1
|
|
|
<?php |
2
|
|
|
/** |
3
|
|
|
* AnimeDb package. |
4
|
|
|
* |
5
|
|
|
* @author Peter Gribanov <[email protected]> |
6
|
|
|
* @copyright Copyright (c) 2011, Peter Gribanov |
7
|
|
|
* @license http://opensource.org/licenses/GPL-3.0 GPL v3 |
8
|
|
|
*/ |
9
|
|
|
|
10
|
|
|
namespace AnimeDb\Bundle\WorldArtFillerBundle\Service; |
11
|
|
|
|
12
|
|
|
use Guzzle\Http\Client; |
13
|
|
|
use Symfony\Component\HttpFoundation\Request; |
14
|
|
|
|
15
|
|
|
class Browser |
16
|
|
|
{ |
17
|
|
|
/** |
18
|
|
|
* Default HTTP User-Agent. |
19
|
|
|
* |
20
|
|
|
* @var string |
21
|
|
|
*/ |
22
|
|
|
const DEFAULT_USER_AGENT = 'Mozilla/5.0 (Windows NT 6.2; WOW64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/29.0.1547.2 Safari/537.36'; |
23
|
|
|
|
24
|
|
|
/** |
25
|
|
|
* Browser. |
26
|
|
|
* |
27
|
|
|
* @var \Guzzle\Http\Client |
28
|
|
|
*/ |
29
|
|
|
protected $browser; |
30
|
|
|
|
31
|
|
|
/** |
32
|
|
|
* Request. |
33
|
|
|
* |
34
|
|
|
* @var \Symfony\Component\HttpFoundation\Request |
35
|
|
|
*/ |
36
|
|
|
protected $request; |
37
|
|
|
|
38
|
|
|
/** |
39
|
|
|
* HTTP host. |
40
|
|
|
* |
41
|
|
|
* @var string |
42
|
|
|
*/ |
43
|
|
|
protected $host; |
44
|
|
|
|
45
|
|
|
/** |
46
|
|
|
* Browser timeout. |
47
|
|
|
* |
48
|
|
|
* @var int |
49
|
|
|
*/ |
50
|
|
|
protected $timeout; |
51
|
|
|
|
52
|
|
|
/** |
53
|
|
|
* Browser proxy list. |
54
|
|
|
* |
55
|
|
|
* @var array |
56
|
|
|
*/ |
57
|
|
|
protected $proxy_list; |
58
|
|
|
|
59
|
|
|
/** |
60
|
|
|
* Construct. |
61
|
|
|
* |
62
|
|
|
* @param string $host |
63
|
|
|
* @param int $timeout |
64
|
|
|
* @param array $proxy_list |
65
|
|
|
*/ |
66
|
|
|
public function __construct($host, $timeout, array $proxy_list) |
67
|
|
|
{ |
68
|
|
|
$this->host = $host; |
69
|
|
|
$this->proxy_list = $proxy_list; |
70
|
|
|
$this->timeout = $timeout; |
71
|
|
|
} |
72
|
|
|
|
73
|
|
|
/** |
74
|
|
|
* Get host. |
75
|
|
|
* |
76
|
|
|
* @return string |
77
|
|
|
*/ |
78
|
|
|
public function getHost() |
79
|
|
|
{ |
80
|
|
|
return $this->host; |
81
|
|
|
} |
82
|
|
|
|
83
|
|
|
/** |
84
|
|
|
* Set request. |
85
|
|
|
* |
86
|
|
|
* @param \Symfony\Component\HttpFoundation\Request $request |
87
|
|
|
*/ |
88
|
|
|
public function setRequest(Request $request = null) |
89
|
|
|
{ |
90
|
|
|
$this->request = $request; |
91
|
|
|
// try to set User-Agent from original request |
92
|
|
|
if ($request && $this->browser) { |
93
|
|
|
$this->browser->setDefaultHeaders([ |
|
|
|
|
94
|
|
|
'User-Agent' => $request->server->get('HTTP_USER_AGENT', self::DEFAULT_USER_AGENT), |
95
|
|
|
]); |
96
|
|
|
} |
97
|
|
|
} |
98
|
|
|
|
99
|
|
|
/** |
100
|
|
|
* Get DOMDocument from path. |
101
|
|
|
* |
102
|
|
|
* Receive content from the URL, cleaning using Tidy and creating DOM document |
103
|
|
|
* |
104
|
|
|
* @param string $path |
105
|
|
|
* |
106
|
|
|
* @return \DOMDocument|null |
107
|
|
|
*/ |
108
|
|
|
public function getDom($path) |
109
|
|
|
{ |
110
|
|
|
$dom = new \DOMDocument('1.0', 'utf8'); |
111
|
|
|
if (($content = $this->getContent($path)) && $dom->loadHTML($content)) { |
112
|
|
|
return $dom; |
113
|
|
|
} else { |
114
|
|
|
return; |
115
|
|
|
} |
116
|
|
|
} |
117
|
|
|
|
118
|
|
|
/** |
119
|
|
|
* Get content from path. |
120
|
|
|
* |
121
|
|
|
* Receive content from the URL and cleaning using Tidy |
122
|
|
|
* |
123
|
|
|
* @param string $path |
124
|
|
|
* |
125
|
|
|
* @return string |
126
|
|
|
*/ |
127
|
|
|
public function getContent($path) |
128
|
|
|
{ |
129
|
|
|
/* @var $response \Guzzle\Http\Message\Response */ |
130
|
|
|
$response = $this->getBrowser()->get($path)->send(); |
131
|
|
|
if ($response->isError()) { |
132
|
|
|
throw new \RuntimeException('Failed to query the server '.$this->host); |
133
|
|
|
} |
134
|
|
|
if ($response->getStatusCode() !== 200 || !($html = $response->getBody(true))) { |
|
|
|
|
135
|
|
|
return; |
136
|
|
|
} |
137
|
|
|
$html = iconv('windows-1251', 'utf-8', $html); |
138
|
|
|
|
139
|
|
|
// clean content |
140
|
|
|
$config = [ |
141
|
|
|
'output-xhtml' => true, |
142
|
|
|
'indent' => true, |
143
|
|
|
'indent-spaces' => 0, |
144
|
|
|
'fix-backslash' => true, |
145
|
|
|
'hide-comments' => true, |
146
|
|
|
'drop-empty-paras' => true, |
147
|
|
|
'wrap' => false, |
148
|
|
|
]; |
149
|
|
|
$tidy = new \tidy(); |
150
|
|
|
$tidy->parseString($html, $config, 'utf8'); |
151
|
|
|
$tidy->cleanRepair(); |
152
|
|
|
$html = $tidy->root()->value; |
|
|
|
|
153
|
|
|
// ignore blocks |
154
|
|
|
$html = preg_replace('/<noembed>.*?<\/noembed>/is', '', $html); |
155
|
|
|
$html = preg_replace('/<noindex>.*?<\/noindex>/is', '', $html); |
156
|
|
|
// remove noembed |
157
|
|
|
return $html; |
158
|
|
|
} |
159
|
|
|
|
160
|
|
|
/** |
161
|
|
|
* Get HTTP browser. |
162
|
|
|
* |
163
|
|
|
* @param \Guzzle\Http\Client |
164
|
|
|
*/ |
165
|
|
|
protected function getBrowser() |
166
|
|
|
{ |
167
|
|
|
if (!($this->browser instanceof Client)) { |
168
|
|
|
$this->browser = new Client($this->host); |
169
|
|
|
|
170
|
|
|
// try to set User-Agent from original request |
171
|
|
|
$user_agent = self::DEFAULT_USER_AGENT; |
172
|
|
|
if ($this->request) { |
173
|
|
|
$user_agent = $this->request->server->get('HTTP_USER_AGENT', self::DEFAULT_USER_AGENT); |
174
|
|
|
} |
175
|
|
|
$this->browser->setDefaultHeaders(['User-Agent' => $user_agent]); |
|
|
|
|
176
|
|
|
|
177
|
|
|
// configure browser client |
178
|
|
|
$this->browser->setDefaultOption('timeout', $this->timeout); |
179
|
|
|
if ($this->proxy_list) { |
|
|
|
|
180
|
|
|
$this->browser->setDefaultOption('proxy', $this->proxy_list[array_rand($this->proxy_list)]); |
181
|
|
|
} |
182
|
|
|
} |
183
|
|
|
|
184
|
|
|
return $this->browser; |
185
|
|
|
} |
186
|
|
|
} |
187
|
|
|
|
This method has been deprecated.