1
|
|
|
<?php |
2
|
|
|
/** |
3
|
|
|
* TLDExtract: Library for extraction of domain parts e.g. TLD. Domain parser that uses Public Suffix List. |
4
|
|
|
* |
5
|
|
|
* @link https://github.com/layershifter/TLDExtract |
6
|
|
|
* |
7
|
|
|
* @copyright Copyright (c) 2016, Alexander Fedyashov |
8
|
|
|
* @license https://raw.githubusercontent.com/layershifter/TLDExtract/master/LICENSE Apache 2.0 License |
9
|
|
|
*/ |
10
|
|
|
|
11
|
|
|
namespace LayerShifter\TLDExtract; |
12
|
|
|
|
13
|
|
|
use LayerShifter\TLDDatabase\Store; |
14
|
|
|
use LayerShifter\TLDExtract\Exceptions\RuntimeException; |
15
|
|
|
use LayerShifter\TLDSupport\Helpers\Arr; |
16
|
|
|
use LayerShifter\TLDSupport\Helpers\IP; |
17
|
|
|
use LayerShifter\TLDSupport\Helpers\Str; |
18
|
|
|
|
19
|
|
|
/** |
20
|
|
|
* Extract class accurately extracts subdomain, domain and TLD components from URLs. |
21
|
|
|
* |
22
|
|
|
* @see Result for more information on the returned data structure. |
23
|
|
|
*/ |
24
|
|
|
class Extract |
25
|
|
|
{ |
26
|
|
|
|
27
|
|
|
/** |
28
|
|
|
* @const int If this option provided, extract will consider ICCAN suffixes. |
29
|
|
|
*/ |
30
|
|
|
const MODE_ALLOW_ICCAN = 2; |
31
|
|
|
/** |
32
|
|
|
* @const int If this option provided, extract will consider private suffixes. |
33
|
|
|
*/ |
34
|
|
|
const MODE_ALLOW_PRIVATE = 4; |
35
|
|
|
/** |
36
|
|
|
* @const int If this option provided, extract will consider custom domains. |
37
|
|
|
*/ |
38
|
|
|
const MODE_ALLOW_NOT_EXISTING_SUFFIXES = 8; |
39
|
|
|
/** |
40
|
|
|
* @const string RFC 3986 compliant scheme regex pattern. |
41
|
|
|
* |
42
|
|
|
* @see https://tools.ietf.org/html/rfc3986#section-3.1 |
43
|
|
|
*/ |
44
|
|
|
const SCHEMA_PATTERN = '#^([a-zA-Z][a-zA-Z0-9+\-.]*:)?//#'; |
45
|
|
|
|
46
|
|
|
/** |
47
|
|
|
* @var int Value of extraction options. |
48
|
|
|
*/ |
49
|
|
|
private $extractionMode; |
50
|
|
|
/** |
51
|
|
|
* @var string Name of class that will store results of parsing. |
52
|
|
|
*/ |
53
|
|
|
private $resultClassName; |
54
|
|
|
/** |
55
|
|
|
* @var Store Object of TLDDatabase\Store class. |
56
|
|
|
*/ |
57
|
|
|
private $suffixStore; |
58
|
|
|
|
59
|
|
|
/** |
60
|
|
|
* Factory constructor. |
61
|
|
|
* |
62
|
|
|
* @param null|string $databaseFile Optional, name of file with Public Suffix List database |
63
|
|
|
* @param null|string $resultClassName Optional, name of class that will store results of parsing |
64
|
|
|
* @param null|int $extractionMode Optional, option that will control extraction process |
65
|
|
|
* |
66
|
|
|
* @throws RuntimeException |
67
|
|
|
*/ |
68
|
|
|
public function __construct($databaseFile = null, $resultClassName = null, $extractionMode = null) |
69
|
|
|
{ |
70
|
|
|
$this->suffixStore = new Store($databaseFile); |
71
|
|
|
$this->resultClassName = Result::class; |
72
|
|
|
|
73
|
|
|
// Checks for resultClassName argument. |
74
|
|
|
|
75
|
|
|
if (null !== $resultClassName) { |
76
|
|
|
if (!class_exists($resultClassName)) { |
77
|
|
|
throw new RuntimeException(sprintf('Class "%s" is not defined', $resultClassName)); |
78
|
|
|
} |
79
|
|
|
|
80
|
|
|
if (!in_array(ResultInterface::class, class_implements($resultClassName), true)) { |
81
|
|
|
throw new RuntimeException(sprintf('Class "%s" not implements ResultInterface', $resultClassName)); |
82
|
|
|
} |
83
|
|
|
|
84
|
|
|
$this->resultClassName = $resultClassName; |
85
|
|
|
} |
86
|
|
|
|
87
|
|
|
$this->setExtractionMode($extractionMode); |
88
|
|
|
} |
89
|
|
|
|
90
|
|
|
/** |
91
|
|
|
* Sets extraction mode, option that will control extraction process. |
92
|
|
|
* |
93
|
|
|
* @param int $extractionMode One of MODE_* constants |
94
|
|
|
* |
95
|
|
|
* @throws RuntimeException |
96
|
|
|
*/ |
97
|
|
|
public function setExtractionMode($extractionMode = null) |
98
|
|
|
{ |
99
|
|
|
if (null === $extractionMode) { |
100
|
|
|
$this->extractionMode = static::MODE_ALLOW_ICCAN |
101
|
|
|
| static::MODE_ALLOW_PRIVATE |
102
|
|
|
| static::MODE_ALLOW_NOT_EXISTING_SUFFIXES; |
103
|
|
|
|
104
|
|
|
return; |
105
|
|
|
} |
106
|
|
|
|
107
|
|
|
if (!is_int($extractionMode)) { |
108
|
|
|
throw new RuntimeException('Invalid argument type, extractionMode must be integer'); |
109
|
|
|
} |
110
|
|
|
|
111
|
|
|
if (!in_array($extractionMode, [ |
112
|
|
|
static::MODE_ALLOW_ICCAN, |
113
|
|
|
static::MODE_ALLOW_PRIVATE, |
114
|
|
|
static::MODE_ALLOW_NOT_EXISTING_SUFFIXES, |
115
|
|
|
static::MODE_ALLOW_ICCAN | static::MODE_ALLOW_PRIVATE, |
116
|
|
|
static::MODE_ALLOW_ICCAN | static::MODE_ALLOW_NOT_EXISTING_SUFFIXES, |
117
|
|
|
static::MODE_ALLOW_ICCAN | static::MODE_ALLOW_PRIVATE | static::MODE_ALLOW_NOT_EXISTING_SUFFIXES, |
118
|
|
|
static::MODE_ALLOW_PRIVATE | static::MODE_ALLOW_NOT_EXISTING_SUFFIXES |
119
|
|
|
], true) |
120
|
|
|
) { |
121
|
|
|
throw new RuntimeException( |
122
|
|
|
'Invalid argument type, extractionMode must be one of defined constants of their combination' |
123
|
|
|
); |
124
|
|
|
} |
125
|
|
|
|
126
|
|
|
$this->extractionMode = $extractionMode; |
127
|
|
|
} |
128
|
|
|
|
129
|
|
|
/** |
130
|
|
|
* Extract the subdomain, host and gTLD/ccTLD components from a URL. |
131
|
|
|
* |
132
|
|
|
* @param string $url URL that will be extracted |
133
|
|
|
* |
134
|
|
|
* @return ResultInterface |
135
|
|
|
*/ |
136
|
|
|
public function parse($url) |
137
|
|
|
{ |
138
|
|
|
$hostname = $this->extractHostname($url); |
139
|
|
|
|
140
|
|
|
// If received hostname is valid IP address, result will be formed from it. |
141
|
|
|
|
142
|
|
|
if (IP::isValid($hostname)) { |
143
|
|
|
return new $this->resultClassName(null, $hostname, null); |
144
|
|
|
} |
145
|
|
|
|
146
|
|
|
list($subDomain, $host, $suffix) = $this->extractParts($hostname); |
147
|
|
|
|
148
|
|
|
return new $this->resultClassName($subDomain, $host, $suffix); |
149
|
|
|
} |
150
|
|
|
|
151
|
|
|
/** |
152
|
|
|
* Method that extracts the hostname or IP address from a URL. |
153
|
|
|
* |
154
|
|
|
* @param string $url URL for extraction |
155
|
|
|
* |
156
|
|
|
* @return null|string Hostname or IP address |
157
|
|
|
*/ |
158
|
|
|
private function extractHostname($url) |
159
|
|
|
{ |
160
|
|
|
$url = trim(Str::lower($url)); |
161
|
|
|
|
162
|
|
|
// Removes scheme and path i.e. "https://github.com/layershifter" to "github.com/layershifter". |
163
|
|
|
|
164
|
|
|
$url = preg_replace(static::SCHEMA_PATTERN, '', $url); |
165
|
|
|
|
166
|
|
|
// Removes path and query part of URL i.e. "github.com/layershifter" to "github.com". |
167
|
|
|
|
168
|
|
|
$url = $this->fixQueryPart($url); |
169
|
|
|
$hostname = Arr::first(explode('/', $url, 2)); |
170
|
|
|
|
171
|
|
|
// Removes username from URL i.e. [email protected] to github.com. |
172
|
|
|
|
173
|
|
|
$hostname = Arr::last(explode('@', $hostname)); |
174
|
|
|
|
175
|
|
|
// Remove ports from hosts, also check for IPv6 literals like "[3ffe:2a00:100:7031::1]". |
176
|
|
|
// |
177
|
|
|
// @see http://www.ietf.org/rfc/rfc2732.txt |
178
|
|
|
|
179
|
|
|
$lastBracketPosition = Str::strrpos($hostname, ']'); |
180
|
|
|
|
181
|
|
|
if ($lastBracketPosition !== false && Str::startsWith($hostname, '[')) { |
182
|
|
|
return Str::substr($hostname, 1, $lastBracketPosition - 1); |
183
|
|
|
} |
184
|
|
|
|
185
|
|
|
// This is either a normal hostname or an IPv4 address, just remove the port. |
186
|
|
|
|
187
|
|
|
$hostname = Arr::first(explode(':', $hostname)); |
188
|
|
|
|
189
|
|
|
// If string is empty, null will be returned. |
190
|
|
|
|
191
|
|
|
return '' === $hostname ? null : $hostname; |
192
|
|
|
} |
193
|
|
|
|
194
|
|
|
/** |
195
|
|
|
* Extracts subdomain, host and suffix from input string. Based on algorithm described in |
196
|
|
|
* https://publicsuffix.org/list/. |
197
|
|
|
* |
198
|
|
|
* @param string $hostname Hostname for extraction |
199
|
|
|
* |
200
|
|
|
* @return array|string[] An array that contains subdomain, host and suffix. |
201
|
|
|
*/ |
202
|
|
|
public function extractParts($hostname) |
203
|
|
|
{ |
204
|
|
|
$suffix = $this->extractSuffix($hostname); |
205
|
|
|
|
206
|
|
|
if ($suffix === $hostname) { |
207
|
|
|
return [null, $hostname, null]; |
208
|
|
|
} |
209
|
|
|
|
210
|
|
|
if (null !== $suffix) { |
211
|
|
|
$hostname = Str::substr($hostname, 0, -Str::length($suffix) - 1); |
212
|
|
|
} |
213
|
|
|
|
214
|
|
|
$lastDot = Str::strrpos($hostname, '.'); |
215
|
|
|
|
216
|
|
|
if (false === $lastDot) { |
217
|
|
|
return [null, $hostname, $suffix]; |
218
|
|
|
} |
219
|
|
|
|
220
|
|
|
$subDomain = Str::substr($hostname, 0, $lastDot); |
|
|
|
|
221
|
|
|
$host = Str::substr($hostname, $lastDot + 1); |
222
|
|
|
|
223
|
|
|
return [ |
224
|
|
|
$subDomain, |
225
|
|
|
$host, |
226
|
|
|
$suffix |
227
|
|
|
]; |
228
|
|
|
} |
229
|
|
|
|
230
|
|
|
/** |
231
|
|
|
* Extracts suffix from hostname using Public Suffix List database. |
232
|
|
|
* |
233
|
|
|
* @param string $hostname Hostname for extraction |
234
|
|
|
* |
235
|
|
|
* @return null|string |
236
|
|
|
*/ |
237
|
|
|
private function extractSuffix($hostname) |
238
|
|
|
{ |
239
|
|
|
// If hostname has leading dot, it's invalid. |
240
|
|
|
// If hostname is a single label domain makes, it's invalid. |
241
|
|
|
|
242
|
|
|
if (Str::startsWith($hostname, '.') || Str::strpos($hostname, '.') === false) { |
243
|
|
|
return null; |
244
|
|
|
} |
245
|
|
|
|
246
|
|
|
// If domain is in punycode, it will be converted to IDN. |
247
|
|
|
|
248
|
|
|
$isPunycoded = Str::strpos($hostname, 'xn--') !== false; |
249
|
|
|
|
250
|
|
|
if ($isPunycoded) { |
251
|
|
|
$hostname = idn_to_utf8($hostname); |
252
|
|
|
} |
253
|
|
|
|
254
|
|
|
$suffix = $this->parseSuffix($hostname); |
255
|
|
|
|
256
|
|
|
if (null === $suffix) { |
257
|
|
|
if (!($this->extractionMode & static::MODE_ALLOW_NOT_EXISTING_SUFFIXES)) { |
258
|
|
|
return null; |
259
|
|
|
} |
260
|
|
|
|
261
|
|
|
$suffix = Str::substr($hostname, Str::strrpos($hostname, '.') + 1); |
262
|
|
|
} |
263
|
|
|
|
264
|
|
|
// If domain is punycoded, suffix will be converted to punycode. |
265
|
|
|
|
266
|
|
|
return $isPunycoded ? idn_to_ascii($suffix) : $suffix; |
267
|
|
|
} |
268
|
|
|
|
269
|
|
|
/** |
270
|
|
|
* Extracts suffix from hostname using Public Suffix List database. |
271
|
|
|
* |
272
|
|
|
* @param string $hostname Hostname for extraction |
273
|
|
|
* |
274
|
|
|
* @return null|string |
275
|
|
|
*/ |
276
|
|
|
private function parseSuffix($hostname) |
277
|
|
|
{ |
278
|
|
|
$hostnameParts = explode('.', $hostname); |
279
|
|
|
$realSuffix = null; |
280
|
|
|
|
281
|
|
|
for ($i = 0, $count = count($hostnameParts); $i < $count; $i++) { |
282
|
|
|
$possibleSuffix = implode('.', array_slice($hostnameParts, $i)); |
283
|
|
|
$exceptionSuffix = '!' . $possibleSuffix; |
284
|
|
|
|
285
|
|
|
if ($this->suffixExists($exceptionSuffix)) { |
286
|
|
|
$realSuffix = implode('.', array_slice($hostnameParts, $i + 1)); |
287
|
|
|
|
288
|
|
|
break; |
289
|
|
|
} |
290
|
|
|
|
291
|
|
|
if ($this->suffixExists($possibleSuffix)) { |
292
|
|
|
$realSuffix = $possibleSuffix; |
293
|
|
|
|
294
|
|
|
break; |
295
|
|
|
} |
296
|
|
|
|
297
|
|
|
$wildcardTld = '*.' . implode('.', array_slice($hostnameParts, $i + 1)); |
298
|
|
|
|
299
|
|
|
if ($this->suffixExists($wildcardTld)) { |
300
|
|
|
$realSuffix = $possibleSuffix; |
301
|
|
|
|
302
|
|
|
break; |
303
|
|
|
} |
304
|
|
|
} |
305
|
|
|
|
306
|
|
|
return $realSuffix; |
307
|
|
|
} |
308
|
|
|
|
309
|
|
|
/** |
310
|
|
|
* Method that checks existence of entry in Public Suffix List database, including provided options. |
311
|
|
|
* |
312
|
|
|
* @param string $entry Entry for check in Public Suffix List database |
313
|
|
|
* |
314
|
|
|
* @return bool |
315
|
|
|
*/ |
316
|
|
|
private function suffixExists($entry) |
317
|
|
|
{ |
318
|
|
|
if (!$this->suffixStore->isExists($entry)) { |
319
|
|
|
return false; |
320
|
|
|
} |
321
|
|
|
|
322
|
|
|
$type = $this->suffixStore->getType($entry); |
323
|
|
|
|
324
|
|
|
if ($this->extractionMode & static::MODE_ALLOW_ICCAN && $type === Store::TYPE_ICCAN) { |
325
|
|
|
return true; |
326
|
|
|
} |
327
|
|
|
|
328
|
|
|
return $this->extractionMode & static::MODE_ALLOW_PRIVATE && $type === Store::TYPE_PRIVATE; |
329
|
|
|
} |
330
|
|
|
|
331
|
|
|
/** |
332
|
|
|
* Fixes URL from "github.com?layershifter" to "github.com/?layershifter". |
333
|
|
|
* |
334
|
|
|
* @see https://github.com/layershifter/TLDExtract/issues/5 |
335
|
|
|
* |
336
|
|
|
* @param string $url |
337
|
|
|
* |
338
|
|
|
* @return string |
339
|
|
|
*/ |
340
|
|
|
private function fixQueryPart($url) |
341
|
|
|
{ |
342
|
|
|
$position = Str::strpos($url, '?'); |
343
|
|
|
|
344
|
|
|
if ($position === false) { |
345
|
|
|
return $url; |
346
|
|
|
} |
347
|
|
|
|
348
|
|
|
return Str::substr($url, 0, $position) . '/' . Str::substr($url, $position); |
|
|
|
|
349
|
|
|
} |
350
|
|
|
} |
351
|
|
|
|
If a method or function can return multiple different values and unless you are sure that you only can receive a single value in this context, we recommend to add an additional type check:
If this a common case that PHP Analyzer should handle natively, please let us know by opening an issue.