|
1
|
|
|
<?php |
|
2
|
|
|
|
|
3
|
|
|
/* |
|
4
|
|
|
* This file is part of the bisarca/robots-txt package. |
|
5
|
|
|
* |
|
6
|
|
|
* (c) Emanuele Minotto <[email protected]> |
|
7
|
|
|
* |
|
8
|
|
|
* For the full copyright and license information, please view the LICENSE |
|
9
|
|
|
* file that was distributed with this source code. |
|
10
|
|
|
*/ |
|
11
|
|
|
|
|
12
|
|
|
namespace Bisarca\RobotsTxt; |
|
13
|
|
|
|
|
14
|
|
|
use Bisarca\RobotsTxt\Directive\DirectiveInterface; |
|
15
|
|
|
use Bisarca\RobotsTxt\Directive\NonGroupInterface; |
|
16
|
|
|
use Bisarca\RobotsTxt\Directive\StartOfGroupInterface; |
|
17
|
|
|
use Bisarca\RobotsTxt\Exception\ExceptionInterface; |
|
18
|
|
|
use Bisarca\RobotsTxt\Exception\MissingDirectiveException; |
|
19
|
|
|
|
|
20
|
|
|
/** |
|
21
|
|
|
* Robots.txt file parser. |
|
22
|
|
|
*/ |
|
23
|
|
|
class Parser |
|
24
|
|
|
{ |
|
25
|
|
|
/** |
|
26
|
|
|
* Registered directives. |
|
27
|
|
|
* |
|
28
|
|
|
* @var string[] |
|
29
|
|
|
*/ |
|
30
|
|
|
const DIRECTIVES = [ |
|
31
|
|
|
'allow' => Directive\Allow::class, |
|
32
|
|
|
'comment' => Directive\Comment::class, |
|
33
|
|
|
'disallow' => Directive\Disallow::class, |
|
34
|
|
|
'host' => Directive\Host::class, |
|
35
|
|
|
'sitemap' => Directive\Sitemap::class, |
|
36
|
|
|
'user-agent' => Directive\UserAgent::class, |
|
37
|
|
|
]; |
|
38
|
|
|
|
|
39
|
|
|
/** |
|
40
|
|
|
* Parse robots.txt content. |
|
41
|
|
|
* |
|
42
|
|
|
* @param string $content |
|
43
|
|
|
* |
|
44
|
|
|
* @return Rulesets |
|
45
|
|
|
*/ |
|
46
|
|
|
public function parse(string $content): Rulesets |
|
47
|
|
|
{ |
|
48
|
|
|
$rows = $this->extractRows($content); |
|
49
|
|
|
$groups = []; |
|
50
|
|
|
|
|
51
|
|
|
$counter = -1; |
|
52
|
|
|
$type = false; |
|
53
|
|
|
|
|
54
|
|
|
foreach ($rows as $row) { |
|
55
|
|
|
try { |
|
56
|
|
|
$directive = $this->createDirective($row); |
|
57
|
|
|
} catch (ExceptionInterface $exception) { |
|
58
|
|
|
continue; |
|
59
|
|
|
} |
|
60
|
|
|
|
|
61
|
|
|
$previous = $type; |
|
62
|
|
|
$type = $directive instanceof StartOfGroupInterface; |
|
63
|
|
|
|
|
64
|
|
|
if ( |
|
65
|
|
|
$directive instanceof NonGroupInterface || |
|
66
|
|
|
( |
|
67
|
|
|
$type && |
|
68
|
|
|
!($type && $previous) |
|
69
|
|
|
) |
|
70
|
|
|
) { |
|
71
|
|
|
++$counter; |
|
72
|
|
|
} |
|
73
|
|
|
|
|
74
|
|
|
if (!isset($groups[$counter])) { |
|
75
|
|
|
$groups[$counter] = []; |
|
76
|
|
|
} |
|
77
|
|
|
|
|
78
|
|
|
$groups[$counter][] = $directive; |
|
79
|
|
|
} |
|
80
|
|
|
|
|
81
|
|
|
// any group-member records without a preceding |
|
82
|
|
|
// start-of-group record are ignored |
|
83
|
|
|
unset($groups[-1]); |
|
84
|
|
|
|
|
85
|
|
|
foreach ($groups as $index => $group) { |
|
86
|
|
|
$groups[$index] = new Ruleset(...$group); |
|
87
|
|
|
} |
|
88
|
|
|
|
|
89
|
|
|
return new Rulesets(...$groups); |
|
90
|
|
|
} |
|
91
|
|
|
|
|
92
|
|
|
/** |
|
93
|
|
|
* Extract single rows from main content. |
|
94
|
|
|
* |
|
95
|
|
|
* @param string $content |
|
96
|
|
|
* |
|
97
|
|
|
* @return array |
|
98
|
|
|
*/ |
|
99
|
|
|
private function extractRows(string $content): array |
|
100
|
|
|
{ |
|
101
|
|
|
// split by EOL |
|
102
|
|
|
$rows = explode(PHP_EOL, $content); |
|
103
|
|
|
|
|
104
|
|
|
// remove comments and wrapper spaces |
|
105
|
|
|
$rows = array_map(function ($row) { |
|
106
|
|
|
return trim(preg_replace('/^(.*)#.*/', '$1', $row)); |
|
107
|
|
|
}, $rows); |
|
108
|
|
|
|
|
109
|
|
|
// empty lines aren't useful |
|
110
|
|
|
return array_filter($rows); |
|
111
|
|
|
} |
|
112
|
|
|
|
|
113
|
|
|
/** |
|
114
|
|
|
* Creates a directive from the raw line contained in the robots.txt file. |
|
115
|
|
|
* |
|
116
|
|
|
* @param string $row Raw line |
|
117
|
|
|
* |
|
118
|
|
|
* @throws MissingDirectiveException If no directive is available |
|
119
|
|
|
* |
|
120
|
|
|
* @return DirectiveInterface |
|
121
|
|
|
*/ |
|
122
|
|
|
private function createDirective(string $row): DirectiveInterface |
|
123
|
|
|
{ |
|
124
|
|
|
$directives = array_filter( |
|
125
|
|
|
self::DIRECTIVES, |
|
126
|
|
|
function ($field) use ($row) { |
|
127
|
|
|
return preg_match(sprintf('/^%s:\s+.+/i', $field), $row); |
|
128
|
|
|
}, |
|
129
|
|
|
ARRAY_FILTER_USE_KEY |
|
130
|
|
|
); |
|
131
|
|
|
$directives = array_values($directives); |
|
132
|
|
|
|
|
133
|
|
|
// no directives found for this row |
|
134
|
|
|
// no action required |
|
135
|
|
|
if (empty($directives)) { |
|
136
|
|
|
throw MissingDirectiveException::create($row); |
|
137
|
|
|
} |
|
138
|
|
|
|
|
139
|
|
|
// directives should be sorted by priority |
|
140
|
|
|
return new $directives[0]($row); |
|
141
|
|
|
} |
|
142
|
|
|
} |
|
143
|
|
|
|
bisarca /
robots-txt
