Code Coverage |
||||||||||
Lines |
Functions and Methods |
Classes and Traits |
||||||||
| Total | |
78.82% |
67 / 85 |
|
50.00% |
4 / 8 |
CRAP | |
0.00% |
0 / 1 |
| ThumbExtractor | |
78.82% |
67 / 85 |
|
50.00% |
4 / 8 |
40.13 | |
0.00% |
0 / 1 |
| __construct | |
100.00% |
1 / 1 |
|
100.00% |
1 / 1 |
1 | |||
| findThumbs | |
100.00% |
4 / 4 |
|
100.00% |
1 / 1 |
1 | |||
| select | |
100.00% |
12 / 12 |
|
100.00% |
1 / 1 |
1 | |||
| sort | |
17.65% |
3 / 17 |
|
0.00% |
0 / 1 |
18.96 | |||
| filter | |
96.43% |
27 / 28 |
|
0.00% |
0 / 1 |
10 | |||
| extractTitleFromAnchorElement | |
88.89% |
8 / 9 |
|
0.00% |
0 / 1 |
5.03 | |||
| matchesSelectors | |
100.00% |
4 / 4 |
|
100.00% |
1 / 1 |
3 | |||
| closest | |
80.00% |
8 / 10 |
|
0.00% |
0 / 1 |
5.20 | |||
| 1 | <?php |
| 2 | declare( strict_types = 1 ); |
| 3 | |
| 4 | namespace MediaWiki\Extension\MultimediaViewer; |
| 5 | |
| 6 | use MediaWiki\FileRepo\File\File; |
| 7 | use MediaWiki\Title\Title; |
| 8 | use Wikimedia\Parsoid\Core\DOMCompat; |
| 9 | use Wikimedia\Parsoid\DOM\DocumentFragment; |
| 10 | use Wikimedia\Parsoid\DOM\Element; |
| 11 | use Wikimedia\Parsoid\Utils\DOMUtils; |
| 12 | use Wikimedia\Zest\Zest; |
| 13 | |
| 14 | /** |
| 15 | * Extract thumbnail image data from a wiki page. |
| 16 | * |
| 17 | * Applies the following set of filters: |
| 18 | * - CSS selectors that indicate navigational icons and SVG infobox images |
| 19 | * - CSS selectors as passed by a config variable via the constructor |
| 20 | * - file extension allowlist |
| 21 | * - minimum image size |
| 22 | */ |
| 23 | class ThumbExtractor { |
| 24 | /** |
| 25 | * Exclusions of thumbnails in known non-content areas. |
| 26 | * |
| 27 | * @const string[] |
| 28 | */ |
| 29 | private const DISALLOWED_SELECTORS = [ |
| 30 | // Selectors below this line should be kept in sync with MultimediaViewer's |
| 31 | // isAllowedThumb implementation. They apply equally to MultimediaViewer |
| 32 | // and the carousel. |
| 33 | // |
| 34 | // This is inside an informational template like {{refimprove}} on enwiki |
| 35 | '.metadata', |
| 36 | // MediaViewer has been specifically disabled for this image |
| 37 | '.noviewer', |
| 38 | // We are on an error page for a non-existing article, the image is part of some template |
| 39 | '.noarticletext', |
| 40 | '#siteNotice', |
| 41 | // Thumbnails of a slideshow gallery |
| 42 | 'ul.mw-gallery-slideshow li.gallerybox', |
| 43 | |
| 44 | // Selectors below this line are carousel-only. They must not be added |
| 45 | // to MultimediaViewer's isAllowedThumb, since they should not affect |
| 46 | // the viewer itself. |
| 47 | // |
| 48 | // Image is excluded from the carousel only; still shown in the viewer |
| 49 | '.nocarousel', |
| 50 | ]; |
| 51 | |
| 52 | /** |
| 53 | * @param string[] $allowedExtensions A file extension allowlist, |
| 54 | * typically based on https://commons.wikimedia.org/wiki/Special:MediaStatistics |
| 55 | * @param string[] $excludedImageSelectors CSS selectors whose ancestor presence causes an image to be excluded |
| 56 | * @param int $minWidth Minimum image width in pixels |
| 57 | * @param int $minHeight Minimum image height in pixels |
| 58 | * @param string $articlePath ArticlePath config value |
| 59 | */ |
| 60 | public function __construct( |
| 61 | private array $allowedExtensions, |
| 62 | private array $excludedImageSelectors, |
| 63 | private int $minWidth, |
| 64 | private int $minHeight, |
| 65 | private string $articlePath, |
| 66 | ) { |
| 67 | } |
| 68 | |
| 69 | /** |
| 70 | * Select and filter thumbnail elements from a DOM body. |
| 71 | * |
| 72 | * Applies CSS selector-based selection and exclusion filtering, but does |
| 73 | * not match against files. Useful when file metadata is unavailable (e.g. |
| 74 | * when working with proxied HTML from MobileFrontendContentProvider). |
| 75 | * |
| 76 | * @param DocumentFragment $body A wiki page's DOM body fragment |
| 77 | * @return Element[] The filtered thumbnail elements |
| 78 | */ |
| 79 | public function findThumbs( DocumentFragment $body ): array { |
| 80 | $thumbs = $this->select( $body ); |
| 81 | $thumbs = $this->sort( $body, $thumbs ); |
| 82 | $thumbs = $this->filter( $thumbs ); |
| 83 | return $thumbs; |
| 84 | } |
| 85 | |
| 86 | /** |
| 87 | * Select image thumbnails from a DOM body via CSS selectors. |
| 88 | * |
| 89 | * @param DocumentFragment $body A wiki page's DOM body fragment |
| 90 | * @return Element[] The extracted elements |
| 91 | */ |
| 92 | private function select( DocumentFragment $body ): array { |
| 93 | $selectors = implode( ', ', [ |
| 94 | // Parsoid thumbs |
| 95 | '[typeof~="mw:File"] a.mw-file-description img', |
| 96 | '[typeof~="mw:File"] a.mw-file-description .lazy-image-placeholder', |
| 97 | // Legacy parser thumbs |
| 98 | '.gallery .image img', |
| 99 | '.gallery .image .lazy-image-placeholder', |
| 100 | 'a.image img', |
| 101 | 'a.image .lazy-image-placeholder', |
| 102 | 'a.mw-file-description img', |
| 103 | 'a.mw-file-description .lazy-image-placeholder', |
| 104 | '#file a img', |
| 105 | ] ); |
| 106 | |
| 107 | return iterator_to_array( DOMCompat::querySelectorAll( $body, $selectors ) ); |
| 108 | } |
| 109 | |
| 110 | /** |
| 111 | * @param DocumentFragment $body A wiki page's DOM body fragment |
| 112 | * @param Element[] $thumbs |
| 113 | * @return Element[] |
| 114 | */ |
| 115 | private function sort( DocumentFragment $body, array $thumbs ): array { |
| 116 | // $thumbs is not guaranteed to be in the correct order in |
| 117 | // which the nodes appear in the document |
| 118 | if ( $thumbs && method_exists( $thumbs[ 0 ], 'compareDocumentPosition' ) ) { |
| 119 | // PHP 8.4+ has built-in position comparison |
| 120 | usort( |
| 121 | $thumbs, |
| 122 | static fn ( $a, $b ) => $a->compareDocumentPosition( $b ) & 2 ? 1 : -1, |
| 123 | ); |
| 124 | } elseif ( count( $thumbs ) > 1 ) { |
| 125 | // Older versions don't; we'll determine the relative |
| 126 | // position of thumbs in the document by locating it |
| 127 | // in the HTML and sort accordingly |
| 128 | $doc = $body->ownerDocument; |
| 129 | $bodyHtml = $doc->saveHTML( $body ); |
| 130 | $positions = array_map( |
| 131 | static fn ( $thumb ) => strpos( $bodyHtml, $doc->saveHTML( $thumb ) ), |
| 132 | $thumbs, |
| 133 | ); |
| 134 | uksort( |
| 135 | $thumbs, |
| 136 | static fn ( $a, $b ) => $positions[ $a ] <=> $positions[ $b ], |
| 137 | ); |
| 138 | } |
| 139 | |
| 140 | return $thumbs; |
| 141 | } |
| 142 | |
| 143 | /** |
| 144 | * Filter files that don't match the given constraints |
| 145 | * (e.g. don't have an allowed extension or whose size, is too small, or match |
| 146 | * disallowed selectors) |
| 147 | * |
| 148 | * @param Element[] $thumbs |
| 149 | * @return Element[] |
| 150 | */ |
| 151 | private function filter( array $thumbs ): array { |
| 152 | return array_values( array_filter( |
| 153 | $thumbs, |
| 154 | function ( $thumb ): bool { |
| 155 | // Find the parent <a> to get the file name. Rely on |
| 156 | // DOMUtils::nodeName() to abstract over differences in |
| 157 | // PHP versions (which may use uppercase or lowercase tag names). |
| 158 | $anchor = DOMCompat::getParentElement( $thumb ); |
| 159 | $title = $this->extractTitleFromAnchorElement( $anchor ); |
| 160 | if ( !$title || $title->getNamespace() !== NS_FILE ) { |
| 161 | return false; |
| 162 | } |
| 163 | |
| 164 | $filename = $title->getDBkey(); |
| 165 | $n = strrpos( $filename, '.' ); |
| 166 | $extension = File::normalizeExtension( $n ? substr( $filename, $n + 1 ) : '' ); |
| 167 | if ( !in_array( $extension, $this->allowedExtensions ) ) { |
| 168 | return false; |
| 169 | } |
| 170 | |
| 171 | // if width/height is set, ensure it matches the minima defined |
| 172 | $width = DOMCompat::getAttribute( $thumb, 'width' ) ?? |
| 173 | DOMCompat::getAttribute( $thumb, 'data-width' ); |
| 174 | $height = DOMCompat::getAttribute( $thumb, 'height' ) ?? |
| 175 | DOMCompat::getAttribute( $thumb, 'data-height' ); |
| 176 | if ( |
| 177 | ( $width !== null && ( (int)$width ) < $this->minWidth ) || |
| 178 | ( $height !== null && ( (int)$height ) < $this->minHeight ) |
| 179 | ) { |
| 180 | return false; |
| 181 | } |
| 182 | |
| 183 | $disallowedSelectors = array_merge( |
| 184 | self::DISALLOWED_SELECTORS, |
| 185 | $this->excludedImageSelectors |
| 186 | ); |
| 187 | if ( $this->matchesSelectors( $thumb, $disallowedSelectors ) ) { |
| 188 | return false; |
| 189 | } |
| 190 | |
| 191 | return true; |
| 192 | } |
| 193 | ) ); |
| 194 | } |
| 195 | |
| 196 | public function extractTitleFromAnchorElement( ?Element $anchor ): ?Title { |
| 197 | if ( !$anchor || DOMUtils::nodeName( $anchor ) !== 'a' ) { |
| 198 | return null; |
| 199 | } |
| 200 | |
| 201 | // Decode percent-encoded characters in the src URL so that |
| 202 | // links containing special characters (commas, Unicode, |
| 203 | // parentheses, etc.) can match. The href attribute can contain |
| 204 | // contains encoded forms like %2C or %C3%A2 (although the |
| 205 | // legacy parser and Parsoid handle them differently), whereas |
| 206 | // titles are in decoded form. |
| 207 | // See T428610. |
| 208 | $href = rawurldecode( $anchor->getAttribute( 'href' ) ); |
| 209 | |
| 210 | // Extract file name from href |
| 211 | // Parsoid: href="//en.wikipedia.org/wiki/File:Example.jpg" |
| 212 | // Legacy: href="/wiki/File:Example.jpg" |
| 213 | $path = parse_url( $href, PHP_URL_PATH ); |
| 214 | $articlePathRegex = '/' . str_replace( '\\$1', '(.+)', preg_quote( $this->articlePath, '/' ) ) . '/'; |
| 215 | |
| 216 | if ( |
| 217 | !preg_match( $articlePathRegex, $path, $match ) && |
| 218 | // Fallback for content this wiki did not render: pages proxied |
| 219 | // from a foreign wiki (MobileFrontendContentProvider) carry the |
| 220 | // *source* wiki's article path ("/wiki/$1" on Wikimedia sites), |
| 221 | // and Parsoid emits relative "./Title" hrefs. |
| 222 | !preg_match( '#(?:^\./|/wiki/)(.+)$#', $path, $match ) |
| 223 | ) { |
| 224 | return null; |
| 225 | } |
| 226 | |
| 227 | return Title::newFromText( $match[1] ); |
| 228 | } |
| 229 | |
| 230 | /** |
| 231 | * Check if the thumbnail matches any of the given CSS selectors. |
| 232 | * |
| 233 | * @param Element $thumb An image thumbnail element |
| 234 | * @param string[] $selectors CSS selectors whose ancestor presence causes an image to be excluded |
| 235 | * @return bool |
| 236 | */ |
| 237 | private function matchesSelectors( Element $thumb, array $selectors ): bool { |
| 238 | foreach ( $selectors as $selector ) { |
| 239 | if ( $this->closest( $thumb, $selector ) !== null ) { |
| 240 | return true; |
| 241 | } |
| 242 | } |
| 243 | |
| 244 | return false; |
| 245 | } |
| 246 | |
| 247 | /** |
| 248 | * Find the closest parent of a DOM element (including itself) |
| 249 | * that matches a CSS selector. |
| 250 | * |
| 251 | * @param Element|null $element A DOM element |
| 252 | * @param string $selector A CSS selector to match |
| 253 | * @return Element|null The matching element, or null if none |
| 254 | */ |
| 255 | private function closest( ?Element $element, string $selector ): ?Element { |
| 256 | if ( $element === null ) { |
| 257 | return null; |
| 258 | } |
| 259 | |
| 260 | // Check the element itself first |
| 261 | if ( Zest::matches( $element, $selector ) ) { |
| 262 | return $element; |
| 263 | } |
| 264 | |
| 265 | // Traverse up through parents |
| 266 | $parent = DOMCompat::getParentElement( $element ); |
| 267 | while ( $parent !== null ) { |
| 268 | if ( Zest::matches( $parent, $selector ) ) { |
| 269 | return $parent; |
| 270 | } |
| 271 | $parent = DOMCompat::getParentElement( $parent ); |
| 272 | } |
| 273 | |
| 274 | return null; |
| 275 | } |
| 276 | } |