|
1
|
|
|
<?php |
|
2
|
|
|
|
|
3
|
|
|
namespace Pilipinews\Website\Sunstar; |
|
4
|
|
|
|
|
5
|
|
|
use Pilipinews\Common\Article; |
|
6
|
|
|
use Pilipinews\Common\Crawler as DomCrawler; |
|
7
|
|
|
use Pilipinews\Common\Interfaces\ScraperInterface; |
|
8
|
|
|
use Pilipinews\Common\Scraper as AbstractScraper; |
|
9
|
|
|
|
|
10
|
|
|
/** |
|
11
|
|
|
* Sunstar News Scraper |
|
12
|
|
|
* |
|
13
|
|
|
* @package Pilipinews |
|
14
|
|
|
* @author Rougin Gutib <[email protected]> |
|
15
|
|
|
*/ |
|
16
|
|
|
class Scraper extends AbstractScraper implements ScraperInterface |
|
17
|
|
|
{ |
|
18
|
|
|
/** |
|
19
|
|
|
* @var array |
|
20
|
|
|
*/ |
|
21
|
|
|
protected $elements = array('.article-header', '.articleBody'); |
|
22
|
|
|
|
|
23
|
|
|
/** |
|
24
|
|
|
* @var string[] |
|
25
|
|
|
*/ |
|
26
|
|
|
protected $removables = array('.subSection', '.titleArticle', '.pagingWrap', 'script', '#fb-root'); |
|
27
|
|
|
|
|
28
|
|
|
/** |
|
29
|
|
|
* @var string[] |
|
30
|
|
|
*/ |
|
31
|
|
|
protected $texts = array("PHOTO: https://www.sunstar.com.ph/\n", 'Please refresh page for updates.', 'ARTICLE_MOBILE_AD_CODE'); |
|
32
|
|
|
|
|
33
|
|
|
/** |
|
34
|
|
|
* Returns the contents of an article. |
|
35
|
|
|
* |
|
36
|
|
|
* @param string $link |
|
37
|
|
|
* @return \Pilipinews\Common\Article |
|
38
|
|
|
*/ |
|
39
|
30 |
|
public function scrape($link) |
|
40
|
|
|
{ |
|
41
|
30 |
|
$this->prepare((string) strtolower($link)); |
|
42
|
|
|
|
|
43
|
30 |
|
$title = $this->title('title', ' - SUNSTAR'); |
|
44
|
|
|
|
|
45
|
30 |
|
$this->remove((array) $this->removables); |
|
46
|
|
|
|
|
47
|
30 |
|
$this->crawler = $this->carousel($this->crawler); |
|
48
|
|
|
|
|
49
|
30 |
|
$body = $this->body($this->elements); |
|
50
|
|
|
|
|
51
|
30 |
|
$body = $this->video($this->image($body)); |
|
52
|
|
|
|
|
53
|
30 |
|
$html = $this->html($body, $this->texts); |
|
54
|
|
|
|
|
55
|
30 |
|
return new Article($title, $html, $link); |
|
56
|
|
|
} |
|
57
|
|
|
|
|
58
|
|
|
/** |
|
59
|
|
|
* Returns the article content based on a given element. |
|
60
|
|
|
* |
|
61
|
|
|
* @param string|string[] $element |
|
62
|
|
|
* @return \Pilipinews\Common\Crawler |
|
63
|
|
|
*/ |
|
64
|
30 |
|
protected function body($elements) |
|
65
|
|
|
{ |
|
66
|
30 |
|
is_string($elements) && $elements = array($elements); |
|
67
|
|
|
|
|
68
|
30 |
|
foreach ((array) $elements as $key => $element) |
|
69
|
|
|
{ |
|
70
|
30 |
|
$body = $this->crawler->filter($element)->last()->html(); |
|
71
|
|
|
|
|
72
|
30 |
|
$body = (string) trim(preg_replace('/\s+/', ' ', $body)); |
|
73
|
|
|
|
|
74
|
30 |
|
$elements[$key] = str_replace(' ', ' ', (string) $body); |
|
75
|
10 |
|
} |
|
76
|
|
|
|
|
77
|
30 |
|
return new DomCrawler(implode('<br><br><br>', $elements)); |
|
78
|
|
|
} |
|
79
|
|
|
|
|
80
|
|
|
/** |
|
81
|
|
|
* Converts carousel elements to readable string. |
|
82
|
|
|
* |
|
83
|
|
|
* @param \Pilipinews\Common\Crawler $crawler |
|
84
|
|
|
* @return \Pilipinews\Common\Crawler |
|
85
|
|
|
*/ |
|
86
|
10 |
|
protected function carousel(DomCrawler $crawler) |
|
87
|
|
|
{ |
|
88
|
20 |
|
$callback = function (DomCrawler $crawler) |
|
89
|
|
|
{ |
|
90
|
3 |
|
$texts = $crawler->filter('.img-caption'); |
|
91
|
|
|
|
|
92
|
2 |
|
$function = function ($result, $index) use ($texts) |
|
93
|
|
|
{ |
|
94
|
3 |
|
$text = $texts->eq($index)->text(); |
|
95
|
|
|
|
|
96
|
3 |
|
$image = $result->attr('src') . ' - ' . $text; |
|
97
|
|
|
|
|
98
|
3 |
|
return '<p>PHOTO: ' . $image . '</p>'; |
|
99
|
3 |
|
}; |
|
100
|
|
|
|
|
101
|
3 |
|
$items = $crawler->filter('img'); |
|
102
|
|
|
|
|
103
|
3 |
|
$image = $items->each($function); |
|
104
|
|
|
|
|
105
|
3 |
|
return implode("<br><br>", $image); |
|
106
|
30 |
|
}; |
|
107
|
|
|
|
|
108
|
30 |
|
return $this->replace($crawler, '.owl-carousel', $callback); |
|
109
|
|
|
} |
|
110
|
|
|
|
|
111
|
|
|
/** |
|
112
|
|
|
* Converts image elements to readable string. |
|
113
|
|
|
* |
|
114
|
|
|
* @param \Pilipinews\Common\Crawler $crawler |
|
115
|
|
|
* @return \Pilipinews\Common\Crawler |
|
116
|
|
|
*/ |
|
117
|
10 |
|
protected function image(DomCrawler $crawler) |
|
118
|
|
|
{ |
|
119
|
20 |
|
$callback = function (DomCrawler $crawler) |
|
120
|
|
|
{ |
|
121
|
27 |
|
$break = (string) '<br><br><br>'; |
|
122
|
|
|
|
|
123
|
27 |
|
$result = $crawler->filter('img')->first(); |
|
124
|
|
|
|
|
125
|
27 |
|
$image = $result->attr('src') . $break; |
|
126
|
|
|
|
|
127
|
27 |
|
return (string) $break . 'PHOTO: ' . $image; |
|
128
|
30 |
|
}; |
|
129
|
|
|
|
|
130
|
30 |
|
return $this->replace($crawler, '.imgArticle', $callback); |
|
131
|
|
|
} |
|
132
|
|
|
|
|
133
|
|
|
/** |
|
134
|
|
|
* Initializes the crawler instance. |
|
135
|
|
|
* |
|
136
|
|
|
* @param string $link |
|
137
|
|
|
* @return void |
|
138
|
|
|
*/ |
|
139
|
30 |
|
protected function prepare($link) |
|
140
|
|
|
{ |
|
141
|
30 |
|
$this->crawler = new DomCrawler(Client::request($link)); |
|
142
|
30 |
|
} |
|
143
|
|
|
|
|
144
|
|
|
/** |
|
145
|
|
|
* Converts video elements to readable string. |
|
146
|
|
|
* |
|
147
|
|
|
* @param \Pilipinews\Common\Crawler $crawler |
|
148
|
|
|
* @return \Pilipinews\Common\Crawler |
|
149
|
|
|
*/ |
|
150
|
|
|
protected function video(DomCrawler $crawler) |
|
151
|
|
|
{ |
|
152
|
30 |
|
$callback = function (DomCrawler $crawler) |
|
153
|
|
|
{ |
|
154
|
6 |
|
$link = trim($crawler->attr('data-href')); |
|
155
|
|
|
|
|
156
|
6 |
|
$break = '<br><br><br>'; |
|
157
|
|
|
|
|
158
|
6 |
|
return $break . 'VIDEO: ' . $link . $break; |
|
159
|
30 |
|
}; |
|
160
|
|
|
|
|
161
|
10 |
|
return $this->replace($crawler, '.fb-video', $callback); |
|
162
|
|
|
} |
|
163
|
|
|
} |
|
164
|
|
|
|
pilipinews /
sunstar
