|
1
|
|
|
<?php |
|
2
|
|
|
|
|
3
|
|
|
namespace Pilipinews\Website\Rappler; |
|
4
|
|
|
|
|
5
|
|
|
use Pilipinews\Common\Article; |
|
6
|
|
|
use Pilipinews\Common\Interfaces\ScraperInterface; |
|
7
|
|
|
use Pilipinews\Common\Scraper as AbstractScraper; |
|
8
|
|
|
use Pilipinews\Common\Crawler as DomCrawler; |
|
9
|
|
|
|
|
10
|
|
|
/** |
|
11
|
|
|
* Rappler News Scraper |
|
12
|
|
|
* |
|
13
|
|
|
* @package Pilipinews |
|
14
|
|
|
* @author Rougin Gutib <[email protected]> |
|
15
|
|
|
*/ |
|
16
|
|
|
class Scraper extends AbstractScraper implements ScraperInterface |
|
17
|
|
|
{ |
|
18
|
|
|
/** |
|
19
|
|
|
* @var string[] |
|
20
|
|
|
*/ |
|
21
|
|
|
protected $removables = array('.author-box'); |
|
22
|
|
|
|
|
23
|
|
|
/** |
|
24
|
|
|
* @var string[] |
|
25
|
|
|
*/ |
|
26
|
|
|
protected $texts = array( |
|
27
|
|
|
"What's the weather like in your area? Report the situation through Rappler's Agos (http://agos.rappler.com/) or tweet us at @rapplerdotcom (https://twitter.com/rapplerdotcom).", |
|
28
|
|
|
"Not on the list? Help us crowdsource class suspensions by posting in the comments section or tweeting @rapplerdotcom (https://twitter.com/rapplerdotcom).\n\nFor more information: (https://www.facebook.com/gov.abet/posts/10152811185356858)When are classes cancelled or suspended? (https://www.rappler.com/move-ph/31299-classes-cancelled-suspended)", |
|
29
|
|
|
"\n\nPlease refresh this page for updates." |
|
30
|
|
|
); |
|
31
|
|
|
|
|
32
|
|
|
/** |
|
33
|
|
|
* Returns the contents of an article. |
|
34
|
|
|
* |
|
35
|
|
|
* @param string $link |
|
36
|
|
|
* @return \Pilipinews\Common\Article |
|
37
|
|
|
*/ |
|
38
|
30 |
|
public function scrape($link) |
|
39
|
|
|
{ |
|
40
|
30 |
|
$this->prepare(mb_strtolower($link)); |
|
41
|
|
|
|
|
42
|
30 |
|
$title = $this->title('h1'); |
|
43
|
|
|
|
|
44
|
30 |
|
$this->remove((array) $this->removables); |
|
45
|
|
|
|
|
46
|
30 |
|
$body = $this->body('.ArticleWrapper__ArticleBodyWrapper-sc-36pn73-0'); |
|
47
|
|
|
|
|
48
|
30 |
|
$body = $this->image($body); |
|
49
|
|
|
|
|
50
|
30 |
|
$body = $this->scribd($body); |
|
51
|
|
|
|
|
52
|
30 |
|
$body = $this->video($body); |
|
53
|
|
|
|
|
54
|
30 |
|
$body = $this->tweet($body); |
|
55
|
|
|
|
|
56
|
30 |
|
$html = $this->html($body, $this->texts); |
|
57
|
|
|
|
|
58
|
30 |
|
$html = htmlspecialchars_decode($html); |
|
59
|
|
|
|
|
60
|
30 |
|
return new Article($title, $html, $link); |
|
61
|
|
|
} |
|
62
|
|
|
|
|
63
|
|
|
/** |
|
64
|
|
|
* Converts image elements to readable string. |
|
65
|
|
|
* |
|
66
|
|
|
* @param \Pilipinews\Common\Crawler $crawler |
|
67
|
|
|
* @return \Pilipinews\Common\Crawler |
|
68
|
|
|
*/ |
|
69
|
|
|
protected function image(DomCrawler $crawler) |
|
70
|
|
|
{ |
|
71
|
30 |
|
$callback = function (DomCrawler $crawler, $html) |
|
|
|
|
|
|
72
|
|
|
{ |
|
73
|
|
|
$image = $crawler->previousAll()->first(); |
|
74
|
|
|
|
|
75
|
|
|
$photo = $image->filter('img')->attr('data-original'); |
|
76
|
|
|
|
|
77
|
|
|
$node = $image->getNode((integer) 0); |
|
78
|
|
|
|
|
79
|
|
|
$node->parentNode->removeChild($node); |
|
80
|
|
|
|
|
81
|
|
|
if ($text = trim($crawler->first()->text())) |
|
82
|
|
|
{ |
|
83
|
|
|
$text = ' - ' . $text; |
|
84
|
|
|
} |
|
85
|
|
|
|
|
86
|
|
|
return '<p>PHOTO: ' . $photo . $text . '</p>'; |
|
87
|
30 |
|
}; |
|
88
|
|
|
|
|
89
|
30 |
|
return $this->replace($crawler, 'p.caption', $callback); |
|
90
|
|
|
} |
|
91
|
|
|
|
|
92
|
|
|
/** |
|
93
|
|
|
* Converts embedded Scribd elements to readable string. |
|
94
|
|
|
* |
|
95
|
|
|
* @param \Pilipinews\Common\Crawler $crawler |
|
96
|
|
|
* @return \Pilipinews\Common\Crawler |
|
97
|
|
|
*/ |
|
98
|
|
|
protected function scribd(DomCrawler $crawler) |
|
99
|
|
|
{ |
|
100
|
30 |
|
$callback = function (DomCrawler $crawler, $html) |
|
|
|
|
|
|
101
|
|
|
{ |
|
102
|
|
|
$title = (string) $crawler->attr('title'); |
|
103
|
|
|
|
|
104
|
|
|
$link = (string) $crawler->attr('src'); |
|
105
|
|
|
|
|
106
|
|
|
return '<p>' . $title . ' (' . $link . ')</p>'; |
|
107
|
30 |
|
}; |
|
108
|
|
|
|
|
109
|
30 |
|
$class = (string) '.scribd_iframe_embed'; |
|
110
|
|
|
|
|
111
|
30 |
|
return $this->replace($crawler, $class, $callback); |
|
112
|
|
|
} |
|
113
|
|
|
|
|
114
|
|
|
/** |
|
115
|
|
|
* Converts embedded iframe elements to readable string. |
|
116
|
|
|
* |
|
117
|
|
|
* @param \Pilipinews\Common\Crawler $crawler |
|
118
|
|
|
* @return \Pilipinews\Common\Crawler |
|
119
|
|
|
*/ |
|
120
|
|
|
protected function video(DomCrawler $crawler) |
|
121
|
|
|
{ |
|
122
|
30 |
|
$callback = function (DomCrawler $crawler, $html) |
|
|
|
|
|
|
123
|
|
|
{ |
|
124
|
|
|
return '<p>VIDEO: ' . $crawler->attr('src') . '</p>'; |
|
125
|
30 |
|
}; |
|
126
|
|
|
|
|
127
|
|
|
return $this->replace($crawler, 'iframe', $callback); |
|
128
|
|
|
} |
|
129
|
|
|
} |
|
130
|
|
|
|
pilipinews /
rappler
Loading history...
This check looks for parameters that have been defined for a function or method, but which are not used in the method body.