1
|
|
|
<?php |
2
|
|
|
namespace vipnytt\RobotsTxtParser; |
3
|
|
|
|
4
|
|
|
use vipnytt\RobotsTxtParser\Exceptions\TxtParserException; |
5
|
|
|
|
6
|
|
|
/** |
7
|
|
|
* Class TxtParser |
8
|
|
|
* |
9
|
|
|
* @package vipnytt\RobotsTxtParser |
10
|
|
|
*/ |
11
|
|
|
class TxtParser |
12
|
|
|
{ |
13
|
|
|
/** |
14
|
|
|
* Robots.txt max length in bytes |
15
|
|
|
*/ |
16
|
|
|
const DEFAULT_BYTE_LIMIT = 500000; |
17
|
|
|
|
18
|
|
|
/** |
19
|
|
|
* Max rule length |
20
|
|
|
*/ |
21
|
|
|
const RULE_MAX_LENGTH = 500; |
22
|
|
|
|
23
|
|
|
/** |
24
|
|
|
* Directives |
25
|
|
|
*/ |
26
|
|
|
const DIRECTIVE_ALLOW = 'allow'; |
27
|
|
|
const DIRECTIVE_CACHE_DELAY = 'cache-delay'; // unofficial |
28
|
|
|
const DIRECTIVE_CLEAN_PARAM = 'clean-param'; // Yandex only |
29
|
|
|
const DIRECTIVE_CRAWL_DELAY = 'crawl-delay'; |
30
|
|
|
const DIRECTIVE_DISALLOW = 'disallow'; |
31
|
|
|
const DIRECTIVE_HOST = 'host'; // Yandex only |
32
|
|
|
const DIRECTIVE_SITEMAP = 'sitemap'; |
33
|
|
|
const DIRECTIVE_USER_AGENT = 'user-agent'; |
34
|
|
|
|
35
|
|
|
/** |
36
|
|
|
* Default User-Agent |
37
|
|
|
*/ |
38
|
|
|
const FALLBACK_USER_AGENT = '*'; |
39
|
|
|
|
40
|
|
|
/** |
41
|
|
|
* RAW robots.txt content |
42
|
|
|
* @var string |
43
|
|
|
*/ |
44
|
|
|
private $raw = ''; |
45
|
|
|
|
46
|
|
|
/** |
47
|
|
|
* Rule array |
48
|
|
|
* @var array |
49
|
|
|
*/ |
50
|
|
|
private $rules = []; |
51
|
|
|
|
52
|
|
|
/** |
53
|
|
|
* User-Agents |
54
|
|
|
* @var array |
55
|
|
|
*/ |
56
|
|
|
private $userAgents = [self::FALLBACK_USER_AGENT]; |
57
|
|
|
|
58
|
|
|
/** |
59
|
|
|
* Current line |
60
|
|
|
* @var string |
61
|
|
|
*/ |
62
|
|
|
private $line = ''; |
63
|
|
|
|
64
|
|
|
/** |
65
|
|
|
* Previous directive |
66
|
|
|
* @var string |
67
|
|
|
*/ |
68
|
|
|
private $previous; |
69
|
|
|
|
70
|
|
|
/** |
71
|
|
|
* Current Directive |
72
|
|
|
* @var string |
73
|
|
|
*/ |
74
|
|
|
private $directive; |
75
|
|
|
|
76
|
|
|
/** |
77
|
|
|
* Current Rule |
78
|
|
|
* @var array|string |
79
|
|
|
*/ |
80
|
|
|
private $rule; |
81
|
|
|
|
82
|
|
|
/** |
83
|
|
|
* Constructor |
84
|
|
|
* |
85
|
|
|
* @param string $content - file content |
86
|
|
|
* @param string|null $encoding - character encoding |
87
|
|
|
* @param int|null $byteLimit - maximum of bytes to parse |
88
|
|
|
* @throws TxtParserException |
89
|
|
|
*/ |
90
|
|
|
public function __construct($content, $encoding = null, $byteLimit = self::DEFAULT_BYTE_LIMIT) |
91
|
|
|
{ |
92
|
|
|
if ($encoding === null) { |
93
|
|
|
$encoding = mb_detect_encoding($content); |
94
|
|
|
} |
95
|
|
|
if (!mb_internal_encoding($encoding)) { |
96
|
|
|
throw new TxtParserException('Unable to set internal character encoding to `' . $encoding . '`'); |
97
|
|
|
} |
98
|
|
|
|
99
|
|
|
$this->raw = is_int($byteLimit) ? mb_strcut($content, 0, $byteLimit, $encoding) : $content; |
100
|
|
|
$this->parseTxt(); |
101
|
|
|
} |
102
|
|
|
|
103
|
|
|
/** |
104
|
|
|
* Parse robots.txt |
105
|
|
|
* |
106
|
|
|
* @return void |
107
|
|
|
*/ |
108
|
|
|
private function parseTxt() |
109
|
|
|
{ |
110
|
|
|
$lines = array_filter(array_map('trim', mb_split('\n', $this->raw))); |
111
|
|
|
// Parse each line individually |
112
|
|
|
foreach ($lines as $this->line) { |
113
|
|
|
// Limit rule length and remove comments |
114
|
|
|
$this->line = mb_split('#', mb_substr($this->line, 0, self::RULE_MAX_LENGTH), 2)[0]; |
115
|
|
|
// Parse line |
116
|
|
|
if ( |
117
|
|
|
($this->generateRulePair()) === false |
118
|
|
|
|| ($result = $this->parseLine()) === false |
119
|
|
|
) { |
120
|
|
|
continue; |
121
|
|
|
} |
122
|
|
|
// Add rule |
123
|
|
|
$this->previous = $this->directive; |
124
|
|
|
$this->rule = $result; |
125
|
|
|
$this->rules = array_merge_recursive($this->assignUserAgent(), $this->rules); |
126
|
|
|
} |
127
|
|
|
} |
128
|
|
|
|
129
|
|
|
/** |
130
|
|
|
* Generate Directive:Rule pair |
131
|
|
|
* |
132
|
|
|
* @return bool |
133
|
|
|
*/ |
134
|
|
|
private function generateRulePair() |
135
|
|
|
{ |
136
|
|
|
// Split by directive and rule |
137
|
|
|
$pair = array_map('trim', mb_split(':', $this->line, 2)); |
138
|
|
|
// Check if the line contains a rule |
139
|
|
|
if ( |
140
|
|
|
empty($pair[1]) |
141
|
|
|
|| empty($pair[0]) |
142
|
|
|
|| !in_array(($pair[0] = mb_strtolower($pair[0])), $this->directives()) |
143
|
|
|
) { |
144
|
|
|
// Line does not contain any supported directive |
145
|
|
|
return false; |
146
|
|
|
} |
147
|
|
|
$this->directive = $pair[0]; |
148
|
|
|
$this->rule = $pair[1]; |
149
|
|
|
return true; |
150
|
|
|
} |
151
|
|
|
|
152
|
|
|
/** |
153
|
|
|
* Directives and sub directives |
154
|
|
|
* |
155
|
|
|
* @param string|null $parent |
156
|
|
|
* @return array |
157
|
|
|
*/ |
158
|
|
|
private function directives($parent = null) |
159
|
|
|
{ |
160
|
|
|
$array = [ |
161
|
|
|
self::DIRECTIVE_ALLOW => [ |
162
|
|
|
self::DIRECTIVE_CLEAN_PARAM, |
163
|
|
|
self::DIRECTIVE_HOST, |
164
|
|
|
], |
165
|
|
|
self::DIRECTIVE_CACHE_DELAY => [], |
166
|
|
|
self::DIRECTIVE_CLEAN_PARAM => [], |
167
|
|
|
self::DIRECTIVE_CRAWL_DELAY => [], |
168
|
|
|
self::DIRECTIVE_DISALLOW => [ |
169
|
|
|
self::DIRECTIVE_CLEAN_PARAM, |
170
|
|
|
self::DIRECTIVE_HOST, |
171
|
|
|
], |
172
|
|
|
self::DIRECTIVE_HOST => [], |
173
|
|
|
self::DIRECTIVE_SITEMAP => [], |
174
|
|
|
self::DIRECTIVE_USER_AGENT => [ |
175
|
|
|
self::DIRECTIVE_ALLOW, |
176
|
|
|
self::DIRECTIVE_CACHE_DELAY, |
177
|
|
|
self::DIRECTIVE_CRAWL_DELAY, |
178
|
|
|
self::DIRECTIVE_DISALLOW, |
179
|
|
|
], |
180
|
|
|
]; |
181
|
|
|
if ($parent !== null) { |
182
|
|
|
return isset($array[$parent]) ? $array[$parent] : []; |
183
|
|
|
} |
184
|
|
|
return array_keys($array); |
185
|
|
|
} |
186
|
|
|
|
187
|
|
|
/** |
188
|
|
|
* Parse line |
189
|
|
|
* |
190
|
|
|
* @param string|null $parent |
191
|
|
|
* @return array|false |
192
|
|
|
*/ |
193
|
|
|
private function parseLine($parent = null) |
194
|
|
|
{ |
195
|
|
|
if ( |
196
|
|
|
($this->generateRulePair()) === false |
197
|
|
|
|| !in_array($this->directive, $this->directives($parent)) |
198
|
|
|
) { |
199
|
|
|
return false; |
200
|
|
|
} |
201
|
|
|
// Cache directive/rule variables to after inline directives has been parsed |
202
|
|
|
$directive = $this->directive; |
203
|
|
|
$rule = $this->rule; |
204
|
|
|
$this->line = (string)$this->rule; |
205
|
|
|
if (($inline = $this->parseLine($this->directive)) !== false) { |
206
|
|
|
$rule = $inline; |
207
|
|
|
}; |
208
|
|
|
$this->directive = $directive; |
209
|
|
|
$this->rule = $rule; |
210
|
|
|
return $this->add(); |
211
|
|
|
} |
212
|
|
|
|
213
|
|
|
/** |
214
|
|
|
* Add value to directive |
215
|
|
|
* |
216
|
|
|
* @return array|false |
217
|
|
|
*/ |
218
|
|
|
private function add() |
219
|
|
|
{ |
220
|
|
|
switch ($this->directive) { |
221
|
|
|
case self::DIRECTIVE_ALLOW: |
222
|
|
|
case self::DIRECTIVE_DISALLOW: |
223
|
|
|
return $this->addDisAllow(); |
224
|
|
|
case self::DIRECTIVE_CACHE_DELAY: |
225
|
|
|
case self::DIRECTIVE_CRAWL_DELAY: |
226
|
|
|
return $this->addFloat(); |
227
|
|
|
case self::DIRECTIVE_CLEAN_PARAM: |
228
|
|
|
return $this->addCleanParam(); |
229
|
|
|
case self::DIRECTIVE_HOST: |
230
|
|
|
return $this->addHost(); |
231
|
|
|
case self::DIRECTIVE_SITEMAP: |
232
|
|
|
return $this->addSitemap(); |
233
|
|
|
case self::DIRECTIVE_USER_AGENT: |
234
|
|
|
return $this->setUserAgent(); |
235
|
|
|
} |
236
|
|
|
return false; |
237
|
|
|
} |
238
|
|
|
|
239
|
|
|
/** |
240
|
|
|
* Add an Allow or Disallow rule |
241
|
|
|
* |
242
|
|
|
* @return array |
243
|
|
|
*/ |
244
|
|
|
private function addDisAllow() |
245
|
|
|
{ |
246
|
|
|
// If inline directive, pass the array |
247
|
|
|
if (is_array($this->rule)) { |
248
|
|
|
return [ |
249
|
|
|
$this->directive => $this->rule |
250
|
|
|
]; |
251
|
|
|
} |
252
|
|
|
// Return an array of paths |
253
|
|
|
return [ |
254
|
|
|
$this->directive => [ |
255
|
|
|
'path' => [ |
256
|
|
|
$this->rule |
257
|
|
|
] |
258
|
|
|
] |
259
|
|
|
]; |
260
|
|
|
} |
261
|
|
|
|
262
|
|
|
/** |
263
|
|
|
* Add float value |
264
|
|
|
* |
265
|
|
|
* @return array|false |
266
|
|
|
*/ |
267
|
|
|
private function addFloat() |
268
|
|
|
{ |
269
|
|
|
if (empty(($float = floatval($this->rule)))) { |
270
|
|
|
return false; |
271
|
|
|
} |
272
|
|
|
return [ |
273
|
|
|
$this->directive => $float, |
274
|
|
|
]; |
275
|
|
|
} |
276
|
|
|
|
277
|
|
|
/** |
278
|
|
|
* Add Clean-Param record |
279
|
|
|
* |
280
|
|
|
* @return array|false |
281
|
|
|
*/ |
282
|
|
|
private function addCleanParam() |
283
|
|
|
{ |
284
|
|
|
if (!is_string($this->rule)) { |
285
|
|
|
return false; |
286
|
|
|
} |
287
|
|
|
$result = []; |
288
|
|
|
$cleanParam = $this->explodeCleanParamRule($this->rule); |
289
|
|
|
foreach ($cleanParam['param'] as $param) { |
290
|
|
|
$result[$this->directive]['path'][$cleanParam['path']]['param'][] = $param; |
291
|
|
|
} |
292
|
|
|
return $result; |
293
|
|
|
} |
294
|
|
|
|
295
|
|
|
/** |
296
|
|
|
* Explode Clean-Param rule |
297
|
|
|
* |
298
|
|
|
* @param string $rule |
299
|
|
|
* @return array |
300
|
|
|
*/ |
301
|
|
|
private function explodeCleanParamRule($rule) |
302
|
|
|
{ |
303
|
|
|
// split into parameter and path |
304
|
|
|
$array = array_map('trim', mb_split('\s+', $rule, 2)); |
305
|
|
|
$cleanParam = []; |
306
|
|
|
// strip any invalid characters from path prefix |
307
|
|
|
$cleanParam['path'] = isset($array[1]) ? $this->urlEncode(mb_ereg_replace('[^A-Za-z0-9\.-\/\*\_]', '', $array[1])) : "/*"; |
308
|
|
|
$param = array_map('trim', mb_split('&', $array[0])); |
309
|
|
|
foreach ($param as $key) { |
310
|
|
|
$cleanParam['param'][] = $key; |
311
|
|
|
} |
312
|
|
|
return $cleanParam; |
313
|
|
|
} |
314
|
|
|
|
315
|
|
|
/** |
316
|
|
|
* URL encoder according to RFC 3986 |
317
|
|
|
* Returns a string containing the encoded URL with disallowed characters converted to their percentage encodings. |
318
|
|
|
* @link http://publicmind.in/blog/url-encoding/ |
319
|
|
|
* |
320
|
|
|
* @param string $url |
321
|
|
|
* @return string |
322
|
|
|
*/ |
323
|
|
|
private function urlEncode($url) |
324
|
|
|
{ |
325
|
|
|
$reserved = [ |
326
|
|
|
":" => '!%3A!ui', |
327
|
|
|
"/" => '!%2F!ui', |
328
|
|
|
"?" => '!%3F!ui', |
329
|
|
|
"#" => '!%23!ui', |
330
|
|
|
"[" => '!%5B!ui', |
331
|
|
|
"]" => '!%5D!ui', |
332
|
|
|
"@" => '!%40!ui', |
333
|
|
|
"!" => '!%21!ui', |
334
|
|
|
"$" => '!%24!ui', |
335
|
|
|
"&" => '!%26!ui', |
336
|
|
|
"'" => '!%27!ui', |
337
|
|
|
"(" => '!%28!ui', |
338
|
|
|
")" => '!%29!ui', |
339
|
|
|
"*" => '!%2A!ui', |
340
|
|
|
"+" => '!%2B!ui', |
341
|
|
|
"," => '!%2C!ui', |
342
|
|
|
";" => '!%3B!ui', |
343
|
|
|
"=" => '!%3D!ui', |
344
|
|
|
"%" => '!%25!ui' |
345
|
|
|
]; |
346
|
|
|
return preg_replace(array_values($reserved), array_keys($reserved), rawurlencode($url)); |
347
|
|
|
} |
348
|
|
|
|
349
|
|
|
/** |
350
|
|
|
* Add Host |
351
|
|
|
* |
352
|
|
|
* @return array|false |
353
|
|
|
*/ |
354
|
|
|
private function addHost() |
355
|
|
|
{ |
356
|
|
|
if ( |
357
|
|
|
!is_string($this->rule) |
358
|
|
|
|| ($parsed = parse_url(($this->rule = $this->urlEncode($this->rule)))) === false |
359
|
|
|
) { |
360
|
|
|
return false; |
361
|
|
|
} |
362
|
|
|
$host = isset($parsed['host']) ? $parsed['host'] : $parsed['path']; |
363
|
|
|
if ( |
364
|
|
|
!$this->urlValidateHost($host) |
365
|
|
|
|| isset($parsed['scheme']) && !$this->urlValidateScheme($parsed['scheme']) |
366
|
|
|
) { |
367
|
|
|
return false; |
368
|
|
|
} |
369
|
|
|
$scheme = isset($parsed['scheme']) ? $parsed['scheme'] . '://' : ''; |
370
|
|
|
$port = isset($parsed['port']) ? ':' . $parsed['port'] : ''; |
371
|
|
|
return [ |
372
|
|
|
self::DIRECTIVE_HOST => [ |
373
|
|
|
$scheme . $host . $port, |
374
|
|
|
] |
375
|
|
|
]; |
376
|
|
|
} |
377
|
|
|
|
378
|
|
|
/** |
379
|
|
|
* Validate host name |
380
|
|
|
* |
381
|
|
|
* @link http://stackoverflow.com/questions/1755144/how-to-validate-domain-name-in-php |
382
|
|
|
* |
383
|
|
|
* @param string $host |
384
|
|
|
* @return bool |
385
|
|
|
*/ |
386
|
|
|
private static function urlValidateHost($host) |
387
|
|
|
{ |
388
|
|
|
return ( |
389
|
|
|
mb_ereg_match("/^([a-z\d](-*[a-z\d])*)(\.([a-z\d](-*[a-z\d])*))*$/i", $host) //valid chars check |
390
|
|
|
&& mb_ereg_match("/^.{1,253}$/", $host) //overall length check |
391
|
|
|
&& mb_ereg_match("/^[^\.]{1,63}(\.[^\.]{1,63})*$/", $host) //length of each label |
392
|
|
|
&& !filter_var($host, FILTER_VALIDATE_IP) //is not an IP address |
393
|
|
|
); |
394
|
|
|
} |
395
|
|
|
|
396
|
|
|
/** |
397
|
|
|
* Validate URL scheme |
398
|
|
|
* |
399
|
|
|
* @param string $scheme |
400
|
|
|
* @return bool |
401
|
|
|
*/ |
402
|
|
|
private static function urlValidateScheme($scheme) |
403
|
|
|
{ |
404
|
|
|
return in_array($scheme, [ |
405
|
|
|
'http', 'https', |
406
|
|
|
'ftp', 'sftp' |
407
|
|
|
] |
408
|
|
|
); |
409
|
|
|
} |
410
|
|
|
|
411
|
|
|
/** |
412
|
|
|
* Add Sitemap |
413
|
|
|
* |
414
|
|
|
* @return array|false |
415
|
|
|
*/ |
416
|
|
|
private function addSitemap() |
417
|
|
|
{ |
418
|
|
|
if ( |
419
|
|
|
!is_string($this->rule) |
420
|
|
|
|| !$this->urlValidate(($url = $this->urlEncode($this->rule))) |
421
|
|
|
) { |
422
|
|
|
return false; |
423
|
|
|
} |
424
|
|
|
return [ |
425
|
|
|
self::DIRECTIVE_SITEMAP => [ |
426
|
|
|
$url |
427
|
|
|
] |
428
|
|
|
]; |
429
|
|
|
} |
430
|
|
|
|
431
|
|
|
/** |
432
|
|
|
* Validate URL |
433
|
|
|
* |
434
|
|
|
* @param string $url |
435
|
|
|
* @return bool |
436
|
|
|
*/ |
437
|
|
|
public function urlValidate($url) |
438
|
|
|
{ |
439
|
|
|
return ( |
440
|
|
|
filter_var($url, FILTER_VALIDATE_URL) |
441
|
|
|
&& ($parsed = parse_url($url)) !== false |
442
|
|
|
&& $this->urlValidateHost($parsed['host']) |
443
|
|
|
&& $this->urlValidateScheme($parsed['scheme']) |
444
|
|
|
); |
445
|
|
|
} |
446
|
|
|
|
447
|
|
|
/** |
448
|
|
|
* Set User-Agent(s) |
449
|
|
|
* |
450
|
|
|
* @return array |
451
|
|
|
*/ |
452
|
|
|
private function setUserAgent() |
453
|
|
|
{ |
454
|
|
|
switch ($this->previous) { |
455
|
|
|
case self::DIRECTIVE_USER_AGENT: |
456
|
|
|
$this->userAgents[] = $this->rule; |
457
|
|
|
break; |
458
|
|
|
default: |
459
|
|
|
$this->userAgents = [ |
460
|
|
|
$this->rule |
461
|
|
|
]; |
462
|
|
|
} |
463
|
|
|
return []; |
464
|
|
|
} |
465
|
|
|
|
466
|
|
|
/** |
467
|
|
|
* Assign User-Agent dependent rules to the User-Agent arrays |
468
|
|
|
* |
469
|
|
|
* @return array |
470
|
|
|
*/ |
471
|
|
|
private function assignUserAgent() |
472
|
|
|
{ |
473
|
|
|
if (in_array($this->directive, $this->directives(self::DIRECTIVE_USER_AGENT))) { |
474
|
|
|
$rule = []; |
475
|
|
|
foreach ($this->userAgents as $userAgent) { |
476
|
|
|
$rule[self::DIRECTIVE_USER_AGENT][$userAgent] = $this->rule; |
477
|
|
|
} |
478
|
|
|
return $rule; |
479
|
|
|
} |
480
|
|
|
return $this->rule; |
481
|
|
|
} |
482
|
|
|
|
483
|
|
|
/** |
484
|
|
|
* Get rules |
485
|
|
|
*/ |
486
|
|
|
public function getRules() |
487
|
|
|
{ |
488
|
|
|
return $this->rules; |
489
|
|
|
} |
490
|
|
|
} |
491
|
|
|
|