1
|
|
|
<?php |
2
|
|
|
/** |
3
|
|
|
* File containing the {@see AppUtils\ConvertHelper_URLFinder} class. |
4
|
|
|
* |
5
|
|
|
* @package Application Utils |
6
|
|
|
* @subpackage ConvertHelper |
7
|
|
|
* @see AppUtils\ConvertHelper_URLFinder |
8
|
|
|
*/ |
9
|
|
|
|
10
|
|
|
declare(strict_types=1); |
11
|
|
|
|
12
|
|
|
namespace AppUtils; |
13
|
|
|
|
14
|
|
|
/** |
15
|
|
|
* Can find any URLs in a string, be it plain text or HTML, XML. |
16
|
|
|
* |
17
|
|
|
* @package Application Utils |
18
|
|
|
* @subpackage ConvertHelper |
19
|
|
|
* @author Sebastian Mordziol <[email protected]> |
20
|
|
|
* |
21
|
|
|
* @see ConvertHelper::createURLFinder() |
22
|
|
|
*/ |
23
|
|
|
class ConvertHelper_URLFinder implements Interface_Optionable |
24
|
|
|
{ |
25
|
|
|
const ERROR_INVALID_DETECTOR_CLASS = 87901; |
26
|
|
|
|
27
|
|
|
use Traits_Optionable; |
28
|
|
|
|
29
|
|
|
/** |
30
|
|
|
* @var array<string,URLInfo> |
31
|
|
|
*/ |
32
|
|
|
private $urls = array(); |
33
|
|
|
|
34
|
|
|
/** |
35
|
|
|
* @var array<string,URLInfo> |
36
|
|
|
*/ |
37
|
|
|
private $emails = array(); |
38
|
|
|
|
39
|
|
|
/** |
40
|
|
|
* @var string[] |
41
|
|
|
*/ |
42
|
|
|
private $enabledDetectorClasses = array( |
43
|
|
|
ConvertHelper_URLFinder_Detector_Tel::class => true, |
44
|
|
|
ConvertHelper_URLFinder_Detector_HTMLAttributes::class => false, |
45
|
|
|
ConvertHelper_URLFinder_Detector_IPV4::class => true |
46
|
|
|
); |
47
|
|
|
|
48
|
|
|
/** |
49
|
|
|
* @var string[] |
50
|
|
|
*/ |
51
|
|
|
private $matches = array(); |
52
|
|
|
|
53
|
|
|
/** |
54
|
|
|
* @var string[] |
55
|
|
|
*/ |
56
|
|
|
private $boundaries = array( |
57
|
|
|
"\n", |
58
|
|
|
"\t", |
59
|
|
|
"\r", |
60
|
|
|
'"', |
61
|
|
|
"'", |
62
|
|
|
'|', |
63
|
|
|
',', |
64
|
|
|
';', |
65
|
|
|
'<', |
66
|
|
|
'>' |
67
|
|
|
); |
68
|
|
|
|
69
|
|
|
/** |
70
|
|
|
* @var ConvertHelper_URLFinder_Detector[] |
71
|
|
|
*/ |
72
|
|
|
private $detectors = array(); |
73
|
|
|
|
74
|
|
|
/** |
75
|
|
|
* @var string |
76
|
|
|
*/ |
77
|
|
|
private $subject; |
78
|
|
|
|
79
|
|
|
/** |
80
|
|
|
* @var bool |
81
|
|
|
*/ |
82
|
|
|
private $parsed = false; |
83
|
|
|
|
84
|
|
|
public function __construct(string $subject) |
85
|
|
|
{ |
86
|
|
|
$this->subject = $subject; |
87
|
|
|
} |
88
|
|
|
|
89
|
|
|
public function getDefaultOptions() : array |
90
|
|
|
{ |
91
|
|
|
return array( |
92
|
|
|
'include-emails' => false, |
93
|
|
|
'omit-mailto' => false, |
94
|
|
|
'sorting' => false, |
95
|
|
|
'normalize' => false |
96
|
|
|
); |
97
|
|
|
} |
98
|
|
|
|
99
|
|
|
/** |
100
|
|
|
* Whether all URLs should be normalized (parameters ordered alphabetically, |
101
|
|
|
* whitespace removal). This ensures that URL duplicates are detected even |
102
|
|
|
* if they have a different order of parameters. |
103
|
|
|
* |
104
|
|
|
* @param bool $enabled |
105
|
|
|
* @return $this |
106
|
|
|
*/ |
107
|
|
|
public function enableNormalizing(bool $enabled=true) : ConvertHelper_URLFinder |
108
|
|
|
{ |
109
|
|
|
$this->setOption('normalize', $enabled); |
110
|
|
|
return $this; |
111
|
|
|
} |
112
|
|
|
|
113
|
|
|
/** |
114
|
|
|
* Whether to enable sorting the URLs alphabetically (disabled by default). |
115
|
|
|
* |
116
|
|
|
* @param bool $enabled |
117
|
|
|
* @return $this |
118
|
|
|
*/ |
119
|
|
|
public function enableSorting(bool $enabled=true) : ConvertHelper_URLFinder |
120
|
|
|
{ |
121
|
|
|
$this->setOption('sorting', $enabled); |
122
|
|
|
return $this; |
123
|
|
|
} |
124
|
|
|
|
125
|
|
|
/** |
126
|
|
|
* Whether to include email addresses in the search. |
127
|
|
|
* This is only relevant when using the getURLs() |
128
|
|
|
* method. |
129
|
|
|
* |
130
|
|
|
* @param bool $include |
131
|
|
|
* @return ConvertHelper_URLFinder |
132
|
|
|
*/ |
133
|
|
|
public function includeEmails(bool $include=true) : ConvertHelper_URLFinder |
134
|
|
|
{ |
135
|
|
|
$this->setOption('include-emails', $include); |
136
|
|
|
return $this; |
137
|
|
|
} |
138
|
|
|
|
139
|
|
|
/** |
140
|
|
|
* Whether to omit the mailto: that is automatically added to all email addresses. |
141
|
|
|
* |
142
|
|
|
* @param bool $omit |
143
|
|
|
* @return ConvertHelper_URLFinder |
144
|
|
|
*/ |
145
|
|
|
public function omitMailto(bool $omit=true) : ConvertHelper_URLFinder |
146
|
|
|
{ |
147
|
|
|
$this->setOption('omit-mailto', $omit); |
148
|
|
|
return $this; |
149
|
|
|
} |
150
|
|
|
|
151
|
|
|
/** |
152
|
|
|
* Splits the string by a list of word boundaries, so that all relevant |
153
|
|
|
* words are separated into individual lines. Each line is then checked |
154
|
|
|
* to keep only strings that are more or less likely to contain a domain name. |
155
|
|
|
* |
156
|
|
|
* @param string $subject |
157
|
|
|
* @return string[] |
158
|
|
|
*/ |
159
|
|
|
private function splitSubject(string $subject) : array |
160
|
|
|
{ |
161
|
|
|
$subject = str_replace($this->boundaries, ' ', $subject); |
162
|
|
|
$lines = ConvertHelper::explodeTrim(' ', $subject); |
163
|
|
|
|
164
|
|
|
unset($subject); |
165
|
|
|
|
166
|
|
|
$keep = array(); |
167
|
|
|
|
168
|
|
|
foreach ($lines as $line) |
169
|
|
|
{ |
170
|
|
|
$line = $this->analyzeLine($line); |
171
|
|
|
|
172
|
|
|
if($line !== null) { |
173
|
|
|
$keep[] = $line; |
174
|
|
|
} |
175
|
|
|
} |
176
|
|
|
|
177
|
|
|
return array_unique($keep); |
178
|
|
|
} |
179
|
|
|
|
180
|
|
|
/** |
181
|
|
|
* Analyzes a single line to see if it is likely to contain a domain name. |
182
|
|
|
* |
183
|
|
|
* @param string $line |
184
|
|
|
* @return string|null |
185
|
|
|
*/ |
186
|
|
|
private function analyzeLine(string $line) : ?string |
187
|
|
|
{ |
188
|
|
|
// Strip punctuation from the beginning and end, |
189
|
|
|
// to exclude the end of phrases, e.g. "domain.com." |
190
|
|
|
$line = trim($line, '.'); |
191
|
|
|
|
192
|
|
|
// Handle detecting an URI scheme |
193
|
|
|
if(strstr($line, ':') !== false) |
194
|
|
|
{ |
195
|
|
|
$scheme = URLInfo_Schemes::detectScheme($line); |
196
|
|
|
|
197
|
|
|
if ($scheme !== null) |
198
|
|
|
{ |
199
|
|
|
return $line; |
200
|
|
|
} |
201
|
|
|
} |
202
|
|
|
|
203
|
|
|
// From here on out, the only things we can still |
204
|
|
|
// detect are email addresses and domain names. |
205
|
|
|
|
206
|
|
|
// No dot? Then it's certainly not a domain name. |
207
|
|
|
if(strstr($line, '.') === false) { |
208
|
|
|
return null; |
209
|
|
|
} |
210
|
|
|
|
211
|
|
|
// Check again if there are still dots present |
212
|
|
|
if(strstr($line, '.') !== false) { |
213
|
|
|
return $line; |
214
|
|
|
} |
215
|
|
|
|
216
|
|
|
return null; |
217
|
|
|
} |
218
|
|
|
|
219
|
|
|
/** |
220
|
|
|
* Filters the subject string before trying to detect regular HTTP/HTTPS |
221
|
|
|
* URLs as well as email addresses that are domain-based. |
222
|
|
|
* |
223
|
|
|
* @param string $subject |
224
|
|
|
* @return string |
225
|
|
|
*/ |
226
|
|
|
private function filterSubjectBefore(string $subject) : string |
227
|
|
|
{ |
228
|
|
|
$subject = stripslashes($subject); |
229
|
|
|
|
230
|
|
|
foreach($this->detectors as $detector) |
231
|
|
|
{ |
232
|
|
|
// Avoid processing the string if it is not needed. |
233
|
|
|
if($detector->getRunPosition() !== ConvertHelper_URLFinder_Detector::RUN_BEFORE || !$detector->isValidFor($subject)) { |
234
|
|
|
continue; |
235
|
|
|
} |
236
|
|
|
|
237
|
|
|
$subject = $detector->processString($subject); |
238
|
|
|
|
239
|
|
|
$this->matches = array_merge($this->matches, $detector->getMatches()); |
240
|
|
|
} |
241
|
|
|
|
242
|
|
|
return $subject; |
243
|
|
|
} |
244
|
|
|
|
245
|
|
|
/** |
246
|
|
|
* @param string $className |
247
|
|
|
* @return ConvertHelper_URLFinder_Detector |
248
|
|
|
* @throws ConvertHelper_Exception |
249
|
|
|
*/ |
250
|
|
|
private function createDetector(string $className) : ConvertHelper_URLFinder_Detector |
251
|
|
|
{ |
252
|
|
|
$detector = new $className(); |
253
|
|
|
|
254
|
|
|
if($detector instanceof ConvertHelper_URLFinder_Detector) |
255
|
|
|
{ |
256
|
|
|
return $detector; |
257
|
|
|
} |
258
|
|
|
|
259
|
|
|
throw new ConvertHelper_Exception( |
260
|
|
|
'Not a valid detector class.', |
261
|
|
|
sprintf( |
262
|
|
|
'The class [%s] is not an instance of [%s].', |
263
|
|
|
$className, |
264
|
|
|
ConvertHelper_URLFinder_Detector::class |
265
|
|
|
), |
266
|
|
|
self::ERROR_INVALID_DETECTOR_CLASS |
267
|
|
|
); |
268
|
|
|
} |
269
|
|
|
|
270
|
|
|
/** |
271
|
|
|
* Fetches all URLs that can be found in the subject string. |
272
|
|
|
* |
273
|
|
|
* @return string[] |
274
|
|
|
*/ |
275
|
|
|
public function getURLs() : array |
276
|
|
|
{ |
277
|
|
|
$this->parse(); |
278
|
|
|
|
279
|
|
|
$result = $this->getItemsAsString($this->urls); |
280
|
|
|
|
281
|
|
|
if($this->getBoolOption('include-emails')) |
282
|
|
|
{ |
283
|
|
|
$result = array_merge($result, $this->getEmails()); |
284
|
|
|
} |
285
|
|
|
|
286
|
|
|
if($this->getBoolOption('sorting')) |
287
|
|
|
{ |
288
|
|
|
usort($result, function(string $a, string $b) { |
289
|
|
|
return strnatcasecmp($a, $b); |
290
|
|
|
}); |
291
|
|
|
} |
292
|
|
|
|
293
|
|
|
return $result; |
294
|
|
|
} |
295
|
|
|
|
296
|
|
|
/** |
297
|
|
|
* @param array<string,URLInfo> $collection |
298
|
|
|
* @return string[] |
299
|
|
|
*/ |
300
|
|
|
private function getItemsAsString(array $collection) : array |
301
|
|
|
{ |
302
|
|
|
$normalize = $this->getBoolOption('normalize'); |
303
|
|
|
|
304
|
|
|
$result = array(); |
305
|
|
|
|
306
|
|
|
foreach($collection as $url => $info) { |
307
|
|
|
if($normalize) { |
308
|
|
|
$url = $info->getNormalized(); |
309
|
|
|
} |
310
|
|
|
|
311
|
|
|
if(!in_array($url, $result)) { |
312
|
|
|
$result[] = $url; |
313
|
|
|
} |
314
|
|
|
} |
315
|
|
|
|
316
|
|
|
return $result; |
317
|
|
|
} |
318
|
|
|
|
319
|
|
|
/** |
320
|
|
|
* Instantiates the selected detector classes, which are |
321
|
|
|
* used to detect specific elements in the target string |
322
|
|
|
* (beyond regular URLs and Email addresses). |
323
|
|
|
* |
324
|
|
|
* @throws ConvertHelper_Exception |
325
|
|
|
*/ |
326
|
|
|
private function initDetectors() : void |
327
|
|
|
{ |
328
|
|
|
foreach($this->enabledDetectorClasses as $className => $enabled) |
329
|
|
|
{ |
330
|
|
|
if($enabled) { |
331
|
|
|
$this->detectors[] = $this->createDetector($className); |
332
|
|
|
} |
333
|
|
|
} |
334
|
|
|
} |
335
|
|
|
|
336
|
|
|
/** |
337
|
|
|
* Parses the specified string to detect all URLs and Email addresses. |
338
|
|
|
* For accurate results, this does not use a regex, but splits the |
339
|
|
|
* string into a list of strings that are likely to be either an URL |
340
|
|
|
* or Email address. Each of these is then checked for a valid scheme |
341
|
|
|
* or domain name extension. |
342
|
|
|
*/ |
343
|
|
|
private function parse() : void |
344
|
|
|
{ |
345
|
|
|
if($this->parsed) { |
346
|
|
|
return; |
347
|
|
|
} |
348
|
|
|
|
349
|
|
|
$this->parsed = true; |
350
|
|
|
|
351
|
|
|
$this->initDetectors(); |
352
|
|
|
$this->detectMatches($this->subject); |
353
|
|
|
|
354
|
|
|
unset($this->subject); |
355
|
|
|
|
356
|
|
|
foreach($this->matches as $match) |
357
|
|
|
{ |
358
|
|
|
$info = parseURL($match); |
359
|
|
|
|
360
|
|
|
if($info->isEmail()) |
361
|
|
|
{ |
362
|
|
|
$this->emails[$this->filterEmailAddress($match)] = $info; |
363
|
|
|
continue; |
364
|
|
|
} |
365
|
|
|
|
366
|
|
|
$this->urls[$match] = $info; |
367
|
|
|
} |
368
|
|
|
} |
369
|
|
|
|
370
|
|
|
/** |
371
|
|
|
* Enables the search for relative URLs in HTML attributes. |
372
|
|
|
* |
373
|
|
|
* @param bool $enable |
374
|
|
|
* @return $this |
375
|
|
|
*/ |
376
|
|
|
public function enableHTMLAttributes(bool $enable=true) : ConvertHelper_URLFinder |
377
|
|
|
{ |
378
|
|
|
$this->enabledDetectorClasses[ConvertHelper_URLFinder_Detector_HTMLAttributes::class] = $enable; |
379
|
|
|
return $this; |
380
|
|
|
} |
381
|
|
|
|
382
|
|
|
/** |
383
|
|
|
* Ensures that the email address has the `mailto:` scheme prepended, |
384
|
|
|
* and lowercases it to avoid case mixups. |
385
|
|
|
* |
386
|
|
|
* @param string $email |
387
|
|
|
* @return string |
388
|
|
|
*/ |
389
|
|
|
private function filterEmailAddress(string $email) : string |
390
|
|
|
{ |
391
|
|
|
if(stristr($email, 'mailto:') === false) { |
392
|
|
|
$email = 'mailto:'.$email; |
393
|
|
|
} |
394
|
|
|
|
395
|
|
|
return strtolower($email); |
396
|
|
|
} |
397
|
|
|
|
398
|
|
|
/** |
399
|
|
|
* Detects all URL and Email matches in the specified string. |
400
|
|
|
* |
401
|
|
|
* @param string $subject |
402
|
|
|
*/ |
403
|
|
|
private function detectMatches(string $subject) : void |
404
|
|
|
{ |
405
|
|
|
$subject = $this->filterSubjectBefore($subject); |
406
|
|
|
|
407
|
|
|
$lines = $this->splitSubject($subject); |
408
|
|
|
$domains = new ConvertHelper_URLFinder_DomainExtensions(); |
409
|
|
|
|
410
|
|
|
foreach ($lines as $line) |
411
|
|
|
{ |
412
|
|
|
$scheme = URLInfo_Schemes::detectScheme($line); |
413
|
|
|
if($scheme !== null) { |
414
|
|
|
$this->matches[] = $line; |
415
|
|
|
continue; |
416
|
|
|
} |
417
|
|
|
|
418
|
|
|
$extension = $this->detectDomainExtension($line); |
419
|
|
|
|
420
|
|
|
if($domains->nameExists($extension)) { |
421
|
|
|
$this->matches[] = $line; |
422
|
|
|
} |
423
|
|
|
} |
424
|
|
|
|
425
|
|
|
$this->filterSubjectAfter($subject); |
426
|
|
|
} |
427
|
|
|
|
428
|
|
|
private function filterSubjectAfter(string $subject) : void |
429
|
|
|
{ |
430
|
|
|
// Sort the matches from longest to shortest, to avoid |
431
|
|
|
// replacing parts of URLs. |
432
|
|
|
$remove = $this->matches; |
433
|
|
|
usort($remove, function(string $a, string $b) { |
434
|
|
|
return strlen($b) - strlen($a); |
435
|
|
|
}); |
436
|
|
|
|
437
|
|
|
$subject = str_replace($remove, ' ', $subject); |
438
|
|
|
|
439
|
|
|
foreach($this->detectors as $detector) |
440
|
|
|
{ |
441
|
|
|
if($detector->getRunPosition() !== ConvertHelper_URLFinder_Detector::RUN_AFTER || !$detector->isValidFor($subject)) { |
442
|
|
|
continue; |
443
|
|
|
} |
444
|
|
|
|
445
|
|
|
$subject = $detector->processString($subject); |
446
|
|
|
|
447
|
|
|
$this->matches = array_merge($this->matches, $detector->getMatches()); |
448
|
|
|
} |
449
|
|
|
} |
450
|
|
|
|
451
|
|
|
/** |
452
|
|
|
* Attempts to extract a valid domain name extension from |
453
|
|
|
* the specified URL. |
454
|
|
|
* |
455
|
|
|
* @param string $url |
456
|
|
|
* @return string |
457
|
|
|
* @see ConvertHelper_URLFinder_DomainExtensions |
458
|
|
|
*/ |
459
|
|
|
private function detectDomainExtension(string $url) : string |
460
|
|
|
{ |
461
|
|
|
$boundaries = array('/', '?'); |
462
|
|
|
|
463
|
|
|
// Remove the path or query parts to access the domain extension only |
464
|
|
|
foreach($boundaries as $boundary) { |
465
|
|
|
if(strstr($url, $boundary)) { |
466
|
|
|
$parts = explode($boundary, $url); |
467
|
|
|
$url = array_shift($parts); |
468
|
|
|
break; |
469
|
|
|
} |
470
|
|
|
} |
471
|
|
|
|
472
|
|
|
$parts = explode('.', $url); |
473
|
|
|
|
474
|
|
|
return array_pop($parts); |
475
|
|
|
} |
476
|
|
|
|
477
|
|
|
/** |
478
|
|
|
* Retrieves all email addresses from the subject string. |
479
|
|
|
* |
480
|
|
|
* @return string[] |
481
|
|
|
* |
482
|
|
|
* @see omitMailto() |
483
|
|
|
*/ |
484
|
|
|
public function getEmails() : array |
485
|
|
|
{ |
486
|
|
|
$this->parse(); |
487
|
|
|
|
488
|
|
|
$result = $this->getItemsAsString($this->emails); |
489
|
|
|
|
490
|
|
|
if($this->getBoolOption('omit-mailto')) { |
491
|
|
|
$keep = array(); |
492
|
|
|
foreach($result as $email) { |
493
|
|
|
$keep[] = str_replace('mailto:', '', $email); |
494
|
|
|
} |
495
|
|
|
|
496
|
|
|
$result = $keep; |
497
|
|
|
} |
498
|
|
|
|
499
|
|
|
if($this->getBoolOption('sorting')) |
500
|
|
|
{ |
501
|
|
|
usort($result, function(string $a, string $b) { |
502
|
|
|
return strnatcasecmp($a, $b); |
503
|
|
|
}); |
504
|
|
|
} |
505
|
|
|
|
506
|
|
|
return $result; |
507
|
|
|
} |
508
|
|
|
|
509
|
|
|
/** |
510
|
|
|
* Retrieves all URLs as URLInfo instances. |
511
|
|
|
* |
512
|
|
|
* @return URLInfo[] |
513
|
|
|
*/ |
514
|
|
|
public function getInfos() : array |
515
|
|
|
{ |
516
|
|
|
$this->parse(); |
517
|
|
|
|
518
|
|
|
$result = array(); |
519
|
|
|
$normalize = $this->getBoolOption('normalize'); |
520
|
|
|
|
521
|
|
|
foreach($this->urls as $url => $info) |
522
|
|
|
{ |
523
|
|
|
if($normalize) { |
524
|
|
|
$url = $info->getNormalized(); |
525
|
|
|
} |
526
|
|
|
|
527
|
|
|
$result[$url] = $info; |
528
|
|
|
} |
529
|
|
|
|
530
|
|
|
if($this->getBoolOption('sorting')) |
531
|
|
|
{ |
532
|
|
|
ksort($result); |
533
|
|
|
} |
534
|
|
|
|
535
|
|
|
return array_values($result); |
536
|
|
|
} |
537
|
|
|
} |
538
|
|
|
|