1
|
|
|
<?php |
2
|
|
|
|
3
|
|
|
namespace MediaMonks\Crawler; |
4
|
|
|
|
5
|
|
|
use MediaMonks\Crawler\Url\Matcher\UrlMatcherInterface; |
6
|
|
|
use MediaMonks\Crawler\Url\Normalizer\UrlNormalizerInterface; |
7
|
|
|
use Symfony\Component\BrowserKit\Client; |
8
|
|
|
use Symfony\Component\DomCrawler\Crawler as DomCrawler; |
9
|
|
|
use Psr\Log\LoggerAwareInterface; |
10
|
|
|
use Psr\Log\LoggerInterface; |
11
|
|
|
use Psr\Log\NullLogger; |
12
|
|
|
|
13
|
|
|
class Crawler implements LoggerAwareInterface |
14
|
|
|
{ |
15
|
|
|
/** |
16
|
|
|
* @var Client |
17
|
|
|
*/ |
18
|
|
|
private $client; |
19
|
|
|
|
20
|
|
|
/** |
21
|
|
|
* @var int |
22
|
|
|
*/ |
23
|
|
|
private $limit = 0; |
24
|
|
|
|
25
|
|
|
/** |
26
|
|
|
* @var bool |
27
|
|
|
*/ |
28
|
|
|
private $stopOnError = false; |
29
|
|
|
|
30
|
|
|
/** |
31
|
|
|
* @var UrlMatcherInterface[] |
32
|
|
|
*/ |
33
|
|
|
private $whitelistUrlMatchers = []; |
34
|
|
|
|
35
|
|
|
/** |
36
|
|
|
* @var UrlMatcherInterface[] |
37
|
|
|
*/ |
38
|
|
|
private $blacklistUrlMatchers = []; |
39
|
|
|
|
40
|
|
|
/** |
41
|
|
|
* @var UrlNormalizerInterface[] |
42
|
|
|
*/ |
43
|
|
|
private $urlNormalizers = []; |
44
|
|
|
|
45
|
|
|
/** |
46
|
|
|
* @var Url |
47
|
|
|
*/ |
48
|
|
|
private $baseUrl; |
49
|
|
|
|
50
|
|
|
/** |
51
|
|
|
* @var array |
52
|
|
|
*/ |
53
|
|
|
private $urlsCrawled = []; |
54
|
|
|
|
55
|
|
|
/** |
56
|
|
|
* @var array |
57
|
|
|
*/ |
58
|
|
|
private $urlsQueued = []; |
59
|
|
|
|
60
|
|
|
/** |
61
|
|
|
* @var array |
62
|
|
|
*/ |
63
|
|
|
private $urlsRejected = []; |
64
|
|
|
|
65
|
|
|
/** |
66
|
|
|
* @var array |
67
|
|
|
*/ |
68
|
|
|
private $urlsReturned = []; |
69
|
|
|
|
70
|
|
|
/** |
71
|
|
|
* @var LoggerInterface |
72
|
|
|
*/ |
73
|
|
|
private $logger = null; |
74
|
|
|
|
75
|
|
|
/** |
76
|
|
|
* @param Client $client |
77
|
|
|
* @param array $options |
78
|
|
|
*/ |
79
|
5 |
|
public function __construct(Client $client = null, array $options = []) |
80
|
|
|
{ |
81
|
5 |
|
if (empty($client)) { |
82
|
3 |
|
$client = new \Goutte\Client(); |
83
|
3 |
|
} |
84
|
|
|
|
85
|
5 |
|
$this->setClient($client); |
86
|
5 |
|
$this->setOptions($options); |
87
|
|
|
|
88
|
5 |
|
return $this; |
89
|
|
|
} |
90
|
|
|
|
91
|
|
|
/** |
92
|
|
|
* @param Client $client |
93
|
|
|
*/ |
94
|
5 |
|
public function setClient(Client $client) |
95
|
|
|
{ |
96
|
5 |
|
$this->client = $client; |
97
|
5 |
|
} |
98
|
|
|
|
99
|
|
|
/** |
100
|
|
|
* @return Client |
101
|
|
|
*/ |
102
|
2 |
|
public function getClient() |
103
|
|
|
{ |
104
|
2 |
|
return $this->client; |
105
|
|
|
} |
106
|
|
|
|
107
|
|
|
/** |
108
|
|
|
* @param array $options |
109
|
|
|
*/ |
110
|
5 |
|
public function setOptions(array $options) |
111
|
|
|
{ |
112
|
5 |
|
if (isset($options['limit'])) { |
113
|
2 |
|
$this->setLimit($options['limit']); |
114
|
1 |
|
} |
115
|
5 |
|
if (isset($options['stop_on_error'])) { |
116
|
1 |
|
$this->setStopOnError($options['stop_on_error']); |
117
|
1 |
|
} |
118
|
5 |
|
if (isset($options['logger'])) { |
119
|
1 |
|
$this->setLogger($options['logger']); |
120
|
1 |
|
} |
121
|
5 |
|
if (isset($options['whitelist_url_matchers'])) { |
122
|
1 |
|
$this->setWhitelistUrlMatchers($options['whitelist_url_matchers']); |
123
|
1 |
|
} |
124
|
5 |
|
if (isset($options['blacklist_url_matchers'])) { |
125
|
1 |
|
$this->setBlacklistUrlMatchers($options['blacklist_url_matchers']); |
126
|
1 |
|
} |
127
|
5 |
|
} |
128
|
|
|
|
129
|
|
|
/** |
130
|
|
|
* @return int |
131
|
|
|
*/ |
132
|
3 |
|
public function getLimit() |
133
|
|
|
{ |
134
|
3 |
|
return $this->limit; |
135
|
|
|
} |
136
|
|
|
|
137
|
|
|
/** |
138
|
|
|
* @param int $limit |
139
|
|
|
* @return $this |
140
|
|
|
*/ |
141
|
2 |
|
public function setLimit($limit) |
142
|
|
|
{ |
143
|
2 |
|
$this->limit = $limit; |
144
|
|
|
|
145
|
2 |
|
return $this; |
146
|
|
|
} |
147
|
|
|
|
148
|
|
|
/** |
149
|
|
|
* @return boolean |
150
|
|
|
*/ |
151
|
3 |
|
public function getStopOnError() |
152
|
|
|
{ |
153
|
3 |
|
return $this->stopOnError; |
154
|
|
|
} |
155
|
|
|
|
156
|
|
|
/** |
157
|
|
|
* @param boolean $stopOnError |
158
|
|
|
* @return Crawler |
159
|
|
|
*/ |
160
|
2 |
|
public function setStopOnError($stopOnError) |
161
|
|
|
{ |
162
|
2 |
|
$this->stopOnError = $stopOnError; |
163
|
|
|
|
164
|
2 |
|
return $this; |
165
|
|
|
} |
166
|
|
|
|
167
|
|
|
/** |
168
|
|
|
* @return array |
169
|
|
|
*/ |
170
|
3 |
|
public function getUrlsCrawled() |
171
|
|
|
{ |
172
|
3 |
|
return $this->urlsCrawled; |
173
|
|
|
} |
174
|
|
|
|
175
|
|
|
/** |
176
|
|
|
* @return array |
177
|
|
|
*/ |
178
|
1 |
|
public function getUrlsQueued() |
179
|
|
|
{ |
180
|
1 |
|
return $this->urlsQueued; |
181
|
|
|
} |
182
|
|
|
|
183
|
|
|
/** |
184
|
|
|
* @return array |
185
|
|
|
*/ |
186
|
1 |
|
public function getUrlsRejected() |
187
|
|
|
{ |
188
|
1 |
|
return $this->urlsRejected; |
189
|
|
|
} |
190
|
|
|
|
191
|
|
|
/** |
192
|
|
|
* @return array |
193
|
|
|
*/ |
194
|
1 |
|
public function getUrlsReturned() |
195
|
|
|
{ |
196
|
1 |
|
return $this->urlsReturned; |
197
|
|
|
} |
198
|
|
|
|
199
|
|
|
/** |
200
|
|
|
* @param $urlMatchers |
201
|
|
|
* @return $this |
202
|
|
|
*/ |
203
|
1 |
|
public function setWhitelistUrlMatchers(array $urlMatchers) |
204
|
|
|
{ |
205
|
1 |
|
$this->clearWhitelistUrlMatchers(); |
206
|
1 |
|
foreach ($urlMatchers as $matcher) { |
207
|
1 |
|
$this->addWhitelistUrlMatcher($matcher); |
208
|
1 |
|
} |
209
|
|
|
|
210
|
1 |
|
return $this; |
211
|
|
|
} |
212
|
|
|
|
213
|
|
|
/** |
214
|
|
|
* @return Url\Matcher\UrlMatcherInterface[] |
215
|
|
|
*/ |
216
|
2 |
|
public function getWhitelistUrlMatchers() |
217
|
|
|
{ |
218
|
2 |
|
return $this->whitelistUrlMatchers; |
219
|
|
|
} |
220
|
|
|
|
221
|
|
|
/** |
222
|
|
|
* @param UrlMatcherInterface $urlMatcher |
223
|
|
|
* @return $this |
224
|
|
|
*/ |
225
|
1 |
|
public function addWhitelistUrlMatcher(UrlMatcherInterface $urlMatcher) |
226
|
|
|
{ |
227
|
1 |
|
$this->whitelistUrlMatchers[] = $urlMatcher; |
228
|
|
|
|
229
|
1 |
|
return $this; |
230
|
|
|
} |
231
|
|
|
|
232
|
|
|
/** |
233
|
|
|
* @return $this |
234
|
|
|
*/ |
235
|
1 |
|
public function clearWhitelistUrlMatchers() |
236
|
|
|
{ |
237
|
1 |
|
$this->whitelistUrlMatchers = []; |
238
|
|
|
|
239
|
1 |
|
return $this; |
240
|
|
|
} |
241
|
|
|
|
242
|
|
|
/** |
243
|
|
|
* @param array $urlMatchers |
244
|
|
|
* @return $this |
245
|
|
|
*/ |
246
|
1 |
|
public function setBlacklistUrlMatchers(array $urlMatchers) |
247
|
|
|
{ |
248
|
1 |
|
$this->clearBlacklistUrlMatchers(); |
249
|
1 |
|
foreach ($urlMatchers as $matcher) { |
250
|
1 |
|
$this->addBlacklistUrlMatcher($matcher); |
251
|
1 |
|
} |
252
|
|
|
|
253
|
1 |
|
return $this; |
254
|
|
|
} |
255
|
|
|
|
256
|
|
|
/** |
257
|
|
|
* @return Url\Matcher\UrlMatcherInterface[] |
258
|
|
|
*/ |
259
|
2 |
|
public function getBlacklistUrlMatchers() |
260
|
|
|
{ |
261
|
2 |
|
return $this->blacklistUrlMatchers; |
262
|
|
|
} |
263
|
|
|
|
264
|
|
|
/** |
265
|
|
|
* @param UrlMatcherInterface $urlMatcher |
266
|
|
|
* @return $this |
267
|
|
|
*/ |
268
|
1 |
|
public function addBlacklistUrlMatcher(UrlMatcherInterface $urlMatcher) |
269
|
|
|
{ |
270
|
1 |
|
$this->blacklistUrlMatchers[] = $urlMatcher; |
271
|
|
|
|
272
|
1 |
|
return $this; |
273
|
|
|
} |
274
|
|
|
|
275
|
|
|
/** |
276
|
|
|
* @return $this |
277
|
|
|
*/ |
278
|
1 |
|
public function clearBlacklistUrlMatchers() |
279
|
|
|
{ |
280
|
1 |
|
$this->blacklistUrlMatchers = []; |
281
|
|
|
|
282
|
1 |
|
return $this; |
283
|
|
|
} |
284
|
|
|
|
285
|
|
|
/** |
286
|
|
|
* @param array $normalizers |
287
|
|
|
* @return $this |
288
|
|
|
*/ |
289
|
|
|
public function setUrlNormalizers(array $normalizers) |
290
|
|
|
{ |
291
|
|
|
$this->clearUrlNormalizers(); |
292
|
|
|
|
293
|
|
|
foreach ($normalizers as $normalizer) { |
294
|
|
|
$this->addUrlNormalizer($normalizer); |
295
|
|
|
} |
296
|
|
|
|
297
|
|
|
return $this; |
298
|
|
|
} |
299
|
|
|
|
300
|
|
|
/** |
301
|
|
|
* @param UrlNormalizerInterface $normalizer |
302
|
|
|
* @return $this |
303
|
|
|
*/ |
304
|
|
|
public function addUrlNormalizer(UrlNormalizerInterface $normalizer) |
305
|
|
|
{ |
306
|
|
|
$this->urlNormalizers[] = $normalizer; |
307
|
|
|
|
308
|
|
|
return $this; |
309
|
|
|
} |
310
|
|
|
|
311
|
|
|
/** |
312
|
|
|
* @return $this |
313
|
|
|
*/ |
314
|
|
|
public function clearUrlNormalizers() |
315
|
|
|
{ |
316
|
|
|
$this->urlNormalizers = []; |
317
|
|
|
|
318
|
|
|
return $this; |
319
|
|
|
} |
320
|
|
|
|
321
|
|
|
/** |
322
|
|
|
* @return LoggerInterface |
323
|
|
|
*/ |
324
|
5 |
|
public function getLogger() |
325
|
|
|
{ |
326
|
5 |
|
if (is_null($this->logger)) { |
327
|
3 |
|
$this->logger = new NullLogger(); |
328
|
3 |
|
} |
329
|
|
|
|
330
|
5 |
|
return $this->logger; |
331
|
|
|
} |
332
|
|
|
|
333
|
|
|
/** |
334
|
|
|
* @param LoggerInterface $logger |
335
|
|
|
* @return $this |
336
|
|
|
*/ |
337
|
2 |
|
public function setLogger(LoggerInterface $logger) |
338
|
|
|
{ |
339
|
2 |
|
$this->logger = $logger; |
340
|
|
|
|
341
|
2 |
|
return $this; |
342
|
|
|
} |
343
|
|
|
|
344
|
|
|
/** |
345
|
|
|
* @param Url $url |
346
|
|
|
*/ |
347
|
2 |
|
protected function addUrlToQueue(Url $url) |
348
|
|
|
{ |
349
|
2 |
|
$this->urlsQueued[(string)$url] = $url; |
350
|
2 |
|
} |
351
|
|
|
|
352
|
|
|
/** |
353
|
|
|
* @param $url |
354
|
|
|
* @return Url |
355
|
|
|
*/ |
356
|
2 |
|
protected function createHttpUrlString($url) |
357
|
|
|
{ |
358
|
2 |
|
return Url::createFromString($url); |
359
|
|
|
} |
360
|
|
|
|
361
|
|
|
/** |
362
|
|
|
* @param Url $url |
363
|
|
|
*/ |
364
|
2 |
|
protected function reset(Url $url) |
365
|
|
|
{ |
366
|
2 |
|
$this->baseUrl = $url; |
367
|
2 |
|
$this->urlsCrawled = []; |
368
|
2 |
|
$this->urlsQueued = []; |
369
|
|
|
|
370
|
2 |
|
$this->addUrlToQueue($url); |
371
|
2 |
|
} |
372
|
|
|
|
373
|
|
|
/** |
374
|
|
|
* @param string $url |
375
|
|
|
* @return \Generator|void |
376
|
|
|
*/ |
377
|
2 |
|
public function crawl($url) |
378
|
|
|
{ |
379
|
2 |
|
$url = $this->createHttpUrlString($url); |
380
|
2 |
|
$this->reset($url); |
381
|
|
|
|
382
|
2 |
|
while (count($this->urlsQueued) > 0) { |
383
|
|
|
|
384
|
2 |
|
$url = array_shift($this->urlsQueued); |
385
|
|
|
|
386
|
|
|
try { |
387
|
2 |
|
$crawler = $this->requestPage((string)$url); |
388
|
2 |
|
} catch (\Exception $e) { |
389
|
|
|
$this->getLogger()->error(sprintf('Error requesting page %s: %s', $url, $e->getMessage())); |
390
|
|
|
|
391
|
|
|
if ($this->getStopOnError()) { |
392
|
|
|
return; |
393
|
|
|
} |
394
|
|
|
|
395
|
|
|
continue; |
396
|
|
|
} |
397
|
|
|
|
398
|
2 |
|
$this->urlsCrawled[] = (string)$url; |
399
|
2 |
|
$this->updateQueue($crawler); |
400
|
|
|
|
401
|
2 |
|
if ($this->shouldReturnUrl($url)) { |
402
|
2 |
|
$this->getLogger()->debug(sprintf('Return url "%s"', $url)); |
403
|
|
|
|
404
|
2 |
|
$this->urlsReturned[] = (string)$url; |
405
|
|
|
|
406
|
2 |
|
yield new Page($url, $crawler); |
407
|
2 |
|
} |
408
|
|
|
|
409
|
2 |
|
if ($this->isLimitReached()) { |
410
|
|
|
$this->getLogger()->info(sprintf('Crawl limit of %d was reach', $this->limit)); |
411
|
|
|
|
412
|
|
|
return; |
413
|
|
|
} |
414
|
2 |
|
} |
415
|
2 |
|
} |
416
|
|
|
|
417
|
|
|
/** |
418
|
|
|
* @param DomCrawler $crawler |
419
|
|
|
*/ |
420
|
2 |
|
protected function updateQueue(DomCrawler $crawler) |
421
|
|
|
{ |
422
|
2 |
|
foreach ($this->extractUrlsFromCrawler($crawler) as $url) { |
423
|
1 |
|
if (!in_array($url, $this->urlsRejected)) { |
424
|
1 |
|
$this->getLogger()->debug(sprintf('Found url %s in page', $url)); |
425
|
|
|
try { |
426
|
1 |
|
$url = $this->normalizeUrl($this->createHttpUrlString($url)); |
427
|
|
|
|
428
|
1 |
|
if ($this->shouldCrawlUrl($url)) { |
429
|
1 |
|
$this->addUrlToQueue($url); |
430
|
1 |
|
} |
431
|
1 |
|
} catch (\Exception $e) { |
432
|
|
|
$this->getLogger()->warning( |
433
|
|
|
sprintf('Url %s could not be converted to an object: %s', $url, $e->getMessage()) |
434
|
|
|
); |
435
|
|
|
$this->urlsRejected[] = $url; |
436
|
|
|
} |
437
|
1 |
|
} |
438
|
2 |
|
} |
439
|
2 |
|
} |
440
|
|
|
|
441
|
|
|
/** |
442
|
|
|
* @param Url $url |
443
|
|
|
* @return Url |
444
|
|
|
*/ |
445
|
1 |
|
protected function normalizeUrl(Url $url) |
446
|
|
|
{ |
447
|
1 |
|
foreach($this->urlNormalizers as $normalizer) { |
448
|
|
|
$url = $normalizer->normalize($url); |
449
|
1 |
|
} |
450
|
|
|
|
451
|
1 |
|
return $url; |
452
|
|
|
} |
453
|
|
|
|
454
|
|
|
/** |
455
|
|
|
* @param Url $url |
456
|
|
|
* @return bool |
457
|
|
|
*/ |
458
|
2 |
|
protected function shouldReturnUrl(Url $url) |
459
|
|
|
{ |
460
|
2 |
|
if (!empty($this->whitelistUrlMatchers)) { |
461
|
|
|
foreach ($this->whitelistUrlMatchers as $matcher) { |
462
|
|
|
if ($matcher->matches($url)) { |
463
|
|
|
return true; |
464
|
|
|
} |
465
|
|
|
} |
466
|
|
|
$this->getLogger()->info(sprintf('Skipped "%s" because it is not whitelisted', $url)); |
467
|
|
|
|
468
|
|
|
return false; |
469
|
|
|
} |
470
|
|
|
|
471
|
2 |
|
foreach ($this->blacklistUrlMatchers as $matcher) { |
472
|
|
|
if ($matcher->matches($url)) { |
473
|
|
|
$this->getLogger()->info(sprintf('Skipped "%s" because it is blacklisted', $url)); |
474
|
|
|
|
475
|
|
|
return false; |
476
|
|
|
} |
477
|
2 |
|
} |
478
|
|
|
|
479
|
2 |
|
return true; |
480
|
|
|
} |
481
|
|
|
|
482
|
|
|
/** |
483
|
|
|
* @param Url $url |
484
|
|
|
* @return bool |
485
|
|
|
*/ |
486
|
1 |
|
protected function shouldCrawlUrl(Url $url) |
487
|
|
|
{ |
488
|
1 |
|
$urlString = (string)$url; |
489
|
1 |
|
if (in_array($urlString, $this->urlsRejected)) { |
490
|
|
|
return false; |
491
|
|
|
} |
492
|
1 |
|
if (in_array($urlString, $this->urlsCrawled)) { |
493
|
|
|
return false; |
494
|
|
|
} |
495
|
1 |
|
if (isset($this->urlsQueued[$urlString])) { |
496
|
|
|
return false; |
497
|
|
|
} |
498
|
|
|
|
499
|
1 |
|
if (!$this->isUrlPartOfBaseUrl($url)) { |
500
|
|
|
$this->urlsRejected[] = (string)$url; |
501
|
|
|
return false; |
502
|
|
|
} |
503
|
|
|
|
504
|
1 |
|
return true; |
505
|
|
|
} |
506
|
|
|
|
507
|
|
|
/** |
508
|
|
|
* @param Url $url |
509
|
|
|
* @return bool |
510
|
|
|
*/ |
511
|
1 |
|
protected function isUrlPartOfBaseUrl(Url $url) |
512
|
|
|
{ |
513
|
1 |
|
$baseUrlString = (string)$this->baseUrl; |
514
|
1 |
|
$this->getLogger()->debug($baseUrlString.' - '.$url); |
515
|
1 |
|
if (strpos((string)$url, $baseUrlString) === false) { |
516
|
|
|
return false; |
517
|
|
|
} |
518
|
|
|
|
519
|
1 |
|
return true; |
520
|
|
|
} |
521
|
|
|
|
522
|
|
|
/** |
523
|
|
|
* @return bool |
524
|
|
|
*/ |
525
|
2 |
|
private function isLimitReached() |
526
|
|
|
{ |
527
|
2 |
|
return (!empty($this->limit) && count($this->urlsReturned) === $this->limit); |
528
|
|
|
} |
529
|
|
|
|
530
|
|
|
/** |
531
|
|
|
* @param DomCrawler $crawler |
532
|
|
|
* @return array |
533
|
|
|
*/ |
534
|
2 |
|
protected function extractUrlsFromCrawler(DomCrawler $crawler) |
535
|
|
|
{ |
536
|
2 |
|
return $crawler->filter('a')->each( |
537
|
1 |
|
function (DomCrawler $node) { |
538
|
1 |
|
return $node->link()->getUri(); |
539
|
|
|
} |
540
|
2 |
|
); |
541
|
|
|
} |
542
|
|
|
|
543
|
|
|
/** |
544
|
|
|
* @param $url |
545
|
|
|
* @return DomCrawler |
546
|
|
|
*/ |
547
|
2 |
|
protected function requestPage($url) |
548
|
|
|
{ |
549
|
2 |
|
$this->getLogger()->info(sprintf('Crawling page %s', $url)); |
550
|
2 |
|
$crawler = $this->client->request('GET', $url); |
551
|
2 |
|
$this->getLogger()->info(sprintf('Crawled page %s', $url)); |
552
|
|
|
|
553
|
2 |
|
return $crawler; |
554
|
|
|
} |
555
|
|
|
} |
556
|
|
|
|