|
1
|
|
|
<?php |
|
2
|
|
|
namespace vipnytt\RobotsTxtParser; |
|
3
|
|
|
|
|
4
|
|
|
use GuzzleHttp; |
|
5
|
|
|
use vipnytt\RobotsTxtParser\Client; |
|
6
|
|
|
use vipnytt\RobotsTxtParser\Parser\RobotsTxtInterface; |
|
7
|
|
|
|
|
8
|
|
|
/** |
|
9
|
|
|
* Class Download |
|
10
|
|
|
* |
|
11
|
|
|
* @package vipnytt\RobotsTxtParser |
|
12
|
|
|
*/ |
|
13
|
|
|
class Download implements RobotsTxtInterface |
|
14
|
|
|
{ |
|
15
|
|
|
/** |
|
16
|
|
|
* Base uri |
|
17
|
|
|
* @var string |
|
18
|
|
|
*/ |
|
19
|
|
|
protected $baseUri; |
|
20
|
|
|
|
|
21
|
|
|
/** |
|
22
|
|
|
* HTTP Status code |
|
23
|
|
|
* @var int |
|
24
|
|
|
*/ |
|
25
|
|
|
protected $statusCode; |
|
26
|
|
|
|
|
27
|
|
|
/** |
|
28
|
|
|
* Robots.txt contents |
|
29
|
|
|
* @var string |
|
30
|
|
|
*/ |
|
31
|
|
|
protected $contents; |
|
32
|
|
|
|
|
33
|
|
|
/** |
|
34
|
|
|
* Robots.txt character encoding |
|
35
|
|
|
* @var string |
|
36
|
|
|
*/ |
|
37
|
|
|
protected $encoding; |
|
38
|
|
|
|
|
39
|
|
|
/** |
|
40
|
|
|
* Download constructor. |
|
41
|
|
|
* |
|
42
|
|
|
* @param string $baseUri |
|
43
|
|
|
* @param array $guzzleConfig |
|
44
|
|
|
*/ |
|
45
|
|
|
public function __construct($baseUri, $guzzleConfig = []) |
|
46
|
|
|
{ |
|
47
|
|
|
$this->baseUri = $baseUri; |
|
48
|
|
|
try { |
|
49
|
|
|
$client = new GuzzleHttp\Client( |
|
50
|
|
|
array_merge_recursive( |
|
51
|
|
|
[ |
|
52
|
|
|
'allow_redirects' => [ |
|
53
|
|
|
'max' => self::MAX_REDIRECTS, |
|
54
|
|
|
'referer' => true, |
|
55
|
|
|
'strict' => true, |
|
56
|
|
|
'track_redirects' => true, |
|
57
|
|
|
], |
|
58
|
|
|
'base_uri' => $baseUri, |
|
59
|
|
|
'headers' => [ |
|
60
|
|
|
'Accept' => 'text/plain;q=1.0, text/*;q=0.8, */*;q=0.1', |
|
61
|
|
|
'Accept-Charset' => 'utf-8;q=1.0, *;q=0.1', |
|
62
|
|
|
'Accept-Encoding' => 'identity;q=1.0, *;q=0.1', |
|
63
|
|
|
'User-Agent' => 'RobotsTxtParser-VIPnytt/1.0 (+https://github.com/VIPnytt/RobotsTxtParser/blob/master/README.md)', |
|
64
|
|
|
], |
|
65
|
|
|
'http_errors' => false, |
|
66
|
|
|
'timeout' => 60, |
|
67
|
|
|
'verify' => true, |
|
68
|
|
|
], |
|
69
|
|
|
$guzzleConfig |
|
70
|
|
|
) |
|
71
|
|
|
); |
|
72
|
|
|
$response = $client->request('GET', '/robots.txt'); |
|
73
|
|
|
$this->statusCode = $response->getStatusCode(); |
|
74
|
|
|
$this->contents = $response->getBody()->getContents(); |
|
75
|
|
|
$this->encoding = $this->headerEncoding($response->getHeader('content-type')); |
|
76
|
|
|
} catch (GuzzleHttp\Exception\TransferException $e) { |
|
77
|
|
|
$this->statusCode = 523; |
|
78
|
|
|
$this->contents = ''; |
|
79
|
|
|
$this->encoding = self::ENCODING; |
|
80
|
|
|
} |
|
81
|
|
|
} |
|
82
|
|
|
|
|
83
|
|
|
/** |
|
84
|
|
|
* HTTP header encoding |
|
85
|
|
|
* |
|
86
|
|
|
* @param array $headers |
|
87
|
|
|
* @return string |
|
88
|
|
|
*/ |
|
89
|
|
|
protected function headerEncoding($headers) |
|
90
|
|
|
{ |
|
91
|
|
|
foreach ($headers as $header) { |
|
92
|
|
|
$split = array_map('trim', mb_split(';', $header)); |
|
93
|
|
|
foreach ($split as $string) { |
|
94
|
|
|
if (mb_stripos($string, 'charset=') === 0) { |
|
95
|
|
|
$encoding = mb_split('=', $string, 2)[1]; |
|
96
|
|
|
if (in_array(mb_strtolower($encoding), array_map('mb_strtolower', mb_list_encodings()))) { |
|
97
|
|
|
return $encoding; |
|
98
|
|
|
} |
|
99
|
|
|
} |
|
100
|
|
|
} |
|
101
|
|
|
} |
|
102
|
|
|
return $this->detectEncoding(); |
|
103
|
|
|
} |
|
104
|
|
|
|
|
105
|
|
|
/** |
|
106
|
|
|
* Manually detect encoding |
|
107
|
|
|
* |
|
108
|
|
|
* @return string |
|
109
|
|
|
*/ |
|
110
|
|
|
protected function detectEncoding() |
|
111
|
|
|
{ |
|
112
|
|
|
if (($encoding = mb_detect_encoding($this->getContents())) !== false) { |
|
113
|
|
|
return $encoding; |
|
114
|
|
|
} |
|
115
|
|
|
return self::ENCODING; |
|
116
|
|
|
} |
|
117
|
|
|
|
|
118
|
|
|
/** |
|
119
|
|
|
* URL content |
|
120
|
|
|
* |
|
121
|
|
|
* @return string |
|
122
|
|
|
*/ |
|
123
|
|
|
public function getContents() |
|
124
|
|
|
{ |
|
125
|
|
|
return $this->contents; |
|
126
|
|
|
} |
|
127
|
|
|
|
|
128
|
|
|
/** |
|
129
|
|
|
* Parser client |
|
130
|
|
|
* |
|
131
|
|
|
* @param int|null $byteLimit |
|
132
|
|
|
* @return Client |
|
133
|
|
|
*/ |
|
134
|
|
|
public function parserClient($byteLimit = self::BYTE_LIMIT) |
|
135
|
|
|
{ |
|
136
|
|
|
return new Client($this->baseUri, $this->getStatusCode(), $this->getContents(), $this->getEncoding(), $byteLimit); |
|
137
|
|
|
} |
|
138
|
|
|
|
|
139
|
|
|
/** |
|
140
|
|
|
* Status code |
|
141
|
|
|
* |
|
142
|
|
|
* @return int |
|
143
|
|
|
*/ |
|
144
|
|
|
public function getStatusCode() |
|
145
|
|
|
{ |
|
146
|
|
|
return $this->statusCode; |
|
147
|
|
|
} |
|
148
|
|
|
|
|
149
|
|
|
/** |
|
150
|
|
|
* Encoding |
|
151
|
|
|
* |
|
152
|
|
|
* @return string |
|
153
|
|
|
*/ |
|
154
|
|
|
public function getEncoding() |
|
155
|
|
|
{ |
|
156
|
|
|
return $this->encoding; |
|
157
|
|
|
} |
|
158
|
|
|
} |
|
159
|
|
|
|
VIPnytt /
RobotsTxtParser
