Code Coverage |
||||||||||
Lines |
Functions and Methods |
Classes and Traits |
||||||||
| Total | |
91.03% |
71 / 78 |
|
57.14% |
4 / 7 |
CRAP | |
0.00% |
0 / 1 |
| Parser | |
91.03% |
71 / 78 |
|
57.14% |
4 / 7 |
40.10 | |
0.00% |
0 / 1 |
| __construct | |
100.00% |
1 / 1 |
|
100.00% |
1 / 1 |
1 | |||
| parseDocument | |
100.00% |
7 / 7 |
|
100.00% |
1 / 1 |
1 | |||
| looksLikeXhtml | |
100.00% |
8 / 8 |
|
100.00% |
1 / 1 |
2 | |||
| mirrorSelectedContent | |
94.74% |
18 / 19 |
|
0.00% |
0 / 1 |
10.01 | |||
| findFirstDescendant | |
100.00% |
10 / 10 |
|
100.00% |
1 / 1 |
6 | |||
| findSelectedOption | |
94.44% |
17 / 18 |
|
0.00% |
0 / 1 |
10.02 | |||
| parseFragment | |
66.67% |
10 / 15 |
|
0.00% |
0 / 1 |
12.00 | |||
| 1 | <?php |
| 2 | |
| 3 | declare(strict_types=1); |
| 4 | |
| 5 | namespace Phpdftk\Html; |
| 6 | |
| 7 | use Phpdftk\Html\Dom\Document; |
| 8 | use Phpdftk\Html\Dom\DocumentFragment; |
| 9 | use Phpdftk\Html\Dom\Element; |
| 10 | use Phpdftk\Html\Tokenizer\Tokenizer; |
| 11 | use Phpdftk\Html\TreeConstruction\TreeBuilder; |
| 12 | |
| 13 | /** |
| 14 | * WHATWG HTML5 parser entry point. Hand-rolls the tokenizer (§13.2.5) and |
| 15 | * tree-construction state machine (§13.2.6) — no `libxml`, no DOM extension. |
| 16 | * |
| 17 | * The public surface is intentionally tiny: parseDocument() for a full HTML |
| 18 | * document and parseFragment() for innerHTML-style operations and HTML |
| 19 | * embedded in SVG <foreignObject>. |
| 20 | * |
| 21 | * Implementation is staged across Phase 1B sub-phases: |
| 22 | * - 1B.1: public DOM types and parser shell (this file). |
| 23 | * - 1B.2: tokenizer state machine. |
| 24 | * - 1B.3: tree-construction insertion modes. |
| 25 | * - 1B.4: declarative-shadow-DOM tree-construction integration. |
| 26 | * - 1B.5: html5lib-tests integration to 100%. |
| 27 | */ |
| 28 | final class Parser |
| 29 | { |
| 30 | public function __construct(public readonly ParserOptions $options = new ParserOptions()) {} |
| 31 | |
| 32 | /** |
| 33 | * Parse a complete HTML document. |
| 34 | * |
| 35 | * @param string $html The HTML source. |
| 36 | * @param string|null $encoding Optional override for encoding sniffing. |
| 37 | */ |
| 38 | public function parseDocument(string $html, ?string $encoding = null): Document |
| 39 | { |
| 40 | // HTML §13.2.3 — the tokenizer works in UTF-8, so a document in a |
| 41 | // legacy encoding has to be transcoded FIRST. Skipping this did |
| 42 | // not merely mislabel such a document, it mis-read it: every |
| 43 | // non-ASCII byte became U+FFFD, so `<meta charset=windows-1251>` |
| 44 | // rendered replacement boxes where its text should be. |
| 45 | $html = EncodingSniffer::toUtf8($html, $encoding); |
| 46 | $tokenizer = new Tokenizer($html); |
| 47 | $builder = new TreeBuilder($this->options); |
| 48 | $builder->xhtmlSelfClosing = self::looksLikeXhtml($html); |
| 49 | $doc = $builder->build($tokenizer); |
| 50 | $this->mirrorSelectedContent($doc); |
| 51 | return $doc; |
| 52 | } |
| 53 | |
| 54 | /** |
| 55 | * Is this source XHTML rather than `text/html`? |
| 56 | * |
| 57 | * XHTML is XML, so `<div/>` is an empty element; the HTML tokenizer |
| 58 | * instead treats the trailing slash as a parse error and opens a div |
| 59 | * that swallows the rest of the document. Sniffing lets a single |
| 60 | * `parseDocument()` entry point serve both without the caller having |
| 61 | * to know the MIME type. |
| 62 | * |
| 63 | * The signals are the ones only an XHTML document carries: an XML |
| 64 | * declaration, or an XHTML `<!DOCTYPE>` public identifier. The |
| 65 | * `xmlns` attribute alone is NOT enough — it is legal (and common) |
| 66 | * in `text/html`, where the HTML rules still apply. |
| 67 | */ |
| 68 | private static function looksLikeXhtml(string $html): bool |
| 69 | { |
| 70 | $prefix = substr($html, 0, 1024); |
| 71 | // Skip a UTF-8 BOM and leading whitespace before the prolog test. |
| 72 | $trimmed = ltrim($prefix, "\xEF\xBB\xBF \t\r\n\f"); |
| 73 | if (str_starts_with($trimmed, '<?xml')) { |
| 74 | return true; |
| 75 | } |
| 76 | return (bool) preg_match( |
| 77 | '/<!DOCTYPE\s+html\s+PUBLIC\s+"-\/\/W3C\/\/DTD\s+XHTML/i', |
| 78 | $prefix, |
| 79 | ); |
| 80 | } |
| 81 | |
| 82 | /** |
| 83 | * Post-parse pass for the customizable-select `<selectedcontent>` element: |
| 84 | * mirror the selected option's children into each `<selectedcontent>` |
| 85 | * descendant of every `<select>`. Per the customizable-select draft, this |
| 86 | * happens at runtime — but parsing-time test fixtures (html5lib-tests |
| 87 | * webkit02 #45–#48) bake the mirror into their expected output, so we |
| 88 | * perform it once after the tree is built. |
| 89 | */ |
| 90 | private function mirrorSelectedContent(Document $doc): void |
| 91 | { |
| 92 | $stack = [$doc]; |
| 93 | $selects = []; |
| 94 | while ($stack !== []) { |
| 95 | $node = array_pop($stack); |
| 96 | foreach ($node->childNodes() as $child) { |
| 97 | if ($child instanceof Element) { |
| 98 | if ($child->localName === 'select' |
| 99 | && $child->namespaceURI === \Phpdftk\Html\Dom\Document::HTML_NS |
| 100 | ) { |
| 101 | $selects[] = $child; |
| 102 | } |
| 103 | $stack[] = $child; |
| 104 | } |
| 105 | } |
| 106 | } |
| 107 | foreach ($selects as $select) { |
| 108 | $selectedContent = $this->findFirstDescendant($select, 'selectedcontent'); |
| 109 | if ($selectedContent === null) { |
| 110 | continue; |
| 111 | } |
| 112 | $option = $this->findSelectedOption($select); |
| 113 | if ($option === null) { |
| 114 | continue; |
| 115 | } |
| 116 | foreach ($option->childNodes() as $child) { |
| 117 | $selectedContent->appendChild($child->cloneNode(true)); |
| 118 | } |
| 119 | } |
| 120 | } |
| 121 | |
| 122 | private function findFirstDescendant(Element $root, string $localName): ?Element |
| 123 | { |
| 124 | $stack = [$root]; |
| 125 | while ($stack !== []) { |
| 126 | $node = array_pop($stack); |
| 127 | foreach ($node->childNodes() as $child) { |
| 128 | if ($child instanceof Element) { |
| 129 | if ($child->localName === $localName |
| 130 | && $child->namespaceURI === \Phpdftk\Html\Dom\Document::HTML_NS |
| 131 | ) { |
| 132 | return $child; |
| 133 | } |
| 134 | $stack[] = $child; |
| 135 | } |
| 136 | } |
| 137 | } |
| 138 | return null; |
| 139 | } |
| 140 | |
| 141 | private function findSelectedOption(Element $select): ?Element |
| 142 | { |
| 143 | $firstOption = null; |
| 144 | $stack = [$select]; |
| 145 | while ($stack !== []) { |
| 146 | $node = array_pop($stack); |
| 147 | foreach ($node->childNodes() as $child) { |
| 148 | if (!$child instanceof Element) { |
| 149 | continue; |
| 150 | } |
| 151 | if ($child->localName === 'option' |
| 152 | && $child->namespaceURI === \Phpdftk\Html\Dom\Document::HTML_NS |
| 153 | ) { |
| 154 | if ($child->hasAttribute('selected')) { |
| 155 | return $child; |
| 156 | } |
| 157 | $firstOption ??= $child; |
| 158 | } |
| 159 | // Don't recurse into nested selects. |
| 160 | if ($child->localName === 'select' |
| 161 | && $child->namespaceURI === \Phpdftk\Html\Dom\Document::HTML_NS |
| 162 | && $child !== $select |
| 163 | ) { |
| 164 | continue; |
| 165 | } |
| 166 | $stack[] = $child; |
| 167 | } |
| 168 | } |
| 169 | return $firstOption; |
| 170 | } |
| 171 | |
| 172 | /** |
| 173 | * Parse an HTML fragment in the context of a host element per WHATWG |
| 174 | * §13.4. The context element determines the initial tokenizer state |
| 175 | * (e.g. RCDATA for <title>/<textarea>, RAWTEXT for <style>/<script>, |
| 176 | * PLAINTEXT for <plaintext>) and the initial insertion mode (via the |
| 177 | * "reset insertion mode appropriately" walk with the context as the |
| 178 | * implicit bottom of the stack). |
| 179 | */ |
| 180 | public function parseFragment(string $html, Element $context): DocumentFragment |
| 181 | { |
| 182 | // Step 1: new document, inherit mode from context's owner. |
| 183 | $doc = new \Phpdftk\Html\Dom\Document(); |
| 184 | $doc->mode = $context->ownerDocument->mode; |
| 185 | |
| 186 | // Step 2: tokenizer with context-aware initial state. |
| 187 | $tokenizer = new Tokenizer($html); |
| 188 | if ($context->namespaceURI === \Phpdftk\Html\Dom\Document::HTML_NS) { |
| 189 | $tokenizer->state = match ($context->localName) { |
| 190 | 'title', 'textarea' => \Phpdftk\Html\Tokenizer\TokenizerState::Rcdata, |
| 191 | 'style', 'xmp', 'iframe', 'noembed', 'noframes' => \Phpdftk\Html\Tokenizer\TokenizerState::Rawtext, |
| 192 | 'script' => \Phpdftk\Html\Tokenizer\TokenizerState::ScriptData, |
| 193 | 'noscript' => $this->options->scriptingEnabled |
| 194 | ? \Phpdftk\Html\Tokenizer\TokenizerState::Rawtext |
| 195 | : \Phpdftk\Html\Tokenizer\TokenizerState::Data, |
| 196 | 'plaintext' => \Phpdftk\Html\Tokenizer\TokenizerState::Plaintext, |
| 197 | default => \Phpdftk\Html\Tokenizer\TokenizerState::Data, |
| 198 | }; |
| 199 | } |
| 200 | |
| 201 | // Step 3-9 delegated to TreeBuilder::buildFragment, which configures |
| 202 | // the initial state (html root, form pointer, template stack, reset |
| 203 | // insertion mode based on context) and runs the parse. |
| 204 | $builder = new TreeBuilder($this->options, $doc); |
| 205 | return $builder->buildFragment($tokenizer, $context); |
| 206 | } |
| 207 | } |