|
1
|
|
|
<?php |
|
2
|
|
|
namespace vipnytt\RobotsTxtParser; |
|
3
|
|
|
|
|
4
|
|
|
use vipnytt\RobotsTxtParser\Client\Directives\HostClient; |
|
5
|
|
|
use vipnytt\RobotsTxtParser\Client\Directives\SitemapClient; |
|
6
|
|
|
use vipnytt\RobotsTxtParser\Client\Directives\UserAgentClient; |
|
7
|
|
|
use vipnytt\RobotsTxtParser\Client\Encoding\EncodingConverter; |
|
8
|
|
|
use vipnytt\RobotsTxtParser\Parser\RobotsTxtParser; |
|
9
|
|
|
use vipnytt\RobotsTxtParser\Parser\UrlParser; |
|
10
|
|
|
|
|
11
|
|
|
/** |
|
12
|
|
|
* Class Input |
|
13
|
|
|
* |
|
14
|
|
|
* @link https://developers.google.com/webmasters/control-crawl-index/docs/robots_txt |
|
15
|
|
|
* @link https://yandex.com/support/webmaster/controlling-robot/robots-txt.xml |
|
16
|
|
|
* @link http://www.robotstxt.org/robotstxt.html |
|
17
|
|
|
* @link https://www.w3.org/TR/html4/appendix/notes.html#h-B.4.1.1 |
|
18
|
|
|
* @link http://www.conman.org/people/spc/robots2.html |
|
19
|
|
|
* |
|
20
|
|
|
* @package vipnytt\RobotsTxtParser\Client |
|
21
|
|
|
*/ |
|
22
|
|
|
class Input extends RobotsTxtParser |
|
23
|
|
|
{ |
|
24
|
|
|
use UrlParser; |
|
25
|
|
|
|
|
26
|
|
|
/** |
|
27
|
|
|
* Base uri |
|
28
|
|
|
* @var string |
|
29
|
|
|
*/ |
|
30
|
|
|
protected $base; |
|
31
|
|
|
|
|
32
|
|
|
/** |
|
33
|
|
|
* Status code |
|
34
|
|
|
* @var int|null |
|
35
|
|
|
*/ |
|
36
|
|
|
protected $statusCode; |
|
37
|
|
|
|
|
38
|
|
|
/** |
|
39
|
|
|
* Robots.txt content |
|
40
|
|
|
* @var string |
|
41
|
|
|
*/ |
|
42
|
|
|
protected $content; |
|
43
|
|
|
|
|
44
|
|
|
/** |
|
45
|
|
|
* UserAgentClient class cache |
|
46
|
|
|
* @var UserAgentClient[] |
|
47
|
|
|
*/ |
|
48
|
|
|
private $userAgentClients = []; |
|
49
|
|
|
|
|
50
|
|
|
/** |
|
51
|
|
|
* Core constructor. |
|
52
|
|
|
* |
|
53
|
|
|
* @param string $baseUri |
|
54
|
|
|
* @param int $statusCode |
|
55
|
|
|
* @param string|null $content |
|
56
|
|
|
* @param string $encoding |
|
57
|
|
|
* @param int|null $byteLimit |
|
58
|
|
|
*/ |
|
59
|
|
|
public function __construct($baseUri, $statusCode, $content, $encoding = self::ENCODING, $byteLimit = self::BYTE_LIMIT) |
|
60
|
|
|
{ |
|
61
|
|
|
$this->base = $this->urlBase($this->urlEncode($baseUri)); |
|
62
|
|
|
$this->statusCode = $statusCode; |
|
63
|
|
|
$this->content = $content; |
|
64
|
|
|
$this->convertEncoding($encoding); |
|
65
|
|
|
$this->limitBytes($byteLimit); |
|
66
|
|
|
parent::__construct($this->content); |
|
67
|
|
|
} |
|
68
|
|
|
|
|
69
|
|
|
/** |
|
70
|
|
|
* Convert character encoding |
|
71
|
|
|
* |
|
72
|
|
|
* @param string $encoding |
|
73
|
|
|
* @return string |
|
74
|
|
|
*/ |
|
75
|
|
|
private function convertEncoding($encoding) |
|
76
|
|
|
{ |
|
77
|
|
|
mb_internal_encoding(self::ENCODING); |
|
78
|
|
|
$convert = new EncodingConverter($this->content, $encoding); |
|
79
|
|
|
if (($result = $convert->auto()) !== false) { |
|
80
|
|
|
return $this->content = $result; |
|
81
|
|
|
} |
|
82
|
|
|
return $this->content; |
|
83
|
|
|
} |
|
84
|
|
|
|
|
85
|
|
|
/** |
|
86
|
|
|
* Byte limit |
|
87
|
|
|
* |
|
88
|
|
|
* @param int|null $bytes |
|
89
|
|
|
* @return string |
|
90
|
|
|
*/ |
|
91
|
|
|
private function limitBytes($bytes) |
|
92
|
|
|
{ |
|
93
|
|
|
if ($bytes === null) { |
|
94
|
|
|
return $this->content; |
|
95
|
|
|
} elseif ($bytes < 5000) { |
|
96
|
|
|
trigger_error('Byte limit is set dangerously low!', E_USER_WARNING); |
|
97
|
|
|
} |
|
98
|
|
|
return $this->content = mb_strcut($this->content, 0, intval($bytes)); |
|
99
|
|
|
} |
|
100
|
|
|
|
|
101
|
|
|
/** |
|
102
|
|
|
* Sitemaps |
|
103
|
|
|
* |
|
104
|
|
|
* @return SitemapClient |
|
105
|
|
|
*/ |
|
106
|
|
|
public function sitemap() |
|
107
|
|
|
{ |
|
108
|
|
|
$export = $this->sitemap->export(); |
|
109
|
|
|
return new SitemapClient(isset($export[self::DIRECTIVE_SITEMAP]) ? $export[self::DIRECTIVE_SITEMAP] : []); |
|
110
|
|
|
} |
|
111
|
|
|
|
|
112
|
|
|
/** |
|
113
|
|
|
* Host |
|
114
|
|
|
* |
|
115
|
|
|
* @return HostClient |
|
116
|
|
|
*/ |
|
117
|
|
|
public function host() |
|
118
|
|
|
{ |
|
119
|
|
|
$export = $this->host->export(); |
|
120
|
|
|
return new HostClient(isset($export[self::DIRECTIVE_HOST][0]) ? $export[self::DIRECTIVE_HOST][0] : null); |
|
121
|
|
|
} |
|
122
|
|
|
|
|
123
|
|
|
/** |
|
124
|
|
|
* Get Clean-param |
|
125
|
|
|
* |
|
126
|
|
|
* @return array |
|
127
|
|
|
*/ |
|
128
|
|
|
public function getCleanParam() |
|
129
|
|
|
{ |
|
130
|
|
|
$export = $this->cleanParam->export(); |
|
131
|
|
|
if (isset($export[self::DIRECTIVE_CLEAN_PARAM])) { |
|
132
|
|
|
return $export[self::DIRECTIVE_CLEAN_PARAM]; |
|
133
|
|
|
} |
|
134
|
|
|
return []; |
|
135
|
|
|
} |
|
136
|
|
|
|
|
137
|
|
|
/** |
|
138
|
|
|
* Get User-agent list |
|
139
|
|
|
* |
|
140
|
|
|
* @return array |
|
141
|
|
|
*/ |
|
142
|
|
|
public function getUserAgents() |
|
143
|
|
|
{ |
|
144
|
|
|
return $this->userAgent->userAgents; |
|
145
|
|
|
} |
|
146
|
|
|
|
|
147
|
|
|
/** |
|
148
|
|
|
* Client User-agent specific rules |
|
149
|
|
|
* |
|
150
|
|
|
* @param string $string |
|
151
|
|
|
* @return UserAgentClient |
|
152
|
|
|
*/ |
|
153
|
|
|
public function userAgent($string = self::USER_AGENT) |
|
154
|
|
|
{ |
|
155
|
|
|
if (isset($this->userAgentClients[$string])) { |
|
156
|
|
|
return $this->userAgentClients[$string]; |
|
157
|
|
|
} |
|
158
|
|
|
$this->userAgentClients[$string] = new UserAgentClient($string, $this->userAgent, $this->base, $this->statusCode); |
|
159
|
|
|
return $this->userAgentClients[$string]; |
|
160
|
|
|
} |
|
161
|
|
|
} |
|
162
|
|
|
|
VIPnytt /
RobotsTxtParser
