|
1
|
|
|
<?php |
|
2
|
|
|
|
|
3
|
|
|
namespace Jclyons52\PagePreview; |
|
4
|
|
|
|
|
5
|
|
|
use Jclyons52\PHPQuery\Document; |
|
6
|
|
|
|
|
7
|
|
|
class Crawler |
|
8
|
|
|
{ |
|
9
|
|
|
/** |
|
10
|
|
|
* PHPQuery document object that will be used to select elements |
|
11
|
|
|
* @var \Jclyons52\PHPQuery\Document |
|
12
|
|
|
*/ |
|
13
|
|
|
private $document; |
|
14
|
|
|
|
|
15
|
48 |
|
public function __construct(Document $document) |
|
16
|
|
|
{ |
|
17
|
48 |
|
$this->document = $document; |
|
18
|
48 |
|
} |
|
19
|
|
|
|
|
20
|
|
|
/** |
|
21
|
|
|
* @return array |
|
22
|
|
|
*/ |
|
23
|
45 |
|
public function getPreviewData($url) |
|
24
|
|
|
{ |
|
25
|
45 |
|
$title = $this->title(); |
|
26
|
|
|
|
|
27
|
45 |
|
$images = $this->formatImages($url); |
|
28
|
|
|
|
|
29
|
45 |
|
$description = $this->meta('description'); |
|
30
|
|
|
|
|
31
|
45 |
|
$meta = $this->meta(); |
|
32
|
|
|
|
|
33
|
45 |
|
$keywords = $this->metaKeywords(); |
|
34
|
|
|
|
|
35
|
45 |
|
if ($keywords !== []) { |
|
36
|
39 |
|
$meta['keywords'] = $keywords; |
|
37
|
26 |
|
} |
|
38
|
|
|
|
|
39
|
|
|
return [ |
|
40
|
45 |
|
'title' => $title, |
|
41
|
45 |
|
'images' => $images, |
|
42
|
45 |
|
'description' => $description, |
|
43
|
45 |
|
'url' => $url->original, |
|
44
|
45 |
|
'meta' => $meta, |
|
45
|
30 |
|
]; |
|
46
|
30 |
|
} |
|
47
|
|
|
|
|
48
|
|
|
/** |
|
49
|
|
|
* get source attributes of all image tags on the page |
|
50
|
|
|
* @return array<String> |
|
51
|
|
|
*/ |
|
52
|
48 |
|
public function images() |
|
53
|
2 |
|
{ |
|
54
|
48 |
|
$images = $this->document->querySelectorAll('img'); |
|
55
|
|
|
|
|
56
|
48 |
|
if ($images === []) { |
|
57
|
3 |
|
return []; |
|
58
|
|
|
} |
|
59
|
|
|
|
|
60
|
45 |
|
$urls = $images->attr('src'); |
|
61
|
|
|
|
|
62
|
|
|
$urls = array_filter($urls, function ($url) { |
|
63
|
45 |
|
$url = trim($url); |
|
64
|
45 |
|
return $url; |
|
65
|
45 |
|
}); |
|
66
|
|
|
|
|
67
|
45 |
|
return $urls; |
|
68
|
|
|
} |
|
69
|
|
|
|
|
70
|
|
|
/** |
|
71
|
|
|
* @return string |
|
72
|
|
|
*/ |
|
73
|
45 |
|
private function title() |
|
74
|
|
|
{ |
|
75
|
45 |
|
return $this->document->querySelector('title')->text(); |
|
76
|
|
|
} |
|
77
|
|
|
|
|
78
|
|
|
/** |
|
79
|
|
|
* @return mixed |
|
80
|
|
|
*/ |
|
81
|
45 |
|
private function metaKeywords() |
|
82
|
|
|
{ |
|
83
|
45 |
|
$keywordsElement = $this->document->querySelector('meta[name="keywords"]'); |
|
84
|
|
|
|
|
85
|
45 |
|
if (!$keywordsElement) { |
|
86
|
6 |
|
return []; |
|
87
|
|
|
} |
|
88
|
|
|
|
|
89
|
39 |
|
$keywordString = $keywordsElement->attr('content'); |
|
90
|
|
|
|
|
91
|
39 |
|
$keywords = explode(',', $keywordString); |
|
92
|
|
|
|
|
93
|
|
|
return array_map(function ($word) { |
|
94
|
39 |
|
return trim($word); |
|
95
|
|
|
|
|
96
|
39 |
|
}, $keywords); |
|
97
|
|
|
} |
|
98
|
|
|
|
|
99
|
|
|
/** |
|
100
|
|
|
* @param string $element |
|
101
|
|
|
* @return array |
|
102
|
|
|
*/ |
|
103
|
45 |
|
private function meta($element = null) |
|
104
|
|
|
{ |
|
105
|
45 |
|
$selector = "meta"; |
|
106
|
45 |
|
if ($element === null) { |
|
107
|
45 |
|
$metaTags = $this->document->querySelectorAll($selector); |
|
108
|
45 |
|
return $this->metaTagsToArray($metaTags); |
|
109
|
|
|
} |
|
110
|
|
|
|
|
111
|
45 |
|
$selector .= "[name='{$element}']"; |
|
112
|
45 |
|
$metaTags = $this->document->querySelector($selector); |
|
113
|
45 |
|
if ($metaTags === null) { |
|
114
|
6 |
|
return null; |
|
115
|
|
|
} |
|
116
|
39 |
|
return $metaTags->attr('content'); |
|
117
|
|
|
} |
|
118
|
|
|
|
|
119
|
|
|
/** |
|
120
|
|
|
* @param \Jclyons52\PHPQuery\Support\NodeCollection $metaTags |
|
121
|
|
|
* @return array |
|
122
|
|
|
*/ |
|
123
|
45 |
|
private function metaTagsToArray($metaTags) |
|
124
|
|
|
{ |
|
125
|
45 |
|
$values = []; |
|
126
|
45 |
|
foreach ($metaTags as $meta) { |
|
127
|
45 |
|
$name = $meta->attr('name'); |
|
128
|
45 |
|
if ($name === '') { |
|
129
|
45 |
|
$name = $meta->attr('property'); |
|
130
|
30 |
|
} |
|
131
|
45 |
|
$content = $meta->attr('content'); |
|
132
|
45 |
|
if ($name === '' || $content == '') { |
|
133
|
45 |
|
continue; |
|
134
|
|
|
} |
|
135
|
39 |
|
$values[$name] = $content; |
|
136
|
30 |
|
} |
|
137
|
45 |
|
return $values; |
|
138
|
|
|
} |
|
139
|
|
|
|
|
140
|
|
|
/** |
|
141
|
|
|
* @param $url Url |
|
142
|
|
|
* @return mixed |
|
143
|
|
|
*/ |
|
144
|
|
|
private function formatImages($url) |
|
145
|
|
|
{ |
|
146
|
45 |
|
return array_map(function ($imageUrl) use ($url) { |
|
147
|
42 |
|
return $url->formatRelativeToAbsolute($imageUrl); |
|
148
|
45 |
|
}, $this->images()); |
|
149
|
|
|
} |
|
150
|
|
|
} |
|
151
|
|
|
|
jclyons52 /
page-preview
