPlugin,
tests: { 'test one': testOne, 'test x': testTwo }
});
pluginTester({
plugin: yourPlugin,
tests: { 'test five': testOne }
});
Will result in test blocks with names like:
1. Test one
2. Test two
3. Test one
4. Test x
5. Test five
However, setting this option to true
will restart the numbering:
/* file: test/unit.test.js */
import { pluginTester } from 'babel-plugin-tester';
import yourPlugin from '../src/your-plugin';
pluginTester({
plugin: yourPlugin,
tests: { 'test one': testOne, 'test two': testTwo }
});
pluginTester({
plugin: yourPlugin,
restartTitleNumbering: true,
tests: { 'test one': testOne, 'test x': testTwo }
});
pluginTester({
plugin: yourPlugin,
tests: { 'test five': testOne }
});
Which will result in test blocks with names like:
1. Test one
2. Test two
1. Test one
2. Test x
3. Test five
This option is false
by default.
There are two ways to create tests: using the tests
option to provide
one or more test objects or using the fixtures
option described here.
Both can be used simultaneously.
The fixtures
option must be a path to a directory with a structure similar to
the following:
fixtures
βββ first-test # test title will be: "1. first test"
βΒ Β βββ code.js # required
βΒ Β βββ output.js # required (unless using the `throws` option)
βββ second-test # test title will be: "2. second test"
β βββ .babelrc.js # optional
β βββ options.json # optional
β βββ code.ts # required (other file extensions are allowed too)
β βββ output.js # required (unless using the `throws` option)
βββ nested
βββ options.json # optional
βββ third-test # test title will be: "3. nested > third test"
β βββ code.mjs # required (other file extensions are allowed too)
β βββ output.js # required (unless using the `throws` option)
β βββ options.js # optional (overrides props in nested/options.json)
βββ x-fourth-test # test title will be: "4. nested > x fourth test"
βββ exec.js # required (alternative to code/output structure)
Assuming the fixtures
directory is in the same directory as your test file,
you could use it with the following configuration:
pluginTester({
plugin,
fixtures: path.join(__dirname, 'fixtures')
});
If
fixtures
is not an absolute path, it will bepath.join
'd with the directory name offilepath
.
.babelrc
,.babelrc.json
,.babelrc.js
,.babelrc.cjs
, and.babelrc.mjs
config files in fixture directories are technically supported out-of-the-box, though.mjs
config files will cause a segfault in Jest until this issue with V8/Chromium is resolved.
And it would run four tests, one for each directory in fixtures
containing a
file starting with "code" or "exec".
This file's contents will be used as the source code input into babel at
transform time. Any file extension can be used, even a multi-part extension
(e.g. .test.js
in code.test.js
) as long as the file name starts with
code.
; the expected output file will have the same file extension suffix
(i.e. .js
in code.test.js
) as this file unless changed with the
fixtureOutputExt
option.
After being transformed by babel, the resulting output will have whitespace trimmed, line endings converted, and then get formatted by prettier.
Note that this file cannot appear in the same directory as exec.js
. If
more than one code.*
file exists in a directory, the first one will be used
and the rest will be silently ignored.
This file, if provided, will have its contents compared with babel's output,
which is code.js
transformed by babel and formatted with prettier.
If this file is missing and neither throws
nor exec.js
are being
used, this file will be automatically generated from babel's output.
Additionally, the name and extension of this file can be changed with the
fixtureOutputName
and fixtureOutputExt
options.
Before being compared to babel's output, this file's contents will have whitespace trimmed and line endings converted.
Note that this file cannot appear in the same directory as exec.js
.
This file's contents will be used as the input into babel at transform time just
like the code.js
file, except the output will be evaluated in the
same CJS context as the test runner itself, meaning it supports features
like a/sync IIFEs, debugging breakpoints (!), and has access to mocked modules,
expect
, require
, __dirname
and __filename
(derived from this file's
path), and other globals/features provided by your test framework. However, the
context does not support import
, top-level await, or any other ESM syntax.
Hence, while any file extension can be used (e.g. .ts
, .vue
, .jsx
), this
file will always be evaluated as CJS.
The test will always pass unless an exception is thrown (e.g. when an expect()
fails).
Use this to make advanced assertions on the output. For example, to test that
babel-plugin-proposal-throw-expressions actually throws, your exec.js
file might contain:
expect(() => throw new Error('throw expression')).toThrow('throw expression');
Keep in mind that, despite sharing a global context, execution will occur in a separate realm, which means native/intrinsic types will be different. This can lead to unexpectedly failing tests. For example:
expect(require(`${__dirname}/imported-file.json`)).toStrictEqual({ data: 'imported' });This may fail in some test frameworks with the message "serializes to the same string". This is because the former object's
Object
prototype comes from a different realm than the second object'sObject
prototype, meaning the two objects are not technically strictly equal. However, something like the following, which creates two objects in the same realm, will pass:expect( Object.fromEntries( Object.entries(require(`${__dirname}/imported-file.json`)) ) ).toStrictEqual({ data: 'imported' });Or:
expect(JSON.stringify(require(`${__dirname}/imported-file.json`))).toBe( JSON.stringify({ data: 'imported' }) );Or even:
expect(require(`${__dirname}/imported-file.json`)).toEqual({ data: 'imported' });
After being transformed by babel but before being evaluated, the babel output will have whitespace trimmed, line endings converted, and then get formatted by prettier.
Note that this file cannot appear in the same directory as code.js
or
output.js
. If more than one exec.*
file exists in a directory, the
first one will be used and the rest will be silently ignored.
For each fixture, the contents of the entirely optional options.json
file are
lodash.mergeWith
'd with the options provided to
babel-plugin-tester, with the former taking precedence. Note that arrays will be
concatenated and explicitly undefined values will unset previously defined
values during merging.
For added flexibility, options.json
can be specified as options.js
instead
so long as a JSON object is exported via module.exports
. If both files
exist in the same directory, options.js
will take precedence and
options.json
will be ignored entirely.
Fixtures support deeply nested directory structures as well as shared or "root"
options.json
files. For example, placing an options.json
file in the
fixtures/nested
directory would make its contents the "global configuration"
for all fixtures under fixtures/nested
. That is: each fixture would
lodash.mergeWith
the options provided to
babel-plugin-tester, fixtures/nested/options.json
, and the contents of their
local options.json
file as described above.
What follows are the properties you may use if you provide an options file, all of which are optional:
This is used to configure babel. Properties specified here override
(lodash.mergeWith
) those from the babelOptions
option provided to babel-plugin-tester. Note that arrays will be concatenated
and explicitly undefined values will unset previously defined values during
merging.
This is used to pass options into your plugin at transform time. Properties
specified here override (lodash.mergeWith
) those from the
pluginOptions
option provided to babel-plugin-tester. Note that arrays
will be concatenated and explicitly undefined values will unset previously
defined values during merging.
Unlike with babel-plugin-tester's options, you can safely mix plugin-specific
properties (like pluginOptions
) with preset-specific properties (like
presetOptions
) in your options files.
This is used to pass options into your preset at transform time. Properties
specified here override (lodash.mergeWith
) those from the
presetOptions
option provided to babel-plugin-tester. Note that arrays
will be concatenated and explicitly undefined values will unset previously
defined values during merging.
Unlike with babel-plugin-tester's options, you can safely mix plugin-specific
properties (like pluginOptions
) with preset-specific properties (like
presetOptions
) in your options files.
If provided, this will be used as the title of the test. Otherwise, the directory name will be used as the title by default (with spaces replacing dashes).
Use this to run only the specified fixture. Useful while developing to help
focus on a small number of fixtures. Can be used in multiple options.json
files.
Requires Jest, an equivalent interface (like Vitest), or a manually-defined
it
object exposing an appropriateonly
method.
Use this to skip running the specified fixture. Useful for when you are working
on a feature that is not yet supported. Can be used in multiple options.json
files.
Requires Jest, an equivalent interface (like Vitest), or a manually-defined
it
object exposing an appropriateskip
method.
When using certain values, this property must be used in
options.js
instead ofoptions.json
.
Use this to assert that a particular code.js
file should cause babel to throw
an error during transformation. For example:
{
// ...
throws: true,
throws: 'should have this exact message',
throws: /should pass this regex/,
throws: SyntaxError, // Should be an instance of this class
throws: err => {
if (err instanceof SyntaxError && /message/.test(err.message)) {
return true; // Test will fail if this function's return value !== true
}
},
}
Be careful using
instanceof
across realms as it can lead to strange behavior with frontend frames/windows and with tools that rely on Node's VM module (like Jest). Prefer name checks and utilities likeisNativeError
,Array.isArray
, and overridingSymbol.hasInstance
instead.
If the value of throws
is a class, that class must be a subtype of
Error
or the behavior of babel-plugin-tester is undefined.
For backwards compatibility reasons,
throws
is synonymous witherror
. They can be used interchangeably, withthrows
taking precedence.
Note that this property cannot be present when using an exec.js
or
output.js
file or when using the outputRaw
option.
As it requires a function value, this property must be used in
options.js
instead ofoptions.json
.
This function will be run before a particular fixture's tests are run. It can
return a function which will be treated as a teardown
function. It can
also return a promise. If that promise resolves to a function, that will be
treated as a teardown
function.
This function, if provided, will run after any setup
function provided
as a babel-plugin-tester option. See here for the complete run order.
As it requires a function value, this property must be used in
options.js
instead ofoptions.json
.
This function will be run after a fixture's tests finish running. You can define
this via teardown
or you can return it from the setup
function. This
can likewise return a promise if it is asynchronous.
This function, if provided, will be run after any teardown function returned
by the setup
property, both of which will run before any
teardown
function provided as a babel-plugin-tester option. See
here for the complete run order.
As it requires a function value, this property must be used in
options.js
instead ofoptions.json
.
This function is used to format all babel outputs, and defaults to a function that invokes prettier. If a prettier configuration file is found, then that will be used. Otherwise, prettier will use its own default configuration.
You can also entirely disable formatting.
This will override the formatResult
function provided to
babel-plugin-tester.
As it requires a function value, this property must be used in
options.js
instead ofoptions.json
.
This option is similar in intent to output.js
except it tests against
the entire BabelFileResult
object returned by babel's transform
function instead of only the code
property of BabelFileResult
.
outputRaw
must be a function with the following signature:
outputRaw: (output: BabelFileResult) => void
Where the output
parameter is an instance of BabelFileResult
:
interface BabelFileResult {
ast?: Node | undefined;
code?: string | undefined;
ignored?: boolean | undefined;
map?: object | undefined;
metadata?: BabelFileMetadata | undefined;
}
So long as the outputRaw
function does not throw, it will never cause the test
to fail. On the other hand, if the outputRaw
function throws, such as when
expect(output.metadata).toStrictEqual({ ... })
fails, the test will fail
regardless of other options.
The output
parameter is not trimmed, converted, stripped, or modified at all.
Note that outputRaw
does not replace output.js
et al, it only adds
additional (custom) expectations to your test. Further note that this option
can appear alongside any other fixtures
option except throws
.
Use this to provide your own fixture output file name. Defaults to "output"
.
Use this to provide your own fixture output file extension. Including the
leading period is optional; that is: if you want output.jsx
,
fixtureOutputExt
can be set to either "jsx"
or ".jsx"
. If omitted, the
input fixture's file extension will be used instead.
This is particularly useful if you are testing TypeScript input.
There are two ways to create tests: using the fixtures
option that
leverages the filesystem or using the tests
option described here. Both can be
used simultaneously.
Using the tests
option, you can provide test objects describing your
expected transformations. You can provide tests
as an object of test objects
or an array of test objects. If you provide an object, the object's keys will be
used as the default title of each test. If you provide an array, each test's
default title will be derived from its index and
pluginName
/presetName
.
See the example for more details.
A minimal test object can be:
What follows are the properties you may use if you provide an object, most of which are optional:
This is used to configure babel. Properties specified here override
(lodash.mergeWith
) those from the babelOptions
option provided to babel-plugin-tester. Note that arrays will be concatenated
and explicitly undefined values will unset previously defined values during
merging.
This is used to pass options into your plugin at transform time. Properties
specified here override (lodash.mergeWith
) those from the
pluginOptions
option provided to babel-plugin-tester. Note that arrays
will be concatenated and explicitly undefined values will unset previously
defined values during merging.
Unlike with babel-plugin-tester's options, you can safely mix plugin-specific
properties (like pluginOptions
) with preset-specific properties (like
presetOptions
) in your test objects.
This is used to pass options into your preset at transform time. Properties
specified here override (lodash.mergeWith
) those from the
presetOptions
option provided to babel-plugin-tester. Note that arrays
will be concatenated and explicitly undefined values will unset previously
defined values during merging.
Unlike with babel-plugin-tester's options, you can safely mix plugin-specific
properties (like pluginOptions
) with preset-specific properties (like
presetOptions
) in your test objects.
If provided, this will be used as the title of the test. Otherwise, the title will be determined from test object by default.
Use this to run only the specified test. Useful while developing to help focus on a small number of tests. Can be used on multiple tests.
Requires Jest, an equivalent interface (like Vitest), or a manually-defined
it
object exposing an appropriateonly
method.
Use this to skip running the specified test. Useful for when you are working on a feature that is not yet supported. Can be used on multiple tests.
Requires Jest, an equivalent interface (like Vitest), or a manually-defined
it
object exposing an appropriateskip
method.
Use this to assert that a particular test object should cause babel to throw an error during transformation. For example:
{
// ...
throws: true,
throws: 'should have this exact message',
throws: /should pass this regex/,
throws: SyntaxError, // Should be an instance of this class
throws: err => {
if (err instanceof SyntaxError && /message/.test(err.message)) {
return true; // Test will fail if this function's return value !== true
}
},
}
Be careful using
instanceof
across realms as it can lead to strange behavior with frontend frames/windows and with tools that rely on Node's VM module (like Jest). Prefer name checks and utilities likeisNativeError
,Array.isArray
, and overridingSymbol.hasInstance
instead.
If the value of throws
is a class, that class must be a subtype of
Error
or the behavior of babel-plugin-tester is undefined.
For backwards compatibility reasons,
throws
is synonymous witherror
. They can be used interchangeably, withthrows
taking precedence.
Note that this property cannot be present when using the output
,
outputRaw
, outputFixture
, exec
, execFixture
, or
snapshot
properties.
This function will be run before a particular test is run. It can return a
function which will be treated as a teardown
function. It can also
return a promise. If that promise resolves to a function, that will be treated
as a teardown
function.
This function, if provided, will run after any setup
function provided
as a babel-plugin-tester option. See here for the complete run order.
This function will be run after a test finishes running. You can define this via
teardown
or you can return it from the setup
function. This can
likewise return a promise if it is asynchronous.
This function, if provided, will be run after any teardown function returned
by the setup
property, both of which will run before any
teardown
function provided as a babel-plugin-tester option. See
here for the complete run order.
This function is used to format all babel outputs, and defaults to a function that invokes prettier. If a prettier configuration file is found, then that will be used. Otherwise, prettier will use its own default configuration.
You can also entirely disable formatting.
This will override the formatResult
function provided to
babel-plugin-tester.
If you would prefer to take a snapshot of babel's output rather than compare it
to something you provide manually, specify snapshot: true
. This will cause
babel-plugin-tester to generate a snapshot containing both the source code
and babel's output.
Defaults to false
.
Note that this property cannot appear in the same test object as the
output
, outputFixture
, exec
, execFixture
, or
throws
properties. However, it can be used with outputRaw
.
Requires Jest, an appropriate shim or equivalent interface (like Vitest), or a manually-defined
expect
object exposing an appropriatetoMatchSnapshot
method.
The code that you want babel to transform using your plugin or preset. This must
be provided unless you are using the codeFixture
or exec
properties instead. If you do not provide the output
or
outputFixture
properties, and snapshot
is not truthy, then the
assertion is that this code is unchanged by the transformation.
Before being transformed by babel, any indentation will be stripped as a convenience for template literals. After being transformed, the resulting output will have whitespace trimmed, line endings converted, and then get formatted by prettier.
Note that this property cannot appear in the same test object as the
codeFixture
, exec
, or execFixture
properties.
The value of this property will be compared with the output from babel's
transform
function.
Before being compared to babel's output, this value will have whitespace trimmed, line endings converted, and any indentation stripped as a convenience for template literals.
Note that this property cannot appear in the same test object as the
outputFixture
, exec
, execFixture
, throws
, or
snapshot
properties. However, it can be used with outputRaw
.
This property is similar to output
and related properties except it
tests against the entire BabelFileResult
object returned by babel's
transform
function instead of only the code
property of
BabelFileResult
.
outputRaw
must be a function with the following signature:
outputRaw: (output: BabelFileResult) => void
Where the output
parameter is an instance of BabelFileResult
:
interface BabelFileResult {
ast?: Node | undefined;
code?: string | undefined;
ignored?: boolean | undefined;
map?: object | undefined;
metadata?: BabelFileMetadata | undefined;
}
So long as the outputRaw
function does not throw, this property will never
cause the test to fail. On the other hand, if the outputRaw
function throws,
such as when expect(output.metadata).toStrictEqual({ ... })
fails, the test
will fail regardless of other properties.
The output
parameter is not trimmed, converted, stripped, or modified at all.
Note that outputRaw
does not replace output
et al, it only adds
additional (custom) expectations to your test. Further note that outputRaw
can appear in the same test object as any other property except
throws
.
The provided source will be transformed just like the code
property,
except the output will be evaluated in the same CJS context as the
test runner itself, meaning it supports features like a/sync IIFEs, debugging
breakpoints (!), and has access to mocked modules, expect
, require
,
__dirname
and __filename
(derived from available path info and falling back
on filepath
), and other globals/features provided by your test
framework. However, the context does not support import
, top-level await, or
any other ESM syntax. Hence, while any file extension can be used (e.g. .ts
,
.vue
, .jsx
), this file will always be evaluated as CJS.
The test will always pass unless an exception is thrown (e.g. when an expect()
fails).
Use this to make advanced assertions on the output. For example, you can test that babel-plugin-proposal-throw-expressions actually throws using the following:
{
// ...
exec: `
expect(() => throw new Error('throw expression')).toThrow('throw expression');
`;
}
Keep in mind that, despite sharing a global context, execution will occur in a separate realm, which means native/intrinsic types will be different. This can lead to unexpectedly failing tests. For example:
expect(require(`${__dirname}/imported-file.json`)).toStrictEqual({ data: 'imported' });This may fail in some test frameworks with the message "serializes to the same string". This is because the former object's
Object
prototype comes from a different realm than the second object'sObject
prototype, meaning the two objects are not technically strictly equal. However, something like the following, which creates two objects in the same realm, will pass:expect( Object.fromEntries( Object.entries(require(`${__dirname}/imported-file.json`)) ) ).toStrictEqual({ data: 'imported' });Or:
expect(JSON.stringify(require(`${__dirname}/imported-file.json`))).toBe( JSON.stringify({ data: 'imported' }) );Or even:
expect(require(`${__dirname}/imported-file.json`)).toEqual({ data: 'imported' });
After being transformed by babel but before being evaluated, the babel output will have whitespace trimmed, line endings converted, and then get formatted by prettier.
Note that this property cannot appear in the same test object as the
execFixture
, code
, codeFixture
, output
,
outputFixture
, throws
, or snapshot
properties. However,
it can be used with outputRaw
.
If you would rather put your code
in a separate file, you can specify a
file path here instead. If it is an absolute path, then that's the file that
will be loaded. Otherwise, codeFixture
will be path.join
'd with the
directory name of filepath
.
After being transformed by babel, the resulting output will have whitespace trimmed, line endings converted, and then get formatted by prettier.
Like code
, this property cannot appear in the same test object as the
exec
or execFixture
properties, nor the code
property.
If you find you are using this property more than a couple of times, consider using
fixtures
instead.
For backwards compatibility reasons,
codeFixture
is synonymous withfixture
. They can be used interchangeably, though care must be taken not to confuse the test object propertyfixture
with the babel-plugin-tester optionfixtures
, the latter being plural.
If you would rather put your output
in a separate file, you can specify
a file path here instead. If it is an absolute path, then that's the file that
will be loaded. Otherwise, outputFixture
will be path.join
'd with the
directory name of filepath
.
Before being compared to babel's output, this file's contents will have whitespace trimmed and line endings converted.
Like output
, this property cannot appear in the same test object as the
exec
, execFixture
, throws
, or snapshot
properties, nor the output
property. However, it can be used with
outputRaw
.
If you find you are using this property more than a couple of times, consider using
fixtures
instead.
If you would rather put your exec
in a separate file, you can specify a
file path here instead. If it is an absolute path, then that's the file that
will be loaded. Otherwise, execFixture
will be path.join
'd with the
directory name of filepath
.
After being transformed by babel but before being evaluated, the babel output will have whitespace trimmed, line endings converted, and then get formatted by prettier.
Like exec
, this property cannot appear in the same test object as the
code
, codeFixture
, output
, outputFixture
,
throws
, or snapshot
properties, nor the exec
property.
However, it can be used with outputRaw
.
If you find you are using this property more than a couple of times, consider using
fixtures
instead.
import { pluginTester } from 'babel-plugin-tester';
import identifierReversePlugin from '../identifier-reverse-plugin';
// NOTE: you can use beforeAll, afterAll, beforeEach, and afterEach as usual,
// but initial configuration tasks, like loading content from fixture files,
// will complete *at the point the pluginTester function is called* which means
// BEFORE beforeAll and other Jest hooks are run.
pluginTester({
plugin: identifierReversePlugin,
// Defaults to false, but with this line we set the default to true across
// *all* tests.
snapshot: true,
tests: [
{
code: "'hello';"
// Snapshot should show that prettier has changed the single quotes to
// double quotes (using prettier's default configuration).
},
{
// This test will pass if and only if code has not changed.
code: '"hello";'
// To prevent false negatives (like with reckless use of `npx jest -u`),
// snapshots of code that does not change are forbidden. Snapshots
// succeed only when babel output !== code input.
snapshot: false;
},
{
code: 'var hello = "hi";',
output: 'var olleh = "hi";',
// You can't take a snapshot and also manually specify an output string.
// It's either one or the other.
snapshot: false
},
// A valid test can be a test object or a simple string.
`
function sayHi(person) {
return 'Hello ' + person + '!'
}
console.log(sayHi('Jenny'))
`
]
});
import path from 'node:path';
import { pluginTester } from 'babel-plugin-tester';
import identifierReversePlugin from '../identifier-reverse-plugin';
pluginTester({
// One (and ONLY ONE) of the two following lines MUST be included.
plugin: identifierReversePlugin,
//preset: coolNewBabelPreset,
// Usually unnecessary if it is returned by the plugin. This will default to
// 'unknown plugin' if a name cannot otherwise be inferred.
pluginName: 'identifier reverse',
// Unlike with pluginName, there is no presetName inference. This will default
// to 'unknown preset' if a name is not provided.
//presetName: 'cool-new-babel-preset',
// Used to test specific plugin options.
pluginOptions: {
optionA: true
},
//presetOptions: {
// optionB: false,
//}
// Defaults to the plugin name.
title: 'describe block title',
// Only useful if you are using fixtures, codeFixture, outputFixture, or
// execFixture options. Defaults to the absolute path of the file the
// pluginTester function was invoked from, which in this case is equivalent
// to the following line:
filepath: __filename,
// These are the defaults that will be lodash.mergeWith'd with the provided
// babelOptions option.
babelOptions: {
parserOpts: {},
generatorOpts: {},
babelrc: false,
configFile: false
},
// Defaults to false but we're being explicit here: do not use snapshots
// across all tests. Note that snapshots are only guaranteed to work with
// Jest.
snapshot: false,
// Defaults to a function that formats with prettier.
formatResult: customFormatFunction,
// You can provide tests as an object:
tests: {
// The key is the title. The value is the code that is unchanged (because
// snapshot === false across all tests). Test title will be: "1. does not
// change code with no identifiers".
'does not change code with no identifiers': '"hello";',
// Test title will be: "2. changes this code".
'changes this code': {
// Input to the plugin.
code: 'var hello = "hi";',
// Expected output.
output: 'var olleh = "hi";'
}
},
// Alternatively, you can provide tests as an array:
tests: [
// Should be unchanged by the plugin (because snapshot === false across all
// tests). Test title will be: "1. identifier reverse".
'"hello";',
{
// Test title will be: "2. identifier reverse".
code: 'var hello = "hi";',
output: 'var olleh = "hi";'
},
{
// Test title will be: "3. unchanged code".
title: 'unchanged code',
// Because this is an absolute path, the filepath option above will not
// be used to resolve this path.
codeFixture: path.join(
__dirname,
'..',
'fixtures',
'codeFixture-unchanging.js'
)
// No output, outputFixture, or snapshot, so the assertion will be that
// the plugin does not change this code.
},
{
// Because these are not absolute paths, they will be joined with the
// directory of the filepath option provided above.
codeFixture: path.join('..', 'fixtures', 'codeFixture.js'),
// Because outputFixture is provided, the assertion will be that the
// plugin will change the contents of "codeFixture.js" to the contents of
// "outputFixture.js".
outputFixture: path.join('..', 'fixtures', 'outputFixture.js')
},
{
// As a convenience, this will have the indentation striped and it will
// be trimmed.
code: `
function sayHi(person) {
return 'Hello ' + person + '!';
}
`,
// This will take a Jest snapshot, overwriting the default/global
// settings (set above). The snapshot will contain both source code and
// the transformed output, making the snapshot file easier to understand.
snapshot: true
},
{
code: 'var hello = "hi";',
output: 'var olleh = "hi";',
// This can be used to overwrite pluginOptions (set above).
pluginOptions: {
optionA: false
}
// This can be used to overwrite presetOptions (set above).
//presetOptions: {
// optionB: true
//}
},
{
title: 'unchanged code',
code: '"no change";',
setup() {
// Runs before this test.
return function teardown() {
// Runs after this tests.
};
// Can also return a promise.
},
teardown() {
// Runs after this test.
// Can return a promise.
}
},
{
// This source will be transformed just like the code property, except the
// produced code will be evaluated in the same CJS context as the test
// runner. This lets us make more advanced assertions on the output.
exec: `
const hello = "hi";
// The plugin will reverse ALL identifiers, even globals like "expect"!
tcepxe(hello)['toBe']("hi");
`
}
]
});
See fixtures
for an example directory layout or check out the use of
babel-plugin-tester fixtures in some of these other projects:
- babel-plugin-transform-rewrite-imports
- babel-plugin-explicit-exports-references
- babel-plugin-transform-default-named-imports
This package was originally tested on and built to work with Jest, but it
is also tested against Vitest, Mocha, Jasmine, and
node:test
. See below for details.
All babel-plugin-tester features work with Jest. No further action is necessary π
All babel-plugin-tester features work with Vitest, though Vitest does not
provide global APIs by default. You can either supply some interoperability code
(see Jasmine or node:test
below for an example) or run Vitest with the
--globals
CLI option.
Most babel-plugin-tester features work with Mocha, except Mocha does not natively support snapshots.
Most babel-plugin-tester features work with Jasmine if you define the appropriate globals:
import { pluginTester } from 'babel-plugin-tester';
globalThis.it.skip = globalThis.xit;
globalThis.it.only = globalThis.fit;
pluginTester(...);
However, Jasmine does not natively support snapshots.
Most babel-plugin-tester features work with node:test
if you define the
appropriate globals:
import { describe, it } from 'node:test';
import { pluginTester } from 'babel-plugin-tester';
globalThis.describe = describe;
globalThis.it = it;
// globalThis.it.skip = ... (weirdly, this is already defined)
globalThis.it.only = (...args) => it(args[0], { only: true }, args[1]);
pluginTester(...);
However, node:test
does not natively support snapshots.
Other testing frameworks and test runners should also work so long as they
define standard describe
and it
globals with async support, or appropriate
interoperability code is used like in the above Jasmine and node:test
examples.
babelOptions.babelrc
and babelOptions.configFile
are set to
false
by default. This way, you can manually import (or provide an object
literal) the exact configuration you want to apply rather than relying on
babel's somewhat complex configuration loading rules. However, if your
plugin, preset, or project relies on a complicated external setup to do its
work, and you do not mind the default run order, you can leverage
babel's automatic configuration loading via the babelOptions.babelrc
and/or babelOptions.configFile
options.
Fixtures provided via the
fixtures
option do not need to provide a separatebabelOptions.filename
since it will be set automatically. This section only applies to test objects.
When relying on babelOptions.babelrc
, you must also provide a
babelOptions.filename
for each test object that does not include a
codeFixture
or execFixture
property. For example:
pluginTester({
plugin,
tests: [
{
code: '"blah"',
// This configuration is set at the test level
babelOptions: {
babelrc: true,
filename: path.join(__dirname, 'some-file.js')
}
},
{
code: '"hi"',
// This configuration is set at the test level
babelOptions: {
babelrc: true,
filename: path.join(__dirname, 'some-other-file.js')
}
},
{
// babelOptions.filename will be set to the value of codeFixture for you
// unless you set it manually here at the test level
codeFixture: path.join(__dirname, 'fixtures', 'my-file.js')
},
{
// babelOptions.filename will be set to the value of execFixture for you
// unless you set it manually here at the test level
execFixture: path.join(__dirname, 'fixtures', 'my-script.js')
}
]
});
This file does not actually have to exist either, so you can use whatever value
you want for filename
as long as the .babelrc
file is resolved
properly. Hence, the above example could be simplified further:
pluginTester({
plugin,
// This configuration is global: it applies to *all* tests by default!
babelOptions: {
babelrc: true,
// The value of filename does not have to point to a file that exists
filename: __filename
},
tests: [
'"blah"',
'"hi"',
{
// babelOptions.filename will be set to the value of codeFixture for you
// unless you set it manually here at the test level
codeFixture: path.join(__dirname, 'fixtures', 'my-file.js')
},
{
// babelOptions.filename will be set to the value of execFixture for you
// unless you set it manually here at the test level
execFixture: path.join(__dirname, 'fixtures', 'my-script.js')
}
]
});
Inferring pluginName
during testing requires invoking the plugin
at least twice: once outside of babel to check for the plugin's name and then
again when run by babel. This is irrelevant to babel-plugin-tester (even if your
plugin crashes when run outside of babel) and to the overwhelming majority of
babel plugins in existence. This only becomes a problem if your plugin is
aggressively stateful, which is against the babel handbook on plugin
design.
For example, the following plugin which replaces an import specifier using a regular expression will exhibit strange behavior due to being invoked twice:
/* -*-*- BAD CODE DO NOT USE -*-*- */
let source;
// vvv When first invoked outside of babel, all passed arguments are mocks vvv
function badNotGoodPlugin({ assertVersion, types: t }) {
// ^^^ Which means assertVersion is mocked and t is undefined ^^^
assertVersion(7);
// vvv So don't memoize `t` here (which among other things is poor design) vvv
if (!source) {
source = (value, original, replacement) => {
return t.stringLiteral(value.replace(original, replacement));
};
}
return {
name: 'bad-bad-not-good',
visitor: {
ImportDeclaration(path, state) {
path.node.source = source(
path.node.source.value,
state.opts.originalRegExp,
state.opts.replacementString
);
}
}
};
}
pluginTester({
plugin: badNotGoodPlugin,
pluginOptions: { originalRegExp: /^y$/, replacementString: 'z' },
tests: [{ code: 'import { x } from "y";', output: 'import { x } from "z";' }]
});
// Result: error!
// TypeError: Cannot read properties of undefined (reading 'stringLiteral')
If you still want to use global state despite the handbook's advice, either initialize global state within your visitor:
let source;
function okayPlugin({ assertVersion, types: t }) {
assertVersion(7);
return {
name: 'okay',
visitor: {
Program: {
enter() {
// vvv Initialize global state in a safe place vvv
if (!source) {
source = (value, original, replacement) => {
return t.stringLiteral(value.replace(original, replacement));
};
}
}
},
ImportDeclaration(path, state) {
path.node.source = source(
path.node.source.value,
state.opts.originalRegExp,
state.opts.replacementString
);
}
}
};
}
pluginTester({
plugin: okayPlugin,
pluginOptions: { originalRegExp: /^y$/, replacementString: 'z' },
tests: [{ code: 'import { x } from "y";', output: 'import { x } from "z";' }]
});
// Result: works!
Or do things the proper way and just use local state instead:
function betterPlugin({ assertVersion, types: t }) {
assertVersion(7);
// vvv Use local state instead so t is memoized properly vvv
const source = (value, original, replacement) => {
return t.stringLiteral(value.replace(original, replacement));
};
return {
name: 'better',
visitor: {
ImportDeclaration(path, state) {
path.node.source = source(
path.node.source.value,
state.opts.originalRegExp,
state.opts.replacementString
);
}
}
};
}
pluginTester({
plugin: betterPlugin,
pluginOptions: { originalRegExp: /^y$/, replacementString: 'z' },
tests: [{ code: 'import { x } from "y";', output: 'import { x } from "z";' }]
});
// Result: works!
If you are using Jest and snapshots, then the snapshot output could have a bunch
of bothersome \"
to escape quotes. This is because, when Jest serializes a
string, it will wrap everything in double quotes. This is not a huge deal, but
it makes the snapshots harder to read, so we automatically add a snapshot
serializer for you to remove those. Note that this serializer is added globally
and thus will affect all snapshots taken in the test file, even those outside
of babel-plugin-tester.
If you would like to disable this feature, then use the "pure" import (also disables formatting of babel output with prettier):
- import { pluginTester } from 'babel-plugin-tester'
+ import { pluginTester } from 'babel-plugin-tester/pure'
It seems recent versions of Jest already ship with easier-to-read snapshots, making this serializer redundant. Therefore, the built-in custom serializer will likely be removed entirely in a future version of babel-plugin-tester.
By default, a formatter is used which formats all babel output with prettier. It will look for a prettier configuration file relative to the file that's being tested or the current working directory. If it cannot find one, then it uses the default configuration for prettier.
This makes your snapshots easier to read and your expectations easier to write,
but if you would like to disable this feature, you can either use the pure
import to disable automatic formatting (along with snapshot serialization)
or you can override the formatResult
option manually like so:
pluginTester({
// ...
formatResult: (r) => r
// ...
});
This package uses debug under the hood. To view all possible debugging
output, including the results of all babel transformations, set the
DEBUG='babel-plugin-tester,babel-plugin-tester:*'
environment variable
when running your tests.
For example:
# Those using Windows (but not WSL) have to set environment variable differently
NODE_ENV='test' DEBUG='babel-plugin-tester,babel-plugin-tester:*' DEBUG_DEPTH='1' npx jest
The following debug namespaces are available for activation:
babel-plugin-tester:index
babel-plugin-tester:formatter
babel-plugin-tester:serializer
babel-plugin-tester:tester
babel-plugin-tester:tester:resolve-base
babel-plugin-tester:tester:resolve-env
babel-plugin-tester:tester:normalize
babel-plugin-tester:tester:normalize:create-desc
babel-plugin-tester:tester:normalize:create-fix
babel-plugin-tester:tester:normalize:create-obj
babel-plugin-tester:tester:register
babel-plugin-tester:tester:wrapper
babel-plugin-tester:tester:test
babel-plugin-tester:tester:validate
babel-plugin-tester:tester:read-opts
babel-plugin-tester:tester:read-code
babel-plugin-tester:tester:eol
babel-plugin-tester:tester:finalize
The babel-plugin-tester:tester
namespace and its sub-namespaces each have an
additional verbose
sub-namespace that can be activated or deactivated at will,
e.g. babel-plugin-tester:tester:verbose
and
babel-plugin-tester:tester:wrapper:verbose
.
For example, to view all debug output except verbose output:
# Those using Windows (but not WSL) have to set environment variable differently
NODE_ENV='test' DEBUG='babel-plugin-tester,babel-plugin-tester:*,-*:verbose' npx jest
The optional TEST_ONLY
and TEST_SKIP
environment variables are recognized by
babel-plugin-tester, allowing you to control which tests are run in an adhoc
fashion without having to modify your test configuration code.
The values of these variables will be transformed into regular expressions via
RegExp(value, 'u')
and matched against each test/fixture title (not including
automatically assigned numbers prefixed to titles). Tests with titles that
match TEST_ONLY
will be run while all others are skipped. On the other hand,
tests with titles that match TEST_SKIP
will be skipped while others are run.
For example, to skip the test titled "this is the name of a failing unit test":
TEST_SKIP='name of a failing' npx jest
Given both TEST_ONLY
and TEST_SKIP
, tests matched by TEST_SKIP
will
always be skipped, even if they are also matched by TEST_ONLY
. These
environment variables also override both the fixture-specific
skip
/only
and test object skip
/only
properties.
In addition to TEST_ONLY
and TEST_SKIP
, you can also target tests
specifically by their automatically assigned number using TEST_NUM_ONLY
and TEST_NUM_SKIP
. These environment variables function identically to their
counterparts except they accept one or more numbers separated by commas (spaces
around commas are ignored) instead of regular expressions. Inclusive ranges
(e.g. 4-9
) are also supported.
For example, the following will skip tests numbered 1, 3, 5, and 6-10 (inclusive):
# Spaces around commas don't matter and sequential/final commas are ignored
TEST_NUM_SKIP='5,1, 6-10,, 3,' npx jest
TEST_NUM_ONLY
and TEST_NUM_SKIP
are meaningless if titleNumbering
is
false
or your tests are otherwise unnumbered, and may match multiple tests if
automatic numbering is restarted.
For each test object and fixture test, setup and teardown functions are run in the following order:
- Base
setup
. - Test object
setup
/ fixturesetup
. - Test object / fixture test is run.
- Any function returned by test object
setup
/ fixturesetup
. - Test object
teardown
/ fixtureteardown
. - Any function returned by base
setup
. - Base
teardown
.
The API was inspired by:
- ESLint's RuleTester.
- @thejameskyle's tweet.
- Babel's own
@babel/helper-plugin-test-runner
.
Looking to contribute? Look for the Good First Issue label.
Please file an issue for bugs, missing documentation, or unexpected behavior.
Please file an issue to suggest new features. Vote on feature requests by adding a π. This helps maintainers prioritize what to work on.
Thanks goes to these people (emoji key):
This project follows the all-contributors specification. Contributions of any kind welcome!
MIT