1
|
|
|
<?php |
2
|
|
|
|
3
|
|
|
namespace PiedWeb\UrlHarvester; |
4
|
|
|
|
5
|
|
|
use phpUri; |
6
|
|
|
|
7
|
|
|
class ExtractLinks |
8
|
|
|
{ |
9
|
|
|
const SELECT_A = 'a[href]'; |
10
|
|
|
|
11
|
|
|
const SELECT_ALL = 'href,src'; |
12
|
|
|
|
13
|
|
|
/** |
14
|
|
|
* @var \simple_html_dom |
15
|
|
|
*/ |
16
|
|
|
private $dom; |
17
|
|
|
|
18
|
|
|
/** |
19
|
|
|
* @var string |
20
|
|
|
*/ |
21
|
|
|
private $baseUrl; |
22
|
|
|
|
23
|
|
|
/** |
24
|
|
|
* @var string |
25
|
|
|
*/ |
26
|
|
|
private $selector; |
27
|
|
|
|
28
|
|
|
/** |
29
|
|
|
* @param string $dom HTML code from the page |
30
|
|
|
* @param string $baseUrl To get absolute urls |
31
|
|
|
* @param string $selector |
32
|
|
|
* |
33
|
|
|
* @return array |
34
|
|
|
*/ |
35
|
9 |
|
public static function get(\simple_html_dom $dom, string $baseUrl, $selector = self::SELECT_A) |
36
|
|
|
{ |
37
|
9 |
|
$self = new self(); |
38
|
|
|
|
39
|
9 |
|
$self->selector = $selector; |
40
|
9 |
|
$self->baseUrl = $baseUrl; |
41
|
9 |
|
$self->dom = $dom; |
42
|
|
|
|
43
|
9 |
|
return $self->extractLinks(); |
44
|
|
|
} |
45
|
|
|
|
46
|
9 |
|
private function __construct() |
47
|
|
|
{ |
48
|
9 |
|
} |
49
|
|
|
|
50
|
|
|
/** |
51
|
|
|
* @return array |
52
|
|
|
*/ |
53
|
9 |
|
private function extractLinks() |
54
|
|
|
{ |
55
|
9 |
|
$links = []; |
56
|
9 |
|
$elements = $this->getElements(); |
57
|
|
|
|
58
|
9 |
|
if ($elements) { |
59
|
9 |
|
foreach ($elements as $element) { |
60
|
9 |
|
$href = $this->getUrl($element); |
61
|
|
|
|
62
|
9 |
|
if (null !== $href) { |
63
|
9 |
|
$links[] = new Link($href, $element); |
64
|
|
|
} |
65
|
|
|
} |
66
|
|
|
} |
67
|
|
|
|
68
|
9 |
|
return $links; |
69
|
|
|
} |
70
|
|
|
|
71
|
9 |
|
private function getElements() |
72
|
|
|
{ |
73
|
9 |
|
if (self::SELECT_A == $this->selector) { |
74
|
6 |
|
return $this->dom->find($this->selector); |
75
|
|
|
} else { |
76
|
3 |
|
return $this->dom->find('['.implode('],*[', explode(',', $this->selector)).']'); |
77
|
|
|
} |
78
|
|
|
} |
79
|
|
|
|
80
|
|
|
/** |
81
|
|
|
* @return string|null |
82
|
|
|
*/ |
83
|
9 |
|
private function getUrl($element) |
84
|
|
|
{ |
85
|
9 |
|
if (self::SELECT_A == $this->selector) { |
86
|
6 |
|
$href = $element->href; |
87
|
|
|
} else { |
88
|
3 |
|
$attributes = explode(',', $this->selector); |
89
|
3 |
|
foreach ($attributes as $attribute) { |
90
|
3 |
|
$href = $element->$attribute; |
91
|
3 |
|
if (null !== $href) { |
92
|
3 |
|
break; |
93
|
|
|
} |
94
|
|
|
} |
95
|
|
|
} |
96
|
|
|
|
97
|
9 |
|
$href = $this->isItALink($href) ? $href : null; |
|
|
|
|
98
|
9 |
|
$parsed = phpUri::parse($this->baseUrl)->join($href); |
99
|
9 |
|
$href = null !== $href ? ($parsed ? $parsed : $href) : null; |
100
|
|
|
|
101
|
9 |
|
return $href; |
102
|
|
|
} |
103
|
|
|
|
104
|
9 |
|
private function isItALink($href) |
105
|
|
|
{ |
106
|
|
|
return |
107
|
9 |
|
0 !== stripos($href, 'mailto:') |
108
|
9 |
|
&& 0 !== strpos($href, 'javascript:'); |
109
|
|
|
} |
110
|
|
|
} |
111
|
|
|
|