|
1
|
|
|
<?php |
|
2
|
|
|
|
|
3
|
|
|
namespace Pilipinews\Common; |
|
4
|
|
|
|
|
5
|
|
|
use Symfony\Component\DomCrawler\Crawler; |
|
6
|
|
|
|
|
7
|
|
|
/** |
|
8
|
|
|
* Abstract Scraper |
|
9
|
|
|
* |
|
10
|
|
|
* @package Pilipinews |
|
11
|
|
|
* @author Rougin Royce Gutib <[email protected]> |
|
12
|
|
|
*/ |
|
13
|
|
|
abstract class Scraper |
|
14
|
|
|
{ |
|
15
|
|
|
/** |
|
16
|
|
|
* @var \Symfony\Component\DomCrawler\Crawler |
|
17
|
|
|
*/ |
|
18
|
|
|
protected $crawler; |
|
19
|
|
|
|
|
20
|
|
|
/** |
|
21
|
|
|
* Returns the article content based on a given element. |
|
22
|
|
|
* |
|
23
|
|
|
* @param string $element |
|
24
|
|
|
* @return \Symfony\Component\DomCrawler\Crawler |
|
25
|
|
|
*/ |
|
26
|
9 |
|
protected function body($element) |
|
27
|
|
|
{ |
|
28
|
9 |
|
$body = $this->crawler->filter($element)->first(); |
|
29
|
|
|
|
|
30
|
9 |
|
$charset = 'ISO-8859-1//TRANSLIT//IGNORE'; |
|
31
|
|
|
|
|
32
|
9 |
|
$body = iconv('UTF-8', $charset, $body->html()); |
|
33
|
|
|
|
|
34
|
9 |
|
$body = trim(preg_replace('/\s+/', ' ', $body)); |
|
35
|
|
|
|
|
36
|
9 |
|
return new Crawler(str_replace(' ', ' ', $body)); |
|
37
|
|
|
} |
|
38
|
|
|
|
|
39
|
|
|
/** |
|
40
|
|
|
* Returns the HTML format of the body from the crawler. |
|
41
|
|
|
* |
|
42
|
|
|
* @param \Symfony\Component\DomCrawler\Crawler $crawler |
|
43
|
|
|
* @param string[] $removables |
|
44
|
|
|
* @return string |
|
45
|
|
|
*/ |
|
46
|
9 |
|
protected function html(Crawler $crawler, $removables = array()) |
|
47
|
2 |
|
{ |
|
48
|
9 |
|
$converter = new Converter; |
|
49
|
|
|
|
|
50
|
9 |
|
$html = $converter->convert($crawler->html()); |
|
51
|
|
|
|
|
52
|
9 |
|
foreach ((array) $removables as $keyword) { |
|
53
|
6 |
|
$html = str_replace($keyword, '', $html); |
|
54
|
|
|
|
|
55
|
6 |
|
$html = str_replace("\n\n\n", '', $html); |
|
56
|
6 |
|
} |
|
57
|
|
|
|
|
58
|
9 |
|
return trim(preg_replace('/\s\s+/', "\n\n", $html)); |
|
59
|
2 |
|
} |
|
60
|
|
|
|
|
61
|
|
|
/** |
|
62
|
|
|
* Initializes the crawler instance. |
|
63
|
|
|
* |
|
64
|
|
|
* @param string $link |
|
65
|
|
|
* @return void |
|
66
|
|
|
*/ |
|
67
|
3 |
|
protected function prepare($link) |
|
68
|
|
|
{ |
|
69
|
3 |
|
$response = Client::request((string) $link); |
|
70
|
|
|
|
|
71
|
3 |
|
$this->crawler = new Crawler($response); |
|
72
|
3 |
|
} |
|
73
|
|
|
|
|
74
|
|
|
/** |
|
75
|
|
|
* Removes specified HTML tags from body. |
|
76
|
|
|
* |
|
77
|
|
|
* @param string[] $elements |
|
78
|
|
|
* @return void |
|
79
|
|
|
*/ |
|
80
|
6 |
|
protected function remove($elements) |
|
81
|
|
|
{ |
|
82
|
3 |
|
$callback = function ($crawler) { |
|
83
|
3 |
|
$node = $crawler->getNode((integer) 0); |
|
84
|
|
|
|
|
85
|
3 |
|
$node->parentNode->removeChild($node); |
|
86
|
9 |
|
}; |
|
87
|
|
|
|
|
88
|
9 |
|
foreach ((array) $elements as $removable) { |
|
89
|
9 |
|
$this->crawler->filter($removable)->each($callback); |
|
90
|
6 |
|
} |
|
91
|
9 |
|
} |
|
92
|
|
|
|
|
93
|
|
|
/** |
|
94
|
|
|
* Replaces a specified HTML tag based from the given callback. |
|
95
|
|
|
* |
|
96
|
|
|
* @param \Symfony\Component\DomCrawler\Crawler $crawler |
|
97
|
|
|
* @param string $element |
|
98
|
|
|
* @param callable $callback |
|
99
|
|
|
* @return \Symfony\Component\DomCrawler\Crawler |
|
100
|
|
|
*/ |
|
101
|
4 |
|
protected function replace(Crawler $crawler, $element, $callback) |
|
102
|
|
|
{ |
|
103
|
2 |
|
$function = function (Crawler $crawler) use ($callback) { |
|
104
|
6 |
|
$node = $crawler->getNode(0); |
|
105
|
|
|
|
|
106
|
6 |
|
$html = $node->ownerDocument->saveHtml($node); |
|
107
|
|
|
|
|
108
|
6 |
|
$text = $callback($crawler, (string) $html); |
|
109
|
|
|
|
|
110
|
6 |
|
return array((string) $html, (string) $text); |
|
111
|
6 |
|
}; |
|
112
|
|
|
|
|
113
|
6 |
|
$items = $crawler->filter($element)->each($function); |
|
114
|
|
|
|
|
115
|
6 |
|
$html = (string) $crawler->html(); |
|
116
|
|
|
|
|
117
|
6 |
|
foreach ((array) $items as $item) { |
|
118
|
6 |
|
$html = str_replace($item[0], $item[1], $html); |
|
119
|
4 |
|
} |
|
120
|
|
|
|
|
121
|
6 |
|
return new Crawler((string) $html); |
|
122
|
|
|
} |
|
123
|
|
|
|
|
124
|
|
|
/** |
|
125
|
|
|
* Returns the title text based from given HTML tag. |
|
126
|
|
|
* |
|
127
|
|
|
* @param string $element |
|
128
|
|
|
* @param string $removable |
|
129
|
|
|
* @return string |
|
130
|
|
|
*/ |
|
131
|
9 |
|
protected function title($element, $removable = '') |
|
132
|
|
|
{ |
|
133
|
9 |
|
$converter = new Converter; |
|
134
|
|
|
|
|
135
|
9 |
|
$crawler = $this->crawler->filter($element); |
|
136
|
|
|
|
|
137
|
9 |
|
$html = $crawler->first()->html(); |
|
138
|
|
|
|
|
139
|
9 |
|
$html = str_replace($removable, '', $html); |
|
140
|
|
|
|
|
141
|
9 |
|
return $converter->convert((string) $html); |
|
142
|
|
|
} |
|
143
|
|
|
|
|
144
|
|
|
/** |
|
145
|
|
|
* Parses embedded Twitter tweet in the HTML. |
|
146
|
|
|
* |
|
147
|
|
|
* @param \Symfony\Component\DomCrawler\Crawler $crawler |
|
148
|
|
|
* @return \Symfony\Component\DomCrawler\Crawler |
|
149
|
|
|
*/ |
|
150
|
|
|
protected function tweet(Crawler $crawler) |
|
151
|
|
|
{ |
|
152
|
6 |
|
$callback = function (Crawler $crawler) { |
|
153
|
6 |
|
$parsed = (string) $crawler->text(); |
|
154
|
|
|
|
|
155
|
6 |
|
$text = str_replace('📸: ', '', $parsed); |
|
156
|
|
|
|
|
157
|
6 |
|
return '<p>TWEET: ' . $text . '</p>'; |
|
158
|
6 |
|
}; |
|
159
|
|
|
|
|
160
|
4 |
|
$class = '.twitter-tweet'; |
|
161
|
|
|
|
|
162
|
4 |
|
return $this->replace($crawler, $class, $callback); |
|
163
|
|
|
} |
|
164
|
|
|
} |
|
165
|
|
|
|
pilipinews /
common
