1
|
|
|
<?php |
2
|
|
|
|
3
|
|
|
declare(strict_types=1); |
4
|
|
|
|
5
|
|
|
namespace AOE\Crawler\Controller; |
6
|
|
|
|
7
|
|
|
/*************************************************************** |
8
|
|
|
* Copyright notice |
9
|
|
|
* |
10
|
|
|
* (c) 2019 AOE GmbH <[email protected]> |
11
|
|
|
* |
12
|
|
|
* All rights reserved |
13
|
|
|
* |
14
|
|
|
* This script is part of the TYPO3 project. The TYPO3 project is |
15
|
|
|
* free software; you can redistribute it and/or modify |
16
|
|
|
* it under the terms of the GNU General Public License as published by |
17
|
|
|
* the Free Software Foundation; either version 3 of the License, or |
18
|
|
|
* (at your option) any later version. |
19
|
|
|
* |
20
|
|
|
* The GNU General Public License can be found at |
21
|
|
|
* http://www.gnu.org/copyleft/gpl.html. |
22
|
|
|
* |
23
|
|
|
* This script is distributed in the hope that it will be useful, |
24
|
|
|
* but WITHOUT ANY WARRANTY; without even the implied warranty of |
25
|
|
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
26
|
|
|
* GNU General Public License for more details. |
27
|
|
|
* |
28
|
|
|
* This copyright notice MUST APPEAR in all copies of the script! |
29
|
|
|
***************************************************************/ |
30
|
|
|
|
31
|
|
|
use AOE\Crawler\Configuration\ExtensionConfigurationProvider; |
32
|
|
|
use AOE\Crawler\Domain\Repository\ConfigurationRepository; |
33
|
|
|
use AOE\Crawler\Domain\Repository\ProcessRepository; |
34
|
|
|
use AOE\Crawler\Domain\Repository\QueueRepository; |
35
|
|
|
use AOE\Crawler\Event\EventDispatcher; |
36
|
|
|
use AOE\Crawler\QueueExecutor; |
37
|
|
|
use AOE\Crawler\Utility\SignalSlotUtility; |
38
|
|
|
use Psr\Http\Message\UriInterface; |
39
|
|
|
use Psr\Log\LoggerAwareInterface; |
40
|
|
|
use Psr\Log\LoggerAwareTrait; |
41
|
|
|
use TYPO3\CMS\Backend\Tree\View\PageTreeView; |
42
|
|
|
use TYPO3\CMS\Backend\Utility\BackendUtility; |
43
|
|
|
use TYPO3\CMS\Core\Authentication\BackendUserAuthentication; |
44
|
|
|
use TYPO3\CMS\Core\Core\Bootstrap; |
45
|
|
|
use TYPO3\CMS\Core\Core\Environment; |
46
|
|
|
use TYPO3\CMS\Core\Database\Connection; |
47
|
|
|
use TYPO3\CMS\Core\Database\ConnectionPool; |
48
|
|
|
use TYPO3\CMS\Core\Database\Query\Restriction\DeletedRestriction; |
49
|
|
|
use TYPO3\CMS\Core\Http\Uri; |
50
|
|
|
use TYPO3\CMS\Core\Imaging\Icon; |
51
|
|
|
use TYPO3\CMS\Core\Imaging\IconFactory; |
52
|
|
|
use TYPO3\CMS\Core\Routing\SiteMatcher; |
53
|
|
|
use TYPO3\CMS\Core\Site\Entity\Site; |
54
|
|
|
use TYPO3\CMS\Core\Type\Bitmask\Permission; |
55
|
|
|
use TYPO3\CMS\Core\TypoScript\Parser\TypoScriptParser; |
56
|
|
|
use TYPO3\CMS\Core\Utility\DebugUtility; |
57
|
|
|
use TYPO3\CMS\Core\Utility\GeneralUtility; |
58
|
|
|
use TYPO3\CMS\Core\Utility\MathUtility; |
59
|
|
|
use TYPO3\CMS\Extbase\Object\ObjectManager; |
60
|
|
|
use TYPO3\CMS\Frontend\Page\CacheHashCalculator; |
61
|
|
|
use TYPO3\CMS\Frontend\Page\PageRepository; |
62
|
|
|
|
63
|
|
|
/** |
64
|
|
|
* Class CrawlerController |
65
|
|
|
* |
66
|
|
|
* @package AOE\Crawler\Controller |
67
|
|
|
*/ |
68
|
|
|
class CrawlerController implements LoggerAwareInterface |
69
|
|
|
{ |
70
|
|
|
use LoggerAwareTrait; |
71
|
|
|
|
72
|
|
|
public const CLI_STATUS_NOTHING_PROCCESSED = 0; |
73
|
|
|
public const CLI_STATUS_REMAIN = 1; //queue not empty |
74
|
|
|
public const CLI_STATUS_PROCESSED = 2; //(some) queue items where processed |
75
|
|
|
public const CLI_STATUS_ABORTED = 4; //instance didn't finish |
76
|
|
|
public const CLI_STATUS_POLLABLE_PROCESSED = 8; |
77
|
|
|
|
78
|
|
|
/** |
79
|
|
|
* @var integer |
80
|
|
|
*/ |
81
|
|
|
public $setID = 0; |
82
|
|
|
|
83
|
|
|
/** |
84
|
|
|
* @var string |
85
|
|
|
*/ |
86
|
|
|
public $processID = ''; |
87
|
|
|
|
88
|
|
|
/** |
89
|
|
|
* @var array |
90
|
|
|
*/ |
91
|
|
|
public $duplicateTrack = []; |
92
|
|
|
|
93
|
|
|
/** |
94
|
|
|
* @var array |
95
|
|
|
*/ |
96
|
|
|
public $downloadUrls = []; |
97
|
|
|
|
98
|
|
|
/** |
99
|
|
|
* @var array |
100
|
|
|
*/ |
101
|
|
|
public $incomingProcInstructions = []; |
102
|
|
|
|
103
|
|
|
/** |
104
|
|
|
* @var array |
105
|
|
|
*/ |
106
|
|
|
public $incomingConfigurationSelection = []; |
107
|
|
|
|
108
|
|
|
/** |
109
|
|
|
* @var bool |
110
|
|
|
*/ |
111
|
|
|
public $registerQueueEntriesInternallyOnly = false; |
112
|
|
|
|
113
|
|
|
/** |
114
|
|
|
* @var array |
115
|
|
|
*/ |
116
|
|
|
public $queueEntries = []; |
117
|
|
|
|
118
|
|
|
/** |
119
|
|
|
* @var array |
120
|
|
|
*/ |
121
|
|
|
public $urlList = []; |
122
|
|
|
|
123
|
|
|
/** |
124
|
|
|
* @var array |
125
|
|
|
*/ |
126
|
|
|
public $extensionSettings = []; |
127
|
|
|
|
128
|
|
|
/** |
129
|
|
|
* Mount Point |
130
|
|
|
* |
131
|
|
|
* @var bool |
132
|
|
|
* Todo: Check what this is used for and adjust the type hint or code, as bool doesn't match the current code. |
133
|
|
|
*/ |
134
|
|
|
public $MP = false; |
135
|
|
|
|
136
|
|
|
/** |
137
|
|
|
* @var string |
138
|
|
|
*/ |
139
|
|
|
protected $processFilename; |
140
|
|
|
|
141
|
|
|
/** |
142
|
|
|
* Holds the internal access mode can be 'gui','cli' or 'cli_im' |
143
|
|
|
* |
144
|
|
|
* @var string |
145
|
|
|
*/ |
146
|
|
|
protected $accessMode; |
147
|
|
|
|
148
|
|
|
/** |
149
|
|
|
* @var BackendUserAuthentication|null |
150
|
|
|
*/ |
151
|
|
|
private $backendUser; |
152
|
|
|
|
153
|
|
|
/** |
154
|
|
|
* @var integer |
155
|
|
|
*/ |
156
|
|
|
private $scheduledTime = 0; |
157
|
|
|
|
158
|
|
|
/** |
159
|
|
|
* @var integer |
160
|
|
|
*/ |
161
|
|
|
private $reqMinute = 0; |
162
|
|
|
|
163
|
|
|
/** |
164
|
|
|
* @var bool |
165
|
|
|
*/ |
166
|
|
|
private $submitCrawlUrls = false; |
167
|
|
|
|
168
|
|
|
/** |
169
|
|
|
* @var bool |
170
|
|
|
*/ |
171
|
|
|
private $downloadCrawlUrls = false; |
172
|
|
|
|
173
|
|
|
/** |
174
|
|
|
* @var QueueRepository |
175
|
|
|
*/ |
176
|
|
|
protected $queueRepository; |
177
|
|
|
|
178
|
|
|
/** |
179
|
|
|
* @var ProcessRepository |
180
|
|
|
*/ |
181
|
|
|
protected $processRepository; |
182
|
|
|
|
183
|
|
|
/** |
184
|
|
|
* @var ConfigurationRepository |
185
|
|
|
*/ |
186
|
|
|
protected $configurationRepository; |
187
|
|
|
|
188
|
|
|
/** |
189
|
|
|
* @var string |
190
|
|
|
*/ |
191
|
|
|
protected $tableName = 'tx_crawler_queue'; |
192
|
|
|
|
193
|
|
|
/** |
194
|
|
|
* @var QueueExecutor |
195
|
|
|
*/ |
196
|
|
|
protected $queueExecutor; |
197
|
|
|
|
198
|
|
|
/** |
199
|
|
|
* @var int |
200
|
|
|
*/ |
201
|
|
|
protected $maximumUrlsToCompile = 10000; |
202
|
|
|
|
203
|
|
|
/** |
204
|
|
|
* @var IconFactory |
205
|
|
|
*/ |
206
|
|
|
protected $iconFactory; |
207
|
|
|
|
208
|
|
|
/** |
209
|
|
|
* Method to set the accessMode can be gui, cli or cli_im |
210
|
|
|
* |
211
|
|
|
* @return string |
212
|
|
|
*/ |
213
|
1 |
|
public function getAccessMode() |
214
|
|
|
{ |
215
|
1 |
|
return $this->accessMode; |
216
|
|
|
} |
217
|
|
|
|
218
|
|
|
/** |
219
|
|
|
* @param string $accessMode |
220
|
|
|
*/ |
221
|
1 |
|
public function setAccessMode($accessMode): void |
222
|
|
|
{ |
223
|
1 |
|
$this->accessMode = $accessMode; |
224
|
1 |
|
} |
225
|
|
|
|
226
|
|
|
/** |
227
|
|
|
* Set disabled status to prevent processes from being processed |
228
|
|
|
* |
229
|
|
|
* @param bool $disabled (optional, defaults to true) |
230
|
|
|
* @return void |
231
|
|
|
*/ |
232
|
2 |
|
public function setDisabled($disabled = true): void |
233
|
|
|
{ |
234
|
2 |
|
if ($disabled) { |
235
|
1 |
|
GeneralUtility::writeFile($this->processFilename, ''); |
236
|
|
|
} else { |
237
|
1 |
|
if (is_file($this->processFilename)) { |
238
|
1 |
|
unlink($this->processFilename); |
239
|
|
|
} |
240
|
|
|
} |
241
|
2 |
|
} |
242
|
|
|
|
243
|
|
|
/** |
244
|
|
|
* Get disable status |
245
|
|
|
* |
246
|
|
|
* @return bool true if disabled |
247
|
|
|
*/ |
248
|
2 |
|
public function getDisabled() |
249
|
|
|
{ |
250
|
2 |
|
return is_file($this->processFilename); |
251
|
|
|
} |
252
|
|
|
|
253
|
|
|
/** |
254
|
|
|
* @param string $filenameWithPath |
255
|
|
|
* |
256
|
|
|
* @return void |
257
|
|
|
*/ |
258
|
3 |
|
public function setProcessFilename($filenameWithPath): void |
259
|
|
|
{ |
260
|
3 |
|
$this->processFilename = $filenameWithPath; |
261
|
3 |
|
} |
262
|
|
|
|
263
|
|
|
/** |
264
|
|
|
* @return string |
265
|
|
|
*/ |
266
|
1 |
|
public function getProcessFilename() |
267
|
|
|
{ |
268
|
1 |
|
return $this->processFilename; |
269
|
|
|
} |
270
|
|
|
|
271
|
|
|
/************************************ |
272
|
|
|
* |
273
|
|
|
* Getting URLs based on Page TSconfig |
274
|
|
|
* |
275
|
|
|
************************************/ |
276
|
|
|
|
277
|
48 |
|
public function __construct() |
278
|
|
|
{ |
279
|
48 |
|
$objectManager = GeneralUtility::makeInstance(ObjectManager::class); |
280
|
48 |
|
$this->queueRepository = $objectManager->get(QueueRepository::class); |
281
|
48 |
|
$this->processRepository = $objectManager->get(ProcessRepository::class); |
282
|
48 |
|
$this->configurationRepository = $objectManager->get(ConfigurationRepository::class); |
283
|
48 |
|
$this->queueExecutor = $objectManager->get(QueueExecutor::class); |
284
|
48 |
|
$this->iconFactory = GeneralUtility::makeInstance(IconFactory::class); |
285
|
|
|
|
286
|
48 |
|
$this->processFilename = Environment::getVarPath() . '/lock/tx_crawler.proc'; |
287
|
|
|
|
288
|
|
|
/** @var ExtensionConfigurationProvider $configurationProvider */ |
289
|
48 |
|
$configurationProvider = GeneralUtility::makeInstance(ExtensionConfigurationProvider::class); |
290
|
48 |
|
$settings = $configurationProvider->getExtensionConfiguration(); |
291
|
48 |
|
$this->extensionSettings = is_array($settings) ? $settings : []; |
|
|
|
|
292
|
|
|
|
293
|
|
|
// set defaults: |
294
|
48 |
|
if (MathUtility::convertToPositiveInteger($this->extensionSettings['countInARun']) == 0) { |
295
|
|
|
$this->extensionSettings['countInARun'] = 100; |
296
|
|
|
} |
297
|
|
|
|
298
|
48 |
|
$this->extensionSettings['processLimit'] = MathUtility::forceIntegerInRange($this->extensionSettings['processLimit'], 1, 99, 1); |
299
|
48 |
|
$this->maximumUrlsToCompile = MathUtility::forceIntegerInRange($this->extensionSettings['maxCompileUrls'], 1, 1000000000, 10000); |
300
|
48 |
|
} |
301
|
|
|
|
302
|
|
|
/** |
303
|
|
|
* @return BackendUserAuthentication |
304
|
|
|
*/ |
305
|
1 |
|
private function getBackendUser() |
306
|
|
|
{ |
307
|
|
|
// Make sure the _cli_ user is loaded |
308
|
1 |
|
Bootstrap::initializeBackendAuthentication(); |
309
|
1 |
|
if ($this->backendUser === null) { |
310
|
1 |
|
$this->backendUser = $GLOBALS['BE_USER']; |
311
|
|
|
} |
312
|
1 |
|
return $this->backendUser; |
313
|
|
|
} |
314
|
|
|
|
315
|
|
|
/** |
316
|
|
|
* Sets the extensions settings (unserialized pendant of $TYPO3_CONF_VARS['EXT']['extConf']['crawler']). |
317
|
|
|
* |
318
|
|
|
* @param array $extensionSettings |
319
|
|
|
* @return void |
320
|
|
|
*/ |
321
|
12 |
|
public function setExtensionSettings(array $extensionSettings): void |
322
|
|
|
{ |
323
|
12 |
|
$this->extensionSettings = $extensionSettings; |
324
|
12 |
|
} |
325
|
|
|
|
326
|
|
|
/** |
327
|
|
|
* Check if the given page should be crawled |
328
|
|
|
* |
329
|
|
|
* @param array $pageRow |
330
|
|
|
* @return false|string false if the page should be crawled (not excluded), true / skipMessage if it should be skipped |
331
|
|
|
*/ |
332
|
8 |
|
public function checkIfPageShouldBeSkipped(array $pageRow) |
333
|
|
|
{ |
334
|
8 |
|
$skipPage = false; |
335
|
8 |
|
$skipMessage = 'Skipped'; // message will be overwritten later |
336
|
|
|
|
337
|
|
|
// if page is hidden |
338
|
8 |
|
if (!$this->extensionSettings['crawlHiddenPages']) { |
339
|
8 |
|
if ($pageRow['hidden']) { |
340
|
1 |
|
$skipPage = true; |
341
|
1 |
|
$skipMessage = 'Because page is hidden'; |
342
|
|
|
} |
343
|
|
|
} |
344
|
|
|
|
345
|
8 |
|
if (!$skipPage) { |
346
|
7 |
|
if (GeneralUtility::inList('3,4', $pageRow['doktype']) || $pageRow['doktype'] >= 199) { |
347
|
3 |
|
$skipPage = true; |
348
|
3 |
|
$skipMessage = 'Because doktype is not allowed'; |
349
|
|
|
} |
350
|
|
|
} |
351
|
|
|
|
352
|
8 |
|
if (!$skipPage) { |
353
|
4 |
|
foreach ($GLOBALS['TYPO3_CONF_VARS']['EXTCONF']['crawler']['excludeDoktype'] ?? [] as $key => $doktypeList) { |
354
|
1 |
|
if (GeneralUtility::inList($doktypeList, $pageRow['doktype'])) { |
355
|
1 |
|
$skipPage = true; |
356
|
1 |
|
$skipMessage = 'Doktype was excluded by "' . $key . '"'; |
357
|
1 |
|
break; |
358
|
|
|
} |
359
|
|
|
} |
360
|
|
|
} |
361
|
|
|
|
362
|
8 |
|
if (!$skipPage) { |
363
|
|
|
// veto hook |
364
|
3 |
|
foreach ($GLOBALS['TYPO3_CONF_VARS']['EXTCONF']['crawler']['pageVeto'] ?? [] as $key => $func) { |
365
|
|
|
$params = [ |
366
|
|
|
'pageRow' => $pageRow, |
367
|
|
|
]; |
368
|
|
|
// expects "false" if page is ok and "true" or a skipMessage if this page should _not_ be crawled |
369
|
|
|
$veto = GeneralUtility::callUserFunction($func, $params, $this); |
370
|
|
|
if ($veto !== false) { |
371
|
|
|
$skipPage = true; |
372
|
|
|
if (is_string($veto)) { |
373
|
|
|
$skipMessage = $veto; |
374
|
|
|
} else { |
375
|
|
|
$skipMessage = 'Veto from hook "' . htmlspecialchars($key) . '"'; |
376
|
|
|
} |
377
|
|
|
// no need to execute other hooks if a previous one return a veto |
378
|
|
|
break; |
379
|
|
|
} |
380
|
|
|
} |
381
|
|
|
} |
382
|
|
|
|
383
|
8 |
|
return $skipPage ? $skipMessage : false; |
384
|
|
|
} |
385
|
|
|
|
386
|
|
|
/** |
387
|
|
|
* Wrapper method for getUrlsForPageId() |
388
|
|
|
* It returns an array of configurations and no urls! |
389
|
|
|
* |
390
|
|
|
* @param array $pageRow Page record with at least dok-type and uid columns. |
391
|
|
|
* @param string $skipMessage |
392
|
|
|
* @return array |
393
|
|
|
* @see getUrlsForPageId() |
394
|
|
|
*/ |
395
|
4 |
|
public function getUrlsForPageRow(array $pageRow, &$skipMessage = '') |
396
|
|
|
{ |
397
|
4 |
|
$message = $this->checkIfPageShouldBeSkipped($pageRow); |
398
|
|
|
|
399
|
4 |
|
if ($message === false) { |
400
|
3 |
|
$res = $this->getUrlsForPageId($pageRow['uid']); |
401
|
3 |
|
$skipMessage = ''; |
402
|
|
|
} else { |
403
|
1 |
|
$skipMessage = $message; |
404
|
1 |
|
$res = []; |
405
|
|
|
} |
406
|
|
|
|
407
|
4 |
|
return $res; |
408
|
|
|
} |
409
|
|
|
|
410
|
|
|
/** |
411
|
|
|
* This method is used to count if there are ANY unprocessed queue entries |
412
|
|
|
* of a given page_id and the configuration which matches a given hash. |
413
|
|
|
* If there if none, we can skip an inner detail check |
414
|
|
|
* |
415
|
|
|
* @param int $uid |
416
|
|
|
* @param string $configurationHash |
417
|
|
|
* @return boolean |
418
|
|
|
*/ |
419
|
5 |
|
protected function noUnprocessedQueueEntriesForPageWithConfigurationHashExist($uid, $configurationHash) |
420
|
|
|
{ |
421
|
5 |
|
$queryBuilder = GeneralUtility::makeInstance(ConnectionPool::class)->getQueryBuilderForTable($this->tableName); |
422
|
5 |
|
$noUnprocessedQueueEntriesFound = true; |
423
|
|
|
|
424
|
|
|
$result = $queryBuilder |
425
|
5 |
|
->count('*') |
426
|
5 |
|
->from($this->tableName) |
427
|
5 |
|
->where( |
428
|
5 |
|
$queryBuilder->expr()->eq('page_id', (int)$uid), |
429
|
5 |
|
$queryBuilder->expr()->eq('configuration_hash', $queryBuilder->createNamedParameter($configurationHash)), |
430
|
5 |
|
$queryBuilder->expr()->eq('exec_time', 0) |
431
|
|
|
) |
432
|
5 |
|
->execute() |
433
|
5 |
|
->fetchColumn(); |
434
|
|
|
|
435
|
5 |
|
if ($result) { |
436
|
3 |
|
$noUnprocessedQueueEntriesFound = false; |
437
|
|
|
} |
438
|
|
|
|
439
|
5 |
|
return $noUnprocessedQueueEntriesFound; |
440
|
|
|
} |
441
|
|
|
|
442
|
|
|
/** |
443
|
|
|
* Creates a list of URLs from input array (and submits them to queue if asked for) |
444
|
|
|
* See Web > Info module script + "indexed_search"'s crawler hook-client using this! |
445
|
|
|
* |
446
|
|
|
* @param array $vv Information about URLs from pageRow to crawl. |
447
|
|
|
* @param array $pageRow Page row |
448
|
|
|
* @param int $scheduledTime Unix time to schedule indexing to, typically time() |
449
|
|
|
* @param int $reqMinute Number of requests per minute (creates the interleave between requests) |
450
|
|
|
* @param bool $submitCrawlUrls If set, submits the URLs to queue |
451
|
|
|
* @param bool $downloadCrawlUrls If set (and submitcrawlUrls is false) will fill $downloadUrls with entries) |
452
|
|
|
* @param array $duplicateTrack Array which is passed by reference and contains the an id per url to secure we will not crawl duplicates |
453
|
|
|
* @param array $downloadUrls Array which will be filled with URLS for download if flag is set. |
454
|
|
|
* @param array $incomingProcInstructions Array of processing instructions |
455
|
|
|
* @return string List of URLs (meant for display in backend module) |
456
|
|
|
*/ |
457
|
2 |
|
public function urlListFromUrlArray( |
458
|
|
|
array $vv, |
459
|
|
|
array $pageRow, |
460
|
|
|
$scheduledTime, |
461
|
|
|
$reqMinute, |
462
|
|
|
$submitCrawlUrls, |
463
|
|
|
$downloadCrawlUrls, |
464
|
|
|
array &$duplicateTrack, |
465
|
|
|
array &$downloadUrls, |
466
|
|
|
array $incomingProcInstructions |
467
|
|
|
) { |
468
|
2 |
|
if (!is_array($vv['URLs'])) { |
469
|
|
|
return 'ERROR - no URL generated'; |
470
|
|
|
} |
471
|
2 |
|
$urlLog = []; |
472
|
2 |
|
$pageId = (int)$pageRow['uid']; |
473
|
2 |
|
$configurationHash = $this->getConfigurationHash($vv); |
474
|
2 |
|
$skipInnerCheck = $this->noUnprocessedQueueEntriesForPageWithConfigurationHashExist($pageId, $configurationHash); |
475
|
|
|
|
476
|
2 |
|
foreach ($vv['URLs'] as $urlQuery) { |
477
|
2 |
|
if (!$this->drawURLs_PIfilter($vv['subCfg']['procInstrFilter'], $incomingProcInstructions)) { |
478
|
|
|
continue; |
479
|
|
|
} |
480
|
2 |
|
$url = (string)$this->getUrlFromPageAndQueryParameters( |
481
|
2 |
|
$pageId, |
482
|
2 |
|
$urlQuery, |
483
|
2 |
|
$vv['subCfg']['baseUrl'] ?? null, |
484
|
2 |
|
$vv['subCfg']['force_ssl'] ?? 0 |
485
|
|
|
); |
486
|
|
|
|
487
|
|
|
// Create key by which to determine unique-ness: |
488
|
2 |
|
$uKey = $url . '|' . $vv['subCfg']['userGroups'] . '|' . $vv['subCfg']['procInstrFilter']; |
489
|
|
|
|
490
|
2 |
|
if (isset($duplicateTrack[$uKey])) { |
491
|
|
|
//if the url key is registered just display it and do not resubmit is |
492
|
|
|
$urlLog[] = '<em><span class="text-muted">' . htmlspecialchars($url) . '</span></em>'; |
493
|
|
|
} else { |
494
|
|
|
// Scheduled time: |
495
|
2 |
|
$schTime = $scheduledTime + round(count($duplicateTrack) * (60 / $reqMinute)); |
496
|
2 |
|
$schTime = intval($schTime / 60) * 60; |
497
|
2 |
|
$formattedDate = BackendUtility::datetime($schTime); |
498
|
2 |
|
$this->urlList[] = '[' . $formattedDate . '] ' . $url; |
499
|
2 |
|
$urlList = '[' . $formattedDate . '] ' . htmlspecialchars($url); |
500
|
|
|
|
501
|
|
|
// Submit for crawling! |
502
|
2 |
|
if ($submitCrawlUrls) { |
503
|
2 |
|
$added = $this->addUrl( |
504
|
2 |
|
$pageId, |
505
|
2 |
|
$url, |
506
|
2 |
|
$vv['subCfg'], |
507
|
2 |
|
$scheduledTime, |
508
|
2 |
|
$configurationHash, |
509
|
2 |
|
$skipInnerCheck |
510
|
|
|
); |
511
|
2 |
|
if ($added === false) { |
512
|
2 |
|
$urlList .= ' (URL already existed)'; |
513
|
|
|
} |
514
|
|
|
} elseif ($downloadCrawlUrls) { |
515
|
|
|
$downloadUrls[$url] = $url; |
516
|
|
|
} |
517
|
2 |
|
$urlLog[] = $urlList; |
518
|
|
|
} |
519
|
2 |
|
$duplicateTrack[$uKey] = true; |
520
|
|
|
} |
521
|
|
|
|
522
|
2 |
|
return implode('<br>', $urlLog); |
523
|
|
|
} |
524
|
|
|
|
525
|
|
|
/** |
526
|
|
|
* Returns true if input processing instruction is among registered ones. |
527
|
|
|
* |
528
|
|
|
* @param string $piString PI to test |
529
|
|
|
* @param array $incomingProcInstructions Processing instructions |
530
|
|
|
* @return boolean |
531
|
|
|
*/ |
532
|
5 |
|
public function drawURLs_PIfilter($piString, array $incomingProcInstructions) |
533
|
|
|
{ |
534
|
5 |
|
if (empty($incomingProcInstructions)) { |
535
|
1 |
|
return true; |
536
|
|
|
} |
537
|
|
|
|
538
|
4 |
|
foreach ($incomingProcInstructions as $pi) { |
539
|
4 |
|
if (GeneralUtility::inList($piString, $pi)) { |
540
|
2 |
|
return true; |
541
|
|
|
} |
542
|
|
|
} |
543
|
2 |
|
return false; |
544
|
|
|
} |
545
|
|
|
|
546
|
3 |
|
public function getPageTSconfigForId($id): array |
547
|
|
|
{ |
548
|
3 |
|
if (!$this->MP) { |
549
|
3 |
|
$pageTSconfig = BackendUtility::getPagesTSconfig($id); |
|
|
|
|
550
|
|
|
} else { |
551
|
|
|
// TODO: Please check, this makes no sense to split a boolean value. |
552
|
|
|
[, $mountPointId] = explode('-', $this->MP); |
|
|
|
|
553
|
|
|
$pageTSconfig = BackendUtility::getPagesTSconfig($mountPointId); |
|
|
|
|
554
|
|
|
} |
555
|
|
|
|
556
|
|
|
// Call a hook to alter configuration |
557
|
3 |
|
if (is_array($GLOBALS['TYPO3_CONF_VARS']['EXTCONF']['crawler']['getPageTSconfigForId'])) { |
558
|
|
|
$params = [ |
559
|
|
|
'pageId' => $id, |
560
|
|
|
'pageTSConfig' => &$pageTSconfig, |
561
|
|
|
]; |
562
|
|
|
foreach ($GLOBALS['TYPO3_CONF_VARS']['EXTCONF']['crawler']['getPageTSconfigForId'] as $userFunc) { |
563
|
|
|
GeneralUtility::callUserFunction($userFunc, $params, $this); |
564
|
|
|
} |
565
|
|
|
} |
566
|
3 |
|
return $pageTSconfig; |
567
|
|
|
} |
568
|
|
|
|
569
|
|
|
/** |
570
|
|
|
* This methods returns an array of configurations. |
571
|
|
|
* Adds no urls! |
572
|
|
|
*/ |
573
|
2 |
|
public function getUrlsForPageId(int $pageId): array |
574
|
|
|
{ |
575
|
|
|
// Get page TSconfig for page ID |
576
|
2 |
|
$pageTSconfig = $this->getPageTSconfigForId($pageId); |
577
|
|
|
|
578
|
2 |
|
$res = []; |
579
|
|
|
|
580
|
|
|
// Fetch Crawler Configuration from pageTSconfig |
581
|
2 |
|
$crawlerCfg = $pageTSconfig['tx_crawler.']['crawlerCfg.']['paramSets.'] ?? []; |
582
|
2 |
|
foreach ($crawlerCfg as $key => $values) { |
583
|
1 |
|
if (!is_array($values)) { |
584
|
1 |
|
continue; |
585
|
|
|
} |
586
|
1 |
|
$key = str_replace('.', '', $key); |
587
|
|
|
// Sub configuration for a single configuration string: |
588
|
1 |
|
$subCfg = (array)$crawlerCfg[$key . '.']; |
589
|
1 |
|
$subCfg['key'] = $key; |
590
|
|
|
|
591
|
1 |
|
if (strcmp($subCfg['procInstrFilter'], '')) { |
592
|
1 |
|
$subCfg['procInstrFilter'] = implode(',', GeneralUtility::trimExplode(',', $subCfg['procInstrFilter'])); |
593
|
|
|
} |
594
|
1 |
|
$pidOnlyList = implode(',', GeneralUtility::trimExplode(',', $subCfg['pidsOnly'], true)); |
595
|
|
|
|
596
|
|
|
// process configuration if it is not page-specific or if the specific page is the current page: |
597
|
|
|
// TODO: Check if $pidOnlyList can be kept as Array instead of imploded |
598
|
1 |
|
if (!strcmp((string)$subCfg['pidsOnly'], '') || GeneralUtility::inList($pidOnlyList, strval($pageId))) { |
599
|
|
|
|
600
|
|
|
// Explode, process etc.: |
601
|
1 |
|
$res[$key] = []; |
602
|
1 |
|
$res[$key]['subCfg'] = $subCfg; |
603
|
1 |
|
$res[$key]['paramParsed'] = GeneralUtility::explodeUrl2Array($crawlerCfg[$key]); |
604
|
1 |
|
$res[$key]['paramExpanded'] = $this->expandParameters($res[$key]['paramParsed'], $pageId); |
605
|
1 |
|
$res[$key]['origin'] = 'pagets'; |
606
|
|
|
|
607
|
|
|
// recognize MP value |
608
|
1 |
|
if (!$this->MP) { |
609
|
1 |
|
$res[$key]['URLs'] = $this->compileUrls($res[$key]['paramExpanded'], ['?id=' . $pageId]); |
610
|
|
|
} else { |
611
|
|
|
$res[$key]['URLs'] = $this->compileUrls($res[$key]['paramExpanded'], ['?id=' . $pageId . '&MP=' . $this->MP]); |
|
|
|
|
612
|
|
|
} |
613
|
|
|
} |
614
|
|
|
} |
615
|
|
|
|
616
|
|
|
// Get configuration from tx_crawler_configuration records up the rootline |
617
|
2 |
|
$crawlerConfigurations = $this->configurationRepository->getCrawlerConfigurationRecordsFromRootLine($pageId); |
618
|
2 |
|
foreach ($crawlerConfigurations as $configurationRecord) { |
619
|
|
|
|
620
|
|
|
// check access to the configuration record |
621
|
1 |
|
if (empty($configurationRecord['begroups']) || $this->getBackendUser()->isAdmin() || $this->hasGroupAccess($this->getBackendUser()->user['usergroup_cached_list'], $configurationRecord['begroups'])) { |
622
|
1 |
|
$pidOnlyList = implode(',', GeneralUtility::trimExplode(',', $configurationRecord['pidsonly'], true)); |
623
|
|
|
|
624
|
|
|
// process configuration if it is not page-specific or if the specific page is the current page: |
625
|
|
|
// TODO: Check if $pidOnlyList can be kept as Array instead of imploded |
626
|
1 |
|
if (!strcmp($configurationRecord['pidsonly'], '') || GeneralUtility::inList($pidOnlyList, strval($pageId))) { |
627
|
1 |
|
$key = $configurationRecord['name']; |
628
|
|
|
|
629
|
|
|
// don't overwrite previously defined paramSets |
630
|
1 |
|
if (!isset($res[$key])) { |
631
|
|
|
|
632
|
|
|
/* @var $TSparserObject \TYPO3\CMS\Core\TypoScript\Parser\TypoScriptParser */ |
633
|
1 |
|
$TSparserObject = GeneralUtility::makeInstance(TypoScriptParser::class); |
634
|
1 |
|
$TSparserObject->parse($configurationRecord['processing_instruction_parameters_ts']); |
635
|
|
|
|
636
|
|
|
$subCfg = [ |
637
|
1 |
|
'procInstrFilter' => $configurationRecord['processing_instruction_filter'], |
638
|
1 |
|
'procInstrParams.' => $TSparserObject->setup, |
639
|
1 |
|
'baseUrl' => $configurationRecord['base_url'], |
640
|
1 |
|
'force_ssl' => (int)$configurationRecord['force_ssl'], |
641
|
1 |
|
'userGroups' => $configurationRecord['fegroups'], |
642
|
1 |
|
'exclude' => $configurationRecord['exclude'], |
643
|
1 |
|
'key' => $key, |
644
|
|
|
]; |
645
|
|
|
|
646
|
1 |
|
if (!in_array($pageId, $this->expandExcludeString($subCfg['exclude']))) { |
647
|
1 |
|
$res[$key] = []; |
648
|
1 |
|
$res[$key]['subCfg'] = $subCfg; |
649
|
1 |
|
$res[$key]['paramParsed'] = GeneralUtility::explodeUrl2Array($configurationRecord['configuration']); |
650
|
1 |
|
$res[$key]['paramExpanded'] = $this->expandParameters($res[$key]['paramParsed'], $pageId); |
651
|
1 |
|
$res[$key]['URLs'] = $this->compileUrls($res[$key]['paramExpanded'], ['?id=' . $pageId]); |
652
|
1 |
|
$res[$key]['origin'] = 'tx_crawler_configuration_' . $configurationRecord['uid']; |
653
|
|
|
} |
654
|
|
|
} |
655
|
|
|
} |
656
|
|
|
} |
657
|
|
|
} |
658
|
|
|
|
659
|
2 |
|
foreach ($GLOBALS['TYPO3_CONF_VARS']['EXTCONF']['crawler']['processUrls'] ?? [] as $func) { |
660
|
|
|
$params = [ |
661
|
|
|
'res' => &$res, |
662
|
|
|
]; |
663
|
|
|
GeneralUtility::callUserFunction($func, $params, $this); |
664
|
|
|
} |
665
|
2 |
|
return $res; |
666
|
|
|
} |
667
|
|
|
|
668
|
|
|
/** |
669
|
|
|
* Find all configurations of subpages of a page |
670
|
|
|
* TODO: Write Functional Tests |
671
|
|
|
*/ |
672
|
1 |
|
public function getConfigurationsForBranch(int $rootid, int $depth): array |
673
|
|
|
{ |
674
|
1 |
|
$configurationsForBranch = []; |
675
|
1 |
|
$pageTSconfig = $this->getPageTSconfigForId($rootid); |
676
|
1 |
|
$sets = $pageTSconfig['tx_crawler.']['crawlerCfg.']['paramSets.'] ?? []; |
677
|
1 |
|
foreach ($sets as $key => $value) { |
678
|
|
|
if (!is_array($value)) { |
679
|
|
|
continue; |
680
|
|
|
} |
681
|
|
|
$configurationsForBranch[] = substr($key, -1) == '.' ? substr($key, 0, -1) : $key; |
682
|
|
|
} |
683
|
1 |
|
$pids = []; |
684
|
1 |
|
$rootLine = BackendUtility::BEgetRootLine($rootid); |
685
|
1 |
|
foreach ($rootLine as $node) { |
686
|
1 |
|
$pids[] = $node['uid']; |
687
|
|
|
} |
688
|
|
|
/* @var PageTreeView $tree */ |
689
|
1 |
|
$tree = GeneralUtility::makeInstance(PageTreeView::class); |
690
|
1 |
|
$perms_clause = $this->getBackendUser()->getPagePermsClause(Permission::PAGE_SHOW); |
691
|
1 |
|
$tree->init(empty($perms_clause) ? '' : ('AND ' . $perms_clause)); |
692
|
1 |
|
$tree->getTree($rootid, $depth, ''); |
693
|
1 |
|
foreach ($tree->tree as $node) { |
694
|
|
|
$pids[] = $node['row']['uid']; |
695
|
|
|
} |
696
|
|
|
|
697
|
1 |
|
$queryBuilder = $this->getQueryBuilder('tx_crawler_configuration'); |
698
|
|
|
$statement = $queryBuilder |
699
|
1 |
|
->select('name') |
700
|
1 |
|
->from('tx_crawler_configuration') |
701
|
1 |
|
->where( |
702
|
1 |
|
$queryBuilder->expr()->in('pid', $queryBuilder->createNamedParameter($pids, Connection::PARAM_INT_ARRAY)) |
703
|
|
|
) |
704
|
1 |
|
->execute(); |
705
|
|
|
|
706
|
1 |
|
while ($row = $statement->fetch()) { |
707
|
1 |
|
$configurationsForBranch[] = $row['name']; |
708
|
|
|
} |
709
|
1 |
|
return $configurationsForBranch; |
710
|
|
|
} |
711
|
|
|
|
712
|
|
|
/** |
713
|
|
|
* Get querybuilder for given table |
714
|
|
|
* |
715
|
|
|
* @param string $table |
716
|
|
|
* @return \TYPO3\CMS\Core\Database\Query\QueryBuilder |
717
|
|
|
*/ |
718
|
18 |
|
private function getQueryBuilder(string $table) |
719
|
|
|
{ |
720
|
18 |
|
return GeneralUtility::makeInstance(ConnectionPool::class)->getQueryBuilderForTable($table); |
721
|
|
|
} |
722
|
|
|
|
723
|
|
|
/** |
724
|
|
|
* Check if a user has access to an item |
725
|
|
|
* (e.g. get the group list of the current logged in user from $GLOBALS['TSFE']->gr_list) |
726
|
|
|
* |
727
|
|
|
* @param string $groupList Comma-separated list of (fe_)group UIDs from a user |
728
|
|
|
* @param string $accessList Comma-separated list of (fe_)group UIDs of the item to access |
729
|
|
|
* @return bool TRUE if at least one of the users group UIDs is in the access list or the access list is empty |
730
|
|
|
* @see \TYPO3\CMS\Frontend\Page\PageRepository::getMultipleGroupsWhereClause() |
731
|
|
|
*/ |
732
|
3 |
|
public function hasGroupAccess($groupList, $accessList) |
733
|
|
|
{ |
734
|
3 |
|
if (empty($accessList)) { |
735
|
1 |
|
return true; |
736
|
|
|
} |
737
|
2 |
|
foreach (GeneralUtility::intExplode(',', $groupList) as $groupUid) { |
738
|
2 |
|
if (GeneralUtility::inList($accessList, $groupUid)) { |
739
|
1 |
|
return true; |
740
|
|
|
} |
741
|
|
|
} |
742
|
1 |
|
return false; |
743
|
|
|
} |
744
|
|
|
|
745
|
|
|
/** |
746
|
|
|
* Will expand the parameters configuration to individual values. This follows a certain syntax of the value of each parameter. |
747
|
|
|
* Syntax of values: |
748
|
|
|
* - Basically: If the value is wrapped in [...] it will be expanded according to the following syntax, otherwise the value is taken literally |
749
|
|
|
* - Configuration is splitted by "|" and the parts are processed individually and finally added together |
750
|
|
|
* - For each configuration part: |
751
|
|
|
* - "[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" |
752
|
|
|
* - "_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" |
753
|
|
|
* _ENABLELANG:1 picks only original records without their language overlays |
754
|
|
|
* - Default: Literal value |
755
|
|
|
* |
756
|
|
|
* @param array $paramArray Array with key (GET var name) and values (value of GET var which is configuration for expansion) |
757
|
|
|
* @param integer $pid Current page ID |
758
|
|
|
* @return array |
759
|
|
|
* |
760
|
|
|
* TODO: Write Functional Tests |
761
|
|
|
*/ |
762
|
9 |
|
public function expandParameters($paramArray, $pid) |
763
|
|
|
{ |
764
|
|
|
// Traverse parameter names: |
765
|
9 |
|
foreach ($paramArray as $p => $v) { |
766
|
9 |
|
$v = trim($v); |
767
|
|
|
|
768
|
|
|
// If value is encapsulated in square brackets it means there are some ranges of values to find, otherwise the value is literal |
769
|
9 |
|
if (strpos($v, '[') === 0 && substr($v, -1) === ']') { |
770
|
|
|
// So, find the value inside brackets and reset the paramArray value as an array. |
771
|
9 |
|
$v = substr($v, 1, -1); |
772
|
9 |
|
$paramArray[$p] = []; |
773
|
|
|
|
774
|
|
|
// Explode parts and traverse them: |
775
|
9 |
|
$parts = explode('|', $v); |
776
|
9 |
|
foreach ($parts as $pV) { |
777
|
|
|
|
778
|
|
|
// Look for integer range: (fx. 1-34 or -40--30 // reads minus 40 to minus 30) |
779
|
9 |
|
if (preg_match('/^(-?[0-9]+)\s*-\s*(-?[0-9]+)$/', trim($pV), $reg)) { |
780
|
1 |
|
$reg = $this->swapIfFirstIsLargerThanSecond($reg); |
781
|
|
|
|
782
|
|
|
// Traverse range, add values: |
783
|
1 |
|
$runAwayBrake = 1000; // Limit to size of range! |
784
|
1 |
|
for ($a = $reg[1]; $a <= $reg[2]; $a++) { |
785
|
1 |
|
$paramArray[$p][] = $a; |
786
|
1 |
|
$runAwayBrake--; |
787
|
1 |
|
if ($runAwayBrake <= 0) { |
788
|
|
|
break; |
789
|
|
|
} |
790
|
|
|
} |
791
|
8 |
|
} elseif (strpos(trim($pV), '_TABLE:') === 0) { |
792
|
|
|
|
793
|
|
|
// Parse parameters: |
794
|
6 |
|
$subparts = GeneralUtility::trimExplode(';', $pV); |
795
|
6 |
|
$subpartParams = []; |
796
|
6 |
|
foreach ($subparts as $spV) { |
797
|
6 |
|
[$pKey, $pVal] = GeneralUtility::trimExplode(':', $spV); |
798
|
6 |
|
$subpartParams[$pKey] = $pVal; |
799
|
|
|
} |
800
|
|
|
|
801
|
|
|
// Table exists: |
802
|
6 |
|
if (isset($GLOBALS['TCA'][$subpartParams['_TABLE']])) { |
803
|
6 |
|
$lookUpPid = isset($subpartParams['_PID']) ? intval($subpartParams['_PID']) : intval($pid); |
804
|
6 |
|
$recursiveDepth = isset($subpartParams['_RECURSIVE']) ? intval($subpartParams['_RECURSIVE']) : 0; |
805
|
6 |
|
$pidField = isset($subpartParams['_PIDFIELD']) ? trim($subpartParams['_PIDFIELD']) : 'pid'; |
806
|
6 |
|
$where = $subpartParams['_WHERE'] ?? ''; |
807
|
6 |
|
$addTable = $subpartParams['_ADDTABLE'] ?? ''; |
808
|
|
|
|
809
|
6 |
|
$fieldName = $subpartParams['_FIELD'] ? $subpartParams['_FIELD'] : 'uid'; |
810
|
6 |
|
if ($fieldName === 'uid' || $GLOBALS['TCA'][$subpartParams['_TABLE']]['columns'][$fieldName]) { |
811
|
6 |
|
$queryBuilder = $this->getQueryBuilder($subpartParams['_TABLE']); |
812
|
|
|
|
813
|
6 |
|
if ($recursiveDepth > 0) { |
814
|
|
|
/** @var \TYPO3\CMS\Core\Database\QueryGenerator $queryGenerator */ |
815
|
2 |
|
$queryGenerator = GeneralUtility::makeInstance(\TYPO3\CMS\Core\Database\QueryGenerator::class); |
816
|
2 |
|
$pidList = $queryGenerator->getTreeList($lookUpPid, $recursiveDepth, 0, 1); |
817
|
2 |
|
$pidArray = GeneralUtility::intExplode(',', $pidList); |
818
|
|
|
} else { |
819
|
4 |
|
$pidArray = [(string)$lookUpPid]; |
820
|
|
|
} |
821
|
|
|
|
822
|
6 |
|
$queryBuilder->getRestrictions() |
823
|
6 |
|
->removeAll() |
824
|
6 |
|
->add(GeneralUtility::makeInstance(DeletedRestriction::class)); |
825
|
|
|
|
826
|
|
|
$queryBuilder |
827
|
6 |
|
->select($fieldName) |
828
|
6 |
|
->from($subpartParams['_TABLE']) |
829
|
6 |
|
->where( |
830
|
6 |
|
$queryBuilder->expr()->in($pidField, $queryBuilder->createNamedParameter($pidArray, Connection::PARAM_INT_ARRAY)), |
831
|
6 |
|
$where |
832
|
|
|
); |
833
|
|
|
|
834
|
6 |
|
if (!empty($addTable)) { |
835
|
|
|
// TODO: Check if this works as intended! |
836
|
|
|
$queryBuilder->add('from', $addTable); |
837
|
|
|
} |
838
|
6 |
|
$transOrigPointerField = $GLOBALS['TCA'][$subpartParams['_TABLE']]['ctrl']['transOrigPointerField']; |
839
|
|
|
|
840
|
6 |
|
if ($subpartParams['_ENABLELANG'] && $transOrigPointerField) { |
841
|
|
|
$queryBuilder->andWhere( |
842
|
|
|
$queryBuilder->expr()->lte( |
843
|
|
|
$transOrigPointerField, |
844
|
|
|
0 |
845
|
|
|
) |
846
|
|
|
); |
847
|
|
|
} |
848
|
|
|
|
849
|
6 |
|
$statement = $queryBuilder->execute(); |
850
|
|
|
|
851
|
6 |
|
$rows = []; |
852
|
6 |
|
while ($row = $statement->fetch()) { |
853
|
6 |
|
$rows[$row[$fieldName]] = $row; |
854
|
|
|
} |
855
|
|
|
|
856
|
6 |
|
if (is_array($rows)) { |
857
|
6 |
|
$paramArray[$p] = array_merge($paramArray[$p], array_keys($rows)); |
858
|
|
|
} |
859
|
|
|
} |
860
|
|
|
} |
861
|
|
|
} else { // Just add value: |
862
|
2 |
|
$paramArray[$p][] = $pV; |
863
|
|
|
} |
864
|
|
|
// Hook for processing own expandParameters place holder |
865
|
9 |
|
if (is_array($GLOBALS['TYPO3_CONF_VARS']['SC_OPTIONS']['crawler/class.tx_crawler_lib.php']['expandParameters'])) { |
866
|
|
|
$_params = [ |
867
|
|
|
'pObj' => &$this, |
868
|
|
|
'paramArray' => &$paramArray, |
869
|
|
|
'currentKey' => $p, |
870
|
|
|
'currentValue' => $pV, |
871
|
|
|
'pid' => $pid, |
872
|
|
|
]; |
873
|
|
|
foreach ($GLOBALS['TYPO3_CONF_VARS']['SC_OPTIONS']['crawler/class.tx_crawler_lib.php']['expandParameters'] as $_funcRef) { |
874
|
|
|
GeneralUtility::callUserFunction($_funcRef, $_params, $this); |
875
|
|
|
} |
876
|
|
|
} |
877
|
|
|
} |
878
|
|
|
|
879
|
|
|
// Make unique set of values and sort array by key: |
880
|
9 |
|
$paramArray[$p] = array_unique($paramArray[$p]); |
881
|
9 |
|
ksort($paramArray); |
882
|
|
|
} else { |
883
|
|
|
// Set the literal value as only value in array: |
884
|
2 |
|
$paramArray[$p] = [$v]; |
885
|
|
|
} |
886
|
|
|
} |
887
|
|
|
|
888
|
9 |
|
return $paramArray; |
889
|
|
|
} |
890
|
|
|
|
891
|
|
|
/** |
892
|
|
|
* Compiling URLs from parameter array (output of expandParameters()) |
893
|
|
|
* The number of URLs will be the multiplication of the number of parameter values for each key |
894
|
|
|
* |
895
|
|
|
* @param array $paramArray Output of expandParameters(): Array with keys (GET var names) and for each an array of values |
896
|
|
|
* @param array $urls URLs accumulated in this array (for recursion) |
897
|
|
|
* @return array |
898
|
|
|
*/ |
899
|
5 |
|
public function compileUrls($paramArray, array $urls) |
900
|
|
|
{ |
901
|
5 |
|
if (empty($paramArray)) { |
902
|
5 |
|
return $urls; |
903
|
|
|
} |
904
|
|
|
// shift first off stack: |
905
|
4 |
|
reset($paramArray); |
906
|
4 |
|
$varName = key($paramArray); |
907
|
4 |
|
$valueSet = array_shift($paramArray); |
908
|
|
|
|
909
|
|
|
// Traverse value set: |
910
|
4 |
|
$newUrls = []; |
911
|
4 |
|
foreach ($urls as $url) { |
912
|
3 |
|
foreach ($valueSet as $val) { |
913
|
3 |
|
$newUrls[] = $url . (strcmp((string)$val, '') ? '&' . rawurlencode($varName) . '=' . rawurlencode((string)$val) : ''); |
914
|
|
|
|
915
|
3 |
|
if (count($newUrls) > $this->maximumUrlsToCompile) { |
916
|
|
|
break; |
917
|
|
|
} |
918
|
|
|
} |
919
|
|
|
} |
920
|
4 |
|
return $this->compileUrls($paramArray, $newUrls); |
921
|
|
|
} |
922
|
|
|
|
923
|
|
|
/************************************ |
924
|
|
|
* |
925
|
|
|
* Crawler log |
926
|
|
|
* |
927
|
|
|
************************************/ |
928
|
|
|
|
929
|
|
|
/** |
930
|
|
|
* Return array of records from crawler queue for input page ID |
931
|
|
|
* |
932
|
|
|
* @param integer $id Page ID for which to look up log entries. |
933
|
|
|
* @param string $filter Filter: "all" => all entries, "pending" => all that is not yet run, "finished" => all complete ones |
934
|
|
|
* @param boolean $doFlush If TRUE, then entries selected at DELETED(!) instead of selected! |
935
|
|
|
* @param boolean $doFullFlush |
936
|
|
|
* @param integer $itemsPerPage Limit the amount of entries per page default is 10 |
937
|
|
|
* @return array |
938
|
|
|
*/ |
939
|
4 |
|
public function getLogEntriesForPageId($id, $filter = '', $doFlush = false, $doFullFlush = false, $itemsPerPage = 10) |
940
|
|
|
{ |
941
|
4 |
|
$queryBuilder = GeneralUtility::makeInstance(ConnectionPool::class)->getQueryBuilderForTable($this->tableName); |
942
|
|
|
$queryBuilder |
943
|
4 |
|
->select('*') |
944
|
4 |
|
->from($this->tableName) |
945
|
4 |
|
->where( |
946
|
4 |
|
$queryBuilder->expr()->eq('page_id', $queryBuilder->createNamedParameter($id, \PDO::PARAM_INT)) |
947
|
|
|
) |
948
|
4 |
|
->orderBy('scheduled', 'DESC'); |
949
|
|
|
|
950
|
4 |
|
$expressionBuilder = GeneralUtility::makeInstance(ConnectionPool::class) |
951
|
4 |
|
->getConnectionForTable($this->tableName) |
952
|
4 |
|
->getExpressionBuilder(); |
953
|
4 |
|
$query = $expressionBuilder->andX(); |
954
|
|
|
// PHPStorm adds the highlight that the $addWhere is immediately overwritten, |
955
|
|
|
// but the $query = $expressionBuilder->andX() ensures that the $addWhere is written correctly with AND |
956
|
|
|
// between the statements, it's not a mistake in the code. |
957
|
4 |
|
$addWhere = '1=1'; |
958
|
4 |
|
switch ($filter) { |
959
|
4 |
|
case 'pending': |
960
|
|
|
$queryBuilder->andWhere($queryBuilder->expr()->eq('exec_time', 0)); |
961
|
|
|
$addWhere .= ' AND ' . $query->add($expressionBuilder->eq('exec_time', 0)); |
962
|
|
|
break; |
963
|
4 |
|
case 'finished': |
964
|
|
|
$queryBuilder->andWhere($queryBuilder->expr()->gt('exec_time', 0)); |
965
|
|
|
$addWhere .= ' AND ' . $query->add($expressionBuilder->gt('exec_time', 0)); |
966
|
|
|
break; |
967
|
4 |
|
case 'all': |
968
|
|
|
$doFullFlush = $doFullFlush ?: false; |
969
|
|
|
break; |
970
|
|
|
} |
971
|
|
|
|
972
|
|
|
// FIXME: Write unit test that ensures that the right records are deleted. |
973
|
4 |
|
if ($doFlush) { |
974
|
|
|
// We do currently ignore PageId by flush. |
975
|
|
|
// To have pending and finished parameters accepted |
976
|
|
|
// 2020.04.11 - Tomas Mikkelsen |
977
|
|
|
// $addWhere = $query->add($expressionBuilder->eq('page_id', (int)$id)); |
978
|
2 |
|
$this->flushQueue($doFullFlush ? '1=1' : $addWhere); |
979
|
2 |
|
return []; |
980
|
|
|
} else { |
981
|
2 |
|
if ($itemsPerPage > 0) { |
982
|
|
|
$queryBuilder |
983
|
2 |
|
->setMaxResults((int)$itemsPerPage); |
984
|
|
|
} |
985
|
|
|
|
986
|
2 |
|
return $queryBuilder->execute()->fetchAll(); |
987
|
|
|
} |
988
|
|
|
} |
989
|
|
|
|
990
|
|
|
/** |
991
|
|
|
* Return array of records from crawler queue for input set ID |
992
|
|
|
* |
993
|
|
|
* @param int $set_id Set ID for which to look up log entries. |
994
|
|
|
* @param string $filter Filter: "all" => all entries, "pending" => all that is not yet run, "finished" => all complete ones |
995
|
|
|
* @param bool $doFlush If TRUE, then entries selected at DELETED(!) instead of selected! |
996
|
|
|
* @param int $itemsPerPage Limit the amount of entires per page default is 10 |
997
|
|
|
* @return array |
998
|
|
|
*/ |
999
|
6 |
|
public function getLogEntriesForSetId(int $set_id, string $filter = '', bool $doFlush = false, bool $doFullFlush = false, int $itemsPerPage = 10) |
1000
|
|
|
{ |
1001
|
6 |
|
$queryBuilder = GeneralUtility::makeInstance(ConnectionPool::class)->getQueryBuilderForTable($this->tableName); |
1002
|
|
|
$queryBuilder |
1003
|
6 |
|
->select('*') |
1004
|
6 |
|
->from($this->tableName) |
1005
|
6 |
|
->where( |
1006
|
6 |
|
$queryBuilder->expr()->eq('set_id', $queryBuilder->createNamedParameter($set_id, \PDO::PARAM_INT)) |
1007
|
|
|
) |
1008
|
6 |
|
->orderBy('scheduled', 'DESC'); |
1009
|
|
|
|
1010
|
6 |
|
$expressionBuilder = GeneralUtility::makeInstance(ConnectionPool::class) |
1011
|
6 |
|
->getConnectionForTable($this->tableName) |
1012
|
6 |
|
->getExpressionBuilder(); |
1013
|
6 |
|
$query = $expressionBuilder->andX(); |
1014
|
|
|
// FIXME: Write Unit tests for Filters |
1015
|
|
|
// PHPStorm adds the highlight that the $addWhere is immediately overwritten, |
1016
|
|
|
// but the $query = $expressionBuilder->andX() ensures that the $addWhere is written correctly with AND |
1017
|
|
|
// between the statements, it's not a mistake in the code. |
1018
|
6 |
|
$addWhere = ''; |
1019
|
6 |
|
switch ($filter) { |
1020
|
6 |
|
case 'pending': |
1021
|
1 |
|
$queryBuilder->andWhere($queryBuilder->expr()->eq('exec_time', 0)); |
1022
|
1 |
|
$addWhere = $query->add($expressionBuilder->eq('exec_time', 0)); |
|
|
|
|
1023
|
1 |
|
break; |
1024
|
5 |
|
case 'finished': |
1025
|
1 |
|
$queryBuilder->andWhere($queryBuilder->expr()->gt('exec_time', 0)); |
1026
|
1 |
|
$addWhere = $query->add($expressionBuilder->gt('exec_time', 0)); |
1027
|
1 |
|
break; |
1028
|
|
|
} |
1029
|
|
|
// FIXME: Write unit test that ensures that the right records are deleted. |
1030
|
6 |
|
if ($doFlush) { |
1031
|
4 |
|
$addWhere = $query->add($expressionBuilder->eq('set_id', (int)$set_id)); |
1032
|
4 |
|
$this->flushQueue($doFullFlush ? '' : $addWhere); |
1033
|
4 |
|
return []; |
1034
|
|
|
} else { |
1035
|
2 |
|
if ($itemsPerPage > 0) { |
1036
|
|
|
$queryBuilder |
1037
|
2 |
|
->setMaxResults((int)$itemsPerPage); |
1038
|
|
|
} |
1039
|
|
|
|
1040
|
2 |
|
return $queryBuilder->execute()->fetchAll(); |
1041
|
|
|
} |
1042
|
|
|
} |
1043
|
|
|
|
1044
|
|
|
/** |
1045
|
|
|
* Removes queue entries |
1046
|
|
|
* |
1047
|
|
|
* @param string $where SQL related filter for the entries which should be removed |
1048
|
|
|
* @return void |
1049
|
|
|
*/ |
1050
|
11 |
|
protected function flushQueue($where = ''): void |
1051
|
|
|
{ |
1052
|
11 |
|
$realWhere = strlen((string)$where) > 0 ? $where : '1=1'; |
1053
|
|
|
|
1054
|
11 |
|
$queryBuilder = $this->getQueryBuilder($this->tableName); |
1055
|
|
|
|
1056
|
11 |
|
if (EventDispatcher::getInstance()->hasObserver('queueEntryFlush')) { |
1057
|
|
|
$groups = $queryBuilder |
1058
|
|
|
->select('DISTINCT set_id') |
1059
|
|
|
->from($this->tableName) |
1060
|
|
|
->where($realWhere) |
1061
|
|
|
->execute() |
1062
|
|
|
->fetchAll(); |
1063
|
|
|
if (is_array($groups)) { |
|
|
|
|
1064
|
|
|
foreach ($groups as $group) { |
1065
|
|
|
$subSet = $queryBuilder |
1066
|
|
|
->select('uid', 'set_id') |
1067
|
|
|
->from($this->tableName) |
1068
|
|
|
->where( |
1069
|
|
|
$realWhere, |
1070
|
|
|
$queryBuilder->expr()->eq('set_id', $group['set_id']) |
1071
|
|
|
) |
1072
|
|
|
->execute() |
1073
|
|
|
->fetchAll(); |
1074
|
|
|
EventDispatcher::getInstance()->post('queueEntryFlush', $group['set_id'], $subSet); |
1075
|
|
|
} |
1076
|
|
|
} |
1077
|
|
|
} |
1078
|
|
|
|
1079
|
|
|
$queryBuilder |
1080
|
11 |
|
->delete($this->tableName) |
1081
|
11 |
|
->where($realWhere) |
1082
|
11 |
|
->execute(); |
1083
|
11 |
|
} |
1084
|
|
|
|
1085
|
|
|
/** |
1086
|
|
|
* Adding call back entries to log (called from hooks typically, see indexed search class "class.crawler.php" |
1087
|
|
|
* |
1088
|
|
|
* @param integer $setId Set ID |
1089
|
|
|
* @param array $params Parameters to pass to call back function |
1090
|
|
|
* @param string $callBack Call back object reference, eg. 'EXT:indexed_search/class.crawler.php:&tx_indexedsearch_crawler' |
1091
|
|
|
* @param integer $page_id Page ID to attach it to |
1092
|
|
|
* @param integer $schedule Time at which to activate |
1093
|
|
|
* @return void |
1094
|
|
|
*/ |
1095
|
|
|
public function addQueueEntry_callBack($setId, $params, $callBack, $page_id = 0, $schedule = 0): void |
1096
|
|
|
{ |
1097
|
|
|
if (!is_array($params)) { |
|
|
|
|
1098
|
|
|
$params = []; |
1099
|
|
|
} |
1100
|
|
|
$params['_CALLBACKOBJ'] = $callBack; |
1101
|
|
|
|
1102
|
|
|
GeneralUtility::makeInstance(ConnectionPool::class)->getConnectionForTable('tx_crawler_queue') |
1103
|
|
|
->insert( |
1104
|
|
|
'tx_crawler_queue', |
1105
|
|
|
[ |
1106
|
|
|
'page_id' => (int)$page_id, |
1107
|
|
|
'parameters' => serialize($params), |
1108
|
|
|
'scheduled' => (int)$schedule ?: $this->getCurrentTime(), |
1109
|
|
|
'exec_time' => 0, |
1110
|
|
|
'set_id' => (int)$setId, |
1111
|
|
|
'result_data' => '', |
1112
|
|
|
] |
1113
|
|
|
); |
1114
|
|
|
} |
1115
|
|
|
|
1116
|
|
|
/************************************ |
1117
|
|
|
* |
1118
|
|
|
* URL setting |
1119
|
|
|
* |
1120
|
|
|
************************************/ |
1121
|
|
|
|
1122
|
|
|
/** |
1123
|
|
|
* Setting a URL for crawling: |
1124
|
|
|
* |
1125
|
|
|
* @param integer $id Page ID |
1126
|
|
|
* @param string $url Complete URL |
1127
|
|
|
* @param array $subCfg Sub configuration array (from TS config) |
1128
|
|
|
* @param integer $tstamp Scheduled-time |
1129
|
|
|
* @param string $configurationHash (optional) configuration hash |
1130
|
|
|
* @param bool $skipInnerDuplicationCheck (optional) skip inner duplication check |
1131
|
|
|
* @return bool |
1132
|
|
|
*/ |
1133
|
6 |
|
public function addUrl( |
1134
|
|
|
$id, |
1135
|
|
|
$url, |
1136
|
|
|
array $subCfg, |
1137
|
|
|
$tstamp, |
1138
|
|
|
$configurationHash = '', |
1139
|
|
|
$skipInnerDuplicationCheck = false |
1140
|
|
|
) { |
1141
|
6 |
|
$urlAdded = false; |
1142
|
6 |
|
$rows = []; |
1143
|
|
|
|
1144
|
|
|
// Creating parameters: |
1145
|
|
|
$parameters = [ |
1146
|
6 |
|
'url' => $url, |
1147
|
|
|
]; |
1148
|
|
|
|
1149
|
|
|
// fe user group simulation: |
1150
|
6 |
|
$uGs = implode(',', array_unique(GeneralUtility::intExplode(',', $subCfg['userGroups'], true))); |
1151
|
6 |
|
if ($uGs) { |
1152
|
1 |
|
$parameters['feUserGroupList'] = $uGs; |
1153
|
|
|
} |
1154
|
|
|
|
1155
|
|
|
// Setting processing instructions |
1156
|
6 |
|
$parameters['procInstructions'] = GeneralUtility::trimExplode(',', $subCfg['procInstrFilter']); |
1157
|
6 |
|
if (is_array($subCfg['procInstrParams.'])) { |
1158
|
3 |
|
$parameters['procInstrParams'] = $subCfg['procInstrParams.']; |
1159
|
|
|
} |
1160
|
|
|
|
1161
|
|
|
// Compile value array: |
1162
|
6 |
|
$parameters_serialized = serialize($parameters); |
1163
|
|
|
$fieldArray = [ |
1164
|
6 |
|
'page_id' => (int)$id, |
1165
|
6 |
|
'parameters' => $parameters_serialized, |
1166
|
6 |
|
'parameters_hash' => GeneralUtility::shortMD5($parameters_serialized), |
1167
|
6 |
|
'configuration_hash' => $configurationHash, |
1168
|
6 |
|
'scheduled' => $tstamp, |
1169
|
6 |
|
'exec_time' => 0, |
1170
|
6 |
|
'set_id' => (int)$this->setID, |
1171
|
6 |
|
'result_data' => '', |
1172
|
6 |
|
'configuration' => $subCfg['key'], |
1173
|
|
|
]; |
1174
|
|
|
|
1175
|
6 |
|
if ($this->registerQueueEntriesInternallyOnly) { |
1176
|
|
|
//the entries will only be registered and not stored to the database |
1177
|
1 |
|
$this->queueEntries[] = $fieldArray; |
1178
|
|
|
} else { |
1179
|
5 |
|
if (!$skipInnerDuplicationCheck) { |
1180
|
|
|
// check if there is already an equal entry |
1181
|
4 |
|
$rows = $this->getDuplicateRowsIfExist($tstamp, $fieldArray); |
1182
|
|
|
} |
1183
|
|
|
|
1184
|
5 |
|
if (empty($rows)) { |
1185
|
4 |
|
$connectionForCrawlerQueue = GeneralUtility::makeInstance(ConnectionPool::class)->getConnectionForTable('tx_crawler_queue'); |
1186
|
4 |
|
$connectionForCrawlerQueue->insert( |
1187
|
4 |
|
'tx_crawler_queue', |
1188
|
4 |
|
$fieldArray |
1189
|
|
|
); |
1190
|
4 |
|
$uid = $connectionForCrawlerQueue->lastInsertId('tx_crawler_queue', 'qid'); |
1191
|
4 |
|
$rows[] = $uid; |
1192
|
4 |
|
$urlAdded = true; |
1193
|
4 |
|
EventDispatcher::getInstance()->post('urlAddedToQueue', strval($this->setID), ['uid' => $uid, 'fieldArray' => $fieldArray]); |
1194
|
|
|
} else { |
1195
|
1 |
|
EventDispatcher::getInstance()->post('duplicateUrlInQueue', strval($this->setID), ['rows' => $rows, 'fieldArray' => $fieldArray]); |
1196
|
|
|
} |
1197
|
|
|
} |
1198
|
|
|
|
1199
|
6 |
|
return $urlAdded; |
1200
|
|
|
} |
1201
|
|
|
|
1202
|
|
|
/** |
1203
|
|
|
* This method determines duplicates for a queue entry with the same parameters and this timestamp. |
1204
|
|
|
* If the timestamp is in the past, it will check if there is any unprocessed queue entry in the past. |
1205
|
|
|
* If the timestamp is in the future it will check, if the queued entry has exactly the same timestamp |
1206
|
|
|
* |
1207
|
|
|
* @param int $tstamp |
1208
|
|
|
* @param array $fieldArray |
1209
|
|
|
* |
1210
|
|
|
* @return array |
1211
|
|
|
*/ |
1212
|
7 |
|
protected function getDuplicateRowsIfExist($tstamp, $fieldArray) |
1213
|
|
|
{ |
1214
|
7 |
|
$rows = []; |
1215
|
|
|
|
1216
|
7 |
|
$currentTime = $this->getCurrentTime(); |
1217
|
|
|
|
1218
|
7 |
|
$queryBuilder = GeneralUtility::makeInstance(ConnectionPool::class)->getQueryBuilderForTable($this->tableName); |
1219
|
|
|
$queryBuilder |
1220
|
7 |
|
->select('qid') |
1221
|
7 |
|
->from('tx_crawler_queue'); |
1222
|
|
|
//if this entry is scheduled with "now" |
1223
|
7 |
|
if ($tstamp <= $currentTime) { |
1224
|
2 |
|
if ($this->extensionSettings['enableTimeslot']) { |
1225
|
1 |
|
$timeBegin = $currentTime - 100; |
1226
|
1 |
|
$timeEnd = $currentTime + 100; |
1227
|
|
|
$queryBuilder |
1228
|
1 |
|
->where( |
1229
|
1 |
|
'scheduled BETWEEN ' . $timeBegin . ' AND ' . $timeEnd . '' |
1230
|
|
|
) |
1231
|
1 |
|
->orWhere( |
1232
|
1 |
|
$queryBuilder->expr()->lte('scheduled', $currentTime) |
1233
|
|
|
); |
1234
|
|
|
} else { |
1235
|
|
|
$queryBuilder |
1236
|
1 |
|
->where( |
1237
|
2 |
|
$queryBuilder->expr()->lte('scheduled', $currentTime) |
1238
|
|
|
); |
1239
|
|
|
} |
1240
|
5 |
|
} elseif ($tstamp > $currentTime) { |
1241
|
|
|
//entry with a timestamp in the future need to have the same schedule time |
1242
|
|
|
$queryBuilder |
1243
|
5 |
|
->where( |
1244
|
5 |
|
$queryBuilder->expr()->eq('scheduled', $tstamp) |
1245
|
|
|
); |
1246
|
|
|
} |
1247
|
|
|
|
1248
|
|
|
$queryBuilder |
1249
|
7 |
|
->andWhere('NOT exec_time') |
1250
|
7 |
|
->andWhere('NOT process_id') |
1251
|
7 |
|
->andWhere($queryBuilder->expr()->eq('page_id', $queryBuilder->createNamedParameter($fieldArray['page_id'], \PDO::PARAM_INT))) |
1252
|
7 |
|
->andWhere($queryBuilder->expr()->eq('parameters_hash', $queryBuilder->createNamedParameter($fieldArray['parameters_hash'], \PDO::PARAM_STR))); |
1253
|
|
|
|
1254
|
7 |
|
$statement = $queryBuilder->execute(); |
1255
|
|
|
|
1256
|
7 |
|
while ($row = $statement->fetch()) { |
1257
|
5 |
|
$rows[] = $row['qid']; |
1258
|
|
|
} |
1259
|
|
|
|
1260
|
7 |
|
return $rows; |
1261
|
|
|
} |
1262
|
|
|
|
1263
|
|
|
/** |
1264
|
|
|
* Returns the current system time |
1265
|
|
|
* |
1266
|
|
|
* @return int |
1267
|
|
|
*/ |
1268
|
|
|
public function getCurrentTime() |
1269
|
|
|
{ |
1270
|
|
|
return time(); |
1271
|
|
|
} |
1272
|
|
|
|
1273
|
|
|
/************************************ |
1274
|
|
|
* |
1275
|
|
|
* URL reading |
1276
|
|
|
* |
1277
|
|
|
************************************/ |
1278
|
|
|
|
1279
|
|
|
/** |
1280
|
|
|
* Read URL for single queue entry |
1281
|
|
|
* |
1282
|
|
|
* @param integer $queueId |
1283
|
|
|
* @param boolean $force If set, will process even if exec_time has been set! |
1284
|
|
|
* @return integer |
1285
|
|
|
*/ |
1286
|
|
|
public function readUrl($queueId, $force = false) |
1287
|
|
|
{ |
1288
|
|
|
$queryBuilder = GeneralUtility::makeInstance(ConnectionPool::class)->getQueryBuilderForTable($this->tableName); |
1289
|
|
|
$ret = 0; |
1290
|
|
|
$this->logger->debug('crawler-readurl start ' . microtime(true)); |
1291
|
|
|
// Get entry: |
1292
|
|
|
$queryBuilder |
1293
|
|
|
->select('*') |
1294
|
|
|
->from('tx_crawler_queue') |
1295
|
|
|
->where( |
1296
|
|
|
$queryBuilder->expr()->eq('qid', $queryBuilder->createNamedParameter($queueId, \PDO::PARAM_INT)) |
1297
|
|
|
); |
1298
|
|
|
if (!$force) { |
1299
|
|
|
$queryBuilder |
1300
|
|
|
->andWhere('exec_time = 0') |
1301
|
|
|
->andWhere('process_scheduled > 0'); |
1302
|
|
|
} |
1303
|
|
|
$queueRec = $queryBuilder->execute()->fetch(); |
1304
|
|
|
|
1305
|
|
|
if (!is_array($queueRec)) { |
1306
|
|
|
return; |
1307
|
|
|
} |
1308
|
|
|
|
1309
|
|
|
SignalSlotUtility::emitSignal( |
1310
|
|
|
__CLASS__, |
1311
|
|
|
SignalSlotUtility::SIGNAL_QUEUEITEM_PREPROCESS, |
1312
|
|
|
[$queueId, &$queueRec] |
1313
|
|
|
); |
1314
|
|
|
|
1315
|
|
|
// Set exec_time to lock record: |
1316
|
|
|
$field_array = ['exec_time' => $this->getCurrentTime()]; |
1317
|
|
|
|
1318
|
|
|
if (isset($this->processID)) { |
1319
|
|
|
//if mulitprocessing is used we need to store the id of the process which has handled this entry |
1320
|
|
|
$field_array['process_id_completed'] = $this->processID; |
1321
|
|
|
} |
1322
|
|
|
|
1323
|
|
|
GeneralUtility::makeInstance(ConnectionPool::class)->getConnectionForTable('tx_crawler_queue') |
1324
|
|
|
->update( |
1325
|
|
|
'tx_crawler_queue', |
1326
|
|
|
$field_array, |
1327
|
|
|
['qid' => (int)$queueId] |
1328
|
|
|
); |
1329
|
|
|
|
1330
|
|
|
$result = $this->queueExecutor->executeQueueItem($queueRec, $this); |
1331
|
|
|
$resultData = unserialize($result['content']); |
1332
|
|
|
|
1333
|
|
|
//atm there's no need to point to specific pollable extensions |
1334
|
|
|
if (is_array($GLOBALS['TYPO3_CONF_VARS']['EXTCONF']['crawler']['pollSuccess'])) { |
1335
|
|
|
foreach ($GLOBALS['TYPO3_CONF_VARS']['EXTCONF']['crawler']['pollSuccess'] as $pollable) { |
1336
|
|
|
// only check the success value if the instruction is runnig |
1337
|
|
|
// it is important to name the pollSuccess key same as the procInstructions key |
1338
|
|
|
if (is_array($resultData['parameters']['procInstructions']) |
1339
|
|
|
&& in_array( |
1340
|
|
|
$pollable, |
1341
|
|
|
$resultData['parameters']['procInstructions'] |
1342
|
|
|
) |
1343
|
|
|
) { |
1344
|
|
|
if (!empty($resultData['success'][$pollable]) && $resultData['success'][$pollable]) { |
1345
|
|
|
$ret |= self::CLI_STATUS_POLLABLE_PROCESSED; |
1346
|
|
|
} |
1347
|
|
|
} |
1348
|
|
|
} |
1349
|
|
|
} |
1350
|
|
|
|
1351
|
|
|
// Set result in log which also denotes the end of the processing of this entry. |
1352
|
|
|
$field_array = ['result_data' => serialize($result)]; |
1353
|
|
|
|
1354
|
|
|
SignalSlotUtility::emitSignal( |
1355
|
|
|
__CLASS__, |
1356
|
|
|
SignalSlotUtility::SIGNAL_QUEUEITEM_POSTPROCESS, |
1357
|
|
|
[$queueId, &$field_array] |
1358
|
|
|
); |
1359
|
|
|
|
1360
|
|
|
GeneralUtility::makeInstance(ConnectionPool::class)->getConnectionForTable('tx_crawler_queue') |
1361
|
|
|
->update( |
1362
|
|
|
'tx_crawler_queue', |
1363
|
|
|
$field_array, |
1364
|
|
|
['qid' => (int)$queueId] |
1365
|
|
|
); |
1366
|
|
|
|
1367
|
|
|
$this->logger->debug('crawler-readurl stop ' . microtime(true)); |
1368
|
|
|
return $ret; |
1369
|
|
|
} |
1370
|
|
|
|
1371
|
|
|
/** |
1372
|
|
|
* Read URL for not-yet-inserted log-entry |
1373
|
|
|
* |
1374
|
|
|
* @param array $field_array Queue field array, |
1375
|
|
|
* |
1376
|
|
|
* @return string |
1377
|
|
|
*/ |
1378
|
|
|
public function readUrlFromArray($field_array) |
1379
|
|
|
{ |
1380
|
|
|
// Set exec_time to lock record: |
1381
|
|
|
$field_array['exec_time'] = $this->getCurrentTime(); |
1382
|
|
|
$connectionForCrawlerQueue = GeneralUtility::makeInstance(ConnectionPool::class)->getConnectionForTable($this->tableName); |
1383
|
|
|
$connectionForCrawlerQueue->insert( |
1384
|
|
|
$this->tableName, |
1385
|
|
|
$field_array |
1386
|
|
|
); |
1387
|
|
|
$queueId = $field_array['qid'] = $connectionForCrawlerQueue->lastInsertId($this->tableName, 'qid'); |
1388
|
|
|
|
1389
|
|
|
$result = $this->queueExecutor->executeQueueItem($field_array, $this); |
1390
|
|
|
|
1391
|
|
|
// Set result in log which also denotes the end of the processing of this entry. |
1392
|
|
|
$field_array = ['result_data' => serialize($result)]; |
1393
|
|
|
|
1394
|
|
|
SignalSlotUtility::emitSignal( |
1395
|
|
|
__CLASS__, |
1396
|
|
|
SignalSlotUtility::SIGNAL_QUEUEITEM_POSTPROCESS, |
1397
|
|
|
[$queueId, &$field_array] |
1398
|
|
|
); |
1399
|
|
|
|
1400
|
|
|
$connectionForCrawlerQueue->update( |
1401
|
|
|
$this->tableName, |
1402
|
|
|
$field_array, |
1403
|
|
|
['qid' => $queueId] |
1404
|
|
|
); |
1405
|
|
|
|
1406
|
|
|
return $result; |
|
|
|
|
1407
|
|
|
} |
1408
|
|
|
|
1409
|
|
|
/***************************** |
1410
|
|
|
* |
1411
|
|
|
* Compiling URLs to crawl - tools |
1412
|
|
|
* |
1413
|
|
|
*****************************/ |
1414
|
|
|
|
1415
|
|
|
/** |
1416
|
|
|
* @param integer $id Root page id to start from. |
1417
|
|
|
* @param integer $depth Depth of tree, 0=only id-page, 1= on sublevel, 99 = infinite |
1418
|
|
|
* @param integer $scheduledTime Unix Time when the URL is timed to be visited when put in queue |
1419
|
|
|
* @param integer $reqMinute Number of requests per minute (creates the interleave between requests) |
1420
|
|
|
* @param boolean $submitCrawlUrls If set, submits the URLs to queue in database (real crawling) |
1421
|
|
|
* @param boolean $downloadCrawlUrls If set (and submitcrawlUrls is false) will fill $downloadUrls with entries) |
1422
|
|
|
* @param array $incomingProcInstructions Array of processing instructions |
1423
|
|
|
* @param array $configurationSelection Array of configuration keys |
1424
|
|
|
* @return string |
1425
|
|
|
*/ |
1426
|
|
|
public function getPageTreeAndUrls( |
1427
|
|
|
$id, |
1428
|
|
|
$depth, |
1429
|
|
|
$scheduledTime, |
1430
|
|
|
$reqMinute, |
1431
|
|
|
$submitCrawlUrls, |
1432
|
|
|
$downloadCrawlUrls, |
1433
|
|
|
array $incomingProcInstructions, |
1434
|
|
|
array $configurationSelection |
1435
|
|
|
) { |
1436
|
|
|
$this->scheduledTime = $scheduledTime; |
1437
|
|
|
$this->reqMinute = $reqMinute; |
1438
|
|
|
$this->submitCrawlUrls = $submitCrawlUrls; |
1439
|
|
|
$this->downloadCrawlUrls = $downloadCrawlUrls; |
1440
|
|
|
$this->incomingProcInstructions = $incomingProcInstructions; |
1441
|
|
|
$this->incomingConfigurationSelection = $configurationSelection; |
1442
|
|
|
|
1443
|
|
|
$this->duplicateTrack = []; |
1444
|
|
|
$this->downloadUrls = []; |
1445
|
|
|
|
1446
|
|
|
// Drawing tree: |
1447
|
|
|
/* @var PageTreeView $tree */ |
1448
|
|
|
$tree = GeneralUtility::makeInstance(PageTreeView::class); |
1449
|
|
|
$perms_clause = $this->getBackendUser()->getPagePermsClause(Permission::PAGE_SHOW); |
1450
|
|
|
$tree->init('AND ' . $perms_clause); |
1451
|
|
|
|
1452
|
|
|
$pageInfo = BackendUtility::readPageAccess($id, $perms_clause); |
1453
|
|
|
if (is_array($pageInfo)) { |
|
|
|
|
1454
|
|
|
// Set root row: |
1455
|
|
|
$tree->tree[] = [ |
1456
|
|
|
'row' => $pageInfo, |
1457
|
|
|
'HTML' => $this->iconFactory->getIconForRecord('pages', $pageInfo, Icon::SIZE_SMALL), |
1458
|
|
|
]; |
1459
|
|
|
} |
1460
|
|
|
|
1461
|
|
|
// Get branch beneath: |
1462
|
|
|
if ($depth) { |
1463
|
|
|
$tree->getTree($id, $depth, ''); |
1464
|
|
|
} |
1465
|
|
|
|
1466
|
|
|
// Traverse page tree: |
1467
|
|
|
$code = ''; |
1468
|
|
|
|
1469
|
|
|
foreach ($tree->tree as $data) { |
1470
|
|
|
$this->MP = false; |
1471
|
|
|
|
1472
|
|
|
// recognize mount points |
1473
|
|
|
if ($data['row']['doktype'] == PageRepository::DOKTYPE_MOUNTPOINT) { |
1474
|
|
|
$queryBuilder = GeneralUtility::makeInstance(ConnectionPool::class)->getQueryBuilderForTable($this->tableName); |
1475
|
|
|
$queryBuilder->getRestrictions()->removeAll()->add(GeneralUtility::makeInstance(DeletedRestriction::class)); |
1476
|
|
|
$mountpage = $queryBuilder |
1477
|
|
|
->select('*') |
1478
|
|
|
->from('pages') |
1479
|
|
|
->where( |
1480
|
|
|
$queryBuilder->expr()->eq('uid', $queryBuilder->createNamedParameter($data['row']['uid'], \PDO::PARAM_INT)) |
1481
|
|
|
) |
1482
|
|
|
->execute() |
1483
|
|
|
->fetchAll(); |
1484
|
|
|
$queryBuilder->resetRestrictions(); |
1485
|
|
|
|
1486
|
|
|
// fetch mounted pages |
1487
|
|
|
$this->MP = $mountpage[0]['mount_pid'] . '-' . $data['row']['uid']; |
|
|
|
|
1488
|
|
|
|
1489
|
|
|
$mountTree = GeneralUtility::makeInstance(PageTreeView::class); |
1490
|
|
|
$mountTree->init('AND ' . $perms_clause); |
1491
|
|
|
$mountTree->getTree($mountpage[0]['mount_pid'], $depth); |
1492
|
|
|
|
1493
|
|
|
foreach ($mountTree->tree as $mountData) { |
1494
|
|
|
$code .= $this->drawURLs_addRowsForPage( |
1495
|
|
|
$mountData['row'], |
1496
|
|
|
$mountData['HTML'] . BackendUtility::getRecordTitle('pages', $mountData['row'], true) |
1497
|
|
|
); |
1498
|
|
|
} |
1499
|
|
|
|
1500
|
|
|
// replace page when mount_pid_ol is enabled |
1501
|
|
|
if ($mountpage[0]['mount_pid_ol']) { |
1502
|
|
|
$data['row']['uid'] = $mountpage[0]['mount_pid']; |
1503
|
|
|
} else { |
1504
|
|
|
// if the mount_pid_ol is not set the MP must not be used for the mountpoint page |
1505
|
|
|
$this->MP = false; |
1506
|
|
|
} |
1507
|
|
|
} |
1508
|
|
|
|
1509
|
|
|
$code .= $this->drawURLs_addRowsForPage( |
1510
|
|
|
$data['row'], |
1511
|
|
|
$data['HTML'] . BackendUtility::getRecordTitle('pages', $data['row'], true) |
1512
|
|
|
); |
1513
|
|
|
} |
1514
|
|
|
|
1515
|
|
|
return $code; |
1516
|
|
|
} |
1517
|
|
|
|
1518
|
|
|
/** |
1519
|
|
|
* Expands exclude string |
1520
|
|
|
* |
1521
|
|
|
* @param string $excludeString Exclude string |
1522
|
|
|
* @return array |
1523
|
|
|
*/ |
1524
|
1 |
|
public function expandExcludeString($excludeString) |
1525
|
|
|
{ |
1526
|
|
|
// internal static caches; |
1527
|
1 |
|
static $expandedExcludeStringCache; |
1528
|
1 |
|
static $treeCache; |
1529
|
|
|
|
1530
|
1 |
|
if (empty($expandedExcludeStringCache[$excludeString])) { |
1531
|
1 |
|
$pidList = []; |
1532
|
|
|
|
1533
|
1 |
|
if (!empty($excludeString)) { |
1534
|
|
|
/** @var PageTreeView $tree */ |
1535
|
|
|
$tree = GeneralUtility::makeInstance(PageTreeView::class); |
1536
|
|
|
$tree->init('AND ' . $this->getBackendUser()->getPagePermsClause(Permission::PAGE_SHOW)); |
1537
|
|
|
|
1538
|
|
|
$excludeParts = GeneralUtility::trimExplode(',', $excludeString); |
1539
|
|
|
|
1540
|
|
|
foreach ($excludeParts as $excludePart) { |
1541
|
|
|
[$pid, $depth] = GeneralUtility::trimExplode('+', $excludePart); |
1542
|
|
|
|
1543
|
|
|
// default is "page only" = "depth=0" |
1544
|
|
|
if (empty($depth)) { |
1545
|
|
|
$depth = (stristr($excludePart, '+')) ? 99 : 0; |
1546
|
|
|
} |
1547
|
|
|
|
1548
|
|
|
$pidList[] = $pid; |
1549
|
|
|
|
1550
|
|
|
if ($depth > 0) { |
1551
|
|
|
if (empty($treeCache[$pid][$depth])) { |
1552
|
|
|
$tree->reset(); |
1553
|
|
|
$tree->getTree($pid, $depth); |
1554
|
|
|
$treeCache[$pid][$depth] = $tree->tree; |
1555
|
|
|
} |
1556
|
|
|
|
1557
|
|
|
foreach ($treeCache[$pid][$depth] as $data) { |
1558
|
|
|
$pidList[] = $data['row']['uid']; |
1559
|
|
|
} |
1560
|
|
|
} |
1561
|
|
|
} |
1562
|
|
|
} |
1563
|
|
|
|
1564
|
1 |
|
$expandedExcludeStringCache[$excludeString] = array_unique($pidList); |
1565
|
|
|
} |
1566
|
|
|
|
1567
|
1 |
|
return $expandedExcludeStringCache[$excludeString]; |
1568
|
|
|
} |
1569
|
|
|
|
1570
|
|
|
/** |
1571
|
|
|
* Create the rows for display of the page tree |
1572
|
|
|
* For each page a number of rows are shown displaying GET variable configuration |
1573
|
|
|
*/ |
1574
|
|
|
public function drawURLs_addRowsForPage(array $pageRow, string $pageTitle): string |
1575
|
|
|
{ |
1576
|
|
|
$skipMessage = ''; |
1577
|
|
|
|
1578
|
|
|
// Get list of configurations |
1579
|
|
|
$configurations = $this->getUrlsForPageRow($pageRow, $skipMessage); |
1580
|
|
|
|
1581
|
|
|
if (!empty($this->incomingConfigurationSelection)) { |
1582
|
|
|
// remove configuration that does not match the current selection |
1583
|
|
|
foreach ($configurations as $confKey => $confArray) { |
1584
|
|
|
if (!in_array($confKey, $this->incomingConfigurationSelection)) { |
1585
|
|
|
unset($configurations[$confKey]); |
1586
|
|
|
} |
1587
|
|
|
} |
1588
|
|
|
} |
1589
|
|
|
|
1590
|
|
|
// Traverse parameter combinations: |
1591
|
|
|
$c = 0; |
1592
|
|
|
$content = ''; |
1593
|
|
|
if (!empty($configurations)) { |
1594
|
|
|
foreach ($configurations as $confKey => $confArray) { |
1595
|
|
|
|
1596
|
|
|
// Title column: |
1597
|
|
|
if (!$c) { |
1598
|
|
|
$titleClm = '<td rowspan="' . count($configurations) . '">' . $pageTitle . '</td>'; |
1599
|
|
|
} else { |
1600
|
|
|
$titleClm = ''; |
1601
|
|
|
} |
1602
|
|
|
|
1603
|
|
|
if (!in_array($pageRow['uid'], $this->expandExcludeString($confArray['subCfg']['exclude']))) { |
1604
|
|
|
|
1605
|
|
|
// URL list: |
1606
|
|
|
$urlList = $this->urlListFromUrlArray( |
1607
|
|
|
$confArray, |
1608
|
|
|
$pageRow, |
1609
|
|
|
$this->scheduledTime, |
1610
|
|
|
$this->reqMinute, |
1611
|
|
|
$this->submitCrawlUrls, |
1612
|
|
|
$this->downloadCrawlUrls, |
1613
|
|
|
$this->duplicateTrack, |
1614
|
|
|
$this->downloadUrls, |
1615
|
|
|
$this->incomingProcInstructions // if empty the urls won't be filtered by processing instructions |
1616
|
|
|
); |
1617
|
|
|
|
1618
|
|
|
// Expanded parameters: |
1619
|
|
|
$paramExpanded = ''; |
1620
|
|
|
$calcAccu = []; |
1621
|
|
|
$calcRes = 1; |
1622
|
|
|
foreach ($confArray['paramExpanded'] as $gVar => $gVal) { |
1623
|
|
|
$paramExpanded .= ' |
1624
|
|
|
<tr> |
1625
|
|
|
<td>' . htmlspecialchars('&' . $gVar . '=') . '<br/>' . |
1626
|
|
|
'(' . count($gVal) . ')' . |
1627
|
|
|
'</td> |
1628
|
|
|
<td nowrap="nowrap">' . nl2br(htmlspecialchars(implode(chr(10), $gVal))) . '</td> |
1629
|
|
|
</tr> |
1630
|
|
|
'; |
1631
|
|
|
$calcRes *= count($gVal); |
1632
|
|
|
$calcAccu[] = count($gVal); |
1633
|
|
|
} |
1634
|
|
|
$paramExpanded = '<table>' . $paramExpanded . '</table>'; |
1635
|
|
|
$paramExpanded .= 'Comb: ' . implode('*', $calcAccu) . '=' . $calcRes; |
1636
|
|
|
|
1637
|
|
|
// Options |
1638
|
|
|
$optionValues = ''; |
1639
|
|
|
if ($confArray['subCfg']['userGroups']) { |
1640
|
|
|
$optionValues .= 'User Groups: ' . $confArray['subCfg']['userGroups'] . '<br/>'; |
1641
|
|
|
} |
1642
|
|
|
if ($confArray['subCfg']['procInstrFilter']) { |
1643
|
|
|
$optionValues .= 'ProcInstr: ' . $confArray['subCfg']['procInstrFilter'] . '<br/>'; |
1644
|
|
|
} |
1645
|
|
|
|
1646
|
|
|
// Compile row: |
1647
|
|
|
$content .= ' |
1648
|
|
|
<tr> |
1649
|
|
|
' . $titleClm . ' |
1650
|
|
|
<td>' . htmlspecialchars($confKey) . '</td> |
1651
|
|
|
<td>' . nl2br(htmlspecialchars(rawurldecode(trim(str_replace('&', chr(10) . '&', GeneralUtility::implodeArrayForUrl('', $confArray['paramParsed'])))))) . '</td> |
1652
|
|
|
<td>' . $paramExpanded . '</td> |
1653
|
|
|
<td nowrap="nowrap">' . $urlList . '</td> |
1654
|
|
|
<td nowrap="nowrap">' . $optionValues . '</td> |
1655
|
|
|
<td nowrap="nowrap">' . DebugUtility::viewArray($confArray['subCfg']['procInstrParams.']) . '</td> |
1656
|
|
|
</tr>'; |
1657
|
|
|
} else { |
1658
|
|
|
$content .= '<tr> |
1659
|
|
|
' . $titleClm . ' |
1660
|
|
|
<td>' . htmlspecialchars($confKey) . '</td> |
1661
|
|
|
<td colspan="5"><em>No entries</em> (Page is excluded in this configuration)</td> |
1662
|
|
|
</tr>'; |
1663
|
|
|
} |
1664
|
|
|
|
1665
|
|
|
$c++; |
1666
|
|
|
} |
1667
|
|
|
} else { |
1668
|
|
|
$message = !empty($skipMessage) ? ' (' . $skipMessage . ')' : ''; |
1669
|
|
|
|
1670
|
|
|
// Compile row: |
1671
|
|
|
$content .= ' |
1672
|
|
|
<tr> |
1673
|
|
|
<td>' . $pageTitle . '</td> |
1674
|
|
|
<td colspan="6"><em>No entries</em>' . $message . '</td> |
1675
|
|
|
</tr>'; |
1676
|
|
|
} |
1677
|
|
|
|
1678
|
|
|
return $content; |
1679
|
|
|
} |
1680
|
|
|
|
1681
|
|
|
/***************************** |
1682
|
|
|
* |
1683
|
|
|
* CLI functions |
1684
|
|
|
* |
1685
|
|
|
*****************************/ |
1686
|
|
|
|
1687
|
|
|
/** |
1688
|
|
|
* Running the functionality of the CLI (crawling URLs from queue) |
1689
|
|
|
*/ |
1690
|
|
|
public function CLI_run(int $countInARun, int $sleepTime, int $sleepAfterFinish): int |
1691
|
|
|
{ |
1692
|
|
|
$result = 0; |
1693
|
|
|
$counter = 0; |
1694
|
|
|
|
1695
|
|
|
// First, run hooks: |
1696
|
|
|
$this->CLI_runHooks(); |
1697
|
|
|
|
1698
|
|
|
// Clean up the queue |
1699
|
|
|
$this->queueRepository->cleanupQueue(); |
1700
|
|
|
|
1701
|
|
|
// Select entries: |
1702
|
|
|
//TODO Shouldn't this reside within the transaction? |
1703
|
|
|
$queryBuilderSelect = GeneralUtility::makeInstance(ConnectionPool::class)->getQueryBuilderForTable($this->tableName); |
1704
|
|
|
$rows = $queryBuilderSelect |
1705
|
|
|
->select('qid', 'scheduled') |
1706
|
|
|
->from($this->tableName) |
1707
|
|
|
->where( |
1708
|
|
|
$queryBuilderSelect->expr()->eq('exec_time', 0), |
1709
|
|
|
$queryBuilderSelect->expr()->eq('process_scheduled', 0), |
1710
|
|
|
$queryBuilderSelect->expr()->lte('scheduled', $this->getCurrentTime()) |
1711
|
|
|
) |
1712
|
|
|
->orderBy('scheduled') |
1713
|
|
|
->addOrderBy('qid') |
1714
|
|
|
->setMaxResults($countInARun) |
1715
|
|
|
->execute() |
1716
|
|
|
->fetchAll(); |
1717
|
|
|
|
1718
|
|
|
if (!empty($rows)) { |
1719
|
|
|
$quidList = []; |
1720
|
|
|
|
1721
|
|
|
foreach ($rows as $r) { |
1722
|
|
|
$quidList[] = $r['qid']; |
1723
|
|
|
} |
1724
|
|
|
|
1725
|
|
|
$processId = $this->CLI_buildProcessId(); |
1726
|
|
|
|
1727
|
|
|
//reserve queue entries for process |
1728
|
|
|
|
1729
|
|
|
//$this->queryBuilder->getConnection()->executeQuery('BEGIN'); |
1730
|
|
|
//TODO make sure we're not taking assigned queue-entires |
1731
|
|
|
|
1732
|
|
|
//save the number of assigned queue entries to determine who many have been processed later |
1733
|
|
|
$queryBuilderUpdate = GeneralUtility::makeInstance(ConnectionPool::class)->getQueryBuilderForTable($this->tableName); |
1734
|
|
|
$numberOfAffectedRows = $queryBuilderUpdate |
1735
|
|
|
->update($this->tableName) |
1736
|
|
|
->where( |
1737
|
|
|
$queryBuilderUpdate->expr()->in('qid', $quidList) |
1738
|
|
|
) |
1739
|
|
|
->set('process_scheduled', $this->getCurrentTime()) |
1740
|
|
|
->set('process_id', $processId) |
1741
|
|
|
->execute(); |
1742
|
|
|
|
1743
|
|
|
GeneralUtility::makeInstance(ConnectionPool::class)->getConnectionForTable('tx_crawler_process') |
1744
|
|
|
->update( |
1745
|
|
|
'tx_crawler_process', |
1746
|
|
|
['assigned_items_count' => (int)$numberOfAffectedRows], |
1747
|
|
|
['process_id' => $processId] |
1748
|
|
|
); |
1749
|
|
|
|
1750
|
|
|
if ($numberOfAffectedRows == count($quidList)) { |
1751
|
|
|
//$this->queryBuilder->getConnection()->executeQuery('COMMIT'); |
1752
|
|
|
} else { |
1753
|
|
|
//$this->queryBuilder->getConnection()->executeQuery('ROLLBACK'); |
1754
|
|
|
$this->CLI_debug("Nothing processed due to multi-process collision (" . $this->CLI_buildProcessId() . ")"); |
1755
|
|
|
return ($result | self::CLI_STATUS_ABORTED); |
1756
|
|
|
} |
1757
|
|
|
|
1758
|
|
|
foreach ($rows as $r) { |
1759
|
|
|
$result |= $this->readUrl($r['qid']); |
1760
|
|
|
|
1761
|
|
|
$counter++; |
1762
|
|
|
usleep((int)$sleepTime); // Just to relax the system |
1763
|
|
|
|
1764
|
|
|
// if during the start and the current read url the cli has been disable we need to return from the function |
1765
|
|
|
// mark the process NOT as ended. |
1766
|
|
|
if ($this->getDisabled()) { |
1767
|
|
|
return ($result | self::CLI_STATUS_ABORTED); |
1768
|
|
|
} |
1769
|
|
|
|
1770
|
|
|
if (!$this->processRepository->isProcessActive($this->CLI_buildProcessId())) { |
1771
|
|
|
$this->CLI_debug("conflict / timeout (" . $this->CLI_buildProcessId() . ")"); |
1772
|
|
|
|
1773
|
|
|
//TODO might need an additional returncode |
1774
|
|
|
$result |= self::CLI_STATUS_ABORTED; |
1775
|
|
|
break; //possible timeout |
1776
|
|
|
} |
1777
|
|
|
} |
1778
|
|
|
|
1779
|
|
|
sleep((int)$sleepAfterFinish); |
1780
|
|
|
|
1781
|
|
|
$msg = 'Rows: ' . $counter; |
1782
|
|
|
$this->CLI_debug($msg . " (" . $this->CLI_buildProcessId() . ")"); |
1783
|
|
|
} else { |
1784
|
|
|
$this->CLI_debug("Nothing within queue which needs to be processed (" . $this->CLI_buildProcessId() . ")"); |
1785
|
|
|
} |
1786
|
|
|
|
1787
|
|
|
if ($counter > 0) { |
1788
|
|
|
$result |= self::CLI_STATUS_PROCESSED; |
1789
|
|
|
} |
1790
|
|
|
|
1791
|
|
|
return $result; |
1792
|
|
|
} |
1793
|
|
|
|
1794
|
|
|
/** |
1795
|
|
|
* Activate hooks |
1796
|
|
|
* |
1797
|
|
|
* @return void |
1798
|
|
|
*/ |
1799
|
|
|
public function CLI_runHooks(): void |
1800
|
|
|
{ |
1801
|
|
|
foreach ($GLOBALS['TYPO3_CONF_VARS']['EXTCONF']['crawler']['cli_hooks'] ?? [] as $objRef) { |
1802
|
|
|
$hookObj = GeneralUtility::makeInstance($objRef); |
1803
|
|
|
if (is_object($hookObj)) { |
1804
|
|
|
$hookObj->crawler_init($this); |
1805
|
|
|
} |
1806
|
|
|
} |
1807
|
|
|
} |
1808
|
|
|
|
1809
|
|
|
/** |
1810
|
|
|
* Try to acquire a new process with the given id |
1811
|
|
|
* also performs some auto-cleanup for orphan processes |
1812
|
|
|
* @param string $id identification string for the process |
1813
|
|
|
* @return boolean |
1814
|
|
|
* @todo preemption might not be the most elegant way to clean up |
1815
|
|
|
*/ |
1816
|
|
|
public function CLI_checkAndAcquireNewProcess($id) |
1817
|
|
|
{ |
1818
|
|
|
$queryBuilder = GeneralUtility::makeInstance(ConnectionPool::class)->getQueryBuilderForTable($this->tableName); |
1819
|
|
|
$ret = true; |
1820
|
|
|
|
1821
|
|
|
$systemProcessId = getmypid(); |
1822
|
|
|
if ($systemProcessId < 1) { |
1823
|
|
|
return false; |
1824
|
|
|
} |
1825
|
|
|
|
1826
|
|
|
$processCount = 0; |
1827
|
|
|
$orphanProcesses = []; |
1828
|
|
|
|
1829
|
|
|
//$this->queryBuilder->getConnection()->executeQuery('BEGIN'); |
1830
|
|
|
|
1831
|
|
|
$statement = $queryBuilder |
1832
|
|
|
->select('process_id', 'ttl') |
1833
|
|
|
->from('tx_crawler_process') |
1834
|
|
|
->where( |
1835
|
|
|
'active = 1 AND deleted = 0' |
1836
|
|
|
) |
1837
|
|
|
->execute(); |
1838
|
|
|
|
1839
|
|
|
$currentTime = $this->getCurrentTime(); |
1840
|
|
|
|
1841
|
|
|
while ($row = $statement->fetch()) { |
1842
|
|
|
if ($row['ttl'] < $currentTime) { |
1843
|
|
|
$orphanProcesses[] = $row['process_id']; |
1844
|
|
|
} else { |
1845
|
|
|
$processCount++; |
1846
|
|
|
} |
1847
|
|
|
} |
1848
|
|
|
|
1849
|
|
|
// if there are less than allowed active processes then add a new one |
1850
|
|
|
if ($processCount < (int)$this->extensionSettings['processLimit']) { |
1851
|
|
|
$this->CLI_debug("add process " . $this->CLI_buildProcessId() . " (" . ($processCount + 1) . "/" . (int)$this->extensionSettings['processLimit'] . ")"); |
1852
|
|
|
|
1853
|
|
|
GeneralUtility::makeInstance(ConnectionPool::class)->getConnectionForTable('tx_crawler_process')->insert( |
1854
|
|
|
'tx_crawler_process', |
1855
|
|
|
[ |
1856
|
|
|
'process_id' => $id, |
1857
|
|
|
'active' => 1, |
1858
|
|
|
'ttl' => $currentTime + (int)$this->extensionSettings['processMaxRunTime'], |
1859
|
|
|
'system_process_id' => $systemProcessId, |
1860
|
|
|
] |
1861
|
|
|
); |
1862
|
|
|
} else { |
1863
|
|
|
$this->CLI_debug("Processlimit reached (" . ($processCount) . "/" . (int)$this->extensionSettings['processLimit'] . ")"); |
1864
|
|
|
$ret = false; |
1865
|
|
|
} |
1866
|
|
|
|
1867
|
|
|
$this->processRepository->deleteProcessesMarkedAsDeleted(); |
1868
|
|
|
$this->CLI_releaseProcesses($orphanProcesses); |
1869
|
|
|
|
1870
|
|
|
return $ret; |
1871
|
|
|
} |
1872
|
|
|
|
1873
|
|
|
/** |
1874
|
|
|
* Release a process and the required resources |
1875
|
|
|
* |
1876
|
|
|
* @param mixed $releaseIds string with a single process-id or array with multiple process-ids |
1877
|
|
|
* @return boolean |
1878
|
|
|
*/ |
1879
|
|
|
public function CLI_releaseProcesses($releaseIds) |
1880
|
|
|
{ |
1881
|
|
|
$queryBuilder = GeneralUtility::makeInstance(ConnectionPool::class)->getQueryBuilderForTable($this->tableName); |
1882
|
|
|
|
1883
|
|
|
if (!is_array($releaseIds)) { |
1884
|
|
|
$releaseIds = [$releaseIds]; |
1885
|
|
|
} |
1886
|
|
|
|
1887
|
|
|
if (empty($releaseIds)) { |
1888
|
|
|
return false; //nothing to release |
1889
|
|
|
} |
1890
|
|
|
|
1891
|
|
|
// some kind of 2nd chance algo - this way you need at least 2 processes to have a real cleanup |
1892
|
|
|
// this ensures that a single process can't mess up the entire process table |
1893
|
|
|
|
1894
|
|
|
// mark all processes as deleted which have no "waiting" queue-entires and which are not active |
1895
|
|
|
|
1896
|
|
|
$queryBuilder |
1897
|
|
|
->update($this->tableName, 'q') |
1898
|
|
|
->where( |
1899
|
|
|
'q.process_id IN(SELECT p.process_id FROM tx_crawler_process as p WHERE p.active = 0)' |
1900
|
|
|
) |
1901
|
|
|
->set('q.process_scheduled', 0) |
1902
|
|
|
->set('q.process_id', '') |
1903
|
|
|
->execute(); |
1904
|
|
|
|
1905
|
|
|
// FIXME: Not entirely sure that this is equivalent to the previous version |
1906
|
|
|
$queryBuilder->resetQueryPart('set'); |
1907
|
|
|
|
1908
|
|
|
$queryBuilder |
1909
|
|
|
->update('tx_crawler_process') |
1910
|
|
|
->where( |
1911
|
|
|
$queryBuilder->expr()->eq('active', 0), |
1912
|
|
|
'process_id IN(SELECT q.process_id FROM tx_crawler_queue as q WHERE q.exec_time = 0)' |
1913
|
|
|
) |
1914
|
|
|
->set('system_process_id', 0) |
1915
|
|
|
->execute(); |
1916
|
|
|
|
1917
|
|
|
// mark all requested processes as non-active |
1918
|
|
|
$queryBuilder |
1919
|
|
|
->update('tx_crawler_process') |
1920
|
|
|
->where( |
1921
|
|
|
'NOT EXISTS ( |
1922
|
|
|
SELECT * FROM tx_crawler_queue |
1923
|
|
|
WHERE tx_crawler_queue.process_id = tx_crawler_process.process_id |
1924
|
|
|
AND tx_crawler_queue.exec_time = 0 |
1925
|
|
|
)', |
1926
|
|
|
$queryBuilder->expr()->in('process_id', $queryBuilder->createNamedParameter($releaseIds, Connection::PARAM_STR_ARRAY)), |
1927
|
|
|
$queryBuilder->expr()->eq('deleted', 0) |
1928
|
|
|
) |
1929
|
|
|
->set('active', 0) |
1930
|
|
|
->execute(); |
1931
|
|
|
$queryBuilder->resetQueryPart('set'); |
1932
|
|
|
$queryBuilder |
1933
|
|
|
->update($this->tableName) |
1934
|
|
|
->where( |
1935
|
|
|
$queryBuilder->expr()->eq('exec_time', 0), |
1936
|
|
|
$queryBuilder->expr()->in('process_id', $queryBuilder->createNamedParameter($releaseIds, Connection::PARAM_STR_ARRAY)) |
1937
|
|
|
) |
1938
|
|
|
->set('process_scheduled', 0) |
1939
|
|
|
->set('process_id', '') |
1940
|
|
|
->execute(); |
1941
|
|
|
|
1942
|
|
|
return true; |
1943
|
|
|
} |
1944
|
|
|
|
1945
|
|
|
/** |
1946
|
|
|
* Create a unique Id for the current process |
1947
|
|
|
* |
1948
|
|
|
* @return string the ID |
1949
|
|
|
*/ |
1950
|
1 |
|
public function CLI_buildProcessId() |
1951
|
|
|
{ |
1952
|
1 |
|
if (!$this->processID) { |
1953
|
|
|
$this->processID = GeneralUtility::shortMD5(microtime(true)); |
1954
|
|
|
} |
1955
|
1 |
|
return $this->processID; |
1956
|
|
|
} |
1957
|
|
|
|
1958
|
|
|
/** |
1959
|
|
|
* Prints a message to the stdout (only if debug-mode is enabled) |
1960
|
|
|
* |
1961
|
|
|
* @param string $msg the message |
1962
|
|
|
*/ |
1963
|
|
|
public function CLI_debug($msg): void |
1964
|
|
|
{ |
1965
|
|
|
if ((int)$this->extensionSettings['processDebug']) { |
1966
|
|
|
echo $msg . "\n"; |
1967
|
|
|
flush(); |
1968
|
|
|
} |
1969
|
|
|
} |
1970
|
|
|
|
1971
|
|
|
/** |
1972
|
|
|
* Cleans up entries that stayed for too long in the queue. These are: |
1973
|
|
|
* - processed entries that are over 1.5 days in age |
1974
|
|
|
* - scheduled entries that are over 7 days old |
1975
|
|
|
* |
1976
|
|
|
* @return void |
1977
|
|
|
*/ |
1978
|
1 |
|
public function cleanUpOldQueueEntries(): void |
1979
|
|
|
{ |
1980
|
1 |
|
$processedAgeInSeconds = $this->extensionSettings['cleanUpProcessedAge'] * 86400; // 24*60*60 Seconds in 24 hours |
1981
|
1 |
|
$scheduledAgeInSeconds = $this->extensionSettings['cleanUpScheduledAge'] * 86400; |
1982
|
|
|
|
1983
|
1 |
|
$now = time(); |
1984
|
1 |
|
$condition = '(exec_time<>0 AND exec_time<' . ($now - $processedAgeInSeconds) . ') OR scheduled<=' . ($now - $scheduledAgeInSeconds); |
1985
|
1 |
|
$this->flushQueue($condition); |
1986
|
1 |
|
} |
1987
|
|
|
|
1988
|
|
|
/** |
1989
|
|
|
* Returns a md5 hash generated from a serialized configuration array. |
1990
|
|
|
* |
1991
|
|
|
* @param array $configuration |
1992
|
|
|
* |
1993
|
|
|
* @return string |
1994
|
|
|
*/ |
1995
|
8 |
|
protected function getConfigurationHash(array $configuration) |
1996
|
|
|
{ |
1997
|
8 |
|
unset($configuration['paramExpanded']); |
1998
|
8 |
|
unset($configuration['URLs']); |
1999
|
8 |
|
return md5(serialize($configuration)); |
2000
|
|
|
} |
2001
|
|
|
|
2002
|
|
|
/** |
2003
|
|
|
* Build a URL from a Page and the Query String. If the page has a Site configuration, it can be built by using |
2004
|
|
|
* the Site instance. |
2005
|
|
|
* |
2006
|
|
|
* @param int $pageId |
2007
|
|
|
* @param string $queryString |
2008
|
|
|
* @param string|null $alternativeBaseUrl |
2009
|
|
|
* @param int $httpsOrHttp see tx_crawler_configuration.force_ssl |
2010
|
|
|
* @return UriInterface |
2011
|
|
|
* @throws \TYPO3\CMS\Core\Exception\SiteNotFoundException |
2012
|
|
|
* @throws \TYPO3\CMS\Core\Routing\InvalidRouteArgumentsException |
2013
|
|
|
*/ |
2014
|
10 |
|
protected function getUrlFromPageAndQueryParameters(int $pageId, string $queryString, ?string $alternativeBaseUrl, int $httpsOrHttp): UriInterface |
2015
|
|
|
{ |
2016
|
10 |
|
$site = GeneralUtility::makeInstance(SiteMatcher::class)->matchByPageId((int)$pageId); |
2017
|
10 |
|
if ($site instanceof Site) { |
2018
|
5 |
|
$queryString = ltrim($queryString, '?&'); |
2019
|
5 |
|
$queryParts = []; |
2020
|
5 |
|
parse_str($queryString, $queryParts); |
2021
|
5 |
|
unset($queryParts['id']); |
2022
|
|
|
// workaround as long as we don't have native language support in crawler configurations |
2023
|
5 |
|
if (isset($queryParts['L'])) { |
2024
|
|
|
$queryParts['_language'] = $queryParts['L']; |
2025
|
|
|
unset($queryParts['L']); |
2026
|
|
|
$siteLanguage = $site->getLanguageById((int)$queryParts['_language']); |
|
|
|
|
2027
|
|
|
} else { |
2028
|
5 |
|
$siteLanguage = $site->getDefaultLanguage(); |
2029
|
|
|
} |
2030
|
5 |
|
$url = $site->getRouter()->generateUri($pageId, $queryParts); |
2031
|
5 |
|
if (!empty($alternativeBaseUrl)) { |
2032
|
3 |
|
$alternativeBaseUrl = new Uri($alternativeBaseUrl); |
2033
|
3 |
|
$url = $url->withHost($alternativeBaseUrl->getHost()); |
2034
|
3 |
|
$url = $url->withScheme($alternativeBaseUrl->getScheme()); |
2035
|
3 |
|
$url = $url->withPort($alternativeBaseUrl->getPort()); |
2036
|
3 |
|
if ($userInfo = $alternativeBaseUrl->getUserInfo()) { |
2037
|
5 |
|
$url = $url->withUserInfo($userInfo); |
2038
|
|
|
} |
2039
|
|
|
} |
2040
|
|
|
} else { |
2041
|
|
|
// Technically this is not possible with site handling, but kept for backwards-compatibility reasons |
2042
|
|
|
// Once EXT:crawler is v10-only compatible, this should be removed completely |
2043
|
5 |
|
$baseUrl = ($alternativeBaseUrl ?: GeneralUtility::getIndpEnv('TYPO3_SITE_URL')); |
2044
|
5 |
|
$cacheHashCalculator = GeneralUtility::makeInstance(CacheHashCalculator::class); |
2045
|
5 |
|
$queryString .= '&cHash=' . $cacheHashCalculator->generateForParameters($queryString); |
2046
|
5 |
|
$url = rtrim($baseUrl, '/') . '/index.php' . $queryString; |
2047
|
5 |
|
$url = new Uri($url); |
2048
|
|
|
} |
2049
|
|
|
|
2050
|
10 |
|
if ($httpsOrHttp === -1) { |
2051
|
2 |
|
$url = $url->withScheme('http'); |
2052
|
8 |
|
} elseif ($httpsOrHttp === 1) { |
2053
|
6 |
|
$url = $url->withScheme('https'); |
2054
|
|
|
} |
2055
|
|
|
|
2056
|
10 |
|
return $url; |
2057
|
|
|
} |
2058
|
|
|
|
2059
|
1 |
|
protected function swapIfFirstIsLargerThanSecond(array $reg): array |
2060
|
|
|
{ |
2061
|
|
|
// Swap if first is larger than last: |
2062
|
1 |
|
if ($reg[1] > $reg[2]) { |
2063
|
|
|
$temp = $reg[2]; |
2064
|
|
|
$reg[2] = $reg[1]; |
2065
|
|
|
$reg[1] = $temp; |
2066
|
|
|
} |
2067
|
|
|
|
2068
|
1 |
|
return $reg; |
2069
|
|
|
} |
2070
|
|
|
} |
2071
|
|
|
|