Syntax
phil/css-sanitizerThe tokenizer, the parser and the serializer are public and can be used without the sanitizer: to read CSS as a browser does, to check its structure, or to write it back.
Tokenizer
Syntax\Tokenizer::tokenize() follows CSS Syntax Level 3, section 4. The input is preprocessed — CR LF, CR and FF become
LF, NULL becomes U+FFFD — and bytes that are not UTF-8 become U+FFFD. It returns a list of Token, the last one of
type Eof. Comments are dropped.
use Phil\CssSanitizer\Syntax\Tokenizer;
foreach (new Tokenizer()->tokenize('a { width: 10px }') as $token) {
echo $token->type->value, ' ';
}
// ident whitespace { whitespace ident colon whitespace dimension whitespace } EOFA Token has a type (TokenType), a value with its escapes resolved — the name of an ident, function, at-keyword
or hash, the text of a string or URL, the character of a delim, the representation of a number as written —, the
unit of a dimension, isId for a hash that could be an identifier, isInteger for a number, and its
sourceLine.
Parser
Syntax\Parser follows section 5:
| Method | Returns |
|---|---|
parseStylesheet($css) | A Stylesheet of QualifiedRule, AtRule and Malformed |
parseDeclarationList($css) | Declaration, AtRule and Malformed, as in a style attribute |
parseComponentValues($css) | Component values: Token, SimpleBlock, FunctionValue |
rules($values, $topLevel) | The rules of component values, such as a block's |
declarations($values) | The declarations of component values, such as a block's |
blockContents($values) | Declaration, AtRule, QualifiedRule and Malformed, as CSS Nesting reads a style rule's block |
A rule's block keeps its component values; whether they hold rules or declarations depends on the rule, so the
caller reads them with rules(), declarations() or blockContents(). blockContents() reads an item that starts
with a name as a declaration unless it has no colon, or a {} block beside other values in a property that is not
custom; then it reads a nested rule, whose prelude ends at its block or, as a parse error, at a semicolon. It reads a
block in time linear in its length. What the specification drops as a parse error — a declaration
without its colon, a rule cut by the end of the source — is kept as Malformed, with its content and line.
new Parser(maxDepth: 64) bounds the nesting of blocks and functions; deeper, it throws NestingTooDeep. The static
Parser::trim() takes white space off the ends of component values, and Parser::splitAtCommas() cuts them at the
commas of their top level.
Serializer
Serializer writes CSS back:
| Method | Writes |
|---|---|
componentValues($values) | Component values |
declarations($declarations) | name: value; …, as a style attribute holds them |
styleSheet($sheet) | A Sheet\StyleSheet: its @import rules, then one rule after the other, two spaces of indentation per level, a rule's declarations before its nested rules |
Serializer::identifier($name) | A name, escaped as CSSOM serializes identifiers |
Serializer::string($text) | A string between double quotes |
Numbers keep the representation they were written with. An empty comment goes between two tokens that would read
back as other tokens — an ident and a number, a number and a unit-like ident —, as section 9 requires, except where a
signed number cannot join what comes before it, so that U+0000-00FF stays as written. Reading the output again gives
the same values. A < in a name or a string is written \3c , and a < delim is kept apart from a / that follows
it: the output never starts an end tag.