|
1
|
|
|
<?php |
|
2
|
|
|
namespace vipnytt\RobotsTxtParser\Parser; |
|
3
|
|
|
|
|
4
|
|
|
use vipnytt\RobotsTxtParser\Exceptions\ParserException; |
|
5
|
|
|
|
|
6
|
|
|
/** |
|
7
|
|
|
* Trait Toolbox |
|
8
|
|
|
* |
|
9
|
|
|
* @package vipnytt\RobotsTxtParser\Parser |
|
10
|
|
|
*/ |
|
11
|
|
|
trait Toolbox |
|
12
|
|
|
{ |
|
13
|
|
|
/** |
|
14
|
|
|
* Check basic rule |
|
15
|
|
|
* |
|
16
|
|
|
* @param string $path |
|
17
|
|
|
* @param array $paths |
|
18
|
|
|
* @return bool |
|
19
|
|
|
*/ |
|
20
|
|
|
protected function checkPath($path, $paths) |
|
21
|
|
|
{ |
|
22
|
|
|
foreach ($paths as $rule) { |
|
23
|
|
|
$escape = ['?' => '\?', '.' => '\.', '*' => '.*']; |
|
24
|
|
|
foreach ($escape as $search => $replace) { |
|
25
|
|
|
$rule = str_replace($search, $replace, $rule); |
|
26
|
|
|
} |
|
27
|
|
|
/** |
|
28
|
|
|
* Warning: preg_match need to be replaced |
|
29
|
|
|
* |
|
30
|
|
|
* Bug report |
|
31
|
|
|
* @link https://github.com/t1gor/Robots.txt-Parser-Class/issues/62 |
|
32
|
|
|
* |
|
33
|
|
|
* An robots.txt parser, where a bug-fix is planned |
|
34
|
|
|
* @link https://github.com/diggin/Diggin_RobotRules |
|
35
|
|
|
* |
|
36
|
|
|
* The solution? |
|
37
|
|
|
* PHP PEG (parsing expression grammar) |
|
38
|
|
|
* @link https://github.com/hafriedlander/php-peg |
|
39
|
|
|
*/ |
|
40
|
|
|
try { |
|
41
|
|
|
if (!preg_match('#' . $rule . '#', $path)) { |
|
42
|
|
|
// Rule does not match |
|
43
|
|
|
continue; |
|
44
|
|
|
} else if (mb_stripos($rule, '$') === false) { |
|
45
|
|
|
// No special parsing required |
|
46
|
|
|
return true; |
|
47
|
|
|
} else if (($wildcardPos = mb_strrpos($rule, '*')) !== false) { |
|
48
|
|
|
// Rule contains both an end anchor ($) and wildcard (*) |
|
49
|
|
|
$afterWildcard = mb_substr($rule, $wildcardPos + 1, mb_strlen($rule) - $wildcardPos - 2); |
|
50
|
|
|
if ($afterWildcard == mb_substr($path, -mb_strlen($afterWildcard))) { |
|
51
|
|
|
return true; |
|
52
|
|
|
} |
|
53
|
|
|
} else if (mb_substr($rule, 0, -1) == $path) { |
|
54
|
|
|
// Rule does contains an end anchor |
|
55
|
|
|
return true; |
|
56
|
|
|
} |
|
57
|
|
|
} catch (\Exception $e) { |
|
58
|
|
|
// An preg_match bug has occurred |
|
59
|
|
|
} |
|
60
|
|
|
} |
|
61
|
|
|
return false; |
|
62
|
|
|
} |
|
63
|
|
|
|
|
64
|
|
|
/** |
|
65
|
|
|
* Generate directive/rule pair |
|
66
|
|
|
* |
|
67
|
|
|
* @param string $line |
|
68
|
|
|
* @param array $whiteList |
|
69
|
|
|
* @return array|false |
|
70
|
|
|
*/ |
|
71
|
|
|
protected function generateRulePair($line, $whiteList) |
|
72
|
|
|
{ |
|
73
|
|
|
$whiteList = array_map('mb_strtolower', $whiteList); |
|
74
|
|
|
// Split by directive and rule |
|
75
|
|
|
$pair = array_map('trim', mb_split(':', $line, 2)); |
|
76
|
|
|
// Check if the line contains a rule |
|
77
|
|
|
if ( |
|
78
|
|
|
empty($pair[1]) || |
|
79
|
|
|
empty($pair[0]) || |
|
80
|
|
|
!in_array(($pair[0] = mb_strtolower($pair[0])), $whiteList) |
|
81
|
|
|
) { |
|
82
|
|
|
// Line does not contain any supported directive |
|
83
|
|
|
return false; |
|
84
|
|
|
} |
|
85
|
|
|
return [ |
|
86
|
|
|
'directive' => $pair[0], |
|
87
|
|
|
'value' => $pair[1], |
|
88
|
|
|
]; |
|
89
|
|
|
} |
|
90
|
|
|
|
|
91
|
|
|
/** |
|
92
|
|
|
* Validate directive |
|
93
|
|
|
* |
|
94
|
|
|
* @param $directive |
|
95
|
|
|
* @param $directives |
|
96
|
|
|
* @return string |
|
97
|
|
|
* @throws ParserException |
|
98
|
|
|
*/ |
|
99
|
|
|
protected function validateDirective($directive, $directives) |
|
100
|
|
|
{ |
|
101
|
|
|
if (!in_array($directive, $directives, true)) { |
|
102
|
|
|
throw new ParserException('Directive is not allowed here'); |
|
103
|
|
|
} |
|
104
|
|
|
return mb_strtolower($directive); |
|
105
|
|
|
} |
|
106
|
|
|
} |
|
107
|
|
|
|
VIPnytt /
RobotsTxtParser
