|
1
|
|
|
<?php |
|
2
|
|
|
|
|
3
|
|
|
namespace Spatie\Robots; |
|
4
|
|
|
|
|
5
|
|
|
class RobotsTxt |
|
6
|
|
|
{ |
|
7
|
|
|
protected static $robotsCache = []; |
|
8
|
|
|
|
|
9
|
|
|
protected $disallowsPerUserAgent = []; |
|
10
|
|
|
|
|
11
|
|
|
public static function readFrom(string $source): self |
|
12
|
|
|
{ |
|
13
|
|
|
$content = @file_get_contents($source) ?? ''; |
|
14
|
|
|
|
|
15
|
|
|
return new self($content); |
|
16
|
|
|
} |
|
17
|
|
|
|
|
18
|
|
|
public function __construct(string $content) |
|
19
|
|
|
{ |
|
20
|
|
|
$this->disallowsPerUserAgent = $this->getDisallowsPerUserAgent($content); |
|
21
|
|
|
} |
|
22
|
|
|
|
|
23
|
|
|
public static function create(string $source): self |
|
24
|
|
|
{ |
|
25
|
|
|
if ( |
|
26
|
|
|
strpos($source, 'http') !== false |
|
27
|
|
|
&& strpos($source, 'robots.txt') !== false |
|
28
|
|
|
) { |
|
29
|
|
|
return self::readFrom($source); |
|
30
|
|
|
} |
|
31
|
|
|
|
|
32
|
|
|
return new self($source); |
|
33
|
|
|
} |
|
34
|
|
|
|
|
35
|
|
|
// Google says (https://developers.google.com/search/reference/robots_txt) : |
|
36
|
|
|
// Only one group of group-member records is valid for a particular crawler. |
|
37
|
|
|
// The crawler must determine the correct group of records by finding the group |
|
38
|
|
|
// with the most specific user-agent that still matches. |
|
39
|
|
|
public function allows(string $url, ?string $userAgent = '*'): bool |
|
40
|
|
|
{ |
|
41
|
|
|
$userAgent = strtolower($userAgent); |
|
42
|
|
|
|
|
43
|
|
|
if ($userAgent === null) { |
|
44
|
|
|
$userAgent = '*'; |
|
45
|
|
|
} |
|
46
|
|
|
|
|
47
|
|
|
$path = parse_url($url, PHP_URL_PATH) ?? ''; |
|
48
|
|
|
|
|
49
|
|
|
if ($userAgent != '*' && isset($this->disallowsPerUserAgent[$userAgent])) { |
|
50
|
|
|
return ! $this->pathIsDenied($path, $this->disallowsPerUserAgent[$userAgent]) ?? true; |
|
51
|
|
|
} elseif ($userAgent != '*' && $wildCardUserAgent = $this->getWildCardUserAgent($userAgent)) { |
|
52
|
|
|
return ! $this->pathIsDenied($path, $this->disallowsPerUserAgent[$wildCardUserAgent]) ?? true; |
|
53
|
|
|
} elseif (isset($this->disallowsPerUserAgent['*'])) { |
|
54
|
|
|
return ! $this->pathIsDenied($path, $this->disallowsPerUserAgent['*']) ?? true; |
|
55
|
|
|
} |
|
56
|
|
|
|
|
57
|
|
|
return true; |
|
58
|
|
|
} |
|
59
|
|
|
|
|
60
|
|
|
protected function getWildCardUserAgent(string $userAgent): ?string |
|
61
|
|
|
{ |
|
62
|
|
|
if ($userAgent !== '*') { |
|
63
|
|
|
for ($i = 1; $i <= strlen($userAgent); $i++) { |
|
64
|
|
|
$wildCardUserAgent = substr($userAgent, 0, $i).'*'; |
|
65
|
|
|
if (isset($this->disallowsPerUserAgent[$wildCardUserAgent])) { |
|
66
|
|
|
return $wildCardUserAgent; |
|
67
|
|
|
} |
|
68
|
|
|
} |
|
69
|
|
|
} |
|
70
|
|
|
|
|
71
|
|
|
return null; |
|
72
|
|
|
} |
|
73
|
|
|
|
|
74
|
|
|
protected function pathIsDenied(string $path, array $rules) |
|
75
|
|
|
{ |
|
76
|
|
|
foreach ($rules as $pattern => $rule) { |
|
77
|
|
|
if ($this->match($pattern, $path)) { |
|
78
|
|
|
return $rule; |
|
79
|
|
|
} |
|
80
|
|
|
} |
|
81
|
|
|
} |
|
82
|
|
|
|
|
83
|
|
|
protected function complexRule($path): boolean |
|
84
|
|
|
{ |
|
85
|
|
|
return strpos($path, ['$', '*']); |
|
86
|
|
|
} |
|
87
|
|
|
|
|
88
|
|
|
protected function match($pattern, $string) |
|
89
|
|
|
{ |
|
90
|
|
|
$pattern = preg_quote($pattern, '/'); |
|
91
|
|
|
$pattern = str_replace('\*', '.*', $pattern); |
|
92
|
|
|
//$pattern = preg_replace('/\\\$$/', '$', $pattern); // is not working |
|
93
|
|
|
$pattern = substr($pattern, -2) == '\$' ? substr($pattern, 0, strlen($pattern) - 2).'$' : $pattern; |
|
94
|
|
|
$pattern = preg_replace('/\/$/', '/?', $pattern); |
|
95
|
|
|
|
|
96
|
|
|
return (bool) preg_match('/^'.$pattern.'/', $string); |
|
97
|
|
|
} |
|
98
|
|
|
|
|
99
|
|
|
protected function getDisallowsPerUserAgent(string $content): array |
|
100
|
|
|
{ |
|
101
|
|
|
$lines = explode(PHP_EOL, $content); |
|
102
|
|
|
|
|
103
|
|
|
$lines = array_filter($lines); |
|
104
|
|
|
|
|
105
|
|
|
$disallowsPerUserAgent = []; |
|
106
|
|
|
|
|
107
|
|
|
$currentUserAgent = null; |
|
108
|
|
|
|
|
109
|
|
|
foreach ($lines as $line) { |
|
110
|
|
|
if ($this->isUserAgentLine($line)) { |
|
111
|
|
|
$disallowsPerUserAgent[$this->parseUserAgent($line)] = []; |
|
112
|
|
|
|
|
113
|
|
|
$currentUserAgent = &$disallowsPerUserAgent[$this->parseUserAgent($line)]; |
|
114
|
|
|
|
|
115
|
|
|
continue; |
|
116
|
|
|
} |
|
117
|
|
|
|
|
118
|
|
|
if ($currentUserAgent === null) { |
|
119
|
|
|
continue; |
|
120
|
|
|
} |
|
121
|
|
|
|
|
122
|
|
|
list($pattern, $rule) = $this->parse($line); |
|
123
|
|
|
if ($pattern !== null) { // other than allow/disallow |
|
124
|
|
|
$currentUserAgent[$pattern] = $rule; |
|
125
|
|
|
} |
|
126
|
|
|
} |
|
127
|
|
|
|
|
128
|
|
|
return $this->orderRules($disallowsPerUserAgent); |
|
129
|
|
|
} |
|
130
|
|
|
|
|
131
|
|
|
// Google says (https://developers.google.com/search/reference/robots_txt) : |
|
132
|
|
|
// At a group-member level, in particular for allow and disallow directives, the most specific rule |
|
133
|
|
|
// based on the length of the [path] entry will trump the less specific (shorter) rule. The order of |
|
134
|
|
|
// precedence for rules with wildcards is undefined. |
|
135
|
|
|
protected function orderRules(array $disallowsPerUserAgent): array |
|
136
|
|
|
{ |
|
137
|
|
|
foreach ($disallowsPerUserAgent as $userAgent => $rules) { |
|
138
|
|
|
array_multisort(array_map('strlen', array_keys($rules)), SORT_DESC, $rules); |
|
|
|
|
|
|
139
|
|
|
$disallowsPerUserAgent[$userAgent] = $rules; |
|
140
|
|
|
} |
|
141
|
|
|
|
|
142
|
|
|
return $disallowsPerUserAgent; |
|
143
|
|
|
} |
|
144
|
|
|
|
|
145
|
|
|
protected function isUserAgentLine(string $line): bool |
|
146
|
|
|
{ |
|
147
|
|
|
return stripos(str_replace(' ', '', $line), 'user-agent:') === 0; |
|
148
|
|
|
} |
|
149
|
|
|
|
|
150
|
|
|
protected function parseUserAgent(string $line): string |
|
151
|
|
|
{ |
|
152
|
|
|
return strtolower(trim(preg_replace('/^User-agent\s*:/i', '', trim($line)))); |
|
153
|
|
|
} |
|
154
|
|
|
|
|
155
|
|
|
protected function parse(string $line): ?array |
|
156
|
|
|
{ |
|
157
|
|
|
$line = trim(preg_replace('/\s+!/', ':', $line)); |
|
158
|
|
|
|
|
159
|
|
|
if (stripos($line, 'disallow:') === 0) { |
|
160
|
|
|
return [trim(preg_replace('/^disallow:/i', '', $line)), true]; |
|
161
|
|
|
} |
|
162
|
|
|
|
|
163
|
|
|
if (stripos($line, 'allow:') === 0) { |
|
164
|
|
|
return [trim(preg_replace('/^allow:/i', '', $line)), false]; |
|
165
|
|
|
} |
|
166
|
|
|
|
|
167
|
|
|
// else: could be crawl-delay, sitemap... |
|
168
|
|
|
} |
|
169
|
|
|
|
|
170
|
|
|
protected function isUrlInDirectory(string $url, string $path): bool |
|
171
|
|
|
{ |
|
172
|
|
|
return strpos($url, $path) === 0; |
|
173
|
|
|
} |
|
174
|
|
|
} |
|
175
|
|
|
|
spatie /
robots-txt
Loading history...