|
1
|
|
|
<?php |
|
2
|
|
|
namespace AOE\Crawler\Controller; |
|
3
|
|
|
|
|
4
|
|
|
/*************************************************************** |
|
5
|
|
|
* Copyright notice |
|
6
|
|
|
* |
|
7
|
|
|
* (c) 2017 AOE GmbH <[email protected]> |
|
8
|
|
|
* |
|
9
|
|
|
* All rights reserved |
|
10
|
|
|
* |
|
11
|
|
|
* This script is part of the TYPO3 project. The TYPO3 project is |
|
12
|
|
|
* free software; you can redistribute it and/or modify |
|
13
|
|
|
* it under the terms of the GNU General Public License as published by |
|
14
|
|
|
* the Free Software Foundation; either version 3 of the License, or |
|
15
|
|
|
* (at your option) any later version. |
|
16
|
|
|
* |
|
17
|
|
|
* The GNU General Public License can be found at |
|
18
|
|
|
* http://www.gnu.org/copyleft/gpl.html. |
|
19
|
|
|
* |
|
20
|
|
|
* This script is distributed in the hope that it will be useful, |
|
21
|
|
|
* but WITHOUT ANY WARRANTY; without even the implied warranty of |
|
22
|
|
|
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the |
|
23
|
|
|
* GNU General Public License for more details. |
|
24
|
|
|
* |
|
25
|
|
|
* This copyright notice MUST APPEAR in all copies of the script! |
|
26
|
|
|
***************************************************************/ |
|
27
|
|
|
|
|
28
|
|
|
use AOE\Crawler\Command\CrawlerCommandLineController; |
|
29
|
|
|
use AOE\Crawler\Command\FlushCommandLineController; |
|
30
|
|
|
use AOE\Crawler\Command\QueueCommandLineController; |
|
31
|
|
|
use AOE\Crawler\Domain\Model\Reason; |
|
32
|
|
|
use AOE\Crawler\Event\EventDispatcher; |
|
33
|
|
|
use AOE\Crawler\Utility\IconUtility; |
|
34
|
|
|
use AOE\Crawler\Utility\SignalSlotUtility; |
|
35
|
|
|
use TYPO3\CMS\Backend\Utility\BackendUtility; |
|
36
|
|
|
use TYPO3\CMS\Backend\Tree\View\PageTreeView; |
|
37
|
|
|
use TYPO3\CMS\Core\Authentication\BackendUserAuthentication; |
|
38
|
|
|
use TYPO3\CMS\Core\Database\DatabaseConnection; |
|
39
|
|
|
use TYPO3\CMS\Core\Log\LogLevel; |
|
40
|
|
|
use TYPO3\CMS\Core\TimeTracker\NullTimeTracker; |
|
41
|
|
|
use TYPO3\CMS\Core\Utility\DebugUtility; |
|
42
|
|
|
use TYPO3\CMS\Core\Utility\ExtensionManagementUtility; |
|
43
|
|
|
use TYPO3\CMS\Core\Utility\GeneralUtility; |
|
44
|
|
|
use TYPO3\CMS\Core\Utility\MathUtility; |
|
45
|
|
|
use TYPO3\CMS\Frontend\Controller\TypoScriptFrontendController; |
|
46
|
|
|
use TYPO3\CMS\Frontend\Page\PageGenerator; |
|
47
|
|
|
use TYPO3\CMS\Frontend\Page\PageRepository; |
|
48
|
|
|
use TYPO3\CMS\Frontend\Utility\EidUtility; |
|
49
|
|
|
|
|
50
|
|
|
/** |
|
51
|
|
|
* Class CrawlerController |
|
52
|
|
|
* |
|
53
|
|
|
* @package AOE\Crawler\Controller |
|
54
|
|
|
*/ |
|
55
|
|
|
class CrawlerController |
|
56
|
|
|
{ |
|
57
|
|
|
const CLI_STATUS_NOTHING_PROCCESSED = 0; |
|
58
|
|
|
const CLI_STATUS_REMAIN = 1; //queue not empty |
|
59
|
|
|
const CLI_STATUS_PROCESSED = 2; //(some) queue items where processed |
|
60
|
|
|
const CLI_STATUS_ABORTED = 4; //instance didn't finish |
|
61
|
|
|
const CLI_STATUS_POLLABLE_PROCESSED = 8; |
|
62
|
|
|
|
|
63
|
|
|
/** |
|
64
|
|
|
* @var integer |
|
65
|
|
|
*/ |
|
66
|
|
|
public $setID = 0; |
|
67
|
|
|
|
|
68
|
|
|
/** |
|
69
|
|
|
* @var string |
|
70
|
|
|
*/ |
|
71
|
|
|
public $processID = ''; |
|
72
|
|
|
|
|
73
|
|
|
/** |
|
74
|
|
|
* One hour is max stalled time for the CLI |
|
75
|
|
|
* If the process had the status "start" for 3600 seconds, it will be regarded stalled and a new process is started |
|
76
|
|
|
* |
|
77
|
|
|
* @var integer |
|
78
|
|
|
*/ |
|
79
|
|
|
public $max_CLI_exec_time = 3600; |
|
80
|
|
|
|
|
81
|
|
|
/** |
|
82
|
|
|
* @var array |
|
83
|
|
|
*/ |
|
84
|
|
|
public $duplicateTrack = []; |
|
85
|
|
|
|
|
86
|
|
|
/** |
|
87
|
|
|
* @var array |
|
88
|
|
|
*/ |
|
89
|
|
|
public $downloadUrls = []; |
|
90
|
|
|
|
|
91
|
|
|
/** |
|
92
|
|
|
* @var array |
|
93
|
|
|
*/ |
|
94
|
|
|
public $incomingProcInstructions = []; |
|
95
|
|
|
|
|
96
|
|
|
/** |
|
97
|
|
|
* @var array |
|
98
|
|
|
*/ |
|
99
|
|
|
public $incomingConfigurationSelection = []; |
|
100
|
|
|
|
|
101
|
|
|
/** |
|
102
|
|
|
* @var bool |
|
103
|
|
|
*/ |
|
104
|
|
|
public $registerQueueEntriesInternallyOnly = false; |
|
105
|
|
|
|
|
106
|
|
|
/** |
|
107
|
|
|
* @var array |
|
108
|
|
|
*/ |
|
109
|
|
|
public $queueEntries = []; |
|
110
|
|
|
|
|
111
|
|
|
/** |
|
112
|
|
|
* @var array |
|
113
|
|
|
*/ |
|
114
|
|
|
public $urlList = []; |
|
115
|
|
|
|
|
116
|
|
|
/** |
|
117
|
|
|
* @var boolean |
|
118
|
|
|
*/ |
|
119
|
|
|
public $debugMode = false; |
|
120
|
|
|
|
|
121
|
|
|
/** |
|
122
|
|
|
* @var array |
|
123
|
|
|
*/ |
|
124
|
|
|
public $extensionSettings = []; |
|
125
|
|
|
|
|
126
|
|
|
/** |
|
127
|
|
|
* Mount Point |
|
128
|
|
|
* |
|
129
|
|
|
* @var boolean |
|
130
|
|
|
*/ |
|
131
|
|
|
public $MP = false; |
|
132
|
|
|
|
|
133
|
|
|
/** |
|
134
|
|
|
* @var string |
|
135
|
|
|
*/ |
|
136
|
|
|
protected $processFilename; |
|
137
|
|
|
|
|
138
|
|
|
/** |
|
139
|
|
|
* Holds the internal access mode can be 'gui','cli' or 'cli_im' |
|
140
|
|
|
* |
|
141
|
|
|
* @var string |
|
142
|
|
|
*/ |
|
143
|
|
|
protected $accessMode; |
|
144
|
|
|
|
|
145
|
|
|
/** |
|
146
|
|
|
* @var DatabaseConnection |
|
147
|
|
|
*/ |
|
148
|
|
|
private $db; |
|
149
|
|
|
|
|
150
|
|
|
/** |
|
151
|
|
|
* @var BackendUserAuthentication |
|
152
|
|
|
*/ |
|
153
|
|
|
private $backendUser; |
|
154
|
|
|
|
|
155
|
|
|
/** |
|
156
|
|
|
* @var integer |
|
157
|
|
|
*/ |
|
158
|
|
|
private $scheduledTime = 0; |
|
159
|
|
|
|
|
160
|
|
|
/** |
|
161
|
|
|
* @var integer |
|
162
|
|
|
*/ |
|
163
|
|
|
private $reqMinute = 0; |
|
164
|
|
|
|
|
165
|
|
|
/** |
|
166
|
|
|
* @var bool |
|
167
|
|
|
*/ |
|
168
|
|
|
private $submitCrawlUrls = false; |
|
169
|
|
|
|
|
170
|
|
|
/** |
|
171
|
|
|
* @var bool |
|
172
|
|
|
*/ |
|
173
|
|
|
private $downloadCrawlUrls = false; |
|
174
|
|
|
|
|
175
|
|
|
/** |
|
176
|
|
|
* Method to set the accessMode can be gui, cli or cli_im |
|
177
|
|
|
* |
|
178
|
|
|
* @return string |
|
179
|
|
|
*/ |
|
180
|
1 |
|
public function getAccessMode() |
|
181
|
|
|
{ |
|
182
|
1 |
|
return $this->accessMode; |
|
183
|
|
|
} |
|
184
|
|
|
|
|
185
|
|
|
/** |
|
186
|
|
|
* @param string $accessMode |
|
187
|
|
|
*/ |
|
188
|
1 |
|
public function setAccessMode($accessMode) |
|
189
|
|
|
{ |
|
190
|
1 |
|
$this->accessMode = $accessMode; |
|
191
|
1 |
|
} |
|
192
|
|
|
|
|
193
|
|
|
/** |
|
194
|
|
|
* Set disabled status to prevent processes from being processed |
|
195
|
|
|
* |
|
196
|
|
|
* @param bool $disabled (optional, defaults to true) |
|
197
|
|
|
* @return void |
|
198
|
|
|
*/ |
|
199
|
3 |
|
public function setDisabled($disabled = true) |
|
200
|
|
|
{ |
|
201
|
3 |
|
if ($disabled) { |
|
202
|
2 |
|
GeneralUtility::writeFile($this->processFilename, ''); |
|
203
|
|
|
} else { |
|
204
|
1 |
|
if (is_file($this->processFilename)) { |
|
205
|
1 |
|
unlink($this->processFilename); |
|
206
|
|
|
} |
|
207
|
|
|
} |
|
208
|
3 |
|
} |
|
209
|
|
|
|
|
210
|
|
|
/** |
|
211
|
|
|
* Get disable status |
|
212
|
|
|
* |
|
213
|
|
|
* @return bool true if disabled |
|
214
|
|
|
*/ |
|
215
|
3 |
|
public function getDisabled() |
|
216
|
|
|
{ |
|
217
|
3 |
|
if (is_file($this->processFilename)) { |
|
218
|
2 |
|
return true; |
|
219
|
|
|
} else { |
|
220
|
1 |
|
return false; |
|
221
|
|
|
} |
|
222
|
|
|
} |
|
223
|
|
|
|
|
224
|
|
|
/** |
|
225
|
|
|
* @param string $filenameWithPath |
|
226
|
|
|
* |
|
227
|
|
|
* @return void |
|
228
|
|
|
*/ |
|
229
|
4 |
|
public function setProcessFilename($filenameWithPath) |
|
230
|
|
|
{ |
|
231
|
4 |
|
$this->processFilename = $filenameWithPath; |
|
232
|
4 |
|
} |
|
233
|
|
|
|
|
234
|
|
|
/** |
|
235
|
|
|
* @return string |
|
236
|
|
|
*/ |
|
237
|
1 |
|
public function getProcessFilename() |
|
238
|
|
|
{ |
|
239
|
1 |
|
return $this->processFilename; |
|
240
|
|
|
} |
|
241
|
|
|
|
|
242
|
|
|
/************************************ |
|
243
|
|
|
* |
|
244
|
|
|
* Getting URLs based on Page TSconfig |
|
245
|
|
|
* |
|
246
|
|
|
************************************/ |
|
247
|
|
|
|
|
248
|
60 |
|
public function __construct() |
|
249
|
|
|
{ |
|
250
|
60 |
|
$this->db = $GLOBALS['TYPO3_DB']; |
|
251
|
60 |
|
$this->backendUser = $GLOBALS['BE_USER']; |
|
252
|
60 |
|
$this->processFilename = PATH_site . 'typo3temp/tx_crawler.proc'; |
|
253
|
|
|
|
|
254
|
60 |
|
$settings = unserialize($GLOBALS['TYPO3_CONF_VARS']['EXT']['extConf']['crawler']); |
|
255
|
60 |
|
$settings = is_array($settings) ? $settings : []; |
|
256
|
|
|
|
|
257
|
|
|
// read ext_em_conf_template settings and set |
|
258
|
60 |
|
$this->setExtensionSettings($settings); |
|
259
|
|
|
|
|
260
|
|
|
// set defaults: |
|
261
|
60 |
|
if (MathUtility::convertToPositiveInteger($this->extensionSettings['countInARun']) == 0) { |
|
262
|
23 |
|
$this->extensionSettings['countInARun'] = 100; |
|
263
|
|
|
} |
|
264
|
|
|
|
|
265
|
60 |
|
$this->extensionSettings['processLimit'] = MathUtility::forceIntegerInRange($this->extensionSettings['processLimit'], 1, 99, 1); |
|
266
|
60 |
|
} |
|
267
|
|
|
|
|
268
|
|
|
/** |
|
269
|
|
|
* Sets the extensions settings (unserialized pendant of $TYPO3_CONF_VARS['EXT']['extConf']['crawler']). |
|
270
|
|
|
* |
|
271
|
|
|
* @param array $extensionSettings |
|
272
|
|
|
* @return void |
|
273
|
|
|
*/ |
|
274
|
68 |
|
public function setExtensionSettings(array $extensionSettings) |
|
275
|
|
|
{ |
|
276
|
68 |
|
$this->extensionSettings = $extensionSettings; |
|
277
|
68 |
|
} |
|
278
|
|
|
|
|
279
|
|
|
/** |
|
280
|
|
|
* Check if the given page should be crawled |
|
281
|
|
|
* |
|
282
|
|
|
* @param array $pageRow |
|
283
|
|
|
* @return false|string false if the page should be crawled (not excluded), true / skipMessage if it should be skipped |
|
284
|
|
|
*/ |
|
285
|
10 |
|
public function checkIfPageShouldBeSkipped(array $pageRow) |
|
286
|
|
|
{ |
|
287
|
10 |
|
$skipPage = false; |
|
288
|
10 |
|
$skipMessage = 'Skipped'; // message will be overwritten later |
|
289
|
|
|
|
|
290
|
|
|
// if page is hidden |
|
291
|
10 |
|
if (!$this->extensionSettings['crawlHiddenPages']) { |
|
292
|
10 |
|
if ($pageRow['hidden']) { |
|
293
|
1 |
|
$skipPage = true; |
|
294
|
1 |
|
$skipMessage = 'Because page is hidden'; |
|
295
|
|
|
} |
|
296
|
|
|
} |
|
297
|
|
|
|
|
298
|
10 |
|
if (!$skipPage) { |
|
299
|
9 |
|
if (GeneralUtility::inList('3,4', $pageRow['doktype']) || $pageRow['doktype'] >= 199) { |
|
300
|
3 |
|
$skipPage = true; |
|
301
|
3 |
|
$skipMessage = 'Because doktype is not allowed'; |
|
302
|
|
|
} |
|
303
|
|
|
} |
|
304
|
|
|
|
|
305
|
10 |
|
if (!$skipPage) { |
|
306
|
6 |
|
if (is_array($GLOBALS['TYPO3_CONF_VARS']['EXTCONF']['crawler']['excludeDoktype'])) { |
|
307
|
2 |
|
foreach ($GLOBALS['TYPO3_CONF_VARS']['EXTCONF']['crawler']['excludeDoktype'] as $key => $doktypeList) { |
|
308
|
1 |
|
if (GeneralUtility::inList($doktypeList, $pageRow['doktype'])) { |
|
309
|
1 |
|
$skipPage = true; |
|
310
|
1 |
|
$skipMessage = 'Doktype was excluded by "' . $key . '"'; |
|
311
|
1 |
|
break; |
|
312
|
|
|
} |
|
313
|
|
|
} |
|
314
|
|
|
} |
|
315
|
|
|
} |
|
316
|
|
|
|
|
317
|
10 |
|
if (!$skipPage) { |
|
318
|
|
|
// veto hook |
|
319
|
5 |
|
if (is_array($GLOBALS['TYPO3_CONF_VARS']['EXTCONF']['crawler']['pageVeto'])) { |
|
320
|
|
|
foreach ($GLOBALS['TYPO3_CONF_VARS']['EXTCONF']['crawler']['pageVeto'] as $key => $func) { |
|
321
|
|
|
$params = [ |
|
322
|
|
|
'pageRow' => $pageRow |
|
323
|
|
|
]; |
|
324
|
|
|
// expects "false" if page is ok and "true" or a skipMessage if this page should _not_ be crawled |
|
325
|
|
|
$veto = GeneralUtility::callUserFunction($func, $params, $this); |
|
326
|
|
|
if ($veto !== false) { |
|
327
|
|
|
$skipPage = true; |
|
328
|
|
|
if (is_string($veto)) { |
|
329
|
|
|
$skipMessage = $veto; |
|
330
|
|
|
} else { |
|
331
|
|
|
$skipMessage = 'Veto from hook "' . htmlspecialchars($key) . '"'; |
|
332
|
|
|
} |
|
333
|
|
|
// no need to execute other hooks if a previous one return a veto |
|
334
|
|
|
break; |
|
335
|
|
|
} |
|
336
|
|
|
} |
|
337
|
|
|
} |
|
338
|
|
|
} |
|
339
|
|
|
|
|
340
|
10 |
|
return $skipPage ? $skipMessage : false; |
|
341
|
|
|
} |
|
342
|
|
|
|
|
343
|
|
|
/** |
|
344
|
|
|
* Wrapper method for getUrlsForPageId() |
|
345
|
|
|
* It returns an array of configurations and no urls! |
|
346
|
|
|
* |
|
347
|
|
|
* @param array $pageRow Page record with at least dok-type and uid columns. |
|
348
|
|
|
* @param string $skipMessage |
|
349
|
|
|
* @return array |
|
350
|
|
|
* @see getUrlsForPageId() |
|
351
|
|
|
*/ |
|
352
|
6 |
|
public function getUrlsForPageRow(array $pageRow, &$skipMessage = '') |
|
353
|
|
|
{ |
|
354
|
6 |
|
$message = $this->checkIfPageShouldBeSkipped($pageRow); |
|
355
|
|
|
|
|
356
|
6 |
|
if ($message === false) { |
|
357
|
5 |
|
$forceSsl = ($pageRow['url_scheme'] === 2) ? true : false; |
|
358
|
5 |
|
$res = $this->getUrlsForPageId($pageRow['uid'], $forceSsl); |
|
359
|
5 |
|
$skipMessage = ''; |
|
360
|
|
|
} else { |
|
361
|
1 |
|
$skipMessage = $message; |
|
362
|
1 |
|
$res = []; |
|
363
|
|
|
} |
|
364
|
|
|
|
|
365
|
6 |
|
return $res; |
|
366
|
|
|
} |
|
367
|
|
|
|
|
368
|
|
|
/** |
|
369
|
|
|
* This method is used to count if there are ANY unprocessed queue entries |
|
370
|
|
|
* of a given page_id and the configuration which matches a given hash. |
|
371
|
|
|
* If there if none, we can skip an inner detail check |
|
372
|
|
|
* |
|
373
|
|
|
* @param int $uid |
|
374
|
|
|
* @param string $configurationHash |
|
375
|
|
|
* @return boolean |
|
376
|
|
|
*/ |
|
377
|
7 |
|
protected function noUnprocessedQueueEntriesForPageWithConfigurationHashExist($uid, $configurationHash) |
|
378
|
|
|
{ |
|
379
|
7 |
|
$configurationHash = $this->db->fullQuoteStr($configurationHash, 'tx_crawler_queue'); |
|
380
|
7 |
|
$res = $this->db->exec_SELECTquery('count(*) as anz', 'tx_crawler_queue', "page_id=" . intval($uid) . " AND configuration_hash=" . $configurationHash . " AND exec_time=0"); |
|
381
|
7 |
|
$row = $this->db->sql_fetch_assoc($res); |
|
382
|
|
|
|
|
383
|
7 |
|
return ($row['anz'] == 0); |
|
384
|
|
|
} |
|
385
|
|
|
|
|
386
|
|
|
/** |
|
387
|
|
|
* Creates a list of URLs from input array (and submits them to queue if asked for) |
|
388
|
|
|
* See Web > Info module script + "indexed_search"'s crawler hook-client using this! |
|
389
|
|
|
* |
|
390
|
|
|
* @param array Information about URLs from pageRow to crawl. |
|
391
|
|
|
* @param array Page row |
|
392
|
|
|
* @param integer Unix time to schedule indexing to, typically time() |
|
393
|
|
|
* @param integer Number of requests per minute (creates the interleave between requests) |
|
394
|
|
|
* @param boolean If set, submits the URLs to queue |
|
395
|
|
|
* @param boolean If set (and submitcrawlUrls is false) will fill $downloadUrls with entries) |
|
396
|
|
|
* @param array Array which is passed by reference and contains the an id per url to secure we will not crawl duplicates |
|
397
|
|
|
* @param array Array which will be filled with URLS for download if flag is set. |
|
398
|
|
|
* @param array Array of processing instructions |
|
399
|
|
|
* @return string List of URLs (meant for display in backend module) |
|
400
|
|
|
* |
|
401
|
|
|
*/ |
|
402
|
4 |
|
public function urlListFromUrlArray( |
|
403
|
|
|
array $vv, |
|
404
|
|
|
array $pageRow, |
|
405
|
|
|
$scheduledTime, |
|
406
|
|
|
$reqMinute, |
|
407
|
|
|
$submitCrawlUrls, |
|
408
|
|
|
$downloadCrawlUrls, |
|
409
|
|
|
array &$duplicateTrack, |
|
410
|
|
|
array &$downloadUrls, |
|
411
|
|
|
array $incomingProcInstructions |
|
412
|
|
|
) { |
|
413
|
4 |
|
$urlList = ''; |
|
414
|
|
|
// realurl support (thanks to Ingo Renner) |
|
415
|
4 |
|
if (ExtensionManagementUtility::isLoaded('realurl') && $vv['subCfg']['realurl']) { |
|
416
|
|
|
|
|
417
|
|
|
/** @var tx_realurl $urlObj */ |
|
418
|
|
|
$urlObj = GeneralUtility::makeInstance('tx_realurl'); |
|
419
|
|
|
|
|
420
|
|
|
if (!empty($vv['subCfg']['baseUrl'])) { |
|
421
|
|
|
$urlParts = parse_url($vv['subCfg']['baseUrl']); |
|
422
|
|
|
$host = strtolower($urlParts['host']); |
|
423
|
|
|
$urlObj->host = $host; |
|
424
|
|
|
|
|
425
|
|
|
// First pass, finding configuration OR pointer string: |
|
426
|
|
|
$urlObj->extConf = isset($GLOBALS['TYPO3_CONF_VARS']['EXTCONF']['realurl'][$urlObj->host]) ? $GLOBALS['TYPO3_CONF_VARS']['EXTCONF']['realurl'][$urlObj->host] : $GLOBALS['TYPO3_CONF_VARS']['EXTCONF']['realurl']['_DEFAULT']; |
|
427
|
|
|
|
|
428
|
|
|
// If it turned out to be a string pointer, then look up the real config: |
|
429
|
|
|
if (is_string($urlObj->extConf)) { |
|
430
|
|
|
$urlObj->extConf = is_array($GLOBALS['TYPO3_CONF_VARS']['EXTCONF']['realurl'][$urlObj->extConf]) ? $GLOBALS['TYPO3_CONF_VARS']['EXTCONF']['realurl'][$urlObj->extConf] : $GLOBALS['TYPO3_CONF_VARS']['EXTCONF']['realurl']['_DEFAULT']; |
|
431
|
|
|
} |
|
432
|
|
|
} |
|
433
|
|
|
|
|
434
|
|
|
if (!$GLOBALS['TSFE']->sys_page) { |
|
435
|
|
|
$GLOBALS['TSFE']->sys_page = GeneralUtility::makeInstance('TYPO3\CMS\Frontend\Page\PageRepository'); |
|
436
|
|
|
} |
|
437
|
|
|
if (!$GLOBALS['TSFE']->csConvObj) { |
|
438
|
|
|
$GLOBALS['TSFE']->csConvObj = GeneralUtility::makeInstance('TYPO3\CMS\Core\Charset\CharsetConverter'); |
|
439
|
|
|
} |
|
440
|
|
|
if (!$GLOBALS['TSFE']->tmpl->rootLine[0]['uid']) { |
|
441
|
|
|
$GLOBALS['TSFE']->tmpl->rootLine[0]['uid'] = $urlObj->extConf['pagePath']['rootpage_id']; |
|
442
|
|
|
} |
|
443
|
|
|
} |
|
444
|
|
|
|
|
445
|
4 |
|
if (is_array($vv['URLs'])) { |
|
446
|
4 |
|
$configurationHash = $this->getConfigurationHash($vv); |
|
447
|
4 |
|
$skipInnerCheck = $this->noUnprocessedQueueEntriesForPageWithConfigurationHashExist($pageRow['uid'], $configurationHash); |
|
448
|
|
|
|
|
449
|
4 |
|
foreach ($vv['URLs'] as $urlQuery) { |
|
450
|
4 |
|
if ($this->drawURLs_PIfilter($vv['subCfg']['procInstrFilter'], $incomingProcInstructions)) { |
|
451
|
|
|
|
|
452
|
|
|
// Calculate cHash: |
|
453
|
4 |
|
if ($vv['subCfg']['cHash']) { |
|
454
|
|
|
/* @var $cacheHash \TYPO3\CMS\Frontend\Page\CacheHashCalculator */ |
|
455
|
|
|
$cacheHash = GeneralUtility::makeInstance('TYPO3\CMS\Frontend\Page\CacheHashCalculator'); |
|
456
|
|
|
$urlQuery .= '&cHash=' . $cacheHash->generateForParameters($urlQuery); |
|
457
|
|
|
} |
|
458
|
|
|
|
|
459
|
|
|
// Create key by which to determine unique-ness: |
|
460
|
4 |
|
$uKey = $urlQuery . '|' . $vv['subCfg']['userGroups'] . '|' . $vv['subCfg']['baseUrl'] . '|' . $vv['subCfg']['procInstrFilter']; |
|
461
|
|
|
|
|
462
|
|
|
// realurl support (thanks to Ingo Renner) |
|
463
|
4 |
|
$urlQuery = 'index.php' . $urlQuery; |
|
464
|
4 |
|
if (ExtensionManagementUtility::isLoaded('realurl') && $vv['subCfg']['realurl']) { |
|
465
|
|
|
$params = [ |
|
466
|
|
|
'LD' => [ |
|
467
|
|
|
'totalURL' => $urlQuery |
|
468
|
|
|
], |
|
469
|
|
|
'TCEmainHook' => true |
|
470
|
|
|
]; |
|
471
|
|
|
$urlObj->encodeSpURL($params); |
|
|
|
|
|
|
472
|
|
|
$urlQuery = $params['LD']['totalURL']; |
|
473
|
|
|
} |
|
474
|
|
|
|
|
475
|
|
|
// Scheduled time: |
|
476
|
4 |
|
$schTime = $scheduledTime + round(count($duplicateTrack) * (60 / $reqMinute)); |
|
477
|
4 |
|
$schTime = floor($schTime / 60) * 60; |
|
478
|
|
|
|
|
479
|
4 |
|
if (isset($duplicateTrack[$uKey])) { |
|
480
|
|
|
|
|
481
|
|
|
//if the url key is registered just display it and do not resubmit is |
|
482
|
|
|
$urlList = '<em><span class="typo3-dimmed">' . htmlspecialchars($urlQuery) . '</span></em><br/>'; |
|
483
|
|
|
} else { |
|
484
|
4 |
|
$urlList = '[' . date('d.m.y H:i', $schTime) . '] ' . htmlspecialchars($urlQuery); |
|
485
|
4 |
|
$this->urlList[] = '[' . date('d.m.y H:i', $schTime) . '] ' . $urlQuery; |
|
486
|
|
|
|
|
487
|
4 |
|
$theUrl = ($vv['subCfg']['baseUrl'] ? $vv['subCfg']['baseUrl'] : GeneralUtility::getIndpEnv('TYPO3_SITE_URL')) . $urlQuery; |
|
488
|
|
|
|
|
489
|
|
|
// Submit for crawling! |
|
490
|
4 |
|
if ($submitCrawlUrls) { |
|
491
|
4 |
|
$added = $this->addUrl( |
|
492
|
4 |
|
$pageRow['uid'], |
|
493
|
4 |
|
$theUrl, |
|
494
|
4 |
|
$vv['subCfg'], |
|
495
|
4 |
|
$scheduledTime, |
|
496
|
4 |
|
$configurationHash, |
|
497
|
4 |
|
$skipInnerCheck |
|
498
|
|
|
); |
|
499
|
4 |
|
if ($added === false) { |
|
500
|
4 |
|
$urlList .= ' (Url already existed)'; |
|
501
|
|
|
} |
|
502
|
|
|
} elseif ($downloadCrawlUrls) { |
|
503
|
|
|
$downloadUrls[$theUrl] = $theUrl; |
|
504
|
|
|
} |
|
505
|
|
|
|
|
506
|
4 |
|
$urlList .= '<br />'; |
|
507
|
|
|
} |
|
508
|
4 |
|
$duplicateTrack[$uKey] = true; |
|
509
|
|
|
} |
|
510
|
|
|
} |
|
511
|
|
|
} else { |
|
512
|
|
|
$urlList = 'ERROR - no URL generated'; |
|
513
|
|
|
} |
|
514
|
|
|
|
|
515
|
4 |
|
return $urlList; |
|
516
|
|
|
} |
|
517
|
|
|
|
|
518
|
|
|
/** |
|
519
|
|
|
* Returns true if input processing instruction is among registered ones. |
|
520
|
|
|
* |
|
521
|
|
|
* @param string $piString PI to test |
|
522
|
|
|
* @param array $incomingProcInstructions Processing instructions |
|
523
|
|
|
* @return boolean |
|
524
|
|
|
*/ |
|
525
|
5 |
|
public function drawURLs_PIfilter($piString, array $incomingProcInstructions) |
|
526
|
|
|
{ |
|
527
|
5 |
|
if (empty($incomingProcInstructions)) { |
|
528
|
1 |
|
return true; |
|
529
|
|
|
} |
|
530
|
|
|
|
|
531
|
4 |
|
foreach ($incomingProcInstructions as $pi) { |
|
532
|
4 |
|
if (GeneralUtility::inList($piString, $pi)) { |
|
533
|
4 |
|
return true; |
|
534
|
|
|
} |
|
535
|
|
|
} |
|
536
|
2 |
|
} |
|
537
|
|
|
|
|
538
|
4 |
|
public function getPageTSconfigForId($id) |
|
539
|
|
|
{ |
|
540
|
4 |
|
if (!$this->MP) { |
|
541
|
4 |
|
$pageTSconfig = BackendUtility::getPagesTSconfig($id); |
|
542
|
|
|
} else { |
|
543
|
|
|
list(, $mountPointId) = explode('-', $this->MP); |
|
544
|
|
|
$pageTSconfig = BackendUtility::getPagesTSconfig($mountPointId); |
|
545
|
|
|
} |
|
546
|
|
|
|
|
547
|
|
|
// Call a hook to alter configuration |
|
548
|
4 |
|
if (is_array($GLOBALS['TYPO3_CONF_VARS']['EXTCONF']['crawler']['getPageTSconfigForId'])) { |
|
549
|
|
|
$params = [ |
|
550
|
|
|
'pageId' => $id, |
|
551
|
|
|
'pageTSConfig' => &$pageTSconfig |
|
552
|
|
|
]; |
|
553
|
|
|
foreach ($GLOBALS['TYPO3_CONF_VARS']['EXTCONF']['crawler']['getPageTSconfigForId'] as $userFunc) { |
|
554
|
|
|
GeneralUtility::callUserFunction($userFunc, $params, $this); |
|
555
|
|
|
} |
|
556
|
|
|
} |
|
557
|
|
|
|
|
558
|
4 |
|
return $pageTSconfig; |
|
559
|
|
|
} |
|
560
|
|
|
|
|
561
|
|
|
/** |
|
562
|
|
|
* This methods returns an array of configurations. |
|
563
|
|
|
* And no urls! |
|
564
|
|
|
* |
|
565
|
|
|
* @param integer $id Page ID |
|
566
|
|
|
* @param bool $forceSsl Use https |
|
567
|
|
|
* @return array |
|
568
|
|
|
*/ |
|
569
|
4 |
|
protected function getUrlsForPageId($id, $forceSsl = false) |
|
570
|
|
|
{ |
|
571
|
|
|
|
|
572
|
|
|
/** |
|
573
|
|
|
* Get configuration from tsConfig |
|
574
|
|
|
*/ |
|
575
|
|
|
|
|
576
|
|
|
// Get page TSconfig for page ID: |
|
577
|
4 |
|
$pageTSconfig = $this->getPageTSconfigForId($id); |
|
578
|
|
|
|
|
579
|
4 |
|
$res = []; |
|
580
|
|
|
|
|
581
|
4 |
|
if (is_array($pageTSconfig) && is_array($pageTSconfig['tx_crawler.']['crawlerCfg.'])) { |
|
582
|
3 |
|
$crawlerCfg = $pageTSconfig['tx_crawler.']['crawlerCfg.']; |
|
583
|
|
|
|
|
584
|
3 |
|
if (is_array($crawlerCfg['paramSets.'])) { |
|
585
|
3 |
|
foreach ($crawlerCfg['paramSets.'] as $key => $values) { |
|
586
|
3 |
|
if (is_array($values)) { |
|
587
|
3 |
|
$key = str_replace('.', '', $key); |
|
588
|
|
|
// Sub configuration for a single configuration string: |
|
589
|
3 |
|
$subCfg = (array)$crawlerCfg['paramSets.'][$key . '.']; |
|
590
|
3 |
|
$subCfg['key'] = $key; |
|
591
|
|
|
|
|
592
|
3 |
|
if (strcmp($subCfg['procInstrFilter'], '')) { |
|
593
|
3 |
|
$subCfg['procInstrFilter'] = implode(',', GeneralUtility::trimExplode(',', $subCfg['procInstrFilter'])); |
|
594
|
|
|
} |
|
595
|
3 |
|
$pidOnlyList = implode(',', GeneralUtility::trimExplode(',', $subCfg['pidsOnly'], true)); |
|
596
|
|
|
|
|
597
|
|
|
// process configuration if it is not page-specific or if the specific page is the current page: |
|
598
|
3 |
|
if (!strcmp($subCfg['pidsOnly'], '') || GeneralUtility::inList($pidOnlyList, $id)) { |
|
599
|
|
|
|
|
600
|
|
|
// add trailing slash if not present |
|
601
|
3 |
|
if (!empty($subCfg['baseUrl']) && substr($subCfg['baseUrl'], -1) != '/') { |
|
602
|
|
|
$subCfg['baseUrl'] .= '/'; |
|
603
|
|
|
} |
|
604
|
|
|
|
|
605
|
|
|
// Explode, process etc.: |
|
606
|
3 |
|
$res[$key] = []; |
|
607
|
3 |
|
$res[$key]['subCfg'] = $subCfg; |
|
608
|
3 |
|
$res[$key]['paramParsed'] = $this->parseParams($values); |
|
|
|
|
|
|
609
|
3 |
|
$res[$key]['paramExpanded'] = $this->expandParameters($res[$key]['paramParsed'], $id); |
|
610
|
3 |
|
$res[$key]['origin'] = 'pagets'; |
|
611
|
|
|
|
|
612
|
|
|
// recognize MP value |
|
613
|
3 |
|
if (!$this->MP) { |
|
614
|
3 |
|
$res[$key]['URLs'] = $this->compileUrls($res[$key]['paramExpanded'], ['?id=' . $id]); |
|
615
|
|
|
} else { |
|
616
|
3 |
|
$res[$key]['URLs'] = $this->compileUrls($res[$key]['paramExpanded'], ['?id=' . $id . '&MP=' . $this->MP]); |
|
617
|
|
|
} |
|
618
|
|
|
} |
|
619
|
|
|
} |
|
620
|
|
|
} |
|
621
|
|
|
} |
|
622
|
|
|
} |
|
623
|
|
|
|
|
624
|
|
|
/** |
|
625
|
|
|
* Get configuration from tx_crawler_configuration records |
|
626
|
|
|
*/ |
|
627
|
|
|
|
|
628
|
|
|
// get records along the rootline |
|
629
|
4 |
|
$rootLine = BackendUtility::BEgetRootLine($id); |
|
630
|
|
|
|
|
631
|
4 |
|
foreach ($rootLine as $page) { |
|
632
|
4 |
|
$configurationRecordsForCurrentPage = BackendUtility::getRecordsByField( |
|
633
|
4 |
|
'tx_crawler_configuration', |
|
634
|
4 |
|
'pid', |
|
635
|
4 |
|
intval($page['uid']), |
|
636
|
4 |
|
BackendUtility::BEenableFields('tx_crawler_configuration') . BackendUtility::deleteClause('tx_crawler_configuration') |
|
637
|
|
|
); |
|
638
|
|
|
|
|
639
|
4 |
|
if (is_array($configurationRecordsForCurrentPage)) { |
|
640
|
1 |
|
foreach ($configurationRecordsForCurrentPage as $configurationRecord) { |
|
641
|
|
|
|
|
642
|
|
|
// check access to the configuration record |
|
643
|
1 |
|
if (empty($configurationRecord['begroups']) || $GLOBALS['BE_USER']->isAdmin() || $this->hasGroupAccess($GLOBALS['BE_USER']->user['usergroup_cached_list'], $configurationRecord['begroups'])) { |
|
644
|
1 |
|
$pidOnlyList = implode(',', GeneralUtility::trimExplode(',', $configurationRecord['pidsonly'], true)); |
|
645
|
|
|
|
|
646
|
|
|
// process configuration if it is not page-specific or if the specific page is the current page: |
|
647
|
1 |
|
if (!strcmp($configurationRecord['pidsonly'], '') || GeneralUtility::inList($pidOnlyList, $id)) { |
|
648
|
1 |
|
$key = $configurationRecord['name']; |
|
649
|
|
|
|
|
650
|
|
|
// don't overwrite previously defined paramSets |
|
651
|
1 |
|
if (!isset($res[$key])) { |
|
652
|
|
|
|
|
653
|
|
|
/* @var $TSparserObject \TYPO3\CMS\Core\TypoScript\Parser\TypoScriptParser */ |
|
654
|
1 |
|
$TSparserObject = GeneralUtility::makeInstance('TYPO3\CMS\Core\TypoScript\Parser\TypoScriptParser'); |
|
655
|
1 |
|
$TSparserObject->parse($configurationRecord['processing_instruction_parameters_ts']); |
|
656
|
|
|
|
|
657
|
1 |
|
$isCrawlingProtocolHttps = $this->isCrawlingProtocolHttps($configurationRecord['force_ssl'], $forceSsl); |
|
658
|
|
|
|
|
659
|
|
|
$subCfg = [ |
|
660
|
1 |
|
'procInstrFilter' => $configurationRecord['processing_instruction_filter'], |
|
661
|
1 |
|
'procInstrParams.' => $TSparserObject->setup, |
|
662
|
1 |
|
'baseUrl' => $this->getBaseUrlForConfigurationRecord( |
|
663
|
1 |
|
$configurationRecord['base_url'], |
|
664
|
1 |
|
$configurationRecord['sys_domain_base_url'], |
|
665
|
1 |
|
$isCrawlingProtocolHttps |
|
666
|
|
|
), |
|
667
|
1 |
|
'realurl' => $configurationRecord['realurl'], |
|
668
|
1 |
|
'cHash' => $configurationRecord['chash'], |
|
669
|
1 |
|
'userGroups' => $configurationRecord['fegroups'], |
|
670
|
1 |
|
'exclude' => $configurationRecord['exclude'], |
|
671
|
1 |
|
'rootTemplatePid' => (int) $configurationRecord['root_template_pid'], |
|
672
|
1 |
|
'key' => $key |
|
673
|
|
|
]; |
|
674
|
|
|
|
|
675
|
|
|
// add trailing slash if not present |
|
676
|
1 |
|
if (!empty($subCfg['baseUrl']) && substr($subCfg['baseUrl'], -1) != '/') { |
|
677
|
|
|
$subCfg['baseUrl'] .= '/'; |
|
678
|
|
|
} |
|
679
|
1 |
|
if (!in_array($id, $this->expandExcludeString($subCfg['exclude']))) { |
|
680
|
1 |
|
$res[$key] = []; |
|
681
|
1 |
|
$res[$key]['subCfg'] = $subCfg; |
|
682
|
1 |
|
$res[$key]['paramParsed'] = $this->parseParams($configurationRecord['configuration']); |
|
683
|
1 |
|
$res[$key]['paramExpanded'] = $this->expandParameters($res[$key]['paramParsed'], $id); |
|
684
|
1 |
|
$res[$key]['URLs'] = $this->compileUrls($res[$key]['paramExpanded'], ['?id=' . $id]); |
|
685
|
4 |
|
$res[$key]['origin'] = 'tx_crawler_configuration_' . $configurationRecord['uid']; |
|
686
|
|
|
} |
|
687
|
|
|
} |
|
688
|
|
|
} |
|
689
|
|
|
} |
|
690
|
|
|
} |
|
691
|
|
|
} |
|
692
|
|
|
} |
|
693
|
|
|
|
|
694
|
4 |
|
if (is_array($GLOBALS['TYPO3_CONF_VARS']['EXTCONF']['crawler']['processUrls'])) { |
|
695
|
|
|
foreach ($GLOBALS['TYPO3_CONF_VARS']['EXTCONF']['crawler']['processUrls'] as $func) { |
|
696
|
|
|
$params = [ |
|
697
|
|
|
'res' => &$res, |
|
698
|
|
|
]; |
|
699
|
|
|
GeneralUtility::callUserFunction($func, $params, $this); |
|
700
|
|
|
} |
|
701
|
|
|
} |
|
702
|
|
|
|
|
703
|
4 |
|
return $res; |
|
704
|
|
|
} |
|
705
|
|
|
|
|
706
|
|
|
/** |
|
707
|
|
|
* Checks if a domain record exist and returns the base-url based on the record. If not the given baseUrl string is used. |
|
708
|
|
|
* |
|
709
|
|
|
* @param string $baseUrl |
|
710
|
|
|
* @param integer $sysDomainUid |
|
711
|
|
|
* @param bool $ssl |
|
712
|
|
|
* @return string |
|
713
|
|
|
*/ |
|
714
|
4 |
|
protected function getBaseUrlForConfigurationRecord($baseUrl, $sysDomainUid, $ssl = false) |
|
715
|
|
|
{ |
|
716
|
4 |
|
$sysDomainUid = intval($sysDomainUid); |
|
717
|
4 |
|
$urlScheme = ($ssl === false) ? 'http' : 'https'; |
|
718
|
|
|
|
|
719
|
4 |
|
if ($sysDomainUid > 0) { |
|
720
|
2 |
|
$res = $this->db->exec_SELECTquery( |
|
721
|
2 |
|
'*', |
|
722
|
2 |
|
'sys_domain', |
|
723
|
2 |
|
'uid = ' . $sysDomainUid . |
|
724
|
2 |
|
BackendUtility::BEenableFields('sys_domain') . |
|
725
|
2 |
|
BackendUtility::deleteClause('sys_domain') |
|
726
|
|
|
); |
|
727
|
2 |
|
$row = $this->db->sql_fetch_assoc($res); |
|
728
|
2 |
|
if ($row['domainName'] != '') { |
|
729
|
1 |
|
return $urlScheme . '://' . $row['domainName']; |
|
730
|
|
|
} |
|
731
|
|
|
} |
|
732
|
3 |
|
return $baseUrl; |
|
733
|
|
|
} |
|
734
|
|
|
|
|
735
|
|
|
public function getConfigurationsForBranch($rootid, $depth) |
|
736
|
|
|
{ |
|
737
|
|
|
$configurationsForBranch = []; |
|
738
|
|
|
|
|
739
|
|
|
$pageTSconfig = $this->getPageTSconfigForId($rootid); |
|
740
|
|
|
if (is_array($pageTSconfig) && is_array($pageTSconfig['tx_crawler.']['crawlerCfg.']) && is_array($pageTSconfig['tx_crawler.']['crawlerCfg.']['paramSets.'])) { |
|
741
|
|
|
$sets = $pageTSconfig['tx_crawler.']['crawlerCfg.']['paramSets.']; |
|
742
|
|
|
if (is_array($sets)) { |
|
743
|
|
|
foreach ($sets as $key => $value) { |
|
744
|
|
|
if (!is_array($value)) { |
|
745
|
|
|
continue; |
|
746
|
|
|
} |
|
747
|
|
|
$configurationsForBranch[] = substr($key, -1) == '.' ? substr($key, 0, -1) : $key; |
|
748
|
|
|
} |
|
749
|
|
|
} |
|
750
|
|
|
} |
|
751
|
|
|
$pids = []; |
|
752
|
|
|
$rootLine = BackendUtility::BEgetRootLine($rootid); |
|
753
|
|
|
foreach ($rootLine as $node) { |
|
754
|
|
|
$pids[] = $node['uid']; |
|
755
|
|
|
} |
|
756
|
|
|
/* @var PageTreeView $tree */ |
|
757
|
|
|
$tree = GeneralUtility::makeInstance(PageTreeView::class); |
|
758
|
|
|
$perms_clause = $GLOBALS['BE_USER']->getPagePermsClause(1); |
|
759
|
|
|
$tree->init('AND ' . $perms_clause); |
|
760
|
|
|
$tree->getTree($rootid, $depth, ''); |
|
761
|
|
|
foreach ($tree->tree as $node) { |
|
762
|
|
|
$pids[] = $node['row']['uid']; |
|
763
|
|
|
} |
|
764
|
|
|
|
|
765
|
|
|
$res = $this->db->exec_SELECTquery( |
|
766
|
|
|
'*', |
|
767
|
|
|
'tx_crawler_configuration', |
|
768
|
|
|
'pid IN (' . implode(',', $pids) . ') ' . |
|
769
|
|
|
BackendUtility::BEenableFields('tx_crawler_configuration') . |
|
770
|
|
|
BackendUtility::deleteClause('tx_crawler_configuration') . ' ' . |
|
771
|
|
|
BackendUtility::versioningPlaceholderClause('tx_crawler_configuration') . ' ' |
|
772
|
|
|
); |
|
773
|
|
|
|
|
774
|
|
|
while ($row = $this->db->sql_fetch_assoc($res)) { |
|
775
|
|
|
$configurationsForBranch[] = $row['name']; |
|
776
|
|
|
} |
|
777
|
|
|
$this->db->sql_free_result($res); |
|
778
|
|
|
return $configurationsForBranch; |
|
779
|
|
|
} |
|
780
|
|
|
|
|
781
|
|
|
/** |
|
782
|
|
|
* Check if a user has access to an item |
|
783
|
|
|
* (e.g. get the group list of the current logged in user from $GLOBALS['TSFE']->gr_list) |
|
784
|
|
|
* |
|
785
|
|
|
* @see \TYPO3\CMS\Frontend\Page\PageRepository::getMultipleGroupsWhereClause() |
|
786
|
|
|
* @param string $groupList Comma-separated list of (fe_)group UIDs from a user |
|
787
|
|
|
* @param string $accessList Comma-separated list of (fe_)group UIDs of the item to access |
|
788
|
|
|
* @return bool TRUE if at least one of the users group UIDs is in the access list or the access list is empty |
|
789
|
|
|
*/ |
|
790
|
3 |
|
public function hasGroupAccess($groupList, $accessList) |
|
791
|
|
|
{ |
|
792
|
3 |
|
if (empty($accessList)) { |
|
793
|
1 |
|
return true; |
|
794
|
|
|
} |
|
795
|
2 |
|
foreach (GeneralUtility::intExplode(',', $groupList) as $groupUid) { |
|
796
|
2 |
|
if (GeneralUtility::inList($accessList, $groupUid)) { |
|
797
|
2 |
|
return true; |
|
798
|
|
|
} |
|
799
|
|
|
} |
|
800
|
1 |
|
return false; |
|
801
|
|
|
} |
|
802
|
|
|
|
|
803
|
|
|
/** |
|
804
|
|
|
* Parse GET vars of input Query into array with key=>value pairs |
|
805
|
|
|
* |
|
806
|
|
|
* @param string $inputQuery Input query string |
|
807
|
|
|
* @return array |
|
808
|
|
|
*/ |
|
809
|
7 |
|
public function parseParams($inputQuery) |
|
810
|
|
|
{ |
|
811
|
|
|
// Extract all GET parameters into an ARRAY: |
|
812
|
7 |
|
$paramKeyValues = []; |
|
813
|
7 |
|
$GETparams = explode('&', $inputQuery); |
|
814
|
|
|
|
|
815
|
7 |
|
foreach ($GETparams as $paramAndValue) { |
|
816
|
4 |
|
list($p, $v) = explode('=', $paramAndValue, 2); |
|
817
|
4 |
|
if (strlen($p)) { |
|
818
|
4 |
|
$paramKeyValues[rawurldecode($p)] = rawurldecode($v); |
|
819
|
|
|
} |
|
820
|
|
|
} |
|
821
|
|
|
|
|
822
|
7 |
|
return $paramKeyValues; |
|
823
|
|
|
} |
|
824
|
|
|
|
|
825
|
|
|
/** |
|
826
|
|
|
* Will expand the parameters configuration to individual values. This follows a certain syntax of the value of each parameter. |
|
827
|
|
|
* Syntax of values: |
|
828
|
|
|
* - Basically: If the value is wrapped in [...] it will be expanded according to the following syntax, otherwise the value is taken literally |
|
829
|
|
|
* - Configuration is splitted by "|" and the parts are processed individually and finally added together |
|
830
|
|
|
* - For each configuration part: |
|
831
|
|
|
* - "[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" |
|
832
|
|
|
* - "_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" |
|
833
|
|
|
* _ENABLELANG:1 picks only original records without their language overlays |
|
834
|
|
|
* - Default: Literal value |
|
835
|
|
|
* |
|
836
|
|
|
* @param array $paramArray Array with key (GET var name) and values (value of GET var which is configuration for expansion) |
|
837
|
|
|
* @param integer $pid Current page ID |
|
838
|
|
|
* @return array |
|
839
|
|
|
*/ |
|
840
|
4 |
|
public function expandParameters($paramArray, $pid) |
|
841
|
|
|
{ |
|
842
|
4 |
|
global $TCA; |
|
843
|
|
|
|
|
844
|
|
|
// Traverse parameter names: |
|
845
|
4 |
|
foreach ($paramArray as $p => $v) { |
|
846
|
1 |
|
$v = trim($v); |
|
847
|
|
|
|
|
848
|
|
|
// If value is encapsulated in square brackets it means there are some ranges of values to find, otherwise the value is literal |
|
849
|
1 |
|
if (substr($v, 0, 1) === '[' && substr($v, -1) === ']') { |
|
850
|
|
|
// So, find the value inside brackets and reset the paramArray value as an array. |
|
851
|
1 |
|
$v = substr($v, 1, -1); |
|
852
|
1 |
|
$paramArray[$p] = []; |
|
853
|
|
|
|
|
854
|
|
|
// Explode parts and traverse them: |
|
855
|
1 |
|
$parts = explode('|', $v); |
|
856
|
1 |
|
foreach ($parts as $pV) { |
|
857
|
|
|
|
|
858
|
|
|
// Look for integer range: (fx. 1-34 or -40--30 // reads minus 40 to minus 30) |
|
859
|
1 |
|
if (preg_match('/^(-?[0-9]+)\s*-\s*(-?[0-9]+)$/', trim($pV), $reg)) { |
|
860
|
|
|
|
|
861
|
|
|
// Swap if first is larger than last: |
|
862
|
|
|
if ($reg[1] > $reg[2]) { |
|
863
|
|
|
$temp = $reg[2]; |
|
864
|
|
|
$reg[2] = $reg[1]; |
|
865
|
|
|
$reg[1] = $temp; |
|
866
|
|
|
} |
|
867
|
|
|
|
|
868
|
|
|
// Traverse range, add values: |
|
869
|
|
|
$runAwayBrake = 1000; // Limit to size of range! |
|
870
|
|
|
for ($a = $reg[1]; $a <= $reg[2];$a++) { |
|
871
|
|
|
$paramArray[$p][] = $a; |
|
872
|
|
|
$runAwayBrake--; |
|
873
|
|
|
if ($runAwayBrake <= 0) { |
|
874
|
|
|
break; |
|
875
|
|
|
} |
|
876
|
|
|
} |
|
877
|
1 |
|
} elseif (substr(trim($pV), 0, 7) == '_TABLE:') { |
|
878
|
|
|
|
|
879
|
|
|
// Parse parameters: |
|
880
|
|
|
$subparts = GeneralUtility::trimExplode(';', $pV); |
|
881
|
|
|
$subpartParams = []; |
|
882
|
|
|
foreach ($subparts as $spV) { |
|
883
|
|
|
list($pKey, $pVal) = GeneralUtility::trimExplode(':', $spV); |
|
884
|
|
|
$subpartParams[$pKey] = $pVal; |
|
885
|
|
|
} |
|
886
|
|
|
|
|
887
|
|
|
// Table exists: |
|
888
|
|
|
if (isset($TCA[$subpartParams['_TABLE']])) { |
|
889
|
|
|
$lookUpPid = isset($subpartParams['_PID']) ? intval($subpartParams['_PID']) : $pid; |
|
890
|
|
|
$pidField = isset($subpartParams['_PIDFIELD']) ? trim($subpartParams['_PIDFIELD']) : 'pid'; |
|
891
|
|
|
$where = isset($subpartParams['_WHERE']) ? $subpartParams['_WHERE'] : ''; |
|
892
|
|
|
$addTable = isset($subpartParams['_ADDTABLE']) ? $subpartParams['_ADDTABLE'] : ''; |
|
893
|
|
|
|
|
894
|
|
|
$fieldName = $subpartParams['_FIELD'] ? $subpartParams['_FIELD'] : 'uid'; |
|
895
|
|
|
if ($fieldName === 'uid' || $TCA[$subpartParams['_TABLE']]['columns'][$fieldName]) { |
|
896
|
|
|
$andWhereLanguage = ''; |
|
897
|
|
|
$transOrigPointerField = $TCA[$subpartParams['_TABLE']]['ctrl']['transOrigPointerField']; |
|
898
|
|
|
|
|
899
|
|
|
if ($subpartParams['_ENABLELANG'] && $transOrigPointerField) { |
|
900
|
|
|
$andWhereLanguage = ' AND ' . $this->db->quoteStr($transOrigPointerField, $subpartParams['_TABLE']) . ' <= 0 '; |
|
901
|
|
|
} |
|
902
|
|
|
|
|
903
|
|
|
$where = $this->db->quoteStr($pidField, $subpartParams['_TABLE']) . '=' . intval($lookUpPid) . ' ' . |
|
904
|
|
|
$andWhereLanguage . $where; |
|
905
|
|
|
|
|
906
|
|
|
$rows = $this->db->exec_SELECTgetRows( |
|
907
|
|
|
$fieldName, |
|
908
|
|
|
$subpartParams['_TABLE'] . $addTable, |
|
909
|
|
|
$where . BackendUtility::deleteClause($subpartParams['_TABLE']), |
|
910
|
|
|
'', |
|
911
|
|
|
'', |
|
912
|
|
|
'', |
|
913
|
|
|
$fieldName |
|
914
|
|
|
); |
|
915
|
|
|
|
|
916
|
|
|
if (is_array($rows)) { |
|
917
|
|
|
$paramArray[$p] = array_merge($paramArray[$p], array_keys($rows)); |
|
918
|
|
|
} |
|
919
|
|
|
} |
|
920
|
|
|
} |
|
921
|
|
|
} else { // Just add value: |
|
922
|
1 |
|
$paramArray[$p][] = $pV; |
|
923
|
|
|
} |
|
924
|
|
|
// Hook for processing own expandParameters place holder |
|
925
|
1 |
|
if (is_array($GLOBALS['TYPO3_CONF_VARS']['SC_OPTIONS']['crawler/class.tx_crawler_lib.php']['expandParameters'])) { |
|
926
|
|
|
$_params = [ |
|
927
|
|
|
'pObj' => &$this, |
|
928
|
|
|
'paramArray' => &$paramArray, |
|
929
|
|
|
'currentKey' => $p, |
|
930
|
|
|
'currentValue' => $pV, |
|
931
|
|
|
'pid' => $pid |
|
932
|
|
|
]; |
|
933
|
|
|
foreach ($GLOBALS['TYPO3_CONF_VARS']['SC_OPTIONS']['crawler/class.tx_crawler_lib.php']['expandParameters'] as $key => $_funcRef) { |
|
934
|
1 |
|
GeneralUtility::callUserFunction($_funcRef, $_params, $this); |
|
935
|
|
|
} |
|
936
|
|
|
} |
|
937
|
|
|
} |
|
938
|
|
|
|
|
939
|
|
|
// Make unique set of values and sort array by key: |
|
940
|
1 |
|
$paramArray[$p] = array_unique($paramArray[$p]); |
|
941
|
1 |
|
ksort($paramArray); |
|
942
|
|
|
} else { |
|
943
|
|
|
// Set the literal value as only value in array: |
|
944
|
1 |
|
$paramArray[$p] = [$v]; |
|
945
|
|
|
} |
|
946
|
|
|
} |
|
947
|
|
|
|
|
948
|
4 |
|
return $paramArray; |
|
949
|
|
|
} |
|
950
|
|
|
|
|
951
|
|
|
/** |
|
952
|
|
|
* Compiling URLs from parameter array (output of expandParameters()) |
|
953
|
|
|
* The number of URLs will be the multiplication of the number of parameter values for each key |
|
954
|
|
|
* |
|
955
|
|
|
* @param array $paramArray Output of expandParameters(): Array with keys (GET var names) and for each an array of values |
|
956
|
|
|
* @param array $urls URLs accumulated in this array (for recursion) |
|
957
|
|
|
* @return array |
|
958
|
|
|
*/ |
|
959
|
7 |
|
public function compileUrls($paramArray, $urls = []) |
|
960
|
|
|
{ |
|
961
|
7 |
|
if (count($paramArray) && is_array($urls)) { |
|
962
|
|
|
// shift first off stack: |
|
963
|
3 |
|
reset($paramArray); |
|
964
|
3 |
|
$varName = key($paramArray); |
|
965
|
3 |
|
$valueSet = array_shift($paramArray); |
|
966
|
|
|
|
|
967
|
|
|
// Traverse value set: |
|
968
|
3 |
|
$newUrls = []; |
|
969
|
3 |
|
foreach ($urls as $url) { |
|
970
|
2 |
|
foreach ($valueSet as $val) { |
|
971
|
2 |
|
$newUrls[] = $url . (strcmp($val, '') ? '&' . rawurlencode($varName) . '=' . rawurlencode($val) : ''); |
|
972
|
|
|
|
|
973
|
2 |
|
if (count($newUrls) > MathUtility::forceIntegerInRange($this->extensionSettings['maxCompileUrls'], 1, 1000000000, 10000)) { |
|
974
|
2 |
|
break; |
|
975
|
|
|
} |
|
976
|
|
|
} |
|
977
|
|
|
} |
|
978
|
3 |
|
$urls = $newUrls; |
|
979
|
3 |
|
$urls = $this->compileUrls($paramArray, $urls); |
|
980
|
|
|
} |
|
981
|
|
|
|
|
982
|
7 |
|
return $urls; |
|
983
|
|
|
} |
|
984
|
|
|
|
|
985
|
|
|
/************************************ |
|
986
|
|
|
* |
|
987
|
|
|
* Crawler log |
|
988
|
|
|
* |
|
989
|
|
|
************************************/ |
|
990
|
|
|
|
|
991
|
|
|
/** |
|
992
|
|
|
* Return array of records from crawler queue for input page ID |
|
993
|
|
|
* |
|
994
|
|
|
* @param integer $id Page ID for which to look up log entries. |
|
995
|
|
|
* @param string$filter Filter: "all" => all entries, "pending" => all that is not yet run, "finished" => all complete ones |
|
996
|
|
|
* @param boolean $doFlush If TRUE, then entries selected at DELETED(!) instead of selected! |
|
997
|
|
|
* @param boolean $doFullFlush |
|
998
|
|
|
* @param integer $itemsPerPage Limit the amount of entries per page default is 10 |
|
999
|
|
|
* @return array |
|
1000
|
|
|
*/ |
|
1001
|
4 |
|
public function getLogEntriesForPageId($id, $filter = '', $doFlush = false, $doFullFlush = false, $itemsPerPage = 10) |
|
1002
|
|
|
{ |
|
1003
|
|
|
switch ($filter) { |
|
1004
|
4 |
|
case 'pending': |
|
1005
|
|
|
$addWhere = ' AND exec_time=0'; |
|
1006
|
|
|
break; |
|
1007
|
4 |
|
case 'finished': |
|
1008
|
|
|
$addWhere = ' AND exec_time>0'; |
|
1009
|
|
|
break; |
|
1010
|
|
|
default: |
|
1011
|
4 |
|
$addWhere = ''; |
|
1012
|
4 |
|
break; |
|
1013
|
|
|
} |
|
1014
|
|
|
|
|
1015
|
|
|
// FIXME: Write unit test that ensures that the right records are deleted. |
|
1016
|
4 |
|
if ($doFlush) { |
|
1017
|
2 |
|
$this->flushQueue(($doFullFlush ? '1=1' : ('page_id=' . intval($id))) . $addWhere); |
|
1018
|
2 |
|
return []; |
|
1019
|
|
|
} else { |
|
1020
|
2 |
|
return $this->db->exec_SELECTgetRows( |
|
1021
|
2 |
|
'*', |
|
1022
|
2 |
|
'tx_crawler_queue', |
|
1023
|
2 |
|
'page_id=' . intval($id) . $addWhere, |
|
1024
|
2 |
|
'', |
|
1025
|
2 |
|
'scheduled DESC', |
|
1026
|
2 |
|
(intval($itemsPerPage) > 0 ? intval($itemsPerPage) : '') |
|
1027
|
|
|
); |
|
1028
|
|
|
} |
|
1029
|
|
|
} |
|
1030
|
|
|
|
|
1031
|
|
|
/** |
|
1032
|
|
|
* Return array of records from crawler queue for input set ID |
|
1033
|
|
|
* |
|
1034
|
|
|
* @param integer $set_id Set ID for which to look up log entries. |
|
1035
|
|
|
* @param string $filter Filter: "all" => all entries, "pending" => all that is not yet run, "finished" => all complete ones |
|
1036
|
|
|
* @param boolean $doFlush If TRUE, then entries selected at DELETED(!) instead of selected! |
|
1037
|
|
|
* @param integer $itemsPerPage Limit the amount of entires per page default is 10 |
|
1038
|
|
|
* @return array |
|
1039
|
|
|
*/ |
|
1040
|
6 |
|
public function getLogEntriesForSetId($set_id, $filter = '', $doFlush = false, $doFullFlush = false, $itemsPerPage = 10) |
|
1041
|
|
|
{ |
|
1042
|
|
|
// FIXME: Write Unit tests for Filters |
|
1043
|
|
|
switch ($filter) { |
|
1044
|
6 |
|
case 'pending': |
|
1045
|
1 |
|
$addWhere = ' AND exec_time=0'; |
|
1046
|
1 |
|
break; |
|
1047
|
5 |
|
case 'finished': |
|
1048
|
1 |
|
$addWhere = ' AND exec_time>0'; |
|
1049
|
1 |
|
break; |
|
1050
|
|
|
default: |
|
1051
|
4 |
|
$addWhere = ''; |
|
1052
|
4 |
|
break; |
|
1053
|
|
|
} |
|
1054
|
|
|
// FIXME: Write unit test that ensures that the right records are deleted. |
|
1055
|
6 |
|
if ($doFlush) { |
|
1056
|
4 |
|
$this->flushQueue($doFullFlush ? '' : ('set_id=' . intval($set_id) . $addWhere)); |
|
1057
|
4 |
|
return []; |
|
1058
|
|
|
} else { |
|
1059
|
2 |
|
return $this->db->exec_SELECTgetRows( |
|
1060
|
2 |
|
'*', |
|
1061
|
2 |
|
'tx_crawler_queue', |
|
1062
|
2 |
|
'set_id=' . intval($set_id) . $addWhere, |
|
1063
|
2 |
|
'', |
|
1064
|
2 |
|
'scheduled DESC', |
|
1065
|
2 |
|
(intval($itemsPerPage) > 0 ? intval($itemsPerPage) : '') |
|
1066
|
|
|
); |
|
1067
|
|
|
} |
|
1068
|
|
|
} |
|
1069
|
|
|
|
|
1070
|
|
|
/** |
|
1071
|
|
|
* Removes queue entries |
|
1072
|
|
|
* |
|
1073
|
|
|
* @param string $where SQL related filter for the entries which should be removed |
|
1074
|
|
|
* @return void |
|
1075
|
|
|
*/ |
|
1076
|
10 |
|
protected function flushQueue($where = '') |
|
1077
|
|
|
{ |
|
1078
|
10 |
|
$realWhere = strlen($where) > 0 ? $where : '1=1'; |
|
1079
|
|
|
|
|
1080
|
10 |
|
if (EventDispatcher::getInstance()->hasObserver('queueEntryFlush')) { |
|
1081
|
|
|
$groups = $this->db->exec_SELECTgetRows('DISTINCT set_id', 'tx_crawler_queue', $realWhere); |
|
1082
|
|
|
if (is_array($groups)) { |
|
1083
|
|
|
foreach ($groups as $group) { |
|
1084
|
|
|
EventDispatcher::getInstance()->post('queueEntryFlush', $group['set_id'], $this->db->exec_SELECTgetRows('uid, set_id', 'tx_crawler_queue', $realWhere . ' AND set_id="' . $group['set_id'] . '"')); |
|
1085
|
|
|
} |
|
1086
|
|
|
} |
|
1087
|
|
|
} |
|
1088
|
|
|
|
|
1089
|
10 |
|
$this->db->exec_DELETEquery('tx_crawler_queue', $realWhere); |
|
1090
|
10 |
|
} |
|
1091
|
|
|
|
|
1092
|
|
|
/** |
|
1093
|
|
|
* Adding call back entries to log (called from hooks typically, see indexed search class "class.crawler.php" |
|
1094
|
|
|
* |
|
1095
|
|
|
* @param integer $setId Set ID |
|
1096
|
|
|
* @param array $params Parameters to pass to call back function |
|
1097
|
|
|
* @param string $callBack Call back object reference, eg. 'EXT:indexed_search/class.crawler.php:&tx_indexedsearch_crawler' |
|
1098
|
|
|
* @param integer $page_id Page ID to attach it to |
|
1099
|
|
|
* @param integer $schedule Time at which to activate |
|
1100
|
|
|
* @return void |
|
1101
|
|
|
*/ |
|
1102
|
|
|
public function addQueueEntry_callBack($setId, $params, $callBack, $page_id = 0, $schedule = 0) |
|
1103
|
|
|
{ |
|
1104
|
|
|
if (!is_array($params)) { |
|
1105
|
|
|
$params = []; |
|
1106
|
|
|
} |
|
1107
|
|
|
$params['_CALLBACKOBJ'] = $callBack; |
|
1108
|
|
|
|
|
1109
|
|
|
// Compile value array: |
|
1110
|
|
|
$fieldArray = [ |
|
1111
|
|
|
'page_id' => intval($page_id), |
|
1112
|
|
|
'parameters' => serialize($params), |
|
1113
|
|
|
'scheduled' => intval($schedule) ? intval($schedule) : $this->getCurrentTime(), |
|
1114
|
|
|
'exec_time' => 0, |
|
1115
|
|
|
'set_id' => intval($setId), |
|
1116
|
|
|
'result_data' => '', |
|
1117
|
|
|
]; |
|
1118
|
|
|
|
|
1119
|
|
|
$this->db->exec_INSERTquery('tx_crawler_queue', $fieldArray); |
|
1120
|
|
|
} |
|
1121
|
|
|
|
|
1122
|
|
|
/************************************ |
|
1123
|
|
|
* |
|
1124
|
|
|
* URL setting |
|
1125
|
|
|
* |
|
1126
|
|
|
************************************/ |
|
1127
|
|
|
|
|
1128
|
|
|
/** |
|
1129
|
|
|
* Setting a URL for crawling: |
|
1130
|
|
|
* |
|
1131
|
|
|
* @param integer $id Page ID |
|
1132
|
|
|
* @param string $url Complete URL |
|
1133
|
|
|
* @param array $subCfg Sub configuration array (from TS config) |
|
1134
|
|
|
* @param integer $tstamp Scheduled-time |
|
1135
|
|
|
* @param string $configurationHash (optional) configuration hash |
|
1136
|
|
|
* @param bool $skipInnerDuplicationCheck (optional) skip inner duplication check |
|
1137
|
|
|
* @return bool |
|
1138
|
|
|
*/ |
|
1139
|
4 |
|
public function addUrl( |
|
1140
|
|
|
$id, |
|
1141
|
|
|
$url, |
|
1142
|
|
|
array $subCfg, |
|
1143
|
|
|
$tstamp, |
|
1144
|
|
|
$configurationHash = '', |
|
1145
|
|
|
$skipInnerDuplicationCheck = false |
|
1146
|
|
|
) { |
|
1147
|
4 |
|
$urlAdded = false; |
|
1148
|
4 |
|
$rows = []; |
|
1149
|
|
|
|
|
1150
|
|
|
// Creating parameters: |
|
1151
|
|
|
$parameters = [ |
|
1152
|
4 |
|
'url' => $url |
|
1153
|
|
|
]; |
|
1154
|
|
|
|
|
1155
|
|
|
// fe user group simulation: |
|
1156
|
4 |
|
$uGs = implode(',', array_unique(GeneralUtility::intExplode(',', $subCfg['userGroups'], true))); |
|
1157
|
4 |
|
if ($uGs) { |
|
1158
|
|
|
$parameters['feUserGroupList'] = $uGs; |
|
1159
|
|
|
} |
|
1160
|
|
|
|
|
1161
|
|
|
// Setting processing instructions |
|
1162
|
4 |
|
$parameters['procInstructions'] = GeneralUtility::trimExplode(',', $subCfg['procInstrFilter']); |
|
1163
|
4 |
|
if (is_array($subCfg['procInstrParams.'])) { |
|
1164
|
4 |
|
$parameters['procInstrParams'] = $subCfg['procInstrParams.']; |
|
1165
|
|
|
} |
|
1166
|
|
|
|
|
1167
|
|
|
// Possible TypoScript Template Parents |
|
1168
|
4 |
|
$parameters['rootTemplatePid'] = $subCfg['rootTemplatePid']; |
|
1169
|
|
|
|
|
1170
|
|
|
// Compile value array: |
|
1171
|
4 |
|
$parameters_serialized = serialize($parameters); |
|
1172
|
|
|
$fieldArray = [ |
|
1173
|
4 |
|
'page_id' => intval($id), |
|
1174
|
4 |
|
'parameters' => $parameters_serialized, |
|
1175
|
4 |
|
'parameters_hash' => GeneralUtility::shortMD5($parameters_serialized), |
|
1176
|
4 |
|
'configuration_hash' => $configurationHash, |
|
1177
|
4 |
|
'scheduled' => $tstamp, |
|
1178
|
4 |
|
'exec_time' => 0, |
|
1179
|
4 |
|
'set_id' => intval($this->setID), |
|
1180
|
4 |
|
'result_data' => '', |
|
1181
|
4 |
|
'configuration' => $subCfg['key'], |
|
1182
|
|
|
]; |
|
1183
|
|
|
|
|
1184
|
4 |
|
if ($this->registerQueueEntriesInternallyOnly) { |
|
1185
|
|
|
//the entries will only be registered and not stored to the database |
|
1186
|
|
|
$this->queueEntries[] = $fieldArray; |
|
1187
|
|
|
} else { |
|
1188
|
4 |
|
if (!$skipInnerDuplicationCheck) { |
|
1189
|
|
|
// check if there is already an equal entry |
|
1190
|
4 |
|
$rows = $this->getDuplicateRowsIfExist($tstamp, $fieldArray); |
|
1191
|
|
|
} |
|
1192
|
|
|
|
|
1193
|
4 |
|
if (count($rows) == 0) { |
|
1194
|
4 |
|
$this->db->exec_INSERTquery('tx_crawler_queue', $fieldArray); |
|
1195
|
4 |
|
$uid = $this->db->sql_insert_id(); |
|
1196
|
4 |
|
$rows[] = $uid; |
|
1197
|
4 |
|
$urlAdded = true; |
|
1198
|
4 |
|
EventDispatcher::getInstance()->post('urlAddedToQueue', $this->setID, ['uid' => $uid, 'fieldArray' => $fieldArray]); |
|
1199
|
|
|
} else { |
|
1200
|
2 |
|
EventDispatcher::getInstance()->post('duplicateUrlInQueue', $this->setID, ['rows' => $rows, 'fieldArray' => $fieldArray]); |
|
1201
|
|
|
} |
|
1202
|
|
|
} |
|
1203
|
|
|
|
|
1204
|
4 |
|
return $urlAdded; |
|
1205
|
|
|
} |
|
1206
|
|
|
|
|
1207
|
|
|
/** |
|
1208
|
|
|
* This method determines duplicates for a queue entry with the same parameters and this timestamp. |
|
1209
|
|
|
* If the timestamp is in the past, it will check if there is any unprocessed queue entry in the past. |
|
1210
|
|
|
* If the timestamp is in the future it will check, if the queued entry has exactly the same timestamp |
|
1211
|
|
|
* |
|
1212
|
|
|
* @param int $tstamp |
|
1213
|
|
|
* @param array $fieldArray |
|
1214
|
|
|
* |
|
1215
|
|
|
* @return array |
|
1216
|
|
|
*/ |
|
1217
|
4 |
|
protected function getDuplicateRowsIfExist($tstamp, $fieldArray) |
|
1218
|
|
|
{ |
|
1219
|
4 |
|
$rows = []; |
|
1220
|
|
|
|
|
1221
|
4 |
|
$currentTime = $this->getCurrentTime(); |
|
1222
|
|
|
|
|
1223
|
|
|
//if this entry is scheduled with "now" |
|
1224
|
4 |
|
if ($tstamp <= $currentTime) { |
|
1225
|
1 |
|
if ($this->extensionSettings['enableTimeslot']) { |
|
1226
|
1 |
|
$timeBegin = $currentTime - 100; |
|
1227
|
1 |
|
$timeEnd = $currentTime + 100; |
|
1228
|
1 |
|
$where = ' ((scheduled BETWEEN ' . $timeBegin . ' AND ' . $timeEnd . ' ) OR scheduled <= ' . $currentTime . ') '; |
|
1229
|
|
|
} else { |
|
1230
|
1 |
|
$where = 'scheduled <= ' . $currentTime; |
|
1231
|
|
|
} |
|
1232
|
3 |
|
} elseif ($tstamp > $currentTime) { |
|
1233
|
|
|
//entry with a timestamp in the future need to have the same schedule time |
|
1234
|
3 |
|
$where = 'scheduled = ' . $tstamp ; |
|
1235
|
|
|
} |
|
1236
|
|
|
|
|
1237
|
4 |
|
if (!empty($where)) { |
|
1238
|
4 |
|
$result = $this->db->exec_SELECTgetRows( |
|
1239
|
4 |
|
'qid', |
|
1240
|
4 |
|
'tx_crawler_queue', |
|
1241
|
|
|
$where . |
|
1242
|
4 |
|
' AND NOT exec_time' . |
|
1243
|
4 |
|
' AND NOT process_id ' . |
|
1244
|
4 |
|
' AND page_id=' . intval($fieldArray['page_id']) . |
|
1245
|
4 |
|
' AND parameters_hash = ' . $this->db->fullQuoteStr($fieldArray['parameters_hash'], 'tx_crawler_queue') |
|
1246
|
|
|
); |
|
1247
|
|
|
|
|
1248
|
4 |
|
if (is_array($result)) { |
|
1249
|
4 |
|
foreach ($result as $value) { |
|
1250
|
2 |
|
$rows[] = $value['qid']; |
|
1251
|
|
|
} |
|
1252
|
|
|
} |
|
1253
|
|
|
} |
|
1254
|
|
|
|
|
1255
|
4 |
|
return $rows; |
|
1256
|
|
|
} |
|
1257
|
|
|
|
|
1258
|
|
|
/** |
|
1259
|
|
|
* Returns the current system time |
|
1260
|
|
|
* |
|
1261
|
|
|
* @return int |
|
1262
|
|
|
*/ |
|
1263
|
1 |
|
public function getCurrentTime() |
|
1264
|
|
|
{ |
|
1265
|
1 |
|
return time(); |
|
1266
|
|
|
} |
|
1267
|
|
|
|
|
1268
|
|
|
/************************************ |
|
1269
|
|
|
* |
|
1270
|
|
|
* URL reading |
|
1271
|
|
|
* |
|
1272
|
|
|
************************************/ |
|
1273
|
|
|
|
|
1274
|
|
|
/** |
|
1275
|
|
|
* Read URL for single queue entry |
|
1276
|
|
|
* |
|
1277
|
|
|
* @param integer $queueId |
|
1278
|
|
|
* @param boolean $force If set, will process even if exec_time has been set! |
|
1279
|
|
|
* @return integer |
|
1280
|
|
|
*/ |
|
1281
|
|
|
public function readUrl($queueId, $force = false) |
|
1282
|
|
|
{ |
|
1283
|
|
|
$ret = 0; |
|
1284
|
|
|
if ($this->debugMode) { |
|
1285
|
|
|
GeneralUtility::devlog('crawler-readurl start ' . microtime(true), __FUNCTION__); |
|
1286
|
|
|
} |
|
1287
|
|
|
// Get entry: |
|
1288
|
|
|
list($queueRec) = $this->db->exec_SELECTgetRows( |
|
1289
|
|
|
'*', |
|
1290
|
|
|
'tx_crawler_queue', |
|
1291
|
|
|
'qid=' . intval($queueId) . ($force ? '' : ' AND exec_time=0 AND process_scheduled > 0') |
|
1292
|
|
|
); |
|
1293
|
|
|
|
|
1294
|
|
|
if (!is_array($queueRec)) { |
|
1295
|
|
|
return; |
|
1296
|
|
|
} |
|
1297
|
|
|
|
|
1298
|
|
|
$parameters = unserialize($queueRec['parameters']); |
|
1299
|
|
|
if ($parameters['rootTemplatePid']) { |
|
1300
|
|
|
$this->initTSFE((int)$parameters['rootTemplatePid']); |
|
1301
|
|
|
} else { |
|
1302
|
|
|
GeneralUtility::sysLog( |
|
1303
|
|
|
'Page with (' . $queueRec['page_id'] . ') could not be crawled, please check your crawler configuration. Perhaps no Root Template Pid is set', |
|
1304
|
|
|
'crawler', |
|
1305
|
|
|
GeneralUtility::SYSLOG_SEVERITY_WARNING |
|
1306
|
|
|
); |
|
1307
|
|
|
} |
|
1308
|
|
|
|
|
1309
|
|
|
SignalSlotUtility::emitSignal( |
|
1310
|
|
|
__CLASS__, |
|
1311
|
|
|
SignalSlotUtility::SIGNNAL_QUEUEITEM_PREPROCESS, |
|
1312
|
|
|
[$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
|
|
|
$this->db->exec_UPDATEquery('tx_crawler_queue', 'qid=' . intval($queueId), $field_array); |
|
1323
|
|
|
|
|
1324
|
|
|
$result = $this->readUrl_exec($queueRec); |
|
1325
|
|
|
$resultData = unserialize($result['content']); |
|
1326
|
|
|
|
|
1327
|
|
|
//atm there's no need to point to specific pollable extensions |
|
1328
|
|
|
if (is_array($GLOBALS['TYPO3_CONF_VARS']['EXTCONF']['crawler']['pollSuccess'])) { |
|
1329
|
|
|
foreach ($GLOBALS['TYPO3_CONF_VARS']['EXTCONF']['crawler']['pollSuccess'] as $pollable) { |
|
1330
|
|
|
// only check the success value if the instruction is runnig |
|
1331
|
|
|
// it is important to name the pollSuccess key same as the procInstructions key |
|
1332
|
|
|
if (is_array($resultData['parameters']['procInstructions']) && in_array( |
|
1333
|
|
|
$pollable, |
|
1334
|
|
|
$resultData['parameters']['procInstructions'] |
|
1335
|
|
|
) |
|
1336
|
|
|
) { |
|
1337
|
|
|
if (!empty($resultData['success'][$pollable]) && $resultData['success'][$pollable]) { |
|
1338
|
|
|
$ret |= self::CLI_STATUS_POLLABLE_PROCESSED; |
|
1339
|
|
|
} |
|
1340
|
|
|
} |
|
1341
|
|
|
} |
|
1342
|
|
|
} |
|
1343
|
|
|
|
|
1344
|
|
|
// Set result in log which also denotes the end of the processing of this entry. |
|
1345
|
|
|
$field_array = ['result_data' => serialize($result)]; |
|
1346
|
|
|
|
|
1347
|
|
|
SignalSlotUtility::emitSignal( |
|
1348
|
|
|
__CLASS__, |
|
1349
|
|
|
SignalSlotUtility::SIGNNAL_QUEUEITEM_POSTPROCESS, |
|
1350
|
|
|
[$queueId, &$field_array] |
|
1351
|
|
|
); |
|
1352
|
|
|
|
|
1353
|
|
|
$this->db->exec_UPDATEquery('tx_crawler_queue', 'qid=' . intval($queueId), $field_array); |
|
1354
|
|
|
|
|
1355
|
|
|
if ($this->debugMode) { |
|
1356
|
|
|
GeneralUtility::devlog('crawler-readurl stop ' . microtime(true), __FUNCTION__); |
|
1357
|
|
|
} |
|
1358
|
|
|
|
|
1359
|
|
|
return $ret; |
|
1360
|
|
|
} |
|
1361
|
|
|
|
|
1362
|
|
|
/** |
|
1363
|
|
|
* Read URL for not-yet-inserted log-entry |
|
1364
|
|
|
* |
|
1365
|
|
|
* @param array $field_array Queue field array, |
|
1366
|
|
|
* |
|
1367
|
|
|
* @return string |
|
1368
|
|
|
*/ |
|
1369
|
|
|
public function readUrlFromArray($field_array) |
|
1370
|
|
|
{ |
|
1371
|
|
|
|
|
1372
|
|
|
// Set exec_time to lock record: |
|
1373
|
|
|
$field_array['exec_time'] = $this->getCurrentTime(); |
|
1374
|
|
|
$this->db->exec_INSERTquery('tx_crawler_queue', $field_array); |
|
1375
|
|
|
$queueId = $field_array['qid'] = $this->db->sql_insert_id(); |
|
1376
|
|
|
|
|
1377
|
|
|
$result = $this->readUrl_exec($field_array); |
|
1378
|
|
|
|
|
1379
|
|
|
// Set result in log which also denotes the end of the processing of this entry. |
|
1380
|
|
|
$field_array = ['result_data' => serialize($result)]; |
|
1381
|
|
|
|
|
1382
|
|
|
SignalSlotUtility::emitSignal( |
|
1383
|
|
|
__CLASS__, |
|
1384
|
|
|
SignalSlotUtility::SIGNNAL_QUEUEITEM_POSTPROCESS, |
|
1385
|
|
|
[$queueId, &$field_array] |
|
1386
|
|
|
); |
|
1387
|
|
|
|
|
1388
|
|
|
$this->db->exec_UPDATEquery('tx_crawler_queue', 'qid=' . intval($queueId), $field_array); |
|
1389
|
|
|
|
|
1390
|
|
|
return $result; |
|
1391
|
|
|
} |
|
1392
|
|
|
|
|
1393
|
|
|
/** |
|
1394
|
|
|
* Read URL for a queue record |
|
1395
|
|
|
* |
|
1396
|
|
|
* @param array $queueRec Queue record |
|
1397
|
|
|
* @return string |
|
1398
|
|
|
*/ |
|
1399
|
|
|
public function readUrl_exec($queueRec) |
|
1400
|
|
|
{ |
|
1401
|
|
|
// Decode parameters: |
|
1402
|
|
|
$parameters = unserialize($queueRec['parameters']); |
|
1403
|
|
|
$result = 'ERROR'; |
|
1404
|
|
|
if (is_array($parameters)) { |
|
1405
|
|
|
if ($parameters['_CALLBACKOBJ']) { // Calling object: |
|
1406
|
|
|
$objRef = $parameters['_CALLBACKOBJ']; |
|
1407
|
|
|
$callBackObj = &GeneralUtility::getUserObj($objRef); |
|
1408
|
|
|
if (is_object($callBackObj)) { |
|
1409
|
|
|
unset($parameters['_CALLBACKOBJ']); |
|
1410
|
|
|
$result = ['content' => serialize($callBackObj->crawler_execute($parameters, $this))]; |
|
1411
|
|
|
} else { |
|
1412
|
|
|
$result = ['content' => 'No object: ' . $objRef]; |
|
1413
|
|
|
} |
|
1414
|
|
|
} else { // Regular FE request: |
|
1415
|
|
|
|
|
1416
|
|
|
// Prepare: |
|
1417
|
|
|
$crawlerId = $queueRec['qid'] . ':' . md5($queueRec['qid'] . '|' . $queueRec['set_id'] . '|' . $GLOBALS['TYPO3_CONF_VARS']['SYS']['encryptionKey']); |
|
1418
|
|
|
|
|
1419
|
|
|
// Get result: |
|
1420
|
|
|
$result = $this->requestUrl($parameters['url'], $crawlerId); |
|
1421
|
|
|
|
|
1422
|
|
|
EventDispatcher::getInstance()->post('urlCrawled', $queueRec['set_id'], ['url' => $parameters['url'], 'result' => $result]); |
|
1423
|
|
|
} |
|
1424
|
|
|
} |
|
1425
|
|
|
|
|
1426
|
|
|
return $result; |
|
1427
|
|
|
} |
|
1428
|
|
|
|
|
1429
|
|
|
/** |
|
1430
|
|
|
* Gets the content of a URL. |
|
1431
|
|
|
* |
|
1432
|
|
|
* @param string $originalUrl URL to read |
|
1433
|
|
|
* @param string $crawlerId Crawler ID string (qid + hash to verify) |
|
1434
|
|
|
* @param integer $timeout Timeout time |
|
1435
|
|
|
* @param integer $recursion Recursion limiter for 302 redirects |
|
1436
|
|
|
* @return array |
|
1437
|
|
|
*/ |
|
1438
|
2 |
|
public function requestUrl($originalUrl, $crawlerId, $timeout = 2, $recursion = 10) |
|
1439
|
|
|
{ |
|
1440
|
2 |
|
if (!$recursion) { |
|
1441
|
|
|
return false; |
|
1442
|
|
|
} |
|
1443
|
|
|
|
|
1444
|
|
|
// Parse URL, checking for scheme: |
|
1445
|
2 |
|
$url = parse_url($originalUrl); |
|
1446
|
|
|
|
|
1447
|
2 |
|
if ($url === false) { |
|
1448
|
|
|
if (TYPO3_DLOG) { |
|
1449
|
|
|
GeneralUtility::devLog(sprintf('Could not parse_url() for string "%s"', $url), 'crawler', 4, ['crawlerId' => $crawlerId]); |
|
1450
|
|
|
} |
|
1451
|
|
|
return false; |
|
1452
|
|
|
} |
|
1453
|
|
|
|
|
1454
|
2 |
|
if (!in_array($url['scheme'], ['','http','https'])) { |
|
1455
|
|
|
if (TYPO3_DLOG) { |
|
1456
|
|
|
GeneralUtility::devLog(sprintf('Scheme does not match for url "%s"', $url), 'crawler', 4, ['crawlerId' => $crawlerId]); |
|
1457
|
|
|
} |
|
1458
|
|
|
return false; |
|
1459
|
|
|
} |
|
1460
|
|
|
|
|
1461
|
|
|
// direct request |
|
1462
|
2 |
|
if ($this->extensionSettings['makeDirectRequests']) { |
|
1463
|
2 |
|
$result = $this->sendDirectRequest($originalUrl, $crawlerId); |
|
1464
|
2 |
|
return $result; |
|
1465
|
|
|
} |
|
1466
|
|
|
|
|
1467
|
|
|
$reqHeaders = $this->buildRequestHeaderArray($url, $crawlerId); |
|
1468
|
|
|
|
|
1469
|
|
|
// thanks to Pierrick Caillon for adding proxy support |
|
1470
|
|
|
$rurl = $url; |
|
1471
|
|
|
|
|
1472
|
|
|
if ($GLOBALS['TYPO3_CONF_VARS']['SYS']['curlUse'] && $GLOBALS['TYPO3_CONF_VARS']['SYS']['curlProxyServer']) { |
|
1473
|
|
|
$rurl = parse_url($GLOBALS['TYPO3_CONF_VARS']['SYS']['curlProxyServer']); |
|
1474
|
|
|
$url['path'] = $url['scheme'] . '://' . $url['host'] . ($url['port'] > 0 ? ':' . $url['port'] : '') . $url['path']; |
|
1475
|
|
|
$reqHeaders = $this->buildRequestHeaderArray($url, $crawlerId); |
|
1476
|
|
|
} |
|
1477
|
|
|
|
|
1478
|
|
|
$host = $rurl['host']; |
|
1479
|
|
|
|
|
1480
|
|
|
if ($url['scheme'] == 'https') { |
|
1481
|
|
|
$host = 'ssl://' . $host; |
|
1482
|
|
|
$port = ($rurl['port'] > 0) ? $rurl['port'] : 443; |
|
1483
|
|
|
} else { |
|
1484
|
|
|
$port = ($rurl['port'] > 0) ? $rurl['port'] : 80; |
|
1485
|
|
|
} |
|
1486
|
|
|
|
|
1487
|
|
|
$startTime = microtime(true); |
|
1488
|
|
|
$fp = fsockopen($host, $port, $errno, $errstr, $timeout); |
|
1489
|
|
|
|
|
1490
|
|
|
if (!$fp) { |
|
1491
|
|
|
if (TYPO3_DLOG) { |
|
1492
|
|
|
GeneralUtility::devLog(sprintf('Error while opening "%s"', $url), 'crawler', 4, ['crawlerId' => $crawlerId]); |
|
1493
|
|
|
} |
|
1494
|
|
|
return false; |
|
1495
|
|
|
} else { |
|
1496
|
|
|
// Request message: |
|
1497
|
|
|
$msg = implode("\r\n", $reqHeaders) . "\r\n\r\n"; |
|
1498
|
|
|
fputs($fp, $msg); |
|
1499
|
|
|
|
|
1500
|
|
|
// Read response: |
|
1501
|
|
|
$d = $this->getHttpResponseFromStream($fp); |
|
1502
|
|
|
fclose($fp); |
|
1503
|
|
|
|
|
1504
|
|
|
$time = microtime(true) - $startTime; |
|
1505
|
|
|
$this->log($originalUrl . ' ' . $time); |
|
1506
|
|
|
|
|
1507
|
|
|
// Implode content and headers: |
|
1508
|
|
|
$result = [ |
|
1509
|
|
|
'request' => $msg, |
|
1510
|
|
|
'headers' => implode('', $d['headers']), |
|
1511
|
|
|
'content' => implode('', (array)$d['content']) |
|
1512
|
|
|
]; |
|
1513
|
|
|
|
|
1514
|
|
|
if (($this->extensionSettings['follow30x']) && ($newUrl = $this->getRequestUrlFrom302Header($d['headers'], $url['user'], $url['pass']))) { |
|
1515
|
|
|
$result = array_merge(['parentRequest' => $result], $this->requestUrl($newUrl, $crawlerId, $recursion--)); |
|
|
|
|
|
|
1516
|
|
|
$newRequestUrl = $this->requestUrl($newUrl, $crawlerId, $timeout, --$recursion); |
|
|
|
|
|
|
1517
|
|
|
|
|
1518
|
|
|
if (is_array($newRequestUrl)) { |
|
1519
|
|
|
$result = array_merge(['parentRequest' => $result], $newRequestUrl); |
|
1520
|
|
|
} else { |
|
1521
|
|
|
if (TYPO3_DLOG) { |
|
1522
|
|
|
GeneralUtility::devLog(sprintf('Error while opening "%s"', $url), 'crawler', 4, ['crawlerId' => $crawlerId]); |
|
1523
|
|
|
} |
|
1524
|
|
|
return false; |
|
1525
|
|
|
} |
|
1526
|
|
|
} |
|
1527
|
|
|
|
|
1528
|
|
|
return $result; |
|
1529
|
|
|
} |
|
1530
|
|
|
} |
|
1531
|
|
|
|
|
1532
|
|
|
/** |
|
1533
|
|
|
* Gets the base path of the website frontend. |
|
1534
|
|
|
* (e.g. if you call http://mydomain.com/cms/index.php in |
|
1535
|
|
|
* the browser the base path is "/cms/") |
|
1536
|
|
|
* |
|
1537
|
|
|
* @return string Base path of the website frontend |
|
1538
|
|
|
*/ |
|
1539
|
|
|
protected function getFrontendBasePath() |
|
1540
|
|
|
{ |
|
1541
|
|
|
$frontendBasePath = '/'; |
|
1542
|
|
|
|
|
1543
|
|
|
// Get the path from the extension settings: |
|
1544
|
|
|
if (isset($this->extensionSettings['frontendBasePath']) && $this->extensionSettings['frontendBasePath']) { |
|
1545
|
|
|
$frontendBasePath = $this->extensionSettings['frontendBasePath']; |
|
1546
|
|
|
// If empty, try to use config.absRefPrefix: |
|
1547
|
|
|
} elseif (isset($GLOBALS['TSFE']->absRefPrefix) && !empty($GLOBALS['TSFE']->absRefPrefix)) { |
|
1548
|
|
|
$frontendBasePath = $GLOBALS['TSFE']->absRefPrefix; |
|
1549
|
|
|
// If not in CLI mode the base path can be determined from $_SERVER environment: |
|
1550
|
|
|
} elseif (!defined('TYPO3_REQUESTTYPE_CLI') || !TYPO3_REQUESTTYPE_CLI) { |
|
1551
|
|
|
$frontendBasePath = GeneralUtility::getIndpEnv('TYPO3_SITE_PATH'); |
|
1552
|
|
|
} |
|
1553
|
|
|
|
|
1554
|
|
|
// Base path must be '/<pathSegements>/': |
|
1555
|
|
|
if ($frontendBasePath != '/') { |
|
1556
|
|
|
$frontendBasePath = '/' . ltrim($frontendBasePath, '/'); |
|
1557
|
|
|
$frontendBasePath = rtrim($frontendBasePath, '/') . '/'; |
|
1558
|
|
|
} |
|
1559
|
|
|
|
|
1560
|
|
|
return $frontendBasePath; |
|
1561
|
|
|
} |
|
1562
|
|
|
|
|
1563
|
|
|
/** |
|
1564
|
|
|
* Executes a shell command and returns the outputted result. |
|
1565
|
|
|
* |
|
1566
|
|
|
* @param string $command Shell command to be executed |
|
1567
|
|
|
* @return string Outputted result of the command execution |
|
1568
|
|
|
*/ |
|
1569
|
|
|
protected function executeShellCommand($command) |
|
1570
|
|
|
{ |
|
1571
|
|
|
$result = shell_exec($command); |
|
1572
|
|
|
return $result; |
|
1573
|
|
|
} |
|
1574
|
|
|
|
|
1575
|
|
|
/** |
|
1576
|
|
|
* Reads HTTP response from the given stream. |
|
1577
|
|
|
* |
|
1578
|
|
|
* @param resource $streamPointer Pointer to connection stream. |
|
1579
|
|
|
* @return array Associative array with the following items: |
|
1580
|
|
|
* headers <array> Response headers sent by server. |
|
1581
|
|
|
* content <array> Content, with each line as an array item. |
|
1582
|
|
|
*/ |
|
1583
|
1 |
|
protected function getHttpResponseFromStream($streamPointer) |
|
1584
|
|
|
{ |
|
1585
|
1 |
|
$response = ['headers' => [], 'content' => []]; |
|
1586
|
|
|
|
|
1587
|
1 |
|
if (is_resource($streamPointer)) { |
|
1588
|
|
|
// read headers |
|
1589
|
1 |
|
while ($line = fgets($streamPointer, '2048')) { |
|
1590
|
1 |
|
$line = trim($line); |
|
1591
|
1 |
|
if ($line !== '') { |
|
1592
|
1 |
|
$response['headers'][] = $line; |
|
1593
|
|
|
} else { |
|
1594
|
1 |
|
break; |
|
1595
|
|
|
} |
|
1596
|
|
|
} |
|
1597
|
|
|
|
|
1598
|
|
|
// read content |
|
1599
|
1 |
|
while ($line = fgets($streamPointer, '2048')) { |
|
1600
|
1 |
|
$response['content'][] = $line; |
|
1601
|
|
|
} |
|
1602
|
|
|
} |
|
1603
|
|
|
|
|
1604
|
1 |
|
return $response; |
|
1605
|
|
|
} |
|
1606
|
|
|
|
|
1607
|
|
|
/** |
|
1608
|
|
|
* @param message |
|
1609
|
|
|
*/ |
|
1610
|
2 |
|
protected function log($message) |
|
1611
|
|
|
{ |
|
1612
|
2 |
|
if (!empty($this->extensionSettings['logFileName'])) { |
|
1613
|
|
|
$fileResult = @file_put_contents($this->extensionSettings['logFileName'], date('Ymd His') . ' ' . $message . PHP_EOL, FILE_APPEND); |
|
1614
|
|
|
if (!$fileResult) { |
|
1615
|
|
|
GeneralUtility::devLog('File "' . $this->extensionSettings['logFileName'] . '" could not be written, please check file permissions.', 'crawler', LogLevel::INFO); |
|
1616
|
|
|
} |
|
1617
|
|
|
} |
|
1618
|
2 |
|
} |
|
1619
|
|
|
|
|
1620
|
|
|
/** |
|
1621
|
|
|
* Builds HTTP request headers. |
|
1622
|
|
|
* |
|
1623
|
|
|
* @param array $url |
|
1624
|
|
|
* @param string $crawlerId |
|
1625
|
|
|
* |
|
1626
|
|
|
* @return array |
|
1627
|
|
|
*/ |
|
1628
|
6 |
|
protected function buildRequestHeaderArray(array $url, $crawlerId) |
|
1629
|
|
|
{ |
|
1630
|
6 |
|
$reqHeaders = []; |
|
1631
|
6 |
|
$reqHeaders[] = 'GET ' . $url['path'] . ($url['query'] ? '?' . $url['query'] : '') . ' HTTP/1.0'; |
|
1632
|
6 |
|
$reqHeaders[] = 'Host: ' . $url['host']; |
|
1633
|
6 |
|
if (stristr($url['query'], 'ADMCMD_previewWS')) { |
|
1634
|
2 |
|
$reqHeaders[] = 'Cookie: $Version="1"; be_typo_user="1"; $Path=/'; |
|
1635
|
|
|
} |
|
1636
|
6 |
|
$reqHeaders[] = 'Connection: close'; |
|
1637
|
6 |
|
if ($url['user'] != '') { |
|
1638
|
2 |
|
$reqHeaders[] = 'Authorization: Basic ' . base64_encode($url['user'] . ':' . $url['pass']); |
|
1639
|
|
|
} |
|
1640
|
6 |
|
$reqHeaders[] = 'X-T3crawler: ' . $crawlerId; |
|
1641
|
6 |
|
$reqHeaders[] = 'User-Agent: TYPO3 crawler'; |
|
1642
|
6 |
|
return $reqHeaders; |
|
1643
|
|
|
} |
|
1644
|
|
|
|
|
1645
|
|
|
/** |
|
1646
|
|
|
* Check if the submitted HTTP-Header contains a redirect location and built new crawler-url |
|
1647
|
|
|
* |
|
1648
|
|
|
* @param array $headers HTTP Header |
|
1649
|
|
|
* @param string $user HTTP Auth. User |
|
1650
|
|
|
* @param string $pass HTTP Auth. Password |
|
1651
|
|
|
* @return bool|string |
|
1652
|
|
|
*/ |
|
1653
|
12 |
|
protected function getRequestUrlFrom302Header($headers, $user = '', $pass = '') |
|
1654
|
|
|
{ |
|
1655
|
12 |
|
$header = []; |
|
1656
|
12 |
|
if (!is_array($headers)) { |
|
1657
|
1 |
|
return false; |
|
1658
|
|
|
} |
|
1659
|
11 |
|
if (!(stristr($headers[0], '301 Moved') || stristr($headers[0], '302 Found') || stristr($headers[0], '302 Moved'))) { |
|
1660
|
2 |
|
return false; |
|
1661
|
|
|
} |
|
1662
|
|
|
|
|
1663
|
9 |
|
foreach ($headers as $hl) { |
|
1664
|
9 |
|
$tmp = explode(": ", $hl); |
|
1665
|
9 |
|
$header[trim($tmp[0])] = trim($tmp[1]); |
|
1666
|
9 |
|
if (trim($tmp[0]) == 'Location') { |
|
1667
|
9 |
|
break; |
|
1668
|
|
|
} |
|
1669
|
|
|
} |
|
1670
|
9 |
|
if (!array_key_exists('Location', $header)) { |
|
1671
|
3 |
|
return false; |
|
1672
|
|
|
} |
|
1673
|
|
|
|
|
1674
|
6 |
|
if ($user != '') { |
|
1675
|
3 |
|
if (!($tmp = parse_url($header['Location']))) { |
|
1676
|
1 |
|
return false; |
|
1677
|
|
|
} |
|
1678
|
2 |
|
$newUrl = $tmp['scheme'] . '://' . $user . ':' . $pass . '@' . $tmp['host'] . $tmp['path']; |
|
1679
|
2 |
|
if ($tmp['query'] != '') { |
|
1680
|
2 |
|
$newUrl .= '?' . $tmp['query']; |
|
1681
|
|
|
} |
|
1682
|
|
|
} else { |
|
1683
|
3 |
|
$newUrl = $header['Location']; |
|
1684
|
|
|
} |
|
1685
|
5 |
|
return $newUrl; |
|
1686
|
|
|
} |
|
1687
|
|
|
|
|
1688
|
|
|
/************************** |
|
1689
|
|
|
* |
|
1690
|
|
|
* tslib_fe hooks: |
|
1691
|
|
|
* |
|
1692
|
|
|
**************************/ |
|
1693
|
|
|
|
|
1694
|
|
|
/** |
|
1695
|
|
|
* Initialization hook (called after database connection) |
|
1696
|
|
|
* Takes the "HTTP_X_T3CRAWLER" header and looks up queue record and verifies if the session comes from the system (by comparing hashes) |
|
1697
|
|
|
* |
|
1698
|
|
|
* @param array $params Parameters from frontend |
|
1699
|
|
|
* @param object $ref TSFE object (reference under PHP5) |
|
1700
|
|
|
* @return void |
|
1701
|
|
|
* |
|
1702
|
|
|
* FIXME: Look like this is not used, in commit 9910d3f40cce15f4e9b7bcd0488bf21f31d53ebc it's added as public, |
|
1703
|
|
|
* FIXME: I think this can be removed. (TNM) |
|
1704
|
|
|
*/ |
|
1705
|
|
|
public function fe_init(&$params, $ref) |
|
|
|
|
|
|
1706
|
|
|
{ |
|
1707
|
|
|
// Authenticate crawler request: |
|
1708
|
|
|
if (isset($_SERVER['HTTP_X_T3CRAWLER'])) { |
|
1709
|
|
|
list($queueId, $hash) = explode(':', $_SERVER['HTTP_X_T3CRAWLER']); |
|
1710
|
|
|
list($queueRec) = $this->db->exec_SELECTgetSingleRow('*', 'tx_crawler_queue', 'qid=' . intval($queueId)); |
|
1711
|
|
|
|
|
1712
|
|
|
// If a crawler record was found and hash was matching, set it up: |
|
1713
|
|
|
if (is_array($queueRec) && $hash === md5($queueRec['qid'] . '|' . $queueRec['set_id'] . '|' . $GLOBALS['TYPO3_CONF_VARS']['SYS']['encryptionKey'])) { |
|
1714
|
|
|
$params['pObj']->applicationData['tx_crawler']['running'] = true; |
|
1715
|
|
|
$params['pObj']->applicationData['tx_crawler']['parameters'] = unserialize($queueRec['parameters']); |
|
1716
|
|
|
$params['pObj']->applicationData['tx_crawler']['log'] = []; |
|
1717
|
|
|
} else { |
|
1718
|
|
|
die('No crawler entry found!'); |
|
|
|
|
|
|
1719
|
|
|
} |
|
1720
|
|
|
} |
|
1721
|
|
|
} |
|
1722
|
|
|
|
|
1723
|
|
|
/***************************** |
|
1724
|
|
|
* |
|
1725
|
|
|
* Compiling URLs to crawl - tools |
|
1726
|
|
|
* |
|
1727
|
|
|
*****************************/ |
|
1728
|
|
|
|
|
1729
|
|
|
/** |
|
1730
|
|
|
* @param integer $id Root page id to start from. |
|
1731
|
|
|
* @param integer $depth Depth of tree, 0=only id-page, 1= on sublevel, 99 = infinite |
|
1732
|
|
|
* @param integer $scheduledTime Unix Time when the URL is timed to be visited when put in queue |
|
1733
|
|
|
* @param integer $reqMinute Number of requests per minute (creates the interleave between requests) |
|
1734
|
|
|
* @param boolean $submitCrawlUrls If set, submits the URLs to queue in database (real crawling) |
|
1735
|
|
|
* @param boolean $downloadCrawlUrls If set (and submitcrawlUrls is false) will fill $downloadUrls with entries) |
|
1736
|
|
|
* @param array $incomingProcInstructions Array of processing instructions |
|
1737
|
|
|
* @param array $configurationSelection Array of configuration keys |
|
1738
|
|
|
* @return string |
|
1739
|
|
|
*/ |
|
1740
|
|
|
public function getPageTreeAndUrls( |
|
1741
|
|
|
$id, |
|
1742
|
|
|
$depth, |
|
1743
|
|
|
$scheduledTime, |
|
1744
|
|
|
$reqMinute, |
|
1745
|
|
|
$submitCrawlUrls, |
|
1746
|
|
|
$downloadCrawlUrls, |
|
1747
|
|
|
array $incomingProcInstructions, |
|
1748
|
|
|
array $configurationSelection |
|
1749
|
|
|
) { |
|
1750
|
|
|
global $BACK_PATH; |
|
1751
|
|
|
global $LANG; |
|
1752
|
|
|
if (!is_object($LANG)) { |
|
1753
|
|
|
$LANG = GeneralUtility::makeInstance('language'); |
|
1754
|
|
|
$LANG->init(0); |
|
1755
|
|
|
} |
|
1756
|
|
|
$this->scheduledTime = $scheduledTime; |
|
1757
|
|
|
$this->reqMinute = $reqMinute; |
|
1758
|
|
|
$this->submitCrawlUrls = $submitCrawlUrls; |
|
1759
|
|
|
$this->downloadCrawlUrls = $downloadCrawlUrls; |
|
1760
|
|
|
$this->incomingProcInstructions = $incomingProcInstructions; |
|
1761
|
|
|
$this->incomingConfigurationSelection = $configurationSelection; |
|
1762
|
|
|
|
|
1763
|
|
|
$this->duplicateTrack = []; |
|
1764
|
|
|
$this->downloadUrls = []; |
|
1765
|
|
|
|
|
1766
|
|
|
// Drawing tree: |
|
1767
|
|
|
/* @var PageTreeView $tree */ |
|
1768
|
|
|
$tree = GeneralUtility::makeInstance(PageTreeView::class); |
|
1769
|
|
|
$perms_clause = $GLOBALS['BE_USER']->getPagePermsClause(1); |
|
1770
|
|
|
$tree->init('AND ' . $perms_clause); |
|
1771
|
|
|
|
|
1772
|
|
|
$pageInfo = BackendUtility::readPageAccess($id, $perms_clause); |
|
1773
|
|
|
if (is_array($pageInfo)) { |
|
1774
|
|
|
// Set root row: |
|
1775
|
|
|
$tree->tree[] = [ |
|
1776
|
|
|
'row' => $pageInfo, |
|
1777
|
|
|
'HTML' => IconUtility::getIconForRecord('pages', $pageInfo) |
|
1778
|
|
|
]; |
|
1779
|
|
|
} |
|
1780
|
|
|
|
|
1781
|
|
|
// Get branch beneath: |
|
1782
|
|
|
if ($depth) { |
|
1783
|
|
|
$tree->getTree($id, $depth, ''); |
|
1784
|
|
|
} |
|
1785
|
|
|
|
|
1786
|
|
|
// Traverse page tree: |
|
1787
|
|
|
$code = ''; |
|
1788
|
|
|
|
|
1789
|
|
|
foreach ($tree->tree as $data) { |
|
1790
|
|
|
$this->MP = false; |
|
1791
|
|
|
|
|
1792
|
|
|
// recognize mount points |
|
1793
|
|
|
if ($data['row']['doktype'] == 7) { |
|
1794
|
|
|
$mountpage = $this->db->exec_SELECTgetRows('*', 'pages', 'uid = ' . $data['row']['uid']); |
|
1795
|
|
|
|
|
1796
|
|
|
// fetch mounted pages |
|
1797
|
|
|
$this->MP = $mountpage[0]['mount_pid'] . '-' . $data['row']['uid']; |
|
|
|
|
|
|
1798
|
|
|
|
|
1799
|
|
|
$mountTree = GeneralUtility::makeInstance(PageTreeView::class); |
|
1800
|
|
|
$mountTree->init('AND ' . $perms_clause); |
|
1801
|
|
|
$mountTree->getTree($mountpage[0]['mount_pid'], $depth, ''); |
|
1802
|
|
|
|
|
1803
|
|
|
foreach ($mountTree->tree as $mountData) { |
|
1804
|
|
|
$code .= $this->drawURLs_addRowsForPage( |
|
1805
|
|
|
$mountData['row'], |
|
1806
|
|
|
$mountData['HTML'] . BackendUtility::getRecordTitle('pages', $mountData['row'], true) |
|
1807
|
|
|
); |
|
1808
|
|
|
} |
|
1809
|
|
|
|
|
1810
|
|
|
// replace page when mount_pid_ol is enabled |
|
1811
|
|
|
if ($mountpage[0]['mount_pid_ol']) { |
|
1812
|
|
|
$data['row']['uid'] = $mountpage[0]['mount_pid']; |
|
1813
|
|
|
} else { |
|
1814
|
|
|
// if the mount_pid_ol is not set the MP must not be used for the mountpoint page |
|
1815
|
|
|
$this->MP = false; |
|
1816
|
|
|
} |
|
1817
|
|
|
} |
|
1818
|
|
|
|
|
1819
|
|
|
$code .= $this->drawURLs_addRowsForPage( |
|
1820
|
|
|
$data['row'], |
|
1821
|
|
|
$data['HTML'] . BackendUtility::getRecordTitle('pages', $data['row'], true) |
|
1822
|
|
|
); |
|
1823
|
|
|
} |
|
1824
|
|
|
|
|
1825
|
|
|
return $code; |
|
1826
|
|
|
} |
|
1827
|
|
|
|
|
1828
|
|
|
/** |
|
1829
|
|
|
* Expands exclude string |
|
1830
|
|
|
* |
|
1831
|
|
|
* @param string $excludeString Exclude string |
|
1832
|
|
|
* @return array |
|
1833
|
|
|
*/ |
|
1834
|
1 |
|
public function expandExcludeString($excludeString) |
|
1835
|
|
|
{ |
|
1836
|
|
|
// internal static caches; |
|
1837
|
1 |
|
static $expandedExcludeStringCache; |
|
1838
|
1 |
|
static $treeCache; |
|
1839
|
|
|
|
|
1840
|
1 |
|
if (empty($expandedExcludeStringCache[$excludeString])) { |
|
1841
|
1 |
|
$pidList = []; |
|
1842
|
|
|
|
|
1843
|
1 |
|
if (!empty($excludeString)) { |
|
1844
|
|
|
/** @var PageTreeView $tree */ |
|
1845
|
|
|
$tree = GeneralUtility::makeInstance(PageTreeView::class); |
|
1846
|
|
|
$tree->init('AND ' . $this->backendUser->getPagePermsClause(1)); |
|
1847
|
|
|
|
|
1848
|
|
|
$excludeParts = GeneralUtility::trimExplode(',', $excludeString); |
|
1849
|
|
|
|
|
1850
|
|
|
foreach ($excludeParts as $excludePart) { |
|
1851
|
|
|
list($pid, $depth) = GeneralUtility::trimExplode('+', $excludePart); |
|
1852
|
|
|
|
|
1853
|
|
|
// default is "page only" = "depth=0" |
|
1854
|
|
|
if (empty($depth)) { |
|
1855
|
|
|
$depth = (stristr($excludePart, '+')) ? 99 : 0; |
|
1856
|
|
|
} |
|
1857
|
|
|
|
|
1858
|
|
|
$pidList[] = $pid; |
|
1859
|
|
|
|
|
1860
|
|
|
if ($depth > 0) { |
|
1861
|
|
|
if (empty($treeCache[$pid][$depth])) { |
|
1862
|
|
|
$tree->reset(); |
|
1863
|
|
|
$tree->getTree($pid, $depth); |
|
1864
|
|
|
$treeCache[$pid][$depth] = $tree->tree; |
|
1865
|
|
|
} |
|
1866
|
|
|
|
|
1867
|
|
|
foreach ($treeCache[$pid][$depth] as $data) { |
|
1868
|
|
|
$pidList[] = $data['row']['uid']; |
|
1869
|
|
|
} |
|
1870
|
|
|
} |
|
1871
|
|
|
} |
|
1872
|
|
|
} |
|
1873
|
|
|
|
|
1874
|
1 |
|
$expandedExcludeStringCache[$excludeString] = array_unique($pidList); |
|
1875
|
|
|
} |
|
1876
|
|
|
|
|
1877
|
1 |
|
return $expandedExcludeStringCache[$excludeString]; |
|
1878
|
|
|
} |
|
1879
|
|
|
|
|
1880
|
|
|
/** |
|
1881
|
|
|
* Create the rows for display of the page tree |
|
1882
|
|
|
* For each page a number of rows are shown displaying GET variable configuration |
|
1883
|
|
|
* |
|
1884
|
|
|
* @param array Page row |
|
1885
|
|
|
* @param string Page icon and title for row |
|
1886
|
|
|
* @return string HTML <tr> content (one or more) |
|
1887
|
|
|
*/ |
|
1888
|
|
|
public function drawURLs_addRowsForPage(array $pageRow, $pageTitleAndIcon) |
|
1889
|
|
|
{ |
|
1890
|
|
|
$skipMessage = ''; |
|
1891
|
|
|
|
|
1892
|
|
|
// Get list of configurations |
|
1893
|
|
|
$configurations = $this->getUrlsForPageRow($pageRow, $skipMessage); |
|
1894
|
|
|
|
|
1895
|
|
|
if (count($this->incomingConfigurationSelection) > 0) { |
|
1896
|
|
|
// remove configuration that does not match the current selection |
|
1897
|
|
|
foreach ($configurations as $confKey => $confArray) { |
|
1898
|
|
|
if (!in_array($confKey, $this->incomingConfigurationSelection)) { |
|
1899
|
|
|
unset($configurations[$confKey]); |
|
1900
|
|
|
} |
|
1901
|
|
|
} |
|
1902
|
|
|
} |
|
1903
|
|
|
|
|
1904
|
|
|
// Traverse parameter combinations: |
|
1905
|
|
|
$c = 0; |
|
1906
|
|
|
$content = ''; |
|
1907
|
|
|
if (count($configurations)) { |
|
1908
|
|
|
foreach ($configurations as $confKey => $confArray) { |
|
1909
|
|
|
|
|
1910
|
|
|
// Title column: |
|
1911
|
|
|
if (!$c) { |
|
1912
|
|
|
$titleClm = '<td rowspan="' . count($configurations) . '">' . $pageTitleAndIcon . '</td>'; |
|
1913
|
|
|
} else { |
|
1914
|
|
|
$titleClm = ''; |
|
1915
|
|
|
} |
|
1916
|
|
|
|
|
1917
|
|
|
if (!in_array($pageRow['uid'], $this->expandExcludeString($confArray['subCfg']['exclude']))) { |
|
1918
|
|
|
|
|
1919
|
|
|
// URL list: |
|
1920
|
|
|
$urlList = $this->urlListFromUrlArray( |
|
1921
|
|
|
$confArray, |
|
1922
|
|
|
$pageRow, |
|
1923
|
|
|
$this->scheduledTime, |
|
1924
|
|
|
$this->reqMinute, |
|
1925
|
|
|
$this->submitCrawlUrls, |
|
1926
|
|
|
$this->downloadCrawlUrls, |
|
1927
|
|
|
$this->duplicateTrack, |
|
1928
|
|
|
$this->downloadUrls, |
|
1929
|
|
|
$this->incomingProcInstructions // if empty the urls won't be filtered by processing instructions |
|
1930
|
|
|
); |
|
1931
|
|
|
|
|
1932
|
|
|
// Expanded parameters: |
|
1933
|
|
|
$paramExpanded = ''; |
|
1934
|
|
|
$calcAccu = []; |
|
1935
|
|
|
$calcRes = 1; |
|
1936
|
|
|
foreach ($confArray['paramExpanded'] as $gVar => $gVal) { |
|
1937
|
|
|
$paramExpanded .= ' |
|
1938
|
|
|
<tr> |
|
1939
|
|
|
<td class="bgColor4-20">' . htmlspecialchars('&' . $gVar . '=') . '<br/>' . |
|
1940
|
|
|
'(' . count($gVal) . ')' . |
|
1941
|
|
|
'</td> |
|
1942
|
|
|
<td class="bgColor4" nowrap="nowrap">' . nl2br(htmlspecialchars(implode(chr(10), $gVal))) . '</td> |
|
1943
|
|
|
</tr> |
|
1944
|
|
|
'; |
|
1945
|
|
|
$calcRes *= count($gVal); |
|
1946
|
|
|
$calcAccu[] = count($gVal); |
|
1947
|
|
|
} |
|
1948
|
|
|
$paramExpanded = '<table class="lrPadding c-list param-expanded">' . $paramExpanded . '</table>'; |
|
1949
|
|
|
$paramExpanded .= 'Comb: ' . implode('*', $calcAccu) . '=' . $calcRes; |
|
1950
|
|
|
|
|
1951
|
|
|
// Options |
|
1952
|
|
|
$optionValues = ''; |
|
1953
|
|
|
if ($confArray['subCfg']['userGroups']) { |
|
1954
|
|
|
$optionValues .= 'User Groups: ' . $confArray['subCfg']['userGroups'] . '<br/>'; |
|
1955
|
|
|
} |
|
1956
|
|
|
if ($confArray['subCfg']['baseUrl']) { |
|
1957
|
|
|
$optionValues .= 'Base Url: ' . $confArray['subCfg']['baseUrl'] . '<br/>'; |
|
1958
|
|
|
} |
|
1959
|
|
|
if ($confArray['subCfg']['procInstrFilter']) { |
|
1960
|
|
|
$optionValues .= 'ProcInstr: ' . $confArray['subCfg']['procInstrFilter'] . '<br/>'; |
|
1961
|
|
|
} |
|
1962
|
|
|
|
|
1963
|
|
|
// Compile row: |
|
1964
|
|
|
$content .= ' |
|
1965
|
|
|
<tr class="bgColor' . ($c % 2 ? '-20' : '-10') . '"> |
|
1966
|
|
|
' . $titleClm . ' |
|
1967
|
|
|
<td>' . htmlspecialchars($confKey) . '</td> |
|
1968
|
|
|
<td>' . nl2br(htmlspecialchars(rawurldecode(trim(str_replace('&', chr(10) . '&', GeneralUtility::implodeArrayForUrl('', $confArray['paramParsed'])))))) . '</td> |
|
1969
|
|
|
<td>' . $paramExpanded . '</td> |
|
1970
|
|
|
<td nowrap="nowrap">' . $urlList . '</td> |
|
1971
|
|
|
<td nowrap="nowrap">' . $optionValues . '</td> |
|
1972
|
|
|
<td nowrap="nowrap">' . DebugUtility::viewArray($confArray['subCfg']['procInstrParams.']) . '</td> |
|
1973
|
|
|
</tr>'; |
|
1974
|
|
|
} else { |
|
1975
|
|
|
$content .= '<tr class="bgColor' . ($c % 2 ? '-20' : '-10') . '"> |
|
1976
|
|
|
' . $titleClm . ' |
|
1977
|
|
|
<td>' . htmlspecialchars($confKey) . '</td> |
|
1978
|
|
|
<td colspan="5"><em>No entries</em> (Page is excluded in this configuration)</td> |
|
1979
|
|
|
</tr>'; |
|
1980
|
|
|
} |
|
1981
|
|
|
|
|
1982
|
|
|
$c++; |
|
1983
|
|
|
} |
|
1984
|
|
|
} else { |
|
1985
|
|
|
$message = !empty($skipMessage) ? ' (' . $skipMessage . ')' : ''; |
|
1986
|
|
|
|
|
1987
|
|
|
// Compile row: |
|
1988
|
|
|
$content .= ' |
|
1989
|
|
|
<tr class="bgColor-20" style="border-bottom: 1px solid black;"> |
|
1990
|
|
|
<td>' . $pageTitleAndIcon . '</td> |
|
1991
|
|
|
<td colspan="6"><em>No entries</em>' . $message . '</td> |
|
1992
|
|
|
</tr>'; |
|
1993
|
|
|
} |
|
1994
|
|
|
|
|
1995
|
|
|
return $content; |
|
1996
|
|
|
} |
|
1997
|
|
|
|
|
1998
|
|
|
/** |
|
1999
|
|
|
* @return int |
|
2000
|
|
|
*/ |
|
2001
|
1 |
|
public function getUnprocessedItemsCount() |
|
2002
|
|
|
{ |
|
2003
|
1 |
|
$res = $this->db->exec_SELECTquery( |
|
2004
|
1 |
|
'count(*) as num', |
|
2005
|
1 |
|
'tx_crawler_queue', |
|
2006
|
1 |
|
'exec_time=0 AND process_scheduled=0 AND scheduled<=' . $this->getCurrentTime() |
|
2007
|
|
|
); |
|
2008
|
|
|
|
|
2009
|
1 |
|
$count = $this->db->sql_fetch_assoc($res); |
|
2010
|
1 |
|
return $count['num']; |
|
2011
|
|
|
} |
|
2012
|
|
|
|
|
2013
|
|
|
/***************************** |
|
2014
|
|
|
* |
|
2015
|
|
|
* CLI functions |
|
2016
|
|
|
* |
|
2017
|
|
|
*****************************/ |
|
2018
|
|
|
|
|
2019
|
|
|
/** |
|
2020
|
|
|
* Main function for running from Command Line PHP script (cron job) |
|
2021
|
|
|
* See ext/crawler/cli/crawler_cli.phpsh for details |
|
2022
|
|
|
* |
|
2023
|
|
|
* @return int number of remaining items or false if error |
|
2024
|
|
|
*/ |
|
2025
|
|
|
public function CLI_main() |
|
2026
|
|
|
{ |
|
2027
|
|
|
$this->setAccessMode('cli'); |
|
2028
|
|
|
$result = self::CLI_STATUS_NOTHING_PROCCESSED; |
|
2029
|
|
|
$cliObj = GeneralUtility::makeInstance(CrawlerCommandLineController::class); |
|
2030
|
|
|
|
|
2031
|
|
|
if (isset($cliObj->cli_args['-h']) || isset($cliObj->cli_args['--help'])) { |
|
2032
|
|
|
$cliObj->cli_validateArgs(); |
|
2033
|
|
|
$cliObj->cli_help(); |
|
2034
|
|
|
exit; |
|
|
|
|
|
|
2035
|
|
|
} |
|
2036
|
|
|
|
|
2037
|
|
|
if (!$this->getDisabled() && $this->CLI_checkAndAcquireNewProcess($this->CLI_buildProcessId())) { |
|
2038
|
|
|
$countInARun = $cliObj->cli_argValue('--countInARun') ? intval($cliObj->cli_argValue('--countInARun')) : $this->extensionSettings['countInARun']; |
|
2039
|
|
|
// Seconds |
|
2040
|
|
|
$sleepAfterFinish = $cliObj->cli_argValue('--sleepAfterFinish') ? intval($cliObj->cli_argValue('--sleepAfterFinish')) : $this->extensionSettings['sleepAfterFinish']; |
|
2041
|
|
|
// Milliseconds |
|
2042
|
|
|
$sleepTime = $cliObj->cli_argValue('--sleepTime') ? intval($cliObj->cli_argValue('--sleepTime')) : $this->extensionSettings['sleepTime']; |
|
2043
|
|
|
|
|
2044
|
|
|
try { |
|
2045
|
|
|
// Run process: |
|
2046
|
|
|
$result = $this->CLI_run($countInARun, $sleepTime, $sleepAfterFinish); |
|
2047
|
|
|
} catch (\Exception $e) { |
|
2048
|
|
|
$this->CLI_debug(get_class($e) . ': ' . $e->getMessage()); |
|
2049
|
|
|
$result = self::CLI_STATUS_ABORTED; |
|
2050
|
|
|
} |
|
2051
|
|
|
|
|
2052
|
|
|
// Cleanup |
|
2053
|
|
|
$this->db->exec_DELETEquery('tx_crawler_process', 'assigned_items_count = 0'); |
|
2054
|
|
|
|
|
2055
|
|
|
//TODO can't we do that in a clean way? |
|
2056
|
|
|
$releaseStatus = $this->CLI_releaseProcesses($this->CLI_buildProcessId()); |
|
|
|
|
|
|
2057
|
|
|
|
|
2058
|
|
|
$this->CLI_debug("Unprocessed Items remaining:" . $this->getUnprocessedItemsCount() . " (" . $this->CLI_buildProcessId() . ")"); |
|
2059
|
|
|
$result |= ($this->getUnprocessedItemsCount() > 0 ? self::CLI_STATUS_REMAIN : self::CLI_STATUS_NOTHING_PROCCESSED); |
|
2060
|
|
|
} else { |
|
2061
|
|
|
$result |= self::CLI_STATUS_ABORTED; |
|
2062
|
|
|
} |
|
2063
|
|
|
|
|
2064
|
|
|
return $result; |
|
2065
|
|
|
} |
|
2066
|
|
|
|
|
2067
|
|
|
/** |
|
2068
|
|
|
* Function executed by crawler_im.php cli script. |
|
2069
|
|
|
* |
|
2070
|
|
|
* @return void |
|
2071
|
|
|
*/ |
|
2072
|
|
|
public function CLI_main_im() |
|
2073
|
|
|
{ |
|
2074
|
|
|
$this->setAccessMode('cli_im'); |
|
2075
|
|
|
|
|
2076
|
|
|
$cliObj = GeneralUtility::makeInstance(QueueCommandLineController::class); |
|
2077
|
|
|
|
|
2078
|
|
|
// Force user to admin state and set workspace to "Live": |
|
2079
|
|
|
$this->backendUser->user['admin'] = 1; |
|
2080
|
|
|
$this->backendUser->setWorkspace(0); |
|
2081
|
|
|
|
|
2082
|
|
|
// Print help |
|
2083
|
|
|
if (!isset($cliObj->cli_args['_DEFAULT'][1])) { |
|
2084
|
|
|
$cliObj->cli_validateArgs(); |
|
2085
|
|
|
$cliObj->cli_help(); |
|
2086
|
|
|
exit; |
|
|
|
|
|
|
2087
|
|
|
} |
|
2088
|
|
|
|
|
2089
|
|
|
$cliObj->cli_validateArgs(); |
|
2090
|
|
|
|
|
2091
|
|
|
if ($cliObj->cli_argValue('-o') === 'exec') { |
|
2092
|
|
|
$this->registerQueueEntriesInternallyOnly = true; |
|
2093
|
|
|
} |
|
2094
|
|
|
|
|
2095
|
|
|
if (isset($cliObj->cli_args['_DEFAULT'][2])) { |
|
2096
|
|
|
// Crawler is called over TYPO3 BE |
|
2097
|
|
|
$pageId = MathUtility::forceIntegerInRange($cliObj->cli_args['_DEFAULT'][2], 0); |
|
2098
|
|
|
} else { |
|
2099
|
|
|
// Crawler is called over cli |
|
2100
|
|
|
$pageId = MathUtility::forceIntegerInRange($cliObj->cli_args['_DEFAULT'][1], 0); |
|
2101
|
|
|
} |
|
2102
|
|
|
|
|
2103
|
|
|
$configurationKeys = $this->getConfigurationKeys($cliObj); |
|
2104
|
|
|
|
|
2105
|
|
|
if (!is_array($configurationKeys)) { |
|
2106
|
|
|
$configurations = $this->getUrlsForPageId($pageId); |
|
2107
|
|
|
if (is_array($configurations)) { |
|
2108
|
|
|
$configurationKeys = array_keys($configurations); |
|
2109
|
|
|
} else { |
|
2110
|
|
|
$configurationKeys = []; |
|
2111
|
|
|
} |
|
2112
|
|
|
} |
|
2113
|
|
|
|
|
2114
|
|
|
if ($cliObj->cli_argValue('-o') === 'queue' || $cliObj->cli_argValue('-o') === 'exec') { |
|
2115
|
|
|
$reason = new Reason(); |
|
2116
|
|
|
$reason->setReason(Reason::REASON_GUI_SUBMIT); |
|
2117
|
|
|
$reason->setDetailText('The cli script of the crawler added to the queue'); |
|
2118
|
|
|
EventDispatcher::getInstance()->post( |
|
2119
|
|
|
'invokeQueueChange', |
|
2120
|
|
|
$this->setID, |
|
2121
|
|
|
['reason' => $reason] |
|
2122
|
|
|
); |
|
2123
|
|
|
} |
|
2124
|
|
|
|
|
2125
|
|
|
if ($this->extensionSettings['cleanUpOldQueueEntries']) { |
|
2126
|
|
|
$this->cleanUpOldQueueEntries(); |
|
2127
|
|
|
} |
|
2128
|
|
|
|
|
2129
|
|
|
$this->setID = (int) GeneralUtility::md5int(microtime()); |
|
2130
|
|
|
$this->getPageTreeAndUrls( |
|
2131
|
|
|
$pageId, |
|
2132
|
|
|
MathUtility::forceIntegerInRange($cliObj->cli_argValue('-d'), 0, 99), |
|
2133
|
|
|
$this->getCurrentTime(), |
|
2134
|
|
|
MathUtility::forceIntegerInRange($cliObj->cli_isArg('-n') ? $cliObj->cli_argValue('-n') : 30, 1, 1000), |
|
2135
|
|
|
$cliObj->cli_argValue('-o') === 'queue' || $cliObj->cli_argValue('-o') === 'exec', |
|
2136
|
|
|
$cliObj->cli_argValue('-o') === 'url', |
|
2137
|
|
|
GeneralUtility::trimExplode(',', $cliObj->cli_argValue('-proc'), true), |
|
2138
|
|
|
$configurationKeys |
|
2139
|
|
|
); |
|
2140
|
|
|
|
|
2141
|
|
|
if ($cliObj->cli_argValue('-o') === 'url') { |
|
2142
|
|
|
$cliObj->cli_echo(implode(chr(10), $this->downloadUrls) . chr(10), true); |
|
2143
|
|
|
} elseif ($cliObj->cli_argValue('-o') === 'exec') { |
|
2144
|
|
|
$cliObj->cli_echo("Executing " . count($this->urlList) . " requests right away:\n\n"); |
|
2145
|
|
|
$cliObj->cli_echo(implode(chr(10), $this->urlList) . chr(10)); |
|
2146
|
|
|
$cliObj->cli_echo("\nProcessing:\n"); |
|
2147
|
|
|
|
|
2148
|
|
|
foreach ($this->queueEntries as $queueRec) { |
|
2149
|
|
|
$p = unserialize($queueRec['parameters']); |
|
2150
|
|
|
$cliObj->cli_echo($p['url'] . ' (' . implode(',', $p['procInstructions']) . ') => '); |
|
2151
|
|
|
|
|
2152
|
|
|
$result = $this->readUrlFromArray($queueRec); |
|
2153
|
|
|
|
|
2154
|
|
|
$requestResult = unserialize($result['content']); |
|
2155
|
|
|
if (is_array($requestResult)) { |
|
2156
|
|
|
$resLog = is_array($requestResult['log']) ? chr(10) . chr(9) . chr(9) . implode(chr(10) . chr(9) . chr(9), $requestResult['log']) : ''; |
|
2157
|
|
|
$cliObj->cli_echo('OK: ' . $resLog . chr(10)); |
|
2158
|
|
|
} else { |
|
2159
|
|
|
$cliObj->cli_echo('Error checking Crawler Result: ' . substr(preg_replace('/\s+/', ' ', strip_tags($result['content'])), 0, 30000) . '...' . chr(10)); |
|
2160
|
|
|
} |
|
2161
|
|
|
} |
|
2162
|
|
|
} elseif ($cliObj->cli_argValue('-o') === 'queue') { |
|
2163
|
|
|
$cliObj->cli_echo("Putting " . count($this->urlList) . " entries in queue:\n\n"); |
|
2164
|
|
|
$cliObj->cli_echo(implode(chr(10), $this->urlList) . chr(10)); |
|
2165
|
|
|
} else { |
|
2166
|
|
|
$cliObj->cli_echo(count($this->urlList) . " entries found for processing. (Use -o to decide action):\n\n", true); |
|
2167
|
|
|
$cliObj->cli_echo(implode(chr(10), $this->urlList) . chr(10), true); |
|
2168
|
|
|
} |
|
2169
|
|
|
} |
|
2170
|
|
|
|
|
2171
|
|
|
/** |
|
2172
|
|
|
* Function executed by crawler_im.php cli script. |
|
2173
|
|
|
* |
|
2174
|
|
|
* @return bool |
|
2175
|
|
|
*/ |
|
2176
|
|
|
public function CLI_main_flush() |
|
2177
|
|
|
{ |
|
2178
|
|
|
$this->setAccessMode('cli_flush'); |
|
2179
|
|
|
$cliObj = GeneralUtility::makeInstance(FlushCommandLineController::class); |
|
2180
|
|
|
|
|
2181
|
|
|
// Force user to admin state and set workspace to "Live": |
|
2182
|
|
|
$this->backendUser->user['admin'] = 1; |
|
2183
|
|
|
$this->backendUser->setWorkspace(0); |
|
2184
|
|
|
|
|
2185
|
|
|
// Print help |
|
2186
|
|
|
if (!isset($cliObj->cli_args['_DEFAULT'][1])) { |
|
2187
|
|
|
$cliObj->cli_validateArgs(); |
|
2188
|
|
|
$cliObj->cli_help(); |
|
2189
|
|
|
exit; |
|
|
|
|
|
|
2190
|
|
|
} |
|
2191
|
|
|
|
|
2192
|
|
|
$cliObj->cli_validateArgs(); |
|
2193
|
|
|
$pageId = MathUtility::forceIntegerInRange($cliObj->cli_args['_DEFAULT'][1], 0); |
|
2194
|
|
|
$fullFlush = ($pageId == 0); |
|
2195
|
|
|
|
|
2196
|
|
|
$mode = $cliObj->cli_argValue('-o'); |
|
2197
|
|
|
|
|
2198
|
|
|
switch ($mode) { |
|
2199
|
|
|
case 'all': |
|
2200
|
|
|
$result = $this->getLogEntriesForPageId($pageId, '', true, $fullFlush); |
|
2201
|
|
|
break; |
|
2202
|
|
|
case 'finished': |
|
2203
|
|
|
case 'pending': |
|
2204
|
|
|
$result = $this->getLogEntriesForPageId($pageId, $mode, true, $fullFlush); |
|
2205
|
|
|
break; |
|
2206
|
|
|
default: |
|
2207
|
|
|
$cliObj->cli_validateArgs(); |
|
2208
|
|
|
$cliObj->cli_help(); |
|
2209
|
|
|
$result = false; |
|
2210
|
|
|
} |
|
2211
|
|
|
|
|
2212
|
|
|
return $result !== false; |
|
2213
|
|
|
} |
|
2214
|
|
|
|
|
2215
|
|
|
/** |
|
2216
|
|
|
* Obtains configuration keys from the CLI arguments |
|
2217
|
|
|
* |
|
2218
|
|
|
* @param QueueCommandLineController $cliObj Command line object |
|
2219
|
|
|
* @return mixed Array of keys or null if no keys found |
|
2220
|
|
|
*/ |
|
2221
|
|
|
protected function getConfigurationKeys(QueueCommandLineController &$cliObj) |
|
2222
|
|
|
{ |
|
2223
|
|
|
$parameter = trim($cliObj->cli_argValue('-conf')); |
|
2224
|
|
|
return ($parameter != '' ? GeneralUtility::trimExplode(',', $parameter) : []); |
|
2225
|
|
|
} |
|
2226
|
|
|
|
|
2227
|
|
|
/** |
|
2228
|
|
|
* Running the functionality of the CLI (crawling URLs from queue) |
|
2229
|
|
|
* |
|
2230
|
|
|
* @param int $countInARun |
|
2231
|
|
|
* @param int $sleepTime |
|
2232
|
|
|
* @param int $sleepAfterFinish |
|
2233
|
|
|
* @return string |
|
2234
|
|
|
*/ |
|
2235
|
|
|
public function CLI_run($countInARun, $sleepTime, $sleepAfterFinish) |
|
2236
|
|
|
{ |
|
2237
|
|
|
$result = 0; |
|
2238
|
|
|
$counter = 0; |
|
2239
|
|
|
|
|
2240
|
|
|
// First, run hooks: |
|
2241
|
|
|
$this->CLI_runHooks(); |
|
2242
|
|
|
|
|
2243
|
|
|
// Clean up the queue |
|
2244
|
|
|
if (intval($this->extensionSettings['purgeQueueDays']) > 0) { |
|
2245
|
|
|
$purgeDate = $this->getCurrentTime() - 24 * 60 * 60 * intval($this->extensionSettings['purgeQueueDays']); |
|
2246
|
|
|
$del = $this->db->exec_DELETEquery( |
|
2247
|
|
|
'tx_crawler_queue', |
|
2248
|
|
|
'exec_time!=0 AND exec_time<' . $purgeDate |
|
2249
|
|
|
); |
|
2250
|
|
|
if (false == $del) { |
|
2251
|
|
|
GeneralUtility::devLog('Records could not be deleted.', 'crawler', LogLevel::INFO); |
|
2252
|
|
|
} |
|
2253
|
|
|
} |
|
2254
|
|
|
|
|
2255
|
|
|
// Select entries: |
|
2256
|
|
|
//TODO Shouldn't this reside within the transaction? |
|
2257
|
|
|
$rows = $this->db->exec_SELECTgetRows( |
|
2258
|
|
|
'qid,scheduled', |
|
2259
|
|
|
'tx_crawler_queue', |
|
2260
|
|
|
'exec_time=0 |
|
2261
|
|
|
AND process_scheduled= 0 |
|
2262
|
|
|
AND scheduled<=' . $this->getCurrentTime(), |
|
2263
|
|
|
'', |
|
2264
|
|
|
'scheduled, qid', |
|
2265
|
|
|
intval($countInARun) |
|
2266
|
|
|
); |
|
2267
|
|
|
|
|
2268
|
|
|
if (count($rows) > 0) { |
|
2269
|
|
|
$quidList = []; |
|
2270
|
|
|
|
|
2271
|
|
|
foreach ($rows as $r) { |
|
|
|
|
|
|
2272
|
|
|
$quidList[] = $r['qid']; |
|
2273
|
|
|
} |
|
2274
|
|
|
|
|
2275
|
|
|
$processId = $this->CLI_buildProcessId(); |
|
2276
|
|
|
|
|
2277
|
|
|
//reserve queue entries for process |
|
2278
|
|
|
$this->db->sql_query('BEGIN'); |
|
2279
|
|
|
//TODO make sure we're not taking assigned queue-entires |
|
2280
|
|
|
$this->db->exec_UPDATEquery( |
|
2281
|
|
|
'tx_crawler_queue', |
|
2282
|
|
|
'qid IN (' . implode(',', $quidList) . ')', |
|
2283
|
|
|
[ |
|
2284
|
|
|
'process_scheduled' => intval($this->getCurrentTime()), |
|
2285
|
|
|
'process_id' => $processId |
|
2286
|
|
|
] |
|
2287
|
|
|
); |
|
2288
|
|
|
|
|
2289
|
|
|
//save the number of assigned queue entrys to determine who many have been processed later |
|
2290
|
|
|
$numberOfAffectedRows = $this->db->sql_affected_rows(); |
|
2291
|
|
|
$this->db->exec_UPDATEquery( |
|
2292
|
|
|
'tx_crawler_process', |
|
2293
|
|
|
"process_id = '" . $processId . "'", |
|
2294
|
|
|
[ |
|
2295
|
|
|
'assigned_items_count' => intval($numberOfAffectedRows) |
|
2296
|
|
|
] |
|
2297
|
|
|
); |
|
2298
|
|
|
|
|
2299
|
|
|
if ($numberOfAffectedRows == count($quidList)) { |
|
2300
|
|
|
$this->db->sql_query('COMMIT'); |
|
2301
|
|
|
} else { |
|
2302
|
|
|
$this->db->sql_query('ROLLBACK'); |
|
2303
|
|
|
$this->CLI_debug("Nothing processed due to multi-process collision (" . $this->CLI_buildProcessId() . ")"); |
|
2304
|
|
|
return ($result | self::CLI_STATUS_ABORTED); |
|
2305
|
|
|
} |
|
2306
|
|
|
|
|
2307
|
|
|
foreach ($rows as $r) { |
|
|
|
|
|
|
2308
|
|
|
$result |= $this->readUrl($r['qid']); |
|
2309
|
|
|
|
|
2310
|
|
|
$counter++; |
|
2311
|
|
|
usleep(intval($sleepTime)); // Just to relax the system |
|
2312
|
|
|
|
|
2313
|
|
|
// if during the start and the current read url the cli has been disable we need to return from the function |
|
2314
|
|
|
// mark the process NOT as ended. |
|
2315
|
|
|
if ($this->getDisabled()) { |
|
2316
|
|
|
return ($result | self::CLI_STATUS_ABORTED); |
|
2317
|
|
|
} |
|
2318
|
|
|
|
|
2319
|
|
|
if (!$this->CLI_checkIfProcessIsActive($this->CLI_buildProcessId())) { |
|
2320
|
|
|
$this->CLI_debug("conflict / timeout (" . $this->CLI_buildProcessId() . ")"); |
|
2321
|
|
|
|
|
2322
|
|
|
//TODO might need an additional returncode |
|
2323
|
|
|
$result |= self::CLI_STATUS_ABORTED; |
|
2324
|
|
|
break; //possible timeout |
|
2325
|
|
|
} |
|
2326
|
|
|
} |
|
2327
|
|
|
|
|
2328
|
|
|
sleep(intval($sleepAfterFinish)); |
|
2329
|
|
|
|
|
2330
|
|
|
$msg = 'Rows: ' . $counter; |
|
2331
|
|
|
$this->CLI_debug($msg . " (" . $this->CLI_buildProcessId() . ")"); |
|
2332
|
|
|
} else { |
|
2333
|
|
|
$this->CLI_debug("Nothing within queue which needs to be processed (" . $this->CLI_buildProcessId() . ")"); |
|
2334
|
|
|
} |
|
2335
|
|
|
|
|
2336
|
|
|
if ($counter > 0) { |
|
2337
|
|
|
$result |= self::CLI_STATUS_PROCESSED; |
|
2338
|
|
|
} |
|
2339
|
|
|
|
|
2340
|
|
|
return $result; |
|
2341
|
|
|
} |
|
2342
|
|
|
|
|
2343
|
|
|
/** |
|
2344
|
|
|
* Activate hooks |
|
2345
|
|
|
* |
|
2346
|
|
|
* @return void |
|
2347
|
|
|
*/ |
|
2348
|
|
|
public function CLI_runHooks() |
|
2349
|
|
|
{ |
|
2350
|
|
|
global $TYPO3_CONF_VARS; |
|
2351
|
|
|
if (is_array($TYPO3_CONF_VARS['EXTCONF']['crawler']['cli_hooks'])) { |
|
2352
|
|
|
foreach ($TYPO3_CONF_VARS['EXTCONF']['crawler']['cli_hooks'] as $objRef) { |
|
2353
|
|
|
$hookObj = &GeneralUtility::getUserObj($objRef); |
|
2354
|
|
|
if (is_object($hookObj)) { |
|
2355
|
|
|
$hookObj->crawler_init($this); |
|
2356
|
|
|
} |
|
2357
|
|
|
} |
|
2358
|
|
|
} |
|
2359
|
|
|
} |
|
2360
|
|
|
|
|
2361
|
|
|
/** |
|
2362
|
|
|
* Try to acquire a new process with the given id |
|
2363
|
|
|
* also performs some auto-cleanup for orphan processes |
|
2364
|
|
|
* @todo preemption might not be the most elegant way to clean up |
|
2365
|
|
|
* |
|
2366
|
|
|
* @param string $id identification string for the process |
|
2367
|
|
|
* @return boolean |
|
2368
|
|
|
*/ |
|
2369
|
|
|
public function CLI_checkAndAcquireNewProcess($id) |
|
2370
|
|
|
{ |
|
2371
|
|
|
$ret = true; |
|
2372
|
|
|
|
|
2373
|
|
|
$systemProcessId = getmypid(); |
|
2374
|
|
|
if ($systemProcessId < 1) { |
|
2375
|
|
|
return false; |
|
2376
|
|
|
} |
|
2377
|
|
|
|
|
2378
|
|
|
$processCount = 0; |
|
2379
|
|
|
$orphanProcesses = []; |
|
2380
|
|
|
|
|
2381
|
|
|
$this->db->sql_query('BEGIN'); |
|
2382
|
|
|
|
|
2383
|
|
|
$res = $this->db->exec_SELECTquery( |
|
2384
|
|
|
'process_id,ttl', |
|
2385
|
|
|
'tx_crawler_process', |
|
2386
|
|
|
'active=1 AND deleted=0' |
|
2387
|
|
|
); |
|
2388
|
|
|
|
|
2389
|
|
|
$currentTime = $this->getCurrentTime(); |
|
2390
|
|
|
|
|
2391
|
|
|
while ($row = $this->db->sql_fetch_assoc($res)) { |
|
2392
|
|
|
if ($row['ttl'] < $currentTime) { |
|
2393
|
|
|
$orphanProcesses[] = $row['process_id']; |
|
2394
|
|
|
} else { |
|
2395
|
|
|
$processCount++; |
|
2396
|
|
|
} |
|
2397
|
|
|
} |
|
2398
|
|
|
|
|
2399
|
|
|
// if there are less than allowed active processes then add a new one |
|
2400
|
|
|
if ($processCount < intval($this->extensionSettings['processLimit'])) { |
|
2401
|
|
|
$this->CLI_debug("add process " . $this->CLI_buildProcessId() . " (" . ($processCount + 1) . "/" . intval($this->extensionSettings['processLimit']) . ")"); |
|
2402
|
|
|
|
|
2403
|
|
|
// create new process record |
|
2404
|
|
|
$this->db->exec_INSERTquery( |
|
2405
|
|
|
'tx_crawler_process', |
|
2406
|
|
|
[ |
|
2407
|
|
|
'process_id' => $id, |
|
2408
|
|
|
'active' => '1', |
|
2409
|
|
|
'ttl' => ($currentTime + intval($this->extensionSettings['processMaxRunTime'])), |
|
2410
|
|
|
'system_process_id' => $systemProcessId |
|
2411
|
|
|
] |
|
2412
|
|
|
); |
|
2413
|
|
|
} else { |
|
2414
|
|
|
$this->CLI_debug("Processlimit reached (" . ($processCount) . "/" . intval($this->extensionSettings['processLimit']) . ")"); |
|
2415
|
|
|
$ret = false; |
|
2416
|
|
|
} |
|
2417
|
|
|
|
|
2418
|
|
|
$this->CLI_releaseProcesses($orphanProcesses, true); // maybe this should be somehow included into the current lock |
|
2419
|
|
|
$this->CLI_deleteProcessesMarkedDeleted(); |
|
2420
|
|
|
|
|
2421
|
|
|
$this->db->sql_query('COMMIT'); |
|
2422
|
|
|
|
|
2423
|
|
|
return $ret; |
|
2424
|
|
|
} |
|
2425
|
|
|
|
|
2426
|
|
|
/** |
|
2427
|
|
|
* Release a process and the required resources |
|
2428
|
|
|
* |
|
2429
|
|
|
* @param mixed $releaseIds string with a single process-id or array with multiple process-ids |
|
2430
|
|
|
* @param boolean $withinLock show whether the DB-actions are included within an existing lock |
|
2431
|
|
|
* @return boolean |
|
2432
|
|
|
*/ |
|
2433
|
|
|
public function CLI_releaseProcesses($releaseIds, $withinLock = false) |
|
2434
|
|
|
{ |
|
2435
|
|
|
if (!is_array($releaseIds)) { |
|
2436
|
|
|
$releaseIds = [$releaseIds]; |
|
2437
|
|
|
} |
|
2438
|
|
|
|
|
2439
|
|
|
if (!count($releaseIds) > 0) { |
|
2440
|
|
|
return false; //nothing to release |
|
2441
|
|
|
} |
|
2442
|
|
|
|
|
2443
|
|
|
if (!$withinLock) { |
|
2444
|
|
|
$this->db->sql_query('BEGIN'); |
|
2445
|
|
|
} |
|
2446
|
|
|
|
|
2447
|
|
|
// some kind of 2nd chance algo - this way you need at least 2 processes to have a real cleanup |
|
2448
|
|
|
// this ensures that a single process can't mess up the entire process table |
|
2449
|
|
|
|
|
2450
|
|
|
// mark all processes as deleted which have no "waiting" queue-entires and which are not active |
|
2451
|
|
|
$this->db->exec_UPDATEquery( |
|
2452
|
|
|
'tx_crawler_queue', |
|
2453
|
|
|
'process_id IN (SELECT process_id FROM tx_crawler_process WHERE active=0 AND deleted=0)', |
|
2454
|
|
|
[ |
|
2455
|
|
|
'process_scheduled' => 0, |
|
2456
|
|
|
'process_id' => '' |
|
2457
|
|
|
] |
|
2458
|
|
|
); |
|
2459
|
|
|
$this->db->exec_UPDATEquery( |
|
2460
|
|
|
'tx_crawler_process', |
|
2461
|
|
|
'active=0 AND deleted=0 |
|
2462
|
|
|
AND NOT EXISTS ( |
|
2463
|
|
|
SELECT * FROM tx_crawler_queue |
|
2464
|
|
|
WHERE tx_crawler_queue.process_id = tx_crawler_process.process_id |
|
2465
|
|
|
AND tx_crawler_queue.exec_time = 0 |
|
2466
|
|
|
)', |
|
2467
|
|
|
[ |
|
2468
|
|
|
'deleted' => '1', |
|
2469
|
|
|
'system_process_id' => 0 |
|
2470
|
|
|
] |
|
2471
|
|
|
); |
|
2472
|
|
|
// mark all requested processes as non-active |
|
2473
|
|
|
$this->db->exec_UPDATEquery( |
|
2474
|
|
|
'tx_crawler_process', |
|
2475
|
|
|
'process_id IN (\'' . implode('\',\'', $releaseIds) . '\') AND deleted=0', |
|
2476
|
|
|
[ |
|
2477
|
|
|
'active' => '0' |
|
2478
|
|
|
] |
|
2479
|
|
|
); |
|
2480
|
|
|
$this->db->exec_UPDATEquery( |
|
2481
|
|
|
'tx_crawler_queue', |
|
2482
|
|
|
'exec_time=0 AND process_id IN ("' . implode('","', $releaseIds) . '")', |
|
2483
|
|
|
[ |
|
2484
|
|
|
'process_scheduled' => 0, |
|
2485
|
|
|
'process_id' => '' |
|
2486
|
|
|
] |
|
2487
|
|
|
); |
|
2488
|
|
|
|
|
2489
|
|
|
if (!$withinLock) { |
|
2490
|
|
|
$this->db->sql_query('COMMIT'); |
|
2491
|
|
|
} |
|
2492
|
|
|
|
|
2493
|
|
|
return true; |
|
2494
|
|
|
} |
|
2495
|
|
|
|
|
2496
|
|
|
/** |
|
2497
|
|
|
* Delete processes marked as deleted |
|
2498
|
|
|
* |
|
2499
|
|
|
* @return void |
|
2500
|
|
|
*/ |
|
2501
|
1 |
|
public function CLI_deleteProcessesMarkedDeleted() |
|
2502
|
|
|
{ |
|
2503
|
1 |
|
$this->db->exec_DELETEquery('tx_crawler_process', 'deleted = 1'); |
|
2504
|
1 |
|
} |
|
2505
|
|
|
|
|
2506
|
|
|
/** |
|
2507
|
|
|
* Check if there are still resources left for the process with the given id |
|
2508
|
|
|
* Used to determine timeouts and to ensure a proper cleanup if there's a timeout |
|
2509
|
|
|
* |
|
2510
|
|
|
* @param string identification string for the process |
|
2511
|
|
|
* @return boolean determines if the process is still active / has resources |
|
2512
|
|
|
* |
|
2513
|
|
|
* FIXME: Please remove Transaction, not needed as only a select query. |
|
2514
|
|
|
*/ |
|
2515
|
|
|
public function CLI_checkIfProcessIsActive($pid) |
|
2516
|
|
|
{ |
|
2517
|
|
|
$ret = false; |
|
2518
|
|
|
$this->db->sql_query('BEGIN'); |
|
2519
|
|
|
$res = $this->db->exec_SELECTquery( |
|
2520
|
|
|
'process_id,active,ttl', |
|
2521
|
|
|
'tx_crawler_process', |
|
2522
|
|
|
'process_id = \'' . $pid . '\' AND deleted=0', |
|
2523
|
|
|
'', |
|
2524
|
|
|
'ttl', |
|
2525
|
|
|
'0,1' |
|
2526
|
|
|
); |
|
2527
|
|
|
if ($row = $this->db->sql_fetch_assoc($res)) { |
|
2528
|
|
|
$ret = intVal($row['active']) == 1; |
|
2529
|
|
|
} |
|
2530
|
|
|
$this->db->sql_query('COMMIT'); |
|
2531
|
|
|
|
|
2532
|
|
|
return $ret; |
|
2533
|
|
|
} |
|
2534
|
|
|
|
|
2535
|
|
|
/** |
|
2536
|
|
|
* Create a unique Id for the current process |
|
2537
|
|
|
* |
|
2538
|
|
|
* @return string the ID |
|
2539
|
|
|
*/ |
|
2540
|
2 |
|
public function CLI_buildProcessId() |
|
2541
|
|
|
{ |
|
2542
|
2 |
|
if (!$this->processID) { |
|
2543
|
1 |
|
$this->processID = GeneralUtility::shortMD5($this->microtime(true)); |
|
2544
|
|
|
} |
|
2545
|
2 |
|
return $this->processID; |
|
2546
|
|
|
} |
|
2547
|
|
|
|
|
2548
|
|
|
/** |
|
2549
|
|
|
* @param bool $get_as_float |
|
2550
|
|
|
* |
|
2551
|
|
|
* @return mixed |
|
2552
|
|
|
*/ |
|
2553
|
|
|
protected function microtime($get_as_float = false) |
|
2554
|
|
|
{ |
|
2555
|
|
|
return microtime($get_as_float); |
|
2556
|
|
|
} |
|
2557
|
|
|
|
|
2558
|
|
|
/** |
|
2559
|
|
|
* Prints a message to the stdout (only if debug-mode is enabled) |
|
2560
|
|
|
* |
|
2561
|
|
|
* @param string $msg the message |
|
2562
|
|
|
*/ |
|
2563
|
|
|
public function CLI_debug($msg) |
|
2564
|
|
|
{ |
|
2565
|
|
|
if (intval($this->extensionSettings['processDebug'])) { |
|
2566
|
|
|
echo $msg . "\n"; |
|
2567
|
|
|
flush(); |
|
2568
|
|
|
} |
|
2569
|
|
|
} |
|
2570
|
|
|
|
|
2571
|
|
|
/** |
|
2572
|
|
|
* Get URL content by making direct request to TYPO3. |
|
2573
|
|
|
* |
|
2574
|
|
|
* @param string $url Page URL |
|
2575
|
|
|
* @param int $crawlerId Crawler-ID |
|
2576
|
|
|
* @return array |
|
2577
|
|
|
*/ |
|
2578
|
2 |
|
protected function sendDirectRequest($url, $crawlerId) |
|
2579
|
|
|
{ |
|
2580
|
2 |
|
$parsedUrl = parse_url($url); |
|
2581
|
2 |
|
if (!is_array($parsedUrl)) { |
|
2582
|
|
|
return []; |
|
2583
|
|
|
} |
|
2584
|
|
|
|
|
2585
|
2 |
|
$requestHeaders = $this->buildRequestHeaderArray($parsedUrl, $crawlerId); |
|
2586
|
|
|
|
|
2587
|
2 |
|
$cmd = escapeshellcmd($this->extensionSettings['phpPath']); |
|
2588
|
2 |
|
$cmd .= ' '; |
|
2589
|
2 |
|
$cmd .= escapeshellarg(ExtensionManagementUtility::extPath('crawler') . 'cli/bootstrap.php'); |
|
2590
|
2 |
|
$cmd .= ' '; |
|
2591
|
2 |
|
$cmd .= escapeshellarg($this->getFrontendBasePath()); |
|
2592
|
2 |
|
$cmd .= ' '; |
|
2593
|
2 |
|
$cmd .= escapeshellarg($url); |
|
2594
|
2 |
|
$cmd .= ' '; |
|
2595
|
2 |
|
$cmd .= escapeshellarg(base64_encode(serialize($requestHeaders))); |
|
2596
|
|
|
|
|
2597
|
2 |
|
$startTime = microtime(true); |
|
2598
|
2 |
|
$content = $this->executeShellCommand($cmd); |
|
2599
|
2 |
|
$this->log($url . ' ' . (microtime(true) - $startTime)); |
|
2600
|
|
|
|
|
2601
|
|
|
$result = [ |
|
2602
|
2 |
|
'request' => implode("\r\n", $requestHeaders) . "\r\n\r\n", |
|
2603
|
2 |
|
'headers' => '', |
|
2604
|
2 |
|
'content' => $content |
|
2605
|
|
|
]; |
|
2606
|
|
|
|
|
2607
|
2 |
|
return $result; |
|
2608
|
|
|
} |
|
2609
|
|
|
|
|
2610
|
|
|
/** |
|
2611
|
|
|
* Cleans up entries that stayed for too long in the queue. These are: |
|
2612
|
|
|
* - processed entries that are over 1.5 days in age |
|
2613
|
|
|
* - scheduled entries that are over 7 days old |
|
2614
|
|
|
* |
|
2615
|
|
|
* @return void |
|
2616
|
|
|
*/ |
|
2617
|
|
|
protected function cleanUpOldQueueEntries() |
|
2618
|
|
|
{ |
|
2619
|
|
|
$processedAgeInSeconds = $this->extensionSettings['cleanUpProcessedAge'] * 86400; // 24*60*60 Seconds in 24 hours |
|
2620
|
|
|
$scheduledAgeInSeconds = $this->extensionSettings['cleanUpScheduledAge'] * 86400; |
|
2621
|
|
|
|
|
2622
|
|
|
$now = time(); |
|
2623
|
|
|
$condition = '(exec_time<>0 AND exec_time<' . ($now - $processedAgeInSeconds) . ') OR scheduled<=' . ($now - $scheduledAgeInSeconds); |
|
2624
|
|
|
$this->flushQueue($condition); |
|
2625
|
|
|
} |
|
2626
|
|
|
|
|
2627
|
|
|
/** |
|
2628
|
|
|
* Initializes a TypoScript Frontend necessary for using TypoScript and TypoLink functions |
|
2629
|
|
|
* |
|
2630
|
|
|
* @param int $id |
|
2631
|
|
|
* @param int $typeNum |
|
2632
|
|
|
* |
|
2633
|
|
|
* @return void |
|
2634
|
|
|
*/ |
|
2635
|
|
|
protected function initTSFE($id = 1, $typeNum = 0) |
|
2636
|
|
|
{ |
|
2637
|
|
|
EidUtility::initTCA(); |
|
2638
|
|
|
if (!is_object($GLOBALS['TT'])) { |
|
2639
|
|
|
$GLOBALS['TT'] = new NullTimeTracker(); |
|
2640
|
|
|
$GLOBALS['TT']->start(); |
|
2641
|
|
|
} |
|
2642
|
|
|
|
|
2643
|
|
|
$GLOBALS['TSFE'] = GeneralUtility::makeInstance(TypoScriptFrontendController::class, $GLOBALS['TYPO3_CONF_VARS'], $id, $typeNum); |
|
2644
|
|
|
$GLOBALS['TSFE']->sys_page = GeneralUtility::makeInstance(PageRepository::class); |
|
2645
|
|
|
$GLOBALS['TSFE']->sys_page->init(true); |
|
2646
|
|
|
$GLOBALS['TSFE']->connectToDB(); |
|
2647
|
|
|
$GLOBALS['TSFE']->initFEuser(); |
|
2648
|
|
|
$GLOBALS['TSFE']->determineId(); |
|
2649
|
|
|
$GLOBALS['TSFE']->initTemplate(); |
|
2650
|
|
|
$GLOBALS['TSFE']->rootLine = $GLOBALS['TSFE']->sys_page->getRootLine($id, ''); |
|
2651
|
|
|
$GLOBALS['TSFE']->getConfigArray(); |
|
2652
|
|
|
PageGenerator::pagegenInit(); |
|
2653
|
|
|
} |
|
2654
|
|
|
|
|
2655
|
|
|
/** |
|
2656
|
|
|
* Returns a md5 hash generated from a serialized configuration array. |
|
2657
|
|
|
* |
|
2658
|
|
|
* @param array $configuration |
|
2659
|
|
|
* |
|
2660
|
|
|
* @return string |
|
2661
|
|
|
*/ |
|
2662
|
9 |
|
protected function getConfigurationHash(array $configuration) { |
|
2663
|
9 |
|
unset($configuration['paramExpanded']); |
|
2664
|
9 |
|
unset($configuration['URLs']); |
|
2665
|
9 |
|
return md5(serialize($configuration)); |
|
2666
|
|
|
} |
|
2667
|
|
|
|
|
2668
|
|
|
/** |
|
2669
|
|
|
* Check whether the Crawling Protocol should be http or https |
|
2670
|
|
|
* |
|
2671
|
|
|
* @param $crawlerConfiguration |
|
2672
|
|
|
* @param $pageConfiguration |
|
2673
|
|
|
* |
|
2674
|
|
|
* @return bool |
|
2675
|
|
|
*/ |
|
2676
|
6 |
|
protected function isCrawlingProtocolHttps($crawlerConfiguration, $pageConfiguration) { |
|
2677
|
|
|
switch($crawlerConfiguration) { |
|
2678
|
6 |
|
case -1: |
|
2679
|
1 |
|
return false; |
|
2680
|
|
|
break; |
|
|
|
|
|
|
2681
|
5 |
|
case 0: |
|
2682
|
3 |
|
return $pageConfiguration; |
|
2683
|
|
|
break; |
|
|
|
|
|
|
2684
|
2 |
|
case 1: |
|
2685
|
1 |
|
return true; |
|
2686
|
|
|
break; |
|
|
|
|
|
|
2687
|
|
|
default: |
|
2688
|
1 |
|
return false; |
|
2689
|
|
|
} |
|
2690
|
|
|
|
|
2691
|
|
|
} |
|
2692
|
|
|
} |
|
2693
|
|
|
|
AOEpeople /
crawler
Loading history...
If you define a variable conditionally, it can happen that it is not defined for all execution paths.
Let’s take a look at an example:
In the above example, the variable $x is defined if you pass “foo” or “bar” as argument for $a. However, since the switch statement has no default case statement, if you pass any other value, the variable $x would be undefined.
Available Fixes
Check for existence of the variable explicitly:
Define a default value for the variable:
Add a value for the missing path: