|
1
|
|
|
<?php |
|
2
|
|
|
namespace vipnytt\RobotsTxtParser\Modules; |
|
3
|
|
|
|
|
4
|
|
|
use vipnytt\RobotsTxtParser\Exceptions\ClientException; |
|
5
|
|
|
use vipnytt\RobotsTxtParser\RobotsTxtInterface; |
|
6
|
|
|
|
|
7
|
|
|
/** |
|
8
|
|
|
* Class UserAgentClient |
|
9
|
|
|
* |
|
10
|
|
|
* @package vipnytt\RobotsTxtParser\Modules |
|
11
|
|
|
*/ |
|
12
|
|
|
class UserAgentClient implements RobotsTxtInterface |
|
13
|
|
|
{ |
|
14
|
|
|
use UrlTools; |
|
15
|
|
|
|
|
16
|
|
|
/** |
|
17
|
|
|
* Rules |
|
18
|
|
|
* @var array |
|
19
|
|
|
*/ |
|
20
|
|
|
protected $rules; |
|
21
|
|
|
|
|
22
|
|
|
/** |
|
23
|
|
|
* User-agent |
|
24
|
|
|
* @var string |
|
25
|
|
|
*/ |
|
26
|
|
|
protected $userAgent; |
|
27
|
|
|
|
|
28
|
|
|
/** |
|
29
|
|
|
* Robots.txt base URL |
|
30
|
|
|
* @var string |
|
31
|
|
|
*/ |
|
32
|
|
|
protected $base; |
|
33
|
|
|
|
|
34
|
|
|
/** |
|
35
|
|
|
* Status code parser |
|
36
|
|
|
* @var StatusCodeParser |
|
37
|
|
|
*/ |
|
38
|
|
|
protected $statusCodeParser; |
|
39
|
|
|
|
|
40
|
|
|
/** |
|
41
|
|
|
* UserAgentClient constructor. |
|
42
|
|
|
* |
|
43
|
|
|
* @param array $rules |
|
44
|
|
|
* @param string $userAgent |
|
45
|
|
|
* @param string $baseUrl |
|
46
|
|
|
* @param int $statusCode |
|
47
|
|
|
*/ |
|
48
|
|
|
public function __construct($rules, $userAgent, $baseUrl, $statusCode) |
|
49
|
|
|
{ |
|
50
|
|
|
$this->statusCodeParser = new StatusCodeParser($statusCode, parse_url($baseUrl, PHP_URL_SCHEME)); |
|
51
|
|
|
$this->userAgent = $userAgent; |
|
52
|
|
|
$this->rules = $rules; |
|
53
|
|
|
$this->base = $baseUrl; |
|
54
|
|
|
} |
|
55
|
|
|
|
|
56
|
|
|
/** |
|
57
|
|
|
* Check if URL is allowed to crawl |
|
58
|
|
|
* |
|
59
|
|
|
* @param string $url |
|
60
|
|
|
* @return bool |
|
61
|
|
|
*/ |
|
62
|
|
|
public function isAllowed($url) |
|
63
|
|
|
{ |
|
64
|
|
|
return $this->check(self::DIRECTIVE_ALLOW, $url); |
|
65
|
|
|
} |
|
66
|
|
|
|
|
67
|
|
|
/** |
|
68
|
|
|
* Check |
|
69
|
|
|
* |
|
70
|
|
|
* @param string $directive |
|
71
|
|
|
* @param string $url - URL to check |
|
72
|
|
|
* @return bool |
|
73
|
|
|
* @throws ClientException |
|
74
|
|
|
*/ |
|
75
|
|
|
protected function check($directive, $url) |
|
76
|
|
|
{ |
|
77
|
|
|
$url = $this->urlConvertToFull($url, $this->base); |
|
78
|
|
|
if (!$this->isUrlApplicable([$url, $this->base])) { |
|
79
|
|
|
throw new ClientException('URL belongs to a different robots.txt, please check it against that one instead'); |
|
80
|
|
|
} |
|
81
|
|
|
$this->statusCodeParser->replaceUnofficial(); |
|
82
|
|
|
if (($result = $this->statusCodeParser->check()) !== null) { |
|
83
|
|
|
return $directive === $result; |
|
84
|
|
|
} |
|
85
|
|
|
$result = self::DIRECTIVE_ALLOW; |
|
86
|
|
|
foreach ([self::DIRECTIVE_DISALLOW, self::DIRECTIVE_ALLOW] as $currentDirective) { |
|
87
|
|
|
if ($this->rules[$currentDirective]->check($url)) { |
|
88
|
|
|
$result = $currentDirective; |
|
89
|
|
|
} |
|
90
|
|
|
} |
|
91
|
|
|
return $directive === $result; |
|
92
|
|
|
} |
|
93
|
|
|
|
|
94
|
|
|
/** |
|
95
|
|
|
* Check if the URL belongs to current robots.txt |
|
96
|
|
|
* |
|
97
|
|
|
* @param $urls |
|
98
|
|
|
* @return bool |
|
99
|
|
|
*/ |
|
100
|
|
|
protected function isUrlApplicable($urls) |
|
101
|
|
|
{ |
|
102
|
|
|
foreach ($urls as $url) { |
|
103
|
|
|
$parsed = parse_url($url); |
|
104
|
|
|
$parsed['port'] = is_int($port = parse_url($url, PHP_URL_PORT)) ? $port : getservbyname($parsed['scheme'], 'tcp'); |
|
105
|
|
|
$assembled = $parsed['scheme'] . '://' . $parsed['host'] . ':' . $parsed['port']; |
|
106
|
|
|
if (!isset($result)) { |
|
107
|
|
|
$result = $assembled; |
|
108
|
|
|
} elseif ($result !== $assembled) { |
|
109
|
|
|
return false; |
|
110
|
|
|
} |
|
111
|
|
|
} |
|
112
|
|
|
return true; |
|
113
|
|
|
} |
|
114
|
|
|
|
|
115
|
|
|
/** |
|
116
|
|
|
* Check if URL is disallowed to crawl |
|
117
|
|
|
* |
|
118
|
|
|
* @param string $url |
|
119
|
|
|
* @return bool |
|
120
|
|
|
*/ |
|
121
|
|
|
public function isDisallowed($url) |
|
122
|
|
|
{ |
|
123
|
|
|
return $this->check(self::DIRECTIVE_DISALLOW, $url); |
|
124
|
|
|
} |
|
125
|
|
|
|
|
126
|
|
|
/** |
|
127
|
|
|
* Get Cache-delay |
|
128
|
|
|
* |
|
129
|
|
|
* @return float|int |
|
130
|
|
|
*/ |
|
131
|
|
|
public function getCacheDelay() |
|
132
|
|
|
{ |
|
133
|
|
|
$exported = $this->rules[self::DIRECTIVE_CACHE_DELAY]->export(); |
|
134
|
|
|
return isset($exported[self::DIRECTIVE_CACHE_DELAY]) ? $exported[self::DIRECTIVE_CACHE_DELAY] : $this->getCrawlDelay(); |
|
135
|
|
|
} |
|
136
|
|
|
|
|
137
|
|
|
/** |
|
138
|
|
|
* Get Crawl-delay |
|
139
|
|
|
* |
|
140
|
|
|
* @return float|int |
|
141
|
|
|
*/ |
|
142
|
|
|
public function getCrawlDelay() |
|
143
|
|
|
{ |
|
144
|
|
|
$exported = $this->rules[self::DIRECTIVE_CRAWL_DELAY]->export(); |
|
145
|
|
|
return isset($exported[self::DIRECTIVE_CRAWL_DELAY]) ? $exported[self::DIRECTIVE_CRAWL_DELAY] : 0; |
|
146
|
|
|
} |
|
147
|
|
|
|
|
148
|
|
|
/** |
|
149
|
|
|
* Rule export |
|
150
|
|
|
* |
|
151
|
|
|
* @return array |
|
152
|
|
|
*/ |
|
153
|
|
|
public function export() |
|
154
|
|
|
{ |
|
155
|
|
|
$result = []; |
|
156
|
|
|
foreach ($this->rules as $directive => $object) { |
|
157
|
|
|
if (!empty($export = $object->export())) { |
|
158
|
|
|
$result[$directive] = $export[$directive]; |
|
159
|
|
|
} |
|
160
|
|
|
} |
|
161
|
|
|
return $result; |
|
162
|
|
|
} |
|
163
|
|
|
} |
|
164
|
|
|
|
VIPnytt /
RobotsTxtParser
