jetpack
/
vendor
/
wp-php-toolkit
/
data-liberation
/
URL
/
class-nativeurlintextprocessorwrapper.php
PHP
6 days ago
class-convertedurl.php
6 days ago
class-cssurlprocessor.php
6 days ago
class-nativeurlintextprocessorwrapper.php
6 days ago
class-urlintextprocessor.php
6 days ago
class-urlrewritecache.php
6 days ago
class-wpurl.php
6 days ago
class-wpwhatwgurl.php
6 days ago
class-wpwhatwgurlsearchparams.php
6 days ago
functions.php
6 days ago
public-suffix-list.php
6 days ago
class-nativeurlintextprocessorwrapper.php
242 lines
| 1 | <?php |
| 2 | // phpcs:disable Generic.Classes.DuplicateClassName.Found,Generic.Files.OneObjectStructurePerFile.MultipleFound |
| 3 | |
| 4 | namespace WordPress\DataLiberation\URL; |
| 5 | |
| 6 | use WP_HTML_Text_Replacement; |
| 7 | |
| 8 | require_once __DIR__ . '/class-urlrewritecache.php'; |
| 9 | |
| 10 | /** |
| 11 | * Public URLInTextProcessor adapter backed by the native candidate scanner. |
| 12 | * |
| 13 | * Starting point: a reprint-shaped benchmark over 5,000 post_content values |
| 14 | * measured wp_rewrite_urls() at about 90s with the PHP URL-in-text processor. |
| 15 | * Using the native scanner plus bounded PHP caches reduced that benchmark to |
| 16 | * about 3.6s with identical output. That result suggests the repeated |
| 17 | * PHP-side URL parsing and rewrite decisions were the largest proven cost. |
| 18 | * |
| 19 | * This wrapper is therefore a conservative first step. The native extension |
| 20 | * finds ASCII URL candidates, while PHP still performs the public API's WHATWG |
| 21 | * validation, replacement bookkeeping, and non-ASCII fallback behavior. Moving |
| 22 | * more of this wrapper into Rust may reduce PHP/native crossings and could |
| 23 | * plausibly improve the optimized path further, but that conclusion is still a |
| 24 | * hypothesis until measured. Porting WPURL::parse(), is_child_url_of(), and |
| 25 | * WPURL::replace_base_url() semantics into native code may provide larger |
| 26 | * gains, but would require separate parity work before it is safe to rely on. |
| 27 | */ |
| 28 | class NativeURLInTextProcessorWrapper { |
| 29 | private const CANDIDATE_CACHE_MAX = 4096; |
| 30 | |
| 31 | private $text; |
| 32 | private $bytes_already_parsed = 0; |
| 33 | private $matched_url; |
| 34 | private $parsed_url; |
| 35 | private $url_starts_at; |
| 36 | private $url_length; |
| 37 | private $did_prepend_protocol; |
| 38 | private $base_url; |
| 39 | private $base_protocol; |
| 40 | private $native_processor; |
| 41 | private $fallback_processor; |
| 42 | private $lexical_updates = array(); |
| 43 | |
| 44 | private static $candidate_cache; |
| 45 | |
| 46 | public function __construct( $text, $base_url = null ) { |
| 47 | $this->text = $text; |
| 48 | $this->base_url = $base_url; |
| 49 | $this->base_protocol = $base_url ? parse_url( $base_url, PHP_URL_SCHEME ) : null; |
| 50 | |
| 51 | if ( preg_match( '/[^\x00-\x7F]/', $text ) ) { |
| 52 | $this->fallback_processor = new PHPURLInTextProcessor( $text, $base_url ); |
| 53 | return; |
| 54 | } |
| 55 | |
| 56 | $native_class = 'WordPress\\DataLiberation\\URL\\NativeURLInTextProcessor'; |
| 57 | $this->native_processor = new $native_class( $text, $base_url ); |
| 58 | } |
| 59 | |
| 60 | public function next_url() { |
| 61 | if ( null !== $this->fallback_processor ) { |
| 62 | return $this->fallback_processor->next_url(); |
| 63 | } |
| 64 | |
| 65 | while ( $this->native_processor->next_url() ) { |
| 66 | $this->matched_url = $this->native_processor->get_raw_url(); |
| 67 | $this->parsed_url = null; |
| 68 | $this->url_starts_at = null; |
| 69 | $this->url_length = null; |
| 70 | $this->did_prepend_protocol = false; |
| 71 | |
| 72 | if ( false === $this->matched_url || '' === $this->matched_url || '::' === $this->matched_url ) { |
| 73 | continue; |
| 74 | } |
| 75 | |
| 76 | $url_starts_at = strpos( $this->text, $this->matched_url, $this->bytes_already_parsed ); |
| 77 | if ( false === $url_starts_at ) { |
| 78 | continue; |
| 79 | } |
| 80 | $this->bytes_already_parsed = $url_starts_at + strlen( $this->matched_url ); |
| 81 | $this->url_starts_at = $url_starts_at; |
| 82 | $this->url_length = strlen( $this->matched_url ); |
| 83 | |
| 84 | $had_protocol = method_exists( $this->native_processor, 'had_protocol' ) |
| 85 | ? $this->native_processor->had_protocol() |
| 86 | : WPURL::has_http_https_protocol( $this->matched_url ); |
| 87 | |
| 88 | $cache_key = $this->base_url . "\0" . ( $had_protocol ? '1' : '0' ) . "\0" . $this->matched_url; |
| 89 | $cached = self::candidate_cache()->get( $cache_key ); |
| 90 | if ( null !== $cached ) { |
| 91 | if ( false === $cached ) { |
| 92 | continue; |
| 93 | } |
| 94 | $this->parsed_url = $cached['parsed_url']; |
| 95 | $this->did_prepend_protocol = $cached['did_prepend_protocol']; |
| 96 | |
| 97 | return true; |
| 98 | } |
| 99 | |
| 100 | $parsed_url = $this->parse_and_validate_candidate( $had_protocol ); |
| 101 | if ( false === $parsed_url ) { |
| 102 | self::candidate_cache()->set( $cache_key, false ); |
| 103 | continue; |
| 104 | } |
| 105 | |
| 106 | self::candidate_cache()->set( |
| 107 | $cache_key, |
| 108 | array( |
| 109 | 'parsed_url' => $parsed_url, |
| 110 | 'did_prepend_protocol' => $this->did_prepend_protocol, |
| 111 | ) |
| 112 | ); |
| 113 | |
| 114 | $this->parsed_url = $parsed_url; |
| 115 | |
| 116 | return true; |
| 117 | } |
| 118 | |
| 119 | return false; |
| 120 | } |
| 121 | |
| 122 | public function get_raw_url() { |
| 123 | if ( null !== $this->fallback_processor ) { |
| 124 | return $this->fallback_processor->get_raw_url(); |
| 125 | } |
| 126 | |
| 127 | return $this->matched_url ?? false; |
| 128 | } |
| 129 | |
| 130 | public function get_parsed_url() { |
| 131 | if ( null !== $this->fallback_processor ) { |
| 132 | return $this->fallback_processor->get_parsed_url(); |
| 133 | } |
| 134 | |
| 135 | return $this->parsed_url ?? false; |
| 136 | } |
| 137 | |
| 138 | public function set_raw_url( $new_url ) { |
| 139 | if ( null !== $this->fallback_processor ) { |
| 140 | return $this->fallback_processor->set_raw_url( $new_url ); |
| 141 | } |
| 142 | |
| 143 | if ( null === $this->matched_url ) { |
| 144 | return false; |
| 145 | } |
| 146 | |
| 147 | if ( $this->did_prepend_protocol ) { |
| 148 | $new_url = substr( $new_url, strpos( $new_url, '://' ) + 3 ); |
| 149 | } |
| 150 | |
| 151 | $this->matched_url = $new_url; |
| 152 | $this->lexical_updates[ $this->url_starts_at ] = new WP_HTML_Text_Replacement( |
| 153 | $this->url_starts_at, |
| 154 | $this->url_length, |
| 155 | $new_url |
| 156 | ); |
| 157 | |
| 158 | return true; |
| 159 | } |
| 160 | |
| 161 | public function get_updated_text() { |
| 162 | if ( null !== $this->fallback_processor ) { |
| 163 | return $this->fallback_processor->get_updated_text(); |
| 164 | } |
| 165 | |
| 166 | $this->apply_lexical_updates(); |
| 167 | |
| 168 | return $this->text; |
| 169 | } |
| 170 | |
| 171 | private function parse_and_validate_candidate( $had_protocol ) { |
| 172 | $preprocessed_url = $this->matched_url; |
| 173 | if ( $this->base_url && $this->base_protocol && ! $had_protocol ) { |
| 174 | $preprocessed_url = WPURL::ensure_protocol( $preprocessed_url, $this->base_protocol ); |
| 175 | $this->did_prepend_protocol = true; |
| 176 | } |
| 177 | |
| 178 | $parsed_url = WPURL::parse( $preprocessed_url, $this->base_url ); |
| 179 | if ( false === $parsed_url ) { |
| 180 | return false; |
| 181 | } |
| 182 | |
| 183 | if ( $parsed_url->protocol && ! in_array( $parsed_url->protocol, array( 'http:', 'https:' ), true ) ) { |
| 184 | return false; |
| 185 | } |
| 186 | |
| 187 | if ( $parsed_url->username || $parsed_url->password ) { |
| 188 | return false; |
| 189 | } |
| 190 | |
| 191 | if ( ! $had_protocol ) { |
| 192 | $last_dot_position = strrpos( $parsed_url->hostname, '.' ); |
| 193 | if ( false === $last_dot_position ) { |
| 194 | return false; |
| 195 | } |
| 196 | |
| 197 | $tld = substr( $parsed_url->hostname, $last_dot_position + 1 ); |
| 198 | if ( ! WPURL::is_known_public_domain( $tld ) ) { |
| 199 | return false; |
| 200 | } |
| 201 | } |
| 202 | |
| 203 | return $parsed_url; |
| 204 | } |
| 205 | |
| 206 | private function apply_lexical_updates() { |
| 207 | if ( ! count( $this->lexical_updates ) ) { |
| 208 | return; |
| 209 | } |
| 210 | |
| 211 | ksort( $this->lexical_updates ); |
| 212 | |
| 213 | $bytes_already_copied = 0; |
| 214 | $output_buffer = ''; |
| 215 | foreach ( $this->lexical_updates as $diff ) { |
| 216 | $shift = strlen( $diff->text ) - $diff->length; |
| 217 | if ( $diff->start < $this->bytes_already_parsed ) { |
| 218 | $this->bytes_already_parsed += $shift; |
| 219 | } |
| 220 | |
| 221 | $output_buffer .= substr( $this->text, $bytes_already_copied, $diff->start - $bytes_already_copied ); |
| 222 | if ( $diff->start === $this->url_starts_at ) { |
| 223 | $this->url_starts_at = strlen( $output_buffer ); |
| 224 | $this->url_length = strlen( $diff->text ); |
| 225 | } |
| 226 | $output_buffer .= $diff->text; |
| 227 | $bytes_already_copied = $diff->start + $diff->length; |
| 228 | } |
| 229 | |
| 230 | $this->text = $output_buffer . substr( $this->text, $bytes_already_copied ); |
| 231 | $this->lexical_updates = array(); |
| 232 | } |
| 233 | |
| 234 | private static function candidate_cache() { |
| 235 | if ( null === self::$candidate_cache ) { |
| 236 | self::$candidate_cache = new URLRewriteCache( self::CANDIDATE_CACHE_MAX ); |
| 237 | } |
| 238 | |
| 239 | return self::$candidate_cache; |
| 240 | } |
| 241 | } |
| 242 |