A PostCSS plugin to lint BEM-style CSS.
BEM-style describes CSS that follows a more-or-less strict set of conventions determining what selectors can be used. Typically, these conventions require that classes begin with the name of the component (or "block") that contains them, and that all characters after the component name follow a specified pattern. Original BEM methodology refers to "blocks", "elements", and "modifiers"; SUIT refers to "components", "descendants", and "modifiers". You might have your own terms for similar concepts.
With this plugin, you can check the validity of selectors against a set of BEM-style conventions. You can use preset patterns (SUIT and BEM, currently) or insert your own. The plugin will register warnings if it finds CSS that does not follow the specified conventions.
npm install postcss-bem-linter
Version 1.0.0+ is compatible with PostCSS 5+. (Earlier versions are compatible with PostCSS 4.)
This plugin registers warnings via PostCSS. Therefore, you'll want to use it with a PostCSS runner that prints warnings (e.g. gulp-postcss
) or another PostCSS plugin that prints warnings (e.g. postcss-reporter
).
Throughout this document, terms like "selector", "selector sequence", and "simple selector" are used according to the definitions in the Selectors Level 3 spec.
postcss-bem-linter can also be used as a stylelint plugin: stylelint-selector-bem-pattern.
By using the stylelint plugin, all of your linting can happen in one step, seamlessly: postcss-bem-linter warnings will output alongside other stylelint warnings. Also, you can take advantage of all the other features that stylelint offers, such as a CLI and Node.js API, different formatters for output, etc.
Default mode:
- Only allow selector sequences that match the defined convention.
- Only allow custom-property names that begin with the defined
ComponentName
.
Weak mode:
- While initial selector sequences (before combinators) must match the defined convention, sequences after combinators are not held to any standard.
Prior to 0.5.0, this plugin checked two other details: that :root
rules only contain custom-properties; and that the :root
selector is not grouped or combined with other selectors. These checks can now be performed by stylelint. So from 0.5.0 onwards, this plugin leaves that business to stylelint to focus on its more unique task.
bemLinter([pattern[, options]])
Patterns consist of regular expressions, and functions that return regular expressions, or strings, which describe valid selector sequences.
Keep in mind:
- Patterns describe sequences, not just simple selectors. So if, for example,
you would like to be able to chain state classes to your component classes, as in
.Component.is-open
, your pattern needs to allow for this chaining. - Pseudo-classes and pseudo-elements will be ignored if they occur at the end of the sequence.
Instead of
.Component:first-child.is-open
, you should use.Component.is-open:first-child
. The former will trigger a warning unless you've written a silly complicated pattern.
The following preset patterns are available:
'suit'
(default), as defined here. Options:namespace
: a namespace to prefix valid classes, as described in the SUIT docs
'bem'
, as defined here.namespace
: a namespace to prefix valid classes, to be separated from the block name with a hyphen, e.g. with namespacefoo
,.foo-dropdown__menu
.
You can use a preset pattern and its options in two ways:
- Pass the preset's name as the first argument, and, if needed, an
options
object as the second, e.g.bemLinter('suit', { namespace: 'twt' })
. - Pass an object as the first and only argument, with the preset's name as the
preset
property and, if needed,presetOptions
, e.g.bemLinter({ preset: 'suit', presetOptions: { namespace: 'twt' })
.
'suit'
is the default pattern; so if you do not pass any pattern
argument,
SUIT conventions will be enforced.
You can define a custom pattern by passing as your first and only argument an object with the following properties:
default: /^[-_a-zA-Z0-9]+$/
Describes valid component names in one of the following forms:
- A regular expression.
- A string that provides a valid pattern for the
RegExp()
constructor.
Describes all valid selector sequences for the stylesheet in one of the following forms:
-
A single function that accepts a component name and returns a regular expression, e.g.
componentSelectors: function(componentName) { return new RegExp('^\\.ns-' + componentName + '(?:-[a-zA-Z]+)?$'); }
-
A single string that provides a valid pattern for the
RegExp()
constructor when{componentName}
is interpolated with the defined component's name, e.g.componentSelectors: '^\\.ns-{componentName}(?:-[a-zA-Z]+)?$'
-
An object consisting of two properties,
initial
andcombined
. Both properties accept the same two forms described above: a function accepting a component name and returning a regular expression; or a string, interpolating the component name with{componentName}
, that will provide a valid pattern for theRegExp()
constructor.initial
describes valid initial selector sequences — those occurring at the beginning of a selector, before any combinators.combined
describes valid selector sequences allowed after combinators. Two important notes aboutcombined
:- If you do not specify a
combined
pattern, it is assumed that combined sequences must match the same pattern as initial sequences. - In weak mode, any combined sequences are accepted, even if you have a
combined
pattern.
- If you do not specify a
Describes valid utility selector sequences. This will be used if the stylesheet defines a group of utilities, as explained below. Can take one of the following forms:
- A regular expression.
- A string that provides a valid pattern for the
RegExp()
constructor.
Describes selector sequences to ignore. You can use this to
systematically ignore selectors matching certain patterns, instead of having to add a
/* postcss-bem-linter: ignore */
comment above each one (see below).
Can take one of the following forms:
- A regular expression.
- An array of regular expressions.
- A string that provides a valid pattern for the
RegExp()
constructor. - An array of such string patterns.
Describes custom properties to ignore. Works the same as ignoreSelectors
, above, so please read about that.
You can also choose a preset to start with and override specific parts of it, specific patterns.
For example, if you want to use SUIT's preset generally but write your own utilitySelectors
pattern,
you can do that with a config object like this:
{
preset: 'suit',
utilitySelectors: /^\.fancyUtilities-[a-z]+$/
}
Given all of the above, you might call the plugin in any of the following ways:
// use 'suit' conventions
bemLinter();
bemLinter('suit');
bemLinter('suit', { namespace: 'twt' });
bemLinter({ preset: 'suit', presetOptions: { namespace: 'twt' }});
// use 'bem' conventions
bemLinter('bem');
bemLinter('bem', { namespace: 'ydx' });
bemLinter({ present: 'bem', presetOptions: { namespace: 'ydx' }});
// define a pattern for component names
bemLinter({
componentName: /^[A-Z]+$/
});
bemLinter({
componentName: '^[A-Z]+$'
});
// define a single pattern for all selector sequences, initial or combined
bemLinter({
componentSelectors: function(componentName) {
return new RegExp('^\\.' + componentName + '(?:-[a-z]+)?$');
}
});
bemLinter({
componentSelectors: '^\\.{componentName}(?:-[a-z]+)?$'
});
// define separate `componentName`, `initial`, `combined`, and `utilities` patterns
bemLinter({
componentName: /^[A-Z]+$/,
componentSelectors: {
initial: function(componentName) {
return new RegExp('^\\.' + componentName + '(?:-[a-z]+)?$');
},
combined: function(componentName) {
return new RegExp('^\\.combined-' + componentName + '-[a-z]+$');
}
},
utilitySelectors: /^\.util-[a-z]+$/
});
bemLinter({
componentName: '^[A-Z]+$',
componentSelectors: {
initial: '^\\.{componentName}(?:-[a-z]+)?$',
combined: '^\\.combined-{componentName}-[a-z]+$'
},
utilitySelectors: '^\.util-[a-z]+$'
});
// start with the `bem` preset but include a special `componentName` pattern
// and `ignoreSelectors` pattern to ignore Modernizr-injected `no-*` classes
bemLinter({
preset: 'bem',
componentName: /^cmpnt_[a-zA-Z]+$/,
ignoreSelectors: /^\.no-.+$/
});
bemLinter({
preset: 'bem',
componentName: '^cmpnt_[a-zA-Z]+$',
ignoreSelectors: '^\.no-.+$'
});
// ... using an array for `ignoreSelectors`
bemLinter({
preset: 'bem',
componentName: /^cmpnt_[a-zA-Z]+$/,
ignoreSelectors: [
/^\.no-.+$/,
/^\.isok-.+$/
]
});
bemLinter({
preset: 'bem',
componentName: '^cmpnt_[a-zA-Z]+$',
ignoreSelectors: [
'^\.no-.+$',
'^\.isok-.+$'
]
});
The plugin will only run if it finds special comments that define a named component or a group of utilities.
These definitions can be provided in two syntaxes: concise and verbose.
- Concise definition syntax:
/** @define ComponentName */
or/** @define utilities */
- Verbose definition syntax:
/* postcss-bem-linter: define ComponentName */
or/* postcss-bem-linter: define utilities */
.
Weak mode is turned on by adding ; weak
to a definition,
e.g. /** @define ComponentName; weak */
or /* postcss-bem-linter: define ComponentName; weak */
.
Concise syntax:
/** @define MyComponent */
:root {
--MyComponent-property: value;
}
.MyComponent {}
.MyComponent-other {}
Verbose syntax:
/** postcss-bem-linter: define FancyComponent */
:root {
--FancyComponent-property: value;
}
.FancyComponent {}
.FancyComponent-other {}
Weak mode:
/** @define MyComponent; weak */
:root {
--MyComponent-property: value;
}
.MyComponent {}
.MyComponent .other {}
Utilities:
/** @define utilities */
.u-sizeFill {}
.u-sm-horse {}
If a component is defined and the component name does not match your componentName
pattern,
the plugin will throw an error.
It's recommended that you keep each defined group of rules in a distinct file, with the definition at the top of the file. If, however, you have a good reason for multiple definitions within a single file, you can do that.
Successive definitions override each other. So the following works:
/* @define Foo */
.Foo {}
/* @define Bar */
.Bar {}
/* @define utilities */
.u-something {}
You can also deliberately end the enforcement of a definition with the following special comments:
/* @end */
or /* postcss-bem-linter: end */
.
/* @define Foo */
.Foo {}
/* @end */
.something-something-something {}
One use-case for this functionality is when linting files after concatenation performed by a CSS processor like Less or Sass, whose syntax is not always compatible with PostCSS. See issue #57.
If you need to ignore a specific selector but do not want to ignore the entire stylesheet or end the enforcement of a definition, there are two ways to accomplish this:
As describe above, you can include an ignoreSelectors
regular expression (or array of regular expressions) in your configuration.
This is the best approach if you want to systematically ignore all selectors matching a pattern (e.g. all Modernizr classes).
If you just want to ignore a single, isolated selector, though,
you can do so by preceding the selector with this comment: /* postcss-bem-linter: ignore */
.
/** @define MyComponent */
.MyComponent {
display: flex;
}
/* postcss-bem-linter: ignore */
.no-flexbox .Component {
display: block;
}
The comment will cause the linter to ignore only the very next selector.
Pass your individual CSS files through the plugin. It will register warnings for
conformance failures, which you can print to the console using
postcss-reporter
or relying
on a PostCSS runner (such as gulp-postcss
).
var postcss = require('postcss');
var bemLinter = require('postcss-bem-linter');
var reporter = require('postcss-reporter');
files.forEach(function(file) {
var css = fs.readFileSync(file, 'utf-8');
postcss()
.use(bemLinter())
.use(reporter())
.process(css)
.then(function(result) { .. });
});
Please note that this project is released with a Contributor Code of Conduct. By participating in this project you agree to abide by its terms.
Install the dependencies.
npm install
Run the tests.
npm test
Watch and automatically re-run the unit tests.
npm start