columns.php
9 years ago
counter.php
9 years ago
crawler-detect.php
9 years ago
cron.php
9 years ago
dashboard.php
9 years ago
frontend.php
9 years ago
functions.php
9 years ago
query.php
9 years ago
settings.php
9 years ago
update.php
9 years ago
widgets.php
9 years ago
crawler-detect.php
969 lines
| 1 | <?php |
| 2 | // exit if accessed directly |
| 3 | if ( ! defined( 'ABSPATH' ) ) |
| 4 | exit; |
| 5 | |
| 6 | /** |
| 7 | * Post_Views_Counter_Crawler_Detect class. |
| 8 | * |
| 9 | * Based on CrawlerDetect php class adjusted to PHP 5.2 |
| 10 | * https://github.com/JayBizzle/Crawler-Detect/blob/master/src/CrawlerDetect.php |
| 11 | * |
| 12 | * @since 1.2.4 |
| 13 | * @class Post_Views_Counter_Crawler_Detect |
| 14 | */ |
| 15 | class Post_Views_Counter_Crawler_Detect { |
| 16 | |
| 17 | /** |
| 18 | * The user agent. |
| 19 | * |
| 20 | * @var null |
| 21 | */ |
| 22 | protected $user_agent = null; |
| 23 | |
| 24 | /** |
| 25 | * Headers that contain a user agent. |
| 26 | * |
| 27 | * @var array |
| 28 | */ |
| 29 | protected $http_headers = array(); |
| 30 | |
| 31 | /** |
| 32 | * Store regex matches. |
| 33 | * |
| 34 | * @var array |
| 35 | */ |
| 36 | protected $matches = array(); |
| 37 | |
| 38 | /** |
| 39 | * Crawlers object. |
| 40 | * |
| 41 | * @var object |
| 42 | */ |
| 43 | protected $crawlers; |
| 44 | |
| 45 | /** |
| 46 | * Exclusions object. |
| 47 | * |
| 48 | * @var object |
| 49 | */ |
| 50 | protected $exclusions; |
| 51 | |
| 52 | /** |
| 53 | * Headers object. |
| 54 | * |
| 55 | * @var object |
| 56 | */ |
| 57 | protected $ua_http_headers; |
| 58 | |
| 59 | /** |
| 60 | * Class constructor. |
| 61 | */ |
| 62 | public function __construct() { |
| 63 | add_action( 'after_setup_theme', array( $this, 'init' ) ); |
| 64 | } |
| 65 | |
| 66 | /** |
| 67 | * Initialize class. |
| 68 | */ |
| 69 | public function init() { |
| 70 | // break on admin side |
| 71 | if ( is_admin() && ! ( defined( 'DOING_AJAX' ) && DOING_AJAX ) ) |
| 72 | return; |
| 73 | |
| 74 | $this->crawlers = $this->get_crawlers_list(); |
| 75 | $this->exclusions = $this->get_exclusions_list(); |
| 76 | $this->ua_http_headers = $this->get_headers_list(); |
| 77 | $this->set_http_headers(); |
| 78 | $this->set_user_agent(); |
| 79 | } |
| 80 | |
| 81 | /** |
| 82 | * Set HTTP headers. |
| 83 | * |
| 84 | * @param array $http_headers |
| 85 | */ |
| 86 | public function set_http_headers( $http_headers = null ) { |
| 87 | // use global _SERVER if $http_headers aren't defined |
| 88 | if ( ! is_array( $http_headers ) || ! count( $http_headers ) ) { |
| 89 | $http_headers = $_SERVER; |
| 90 | } |
| 91 | // clear existing headers |
| 92 | $this->http_headers = array(); |
| 93 | // only save HTTP headers - in PHP land, that means only _SERVER vars that start with HTTP_. |
| 94 | foreach ( $http_headers as $key => $value ) { |
| 95 | if ( substr( $key, 0, 5 ) === 'HTTP_' ) { |
| 96 | $this->http_headers[$key] = $value; |
| 97 | } |
| 98 | } |
| 99 | } |
| 100 | |
| 101 | /** |
| 102 | * Return user agent headers. |
| 103 | * |
| 104 | * @return array |
| 105 | */ |
| 106 | public function get_ua_http_headers() { |
| 107 | return $this->ua_http_headers; |
| 108 | } |
| 109 | |
| 110 | /** |
| 111 | * Return the user agent. |
| 112 | * |
| 113 | * @return string |
| 114 | */ |
| 115 | public function get_user_agent() { |
| 116 | return $this->user_agent; |
| 117 | } |
| 118 | |
| 119 | /** |
| 120 | * Set the user agent. |
| 121 | * |
| 122 | * @param string $user_agent |
| 123 | */ |
| 124 | public function set_user_agent( $user_agent = null ) { |
| 125 | if ( false === empty( $user_agent ) ) { |
| 126 | return $this->user_agent = $user_agent; |
| 127 | } else { |
| 128 | $this->user_agent = null; |
| 129 | foreach ( $this->get_ua_http_headers() as $alt_header ) { |
| 130 | if ( false === empty( $this->http_headers[$alt_header] ) ) { // @todo: should use get_http_header(), but it would be slow. |
| 131 | $this->user_agent .= $this->http_headers[$alt_header] . ' '; |
| 132 | } |
| 133 | } |
| 134 | return $this->user_agent = ( ! empty( $this->user_agent ) ? trim( $this->user_agent ) : null); |
| 135 | } |
| 136 | } |
| 137 | |
| 138 | /** |
| 139 | * Build the user agent regex. |
| 140 | * |
| 141 | * @return string |
| 142 | */ |
| 143 | public function get_regex() { |
| 144 | return '(' . implode( '|', $this->crawlers ) . ')'; |
| 145 | } |
| 146 | |
| 147 | /** |
| 148 | * Build the replacement regex. |
| 149 | * |
| 150 | * @return string |
| 151 | */ |
| 152 | public function get_exclusions() { |
| 153 | return '(' . implode( '|', $this->exclusions ) . ')'; |
| 154 | } |
| 155 | |
| 156 | /** |
| 157 | * Check user agent string against the regex. |
| 158 | * |
| 159 | * @param string $user_agent |
| 160 | * |
| 161 | * @return bool |
| 162 | */ |
| 163 | public function is_crawler( $user_agent = null ) { |
| 164 | $agent = is_null( $user_agent ) ? $this->user_agent : $user_agent; |
| 165 | $agent = preg_replace( '/' . $this->get_exclusions() . '/i', '', $agent ); |
| 166 | if ( strlen( trim( $agent ) ) == 0 ) { |
| 167 | return false; |
| 168 | } else { |
| 169 | $result = preg_match( '/' . $this->get_regex() . '/i', trim( $agent ), $matches ); |
| 170 | } |
| 171 | if ( $matches ) { |
| 172 | $this->matches = $matches; |
| 173 | } |
| 174 | return (bool) $result; |
| 175 | } |
| 176 | |
| 177 | /** |
| 178 | * Return the matches. |
| 179 | * |
| 180 | * @return string |
| 181 | */ |
| 182 | public function get_matches() { |
| 183 | return isset( $this->matches[0] ) ? $this->matches[0] : null; |
| 184 | } |
| 185 | |
| 186 | /** |
| 187 | * Return the regular expressions to match against the user agent. |
| 188 | * |
| 189 | * @return array |
| 190 | */ |
| 191 | protected function get_crawlers_list() { |
| 192 | $data = array( |
| 193 | '.*Java.*outbrain', |
| 194 | '008\/', |
| 195 | '192.comAgent', |
| 196 | '2ip\.ru', |
| 197 | '404checker', |
| 198 | '^bluefish ', |
| 199 | '^FDM ', |
| 200 | '^Goose\/', |
| 201 | '^Java\/', |
| 202 | '^Mget', |
| 203 | '^NG\/[0-9\.]', |
| 204 | '^NING\/', |
| 205 | '^PHP\/[0-9]', |
| 206 | '^RMA\/', |
| 207 | '^Ruby|Ruby\/[0-9]', |
| 208 | '^scrutiny\/', |
| 209 | '^VSE\/[0-9]', |
| 210 | '^WordPress\.com', |
| 211 | '^XRL\/[0-9]', |
| 212 | 'a3logics\.in', |
| 213 | 'A6-Indexer', |
| 214 | 'a\.pr-cy\.ru', |
| 215 | 'Aboundex', |
| 216 | 'aboutthedomain', |
| 217 | 'Accoona-AI-Agent', |
| 218 | 'acoon', |
| 219 | 'acrylicapps\.com\/pulp', |
| 220 | 'adbeat', |
| 221 | 'AddThis', |
| 222 | 'ADmantX', |
| 223 | 'adressendeutschland', |
| 224 | 'Advanced Email Extractor v', |
| 225 | 'agentslug', |
| 226 | 'AHC', |
| 227 | 'aihit', |
| 228 | 'aiohttp\/', |
| 229 | 'Airmail', |
| 230 | 'akula\/', |
| 231 | 'alertra', |
| 232 | 'alexa site audit', |
| 233 | 'alyze\.info', |
| 234 | 'amagit', |
| 235 | 'AndroidDownloadManager', |
| 236 | 'Anemone', |
| 237 | 'Ant\.com', |
| 238 | 'Anturis Agent', |
| 239 | 'AnyEvent-HTTP\/', |
| 240 | 'Apache-HttpClient\/', |
| 241 | 'AportWorm\/[0-9]', |
| 242 | 'AppEngine-Google', |
| 243 | 'Arachmo', |
| 244 | 'arachnode', |
| 245 | 'Arachnophilia', |
| 246 | 'archive-com', |
| 247 | 'aria2', |
| 248 | 'asafaweb.com', |
| 249 | 'AskQuickly', |
| 250 | 'Astute', |
| 251 | 'autocite', |
| 252 | 'Autonomy', |
| 253 | 'B-l-i-t-z-B-O-T', |
| 254 | 'Backlink-Ceck\.de', |
| 255 | 'Bad-Neighborhood', |
| 256 | 'baidu\.com', |
| 257 | 'baypup\/[0-9]', |
| 258 | 'baypup\/colbert', |
| 259 | 'BazQux', |
| 260 | 'BCKLINKS', |
| 261 | 'BDFetch', |
| 262 | 'BegunAdvertising\/', |
| 263 | 'bibnum\.bnf', |
| 264 | 'BigBozz', |
| 265 | 'biglotron', |
| 266 | 'BingLocalSearch', |
| 267 | 'BingPreview', |
| 268 | 'binlar', |
| 269 | 'biz_Directory', |
| 270 | 'Blackboard Safeassign', |
| 271 | 'Bloglovin', |
| 272 | 'BlogPulseLive', |
| 273 | 'BlogSearch', |
| 274 | 'Blogtrottr', |
| 275 | 'boitho\.com-dc', |
| 276 | 'BPImageWalker', |
| 277 | 'Braintree-Webhooks', |
| 278 | 'Branch Metrics API', |
| 279 | 'Branch-Passthrough', |
| 280 | 'Browsershots', |
| 281 | 'BUbiNG', |
| 282 | 'Butterfly\/', |
| 283 | 'BuzzSumo', |
| 284 | 'CakePHP', |
| 285 | 'CapsuleChecker', |
| 286 | 'CaretNail', |
| 287 | 'cb crawl', |
| 288 | 'CC Metadata Scaper', |
| 289 | 'Cerberian Drtrs', |
| 290 | 'CERT\.at-Statistics-Survey', |
| 291 | 'cg-eye', |
| 292 | 'changedetection', |
| 293 | 'Charlotte', |
| 294 | 'CheckHost', |
| 295 | 'chkme\.com', |
| 296 | 'CirrusExplorer\/', |
| 297 | 'CISPA Vulnerability Notification', |
| 298 | 'CJNetworkQuality', |
| 299 | 'clips\.ua\.ac\.be', |
| 300 | 'Cloud mapping experiment', |
| 301 | 'CloudFlare-AlwaysOnline', |
| 302 | 'Cloudinary\/[0-9]', |
| 303 | 'cmcm\.com', |
| 304 | 'coccoc', |
| 305 | 'CommaFeed', |
| 306 | 'Commons-HttpClient', |
| 307 | 'Comodo SSL Checker', |
| 308 | 'contactbigdatafr', |
| 309 | 'convera', |
| 310 | 'copyright sheriff', |
| 311 | 'cosmos\/[0-9]', |
| 312 | 'Covario-IDS', |
| 313 | 'CrawlForMe\/[0-9]', |
| 314 | 'cron-job\.org', |
| 315 | 'Crowsnest', |
| 316 | 'curb', |
| 317 | 'Curious George', |
| 318 | 'curl', |
| 319 | 'cuwhois\/[0-9]', |
| 320 | 'CyberPatrol', |
| 321 | 'cybo\.com', |
| 322 | 'DareBoost', |
| 323 | 'DataparkSearch', |
| 324 | 'dataprovider', |
| 325 | 'Daum(oa)?[ \/][0-9]', |
| 326 | 'DeuSu', |
| 327 | 'developers\.google\.com\/\+\/web\/snippet\/', |
| 328 | 'Digg', |
| 329 | 'Dispatch\/', |
| 330 | 'dlvr', |
| 331 | 'DNS-Tools Header-Analyzer', |
| 332 | 'DNSPod-reporting', |
| 333 | 'docoloc', |
| 334 | 'DomainAppender', |
| 335 | 'dotSemantic', |
| 336 | 'downforeveryoneorjustme', |
| 337 | 'downnotifier\.com', |
| 338 | 'DowntimeDetector', |
| 339 | 'Dragonfly File Reader', |
| 340 | 'drupact', |
| 341 | 'Drupal (\+http:\/\/drupal\.org\/)', |
| 342 | 'dubaiindex', |
| 343 | 'EARTHCOM', |
| 344 | 'Easy-Thumb', |
| 345 | 'ec2linkfinder', |
| 346 | 'eCairn-Grabber', |
| 347 | 'ECCP', |
| 348 | 'ElectricMonk', |
| 349 | 'elefent', |
| 350 | 'EMail Exractor', |
| 351 | 'EmailWolf', |
| 352 | 'Embed PHP Library', |
| 353 | 'Embedly', |
| 354 | 'europarchive\.org', |
| 355 | 'evc-batch\/[0-9]', |
| 356 | 'EventMachine HttpClient', |
| 357 | 'Evidon', |
| 358 | 'Evrinid', |
| 359 | 'ExactSearch', |
| 360 | 'ExaleadCloudview', |
| 361 | 'Excel\/', |
| 362 | 'Exploratodo', |
| 363 | 'ezooms', |
| 364 | 'facebookexternalhit', |
| 365 | 'facebookplatform', |
| 366 | 'fairshare', |
| 367 | 'Faraday v', |
| 368 | 'Faveeo', |
| 369 | 'Favicon downloader', |
| 370 | 'FavOrg', |
| 371 | 'Feed Wrangler', |
| 372 | 'Feedbin', |
| 373 | 'FeedBooster', |
| 374 | 'FeedBucket', |
| 375 | 'FeedBurner', |
| 376 | 'FeedChecker', |
| 377 | 'Feedly', |
| 378 | 'Feedspot', |
| 379 | 'feeltiptop', |
| 380 | 'Fetch API', |
| 381 | 'Fetch\/[0-9]', |
| 382 | 'Fever\/[0-9]', |
| 383 | 'findlink', |
| 384 | 'findthatfile', |
| 385 | 'Flamingo_SearchEngine', |
| 386 | 'FlipboardBrowserProxy', |
| 387 | 'FlipboardProxy', |
| 388 | 'FlipboardRSS', |
| 389 | 'fluffy', |
| 390 | 'flynxapp', |
| 391 | 'forensiq', |
| 392 | 'FoundSeoTool\/[0-9]', |
| 393 | 'free thumbnails', |
| 394 | 'FreeWebMonitoring SiteChecker', |
| 395 | 'Funnelback', |
| 396 | 'g00g1e\.net', |
| 397 | 'GAChecker', |
| 398 | 'geek-tools', |
| 399 | 'Genderanalyzer', |
| 400 | 'Genieo', |
| 401 | 'GentleSource', |
| 402 | 'GetLinkInfo', |
| 403 | 'getprismatic\.com', |
| 404 | 'GetURLInfo\/[0-9]', |
| 405 | 'GigablastOpenSource', |
| 406 | 'Go [\d\.]* package http', |
| 407 | 'Go-http-client', |
| 408 | 'GomezAgent', |
| 409 | 'gooblog', |
| 410 | 'Goodzer\/[0-9]', |
| 411 | 'Google favicon', |
| 412 | 'Google Keyword Suggestion', |
| 413 | 'Google Keyword Tool', |
| 414 | 'Google Page Speed Insights', |
| 415 | 'Google PP Default', |
| 416 | 'Google Search Console', |
| 417 | 'Google Web Preview', |
| 418 | 'Google-Adwords', |
| 419 | 'Google-Apps-Script', |
| 420 | 'Google-Calendar-Importer', |
| 421 | 'Google-HTTP-Java-Client', |
| 422 | 'Google-Publisher-Plugin', |
| 423 | 'Google-SearchByImage', |
| 424 | 'Google-Site-Verification', |
| 425 | 'Google-Structured-Data-Testing-Tool', |
| 426 | 'google_partner_monitoring', |
| 427 | 'GoogleDocs', |
| 428 | 'GoogleHC\/', |
| 429 | 'GoogleProducer', |
| 430 | 'GoScraper', |
| 431 | 'GoSpotCheck', |
| 432 | 'GoSquared-Status-Checker', |
| 433 | 'gosquared-thumbnailer', |
| 434 | 'GotSiteMonitor', |
| 435 | 'Grammarly', |
| 436 | 'grouphigh', |
| 437 | 'grub-client', |
| 438 | 'GTmetrix', |
| 439 | 'Hatena', |
| 440 | 'hawkReader', |
| 441 | 'HEADMasterSEO', |
| 442 | 'HeartRails_Capture', |
| 443 | 'heritrix', |
| 444 | 'hledejLevne\.cz\/[0-9]', |
| 445 | 'Holmes', |
| 446 | 'HootSuite Image proxy', |
| 447 | 'Hootsuite-WebFeed\/[0-9]', |
| 448 | 'HostTracker', |
| 449 | 'ht:\/\/check', |
| 450 | 'htdig', |
| 451 | 'HTMLParser\/', |
| 452 | 'HTTP-Header-Abfrage', |
| 453 | 'http-kit', |
| 454 | 'HTTP-Tiny', |
| 455 | 'HTTP_Compression_Test', |
| 456 | 'http_request2', |
| 457 | 'http_requester', |
| 458 | 'HttpComponents', |
| 459 | 'httphr', |
| 460 | 'HTTPMon', |
| 461 | 'httpscheck', |
| 462 | 'httpssites_power', |
| 463 | 'httpunit', |
| 464 | 'HttpUrlConnection', |
| 465 | 'httrack', |
| 466 | 'hosterstats', |
| 467 | 'huaweisymantec', |
| 468 | 'HubPages.*crawlingpolicy', |
| 469 | 'HubSpot Connect', |
| 470 | 'HubSpot Marketing Grader', |
| 471 | 'HyperZbozi.cz Feeder', |
| 472 | 'ichiro', |
| 473 | 'IdeelaborPlagiaat', |
| 474 | 'IDG Twitter Links Resolver', |
| 475 | 'IDwhois\/[0-9]', |
| 476 | 'Iframely', |
| 477 | 'igdeSpyder', |
| 478 | 'IlTrovatore', |
| 479 | 'ImageEngine\/', |
| 480 | 'Imagga', |
| 481 | 'InAGist', |
| 482 | 'inbound\.li parser', |
| 483 | 'InDesign%20CC', |
| 484 | 'infegy', |
| 485 | 'infohelfer', |
| 486 | 'InfoWizards Reciprocal Link System PRO', |
| 487 | 'inpwrd\.com', |
| 488 | 'Integrity', |
| 489 | 'integromedb', |
| 490 | 'internet_archive', |
| 491 | 'InternetSeer', |
| 492 | 'internetVista monitor', |
| 493 | 'IODC', |
| 494 | 'IOI', |
| 495 | 'ips-agent', |
| 496 | 'iqdb\/', |
| 497 | 'Irokez', |
| 498 | 'isitup\.org', |
| 499 | 'iskanie', |
| 500 | 'iZSearch', |
| 501 | 'janforman', |
| 502 | 'Jigsaw', |
| 503 | 'Jobboerse', |
| 504 | 'jobo', |
| 505 | 'Jobrapido', |
| 506 | 'KeepRight OpenStreetMap Checker', |
| 507 | 'KimonoLabs\/', |
| 508 | 'knows\.is', |
| 509 | 'kouio', |
| 510 | 'KrOWLer', |
| 511 | 'kulturarw3', |
| 512 | 'KumKie', |
| 513 | 'L\.webis', |
| 514 | 'Larbin', |
| 515 | 'LayeredExtractor', |
| 516 | 'LibVLC', |
| 517 | 'libwww', |
| 518 | 'link checker', |
| 519 | 'Link Valet', |
| 520 | 'link_thumbnailer', |
| 521 | 'linkCheck', |
| 522 | 'linkdex', |
| 523 | 'LinkExaminer', |
| 524 | 'linkfluence', |
| 525 | 'linkpeek', |
| 526 | 'LinkTiger', |
| 527 | 'LinkWalker', |
| 528 | 'Lipperhey', |
| 529 | 'livedoor ScreenShot', |
| 530 | 'LoadImpactPageAnalyzer', |
| 531 | 'LoadImpactRload', |
| 532 | 'LongURL API', |
| 533 | 'looksystems\.net', |
| 534 | 'ltx71', |
| 535 | 'lwp-trivial', |
| 536 | 'lycos', |
| 537 | 'LYT\.SR', |
| 538 | 'mabontland', |
| 539 | 'MagpieRSS', |
| 540 | 'Mail.Ru', |
| 541 | 'MailChimp\.com', |
| 542 | 'Mandrill', |
| 543 | 'marketinggrader', |
| 544 | 'Mediapartners-Google', |
| 545 | 'MegaIndex\.ru', |
| 546 | 'Melvil Rawi\/', |
| 547 | 'MergeFlow-PageReader', |
| 548 | 'MetaInspector', |
| 549 | 'Metaspinner', |
| 550 | 'MetaURI', |
| 551 | 'Microsearch', |
| 552 | 'Microsoft Office ', |
| 553 | 'Microsoft Windows Network Diagnostics', |
| 554 | 'Mindjet', |
| 555 | 'Miniflux', |
| 556 | 'Mnogosearch', |
| 557 | 'mogimogi', |
| 558 | 'Mojolicious (Perl)', |
| 559 | 'monitis', |
| 560 | 'Monitority\/[0-9]', |
| 561 | 'montastic', |
| 562 | 'MonTools', |
| 563 | 'Moreover', |
| 564 | 'Morning Paper', |
| 565 | 'mowser', |
| 566 | 'Mrcgiguy', |
| 567 | 'mShots', |
| 568 | 'MVAClient', |
| 569 | 'nagios', |
| 570 | 'Najdi\.si\/', |
| 571 | 'NETCRAFT', |
| 572 | 'NetLyzer FastProbe', |
| 573 | 'netresearch', |
| 574 | 'NetShelter ContentScan', |
| 575 | 'NetTrack', |
| 576 | 'Netvibes', |
| 577 | 'Neustar WPM', |
| 578 | 'NeutrinoAPI', |
| 579 | 'NewsBlur .*Finder', |
| 580 | 'NewsGator', |
| 581 | 'newsme', |
| 582 | 'newspaper\/', |
| 583 | 'NG-Search', |
| 584 | 'nineconnections\.com', |
| 585 | 'NLNZ_IAHarvester', |
| 586 | 'Nmap Scripting Engine', |
| 587 | 'node-superagent', |
| 588 | 'node\.io', |
| 589 | 'nominet\.org\.uk', |
| 590 | 'Norton-Safeweb', |
| 591 | 'Notifixious', |
| 592 | 'notifyninja', |
| 593 | 'nuhk', |
| 594 | 'nutch', |
| 595 | 'Nuzzel', |
| 596 | 'nWormFeedFinder', |
| 597 | 'Nymesis', |
| 598 | 'Ocelli\/[0-9]', |
| 599 | 'oegp', |
| 600 | 'okhttp', |
| 601 | 'Omea Reader', |
| 602 | 'omgili', |
| 603 | 'Online Domain Tools', |
| 604 | 'OpenCalaisSemanticProxy', |
| 605 | 'Openstat\/', |
| 606 | 'OpenVAS', |
| 607 | 'Optimizer', |
| 608 | 'Orbiter', |
| 609 | 'OrgProbe\/[0-9]', |
| 610 | 'ow\.ly', |
| 611 | 'ownCloud News', |
| 612 | 'Page Analyzer', |
| 613 | 'Page Valet', |
| 614 | 'page2rss', |
| 615 | 'page_verifier', |
| 616 | 'PagePeeker', |
| 617 | 'Pagespeed\/[0-9]', |
| 618 | 'Panopta', |
| 619 | 'panscient', |
| 620 | 'parsijoo', |
| 621 | 'PayPal IPN', |
| 622 | 'Pcore-HTTP', |
| 623 | 'Pearltrees', |
| 624 | 'peerindex', |
| 625 | 'Peew', |
| 626 | 'PhantomJS\/', |
| 627 | 'Photon\/', |
| 628 | 'phpcrawl', |
| 629 | 'phpservermon', |
| 630 | 'Pi-Monster', |
| 631 | 'Pingdom\.com', |
| 632 | 'Pingoscope', |
| 633 | 'PingSpot', |
| 634 | 'Pinterest', |
| 635 | 'Pizilla', |
| 636 | 'Ploetz \+ Zeller', |
| 637 | 'Plukkie', |
| 638 | 'PocketParser', |
| 639 | 'Pompos', |
| 640 | 'Porkbun', |
| 641 | 'Port Monitor', |
| 642 | 'postano', |
| 643 | 'PostPost', |
| 644 | 'postrank', |
| 645 | 'PowerPoint\/', |
| 646 | 'Priceonomics Analysis Engine', |
| 647 | 'Prlog', |
| 648 | 'probethenet', |
| 649 | 'Project 25499', |
| 650 | 'Promotion_Tools_www.searchenginepromotionhelp.com', |
| 651 | 'prospectb2b', |
| 652 | 'Protopage', |
| 653 | 'proximic', |
| 654 | 'PTST ', |
| 655 | 'PTST\/[0-9]+', |
| 656 | 'Pulsepoint XT3 web scraper', |
| 657 | 'Python-httplib2', |
| 658 | 'python-requests', |
| 659 | 'Python-urllib', |
| 660 | 'Qirina Hurdler', |
| 661 | 'Qseero', |
| 662 | 'Qualidator.com SiteAnalyzer', |
| 663 | 'Quora Link Preview', |
| 664 | 'Qwantify', |
| 665 | 'Radian6', |
| 666 | 'RankSonicSiteAuditor', |
| 667 | 'Readability', |
| 668 | 'RealPlayer%20Downloader', |
| 669 | 'RebelMouse', |
| 670 | 'redback\/', |
| 671 | 'Redirect Checker Tool', |
| 672 | 'ReederForMac', |
| 673 | 'ResponseCodeTest\/[0-9]', |
| 674 | 'RestSharp', |
| 675 | 'RetrevoPageAnalyzer', |
| 676 | 'Riddler', |
| 677 | 'Rival IQ', |
| 678 | 'Robosourcer', |
| 679 | 'Robozilla\/[0-9]', |
| 680 | 'ROI Hunter', |
| 681 | 'SalesIntelligent', |
| 682 | 'SauceNAO', |
| 683 | 'SBIder', |
| 684 | 'Scoop', |
| 685 | 'scooter', |
| 686 | 'ScoutJet', |
| 687 | 'ScoutURLMonitor', |
| 688 | 'Scrapy', |
| 689 | 'Scrubby', |
| 690 | 'SearchSight', |
| 691 | 'semanticdiscovery', |
| 692 | 'semanticjuice', |
| 693 | 'SEO Browser', |
| 694 | 'Seo Servis', |
| 695 | 'seo-nastroj.cz', |
| 696 | 'Seobility', |
| 697 | 'SEOCentro', |
| 698 | 'SeoCheck', |
| 699 | 'SeopultContentAnalyzer', |
| 700 | 'SEOstats', |
| 701 | 'Server Density Service Monitoring', |
| 702 | 'servernfo\.com', |
| 703 | 'Seznam screenshot-generator', |
| 704 | 'Shelob', |
| 705 | 'Shoppimon Analyzer', |
| 706 | 'ShoppimonAgent\/[0-9]', |
| 707 | 'ShopWiki', |
| 708 | 'ShortLinkTranslate', |
| 709 | 'shrinktheweb', |
| 710 | 'SilverReader', |
| 711 | 'SimplePie', |
| 712 | 'SimplyFast', |
| 713 | 'Site-Shot\/', |
| 714 | 'Site24x7', |
| 715 | 'SiteBar', |
| 716 | 'SiteCondor', |
| 717 | 'siteexplorer\.info', |
| 718 | 'SiteGuardian', |
| 719 | 'Siteimprove\.com', |
| 720 | 'Sitemap(s)? Generator', |
| 721 | 'Siteshooter B0t', |
| 722 | 'SiteTruth', |
| 723 | 'sitexy\.com', |
| 724 | 'SkypeUriPreview', |
| 725 | 'slider\.com', |
| 726 | 'slurp', |
| 727 | 'SMRF URL Expander', |
| 728 | 'Snappy', |
| 729 | 'SniffRSS', |
| 730 | 'sniptracker', |
| 731 | 'Snoopy', |
| 732 | 'sogou web', |
| 733 | 'SortSite', |
| 734 | 'spaziodati', |
| 735 | 'Specificfeeds', |
| 736 | 'speedy', |
| 737 | 'SPEng', |
| 738 | 'Spinn3r', |
| 739 | 'spray-can', |
| 740 | 'Sprinklr ', |
| 741 | 'spyonweb', |
| 742 | 'Sqworm', |
| 743 | 'SSL Labs', |
| 744 | 'StackRambler', |
| 745 | 'Statastico\/', |
| 746 | 'StatusCake', |
| 747 | 'Stratagems Kumo', |
| 748 | 'Stroke.cz', |
| 749 | 'StudioFACA', |
| 750 | 'suchen', |
| 751 | 'summify', |
| 752 | 'Super Monitoring', |
| 753 | 'Surphace Scout', |
| 754 | 'SwiteScraper', |
| 755 | 'Symfony2 BrowserKit', |
| 756 | 'Sysomos', |
| 757 | 'T0PHackTeam', |
| 758 | 'Tarantula\/', |
| 759 | 'teoma', |
| 760 | 'terrainformatica\.com', |
| 761 | 'The Expert HTML Source Viewer', |
| 762 | 'theinternetrules', |
| 763 | 'theoldreader\.com', |
| 764 | 'Thumbshots', |
| 765 | 'ThumbSniper', |
| 766 | 'TinEye', |
| 767 | 'Tiny Tiny RSS', |
| 768 | 'topster', |
| 769 | 'touche.com', |
| 770 | 'Traackr.com', |
| 771 | 'truwoGPS', |
| 772 | 'tweetedtimes\.com', |
| 773 | 'Tweetminster', |
| 774 | 'Twikle', |
| 775 | 'Twingly', |
| 776 | 'Typhoeus', |
| 777 | 'ubermetrics-technologies', |
| 778 | 'uclassify', |
| 779 | 'UdmSearch', |
| 780 | 'UnwindFetchor', |
| 781 | 'updated', |
| 782 | 'Upflow', |
| 783 | 'URLChecker', |
| 784 | 'URLitor.com', |
| 785 | 'urlresolver', |
| 786 | 'Urlstat', |
| 787 | 'UrlTrends Ranking Updater', |
| 788 | 'Vagabondo', |
| 789 | 'via ggpht\.com GoogleImageProxy', |
| 790 | 'visionutils', |
| 791 | 'vkShare', |
| 792 | 'voltron', |
| 793 | 'Vortex\/[0-9]', |
| 794 | 'voyager\/', |
| 795 | 'VSAgent\/[0-9]', |
| 796 | 'VSB-TUO\/[0-9]', |
| 797 | 'VYU2', |
| 798 | 'w3af\.org', |
| 799 | 'W3C-checklink', |
| 800 | 'W3C-mobileOK', |
| 801 | 'W3C_I18n-Checker', |
| 802 | 'W3C_Unicorn', |
| 803 | 'wangling', |
| 804 | 'Wappalyzer', |
| 805 | 'WatchMouse', |
| 806 | 'WbSrch\/', |
| 807 | 'web-capture\.net', |
| 808 | 'Web-Monitoring', |
| 809 | 'Web-sniffer', |
| 810 | 'Webauskunft', |
| 811 | 'WebCapture', |
| 812 | 'webcollage', |
| 813 | 'WebCookies', |
| 814 | 'WebCorp', |
| 815 | 'WebDoc', |
| 816 | 'WebFetch', |
| 817 | 'WebImages', |
| 818 | 'WebIndex', |
| 819 | 'webkit2png', |
| 820 | 'webmastercoffee', |
| 821 | 'webmon ', |
| 822 | 'webscreenie', |
| 823 | 'Webshot', |
| 824 | 'Website Analyzer\/', |
| 825 | 'websitepulse[+ ]checker', |
| 826 | 'Websnapr\/', |
| 827 | 'Websquash\.com', |
| 828 | 'Webthumb\/[0-9]', |
| 829 | 'WebThumbnail', |
| 830 | 'WeCrawlForThePeace', |
| 831 | 'WeLikeLinks', |
| 832 | 'WEPA', |
| 833 | 'WeSEE', |
| 834 | 'wf84', |
| 835 | 'wget', |
| 836 | 'WhatsApp', |
| 837 | 'WhatsMyIP', |
| 838 | 'WhatWeb', |
| 839 | 'Whibse', |
| 840 | 'Whynder Magnet', |
| 841 | 'Windows-RSS-Platform', |
| 842 | 'WinHttpRequest', |
| 843 | 'wkhtmlto', |
| 844 | 'wmtips', |
| 845 | 'Woko', |
| 846 | 'WomlpeFactory', |
| 847 | 'Word\/', |
| 848 | 'WordPress\/', |
| 849 | 'wotbox', |
| 850 | 'WP Engine Install Performance API', |
| 851 | 'WPScan', |
| 852 | 'wscheck', |
| 853 | 'WWW-Mechanize', |
| 854 | 'www\.monitor\.us', |
| 855 | 'XaxisSemanticsClassifier', |
| 856 | 'Xenu Link Sleuth', |
| 857 | 'XING-contenttabreceiver\/[0-9]', |
| 858 | 'XmlSitemapGenerator', |
| 859 | 'xpymep([0-9]?)\.exe', |
| 860 | 'Y!J-(ASR|BSC)', |
| 861 | 'Yaanb', |
| 862 | 'yacy', |
| 863 | 'Yahoo Ad monitoring', |
| 864 | 'Yahoo Link Preview', |
| 865 | 'YahooCacheSystem', |
| 866 | 'YahooSeeker', |
| 867 | 'YahooYSMcm', |
| 868 | 'YandeG', |
| 869 | 'yandex', |
| 870 | 'yanga', |
| 871 | 'yeti', |
| 872 | 'Yo-yo', |
| 873 | 'Yoleo Consumer', |
| 874 | 'yoogliFetchAgent', |
| 875 | 'YottaaMonitor', |
| 876 | 'yourls\.org', |
| 877 | 'Zao', |
| 878 | 'Zemanta Aggregator', |
| 879 | 'Zend\\\\Http\\\\Client', |
| 880 | 'Zend_Http_Client', |
| 881 | 'zgrab', |
| 882 | 'ZnajdzFoto', |
| 883 | 'ZyBorg', |
| 884 | '[a-z0-9\-_]*((?<!cu)bot|crawler|archiver|transcoder|spider|uptime|validator|fetcher)', |
| 885 | ); |
| 886 | |
| 887 | return $data; |
| 888 | } |
| 889 | |
| 890 | /** |
| 891 | * Return the list of strings to remove from the user agent before running the crawler regex. |
| 892 | * |
| 893 | * @return array |
| 894 | */ |
| 895 | public function get_exclusions_list() { |
| 896 | $data = array( |
| 897 | 'Safari.[\d\.]*', |
| 898 | 'Firefox.[\d\.]*', |
| 899 | 'Chrome.[\d\.]*', |
| 900 | 'Chromium.[\d\.]*', |
| 901 | 'MSIE.[\d\.]', |
| 902 | 'Opera\/[\d\.]*', |
| 903 | 'Mozilla.[\d\.]*', |
| 904 | 'AppleWebKit.[\d\.]*', |
| 905 | 'Trident.[\d\.]*', |
| 906 | 'Windows NT.[\d\.]*', |
| 907 | 'Android [\d\.]*', |
| 908 | 'Macintosh.', |
| 909 | 'Ubuntu', |
| 910 | 'Linux', |
| 911 | '[ ]Intel', |
| 912 | 'Mac OS X [\d_]*', |
| 913 | '(like )?Gecko(.[\d\.]*)?', |
| 914 | 'KHTML,', |
| 915 | 'CriOS.[\d\.]*', |
| 916 | 'CPU iPhone OS ([0-9_])* like Mac OS X', |
| 917 | 'CPU OS ([0-9_])* like Mac OS X', |
| 918 | 'iPod', |
| 919 | 'compatible', |
| 920 | 'x86_..', |
| 921 | 'i686', |
| 922 | 'x64', |
| 923 | 'X11', |
| 924 | 'rv:[\d\.]*', |
| 925 | 'Version.[\d\.]*', |
| 926 | 'WOW64', |
| 927 | 'Win64', |
| 928 | 'Dalvik.[\d\.]*', |
| 929 | ' \.NET CLR [\d\.]*', |
| 930 | 'Presto.[\d\.]*', |
| 931 | 'Media Center PC', |
| 932 | 'BlackBerry', |
| 933 | 'Build', |
| 934 | 'Opera Mini\/\d{1,2}\.\d{1,2}\.[\d\.]*\/\d{1,2}\.', |
| 935 | 'Opera', |
| 936 | ' \.NET[\d\.]*', |
| 937 | '\(|\)|;|,', // remove the following characters ( ) : , |
| 938 | ); |
| 939 | |
| 940 | return $data; |
| 941 | } |
| 942 | |
| 943 | /** |
| 944 | * Return all possible HTTP headers that represent the User-Agent string. |
| 945 | * |
| 946 | * @return array |
| 947 | */ |
| 948 | public function get_headers_list() { |
| 949 | $data = array( |
| 950 | // the default User-Agent string. |
| 951 | 'HTTP_USER_AGENT', |
| 952 | // header can occur on devices using Opera Mini. |
| 953 | 'HTTP_X_OPERAMINI_PHONE_UA', |
| 954 | // vodafone specific header: http://www.seoprinciple.com/mobile-web-community-still-angry-at-vodafone/24/ |
| 955 | 'HTTP_X_DEVICE_USER_AGENT', |
| 956 | 'HTTP_X_ORIGINAL_USER_AGENT', |
| 957 | 'HTTP_X_SKYFIRE_PHONE', |
| 958 | 'HTTP_X_BOLT_PHONE_UA', |
| 959 | 'HTTP_DEVICE_STOCK_UA', |
| 960 | 'HTTP_X_UCBROWSER_DEVICE_UA', |
| 961 | // sometimes, bots (especially Google) use a genuine user agent, but fill this header in with their email address |
| 962 | 'HTTP_FROM', |
| 963 | ); |
| 964 | |
| 965 | return $data; |
| 966 | } |
| 967 | |
| 968 | } |
| 969 |