AOEpeople /
crawler
| Total Complexity | 201 |
| Total Lines | 1898 |
| Duplicated Lines | 0 % |
| Coverage | 59.14% |
| Changes | 15 | ||
| Bugs | 0 | Features | 0 |
Complex classes like CrawlerController often do a lot of different things. To break such a class down, we need to identify a cohesive component within that class. A common approach to find such a component is to look for fields/methods that share the same prefixes, or suffixes.
Once you have determined the fields that belong together, you can apply the Extract Class refactoring. If the component makes sense as a sub-class, Extract Subclass is also a candidate, and is often faster.
While breaking up the class, it is a good idea to analyze how other classes use CrawlerController, and based on these observations, apply Extract Interface, too.
| 1 | <?php |
||
| 78 | class CrawlerController implements LoggerAwareInterface |
||
| 79 | { |
||
| 80 | use LoggerAwareTrait; |
||
| 81 | use PublicMethodDeprecationTrait; |
||
| 82 | use PublicPropertyDeprecationTrait; |
||
| 83 | |||
| 84 | public const CLI_STATUS_NOTHING_PROCCESSED = 0; |
||
| 85 | |||
| 86 | //queue not empty |
||
| 87 | public const CLI_STATUS_REMAIN = 1; |
||
| 88 | |||
| 89 | //(some) queue items where processed |
||
| 90 | public const CLI_STATUS_PROCESSED = 2; |
||
| 91 | |||
| 92 | //instance didn't finish |
||
| 93 | public const CLI_STATUS_ABORTED = 4; |
||
| 94 | |||
| 95 | public const CLI_STATUS_POLLABLE_PROCESSED = 8; |
||
| 96 | |||
| 97 | /** |
||
| 98 | * @var integer |
||
| 99 | */ |
||
| 100 | public $setID = 0; |
||
| 101 | |||
| 102 | /** |
||
| 103 | * @var string |
||
| 104 | */ |
||
| 105 | public $processID = ''; |
||
| 106 | |||
| 107 | /** |
||
| 108 | * @var array |
||
| 109 | */ |
||
| 110 | public $duplicateTrack = []; |
||
| 111 | |||
| 112 | /** |
||
| 113 | * @var array |
||
| 114 | */ |
||
| 115 | public $downloadUrls = []; |
||
| 116 | |||
| 117 | /** |
||
| 118 | * @var array |
||
| 119 | */ |
||
| 120 | public $incomingProcInstructions = []; |
||
| 121 | |||
| 122 | /** |
||
| 123 | * @var array |
||
| 124 | */ |
||
| 125 | public $incomingConfigurationSelection = []; |
||
| 126 | |||
| 127 | /** |
||
| 128 | * @var bool |
||
| 129 | */ |
||
| 130 | public $registerQueueEntriesInternallyOnly = false; |
||
| 131 | |||
| 132 | /** |
||
| 133 | * @var array |
||
| 134 | */ |
||
| 135 | public $queueEntries = []; |
||
| 136 | |||
| 137 | /** |
||
| 138 | * @var array |
||
| 139 | */ |
||
| 140 | public $urlList = []; |
||
| 141 | |||
| 142 | /** |
||
| 143 | * @var array |
||
| 144 | */ |
||
| 145 | public $extensionSettings = []; |
||
| 146 | |||
| 147 | /** |
||
| 148 | * Mount Point |
||
| 149 | * |
||
| 150 | * @var bool |
||
| 151 | * Todo: Check what this is used for and adjust the type hint or code, as bool doesn't match the current code. |
||
| 152 | */ |
||
| 153 | public $MP = false; |
||
| 154 | |||
| 155 | /** |
||
| 156 | * @var string |
||
| 157 | * @deprecated |
||
| 158 | */ |
||
| 159 | protected $processFilename; |
||
| 160 | |||
| 161 | /** |
||
| 162 | * Holds the internal access mode can be 'gui','cli' or 'cli_im' |
||
| 163 | * |
||
| 164 | * @var string |
||
| 165 | * @deprecated |
||
| 166 | */ |
||
| 167 | protected $accessMode; |
||
| 168 | |||
| 169 | /** |
||
| 170 | * @var QueueRepository |
||
| 171 | */ |
||
| 172 | protected $queueRepository; |
||
| 173 | |||
| 174 | /** |
||
| 175 | * @var ProcessRepository |
||
| 176 | */ |
||
| 177 | protected $processRepository; |
||
| 178 | |||
| 179 | /** |
||
| 180 | * @var ConfigurationRepository |
||
| 181 | */ |
||
| 182 | protected $configurationRepository; |
||
| 183 | |||
| 184 | /** |
||
| 185 | * @var string |
||
| 186 | */ |
||
| 187 | protected $tableName = 'tx_crawler_queue'; |
||
| 188 | |||
| 189 | /** |
||
| 190 | * @var QueueExecutor |
||
| 191 | */ |
||
| 192 | protected $queueExecutor; |
||
| 193 | |||
| 194 | /** |
||
| 195 | * @var int |
||
| 196 | */ |
||
| 197 | protected $maximumUrlsToCompile = 10000; |
||
| 198 | |||
| 199 | /** |
||
| 200 | * @var IconFactory |
||
| 201 | */ |
||
| 202 | protected $iconFactory; |
||
| 203 | |||
| 204 | /** |
||
| 205 | * @var string[] |
||
| 206 | */ |
||
| 207 | private $deprecatedPublicMethods = [ |
||
|
introduced
by
|
|||
| 208 | 'cleanUpOldQueueEntries' => 'Using CrawlerController::cleanUpOldQueueEntries() is deprecated since 9.0.1 and will be removed in v11.x, please use QueueRepository->cleanUpOldQueueEntries() instead.', |
||
| 209 | 'CLI_debug' => 'Using CrawlerController->CLI_debug() is deprecated since 9.1.3 and will be removed in v11.x', |
||
| 210 | 'CLI_runHooks' => 'Using CrawlerController->CLI_runHooks() is deprecated since 9.1.5 and will be removed in v11.x', |
||
| 211 | 'getAccessMode' => 'Using CrawlerController->getAccessMode() is deprecated since 9.1.3 and will be removed in v11.x', |
||
| 212 | 'getLogEntriesForPageId' => 'Using CrawlerController->getLogEntriesForPageId() is deprecated since 9.1.5 and will be remove in v11.x', |
||
| 213 | 'getLogEntriesForSetId' => 'Using crawlerController::getLogEntriesForSetId() is deprecated since 9.0.1 and will be removed in v11.x', |
||
| 214 | 'flushQueue' => 'Using CrawlerController::flushQueue() is deprecated since 9.0.1 and will be removed in v11.x, please use QueueRepository->flushQueue() instead.', |
||
| 215 | 'setAccessMode' => 'Using CrawlerController->setAccessMode() is deprecated since 9.1.3 and will be removed in v11.x', |
||
| 216 | 'getDisabled' => 'Using CrawlerController->getDisabled() is deprecated since 9.1.3 and will be removed in v11.x, please use Crawler->isDisabled() instead', |
||
| 217 | 'setDisabled' => 'Using CrawlerController->setDisabled() is deprecated since 9.1.3 and will be removed in v11.x, please use Crawler->setDisabled() instead', |
||
| 218 | 'getProcessFilename' => 'Using CrawlerController->getProcessFilename() is deprecated since 9.1.3 and will be removed in v11.x', |
||
| 219 | 'setProcessFilename' => 'Using CrawlerController->setProcessFilename() is deprecated since 9.1.3 and will be removed in v11.x', |
||
| 220 | 'getDuplicateRowsIfExist' => 'Using CrawlerController->getDuplicateRowsIfExist() is deprecated since 9.1.4 and will be remove in v11.x, please use QueueRepository->getDuplicateQueueItemsIfExists() instead', |
||
| 221 | ]; |
||
| 222 | |||
| 223 | /** |
||
| 224 | * @var string[] |
||
| 225 | */ |
||
| 226 | private $deprecatedPublicProperties = [ |
||
|
1 ignored issue
–
show
|
|||
| 227 | 'accessMode' => 'Using CrawlerController->accessMode is deprecated since 9.1.3 and will be removed in v11.x', |
||
| 228 | 'processFilename' => 'Using CrawlerController->accessMode is deprecated since 9.1.3 and will be removed in v11.x', |
||
| 229 | ]; |
||
| 230 | |||
| 231 | /** |
||
| 232 | * @var BackendUserAuthentication|null |
||
| 233 | */ |
||
| 234 | private $backendUser; |
||
| 235 | |||
| 236 | /** |
||
| 237 | * @var integer |
||
| 238 | */ |
||
| 239 | private $scheduledTime = 0; |
||
| 240 | |||
| 241 | /** |
||
| 242 | * @var integer |
||
| 243 | */ |
||
| 244 | private $reqMinute = 0; |
||
| 245 | |||
| 246 | /** |
||
| 247 | * @var bool |
||
| 248 | */ |
||
| 249 | private $submitCrawlUrls = false; |
||
| 250 | |||
| 251 | /** |
||
| 252 | * @var bool |
||
| 253 | */ |
||
| 254 | private $downloadCrawlUrls = false; |
||
| 255 | |||
| 256 | /** |
||
| 257 | * @var PageRepository |
||
| 258 | */ |
||
| 259 | private $pageRepository; |
||
| 260 | |||
| 261 | /** |
||
| 262 | * @var Crawler |
||
| 263 | */ |
||
| 264 | private $crawler; |
||
| 265 | |||
| 266 | /************************************ |
||
| 267 | * |
||
| 268 | * Getting URLs based on Page TSconfig |
||
| 269 | * |
||
| 270 | ************************************/ |
||
| 271 | |||
| 272 | 36 | public function __construct() |
|
| 273 | { |
||
| 274 | 36 | $objectManager = GeneralUtility::makeInstance(ObjectManager::class); |
|
| 275 | 36 | $crawlStrategyFactory = GeneralUtility::makeInstance(CrawlStrategyFactory::class); |
|
| 276 | 36 | $this->queueRepository = $objectManager->get(QueueRepository::class); |
|
| 277 | 36 | $this->processRepository = $objectManager->get(ProcessRepository::class); |
|
| 278 | 36 | $this->configurationRepository = $objectManager->get(ConfigurationRepository::class); |
|
| 279 | 36 | $this->pageRepository = GeneralUtility::makeInstance(PageRepository::class); |
|
| 280 | 36 | $this->queueExecutor = GeneralUtility::makeInstance(QueueExecutor::class, $crawlStrategyFactory); |
|
| 281 | 36 | $this->iconFactory = GeneralUtility::makeInstance(IconFactory::class); |
|
| 282 | 36 | $this->crawler = GeneralUtility::makeInstance(Crawler::class); |
|
| 283 | |||
| 284 | 36 | $this->processFilename = Environment::getVarPath() . '/lock/tx_crawler.proc'; |
|
| 285 | |||
| 286 | /** @var ExtensionConfigurationProvider $configurationProvider */ |
||
| 287 | 36 | $configurationProvider = GeneralUtility::makeInstance(ExtensionConfigurationProvider::class); |
|
| 288 | 36 | $settings = $configurationProvider->getExtensionConfiguration(); |
|
| 289 | 36 | $this->extensionSettings = is_array($settings) ? $settings : []; |
|
| 290 | |||
| 291 | // set defaults: |
||
| 292 | 36 | if (MathUtility::convertToPositiveInteger($this->extensionSettings['countInARun']) === 0) { |
|
| 293 | $this->extensionSettings['countInARun'] = 100; |
||
| 294 | } |
||
| 295 | |||
| 296 | 36 | $this->extensionSettings['processLimit'] = MathUtility::forceIntegerInRange($this->extensionSettings['processLimit'], 1, 99, 1); |
|
| 297 | 36 | $this->setMaximumUrlsToCompile(MathUtility::forceIntegerInRange($this->extensionSettings['maxCompileUrls'], 1, 1000000000, 10000)); |
|
| 298 | 36 | } |
|
| 299 | |||
| 300 | 40 | public function setMaximumUrlsToCompile(int $maximumUrlsToCompile): void |
|
| 301 | { |
||
| 302 | 40 | $this->maximumUrlsToCompile = $maximumUrlsToCompile; |
|
| 303 | 40 | } |
|
| 304 | |||
| 305 | /** |
||
| 306 | * Method to set the accessMode can be gui, cli or cli_im |
||
| 307 | * |
||
| 308 | * @return string |
||
| 309 | * @deprecated |
||
| 310 | */ |
||
| 311 | 1 | public function getAccessMode() |
|
| 312 | { |
||
| 313 | 1 | return $this->accessMode; |
|
| 314 | } |
||
| 315 | |||
| 316 | /** |
||
| 317 | * @param string $accessMode |
||
| 318 | * @deprecated |
||
| 319 | */ |
||
| 320 | 1 | public function setAccessMode($accessMode): void |
|
| 321 | { |
||
| 322 | 1 | $this->accessMode = $accessMode; |
|
| 323 | 1 | } |
|
| 324 | |||
| 325 | /** |
||
| 326 | * Set disabled status to prevent processes from being processed |
||
| 327 | * @deprecated |
||
| 328 | */ |
||
| 329 | 2 | public function setDisabled(?bool $disabled = true): void |
|
| 330 | { |
||
| 331 | 2 | if ($disabled) { |
|
| 332 | 1 | GeneralUtility::writeFile($this->processFilename, 'disabled'); |
|
| 333 | 1 | } elseif (is_file($this->processFilename)) { |
|
| 334 | 1 | unlink($this->processFilename); |
|
| 335 | } |
||
| 336 | 2 | } |
|
| 337 | |||
| 338 | /** |
||
| 339 | * Get disable status |
||
| 340 | * @deprecated |
||
| 341 | */ |
||
| 342 | 2 | public function getDisabled(): bool |
|
| 343 | { |
||
| 344 | 2 | return is_file($this->processFilename); |
|
| 345 | } |
||
| 346 | |||
| 347 | /** |
||
| 348 | * @param string $filenameWithPath |
||
| 349 | * @deprecated |
||
| 350 | */ |
||
| 351 | 3 | public function setProcessFilename($filenameWithPath): void |
|
| 352 | { |
||
| 353 | 3 | $this->processFilename = $filenameWithPath; |
|
| 354 | 3 | } |
|
| 355 | |||
| 356 | /** |
||
| 357 | * @return string |
||
| 358 | * @deprecated |
||
| 359 | */ |
||
| 360 | 1 | public function getProcessFilename() |
|
| 361 | { |
||
| 362 | 1 | return $this->processFilename; |
|
| 363 | } |
||
| 364 | |||
| 365 | /** |
||
| 366 | * Sets the extensions settings (unserialized pendant of $TYPO3_CONF_VARS['EXT']['extConf']['crawler']). |
||
| 367 | */ |
||
| 368 | 14 | public function setExtensionSettings(array $extensionSettings): void |
|
| 369 | { |
||
| 370 | 14 | $this->extensionSettings = $extensionSettings; |
|
| 371 | 14 | } |
|
| 372 | |||
| 373 | /** |
||
| 374 | * Check if the given page should be crawled |
||
| 375 | * |
||
| 376 | * @return false|string false if the page should be crawled (not excluded), true / skipMessage if it should be skipped |
||
| 377 | */ |
||
| 378 | 12 | public function checkIfPageShouldBeSkipped(array $pageRow) |
|
| 379 | { |
||
| 380 | // if page is hidden |
||
| 381 | 12 | if (! $this->extensionSettings['crawlHiddenPages'] && $pageRow['hidden']) { |
|
| 382 | 1 | return 'Because page is hidden'; |
|
| 383 | } |
||
| 384 | |||
| 385 | 11 | if (GeneralUtility::inList('3,4,199,254,255', $pageRow['doktype'])) { |
|
| 386 | 3 | return 'Because doktype is not allowed'; |
|
| 387 | } |
||
| 388 | |||
| 389 | 8 | foreach ($GLOBALS['TYPO3_CONF_VARS']['EXTCONF']['crawler']['excludeDoktype'] ?? [] as $key => $doktypeList) { |
|
| 390 | 1 | if (GeneralUtility::inList($doktypeList, $pageRow['doktype'])) { |
|
| 391 | 1 | return 'Doktype was excluded by "' . $key . '"'; |
|
| 392 | } |
||
| 393 | } |
||
| 394 | |||
| 395 | // veto hook |
||
| 396 | 7 | foreach ($GLOBALS['TYPO3_CONF_VARS']['EXTCONF']['crawler']['pageVeto'] ?? [] as $key => $func) { |
|
| 397 | $params = [ |
||
| 398 | 2 | 'pageRow' => $pageRow, |
|
| 399 | ]; |
||
| 400 | // expects "false" if page is ok and "true" or a skipMessage if this page should _not_ be crawled |
||
| 401 | 2 | $veto = GeneralUtility::callUserFunction($func, $params, $this); |
|
| 402 | 2 | if ($veto !== false) { |
|
| 403 | 2 | if (is_string($veto)) { |
|
| 404 | 1 | return $veto; |
|
| 405 | } |
||
| 406 | 1 | return 'Veto from hook "' . htmlspecialchars($key) . '"'; |
|
| 407 | } |
||
| 408 | } |
||
| 409 | |||
| 410 | 5 | return false; |
|
| 411 | } |
||
| 412 | |||
| 413 | /** |
||
| 414 | * Wrapper method for getUrlsForPageId() |
||
| 415 | * It returns an array of configurations and no urls! |
||
| 416 | * |
||
| 417 | * @param array $pageRow Page record with at least dok-type and uid columns. |
||
| 418 | * @param string $skipMessage |
||
| 419 | * @return array |
||
| 420 | * @see getUrlsForPageId() |
||
| 421 | */ |
||
| 422 | 6 | public function getUrlsForPageRow(array $pageRow, &$skipMessage = '') |
|
| 423 | { |
||
| 424 | 6 | if (! is_int($pageRow['uid'])) { |
|
| 425 | $skipMessage = 'PageUid ' . $pageRow['uid'] . ' was not an integer'; |
||
| 426 | return []; |
||
| 427 | } |
||
| 428 | |||
| 429 | 6 | $message = $this->checkIfPageShouldBeSkipped($pageRow); |
|
| 430 | 6 | if ($message === false) { |
|
| 431 | 5 | $res = $this->getUrlsForPageId($pageRow['uid']); |
|
| 432 | 5 | $skipMessage = ''; |
|
| 433 | } else { |
||
| 434 | 1 | $skipMessage = $message; |
|
| 435 | 1 | $res = []; |
|
| 436 | } |
||
| 437 | |||
| 438 | 6 | return $res; |
|
| 439 | } |
||
| 440 | |||
| 441 | /** |
||
| 442 | * Creates a list of URLs from input array (and submits them to queue if asked for) |
||
| 443 | * See Web > Info module script + "indexed_search"'s crawler hook-client using this! |
||
| 444 | * |
||
| 445 | * @param array $vv Information about URLs from pageRow to crawl. |
||
| 446 | * @param array $pageRow Page row |
||
| 447 | * @param int $scheduledTime Unix time to schedule indexing to, typically time() |
||
| 448 | * @param int $reqMinute Number of requests per minute (creates the interleave between requests) |
||
| 449 | * @param bool $submitCrawlUrls If set, submits the URLs to queue |
||
| 450 | * @param bool $downloadCrawlUrls If set (and submitcrawlUrls is false) will fill $downloadUrls with entries) |
||
| 451 | * @param array $duplicateTrack Array which is passed by reference and contains the an id per url to secure we will not crawl duplicates |
||
| 452 | * @param array $downloadUrls Array which will be filled with URLS for download if flag is set. |
||
| 453 | * @param array $incomingProcInstructions Array of processing instructions |
||
| 454 | * @return string List of URLs (meant for display in backend module) |
||
| 455 | */ |
||
| 456 | 4 | public function urlListFromUrlArray( |
|
| 457 | array $vv, |
||
| 458 | array $pageRow, |
||
| 459 | $scheduledTime, |
||
| 460 | $reqMinute, |
||
| 461 | $submitCrawlUrls, |
||
| 462 | $downloadCrawlUrls, |
||
| 463 | array &$duplicateTrack, |
||
| 464 | array &$downloadUrls, |
||
| 465 | array $incomingProcInstructions |
||
| 466 | ) { |
||
| 467 | 4 | if (! is_array($vv['URLs'])) { |
|
| 468 | return 'ERROR - no URL generated'; |
||
| 469 | } |
||
| 470 | 4 | $urlLog = []; |
|
| 471 | 4 | $pageId = (int) $pageRow['uid']; |
|
| 472 | 4 | $configurationHash = $this->getConfigurationHash($vv); |
|
| 473 | 4 | $skipInnerCheck = $this->queueRepository->noUnprocessedQueueEntriesForPageWithConfigurationHashExist($pageId, $configurationHash); |
|
| 474 | |||
| 475 | 4 | $urlService = new UrlService(); |
|
| 476 | |||
| 477 | 4 | foreach ($vv['URLs'] as $urlQuery) { |
|
| 478 | 4 | if (! $this->drawURLs_PIfilter($vv['subCfg']['procInstrFilter'], $incomingProcInstructions)) { |
|
| 479 | continue; |
||
| 480 | } |
||
| 481 | 4 | $url = (string) $urlService->getUrlFromPageAndQueryParameters( |
|
| 482 | 4 | $pageId, |
|
| 483 | $urlQuery, |
||
| 484 | 4 | $vv['subCfg']['baseUrl'] ?? null, |
|
| 485 | 4 | $vv['subCfg']['force_ssl'] ?? 0 |
|
| 486 | ); |
||
| 487 | |||
| 488 | // Create key by which to determine unique-ness: |
||
| 489 | 4 | $uKey = $url . '|' . $vv['subCfg']['userGroups'] . '|' . $vv['subCfg']['procInstrFilter']; |
|
| 490 | |||
| 491 | 4 | if (isset($duplicateTrack[$uKey])) { |
|
| 492 | //if the url key is registered just display it and do not resubmit is |
||
| 493 | $urlLog[] = '<em><span class="text-muted">' . htmlspecialchars($url) . '</span></em>'; |
||
| 494 | } else { |
||
| 495 | // Scheduled time: |
||
| 496 | 4 | $schTime = $scheduledTime + round(count($duplicateTrack) * (60 / $reqMinute)); |
|
| 497 | 4 | $schTime = intval($schTime / 60) * 60; |
|
| 498 | 4 | $formattedDate = BackendUtility::datetime($schTime); |
|
| 499 | 4 | $this->urlList[] = '[' . $formattedDate . '] ' . $url; |
|
| 500 | 4 | $urlList = '[' . $formattedDate . '] ' . htmlspecialchars($url); |
|
| 501 | |||
| 502 | // Submit for crawling! |
||
| 503 | 4 | if ($submitCrawlUrls) { |
|
| 504 | 4 | $added = $this->addUrl( |
|
| 505 | 4 | $pageId, |
|
| 506 | $url, |
||
| 507 | 4 | $vv['subCfg'], |
|
| 508 | $scheduledTime, |
||
| 509 | $configurationHash, |
||
| 510 | $skipInnerCheck |
||
| 511 | ); |
||
| 512 | 4 | if ($added === false) { |
|
| 513 | 4 | $urlList .= ' (URL already existed)'; |
|
| 514 | } |
||
| 515 | } elseif ($downloadCrawlUrls) { |
||
| 516 | $downloadUrls[$url] = $url; |
||
| 517 | } |
||
| 518 | 4 | $urlLog[] = $urlList; |
|
| 519 | } |
||
| 520 | 4 | $duplicateTrack[$uKey] = true; |
|
| 521 | } |
||
| 522 | |||
| 523 | 4 | return implode('<br>', $urlLog); |
|
| 524 | } |
||
| 525 | |||
| 526 | /** |
||
| 527 | * Returns true if input processing instruction is among registered ones. |
||
| 528 | * |
||
| 529 | * @param string $piString PI to test |
||
| 530 | * @param array $incomingProcInstructions Processing instructions |
||
| 531 | * @return boolean |
||
| 532 | */ |
||
| 533 | 5 | public function drawURLs_PIfilter($piString, array $incomingProcInstructions) |
|
| 534 | { |
||
| 535 | 5 | if (empty($incomingProcInstructions)) { |
|
| 536 | 1 | return true; |
|
| 537 | } |
||
| 538 | |||
| 539 | 4 | foreach ($incomingProcInstructions as $pi) { |
|
| 540 | 4 | if (GeneralUtility::inList($piString, $pi)) { |
|
| 541 | 2 | return true; |
|
| 542 | } |
||
| 543 | } |
||
| 544 | 2 | return false; |
|
| 545 | } |
||
| 546 | |||
| 547 | 5 | public function getPageTSconfigForId($id): array |
|
| 568 | } |
||
| 569 | |||
| 570 | /** |
||
| 571 | * This methods returns an array of configurations. |
||
| 572 | * Adds no urls! |
||
| 573 | */ |
||
| 574 | 4 | public function getUrlsForPageId(int $pageId): array |
|
| 667 | } |
||
| 668 | |||
| 669 | /** |
||
| 670 | * Find all configurations of subpages of a page |
||
| 671 | * TODO: Write Functional Tests |
||
| 672 | */ |
||
| 673 | 1 | public function getConfigurationsForBranch(int $rootid, int $depth): array |
|
| 711 | } |
||
| 712 | |||
| 713 | /** |
||
| 714 | * Check if a user has access to an item |
||
| 715 | * (e.g. get the group list of the current logged in user from $GLOBALS['TSFE']->gr_list) |
||
| 716 | * |
||
| 717 | * @param string $groupList Comma-separated list of (fe_)group UIDs from a user |
||
| 718 | * @param string $accessList Comma-separated list of (fe_)group UIDs of the item to access |
||
| 719 | * @return bool TRUE if at least one of the users group UIDs is in the access list or the access list is empty |
||
| 720 | * @see \TYPO3\CMS\Frontend\Page\PageRepository::getMultipleGroupsWhereClause() |
||
| 721 | */ |
||
| 722 | 3 | public function hasGroupAccess($groupList, $accessList) |
|
| 723 | { |
||
| 724 | 3 | if (empty($accessList)) { |
|
| 725 | 1 | return true; |
|
| 726 | } |
||
| 727 | 2 | foreach (GeneralUtility::intExplode(',', $groupList) as $groupUid) { |
|
| 728 | 2 | if (GeneralUtility::inList($accessList, $groupUid)) { |
|
| 729 | 1 | return true; |
|
| 730 | } |
||
| 731 | } |
||
| 732 | 1 | return false; |
|
| 733 | } |
||
| 734 | |||
| 735 | /** |
||
| 736 | * Will expand the parameters configuration to individual values. This follows a certain syntax of the value of each parameter. |
||
| 737 | * Syntax of values: |
||
| 738 | * - Basically: If the value is wrapped in [...] it will be expanded according to the following syntax, otherwise the value is taken literally |
||
| 739 | * - Configuration is splitted by "|" and the parts are processed individually and finally added together |
||
| 740 | * - For each configuration part: |
||
| 741 | * - "[int]-[int]" = Integer range, will be expanded to all values in between, values included, starting from low to high (max. 1000). Example "1-34" or "-40--30" |
||
| 742 | * - "_TABLE:[TCA table name];[_PID:[optional page id, default is current page]];[_ENABLELANG:1]" = Look up of table records from PID, filtering out deleted records. Example "_TABLE:tt_content; _PID:123" |
||
| 743 | * _ENABLELANG:1 picks only original records without their language overlays |
||
| 744 | * - Default: Literal value |
||
| 745 | * |
||
| 746 | * @param array $paramArray Array with key (GET var name) and values (value of GET var which is configuration for expansion) |
||
| 747 | * @param integer $pid Current page ID |
||
| 748 | * @return array |
||
| 749 | * |
||
| 750 | * TODO: Write Functional Tests |
||
| 751 | */ |
||
| 752 | 11 | public function expandParameters($paramArray, $pid) |
|
| 753 | { |
||
| 754 | // Traverse parameter names: |
||
| 755 | 11 | foreach ($paramArray as $p => $v) { |
|
| 756 | 11 | $v = trim($v); |
|
| 757 | |||
| 758 | // If value is encapsulated in square brackets it means there are some ranges of values to find, otherwise the value is literal |
||
| 759 | 11 | if (strpos($v, '[') === 0 && substr($v, -1) === ']') { |
|
| 760 | // So, find the value inside brackets and reset the paramArray value as an array. |
||
| 761 | 11 | $v = substr($v, 1, -1); |
|
| 762 | 11 | $paramArray[$p] = []; |
|
| 763 | |||
| 764 | // Explode parts and traverse them: |
||
| 765 | 11 | $parts = explode('|', $v); |
|
| 766 | 11 | foreach ($parts as $pV) { |
|
| 767 | |||
| 768 | // Look for integer range: (fx. 1-34 or -40--30 // reads minus 40 to minus 30) |
||
| 769 | 11 | if (preg_match('/^(-?[0-9]+)\s*-\s*(-?[0-9]+)$/', trim($pV), $reg)) { |
|
| 770 | 1 | $reg = $this->swapIfFirstIsLargerThanSecond($reg); |
|
| 771 | |||
| 772 | // Traverse range, add values: |
||
| 773 | // Limit to size of range! |
||
| 774 | 1 | $runAwayBrake = 1000; |
|
| 775 | 1 | for ($a = $reg[1]; $a <= $reg[2]; $a++) { |
|
| 776 | 1 | $paramArray[$p][] = $a; |
|
| 777 | 1 | $runAwayBrake--; |
|
| 778 | 1 | if ($runAwayBrake <= 0) { |
|
| 779 | break; |
||
| 780 | } |
||
| 781 | } |
||
| 782 | 10 | } elseif (strpos(trim($pV), '_TABLE:') === 0) { |
|
| 783 | |||
| 784 | // Parse parameters: |
||
| 785 | 6 | $subparts = GeneralUtility::trimExplode(';', $pV); |
|
| 786 | 6 | $subpartParams = []; |
|
| 787 | 6 | foreach ($subparts as $spV) { |
|
| 788 | 6 | [$pKey, $pVal] = GeneralUtility::trimExplode(':', $spV); |
|
| 789 | 6 | $subpartParams[$pKey] = $pVal; |
|
| 790 | } |
||
| 791 | |||
| 792 | // Table exists: |
||
| 793 | 6 | if (isset($GLOBALS['TCA'][$subpartParams['_TABLE']])) { |
|
| 794 | 6 | $lookUpPid = isset($subpartParams['_PID']) ? intval($subpartParams['_PID']) : intval($pid); |
|
| 795 | 6 | $recursiveDepth = isset($subpartParams['_RECURSIVE']) ? intval($subpartParams['_RECURSIVE']) : 0; |
|
| 796 | 6 | $pidField = isset($subpartParams['_PIDFIELD']) ? trim($subpartParams['_PIDFIELD']) : 'pid'; |
|
| 797 | 6 | $where = $subpartParams['_WHERE'] ?? ''; |
|
| 798 | 6 | $addTable = $subpartParams['_ADDTABLE'] ?? ''; |
|
| 799 | |||
| 800 | 6 | $fieldName = $subpartParams['_FIELD'] ? $subpartParams['_FIELD'] : 'uid'; |
|
| 801 | 6 | if ($fieldName === 'uid' || $GLOBALS['TCA'][$subpartParams['_TABLE']]['columns'][$fieldName]) { |
|
| 802 | 6 | $queryBuilder = $this->getQueryBuilder($subpartParams['_TABLE']); |
|
| 803 | |||
| 804 | 6 | if ($recursiveDepth > 0) { |
|
| 805 | /** @var QueryGenerator $queryGenerator */ |
||
| 806 | 2 | $queryGenerator = GeneralUtility::makeInstance(QueryGenerator::class); |
|
| 807 | 2 | $pidList = $queryGenerator->getTreeList($lookUpPid, $recursiveDepth, 0, 1); |
|
| 808 | 2 | $pidArray = GeneralUtility::intExplode(',', $pidList); |
|
| 809 | } else { |
||
| 810 | 4 | $pidArray = [(string) $lookUpPid]; |
|
| 811 | } |
||
| 812 | |||
| 813 | 6 | $queryBuilder->getRestrictions() |
|
| 814 | 6 | ->removeAll() |
|
| 815 | 6 | ->add(GeneralUtility::makeInstance(DeletedRestriction::class)); |
|
| 816 | |||
| 817 | $queryBuilder |
||
| 818 | 6 | ->select($fieldName) |
|
| 819 | 6 | ->from($subpartParams['_TABLE']) |
|
| 820 | 6 | ->where( |
|
| 821 | 6 | $queryBuilder->expr()->in($pidField, $queryBuilder->createNamedParameter($pidArray, Connection::PARAM_INT_ARRAY)), |
|
| 822 | $where |
||
| 823 | ); |
||
| 824 | |||
| 825 | 6 | if (! empty($addTable)) { |
|
| 826 | // TODO: Check if this works as intended! |
||
| 827 | $queryBuilder->add('from', $addTable); |
||
| 828 | } |
||
| 829 | 6 | $transOrigPointerField = $GLOBALS['TCA'][$subpartParams['_TABLE']]['ctrl']['transOrigPointerField']; |
|
| 830 | |||
| 831 | 6 | if ($subpartParams['_ENABLELANG'] && $transOrigPointerField) { |
|
| 832 | $queryBuilder->andWhere( |
||
| 833 | $queryBuilder->expr()->lte( |
||
| 834 | $transOrigPointerField, |
||
| 835 | 0 |
||
| 836 | ) |
||
| 837 | ); |
||
| 838 | } |
||
| 839 | |||
| 840 | 6 | $statement = $queryBuilder->execute(); |
|
| 841 | |||
| 842 | 6 | $rows = []; |
|
| 843 | 6 | while ($row = $statement->fetch()) { |
|
| 844 | 6 | $rows[$row[$fieldName]] = $row; |
|
| 845 | } |
||
| 846 | |||
| 847 | 6 | if (is_array($rows)) { |
|
| 848 | 6 | $paramArray[$p] = array_merge($paramArray[$p], array_keys($rows)); |
|
| 849 | } |
||
| 850 | } |
||
| 851 | } |
||
| 852 | } else { |
||
| 853 | // Just add value: |
||
| 854 | 4 | $paramArray[$p][] = $pV; |
|
| 855 | } |
||
| 856 | // Hook for processing own expandParameters place holder |
||
| 857 | 11 | if (is_array($GLOBALS['TYPO3_CONF_VARS']['SC_OPTIONS']['crawler/class.tx_crawler_lib.php']['expandParameters'])) { |
|
| 858 | $_params = [ |
||
| 859 | 'pObj' => &$this, |
||
| 860 | 'paramArray' => &$paramArray, |
||
| 861 | 'currentKey' => $p, |
||
| 862 | 'currentValue' => $pV, |
||
| 863 | 'pid' => $pid, |
||
| 864 | ]; |
||
| 865 | foreach ($GLOBALS['TYPO3_CONF_VARS']['SC_OPTIONS']['crawler/class.tx_crawler_lib.php']['expandParameters'] as $_funcRef) { |
||
| 866 | GeneralUtility::callUserFunction($_funcRef, $_params, $this); |
||
| 867 | } |
||
| 868 | } |
||
| 869 | } |
||
| 870 | |||
| 871 | // Make unique set of values and sort array by key: |
||
| 872 | 11 | $paramArray[$p] = array_unique($paramArray[$p]); |
|
| 873 | 11 | ksort($paramArray); |
|
| 874 | } else { |
||
| 875 | // Set the literal value as only value in array: |
||
| 876 | 4 | $paramArray[$p] = [$v]; |
|
| 877 | } |
||
| 878 | } |
||
| 879 | |||
| 880 | 11 | return $paramArray; |
|
| 881 | } |
||
| 882 | |||
| 883 | /** |
||
| 884 | * Compiling URLs from parameter array (output of expandParameters()) |
||
| 885 | * The number of URLs will be the multiplication of the number of parameter values for each key |
||
| 886 | * |
||
| 887 | * @param array $paramArray Output of expandParameters(): Array with keys (GET var names) and for each an array of values |
||
| 888 | * @param array $urls URLs accumulated in this array (for recursion) |
||
| 889 | * @return array |
||
| 890 | */ |
||
| 891 | 8 | public function compileUrls($paramArray, array $urls) |
|
| 892 | { |
||
| 893 | 8 | if (empty($paramArray)) { |
|
| 894 | 8 | return $urls; |
|
| 895 | } |
||
| 896 | 7 | $varName = key($paramArray); |
|
| 897 | 7 | $valueSet = array_shift($paramArray); |
|
| 898 | |||
| 899 | // Traverse value set: |
||
| 900 | 7 | $newUrls = []; |
|
| 901 | 7 | foreach ($urls as $url) { |
|
| 902 | 6 | foreach ($valueSet as $val) { |
|
| 903 | 6 | if (count($newUrls) < $this->getMaximumUrlsToCompile()) { |
|
| 904 | 6 | $newUrls[] = $url . (strcmp((string) $val, '') ? '&' . rawurlencode($varName) . '=' . rawurlencode((string) $val) : ''); |
|
| 905 | } |
||
| 906 | } |
||
| 907 | } |
||
| 908 | 7 | return $this->compileUrls($paramArray, $newUrls); |
|
| 909 | } |
||
| 910 | |||
| 911 | /************************************ |
||
| 912 | * |
||
| 913 | * Crawler log |
||
| 914 | * |
||
| 915 | ************************************/ |
||
| 916 | |||
| 917 | /** |
||
| 918 | * Return array of records from crawler queue for input page ID |
||
| 919 | * |
||
| 920 | * @param integer $id Page ID for which to look up log entries. |
||
| 921 | * @param boolean $doFlush If TRUE, then entries selected at DELETED(!) instead of selected! |
||
| 922 | * @param boolean $doFullFlush |
||
| 923 | * @param integer $itemsPerPage Limit the amount of entries per page default is 10 |
||
| 924 | * @return array |
||
| 925 | * |
||
| 926 | * @deprecated |
||
| 927 | */ |
||
| 928 | 4 | public function getLogEntriesForPageId($id, QueueFilter $queueFilter, $doFlush = false, $doFullFlush = false, $itemsPerPage = 10) |
|
| 929 | { |
||
| 930 | 4 | $queryBuilder = GeneralUtility::makeInstance(ConnectionPool::class)->getQueryBuilderForTable($this->tableName); |
|
| 931 | $queryBuilder |
||
| 932 | 4 | ->select('*') |
|
| 933 | 4 | ->from($this->tableName) |
|
| 934 | 4 | ->where( |
|
| 935 | 4 | $queryBuilder->expr()->eq('page_id', $queryBuilder->createNamedParameter($id, PDO::PARAM_INT)) |
|
| 936 | ) |
||
| 937 | 4 | ->orderBy('scheduled', 'DESC'); |
|
| 938 | |||
| 939 | 4 | $expressionBuilder = GeneralUtility::makeInstance(ConnectionPool::class) |
|
| 940 | 4 | ->getConnectionForTable($this->tableName) |
|
| 941 | 4 | ->getExpressionBuilder(); |
|
| 942 | 4 | $query = $expressionBuilder->andX(); |
|
| 943 | // PHPStorm adds the highlight that the $addWhere is immediately overwritten, |
||
| 944 | // but the $query = $expressionBuilder->andX() ensures that the $addWhere is written correctly with AND |
||
| 945 | // between the statements, it's not a mistake in the code. |
||
| 946 | 4 | switch ($queueFilter) { |
|
| 947 | 4 | case 'pending': |
|
| 948 | $queryBuilder->andWhere($queryBuilder->expr()->eq('exec_time', 0)); |
||
| 949 | break; |
||
| 950 | 4 | case 'finished': |
|
| 951 | $queryBuilder->andWhere($queryBuilder->expr()->gt('exec_time', 0)); |
||
| 952 | break; |
||
| 953 | } |
||
| 954 | |||
| 955 | 4 | if ($doFlush) { |
|
| 956 | 2 | $this->queueRepository->flushQueue($queueFilter); |
|
| 957 | } |
||
| 958 | 4 | if ($itemsPerPage > 0) { |
|
| 959 | $queryBuilder |
||
| 960 | 4 | ->setMaxResults((int) $itemsPerPage); |
|
| 961 | } |
||
| 962 | |||
| 963 | 4 | return $queryBuilder->execute()->fetchAll(); |
|
| 964 | } |
||
| 965 | |||
| 966 | /** |
||
| 967 | * Return array of records from crawler queue for input set ID |
||
| 968 | * |
||
| 969 | * @param int $set_id Set ID for which to look up log entries. |
||
| 970 | * @param string $filter Filter: "all" => all entries, "pending" => all that is not yet run, "finished" => all complete ones |
||
| 971 | * @param bool $doFlush If TRUE, then entries selected at DELETED(!) instead of selected! |
||
| 972 | * @param int $itemsPerPage Limit the amount of entries per page default is 10 |
||
| 973 | * @return array |
||
| 974 | * |
||
| 975 | * @deprecated |
||
| 976 | */ |
||
| 977 | 6 | public function getLogEntriesForSetId(int $set_id, string $filter = '', bool $doFlush = false, bool $doFullFlush = false, int $itemsPerPage = 10) |
|
| 978 | { |
||
| 979 | 6 | $queryBuilder = GeneralUtility::makeInstance(ConnectionPool::class)->getQueryBuilderForTable($this->tableName); |
|
| 980 | $queryBuilder |
||
| 981 | 6 | ->select('*') |
|
| 982 | 6 | ->from($this->tableName) |
|
| 983 | 6 | ->where( |
|
| 984 | 6 | $queryBuilder->expr()->eq('set_id', $queryBuilder->createNamedParameter($set_id, PDO::PARAM_INT)) |
|
| 985 | ) |
||
| 986 | 6 | ->orderBy('scheduled', 'DESC'); |
|
| 987 | |||
| 988 | 6 | $expressionBuilder = GeneralUtility::makeInstance(ConnectionPool::class) |
|
| 989 | 6 | ->getConnectionForTable($this->tableName) |
|
| 990 | 6 | ->getExpressionBuilder(); |
|
| 991 | 6 | $query = $expressionBuilder->andX(); |
|
| 992 | // PHPStorm adds the highlight that the $addWhere is immediately overwritten, |
||
| 993 | // but the $query = $expressionBuilder->andX() ensures that the $addWhere is written correctly with AND |
||
| 994 | // between the statements, it's not a mistake in the code. |
||
| 995 | 6 | $addWhere = ''; |
|
| 996 | 6 | switch ($filter) { |
|
| 997 | 6 | case 'pending': |
|
| 998 | 1 | $queryBuilder->andWhere($queryBuilder->expr()->eq('exec_time', 0)); |
|
| 999 | 1 | $addWhere = $query->add($expressionBuilder->eq('exec_time', 0)); |
|
| 1000 | 1 | break; |
|
| 1001 | 5 | case 'finished': |
|
| 1002 | 1 | $queryBuilder->andWhere($queryBuilder->expr()->gt('exec_time', 0)); |
|
| 1003 | 1 | $addWhere = $query->add($expressionBuilder->gt('exec_time', 0)); |
|
| 1004 | 1 | break; |
|
| 1005 | } |
||
| 1006 | 6 | if ($doFlush) { |
|
| 1007 | 4 | $addWhere = $query->add($expressionBuilder->eq('set_id', (int) $set_id)); |
|
| 1008 | 4 | $this->flushQueue($doFullFlush ? '' : $addWhere); |
|
| 1009 | 4 | return []; |
|
| 1010 | } |
||
| 1011 | 2 | if ($itemsPerPage > 0) { |
|
| 1012 | $queryBuilder |
||
| 1013 | 2 | ->setMaxResults((int) $itemsPerPage); |
|
| 1014 | } |
||
| 1015 | |||
| 1016 | 2 | return $queryBuilder->execute()->fetchAll(); |
|
| 1017 | } |
||
| 1018 | |||
| 1019 | /** |
||
| 1020 | * Adding call back entries to log (called from hooks typically, see indexed search class "class.crawler.php" |
||
| 1021 | * |
||
| 1022 | * @param integer $setId Set ID |
||
| 1023 | * @param array $params Parameters to pass to call back function |
||
| 1024 | * @param string $callBack Call back object reference, eg. 'EXT:indexed_search/class.crawler.php:&tx_indexedsearch_crawler' |
||
| 1025 | * @param integer $page_id Page ID to attach it to |
||
| 1026 | * @param integer $schedule Time at which to activate |
||
| 1027 | */ |
||
| 1028 | public function addQueueEntry_callBack($setId, $params, $callBack, $page_id = 0, $schedule = 0): void |
||
| 1029 | { |
||
| 1030 | if (! is_array($params)) { |
||
| 1031 | $params = []; |
||
| 1032 | } |
||
| 1033 | $params['_CALLBACKOBJ'] = $callBack; |
||
| 1034 | |||
| 1035 | GeneralUtility::makeInstance(ConnectionPool::class)->getConnectionForTable('tx_crawler_queue') |
||
| 1036 | ->insert( |
||
| 1037 | 'tx_crawler_queue', |
||
| 1038 | [ |
||
| 1039 | 'page_id' => (int) $page_id, |
||
| 1040 | 'parameters' => json_encode($params), |
||
| 1041 | 'scheduled' => (int) $schedule ?: $this->getCurrentTime(), |
||
| 1042 | 'exec_time' => 0, |
||
| 1043 | 'set_id' => (int) $setId, |
||
| 1044 | 'result_data' => '', |
||
| 1045 | ] |
||
| 1046 | ); |
||
| 1047 | } |
||
| 1048 | |||
| 1049 | /************************************ |
||
| 1050 | * |
||
| 1051 | * URL setting |
||
| 1052 | * |
||
| 1053 | ************************************/ |
||
| 1054 | |||
| 1055 | /** |
||
| 1056 | * Setting a URL for crawling: |
||
| 1057 | * |
||
| 1058 | * @param integer $id Page ID |
||
| 1059 | * @param string $url Complete URL |
||
| 1060 | * @param array $subCfg Sub configuration array (from TS config) |
||
| 1061 | * @param integer $tstamp Scheduled-time |
||
| 1062 | * @param string $configurationHash (optional) configuration hash |
||
| 1063 | * @param bool $skipInnerDuplicationCheck (optional) skip inner duplication check |
||
| 1064 | * @return bool |
||
| 1065 | */ |
||
| 1066 | 8 | public function addUrl( |
|
| 1067 | $id, |
||
| 1068 | $url, |
||
| 1069 | array $subCfg, |
||
| 1070 | $tstamp, |
||
| 1071 | $configurationHash = '', |
||
| 1072 | $skipInnerDuplicationCheck = false |
||
| 1073 | ) { |
||
| 1074 | 8 | $urlAdded = false; |
|
| 1075 | 8 | $rows = []; |
|
| 1076 | |||
| 1077 | // Creating parameters: |
||
| 1078 | $parameters = [ |
||
| 1079 | 8 | 'url' => $url, |
|
| 1080 | ]; |
||
| 1081 | |||
| 1082 | // fe user group simulation: |
||
| 1083 | 8 | $uGs = implode(',', array_unique(GeneralUtility::intExplode(',', $subCfg['userGroups'], true))); |
|
| 1084 | 8 | if ($uGs) { |
|
| 1085 | 1 | $parameters['feUserGroupList'] = $uGs; |
|
| 1086 | } |
||
| 1087 | |||
| 1088 | // Setting processing instructions |
||
| 1089 | 8 | $parameters['procInstructions'] = GeneralUtility::trimExplode(',', $subCfg['procInstrFilter']); |
|
| 1090 | 8 | if (is_array($subCfg['procInstrParams.'])) { |
|
| 1091 | 5 | $parameters['procInstrParams'] = $subCfg['procInstrParams.']; |
|
| 1092 | } |
||
| 1093 | |||
| 1094 | // Compile value array: |
||
| 1095 | 8 | $parameters_serialized = json_encode($parameters); |
|
| 1096 | $fieldArray = [ |
||
| 1097 | 8 | 'page_id' => (int) $id, |
|
| 1098 | 8 | 'parameters' => $parameters_serialized, |
|
| 1099 | 8 | 'parameters_hash' => GeneralUtility::shortMD5($parameters_serialized), |
|
| 1100 | 8 | 'configuration_hash' => $configurationHash, |
|
| 1101 | 8 | 'scheduled' => $tstamp, |
|
| 1102 | 8 | 'exec_time' => 0, |
|
| 1103 | 8 | 'set_id' => (int) $this->setID, |
|
| 1104 | 8 | 'result_data' => '', |
|
| 1105 | 8 | 'configuration' => $subCfg['key'], |
|
| 1106 | ]; |
||
| 1107 | |||
| 1108 | 8 | if ($this->registerQueueEntriesInternallyOnly) { |
|
| 1109 | //the entries will only be registered and not stored to the database |
||
| 1110 | 1 | $this->queueEntries[] = $fieldArray; |
|
| 1111 | } else { |
||
| 1112 | 7 | if (! $skipInnerDuplicationCheck) { |
|
| 1113 | // check if there is already an equal entry |
||
| 1114 | 6 | $rows = $this->queueRepository->getDuplicateQueueItemsIfExists( |
|
| 1115 | 6 | (bool) $this->extensionSettings['enableTimeslot'], |
|
| 1116 | $tstamp, |
||
| 1117 | 6 | $this->getCurrentTime(), |
|
| 1118 | 6 | $fieldArray['page_id'], |
|
| 1119 | 6 | $fieldArray['parameters_hash'] |
|
| 1120 | ); |
||
| 1121 | } |
||
| 1122 | |||
| 1123 | 7 | if (empty($rows)) { |
|
| 1124 | 6 | $connectionForCrawlerQueue = GeneralUtility::makeInstance(ConnectionPool::class)->getConnectionForTable('tx_crawler_queue'); |
|
| 1125 | 6 | $connectionForCrawlerQueue->insert( |
|
| 1126 | 6 | 'tx_crawler_queue', |
|
| 1127 | $fieldArray |
||
| 1128 | ); |
||
| 1129 | 6 | $uid = $connectionForCrawlerQueue->lastInsertId('tx_crawler_queue', 'qid'); |
|
| 1130 | 6 | $rows[] = $uid; |
|
| 1131 | 6 | $urlAdded = true; |
|
| 1132 | |||
| 1133 | 6 | $signalPayload = ['uid' => $uid, 'fieldArray' => $fieldArray]; |
|
| 1134 | 6 | SignalSlotUtility::emitSignal( |
|
| 1135 | 6 | self::class, |
|
| 1136 | 6 | SignalSlotUtility::SIGNAL_URL_ADDED_TO_QUEUE, |
|
| 1137 | $signalPayload |
||
| 1138 | ); |
||
| 1139 | } else { |
||
| 1140 | 3 | $signalPayload = ['rows' => $rows, 'fieldArray' => $fieldArray]; |
|
| 1141 | 3 | SignalSlotUtility::emitSignal( |
|
| 1142 | 3 | self::class, |
|
| 1143 | 3 | SignalSlotUtility::SIGNAL_DUPLICATE_URL_IN_QUEUE, |
|
| 1144 | $signalPayload |
||
| 1145 | ); |
||
| 1146 | } |
||
| 1147 | } |
||
| 1148 | |||
| 1149 | 8 | return $urlAdded; |
|
| 1150 | } |
||
| 1151 | |||
| 1152 | /** |
||
| 1153 | * Returns the current system time |
||
| 1154 | * |
||
| 1155 | * @return int |
||
| 1156 | */ |
||
| 1157 | 2 | public function getCurrentTime() |
|
| 1158 | { |
||
| 1159 | 2 | return time(); |
|
| 1160 | } |
||
| 1161 | |||
| 1162 | /************************************ |
||
| 1163 | * |
||
| 1164 | * URL reading |
||
| 1165 | * |
||
| 1166 | ************************************/ |
||
| 1167 | |||
| 1168 | /** |
||
| 1169 | * Read URL for single queue entry |
||
| 1170 | * |
||
| 1171 | * @param integer $queueId |
||
| 1172 | * @param boolean $force If set, will process even if exec_time has been set! |
||
| 1173 | * @return integer |
||
| 1174 | */ |
||
| 1175 | public function readUrl($queueId, $force = false) |
||
| 1176 | { |
||
| 1177 | $queryBuilder = GeneralUtility::makeInstance(ConnectionPool::class)->getQueryBuilderForTable($this->tableName); |
||
| 1178 | $ret = 0; |
||
| 1179 | $this->logger->debug('crawler-readurl start ' . microtime(true)); |
||
| 1180 | // Get entry: |
||
| 1181 | $queryBuilder |
||
| 1182 | ->select('*') |
||
| 1183 | ->from('tx_crawler_queue') |
||
| 1184 | ->where( |
||
| 1185 | $queryBuilder->expr()->eq('qid', $queryBuilder->createNamedParameter($queueId, PDO::PARAM_INT)) |
||
| 1186 | ); |
||
| 1187 | if (! $force) { |
||
| 1188 | $queryBuilder |
||
| 1189 | ->andWhere('exec_time = 0') |
||
| 1190 | ->andWhere('process_scheduled > 0'); |
||
| 1191 | } |
||
| 1192 | $queueRec = $queryBuilder->execute()->fetch(); |
||
| 1193 | |||
| 1194 | if (! is_array($queueRec)) { |
||
| 1195 | return; |
||
| 1196 | } |
||
| 1197 | |||
| 1198 | SignalSlotUtility::emitSignal( |
||
| 1199 | self::class, |
||
| 1200 | SignalSlotUtility::SIGNAL_QUEUEITEM_PREPROCESS, |
||
| 1201 | [$queueId, &$queueRec] |
||
| 1202 | ); |
||
| 1203 | |||
| 1204 | // Set exec_time to lock record: |
||
| 1205 | $field_array = ['exec_time' => $this->getCurrentTime()]; |
||
| 1206 | |||
| 1207 | if (isset($this->processID)) { |
||
| 1208 | //if mulitprocessing is used we need to store the id of the process which has handled this entry |
||
| 1209 | $field_array['process_id_completed'] = $this->processID; |
||
| 1210 | } |
||
| 1211 | |||
| 1212 | GeneralUtility::makeInstance(ConnectionPool::class)->getConnectionForTable('tx_crawler_queue') |
||
| 1213 | ->update( |
||
| 1214 | 'tx_crawler_queue', |
||
| 1215 | $field_array, |
||
| 1216 | ['qid' => (int) $queueId] |
||
| 1217 | ); |
||
| 1218 | |||
| 1219 | $result = $this->queueExecutor->executeQueueItem($queueRec, $this); |
||
| 1220 | if ($result['content'] === null) { |
||
| 1221 | $resultData = 'An errors happened'; |
||
| 1222 | } else { |
||
| 1223 | /** @var JsonCompatibilityConverter $jsonCompatibilityConverter */ |
||
| 1224 | $jsonCompatibilityConverter = GeneralUtility::makeInstance(JsonCompatibilityConverter::class); |
||
| 1225 | $resultData = $jsonCompatibilityConverter->convert($result['content']); |
||
| 1226 | } |
||
| 1227 | |||
| 1228 | //atm there's no need to point to specific pollable extensions |
||
| 1229 | if (is_array($GLOBALS['TYPO3_CONF_VARS']['EXTCONF']['crawler']['pollSuccess'])) { |
||
| 1230 | foreach ($GLOBALS['TYPO3_CONF_VARS']['EXTCONF']['crawler']['pollSuccess'] as $pollable) { |
||
| 1231 | // only check the success value if the instruction is runnig |
||
| 1232 | // it is important to name the pollSuccess key same as the procInstructions key |
||
| 1233 | if (is_array($resultData['parameters']['procInstructions']) |
||
| 1234 | && in_array( |
||
| 1235 | $pollable, |
||
| 1236 | $resultData['parameters']['procInstructions'], true |
||
| 1237 | ) |
||
| 1238 | ) { |
||
| 1239 | if (! empty($resultData['success'][$pollable]) && $resultData['success'][$pollable]) { |
||
| 1240 | $ret |= self::CLI_STATUS_POLLABLE_PROCESSED; |
||
| 1241 | } |
||
| 1242 | } |
||
| 1243 | } |
||
| 1244 | } |
||
| 1245 | |||
| 1246 | // Set result in log which also denotes the end of the processing of this entry. |
||
| 1247 | $field_array = ['result_data' => json_encode($result)]; |
||
| 1248 | |||
| 1249 | SignalSlotUtility::emitSignal( |
||
| 1250 | self::class, |
||
| 1251 | SignalSlotUtility::SIGNAL_QUEUEITEM_POSTPROCESS, |
||
| 1252 | [$queueId, &$field_array] |
||
| 1253 | ); |
||
| 1254 | |||
| 1255 | GeneralUtility::makeInstance(ConnectionPool::class)->getConnectionForTable('tx_crawler_queue') |
||
| 1256 | ->update( |
||
| 1257 | 'tx_crawler_queue', |
||
| 1258 | $field_array, |
||
| 1259 | ['qid' => (int) $queueId] |
||
| 1260 | ); |
||
| 1261 | |||
| 1262 | $this->logger->debug('crawler-readurl stop ' . microtime(true)); |
||
| 1263 | return $ret; |
||
| 1264 | } |
||
| 1265 | |||
| 1266 | /** |
||
| 1267 | * Read URL for not-yet-inserted log-entry |
||
| 1268 | * |
||
| 1269 | * @param array $field_array Queue field array, |
||
| 1270 | * |
||
| 1271 | * @return array|bool|mixed|string |
||
| 1272 | */ |
||
| 1273 | public function readUrlFromArray($field_array) |
||
| 1274 | { |
||
| 1275 | // Set exec_time to lock record: |
||
| 1276 | $field_array['exec_time'] = $this->getCurrentTime(); |
||
| 1277 | $connectionForCrawlerQueue = GeneralUtility::makeInstance(ConnectionPool::class)->getConnectionForTable($this->tableName); |
||
| 1278 | $connectionForCrawlerQueue->insert( |
||
| 1279 | $this->tableName, |
||
| 1280 | $field_array |
||
| 1281 | ); |
||
| 1282 | $queueId = $connectionForCrawlerQueue->lastInsertId($this->tableName, 'qid'); |
||
| 1283 | $result = $this->queueExecutor->executeQueueItem($field_array, $this); |
||
| 1284 | |||
| 1285 | // Set result in log which also denotes the end of the processing of this entry. |
||
| 1286 | $field_array = ['result_data' => json_encode($result)]; |
||
| 1287 | |||
| 1288 | SignalSlotUtility::emitSignal( |
||
| 1289 | self::class, |
||
| 1290 | SignalSlotUtility::SIGNAL_QUEUEITEM_POSTPROCESS, |
||
| 1291 | [$queueId, &$field_array] |
||
| 1292 | ); |
||
| 1293 | |||
| 1294 | $connectionForCrawlerQueue->update( |
||
| 1295 | $this->tableName, |
||
| 1296 | $field_array, |
||
| 1297 | ['qid' => $queueId] |
||
| 1298 | ); |
||
| 1299 | |||
| 1300 | return $result; |
||
| 1301 | } |
||
| 1302 | |||
| 1303 | /***************************** |
||
| 1304 | * |
||
| 1305 | * Compiling URLs to crawl - tools |
||
| 1306 | * |
||
| 1307 | *****************************/ |
||
| 1308 | |||
| 1309 | /** |
||
| 1310 | * @param integer $id Root page id to start from. |
||
| 1311 | * @param integer $depth Depth of tree, 0=only id-page, 1= on sublevel, 99 = infinite |
||
| 1312 | * @param integer $scheduledTime Unix Time when the URL is timed to be visited when put in queue |
||
| 1313 | * @param integer $reqMinute Number of requests per minute (creates the interleave between requests) |
||
| 1314 | * @param boolean $submitCrawlUrls If set, submits the URLs to queue in database (real crawling) |
||
| 1315 | * @param boolean $downloadCrawlUrls If set (and submitcrawlUrls is false) will fill $downloadUrls with entries) |
||
| 1316 | * @param array $incomingProcInstructions Array of processing instructions |
||
| 1317 | * @param array $configurationSelection Array of configuration keys |
||
| 1318 | * @return string |
||
| 1319 | */ |
||
| 1320 | public function getPageTreeAndUrls( |
||
| 1321 | $id, |
||
| 1322 | $depth, |
||
| 1323 | $scheduledTime, |
||
| 1324 | $reqMinute, |
||
| 1325 | $submitCrawlUrls, |
||
| 1326 | $downloadCrawlUrls, |
||
| 1327 | array $incomingProcInstructions, |
||
| 1328 | array $configurationSelection |
||
| 1329 | ) { |
||
| 1330 | $this->scheduledTime = $scheduledTime; |
||
| 1331 | $this->reqMinute = $reqMinute; |
||
| 1332 | $this->submitCrawlUrls = $submitCrawlUrls; |
||
| 1333 | $this->downloadCrawlUrls = $downloadCrawlUrls; |
||
| 1334 | $this->incomingProcInstructions = $incomingProcInstructions; |
||
| 1335 | $this->incomingConfigurationSelection = $configurationSelection; |
||
| 1336 | |||
| 1337 | $this->duplicateTrack = []; |
||
| 1338 | $this->downloadUrls = []; |
||
| 1339 | |||
| 1340 | // Drawing tree: |
||
| 1341 | /* @var PageTreeView $tree */ |
||
| 1342 | $tree = GeneralUtility::makeInstance(PageTreeView::class); |
||
| 1343 | $perms_clause = $this->getBackendUser()->getPagePermsClause(Permission::PAGE_SHOW); |
||
| 1344 | $tree->init('AND ' . $perms_clause); |
||
| 1345 | |||
| 1346 | $pageInfo = BackendUtility::readPageAccess($id, $perms_clause); |
||
| 1347 | if (is_array($pageInfo)) { |
||
| 1348 | // Set root row: |
||
| 1349 | $tree->tree[] = [ |
||
| 1350 | 'row' => $pageInfo, |
||
| 1351 | 'HTML' => $this->iconFactory->getIconForRecord('pages', $pageInfo, Icon::SIZE_SMALL), |
||
| 1352 | ]; |
||
| 1353 | } |
||
| 1354 | |||
| 1355 | // Get branch beneath: |
||
| 1356 | if ($depth) { |
||
| 1357 | $tree->getTree($id, $depth, ''); |
||
| 1358 | } |
||
| 1359 | |||
| 1360 | // Traverse page tree: |
||
| 1361 | $code = ''; |
||
| 1362 | |||
| 1363 | foreach ($tree->tree as $data) { |
||
| 1364 | $this->MP = false; |
||
| 1365 | |||
| 1366 | // recognize mount points |
||
| 1367 | if ($data['row']['doktype'] === PageRepository::DOKTYPE_MOUNTPOINT) { |
||
| 1368 | $mountpage = $this->pageRepository->getPage($data['row']['uid']); |
||
| 1369 | |||
| 1370 | // fetch mounted pages |
||
| 1371 | $this->MP = $mountpage[0]['mount_pid'] . '-' . $data['row']['uid']; |
||
| 1372 | |||
| 1373 | $mountTree = GeneralUtility::makeInstance(PageTreeView::class); |
||
| 1374 | $mountTree->init('AND ' . $perms_clause); |
||
| 1375 | $mountTree->getTree($mountpage[0]['mount_pid'], $depth); |
||
| 1376 | |||
| 1377 | foreach ($mountTree->tree as $mountData) { |
||
| 1378 | $code .= $this->drawURLs_addRowsForPage( |
||
| 1379 | $mountData['row'], |
||
| 1380 | $mountData['HTML'] . BackendUtility::getRecordTitle('pages', $mountData['row'], true) |
||
| 1381 | ); |
||
| 1382 | } |
||
| 1383 | |||
| 1384 | // replace page when mount_pid_ol is enabled |
||
| 1385 | if ($mountpage[0]['mount_pid_ol']) { |
||
| 1386 | $data['row']['uid'] = $mountpage[0]['mount_pid']; |
||
| 1387 | } else { |
||
| 1388 | // if the mount_pid_ol is not set the MP must not be used for the mountpoint page |
||
| 1389 | $this->MP = false; |
||
| 1390 | } |
||
| 1391 | } |
||
| 1392 | |||
| 1393 | $code .= $this->drawURLs_addRowsForPage( |
||
| 1394 | $data['row'], |
||
| 1395 | $data['HTML'] . BackendUtility::getRecordTitle('pages', $data['row'], true) |
||
| 1396 | ); |
||
| 1397 | } |
||
| 1398 | |||
| 1399 | return $code; |
||
| 1400 | } |
||
| 1401 | |||
| 1402 | /** |
||
| 1403 | * Expands exclude string |
||
| 1404 | * |
||
| 1405 | * @param string $excludeString Exclude string |
||
| 1406 | * @return array |
||
| 1407 | */ |
||
| 1408 | 2 | public function expandExcludeString($excludeString) |
|
| 1409 | { |
||
| 1410 | // internal static caches; |
||
| 1411 | 2 | static $expandedExcludeStringCache; |
|
| 1412 | 2 | static $treeCache; |
|
| 1413 | |||
| 1414 | 2 | if (empty($expandedExcludeStringCache[$excludeString])) { |
|
| 1415 | 2 | $pidList = []; |
|
| 1416 | |||
| 1417 | 2 | if (! empty($excludeString)) { |
|
| 1418 | /** @var PageTreeView $tree */ |
||
| 1419 | 1 | $tree = GeneralUtility::makeInstance(PageTreeView::class); |
|
| 1420 | 1 | $tree->init('AND ' . $this->getBackendUser()->getPagePermsClause(Permission::PAGE_SHOW)); |
|
| 1421 | |||
| 1422 | 1 | $excludeParts = GeneralUtility::trimExplode(',', $excludeString); |
|
| 1423 | |||
| 1424 | 1 | foreach ($excludeParts as $excludePart) { |
|
| 1425 | 1 | [$pid, $depth] = GeneralUtility::trimExplode('+', $excludePart); |
|
| 1426 | |||
| 1427 | // default is "page only" = "depth=0" |
||
| 1428 | 1 | if (empty($depth)) { |
|
| 1429 | 1 | $depth = (stristr($excludePart, '+')) ? 99 : 0; |
|
| 1430 | } |
||
| 1431 | |||
| 1432 | 1 | $pidList[] = (int) $pid; |
|
| 1433 | |||
| 1434 | 1 | if ($depth > 0) { |
|
| 1435 | if (empty($treeCache[$pid][$depth])) { |
||
| 1436 | $tree->reset(); |
||
| 1437 | $tree->getTree($pid, $depth); |
||
| 1438 | $treeCache[$pid][$depth] = $tree->tree; |
||
| 1439 | } |
||
| 1440 | |||
| 1441 | foreach ($treeCache[$pid][$depth] as $data) { |
||
| 1442 | $pidList[] = (int) $data['row']['uid']; |
||
| 1443 | } |
||
| 1444 | } |
||
| 1445 | } |
||
| 1446 | } |
||
| 1447 | |||
| 1448 | 2 | $expandedExcludeStringCache[$excludeString] = array_unique($pidList); |
|
| 1449 | } |
||
| 1450 | |||
| 1451 | 2 | return $expandedExcludeStringCache[$excludeString]; |
|
| 1452 | } |
||
| 1453 | |||
| 1454 | /** |
||
| 1455 | * Create the rows for display of the page tree |
||
| 1456 | * For each page a number of rows are shown displaying GET variable configuration |
||
| 1457 | */ |
||
| 1458 | public function drawURLs_addRowsForPage(array $pageRow, string $pageTitle): string |
||
| 1459 | { |
||
| 1460 | $skipMessage = ''; |
||
| 1461 | |||
| 1462 | // Get list of configurations |
||
| 1463 | $configurations = $this->getUrlsForPageRow($pageRow, $skipMessage); |
||
| 1464 | $configurations = ConfigurationService::removeDisallowedConfigurations($this->incomingConfigurationSelection, $configurations); |
||
| 1465 | |||
| 1466 | // Traverse parameter combinations: |
||
| 1467 | $c = 0; |
||
| 1468 | $content = ''; |
||
| 1469 | if (! empty($configurations)) { |
||
| 1470 | foreach ($configurations as $confKey => $confArray) { |
||
| 1471 | |||
| 1472 | // Title column: |
||
| 1473 | if (! $c) { |
||
| 1474 | $titleClm = '<td rowspan="' . count($configurations) . '">' . $pageTitle . '</td>'; |
||
| 1475 | } else { |
||
| 1476 | $titleClm = ''; |
||
| 1477 | } |
||
| 1478 | |||
| 1479 | if (! in_array($pageRow['uid'], $this->expandExcludeString($confArray['subCfg']['exclude']), true)) { |
||
| 1480 | |||
| 1481 | // URL list: |
||
| 1482 | $urlList = $this->urlListFromUrlArray( |
||
| 1483 | $confArray, |
||
| 1484 | $pageRow, |
||
| 1485 | $this->scheduledTime, |
||
| 1486 | $this->reqMinute, |
||
| 1487 | $this->submitCrawlUrls, |
||
| 1488 | $this->downloadCrawlUrls, |
||
| 1489 | $this->duplicateTrack, |
||
| 1490 | $this->downloadUrls, |
||
| 1491 | // if empty the urls won't be filtered by processing instructions |
||
| 1492 | $this->incomingProcInstructions |
||
| 1493 | ); |
||
| 1494 | |||
| 1495 | // Expanded parameters: |
||
| 1496 | $paramExpanded = ''; |
||
| 1497 | $calcAccu = []; |
||
| 1498 | $calcRes = 1; |
||
| 1499 | foreach ($confArray['paramExpanded'] as $gVar => $gVal) { |
||
| 1500 | $paramExpanded .= ' |
||
| 1501 | <tr> |
||
| 1502 | <td>' . htmlspecialchars('&' . $gVar . '=') . '<br/>' . |
||
| 1503 | '(' . count($gVal) . ')' . |
||
| 1504 | '</td> |
||
| 1505 | <td nowrap="nowrap">' . nl2br(htmlspecialchars(implode(chr(10), $gVal))) . '</td> |
||
| 1506 | </tr> |
||
| 1507 | '; |
||
| 1508 | $calcRes *= count($gVal); |
||
| 1509 | $calcAccu[] = count($gVal); |
||
| 1510 | } |
||
| 1511 | $paramExpanded = '<table>' . $paramExpanded . '</table>'; |
||
| 1512 | $paramExpanded .= 'Comb: ' . implode('*', $calcAccu) . '=' . $calcRes; |
||
| 1513 | |||
| 1514 | // Options |
||
| 1515 | $optionValues = ''; |
||
| 1516 | if ($confArray['subCfg']['userGroups']) { |
||
| 1517 | $optionValues .= 'User Groups: ' . $confArray['subCfg']['userGroups'] . '<br/>'; |
||
| 1518 | } |
||
| 1519 | if ($confArray['subCfg']['procInstrFilter']) { |
||
| 1520 | $optionValues .= 'ProcInstr: ' . $confArray['subCfg']['procInstrFilter'] . '<br/>'; |
||
| 1521 | } |
||
| 1522 | |||
| 1523 | // Compile row: |
||
| 1524 | $content .= ' |
||
| 1525 | <tr> |
||
| 1526 | ' . $titleClm . ' |
||
| 1527 | <td>' . htmlspecialchars($confKey) . '</td> |
||
| 1528 | <td>' . nl2br(htmlspecialchars(rawurldecode(trim(str_replace('&', chr(10) . '&', GeneralUtility::implodeArrayForUrl('', $confArray['paramParsed'])))))) . '</td> |
||
| 1529 | <td>' . $paramExpanded . '</td> |
||
| 1530 | <td nowrap="nowrap">' . $urlList . '</td> |
||
| 1531 | <td nowrap="nowrap">' . $optionValues . '</td> |
||
| 1532 | <td nowrap="nowrap">' . DebugUtility::viewArray($confArray['subCfg']['procInstrParams.']) . '</td> |
||
| 1533 | </tr>'; |
||
| 1534 | } else { |
||
| 1535 | $content .= '<tr> |
||
| 1536 | ' . $titleClm . ' |
||
| 1537 | <td>' . htmlspecialchars($confKey) . '</td> |
||
| 1538 | <td colspan="5"><em>No entries</em> (Page is excluded in this configuration)</td> |
||
| 1539 | </tr>'; |
||
| 1540 | } |
||
| 1541 | |||
| 1542 | $c++; |
||
| 1543 | } |
||
| 1544 | } else { |
||
| 1545 | $message = ! empty($skipMessage) ? ' (' . $skipMessage . ')' : ''; |
||
| 1546 | |||
| 1547 | // Compile row: |
||
| 1548 | $content .= ' |
||
| 1549 | <tr> |
||
| 1550 | <td>' . $pageTitle . '</td> |
||
| 1551 | <td colspan="6"><em>No entries</em>' . $message . '</td> |
||
| 1552 | </tr>'; |
||
| 1553 | } |
||
| 1554 | |||
| 1555 | return $content; |
||
| 1556 | } |
||
| 1557 | |||
| 1558 | /***************************** |
||
| 1559 | * |
||
| 1560 | * CLI functions |
||
| 1561 | * |
||
| 1562 | *****************************/ |
||
| 1563 | |||
| 1564 | /** |
||
| 1565 | * Running the functionality of the CLI (crawling URLs from queue) |
||
| 1566 | */ |
||
| 1567 | public function CLI_run(int $countInARun, int $sleepTime, int $sleepAfterFinish): int |
||
| 1568 | { |
||
| 1569 | $result = 0; |
||
| 1570 | $counter = 0; |
||
| 1571 | |||
| 1572 | // First, run hooks: |
||
| 1573 | foreach ($GLOBALS['TYPO3_CONF_VARS']['EXTCONF']['crawler']['cli_hooks'] ?? [] as $objRef) { |
||
| 1574 | trigger_error( |
||
| 1575 | 'This hook (crawler/cli_hooks) is deprecated since 9.1.5 and will be removed when dropping support for TYPO3 9LTS and 10LTS', |
||
| 1576 | E_USER_DEPRECATED |
||
| 1577 | ); |
||
| 1578 | $hookObj = GeneralUtility::makeInstance($objRef); |
||
| 1579 | if (is_object($hookObj)) { |
||
| 1580 | $hookObj->crawler_init($this); |
||
| 1581 | } |
||
| 1582 | } |
||
| 1583 | |||
| 1584 | // Clean up the queue |
||
| 1585 | $this->queueRepository->cleanupQueue(); |
||
| 1586 | |||
| 1587 | // Select entries: |
||
| 1588 | $rows = $this->queueRepository->fetchRecordsToBeCrawled($countInARun); |
||
| 1589 | |||
| 1590 | if (! empty($rows)) { |
||
| 1591 | $quidList = []; |
||
| 1592 | |||
| 1593 | foreach ($rows as $r) { |
||
| 1594 | $quidList[] = $r['qid']; |
||
| 1595 | } |
||
| 1596 | |||
| 1597 | $processId = $this->CLI_buildProcessId(); |
||
| 1598 | |||
| 1599 | //save the number of assigned queue entries to determine how many have been processed later |
||
| 1600 | $numberOfAffectedRows = $this->queueRepository->updateProcessIdAndSchedulerForQueueIds($quidList, $processId); |
||
| 1601 | $this->processRepository->updateProcessAssignItemsCount($numberOfAffectedRows, $processId); |
||
| 1602 | |||
| 1603 | if ($numberOfAffectedRows !== count($quidList)) { |
||
| 1604 | return ($result | self::CLI_STATUS_ABORTED); |
||
| 1605 | } |
||
| 1606 | |||
| 1607 | foreach ($rows as $r) { |
||
| 1608 | $result |= $this->readUrl($r['qid']); |
||
| 1609 | |||
| 1610 | $counter++; |
||
| 1611 | // Just to relax the system |
||
| 1612 | usleep((int) $sleepTime); |
||
| 1613 | |||
| 1614 | // if during the start and the current read url the cli has been disable we need to return from the function |
||
| 1615 | // mark the process NOT as ended. |
||
| 1616 | if ($this->crawler->isDisabled()) { |
||
| 1617 | return ($result | self::CLI_STATUS_ABORTED); |
||
| 1618 | } |
||
| 1619 | |||
| 1620 | if (! $this->processRepository->isProcessActive($this->CLI_buildProcessId())) { |
||
| 1621 | $this->CLI_debug('conflict / timeout (' . $this->CLI_buildProcessId() . ')'); |
||
| 1622 | $result |= self::CLI_STATUS_ABORTED; |
||
| 1623 | //possible timeout |
||
| 1624 | break; |
||
| 1625 | } |
||
| 1626 | } |
||
| 1627 | |||
| 1628 | sleep((int) $sleepAfterFinish); |
||
| 1629 | } |
||
| 1630 | |||
| 1631 | if ($counter > 0) { |
||
| 1632 | $result |= self::CLI_STATUS_PROCESSED; |
||
| 1633 | } |
||
| 1634 | |||
| 1635 | return $result; |
||
| 1636 | } |
||
| 1637 | |||
| 1638 | /** |
||
| 1639 | * Activate hooks |
||
| 1640 | * @deprecated |
||
| 1641 | */ |
||
| 1642 | public function CLI_runHooks(): void |
||
| 1643 | { |
||
| 1644 | foreach ($GLOBALS['TYPO3_CONF_VARS']['EXTCONF']['crawler']['cli_hooks'] ?? [] as $objRef) { |
||
| 1645 | $hookObj = GeneralUtility::makeInstance($objRef); |
||
| 1646 | if (is_object($hookObj)) { |
||
| 1647 | $hookObj->crawler_init($this); |
||
| 1648 | } |
||
| 1649 | } |
||
| 1650 | } |
||
| 1651 | |||
| 1652 | /** |
||
| 1653 | * Try to acquire a new process with the given id |
||
| 1654 | * also performs some auto-cleanup for orphan processes |
||
| 1655 | * @param string $id identification string for the process |
||
| 1656 | * @return boolean |
||
| 1657 | * @todo preemption might not be the most elegant way to clean up |
||
| 1658 | */ |
||
| 1659 | public function CLI_checkAndAcquireNewProcess($id) |
||
| 1660 | { |
||
| 1661 | $ret = true; |
||
| 1662 | |||
| 1663 | $systemProcessId = getmypid(); |
||
| 1664 | if (! $systemProcessId) { |
||
| 1665 | return false; |
||
| 1666 | } |
||
| 1667 | |||
| 1668 | $processCount = 0; |
||
| 1669 | $orphanProcesses = []; |
||
| 1670 | |||
| 1671 | $activeProcesses = $this->processRepository->findAllActive(); |
||
| 1672 | $currentTime = $this->getCurrentTime(); |
||
| 1673 | |||
| 1674 | /** @var Process $process */ |
||
| 1675 | foreach ($activeProcesses as $process) { |
||
| 1676 | if ($process->getTtl() < $currentTime) { |
||
| 1677 | $orphanProcesses[] = $process->getProcessId(); |
||
| 1678 | } else { |
||
| 1679 | $processCount++; |
||
| 1680 | } |
||
| 1681 | } |
||
| 1682 | |||
| 1683 | // if there are less than allowed active processes then add a new one |
||
| 1684 | if ($processCount < (int) $this->extensionSettings['processLimit']) { |
||
| 1685 | GeneralUtility::makeInstance(ConnectionPool::class)->getConnectionForTable('tx_crawler_process')->insert( |
||
| 1686 | 'tx_crawler_process', |
||
| 1687 | [ |
||
| 1688 | 'process_id' => $id, |
||
| 1689 | 'active' => 1, |
||
| 1690 | 'ttl' => $currentTime + (int) $this->extensionSettings['processMaxRunTime'], |
||
| 1691 | 'system_process_id' => $systemProcessId, |
||
| 1692 | ] |
||
| 1693 | ); |
||
| 1694 | } else { |
||
| 1695 | $ret = false; |
||
| 1696 | } |
||
| 1697 | |||
| 1698 | $this->processRepository->deleteProcessesMarkedAsDeleted(); |
||
| 1699 | $this->CLI_releaseProcesses($orphanProcesses); |
||
| 1700 | |||
| 1701 | return $ret; |
||
| 1702 | } |
||
| 1703 | |||
| 1704 | /** |
||
| 1705 | * Release a process and the required resources |
||
| 1706 | * |
||
| 1707 | * @param mixed $releaseIds string with a single process-id or array with multiple process-ids |
||
| 1708 | * @return boolean |
||
| 1709 | */ |
||
| 1710 | public function CLI_releaseProcesses($releaseIds) |
||
| 1711 | { |
||
| 1712 | $queryBuilder = GeneralUtility::makeInstance(ConnectionPool::class)->getQueryBuilderForTable($this->tableName); |
||
| 1713 | |||
| 1714 | if (! is_array($releaseIds)) { |
||
| 1715 | $releaseIds = [$releaseIds]; |
||
| 1716 | } |
||
| 1717 | |||
| 1718 | if (empty($releaseIds)) { |
||
| 1719 | //nothing to release |
||
| 1720 | return false; |
||
| 1721 | } |
||
| 1722 | |||
| 1723 | // some kind of 2nd chance algo - this way you need at least 2 processes to have a real cleanup |
||
| 1724 | // this ensures that a single process can't mess up the entire process table |
||
| 1725 | |||
| 1726 | // mark all processes as deleted which have no "waiting" queue-entires and which are not active |
||
| 1727 | |||
| 1728 | $queryBuilder |
||
| 1729 | ->update($this->tableName, 'q') |
||
| 1730 | ->where( |
||
| 1731 | 'q.process_id IN(SELECT p.process_id FROM tx_crawler_process as p WHERE p.active = 0)' |
||
| 1732 | ) |
||
| 1733 | ->set('q.process_scheduled', 0) |
||
| 1734 | ->set('q.process_id', '') |
||
| 1735 | ->execute(); |
||
| 1736 | |||
| 1737 | // FIXME: Not entirely sure that this is equivalent to the previous version |
||
| 1738 | $queryBuilder->resetQueryPart('set'); |
||
| 1739 | |||
| 1740 | $queryBuilder |
||
| 1741 | ->update('tx_crawler_process') |
||
| 1742 | ->where( |
||
| 1743 | $queryBuilder->expr()->eq('active', 0), |
||
| 1744 | 'process_id IN(SELECT q.process_id FROM tx_crawler_queue as q WHERE q.exec_time = 0)' |
||
| 1745 | ) |
||
| 1746 | ->set('system_process_id', 0) |
||
| 1747 | ->execute(); |
||
| 1748 | |||
| 1749 | $this->processRepository->markRequestedProcessesAsNotActive($releaseIds); |
||
| 1750 | $this->queueRepository->unsetProcessScheduledAndProcessIdForQueueEntries($releaseIds); |
||
| 1751 | |||
| 1752 | return true; |
||
| 1753 | } |
||
| 1754 | |||
| 1755 | /** |
||
| 1756 | * Create a unique Id for the current process |
||
| 1757 | * |
||
| 1758 | * @return string the ID |
||
| 1759 | */ |
||
| 1760 | 1 | public function CLI_buildProcessId() |
|
| 1761 | { |
||
| 1762 | 1 | if (! $this->processID) { |
|
| 1763 | $this->processID = GeneralUtility::shortMD5(microtime(true)); |
||
| 1764 | } |
||
| 1765 | 1 | return $this->processID; |
|
| 1766 | } |
||
| 1767 | |||
| 1768 | /** |
||
| 1769 | * Prints a message to the stdout (only if debug-mode is enabled) |
||
| 1770 | * |
||
| 1771 | * @param string $msg the message |
||
| 1772 | * @deprecated |
||
| 1773 | * @codeCoverageIgnore |
||
| 1774 | */ |
||
| 1775 | public function CLI_debug($msg): void |
||
| 1776 | { |
||
| 1777 | if ((int) $this->extensionSettings['processDebug']) { |
||
| 1778 | echo $msg . "\n"; |
||
| 1779 | flush(); |
||
| 1780 | } |
||
| 1781 | } |
||
| 1782 | |||
| 1783 | /** |
||
| 1784 | * Cleans up entries that stayed for too long in the queue. These are: |
||
| 1785 | * - processed entries that are over 1.5 days in age |
||
| 1786 | * - scheduled entries that are over 7 days old |
||
| 1787 | * |
||
| 1788 | * @deprecated |
||
| 1789 | */ |
||
| 1790 | 1 | public function cleanUpOldQueueEntries(): void |
|
| 1791 | { |
||
| 1792 | // 24*60*60 Seconds in 24 hours |
||
| 1793 | 1 | $processedAgeInSeconds = $this->extensionSettings['cleanUpProcessedAge'] * 86400; |
|
| 1794 | 1 | $scheduledAgeInSeconds = $this->extensionSettings['cleanUpScheduledAge'] * 86400; |
|
| 1795 | |||
| 1796 | 1 | $now = time(); |
|
| 1797 | 1 | $condition = '(exec_time<>0 AND exec_time<' . ($now - $processedAgeInSeconds) . ') OR scheduled<=' . ($now - $scheduledAgeInSeconds); |
|
| 1798 | 1 | $this->flushQueue($condition); |
|
| 1799 | 1 | } |
|
| 1800 | |||
| 1801 | /** |
||
| 1802 | * Removes queue entries |
||
| 1803 | * |
||
| 1804 | * @param string $where SQL related filter for the entries which should be removed |
||
| 1805 | * |
||
| 1806 | * @deprecated |
||
| 1807 | */ |
||
| 1808 | 5 | protected function flushQueue($where = ''): void |
|
| 1809 | { |
||
| 1810 | 5 | $realWhere = strlen((string) $where) > 0 ? $where : '1=1'; |
|
| 1811 | |||
| 1812 | 5 | $queryBuilder = $this->getQueryBuilder($this->tableName); |
|
| 1813 | |||
| 1814 | $groups = $queryBuilder |
||
| 1815 | 5 | ->selectLiteral('DISTINCT set_id') |
|
| 1816 | 5 | ->from($this->tableName) |
|
| 1817 | 5 | ->where($realWhere) |
|
| 1818 | 5 | ->execute() |
|
| 1819 | 5 | ->fetchAll(); |
|
| 1820 | 5 | if (is_array($groups)) { |
|
| 1821 | 5 | foreach ($groups as $group) { |
|
| 1822 | $subSet = $queryBuilder |
||
| 1823 | 4 | ->select('qid', 'set_id') |
|
| 1824 | 4 | ->from($this->tableName) |
|
| 1825 | 4 | ->where( |
|
| 1826 | 4 | $realWhere, |
|
| 1827 | 4 | $queryBuilder->expr()->eq('set_id', $group['set_id']) |
|
| 1828 | ) |
||
| 1829 | 4 | ->execute() |
|
| 1830 | 4 | ->fetchAll(); |
|
| 1831 | |||
| 1832 | 4 | $payLoad = ['subSet' => $subSet]; |
|
| 1833 | 4 | SignalSlotUtility::emitSignal( |
|
| 1834 | 4 | self::class, |
|
| 1835 | 4 | SignalSlotUtility::SIGNAL_QUEUE_ENTRY_FLUSH, |
|
| 1836 | $payLoad |
||
| 1837 | ); |
||
| 1838 | } |
||
| 1839 | } |
||
| 1840 | |||
| 1841 | $queryBuilder |
||
| 1842 | 5 | ->delete($this->tableName) |
|
| 1843 | 5 | ->where($realWhere) |
|
| 1844 | 5 | ->execute(); |
|
| 1845 | 5 | } |
|
| 1846 | |||
| 1847 | /** |
||
| 1848 | * This method determines duplicates for a queue entry with the same parameters and this timestamp. |
||
| 1849 | * If the timestamp is in the past, it will check if there is any unprocessed queue entry in the past. |
||
| 1850 | * If the timestamp is in the future it will check, if the queued entry has exactly the same timestamp |
||
| 1851 | * |
||
| 1852 | * @param int $tstamp |
||
| 1853 | * @param array $fieldArray |
||
| 1854 | * |
||
| 1855 | * @return array |
||
| 1856 | * @deprecated |
||
| 1857 | */ |
||
| 1858 | 5 | protected function getDuplicateRowsIfExist($tstamp, $fieldArray) |
|
| 1859 | { |
||
| 1860 | 5 | $rows = []; |
|
| 1861 | |||
| 1862 | 5 | $currentTime = $this->getCurrentTime(); |
|
| 1863 | |||
| 1864 | 5 | $queryBuilder = GeneralUtility::makeInstance(ConnectionPool::class)->getQueryBuilderForTable($this->tableName); |
|
| 1865 | $queryBuilder |
||
| 1866 | 5 | ->select('qid') |
|
| 1867 | 5 | ->from('tx_crawler_queue'); |
|
| 1868 | //if this entry is scheduled with "now" |
||
| 1869 | 5 | if ($tstamp <= $currentTime) { |
|
| 1870 | 2 | if ($this->extensionSettings['enableTimeslot']) { |
|
| 1871 | 1 | $timeBegin = $currentTime - 100; |
|
| 1872 | 1 | $timeEnd = $currentTime + 100; |
|
| 1873 | $queryBuilder |
||
| 1874 | 1 | ->where( |
|
| 1875 | 1 | 'scheduled BETWEEN ' . $timeBegin . ' AND ' . $timeEnd . '' |
|
| 1876 | ) |
||
| 1877 | 1 | ->orWhere( |
|
| 1878 | 1 | $queryBuilder->expr()->lte('scheduled', $currentTime) |
|
| 1879 | ); |
||
| 1880 | } else { |
||
| 1881 | $queryBuilder |
||
| 1882 | 1 | ->where( |
|
| 1883 | 2 | $queryBuilder->expr()->lte('scheduled', $currentTime) |
|
| 1884 | ); |
||
| 1885 | } |
||
| 1886 | 3 | } elseif ($tstamp > $currentTime) { |
|
| 1887 | //entry with a timestamp in the future need to have the same schedule time |
||
| 1888 | $queryBuilder |
||
| 1889 | 3 | ->where( |
|
| 1890 | 3 | $queryBuilder->expr()->eq('scheduled', $tstamp) |
|
| 1891 | ); |
||
| 1892 | } |
||
| 1893 | |||
| 1894 | $queryBuilder |
||
| 1895 | 5 | ->andWhere('NOT exec_time') |
|
| 1896 | 5 | ->andWhere('NOT process_id') |
|
| 1897 | 5 | ->andWhere($queryBuilder->expr()->eq('page_id', $queryBuilder->createNamedParameter($fieldArray['page_id'], PDO::PARAM_INT))) |
|
| 1898 | 5 | ->andWhere($queryBuilder->expr()->eq('parameters_hash', $queryBuilder->createNamedParameter($fieldArray['parameters_hash'], PDO::PARAM_STR))); |
|
| 1899 | |||
| 1900 | 5 | $statement = $queryBuilder->execute(); |
|
| 1901 | |||
| 1902 | 5 | while ($row = $statement->fetch()) { |
|
| 1903 | 5 | $rows[] = $row['qid']; |
|
| 1904 | } |
||
| 1905 | |||
| 1906 | 5 | return $rows; |
|
| 1907 | } |
||
| 1908 | |||
| 1909 | /** |
||
| 1910 | * Returns a md5 hash generated from a serialized configuration array. |
||
| 1911 | * |
||
| 1912 | * @return string |
||
| 1913 | */ |
||
| 1914 | 10 | protected function getConfigurationHash(array $configuration) |
|
| 1915 | { |
||
| 1916 | 10 | unset($configuration['paramExpanded']); |
|
| 1917 | 10 | unset($configuration['URLs']); |
|
| 1918 | 10 | return md5(serialize($configuration)); |
|
| 1919 | } |
||
| 1920 | |||
| 1921 | /** |
||
| 1922 | * Build a URL from a Page and the Query String. If the page has a Site configuration, it can be built by using |
||
| 1923 | * the Site instance. |
||
| 1924 | * |
||
| 1925 | * @param int $httpsOrHttp see tx_crawler_configuration.force_ssl |
||
| 1926 | * @throws SiteNotFoundException |
||
| 1927 | * @throws InvalidRouteArgumentsException |
||
| 1928 | * |
||
| 1929 | * @deprecated Using CrawlerController::getUrlFromPageAndQueryParameters() is deprecated since 9.1.1 and will be removed in v11.x, please use UrlService->getUrlFromPageAndQueryParameters() instead. |
||
| 1930 | * @codeCoverageIgnore |
||
| 1931 | */ |
||
| 1932 | protected function getUrlFromPageAndQueryParameters(int $pageId, string $queryString, ?string $alternativeBaseUrl, int $httpsOrHttp): UriInterface |
||
| 1933 | { |
||
| 1934 | $urlService = new UrlService(); |
||
| 1935 | return $urlService->getUrlFromPageAndQueryParameters($pageId, $queryString, $alternativeBaseUrl, $httpsOrHttp); |
||
| 1936 | } |
||
| 1937 | |||
| 1938 | 1 | protected function swapIfFirstIsLargerThanSecond(array $reg): array |
|
| 1948 | } |
||
| 1949 | |||
| 1950 | 6 | private function getMaximumUrlsToCompile(): int |
|
| 1951 | { |
||
| 1952 | 6 | return $this->maximumUrlsToCompile; |
|
| 1953 | } |
||
| 1954 | |||
| 1955 | /** |
||
| 1956 | * @return BackendUserAuthentication |
||
| 1957 | */ |
||
| 1958 | 2 | private function getBackendUser() |
|
| 1959 | { |
||
| 1960 | // Make sure the _cli_ user is loaded |
||
| 1961 | 2 | Bootstrap::initializeBackendAuthentication(); |
|
| 1962 | 2 | if ($this->backendUser === null) { |
|
| 1963 | 2 | $this->backendUser = $GLOBALS['BE_USER']; |
|
| 1964 | } |
||
| 1965 | 2 | return $this->backendUser; |
|
| 1966 | } |
||
| 1967 | |||
| 1968 | /** |
||
| 1969 | * Get querybuilder for given table |
||
| 1970 | * |
||
| 1971 | * @return QueryBuilder |
||
| 1972 | */ |
||
| 1973 | 12 | private function getQueryBuilder(string $table) |
|
| 1976 | } |
||
| 1977 | } |
||
| 1978 |
Loading history...
