|
1
|
|
|
<?php |
|
2
|
|
|
namespace vipnytt\RobotsTxtParser\Parser; |
|
3
|
|
|
|
|
4
|
|
|
use vipnytt\RobotsTxtParser\Parser\Directives\DirectiveParserCommons; |
|
5
|
|
|
use vipnytt\RobotsTxtParser\Parser\Directives\RootDirectiveHandler; |
|
6
|
|
|
use vipnytt\RobotsTxtParser\RobotsTxtInterface; |
|
7
|
|
|
|
|
8
|
|
|
/** |
|
9
|
|
|
* Class RobotsTxtParser |
|
10
|
|
|
* |
|
11
|
|
|
* @link https://developers.google.com/webmasters/control-crawl-index/docs/robots_txt |
|
12
|
|
|
* @link https://yandex.com/support/webmaster/controlling-robot/robots-txt.xml |
|
13
|
|
|
* @link http://www.robotstxt.org/robotstxt.html |
|
14
|
|
|
* @link https://www.w3.org/TR/html4/appendix/notes.html#h-B.4.1.1 |
|
15
|
|
|
* @link http://www.conman.org/people/spc/robots2.html |
|
16
|
|
|
* |
|
17
|
|
|
* @package vipnytt\RobotsTxtParser\Parser |
|
18
|
|
|
*/ |
|
19
|
|
|
class RobotsTxtParser implements RobotsTxtInterface |
|
20
|
|
|
{ |
|
21
|
|
|
use DirectiveParserCommons; |
|
22
|
|
|
use UrlParser; |
|
23
|
|
|
|
|
24
|
|
|
/** |
|
25
|
|
|
* Directive white list |
|
26
|
|
|
*/ |
|
27
|
|
|
const TOP_LEVEL_DIRECTIVES = [ |
|
28
|
|
|
self::DIRECTIVE_CLEAN_PARAM => 'cleanParam', |
|
29
|
|
|
self::DIRECTIVE_HOST => 'host', |
|
30
|
|
|
self::DIRECTIVE_SITEMAP => 'sitemap', |
|
31
|
|
|
self::DIRECTIVE_USER_AGENT => 'userAgent', |
|
32
|
|
|
]; |
|
33
|
|
|
|
|
34
|
|
|
/** |
|
35
|
|
|
* Root directive handler |
|
36
|
|
|
* @var RootDirectiveHandler |
|
37
|
|
|
*/ |
|
38
|
|
|
protected $handler; |
|
39
|
|
|
|
|
40
|
|
|
/** |
|
41
|
|
|
* Current user-agent(s) |
|
42
|
|
|
* @var array |
|
43
|
|
|
*/ |
|
44
|
|
|
private $userAgents; |
|
45
|
|
|
|
|
46
|
|
|
/** |
|
47
|
|
|
* Previous directive |
|
48
|
|
|
* @var string |
|
49
|
|
|
*/ |
|
50
|
|
|
private $previousDirective; |
|
51
|
|
|
|
|
52
|
|
|
/** |
|
53
|
|
|
* Basic constructor. |
|
54
|
|
|
* |
|
55
|
|
|
* @param string $baseUri |
|
56
|
|
|
* @param string $content |
|
57
|
|
|
*/ |
|
58
|
|
|
public function __construct($baseUri, $content) |
|
59
|
|
|
{ |
|
60
|
|
|
mb_internal_encoding(self::ENCODING); |
|
61
|
|
|
$this->handler = new RootDirectiveHandler($this->urlBase($this->urlEncode($baseUri))); |
|
62
|
|
|
$this->parseTxt($content); |
|
63
|
|
|
} |
|
64
|
|
|
|
|
65
|
|
|
/** |
|
66
|
|
|
* Client robots.txt |
|
67
|
|
|
* |
|
68
|
|
|
* @param string $txt |
|
69
|
|
|
* @return void |
|
70
|
|
|
*/ |
|
71
|
|
|
private function parseTxt($txt) |
|
72
|
|
|
{ |
|
73
|
|
|
$lines = array_filter(array_map('trim', mb_split('\r\n|\n|\r', $txt))); |
|
74
|
|
|
// Client each line individually |
|
75
|
|
|
foreach ($lines as $line) { |
|
76
|
|
|
// Limit rule length |
|
77
|
|
|
$line = mb_substr($line, 0, self::MAX_LENGTH_RULE); |
|
78
|
|
|
// Remove comments |
|
79
|
|
|
$line = mb_split('#', $line, 2)[0]; |
|
80
|
|
|
// Client line |
|
81
|
|
|
$this->add($line); |
|
82
|
|
|
} |
|
83
|
|
|
} |
|
84
|
|
|
|
|
85
|
|
|
/** |
|
86
|
|
|
* Add line |
|
87
|
|
|
* |
|
88
|
|
|
* @param string $line |
|
89
|
|
|
* @return bool |
|
90
|
|
|
*/ |
|
91
|
|
|
public function add($line) |
|
92
|
|
|
{ |
|
93
|
|
|
$previousDirective = $this->previousDirective; |
|
94
|
|
|
$pair = $this->generateRulePair($line, array_keys(self::TOP_LEVEL_DIRECTIVES)); |
|
95
|
|
|
if ($pair === false) { |
|
96
|
|
|
$this->previousDirective = $line; |
|
97
|
|
|
return $this->handler->userAgent()->add($line); |
|
98
|
|
|
} elseif ($pair['directive'] === self::DIRECTIVE_USER_AGENT) { |
|
99
|
|
|
if ($previousDirective !== self::DIRECTIVE_USER_AGENT) { |
|
100
|
|
|
$this->userAgents = []; |
|
101
|
|
|
} |
|
102
|
|
|
$this->userAgents[] = $pair['value']; |
|
103
|
|
|
$this->previousDirective = $pair['directive']; |
|
104
|
|
|
return $this->handler->userAgent()->set($this->userAgents); |
|
105
|
|
|
} |
|
106
|
|
|
$this->previousDirective = $pair['directive']; |
|
107
|
|
|
return $this->handler->{self::TOP_LEVEL_DIRECTIVES[$pair['directive']]}()->add($pair['value']); |
|
108
|
|
|
} |
|
109
|
|
|
|
|
110
|
|
|
/** |
|
111
|
|
|
* Render |
|
112
|
|
|
* |
|
113
|
|
|
* @param string $lineSeparator |
|
114
|
|
|
* @return string |
|
115
|
|
|
*/ |
|
116
|
|
|
public function render($lineSeparator = "\n") |
|
117
|
|
|
{ |
|
118
|
|
|
return implode($lineSeparator, array_merge( |
|
119
|
|
|
$this->handler->host()->render(), |
|
120
|
|
|
$this->handler->cleanParam()->render(), |
|
121
|
|
|
$this->handler->sitemap()->render(), |
|
122
|
|
|
$this->handler->userAgent()->render() |
|
123
|
|
|
)); |
|
124
|
|
|
} |
|
125
|
|
|
|
|
126
|
|
|
/** |
|
127
|
|
|
* Rule array |
|
128
|
|
|
* |
|
129
|
|
|
* @return array |
|
130
|
|
|
*/ |
|
131
|
|
|
public function export() |
|
132
|
|
|
{ |
|
133
|
|
|
return [ |
|
134
|
|
|
self::DIRECTIVE_HOST => $this->handler->host()->client()->export(), |
|
135
|
|
|
self::DIRECTIVE_CLEAN_PARAM => $this->handler->cleanParam()->client()->export(), |
|
136
|
|
|
self::DIRECTIVE_SITEMAP => $this->handler->sitemap()->client()->export(), |
|
137
|
|
|
self::DIRECTIVE_USER_AGENT => $this->handler->userAgent()->export(), |
|
138
|
|
|
]; |
|
139
|
|
|
} |
|
140
|
|
|
} |
|
141
|
|
|
|
VIPnytt /
RobotsTxtParser
