wikimedia/css-sanitizer

Classes to parse and sanitize CSS

v1.0.2 2017-06-13 15:51 UTC

README

Latest Stable Version License

Wikimedia CSS Parser & Sanitizer

This library implements a CSS tokenizer, parser and grammar matcher in PHP that mostly follows the CSS Syntax Module Level 3 candidate recommendation dated 20 February 2014, the CSS Values and Units Module Level 3, and the CSS Selectors Level 3 grammar. It also provides a sanitizer that recognizes various CSS3 modules.

Usage

use Wikimedia\CSS\Parser\Parser;
use Wikimedia\CSS\Sanitizer\StylesheetSanitizer;

/** Parse a stylesheet from a string **/

$parser = Parser::newFromString( $cssText );
$stylesheet = $parser->parseStylesheet();

/** Report any parser errors **/

foreach ( $parser->getParseErrors() as list( $code, $line, $pos ) ) {
	// $code is a string that should be suitable as a key for an i18n library.
	// See errors.md for details.
	$error = lookupI18nMessage( "css-parse-error-$code" );
	echo "Parse error: $error at line $line character $pos\n";
}

/** Apply sanitization to the stylesheet **/

// If you need to customize the defaults, copy the code of this method and
// modify it.
$sanitizer = StylesheetSanitizer::newDefault();
$newStylesheet = $sanitizer->sanitize( $stylesheet );

/** Report any sanitizer errors **/

foreach ( $sanitizer->getSanitizationErrors() as list( $code, $line, $pos ) ) {
	// $code is a string that should be suitable as a key for an i18n library.
	// See errors.md for details.
	$error = lookupI18nMessage( "css-sanitization-error-$code" );
	echo "Sanitization error: $error at line $line character $pos\n";
}

/** Convert the sanitized stylesheet back to text **/

$newText = (string)$newStylesheet;

// Or if you'd rather have it minified too
$minifiedText = Wikimedia\CSS\Util::stringify( $newStylesheet, [ 'minify' => true ] );

Running tests

composer install --prefer-dist
composer test

History

We required a CSS sanitizer with several properties:

  • Strict parsing according to modern standards.
  • Includes line and character position for all errors.
  • Configurable to limit unsafe constructs such as external URL references.
  • Errors are easily localizable.

We could not find a library that fit these requirements, so we created one.