|
1
|
|
|
<?php |
|
2
|
|
|
|
|
3
|
|
|
namespace arc\html; |
|
4
|
|
|
|
|
5
|
|
|
class Parser |
|
6
|
|
|
{ |
|
7
|
|
|
public $options = [ |
|
8
|
|
|
'libxml_options' => 0 |
|
9
|
|
|
]; |
|
10
|
|
|
|
|
11
|
8 |
|
public function __construct( $options = array() ) |
|
12
|
|
|
{ |
|
13
|
8 |
|
$optionList = [ 'libxml_options' ]; |
|
14
|
8 |
|
foreach( $options as $option => $optionValue ) { |
|
15
|
|
|
if ( in_array( $option, $optionList ) ) { |
|
16
|
|
|
$this->{$option} = $optionValue; |
|
17
|
|
|
} |
|
18
|
4 |
|
} |
|
19
|
8 |
|
} |
|
20
|
|
|
|
|
21
|
8 |
|
public function parse( $html, $encoding = null ) |
|
22
|
|
|
{ |
|
23
|
8 |
|
if ( !$html ) { |
|
24
|
|
|
return \arc\html\Proxy( null ); |
|
25
|
|
|
} |
|
26
|
8 |
|
if ( $html instanceof Proxy ) { // already parsed |
|
27
|
|
|
return $html; |
|
28
|
|
|
} |
|
29
|
8 |
|
$html = (string) $html; |
|
30
|
8 |
|
if ( stripos($html, '<body>')!==false ) { |
|
31
|
8 |
|
return $this->parseFull( $html, $encoding ); |
|
32
|
|
|
} else { |
|
33
|
2 |
|
return $this->parsePartial( $html, $encoding ); |
|
34
|
|
|
} |
|
35
|
|
|
} |
|
36
|
|
|
|
|
37
|
2 |
|
private function parsePartial( $html, $encoding ) |
|
38
|
|
|
{ |
|
39
|
2 |
|
$result = $this->parseFull( '<body id="ArcPartialHTML">'.$html.'</body>', $encoding ); |
|
40
|
2 |
|
if ( $result ) { |
|
41
|
2 |
|
$result = new \arc\html\Proxy( $result->find('#ArcPartialHTML')[0]->children(), $this ); |
|
42
|
1 |
|
} else { |
|
43
|
|
|
throw new \arc\UnknownError('Could not parse html.', \arc\exceptions::ILLEGAL_ARGUMENT ); |
|
44
|
|
|
} |
|
45
|
2 |
|
return $result; |
|
46
|
|
|
} |
|
47
|
|
|
|
|
48
|
|
|
private function throwError($prevErrorSetting) |
|
49
|
|
|
{ |
|
50
|
|
|
$errors = libxml_get_errors(); |
|
51
|
|
|
libxml_clear_errors(); |
|
52
|
|
|
libxml_use_internal_errors( $prevErrorSetting ); |
|
53
|
|
|
$message = 'Incorrect html passed.'; |
|
54
|
|
|
foreach ( $errors as $error ) { |
|
55
|
|
|
$message .= "\nline: ".$error->line."; column: ".$error->column."; ".$error->message; |
|
56
|
|
|
} |
|
57
|
|
|
throw new \arc\UnknownError( $message, \arc\exceptions::ILLEGAL_ARGUMENT ); |
|
58
|
|
|
} |
|
59
|
|
|
|
|
60
|
2 |
|
private function insertEncoding($html, $encoding) |
|
61
|
|
|
{ |
|
62
|
2 |
|
$meta = '<meta id="ArcTempEncoding" http-equiv="content-type" content="text/html; charset="'. htmlspecialchars($encoding) .'">'; |
|
63
|
2 |
|
if ( preg_match('/<head([^>]*)>/i', $html) ) { |
|
64
|
2 |
|
$html = preg_replace('/<head([^>]*)>/i', '<head\\1>'.$meta, $html); |
|
65
|
2 |
|
} else if ( preg_match('/<body([^>]*)>/i', $html) ) { |
|
66
|
2 |
|
$html = preg_replace('/<body([^>]*)>/i', '<head>'.$meta.'</head><body\\1>', $html); |
|
67
|
1 |
|
} else { |
|
68
|
|
|
$html = $meta.$html; |
|
69
|
|
|
} |
|
70
|
2 |
|
return $html; |
|
71
|
|
|
} |
|
72
|
|
|
|
|
73
|
2 |
|
private function removeEncoding( $dom ) |
|
74
|
|
|
{ |
|
75
|
2 |
|
$meta = $dom->getElementById('ArcTempEncoding'); |
|
76
|
2 |
|
$meta->parentNode->removeChild($meta); |
|
77
|
2 |
|
} |
|
78
|
|
|
|
|
79
|
8 |
|
private function parseFull( $html, $encoding ) |
|
80
|
|
|
{ |
|
81
|
8 |
|
$dom = new \DomDocument(); |
|
82
|
8 |
|
libxml_disable_entity_loader(); // prevents XXE attacks |
|
83
|
8 |
|
$prevErrorSetting = libxml_use_internal_errors(true); |
|
84
|
8 |
|
if ( $encoding ) { |
|
85
|
2 |
|
$html = $this->insertEncoding($html, $encoding); |
|
86
|
1 |
|
} |
|
87
|
8 |
|
if ( !$dom->loadHTML( $html, $this->options['libxml_options'] ) ) { |
|
88
|
|
|
$this->throwError($prevErrorSetting); |
|
89
|
|
|
} |
|
90
|
8 |
|
if ( $encoding ) { |
|
91
|
2 |
|
$this->removeEncoding( $dom ); |
|
92
|
1 |
|
} |
|
93
|
8 |
|
libxml_use_internal_errors( $prevErrorSetting ); |
|
94
|
8 |
|
return new \arc\html\Proxy( simplexml_import_dom( $dom ), $this ); |
|
95
|
|
|
} |
|
96
|
|
|
|
|
97
|
|
|
} |
|
98
|
|
|
|
Ariadne-CMS /
arc-html
