postcss

Framework for CSS postprocessors

npm install postcss
1 739 downloads in the last day
8 938 downloads in the last week
36 204 downloads in the last month

PostCSS

PostCSS is a framework for CSS postprocessors, to modify CSS with JavaScript.

It takes care of most common CSS tool tasks:

  1. parses CSS;
  2. gives you usable JS API to edit CSS node tree;
  3. dumps modified node tree into CSS string;
  4. generates (or modifies existent) source map for your changes;

You can use this framework to write you own:

  • CSS minifier or beautifizer.
  • CSS polyfills.
  • Grunt plugin to generate sprites, include data-uri images or any other works.
  • Text editor plugin to automate CSS routine.
  • Command-line CSS tool.

Sponsored by Evil Martians.

Built with PostCSS

Quick Example

Let’s fix forgotten content property in ::before and ::after:

var postcss = require('postcss');

var contenter = postcss(function (css) {
    css.eachRule(function (rule) {
        if ( rule.selector.match(/::(before|after)/) ) {
            // In every ::before/::after rule

            // Did we forget content property?
            var good = rule.some(function (i) { return i.prop == 'content'; });

            if ( !good ) {
                // Add content: "" if we forget it
                rule.prepend({ prop: 'content', value: '""' });
            }

        }
    });
});

And then CSS with forgotten content:

a::before {
    width: 10px;
    height: 10px
}

will be fixed by our new contenter:

var fixed = contenter.process(css).css;

to:

a::before {
    content: "";
    width: 10px;
    height: 10px
}

Features

Source Map

PostCSS generates source map for its changes:

result = processor.process(css, { map: true, from: 'from.css', to: 'to.css' });
result.css // String with processed CSS
result.map // Source map

And modifies source map from previous step (like Sass preprocessor):

var sassMap = fs.readFileSync('from.sass.css.map');
processor.process(css, { map: sassMap, from: 'from.sass.css', to: 'to.css' });

Preserves code formatting and indentations

PostCSS will not change any byte of a rule if you don’t modify its node:

postcss(function (css) { }).process(css).css == css;

And when you modify CSS nodes, PostCSS will try to copy coding style:

contenter.process("a::before{color:black}")
// a::before{content:'';color:black}

contenter.process("a::before {\n  color: black;\n  }")
// a::before {
//   content: '';
//   color: black;
//   }

Why PostCSS Better Than …

Preprocessors

Preprocessors (like Sass or Stylus) give us special language with variables, mixins, statements and compile it to CSS. Compass, nib and other mixins libraries use these languages to work with prefixes, sprites and inline images.

But Sass and Stylus languages were created to be syntax-sugar for CSS. Writing really complicated programs using preporcessor languages is very difficult. Autoprefixer is absolutely impossible to implement on top of Sass.

PostCSS gives you comfort and power of JS or CoffeeScript to working with CSS. You can do really magic things with wide range of npm libraries.

But postprocessors are not enemies for preprocessors. Sass and Stylus are still the best way to improve readability and add some syntax sugar to CSS. You can easily combine preprocessors and postprocessors (and PostCSS will also update source map from Sass or Stylus).

RegExp

Some Grunt plugins modify CSS with regular expressions but using a CSS parser and a node tree is a much safer way to edit CSS. Also, regexps will break source maps generated by preprocessors.

CSS Parsers

There are a lot of good CSS parsers, like Gonzales. But they help you only with first step.

Unlike them PostCSS gives you full source map support and useful high level API (for example, safe iterators).

Rework

Rework and PostCSS are very similar, but they has different targets.

Rework was created to build new CSS sublanguage to replace Stylus (like Myth). PostCSS was created for CSS tools, which works in chain with legacy CSS code (like Autoprefixer).

Because of this background difference, PostCSS:

  • better works with source map, because it should update map from previous step (like Sass compiling)
  • saves all your spaces and code style, because it can be worked in text editor plugins
  • has safer parser, because it can be used for legacy code
  • has high level API to clean your processor from common tasks

Usage

Processor

The function postcss(fn) creates a processor from your function:

var postcss = require('postcss');

var processor = postcss(function (css) {
    // Code to modify CSS
});

If you want to combine multiple processors (and parse CSS only once), you can add several functions using the use(fn) method:

var all = postcss().
          use(prefixer).
          use(minifing);

Processor function can change the current CSS node tree:

postcss(function (css) {
    css.append( /* new rule */ )
});

or create a completely new CSS root node and return it instead:

postcss(function (css) {
    var newCSS = postcss.root()
    // Add rules and declarations
    return newCSS;
});

This generated processor transforms some CSS using process(css, opts) method:

var doubler = postcss(function (css) {
    // Clone each declaration
    css.eachDecl(function (decl) {
        decl.parent.prepend( decl.clone() );
    });
});

var css    = "a { color: black; }";
var result = doubler.process(css);

result.css //=> "a { color: black; color: black; }"

You can set the original CSS filename via from option and make syntax error messages much more helpful:

var wrong = "a {";
processor.process(wrong, { from: 'main.css' });
//=> Can't parse CSS: Unclosed block at line 1:1 in main.css

Multiple Inputs

The function postcss() generates processor only for one input. If you need to process several inputs (like in files concatenation) you can use postcss.parse().

Let’s join two CSS with source map support in 5 lines of code:

var file1 = postcss.parse(css1, { from: 'a.css' });
var file2 = postcss.parse(css2, { from: 'b.css' });

file1.rules = file1.rules.concat( file2.rules );

var result = file1.toResult({ to: 'app.css', map: true });

Source Map

PostCSS will generate a source map, if you set map option to true in the process(css, opts) method.

You must set input and output CSS files paths (using from and to options respectively) to generate correct source map.

var result = processor.process(css, {
    map:  true,
    from: 'main.css',
    to:   'main.out.css'
});

result.map //=> '{"version":3,"file":"main.out.css","sources":["main.css"],"names":[],"mappings":"AAAA,KAAI"}'

fs.writeFileSync('main.out.css.map', result.map);

PostCSS can also modify previous source map (for example, from Sass compilation). So if you compile Sass to CSS and then minify this CSS by postprocessor, final source map will contain mapping from Sass code to minified CSS.

Just set map option with an original source map (a string or a JS object):

var result = minifier.process(css, {
    map:   fs.readFileSync('main.sass.css.map'),
    from: 'main.sass.css',
    to:   'main.min.css'
});

result.map //=> Source map from main.sass to main.min.css

PostCSS try to autodetect previous map file. For example, if you process a.css and a.css.map is placed in same dir, PostCSS will read previous map and generate new one. You can disable autodetection by map: false.

PostCSS, by default, will add annotation comment with path to new source map file:

a { }
/*# sourceMappingURL=main.out.css.map */

If you want to remove annotation, set mapAnnotation option to false.

Inline maps are also supported. If input CSS will contain annotation from previous step with map in data:uri, PostCSS will update source map with yours changes and inine new map back to output CSS.

Option inlineMap will force PostCSS to inline new map to CSS:

var result = minifier.process(css, {
    from:     'main.css',
    to:       'main.min.css',
    inlineMap: true
});

result.map //=> undefined, because map is in CSS
result.css //=> "a{}\n/*# sourceMappingURL=data:application/json;base64,eyJ2ZXJzaW9uIjozLCJmaWxlIjoibWFpbi5taW4uY3NzIiwic291cmNlcyI6WyJtYWluLmNzcyJdLCJuYW1lcyI6W10sIm1hcHBpbmdzIjoiQUFBQSxJQUFLIn0= */"

Helpers

Vendor

PostCSS contains heigh optimized code to split vendor prefix:

var vendor = require('postcss/lib/vendor');

vendor.prefix('-moz-tab-size')     //=> '-moz-'
vendor.unprefixed('-moz-tab-size') //=> 'tab-size'

List

To safely split comma- or space-separated values (like in background-image or transform ) with brackets and quotes support you can use list helper:

var list = require('postcss/lib/list');

list.space(image.value)     //=> ['linear-gradient(white, black)', 'blue']
list.comma(transform.value) //=> ['color 200ms', 'background 200ms']

Nodes

Processor function receives Root node with CSS node tree inside.

var processor = postcss(function (cssRoot) {
});

There are 4 types of child nodes: Comment, AtRule, Rule and Declaration. All nodes have toString() and clone() methods.

You can parse CSS and get a Root node by postcss.parse(css, opts) method:

var cssRoot = postcss.parse('a { }');

All node‘s methods return current node, so you can build nice method chains:

root.append( rule1 ).append( rule2 ).toString();

Node Source

Every node stores its origin file (if you set from option to process or parse method) and position:

var root = postcss.parse(css, { from: 'main.css' });
var rule = root.rules[0];

rule.source.file  //=> 'main.css'
rule.source.start //=> { line: 5,  position: 1 }
rule.source.end   //=> { line: 10, position: 5 }

Whitespaces

All nodes (exclude Root) have before property with indentation and all earlier spaces.

Nodes with children (Root, AtRule and Rule) contain also after property with spaces after last child and before } or end of file.

Every Declaration has between property with colon, spaces and comments between property name and value. Rule stores spaces and comments between selector and { in between property. AtRule uses between also to store spaces and comments before { or ; for bodiless at-rule.

var root = postcss.parse("a {\n  color: black;\n}\n");

root.rules[0].between          //=> " " between selector and {
root.rules[0].decls[0].before  //=> "\n  " before color: black
root.rules[0].decls[0].between //=> ": " between property name and value
root.rules[0].after            //=> "\n" before }
root.after                     //=> "\n" from end of file

The simplest way to minify CSS is to set before, between and after properties to an empty string:

var minifier = postcss(function (css) {
    css.eachDecl(function (decl) {
        decl.before  = '';
        decl.between = ':';
    });
    css.eachRule(function (rule) {
        rule.before  = '';
        rule.between = '';
        rule.after   = '';
    });
    css.eachAtRule(function (atRule) {
        atRule.before  = '';
        atRule.between = '';
        atRule.after   = '';
    });
    css.eachComment(function (comment) {
        comment.removeSelf();
    });
});

var css = "a {\n  color:black\n}\n";
minifier.process(css).css //=> "a{color:black}"

Raw Properties

Some CSS values (selectors, comment text, at-rule params and declaration values) can contain comments. PostCSS will clean them from trailing spaces for you:

var root = postcss.parse("a /**/ b {}");
var rule  = root.rules[0];

rule.selector      //=> 'a  b' trimmed and cleaned from comments
rule._selector.raw //=> 'a /**/ b' original raw value

But PostCSS saves raw content to be able to stringify it to CSS, if you don’t change origin value. As you can remember, PostCSS tries to save origin CSS byte-to-byte, when it’s possible:

rule.toString() //=> 'a /**/ b {}' with comment

rule.selector = '.link b';
rule.toString() //=> '.link b {}' you change value and origin comment was gone

Containers

Root, AtRule and Rule nodes can contain children in rules or decls property.

There are common method to work with children:

  • append(newChild) to add child at the end of children list.
  • prepend(newChild) to add child at the beginning of children list.
  • insertBefore(existsChild, newChild) to insert new child before some existent child.
  • insertAfter(existsChild, newChild) to insert new child after some existent child.
  • remove(existsChild) to remove child.
  • index(existsChild) to return child index.
  • some(fn) to return true if fn returns true on any child.
  • every(fn) to return true if fn returns true on all children.

Methods insertBefore, insertAfter and remove can receive child node or child index as an existsChild argument. Have in mind that child index works much faster.

There are two shorcuts to get first and last child:

rule.first //=> First declaration in rule
rule.last  //=> Last declaration in rule

Children

Comment, AtRule, Rule and Declaration nodes should be wrapped in other nodes.

All children contain parent property with parent node:

rule.decls[0].parent == rule;

All children has removeSelf() method:

rule.decls[0].removeSelf();

But remove(index) in parent with child index is much faster:

rule.each(function (decl, i) {
    rule.remove(i);
});

Iterators

All parent nodes have each method to iterate over children nodes:

root = postcss.parse('a { color: black; display: none }');

root.each(function (rule, i) {
    if ( rule.type == 'rule' ) {
        console.log(rule.selector, i); // Will log "a 0"
    }
});

root.rules[0].each(function (decl, i) {
    if ( rule.type != 'comment' ) {
        console.log(decl.prop, i); // Will log "color 0" and "display 1"
    }
});

Unlike for {}-cycle construct or Array#forEach() this iterator is safe. You can mutate children while iteration and it will fix current index:

rule.rules.forEach(function (decl, i) {
    rule.prepend( decl.clone() );
    // Will be infinity cycle, because on prepend current declaration become
    // second and next index will go to current declaration again
});

rule.each(function (decl, i) {
    rule.prepend( decl.clone() );
    // Will work correct (once clone each declaration), because after prepend
    // iterator index will be recalculated
});

Because CSS have nested structure, PostCSS also contains recursive iterator eachInside:

root.eachInside(function (node, i) {
    console.log(node.type ' inside ' + parent.type);
});

There are also shortcuts to recursive iterate all nodes of specific type:

root.eachDecl(function (decl, i) {
    // Each declaration inside root
});

root.eachRule(function (rule, i) {
    // Each rule inside root and any nested at-rules
});

root.eachAtRule(function (atRule, i) {
    // Each at-rule inside root and any nested at-rules
});

root.eachComment(function (comment, i) {
    // Each comment inside root
})

Root Node

Root node contains entire CSS tree. Its children can be only Comment, AtRule or Rule nodes in rules property.

You can create a new root using shortcut:

var root = postcss.root();

Method toString() stringifies entire node tree to CSS string:

root = postcss.parse(css);
root.toString() == css;

Comment Node

/* Block comment */

PostCSS creates Comment nodes only for comments between rules or declarations. Comments inside selectors, at-rules params, declaration values will be stored in Raw property.

Comment has only one property: text with trimmed text inside comment.

comment.text //=> "Block comment"

You can create a new comment using shortcut:

var comment = postcss.comment({ text: 'New comment' });

AtRule Node

@charset 'utf-8';

@font-face {
    font-family: 'Cool'
}

@media print {
    img { display: none }
}

AtRule has two own properties: name and params.

As you see, some at-rules don’t contain any children (like @charset or @import), some of at-rules can contain only declarations (like @font-face or @page), but most of them can contain rules and nested at-rules (like @media, @keyframes and others).

Parser selects AtRule content type by its name. If you create AtRule node manually, it will detect own content type with new child type on first append or other add method call:

var atRule = postcss.atRule({ name: '-x-animations' });
atRule.rules        //=> undefined
atRule.decls        //=> undefined

atRule.append( postcss.rule({ selector: 'from' }) );
atRule.rules.length //=> 1
atRule.decls        //=> undefined

You can create a new at-rule using shortcut:

var atRule = postcss.atRule({ name: 'charset', params: 'utf-8' });

Rule Node

a {
    color: black;
}

Rule node has selector property and contains Declaration and Comment children in decls property.

There is selectors shortcut, which return array:

rule.selector  //=> "a, b"
rule.selectors //=> ['a', 'b']

You can miss Declaration constructor in append and other insert methods:

rule.append({ prop: 'color', value: 'black' });

Property semicolon indicates if last declaration in rule has semicolon or not:

var root = postcss.parse('a { color: black }');
root.rules[0].semicolon //=> false

var root = postcss.parse('a { color: black; }');
root.rules[0].semicolon //=> true

You can create a new rule using shortcut:

var rule = postcss.rule({ selector: 'a' });

Declaration Node

color: black

Declaration node has prop, value and important properties.

You can create a new declaration using this shortcut:

var decl = postcss.decl({ prop: 'color', value: 'black' });

Or use short form in rule’s append() and other add methods:

rule.append({ prop: 'color', value: 'black' });
npm loves you