| 1 |
<?php |
| 2 |
|
| 3 |
namespace FluentCommunity\App\Services; |
| 4 |
|
| 5 |
use FluentCommunity\Framework\Support\Arr; |
| 6 |
|
| 7 |
class RemoteUrlParser |
| 8 |
{ |
| 9 |
|
| 10 |
private static $instance; |
| 11 |
|
| 12 |
/* |
| 13 |
* Create a method to call the getInfoFromRemoteUrl static method as magic method |
| 14 |
*/ |
| 15 |
public static function parse($url) |
| 16 |
{ |
| 17 |
if (!self::$instance) { |
| 18 |
self::$instance = new self(); |
| 19 |
} |
| 20 |
|
| 21 |
$oEmbed = self::$instance->getOembed($url); |
| 22 |
|
| 23 |
if ($oEmbed) { |
| 24 |
return $oEmbed; |
| 25 |
} |
| 26 |
|
| 27 |
return self::$instance->getInfoFromRemoteUrl($url); |
| 28 |
} |
| 29 |
|
| 30 |
public static function extractIframeThumbnail(&$html) |
| 31 |
{ |
| 32 |
if (!is_string($html)) { |
| 33 |
return ''; |
| 34 |
} |
| 35 |
$html = self::sanitizeOembedHtml($html); |
| 36 |
if (!preg_match('/<iframe\s[^>]*\bsrc\s*=\s*([\'"])(.*?)\1/i', $html, $matches)) { |
| 37 |
return ''; |
| 38 |
} |
| 39 |
|
| 40 |
$src = sanitize_url(html_entity_decode($matches[2], ENT_QUOTES | ENT_HTML5, 'UTF-8')); |
| 41 |
if (!$src) { |
| 42 |
return ''; |
| 43 |
} |
| 44 |
|
| 45 |
if (preg_match('#^https?://(?:[\w-]+\.)?youtube\.com/embed/([^?/]+)#i', $src, $ytMatch)) { |
| 46 |
return self::bestYoutubeThumbnail($ytMatch[1]); |
| 47 |
} |
| 48 |
|
| 49 |
$providers = [ |
| 50 |
'#^https?://player\.vimeo\.com/video/([^?/]+).*$#i' => 'https://vumbnail.com/$1.jpg', |
| 51 |
'#^https?://fast\.wistia\.net/embed/iframe/([^?/]+).*$#i' => 'https://fast.wistia.net/embed/medias/$1/swatch', |
| 52 |
'#^https?://(?:www\.)?dailymotion\.com/(?:embed/video|player\.html\?video=)/?([^?/&]+).*$#i' => 'https://www.dailymotion.com/thumbnail/video/$1', |
| 53 |
]; |
| 54 |
|
| 55 |
foreach ($providers as $pattern => $template) { |
| 56 |
$thumb = preg_replace($pattern, $template, $src, 1, $count); |
| 57 |
if ($count) { |
| 58 |
return $thumb; |
| 59 |
} |
| 60 |
} |
| 61 |
|
| 62 |
$parsed = self::parse($src); |
| 63 |
return (!is_wp_error($parsed) && !empty($parsed['image'])) ? $parsed['image'] : ''; |
| 64 |
} |
| 65 |
|
| 66 |
/** |
| 67 |
* Pick the best YouTube thumbnail for a feed preview. |
| 68 |
* |
| 69 |
* Runs on the feed-save path, so it stays cheap: a single HEAD probe for |
| 70 |
* the HD WebP frame (~30-56% smaller than JPG, sharp 16:9), falling back |
| 71 |
* to hqdefault.jpg — the one universally present rung (maxres/sd and even |
| 72 |
* hqdefault.webp 404 for non-HD or legacy uploads). The result is stored |
| 73 |
* on the feed and cached, so the probe is paid once per video. |
| 74 |
*/ |
| 75 |
protected static function bestYoutubeThumbnail($videoId) |
| 76 |
{ |
| 77 |
$videoId = sanitize_text_field($videoId); |
| 78 |
if (!$videoId) { |
| 79 |
return ''; |
| 80 |
} |
| 81 |
|
| 82 |
$fallback = 'https://img.youtube.com/vi/' . $videoId . '/hqdefault.jpg'; |
| 83 |
|
| 84 |
$cacheKey = 'fcom_yt_thumb_' . md5($videoId); |
| 85 |
$cached = get_transient($cacheKey); |
| 86 |
if ($cached !== false) { |
| 87 |
return $cached; |
| 88 |
} |
| 89 |
|
| 90 |
$maxRes = 'https://i.ytimg.com/vi_webp/' . $videoId . '/maxresdefault.webp'; |
| 91 |
$response = wp_remote_head($maxRes, ['timeout' => 1.5, 'redirection' => 0]); |
| 92 |
if (!is_wp_error($response) && wp_remote_retrieve_response_code($response) === 200) { |
| 93 |
set_transient($cacheKey, $maxRes, WEEK_IN_SECONDS); |
| 94 |
return $maxRes; |
| 95 |
} |
| 96 |
|
| 97 |
return $fallback; |
| 98 |
} |
| 99 |
|
| 100 |
public function getOembed($url) |
| 101 |
{ |
| 102 |
$data = (new \WP_oEmbed())->get_data($url, [ |
| 103 |
'discover' => false |
| 104 |
]); |
| 105 |
|
| 106 |
if (empty($data) || is_wp_error($data) || empty($data->provider_name)) { |
| 107 |
return null; |
| 108 |
} |
| 109 |
|
| 110 |
$data = (array)$data; |
| 111 |
|
| 112 |
$provider = strtolower(Arr::get($data, 'provider_name')); |
| 113 |
|
| 114 |
$image = Arr::get($data, 'thumbnail_url'); |
| 115 |
if ($provider === 'youtube') { |
| 116 |
$image = self::bestYoutubeThumbnail(self::getYoutubeVideoId($url)) ?: $image; |
| 117 |
} |
| 118 |
|
| 119 |
return array_filter([ |
| 120 |
'title' => Arr::get($data, 'title'), |
| 121 |
'author_name' => Arr::get($data, 'author_name'), |
| 122 |
'type' => 'oembed', |
| 123 |
'provider' => $provider, |
| 124 |
'content_type' => Arr::get($data, 'type'), |
| 125 |
'url' => $url, |
| 126 |
'html' => self::sanitizeOembedHtml(Arr::get($data, 'html')), |
| 127 |
'image' => $image, |
| 128 |
]); |
| 129 |
} |
| 130 |
|
| 131 |
protected static function getYoutubeVideoId($url) |
| 132 |
{ |
| 133 |
if (preg_match('#(?:youtu\.be/|youtube\.com/(?:embed/|v/|live/|shorts/|watch\?v=))([a-zA-Z0-9_-]+)#i', (string)$url, $match)) { |
| 134 |
return $match[1]; |
| 135 |
} |
| 136 |
|
| 137 |
return ''; |
| 138 |
} |
| 139 |
|
| 140 |
public static function sanitizeOembedHtml($html) |
| 141 |
{ |
| 142 |
if (empty($html)) { |
| 143 |
return $html; |
| 144 |
} |
| 145 |
|
| 146 |
static $allowed = null; |
| 147 |
if ($allowed === null) { |
| 148 |
$allowed = wp_kses_allowed_html('post'); |
| 149 |
$allowed['iframe'] = [ |
| 150 |
'src' => true, |
| 151 |
'width' => true, |
| 152 |
'height' => true, |
| 153 |
'frameborder' => true, |
| 154 |
'allowfullscreen' => true, |
| 155 |
'title' => true, |
| 156 |
'loading' => true, |
| 157 |
'referrerpolicy' => true, |
| 158 |
'sandbox' => true, |
| 159 |
]; |
| 160 |
} |
| 161 |
|
| 162 |
// Strip <script>/<style> incl. contents; wp_kses keeps their inner text otherwise. |
| 163 |
$html = preg_replace('#<(script|style)\b[^>]*>.*?</\1>#is', '', $html); |
| 164 |
$html = preg_replace('#<(script|style)\b[^>]*/?>#i', '', $html); |
| 165 |
|
| 166 |
return wp_kses($html, $allowed, ['https']); |
| 167 |
} |
| 168 |
|
| 169 |
public function getInfoFromRemoteUrl($url) |
| 170 |
{ |
| 171 |
$url = untrailingslashit($url); |
| 172 |
|
| 173 |
if (empty($url)) { |
| 174 |
return new \WP_Error('rest_invalid_url', __('Invalid URL', 'fluent-community'), array('status' => 404)); |
| 175 |
} |
| 176 |
|
| 177 |
$cacheKey = 'fcom_url_details_meta_' . md5($url); |
| 178 |
$cachedReponse = wp_cache_get($cacheKey, 'fluent-community'); |
| 179 |
|
| 180 |
if ($cachedReponse) { |
| 181 |
return $cachedReponse; |
| 182 |
} |
| 183 |
|
| 184 |
$preempted = apply_filters('fluent_community/preview_metadata_pre_fetch', null, $url); |
| 185 |
if ($preempted && is_array($preempted)) { |
| 186 |
wp_cache_set($cacheKey, $preempted, 'fluent-community', apply_filters('rest_url_details_cache_expiration', HOUR_IN_SECONDS)); // phpcs:ignore WordPress.NamingConventions.PrefixAllGlobals.NonPrefixedHooknameFound |
| 187 |
return $preempted; |
| 188 |
} |
| 189 |
|
| 190 |
$remote_url_response = $this->getRemoteBody($url); |
| 191 |
if (is_wp_error($remote_url_response) || empty($remote_url_response)) { |
| 192 |
return $remote_url_response; |
| 193 |
} |
| 194 |
|
| 195 |
$html_head = $this->getDocumentHead($remote_url_response); |
| 196 |
|
| 197 |
$title = $this->getTitle($html_head); |
| 198 |
if (!$title) { |
| 199 |
return new \WP_Error('rest_invalid_url', __('Invalid URL', 'fluent-community'), array('status' => 404)); |
| 200 |
} |
| 201 |
|
| 202 |
$meta_elements = $this->getMetaWithContentElements($html_head); |
| 203 |
|
| 204 |
$data = array_filter([ |
| 205 |
'title' => $title, |
| 206 |
'image' => $this->getImage($meta_elements, $url), |
| 207 |
'description' => $this->getDescription($meta_elements), |
| 208 |
'icon' => $this->getIcon($html_head, $url), |
| 209 |
'type' => 'meta_data', |
| 210 |
'url' => $url |
| 211 |
]); |
| 212 |
|
| 213 |
wp_cache_set($cacheKey, $data, 'fluent-community', apply_filters('rest_url_details_cache_expiration', HOUR_IN_SECONDS)); // phpcs:ignore WordPress.NamingConventions.PrefixAllGlobals.NonPrefixedHooknameFound |
| 214 |
|
| 215 |
return $data; |
| 216 |
} |
| 217 |
|
| 218 |
private function getRemoteBody($url) |
| 219 |
{ |
| 220 |
$modified_user_agent = 'WP-URLDetails/' . get_bloginfo('version') . ' (+' . get_bloginfo('url') . ')'; |
| 221 |
|
| 222 |
$args = array( |
| 223 |
'limit_response_size' => 300 * KB_IN_BYTES, |
| 224 |
'user-agent' => $modified_user_agent, |
| 225 |
); |
| 226 |
|
| 227 |
/** |
| 228 |
* Filters the HTTP request args for URL data retrieval. |
| 229 |
* |
| 230 |
* Can be used to adjust response size limit and other WP_Http::request() args. |
| 231 |
* |
| 232 |
* @param array $args Arguments used for the HTTP request. |
| 233 |
* @param string $url The attempted URL. |
| 234 |
* @since 5.9.0 |
| 235 |
* |
| 236 |
*/ |
| 237 |
$args = apply_filters('rest_url_details_http_request_args', $args, $url); // phpcs:ignore WordPress.NamingConventions.PrefixAllGlobals.NonPrefixedHooknameFound |
| 238 |
|
| 239 |
$response = wp_safe_remote_get($url, $args); |
| 240 |
|
| 241 |
if (\WP_Http::OK !== wp_remote_retrieve_response_code($response)) { |
| 242 |
// Not saving the error response to cache since the error might be temporary. |
| 243 |
return new \WP_Error( |
| 244 |
'no_response', |
| 245 |
__('URL not found. Response returned a non-200 status code for this URL.', 'fluent-community'), |
| 246 |
array('status' => \WP_Http::NOT_FOUND) |
| 247 |
); |
| 248 |
} |
| 249 |
|
| 250 |
$remote_body = wp_remote_retrieve_body($response); |
| 251 |
|
| 252 |
if (empty($remote_body)) { |
| 253 |
return new \WP_Error( |
| 254 |
'no_content', |
| 255 |
__('Unable to retrieve body from response at this URL.', 'fluent-community'), |
| 256 |
array('status' => \WP_Http::NOT_FOUND) |
| 257 |
); |
| 258 |
} |
| 259 |
|
| 260 |
return $remote_body; |
| 261 |
} |
| 262 |
|
| 263 |
private function getDocumentHead($html) |
| 264 |
{ |
| 265 |
$head_html = $html; |
| 266 |
|
| 267 |
// Find the opening `<head>` tag. |
| 268 |
$head_start = strpos($html, '<head'); |
| 269 |
if (false === $head_start) { |
| 270 |
// Didn't find it. Return the original HTML. |
| 271 |
return $html; |
| 272 |
} |
| 273 |
|
| 274 |
// Find the closing `</head>` tag. |
| 275 |
$head_end = strpos($head_html, '</head>'); |
| 276 |
if (false === $head_end) { |
| 277 |
// Didn't find it. Find the opening `<body>` tag. |
| 278 |
$head_end = strpos($head_html, '<body'); |
| 279 |
|
| 280 |
// Didn't find it. Return the original HTML. |
| 281 |
if (false === $head_end) { |
| 282 |
return $html; |
| 283 |
} |
| 284 |
} |
| 285 |
|
| 286 |
// Extract the HTML from opening tag to the closing tag. Then add the closing tag. |
| 287 |
$head_html = substr($head_html, $head_start, $head_end); |
| 288 |
$head_html .= '</head>'; |
| 289 |
|
| 290 |
return $head_html; |
| 291 |
} |
| 292 |
|
| 293 |
private function getDescription($meta_elements) |
| 294 |
{ |
| 295 |
// Bail out if there are no meta elements. |
| 296 |
if (empty($meta_elements[0])) { |
| 297 |
return ''; |
| 298 |
} |
| 299 |
|
| 300 |
$description = $this->getMetadataFromMetaElement( |
| 301 |
$meta_elements, |
| 302 |
'name', |
| 303 |
'(?:description|og:description)' |
| 304 |
); |
| 305 |
|
| 306 |
// Bail out if description not found. |
| 307 |
if ('' === $description) { |
| 308 |
return ''; |
| 309 |
} |
| 310 |
|
| 311 |
return $this->prepare_metadata_for_output($description); |
| 312 |
} |
| 313 |
|
| 314 |
private function getImage($meta_elements, $url) |
| 315 |
{ |
| 316 |
$image = $this->getMetadataFromMetaElement( |
| 317 |
$meta_elements, |
| 318 |
'property', |
| 319 |
'(?:og:image|og:image:url)' |
| 320 |
); |
| 321 |
|
| 322 |
// Bail out if image not found. |
| 323 |
if ('' === $image) { |
| 324 |
return ''; |
| 325 |
} |
| 326 |
|
| 327 |
// Attempt to convert relative URLs to absolute. |
| 328 |
$parsed_url = wp_parse_url($url); |
| 329 |
if (isset($parsed_url['scheme']) && isset($parsed_url['host'])) { |
| 330 |
$root_url = $parsed_url['scheme'] . '://' . $parsed_url['host'] . '/'; |
| 331 |
$image = \WP_Http::make_absolute_url($image, $root_url); |
| 332 |
} |
| 333 |
|
| 334 |
if (!$image) { |
| 335 |
return $image; |
| 336 |
} |
| 337 |
|
| 338 |
return sanitize_url(html_entity_decode($image, ENT_QUOTES | ENT_HTML5, 'UTF-8')); |
| 339 |
} |
| 340 |
|
| 341 |
private function getTitle($html) |
| 342 |
{ |
| 343 |
if (!$html) { |
| 344 |
return ''; |
| 345 |
} |
| 346 |
|
| 347 |
$pattern = '#<title[^>]*>(.*?)<\s*/\s*title>#is'; |
| 348 |
preg_match($pattern, $html, $match_title); |
| 349 |
|
| 350 |
if (empty($match_title[1]) || !is_string($match_title[1])) { |
| 351 |
return ''; |
| 352 |
} |
| 353 |
|
| 354 |
$title = trim($match_title[1]); |
| 355 |
|
| 356 |
return $this->prepare_metadata_for_output($title); |
| 357 |
} |
| 358 |
|
| 359 |
private function getMetaWithContentElements($html) |
| 360 |
{ |
| 361 |
$pattern = '#<meta\s' . |
| 362 |
'[^>]*' . |
| 363 |
'content=(["\']??)(.*)\1' . |
| 364 |
'[^>]*' . |
| 365 |
'\/?>#' . |
| 366 |
'isU'; |
| 367 |
|
| 368 |
preg_match_all($pattern, $html, $elements); |
| 369 |
|
| 370 |
return $elements; |
| 371 |
} |
| 372 |
|
| 373 |
private function prepare_metadata_for_output($metadata) |
| 374 |
{ |
| 375 |
$metadata = html_entity_decode($metadata, ENT_QUOTES, get_bloginfo('charset')); |
| 376 |
$metadata = wp_strip_all_tags($metadata); |
| 377 |
return $metadata; |
| 378 |
} |
| 379 |
|
| 380 |
private function getMetadataFromMetaElement($meta_elements, $attr, $attr_value) |
| 381 |
{ |
| 382 |
// Bail out if there are no meta elements. |
| 383 |
if (empty($meta_elements[0])) { |
| 384 |
return ''; |
| 385 |
} |
| 386 |
|
| 387 |
$metadata = ''; |
| 388 |
$pattern = '#' . |
| 389 |
$attr . '=([\"\']??)\s*' . $attr_value . '\s*\1' . |
| 390 |
'#isU'; |
| 391 |
|
| 392 |
foreach ($meta_elements[0] as $index => $element) { |
| 393 |
preg_match($pattern, $element, $match); |
| 394 |
|
| 395 |
if (empty($match)) { |
| 396 |
continue; |
| 397 |
} |
| 398 |
|
| 399 |
if (isset($meta_elements[2][$index]) && is_string($meta_elements[2][$index])) { |
| 400 |
$metadata = trim($meta_elements[2][$index]); |
| 401 |
} |
| 402 |
|
| 403 |
break; |
| 404 |
} |
| 405 |
|
| 406 |
return $metadata; |
| 407 |
} |
| 408 |
|
| 409 |
private function getIcon($html, $url) |
| 410 |
{ |
| 411 |
// Grab the icon's link element. |
| 412 |
$pattern = '#<link\s[^>]*rel=(?:[\"\']??)\s*(?:icon|shortcut icon|icon shortcut)\s*(?:[\"\']??)[^>]*\/?>#isU'; |
| 413 |
preg_match($pattern, $html, $element); |
| 414 |
if (empty($element[0]) || !is_string($element[0])) { |
| 415 |
return ''; |
| 416 |
} |
| 417 |
$element = trim($element[0]); |
| 418 |
|
| 419 |
// Get the icon's href value. |
| 420 |
$pattern = '#href=([\"\']??)([^\" >]*?)\\1[^>]*#isU'; |
| 421 |
preg_match($pattern, $element, $icon); |
| 422 |
if (empty($icon[2]) || !is_string($icon[2])) { |
| 423 |
return ''; |
| 424 |
} |
| 425 |
$icon = trim($icon[2]); |
| 426 |
|
| 427 |
// If the icon is a data URL, return it. |
| 428 |
$parsed_icon = wp_parse_url($icon); |
| 429 |
if (isset($parsed_icon['scheme']) && 'data' === $parsed_icon['scheme']) { |
| 430 |
return $icon; |
| 431 |
} |
| 432 |
|
| 433 |
// Attempt to convert relative URLs to absolute. |
| 434 |
if (!is_string($url) || '' === $url) { |
| 435 |
return $icon; |
| 436 |
} |
| 437 |
|
| 438 |
$parsed_url = wp_parse_url($url); |
| 439 |
if (isset($parsed_url['scheme']) && isset($parsed_url['host'])) { |
| 440 |
$root_url = $parsed_url['scheme'] . '://' . $parsed_url['host'] . '/'; |
| 441 |
$icon = \WP_Http::make_absolute_url($icon, $root_url); |
| 442 |
} |
| 443 |
|
| 444 |
return $icon; |
| 445 |
} |
| 446 |
} |
| 447 |
|
