|
1
|
|
|
<?php |
|
2
|
|
|
/** |
|
3
|
|
|
* X-Robots-Tag HTTP header parser class |
|
4
|
|
|
* |
|
5
|
|
|
* @author VIP nytt ([email protected]) |
|
6
|
|
|
* @author Jan-Petter Gundersen ([email protected]) |
|
7
|
|
|
* |
|
8
|
|
|
* Project: |
|
9
|
|
|
* @link https://github.com/VIPnytt/X-Robots-Tag-parser |
|
10
|
|
|
* @license https://opensource.org/licenses/MIT MIT license |
|
11
|
|
|
* |
|
12
|
|
|
* Specification: |
|
13
|
|
|
* @link https://developers.google.com/webmasters/control-crawl-index/docs/robots_meta_tag#using-the-x-robots-tag-http-header |
|
14
|
|
|
*/ |
|
15
|
|
|
|
|
16
|
|
|
namespace vipnytt; |
|
17
|
|
|
|
|
18
|
|
|
use vipnytt\XRobotsTagParser\directive; |
|
19
|
|
|
use vipnytt\XRobotsTagParser\URLParser; |
|
20
|
|
|
use vipnytt\XRobotsTagParser\UserAgentParser; |
|
21
|
|
|
|
|
22
|
|
|
class XRobotsTagParser |
|
23
|
|
|
{ |
|
24
|
|
|
const HEADER_RULE_IDENTIFIER = 'x-robots-tag'; |
|
25
|
|
|
const USERAGENT_DEFAULT = ''; |
|
26
|
|
|
|
|
27
|
|
|
const DIRECTIVE_ALL = 'all'; |
|
28
|
|
|
const DIRECTIVE_NONE = 'none'; |
|
29
|
|
|
const DIRECTIVE_NO_ARCHIVE = 'noarchive'; |
|
30
|
|
|
const DIRECTIVE_NO_FOLLOW = 'nofollow'; |
|
31
|
|
|
const DIRECTIVE_NO_IMAGE_INDEX = 'noimageindex'; |
|
32
|
|
|
const DIRECTIVE_NO_INDEX = 'noindex'; |
|
33
|
|
|
const DIRECTIVE_NO_ODP = 'noodp'; |
|
34
|
|
|
const DIRECTIVE_NO_SNIPPET = 'nosnippet'; |
|
35
|
|
|
const DIRECTIVE_NO_TRANSLATE = 'notranslate'; |
|
36
|
|
|
const DIRECTIVE_UNAVAILABLE_AFTER = 'unavailable_after'; |
|
37
|
|
|
|
|
38
|
|
|
private $url = ''; |
|
39
|
|
|
private $userAgent = self::USERAGENT_DEFAULT; |
|
40
|
|
|
|
|
41
|
|
|
private $headers = []; |
|
42
|
|
|
private $currentRule = ''; |
|
43
|
|
|
private $currentUserAgent = self::USERAGENT_DEFAULT; |
|
44
|
|
|
private $currentDirective = ''; |
|
45
|
|
|
private $currentValue = ''; |
|
46
|
|
|
|
|
47
|
|
|
private $options = []; |
|
48
|
|
|
private $rules = []; |
|
49
|
|
|
|
|
50
|
|
|
/** |
|
51
|
|
|
* Constructor |
|
52
|
|
|
* |
|
53
|
|
|
* @param string $url |
|
54
|
|
|
* @param string $userAgent |
|
55
|
|
|
* @param array $options |
|
56
|
|
|
*/ |
|
57
|
|
|
public function __construct($url, $userAgent = self::USERAGENT_DEFAULT, $options = []) |
|
58
|
|
|
{ |
|
59
|
|
|
// Parse URL |
|
60
|
|
|
$urlParser = new URLParser(trim($url)); |
|
61
|
|
|
if (!$urlParser->isValid()) { |
|
62
|
|
|
trigger_error('Invalid URL', E_USER_WARNING); |
|
63
|
|
|
} |
|
64
|
|
|
// Encode URL |
|
65
|
|
|
$this->url = $urlParser->encode(); |
|
66
|
|
|
// Set any optional options |
|
67
|
|
|
$this->options = $options; |
|
68
|
|
|
// Get headers |
|
69
|
|
|
$this->getHeaders(); |
|
70
|
|
|
// Parse rules |
|
71
|
|
|
$this->parse(); |
|
72
|
|
|
// Set User-Agent |
|
73
|
|
|
$parser = new UserAgentParser($userAgent); |
|
74
|
|
|
$this->userAgent = $parser->match(array_keys($this->rules), self::USERAGENT_DEFAULT); |
|
75
|
|
|
} |
|
76
|
|
|
|
|
77
|
|
|
/** |
|
78
|
|
|
* Request HTTP headers |
|
79
|
|
|
* |
|
80
|
|
|
* @return bool |
|
81
|
|
|
*/ |
|
82
|
|
|
private function getHeaders() |
|
83
|
|
|
{ |
|
84
|
|
|
if (isset($this->options['headers'])) { |
|
85
|
|
|
$this->headers = $this->options['headers']; |
|
86
|
|
|
return true; |
|
87
|
|
|
} |
|
88
|
|
|
$this->headers = get_headers($this->url); |
|
89
|
|
|
if ($this->headers === false) { |
|
90
|
|
|
trigger_error('Unable to fetch HTTP headers', E_USER_ERROR); |
|
91
|
|
|
return false; |
|
92
|
|
|
} |
|
93
|
|
|
return true; |
|
94
|
|
|
} |
|
95
|
|
|
|
|
96
|
|
|
/** |
|
97
|
|
|
* Parse HTTP headers |
|
98
|
|
|
* |
|
99
|
|
|
* @return void |
|
100
|
|
|
*/ |
|
101
|
|
|
private function parse() |
|
102
|
|
|
{ |
|
103
|
|
|
foreach ($this->headers as $header) { |
|
104
|
|
|
$parts = explode(':', mb_strtolower($header), 2); |
|
105
|
|
|
if (count($parts) < 2 || $parts[0] != self::HEADER_RULE_IDENTIFIER) { |
|
106
|
|
|
// Header is not a rule |
|
107
|
|
|
continue; |
|
108
|
|
|
} |
|
109
|
|
|
$this->currentRule = trim($parts[1]); |
|
110
|
|
|
$this->detectDirectives(); |
|
111
|
|
|
} |
|
112
|
|
|
|
|
113
|
|
|
} |
|
114
|
|
|
|
|
115
|
|
|
/** |
|
116
|
|
|
* Detect directives in rule |
|
117
|
|
|
* |
|
118
|
|
|
* @return void |
|
119
|
|
|
*/ |
|
120
|
|
|
private function detectDirectives() |
|
121
|
|
|
{ |
|
122
|
|
|
$rules = explode(',', $this->currentRule); |
|
123
|
|
|
foreach ($rules as $rule) { |
|
124
|
|
|
$pair = array_map('trim', explode(':', $rule, 2)); |
|
125
|
|
|
if ($rules[0] === $rule && count($pair) == 2 && !in_array($pair[0], $this->directiveArray())) { |
|
126
|
|
|
$this->currentUserAgent = $pair[0]; |
|
127
|
|
|
$pair = array_map('trim', explode(':', $pair[1], 2)); |
|
128
|
|
|
} |
|
129
|
|
|
if (in_array($pair[0], $this->directiveArray())) { |
|
130
|
|
|
$this->currentDirective = $pair[0]; |
|
131
|
|
|
$this->currentValue = isset($pair[1]) ? $pair[1] : ''; |
|
132
|
|
|
$this->addRule(); |
|
133
|
|
|
} |
|
134
|
|
|
} |
|
135
|
|
|
$this->cleanup(); |
|
136
|
|
|
} |
|
137
|
|
|
|
|
138
|
|
|
/** |
|
139
|
|
|
* Directives supported |
|
140
|
|
|
* |
|
141
|
|
|
* @return array |
|
142
|
|
|
*/ |
|
143
|
|
|
protected function directiveArray() |
|
144
|
|
|
{ |
|
145
|
|
|
return [ |
|
146
|
|
|
self::DIRECTIVE_ALL, |
|
147
|
|
|
self::DIRECTIVE_NONE, |
|
148
|
|
|
self::DIRECTIVE_NO_ARCHIVE, |
|
149
|
|
|
self::DIRECTIVE_NO_FOLLOW, |
|
150
|
|
|
self::DIRECTIVE_NO_IMAGE_INDEX, |
|
151
|
|
|
self::DIRECTIVE_NO_INDEX, |
|
152
|
|
|
self::DIRECTIVE_NO_ODP, |
|
153
|
|
|
self::DIRECTIVE_NO_SNIPPET, |
|
154
|
|
|
self::DIRECTIVE_NO_TRANSLATE, |
|
155
|
|
|
self::DIRECTIVE_UNAVAILABLE_AFTER |
|
156
|
|
|
]; |
|
157
|
|
|
} |
|
158
|
|
|
|
|
159
|
|
|
/** |
|
160
|
|
|
* Add rule |
|
161
|
|
|
* |
|
162
|
|
|
* @return void |
|
163
|
|
|
*/ |
|
164
|
|
|
private function addRule() |
|
165
|
|
|
{ |
|
166
|
|
|
if (!isset($this->rules[$this->currentUserAgent])) { |
|
167
|
|
|
$this->rules[$this->currentUserAgent] = []; |
|
168
|
|
|
} |
|
169
|
|
|
$directive = new directive($this->currentDirective, $this->currentValue, $this->options); |
|
170
|
|
|
$this->rules[$this->currentUserAgent] = array_merge($this->rules[$this->currentUserAgent], $directive->getArray()); |
|
171
|
|
|
} |
|
172
|
|
|
|
|
173
|
|
|
/** |
|
174
|
|
|
* Cleanup before next rule is read |
|
175
|
|
|
* |
|
176
|
|
|
* @return void |
|
177
|
|
|
*/ |
|
178
|
|
|
private function cleanup() |
|
179
|
|
|
{ |
|
180
|
|
|
$this->currentRule = ''; |
|
181
|
|
|
$this->currentUserAgent = self::USERAGENT_DEFAULT; |
|
182
|
|
|
$this->currentDirective = ''; |
|
183
|
|
|
$this->currentValue = ''; |
|
184
|
|
|
} |
|
185
|
|
|
|
|
186
|
|
|
/** |
|
187
|
|
|
* Return all applicable rules |
|
188
|
|
|
* |
|
189
|
|
|
* @return array |
|
190
|
|
|
*/ |
|
191
|
|
|
public function getRules() |
|
192
|
|
|
{ |
|
193
|
|
|
$rules = []; |
|
194
|
|
|
// Default UserAgent |
|
195
|
|
|
if (isset($this->rules[self::USERAGENT_DEFAULT])) { |
|
196
|
|
|
$rules = array_merge($rules, $this->rules[self::USERAGENT_DEFAULT]); |
|
197
|
|
|
} |
|
198
|
|
|
// Matching UserAgent |
|
199
|
|
|
if (isset($this->rules[$this->userAgent])) { |
|
200
|
|
|
$rules = array_merge($rules, $this->rules[$this->userAgent]); |
|
201
|
|
|
} |
|
202
|
|
|
// Result |
|
203
|
|
|
return $rules; |
|
204
|
|
|
} |
|
205
|
|
|
|
|
206
|
|
|
/** |
|
207
|
|
|
* Export all rules for all UserAgents |
|
208
|
|
|
* |
|
209
|
|
|
* @return array |
|
210
|
|
|
*/ |
|
211
|
|
|
public function export() |
|
212
|
|
|
{ |
|
213
|
|
|
return $this->rules; |
|
214
|
|
|
} |
|
215
|
|
|
} |
VIPnytt /
RobotsTagParser
