|
1
|
|
|
<?php |
|
2
|
|
|
|
|
3
|
|
|
declare(strict_types=1); |
|
4
|
|
|
|
|
5
|
|
|
/** |
|
6
|
|
|
* @author Ne-Lexa |
|
7
|
|
|
* @license MIT |
|
8
|
|
|
* |
|
9
|
|
|
* @see https://github.com/Ne-Lexa/google-play-scraper |
|
10
|
|
|
*/ |
|
11
|
|
|
|
|
12
|
|
|
namespace Nelexa\GPlay\Util; |
|
13
|
|
|
|
|
14
|
|
|
/** |
|
15
|
|
|
* @internal |
|
16
|
|
|
*/ |
|
17
|
|
|
class ScraperUtil |
|
18
|
|
|
{ |
|
19
|
|
|
/** |
|
20
|
|
|
* @param string $html |
|
21
|
|
|
* |
|
22
|
|
|
* @return array |
|
23
|
|
|
*/ |
|
24
|
62 |
|
public static function extractScriptData(string $html): array |
|
25
|
|
|
{ |
|
26
|
62 |
|
$scripts = []; |
|
27
|
|
|
|
|
28
|
62 |
|
if (preg_match_all('/>AF_initDataCallback[\s\S]*?<\/script/', $html, $matches)) { |
|
29
|
62 |
|
$scripts = array_reduce( |
|
30
|
62 |
|
$matches[0], |
|
31
|
|
|
static function ($carry, $item) { |
|
32
|
|
|
if ( |
|
33
|
62 |
|
preg_match("/(ds:.*?)'/", $item, $keyMatch) && |
|
34
|
62 |
|
preg_match('/return ([\s\S]*?)}}\);<\//', $item, $valueMatch) |
|
35
|
|
|
) { |
|
36
|
62 |
|
$carry[$keyMatch[1]] = \GuzzleHttp\json_decode($valueMatch[1], true); |
|
37
|
|
|
} |
|
38
|
|
|
|
|
39
|
62 |
|
return $carry; |
|
40
|
62 |
|
}, |
|
41
|
62 |
|
$scripts |
|
42
|
|
|
); |
|
43
|
|
|
} |
|
44
|
|
|
|
|
45
|
62 |
|
return $scripts; |
|
46
|
|
|
} |
|
47
|
|
|
|
|
48
|
|
|
/** |
|
49
|
|
|
* @param string $html |
|
50
|
|
|
* |
|
51
|
|
|
* @return \DOMDocument |
|
52
|
|
|
*/ |
|
53
|
47 |
|
public static function createDomDocument(string $html): \DOMDocument |
|
54
|
|
|
{ |
|
55
|
47 |
|
$doc = new \DOMDocument(); |
|
56
|
47 |
|
$internalErrors = libxml_use_internal_errors(true); |
|
57
|
|
|
|
|
58
|
47 |
|
if (!$doc->loadHTML('<?xml encoding="utf-8"?>' . $html)) { |
|
59
|
|
|
throw new |
|
60
|
|
|
\RuntimeException( |
|
61
|
|
|
'error load html: ' . $html |
|
62
|
|
|
); |
|
63
|
|
|
} |
|
64
|
47 |
|
libxml_use_internal_errors($internalErrors); |
|
65
|
|
|
|
|
66
|
47 |
|
return $doc; |
|
67
|
|
|
} |
|
68
|
|
|
|
|
69
|
|
|
/** |
|
70
|
|
|
* @param string $html |
|
71
|
|
|
* |
|
72
|
|
|
* @return string |
|
73
|
|
|
*/ |
|
74
|
47 |
|
public static function html2text(string $html): string |
|
75
|
|
|
{ |
|
76
|
47 |
|
$doc = self::createDomDocument($html); |
|
77
|
47 |
|
$text = self::convertDomNodeToText($doc); |
|
78
|
47 |
|
$text = preg_replace('/\n{3,}/', "\n\n", trim($text)); |
|
79
|
|
|
|
|
80
|
47 |
|
return trim($text); |
|
81
|
|
|
} |
|
82
|
|
|
|
|
83
|
|
|
/** |
|
84
|
|
|
* @param \DOMNode $node |
|
85
|
|
|
* |
|
86
|
|
|
* @return string |
|
87
|
|
|
*/ |
|
88
|
47 |
|
private static function convertDomNodeToText(\DOMNode $node): string |
|
89
|
|
|
{ |
|
90
|
47 |
|
if ($node instanceof \DOMText) { |
|
91
|
47 |
|
$text = preg_replace('/\s+/', ' ', $node->wholeText); |
|
92
|
|
|
} else { |
|
93
|
47 |
|
$text = ''; |
|
94
|
|
|
|
|
95
|
47 |
|
if ($node->childNodes !== null) { |
|
96
|
47 |
|
foreach ($node->childNodes as $childNode) { |
|
97
|
47 |
|
$text .= self::convertDomNodeToText($childNode); |
|
98
|
|
|
} |
|
99
|
|
|
} |
|
100
|
|
|
|
|
101
|
47 |
|
switch ($node->nodeName) { |
|
102
|
47 |
|
case 'h1': |
|
103
|
47 |
|
case 'h2': |
|
104
|
47 |
|
case 'h3': |
|
105
|
47 |
|
case 'h4': |
|
106
|
47 |
|
case 'h5': |
|
107
|
47 |
|
case 'h6': |
|
108
|
47 |
|
case 'p': |
|
109
|
47 |
|
case 'ul': |
|
110
|
47 |
|
case 'div': |
|
111
|
47 |
|
$text = "\n\n" . $text . "\n\n"; |
|
112
|
47 |
|
break; |
|
113
|
|
|
|
|
114
|
47 |
|
case 'li': |
|
115
|
|
|
$text = '- ' . $text . "\n"; |
|
116
|
|
|
break; |
|
117
|
|
|
|
|
118
|
47 |
|
case 'br': |
|
119
|
45 |
|
$text .= "\n"; |
|
120
|
45 |
|
break; |
|
121
|
|
|
} |
|
122
|
|
|
} |
|
123
|
|
|
|
|
124
|
47 |
|
return $text; |
|
125
|
|
|
} |
|
126
|
|
|
} |
|
127
|
|
|
|
Ne-Lexa /
google-play-scraper
