|
1
|
|
|
<?php |
|
2
|
|
|
|
|
3
|
|
|
namespace Mpclarkson\IconScraper; |
|
4
|
|
|
|
|
5
|
|
|
class Scraper |
|
6
|
|
|
{ |
|
7
|
|
|
protected $url = ''; |
|
8
|
|
|
protected $dataAccess; |
|
9
|
|
|
|
|
10
|
|
|
public function __construct($args = array()) |
|
11
|
|
|
{ |
|
12
|
|
|
if (isset($args['url'])) { |
|
13
|
|
|
$this->url = $args['url']; |
|
14
|
|
|
} |
|
15
|
|
|
|
|
16
|
|
|
$this->dataAccess = new DataAccess(); |
|
17
|
|
|
} |
|
18
|
|
|
|
|
19
|
|
|
/** |
|
20
|
|
|
* @param string $url |
|
21
|
|
|
*/ |
|
22
|
|
|
public static function baseUrl($url, $path = false) |
|
23
|
|
|
{ |
|
24
|
|
|
$return = ''; |
|
25
|
|
|
|
|
26
|
|
|
if (!$url = parse_url($url)) { |
|
27
|
|
|
return false; |
|
28
|
|
|
} |
|
29
|
|
|
|
|
30
|
|
|
// Scheme |
|
31
|
|
|
$scheme = isset($url['scheme']) ? strtolower($url['scheme']) : null; |
|
32
|
|
|
if ($scheme != 'http' && $scheme != 'https') { |
|
33
|
|
|
return false; |
|
34
|
|
|
} |
|
35
|
|
|
$return .= "{$scheme}://"; |
|
36
|
|
|
|
|
37
|
|
|
// Username and password |
|
38
|
|
|
if (isset($url['user'])) { |
|
39
|
|
|
$return .= $url['user']; |
|
40
|
|
|
if (isset($url['pass'])) { |
|
41
|
|
|
$return .= ":{$url['pass']}"; |
|
42
|
|
|
} |
|
43
|
|
|
$return .= '@'; |
|
44
|
|
|
} |
|
45
|
|
|
|
|
46
|
|
|
// Hostname |
|
47
|
|
|
if (!isset($url['host'])) { |
|
48
|
|
|
return false; |
|
49
|
|
|
} |
|
50
|
|
|
|
|
51
|
|
|
$return .= $url['host']; |
|
52
|
|
|
|
|
53
|
|
|
// Port |
|
54
|
|
|
if (isset($url['port'])) { |
|
55
|
|
|
$return .= ":{$url['port']}"; |
|
56
|
|
|
} |
|
57
|
|
|
|
|
58
|
|
|
// Path |
|
59
|
|
|
if ($path && isset($url['path'])) { |
|
60
|
|
|
$return .= $url['path']; |
|
61
|
|
|
} |
|
62
|
|
|
$return .= '/'; |
|
63
|
|
|
|
|
64
|
|
|
return $return; |
|
65
|
|
|
} |
|
66
|
|
|
|
|
67
|
|
|
public function info($url) { |
|
68
|
|
|
if (empty($url) || $url === false) { |
|
69
|
|
|
return false; |
|
70
|
|
|
} |
|
71
|
|
|
|
|
72
|
|
|
$headers = $this->dataAccess->retrieveHeader($url); |
|
73
|
|
|
|
|
74
|
|
|
// leaves only numeric keys |
|
75
|
|
|
$status_lines = array_filter($headers, function ($key) { |
|
76
|
|
|
return is_int($key); |
|
77
|
|
|
}, ARRAY_FILTER_USE_KEY); |
|
78
|
|
|
|
|
79
|
|
|
// uses last returned status line header |
|
80
|
|
|
$exploded = explode(' ', end($status_lines)); |
|
81
|
|
|
|
|
82
|
|
|
if (! array_key_exists(1, $exploded)) { |
|
83
|
|
|
return false; |
|
84
|
|
|
} |
|
85
|
|
|
|
|
86
|
|
|
list(, $status) = $exploded; |
|
87
|
|
|
|
|
88
|
|
|
if (isset($headers['location'])) { |
|
89
|
|
|
$url = $headers['location']; |
|
90
|
|
|
} |
|
91
|
|
|
|
|
92
|
|
|
return ['status' => $status, 'url' => $url]; |
|
93
|
|
|
} |
|
94
|
|
|
|
|
95
|
|
|
/** |
|
96
|
|
|
* @param false|string $url |
|
97
|
|
|
* |
|
98
|
|
|
* @return string |
|
99
|
|
|
*/ |
|
100
|
|
|
public function endRedirect($url) { |
|
101
|
|
|
$out = $this->info($url); |
|
102
|
|
|
return !empty($out['url']) ? $out['url'] : false; |
|
103
|
|
|
} |
|
104
|
|
|
|
|
105
|
|
|
/** |
|
106
|
|
|
* @return array of icons or empty array |
|
107
|
|
|
**/ |
|
108
|
|
|
public function get($url = '') |
|
109
|
|
|
{ |
|
110
|
|
|
|
|
111
|
|
|
// URLs passed to this method take precedence. |
|
112
|
|
|
if (!empty($url)) { |
|
113
|
|
|
$this->url = $url; |
|
114
|
|
|
} |
|
115
|
|
|
|
|
116
|
|
|
$url = rtrim($this->endRedirect($this->baseUrl($this->url, false)), '/'); |
|
117
|
|
|
|
|
118
|
|
|
return $this->getIcons($url); |
|
119
|
|
|
} |
|
120
|
|
|
|
|
121
|
|
|
/** |
|
122
|
|
|
* @param string $url |
|
123
|
|
|
*/ |
|
124
|
|
|
private function getIcons($url) { |
|
125
|
|
|
|
|
126
|
|
|
if (empty($url)) { |
|
127
|
|
|
return []; |
|
128
|
|
|
} |
|
129
|
|
|
|
|
130
|
|
|
$html = $this->dataAccess->retrieveUrl("{$url}/"); |
|
131
|
|
|
preg_match('!<head.*?>.*</head>!ims', $html, $match); |
|
132
|
|
|
|
|
133
|
|
|
if (empty($match) || count($match) == 0) { |
|
134
|
|
|
return []; |
|
135
|
|
|
} |
|
136
|
|
|
|
|
137
|
|
|
$head = $match[0]; |
|
138
|
|
|
|
|
139
|
|
|
$icons = []; |
|
140
|
|
|
|
|
141
|
|
|
$dom = new \DOMDocument(); |
|
142
|
|
|
|
|
143
|
|
|
// Use error supression, because the HTML might be too malformed. |
|
144
|
|
|
if (@$dom->loadHTML($head)) { |
|
145
|
|
|
$links = $dom->getElementsByTagName('link'); |
|
146
|
|
|
|
|
147
|
|
|
foreach ($links as $link) { |
|
148
|
|
|
|
|
149
|
|
|
if ($link->hasAttribute('rel') && $href = $link->getAttribute('href')) { |
|
150
|
|
|
|
|
151
|
|
|
$attribute = $link->getAttribute('rel'); |
|
152
|
|
|
|
|
153
|
|
|
// Make sure the href is an absolute URL. |
|
154
|
|
|
if ($href && filter_var($href, FILTER_VALIDATE_URL) === false) { |
|
155
|
|
|
$href = $url . '/' . $href; //Todo: Improve this |
|
156
|
|
|
} |
|
157
|
|
|
|
|
158
|
|
|
$size = $link->hasAttribute('sizes') ? $link->getAttribute('sizes') : []; |
|
159
|
|
|
$size = !is_array($size) ? explode('x', $size) : $size; |
|
160
|
|
|
|
|
161
|
|
|
$type = false; |
|
162
|
|
|
|
|
163
|
|
|
switch(strtolower($attribute)) { |
|
164
|
|
|
case Icon::APPLE_TOUCH: |
|
165
|
|
|
$type = Icon::APPLE_TOUCH; |
|
166
|
|
|
break; |
|
167
|
|
|
default: |
|
168
|
|
|
if(strpos(strtolower($attribute), 'icon') !== FALSE) { |
|
169
|
|
|
$type = Icon::FAVICON; |
|
170
|
|
|
$size = []; |
|
171
|
|
|
} |
|
172
|
|
|
}; |
|
173
|
|
|
|
|
174
|
|
|
if(!empty($type) && filter_var($href, FILTER_VALIDATE_URL)) { |
|
175
|
|
|
$icons[] = new Icon($type, $href, $size); |
|
176
|
|
|
} |
|
177
|
|
|
} |
|
178
|
|
|
} |
|
179
|
|
|
} |
|
180
|
|
|
|
|
181
|
|
|
//Sort the icons by width |
|
182
|
|
|
usort($icons, function($a, $b) { |
|
183
|
|
|
return $a->getWidth() - $b->getWidth(); |
|
184
|
|
|
}); |
|
185
|
|
|
|
|
186
|
|
|
//If it is empty, try and get one from the root |
|
187
|
|
|
if (empty($icons)) { |
|
188
|
|
|
$icons = $this->getFavicon($url); |
|
189
|
|
|
} |
|
190
|
|
|
|
|
191
|
|
|
return $icons; |
|
192
|
|
|
} |
|
193
|
|
|
|
|
194
|
|
|
private function getFavicon($url) { |
|
195
|
|
|
|
|
196
|
|
|
// Try /favicon.ico first. |
|
197
|
|
|
$info = $this->info("{$url}/favicon.ico"); |
|
198
|
|
|
if ($info['status'] == '200') { |
|
199
|
|
|
$favicon = $info['url']; |
|
200
|
|
|
} |
|
201
|
|
|
|
|
202
|
|
|
// Make sure the favicon is an absolute URL. |
|
203
|
|
|
if (isset($favicon) && filter_var($favicon, FILTER_VALIDATE_URL) === false) { |
|
204
|
|
|
$favicon = $url . '/' . $favicon; |
|
205
|
|
|
} |
|
206
|
|
|
|
|
207
|
|
|
if (isset($favicon)) { |
|
208
|
|
|
return [ |
|
209
|
|
|
new Icon(Icon::FAVICON, $favicon, []) |
|
210
|
|
|
]; |
|
211
|
|
|
} |
|
212
|
|
|
|
|
213
|
|
|
return []; |
|
214
|
|
|
} |
|
215
|
|
|
|
|
216
|
|
|
/** |
|
217
|
|
|
* @return string |
|
218
|
|
|
*/ |
|
219
|
|
|
public function getUrl() |
|
220
|
|
|
{ |
|
221
|
|
|
return $this->url; |
|
222
|
|
|
} |
|
223
|
|
|
|
|
224
|
|
|
/** |
|
225
|
|
|
* @param string $url |
|
226
|
|
|
*/ |
|
227
|
|
|
public function setUrl($url) |
|
228
|
|
|
{ |
|
229
|
|
|
$this->url = $url; |
|
230
|
|
|
} |
|
231
|
|
|
|
|
232
|
|
|
/** |
|
233
|
|
|
* @param DataAccess $dataAccess |
|
234
|
|
|
*/ |
|
235
|
|
|
public function setDataAccess($dataAccess) |
|
236
|
|
|
{ |
|
237
|
|
|
$this->dataAccess = $dataAccess; |
|
238
|
|
|
} |
|
239
|
|
|
} |
|
240
|
|
|
|
mpclarkson /
icon-scraper
