1
|
|
|
<?php |
2
|
|
|
namespace vipnytt\RobotsTxtParser; |
3
|
|
|
|
4
|
|
|
use DateTime; |
5
|
|
|
use GuzzleHttp; |
6
|
|
|
use vipnytt\RobotsTxtParser\Client; |
7
|
|
|
use vipnytt\RobotsTxtParser\Parser\RobotsTxtInterface; |
8
|
|
|
|
9
|
|
|
/** |
10
|
|
|
* Class Download |
11
|
|
|
* |
12
|
|
|
* @package vipnytt\RobotsTxtParser |
13
|
|
|
*/ |
14
|
|
|
class Download implements RobotsTxtInterface |
15
|
|
|
{ |
16
|
|
|
/** |
17
|
|
|
* Base uri |
18
|
|
|
* @var string |
19
|
|
|
*/ |
20
|
|
|
protected $baseUri; |
21
|
|
|
|
22
|
|
|
/** |
23
|
|
|
* Download time |
24
|
|
|
* @var int |
25
|
|
|
*/ |
26
|
|
|
protected $time; |
27
|
|
|
|
28
|
|
|
/** |
29
|
|
|
* Robots.txt max-age |
30
|
|
|
* @var int|null |
31
|
|
|
*/ |
32
|
|
|
protected $maxAge = null; |
33
|
|
|
|
34
|
|
|
/** |
35
|
|
|
* HTTP Status code |
36
|
|
|
* @var int |
37
|
|
|
*/ |
38
|
|
|
protected $statusCode; |
39
|
|
|
|
40
|
|
|
/** |
41
|
|
|
* Robots.txt contents |
42
|
|
|
* @var string |
43
|
|
|
*/ |
44
|
|
|
protected $contents; |
45
|
|
|
|
46
|
|
|
/** |
47
|
|
|
* Robots.txt character encoding |
48
|
|
|
* @var string |
49
|
|
|
*/ |
50
|
|
|
protected $encoding; |
51
|
|
|
|
52
|
|
|
/** |
53
|
|
|
* Parser client class |
54
|
|
|
* @var Client |
55
|
|
|
*/ |
56
|
|
|
protected $parserClient; |
57
|
|
|
|
58
|
|
|
/** |
59
|
|
|
* Download constructor. |
60
|
|
|
* |
61
|
|
|
* @param string $baseUri |
62
|
|
|
* @param array $guzzleConfig |
63
|
|
|
*/ |
64
|
|
|
public function __construct($baseUri, array $guzzleConfig = []) |
65
|
|
|
{ |
66
|
|
|
$this->baseUri = $baseUri; |
67
|
|
|
try { |
68
|
|
|
$client = new GuzzleHttp\Client( |
69
|
|
|
array_merge_recursive( |
70
|
|
|
[ |
71
|
|
|
'allow_redirects' => [ |
72
|
|
|
'max' => self::MAX_REDIRECTS, |
73
|
|
|
'referer' => true, |
74
|
|
|
'strict' => true, |
75
|
|
|
'track_redirects' => true, |
76
|
|
|
], |
77
|
|
|
'base_uri' => $baseUri, |
78
|
|
|
'headers' => [ |
79
|
|
|
'Accept' => 'text/plain;q=1.0, text/*;q=0.8, */*;q=0.1', |
80
|
|
|
'Accept-Charset' => 'utf-8;q=1.0, *;q=0.1', |
81
|
|
|
'Accept-Encoding' => 'identity;q=1.0, *;q=0.1', |
82
|
|
|
'User-Agent' => 'RobotsTxtParser-VIPnytt/1.0 (+https://github.com/VIPnytt/RobotsTxtParser/blob/master/README.md)', |
83
|
|
|
], |
84
|
|
|
'http_errors' => false, |
85
|
|
|
'timeout' => 60, |
86
|
|
|
'verify' => true, |
87
|
|
|
], |
88
|
|
|
$guzzleConfig |
89
|
|
|
) |
90
|
|
|
); |
91
|
|
|
$response = $client->request('GET', '/robots.txt'); |
92
|
|
|
$this->time = time(); |
93
|
|
|
$this->statusCode = $response->getStatusCode(); |
94
|
|
|
$this->contents = $response->getBody()->getContents(); |
95
|
|
|
$this->encoding = $this->headerEncoding($response->getHeader('content-type')); |
96
|
|
|
$this->maxAge = $this->headerMaxAge($response->getHeader('cache-control')); |
97
|
|
|
} catch (GuzzleHttp\Exception\TransferException $e) { |
98
|
|
|
$this->statusCode = 523; |
99
|
|
|
$this->contents = ''; |
100
|
|
|
$this->encoding = self::ENCODING; |
101
|
|
|
$this->maxAge = 0; |
102
|
|
|
} |
103
|
|
|
} |
104
|
|
|
|
105
|
|
|
/** |
106
|
|
|
* Content-Type encoding HTTP header |
107
|
|
|
* |
108
|
|
|
* @param array $headers |
109
|
|
|
* @return string |
110
|
|
|
*/ |
111
|
|
|
protected function headerEncoding(array $headers) |
112
|
|
|
{ |
113
|
|
|
foreach ($headers as $header) { |
114
|
|
|
$split = array_map('trim', mb_split(';', $header)); |
115
|
|
|
foreach ($split as $string) { |
116
|
|
|
if (mb_stripos($string, 'charset=') === 0) { |
117
|
|
|
$encoding = mb_split('=', $string, 2)[1]; |
118
|
|
|
if (in_array(mb_strtolower($encoding), array_map('mb_strtolower', mb_list_encodings()))) { |
119
|
|
|
return $encoding; |
120
|
|
|
} |
121
|
|
|
} |
122
|
|
|
} |
123
|
|
|
} |
124
|
|
|
return $this->detectEncoding(); |
125
|
|
|
} |
126
|
|
|
|
127
|
|
|
/** |
128
|
|
|
* Manually detect encoding |
129
|
|
|
* |
130
|
|
|
* @return string |
131
|
|
|
*/ |
132
|
|
|
protected function detectEncoding() |
133
|
|
|
{ |
134
|
|
|
if (($encoding = mb_detect_encoding($this->getContents())) !== false) { |
135
|
|
|
return $encoding; |
136
|
|
|
} |
137
|
|
|
return self::ENCODING; |
138
|
|
|
} |
139
|
|
|
|
140
|
|
|
/** |
141
|
|
|
* URL content |
142
|
|
|
* |
143
|
|
|
* @return string |
144
|
|
|
*/ |
145
|
|
|
public function getContents() |
146
|
|
|
{ |
147
|
|
|
return $this->contents; |
148
|
|
|
} |
149
|
|
|
|
150
|
|
|
/** |
151
|
|
|
* Cache-Control max-age HTTP header |
152
|
|
|
* |
153
|
|
|
* @param array $headers |
154
|
|
|
* @return integer |
155
|
|
|
*/ |
156
|
|
|
protected function headerMaxAge(array $headers) |
157
|
|
|
{ |
158
|
|
|
foreach ($headers as $header) { |
159
|
|
|
$split = array_map('trim', mb_split(',', $header)); |
160
|
|
|
foreach ($split as $string) { |
161
|
|
|
if (mb_stripos($string, 'max-age=') === 0) { |
162
|
|
|
$maxAge = mb_split('=', $string, 2)[1]; |
163
|
|
|
return intval($maxAge); |
164
|
|
|
} |
165
|
|
|
} |
166
|
|
|
} |
167
|
|
|
return 0; |
168
|
|
|
} |
169
|
|
|
|
170
|
|
|
/** |
171
|
|
|
* Parser client |
172
|
|
|
* |
173
|
|
|
* @param int|null $byteLimit |
174
|
|
|
* @return Client |
175
|
|
|
*/ |
176
|
|
|
public function parserClient($byteLimit = self::BYTE_LIMIT) |
177
|
|
|
{ |
178
|
|
|
if (!is_object($this->parserClient)) { |
179
|
|
|
$this->parserClient = new Client($this->baseUri, $this->getStatusCode(), $this->getContents(), $this->getEncoding(), $byteLimit); |
180
|
|
|
} |
181
|
|
|
return $this->parserClient; |
182
|
|
|
} |
183
|
|
|
|
184
|
|
|
/** |
185
|
|
|
* Status code |
186
|
|
|
* |
187
|
|
|
* @return int |
188
|
|
|
*/ |
189
|
|
|
public function getStatusCode() |
190
|
|
|
{ |
191
|
|
|
return $this->statusCode; |
192
|
|
|
} |
193
|
|
|
|
194
|
|
|
/** |
195
|
|
|
* Encoding |
196
|
|
|
* |
197
|
|
|
* @return string |
198
|
|
|
*/ |
199
|
|
|
public function getEncoding() |
200
|
|
|
{ |
201
|
|
|
return $this->encoding; |
202
|
|
|
} |
203
|
|
|
|
204
|
|
|
/** |
205
|
|
|
* Next update timestamp |
206
|
|
|
* |
207
|
|
|
* @return \DateTime|false |
208
|
|
|
*/ |
209
|
|
|
public function nextUpdate() |
210
|
|
|
{ |
211
|
|
|
$dateTime = new DateTime; |
212
|
|
|
$dateTime->setTimestamp($this->time + self::CACHE_TIME); |
213
|
|
|
return $dateTime; |
214
|
|
|
} |
215
|
|
|
|
216
|
|
|
/** |
217
|
|
|
* Valid until timestamp |
218
|
|
|
* |
219
|
|
|
* @return \DateTime|false |
220
|
|
|
*/ |
221
|
|
|
public function validUntil() |
222
|
|
|
{ |
223
|
|
|
$dateTime = new DateTime; |
224
|
|
|
$dateTime->setTimestamp($this->time + max(self::CACHE_TIME, is_int($this->maxAge) ? $this->maxAge : 0)); |
225
|
|
|
return $dateTime; |
226
|
|
|
} |
227
|
|
|
} |
228
|
|
|
|