|
1
|
|
|
<?php |
|
2
|
|
|
|
|
3
|
|
|
namespace VDB\Spider\Downloader; |
|
4
|
|
|
|
|
5
|
|
|
use VDB\Spider\Downloader\DownloaderInterface; |
|
6
|
|
|
use Symfony\Component\EventDispatcher\Event; |
|
7
|
|
|
use Symfony\Component\EventDispatcher\EventDispatcher; |
|
8
|
|
|
use Symfony\Component\EventDispatcher\EventDispatcherInterface; |
|
9
|
|
|
use Symfony\Component\EventDispatcher\GenericEvent; |
|
10
|
|
|
use VDB\Spider\Event\SpiderEvents; |
|
11
|
|
|
use VDB\Spider\PersistenceHandler\MemoryPersistenceHandler; |
|
12
|
|
|
use VDB\Spider\PersistenceHandler\PersistenceHandlerInterface; |
|
13
|
|
|
use VDB\Spider\RequestHandler\GuzzleRequestHandler; |
|
14
|
|
|
use VDB\Spider\RequestHandler\RequestHandlerInterface; |
|
15
|
|
|
use VDB\Spider\Filter\PostFetchFilterInterface; |
|
16
|
|
|
use VDB\Spider\Resource; |
|
17
|
|
|
use VDB\Spider\Uri\DiscoveredUri; |
|
18
|
|
|
|
|
19
|
|
|
class Downloader implements DownloaderInterface |
|
20
|
|
|
{ |
|
21
|
|
|
/** @var EventDispatcherInterface */ |
|
22
|
|
|
private $dispatcher; |
|
23
|
|
|
|
|
24
|
|
|
/** @var PersistenceHandlerInterface */ |
|
25
|
|
|
private $persistenceHandler; |
|
26
|
|
|
|
|
27
|
|
|
/** @var RequestHandlerInterface */ |
|
28
|
|
|
private $requestHandler; |
|
29
|
|
|
|
|
30
|
|
|
/** @var int the maximum number of downloaded resources. 0 means no limit */ |
|
31
|
|
|
public $downloadLimit = 0; |
|
32
|
|
|
|
|
33
|
|
|
/** @var PostFetchFilterInterface[] */ |
|
34
|
|
|
private $postFetchFilters = array(); |
|
35
|
|
|
|
|
36
|
|
|
/** |
|
37
|
|
|
* @param PostFetchFilterInterface $filter |
|
38
|
|
|
*/ |
|
39
|
|
|
public function addPostFetchFilter(PostFetchFilterInterface $filter) |
|
40
|
|
|
{ |
|
41
|
|
|
$this->postFetchFilters[] = $filter; |
|
42
|
|
|
} |
|
43
|
|
|
|
|
44
|
|
|
/** |
|
45
|
|
|
* @return false|Resource |
|
46
|
|
|
*/ |
|
47
|
|
|
public function download(DiscoveredUri $uri) |
|
48
|
|
|
{ |
|
49
|
|
|
// Fetch the document |
|
50
|
|
|
if (!$resource = $this->fetchResource($uri)) { |
|
51
|
|
|
return false; |
|
52
|
|
|
} |
|
53
|
|
|
|
|
54
|
|
|
$this->getPersistenceHandler()->persist($resource); |
|
55
|
|
|
|
|
56
|
|
|
return $resource; |
|
57
|
|
|
} |
|
58
|
|
|
|
|
59
|
|
|
public function isDownLoadLimitExceeded() |
|
60
|
|
|
{ |
|
61
|
|
|
return $this->downloadLimit !== 0 && $this->getPersistenceHandler()->count() >= $this->downloadLimit; |
|
62
|
|
|
} |
|
63
|
|
|
|
|
64
|
|
|
/** |
|
65
|
|
|
* A shortcut for EventDispatcher::dispatch() |
|
66
|
|
|
* |
|
67
|
|
|
* @param string $eventName |
|
68
|
|
|
* @param null|Event $event |
|
69
|
|
|
*/ |
|
70
|
|
|
private function dispatch($eventName, Event $event = null) |
|
71
|
|
|
{ |
|
72
|
|
|
$this->getDispatcher()->dispatch($eventName, $event); |
|
73
|
|
|
} |
|
74
|
|
|
|
|
75
|
|
|
/** |
|
76
|
|
|
* @param EventDispatcherInterface $eventDispatcher |
|
77
|
|
|
* @return $this |
|
78
|
|
|
*/ |
|
79
|
|
|
public function setDispatcher(EventDispatcherInterface $eventDispatcher) |
|
80
|
|
|
{ |
|
81
|
|
|
$this->dispatcher = $eventDispatcher; |
|
82
|
|
|
|
|
83
|
|
|
return $this; |
|
84
|
|
|
} |
|
85
|
|
|
|
|
86
|
|
|
/** |
|
87
|
|
|
* @return EventDispatcherInterface |
|
88
|
|
|
*/ |
|
89
|
|
|
public function getDispatcher() |
|
90
|
|
|
{ |
|
91
|
|
|
if (!$this->dispatcher) { |
|
92
|
|
|
$this->dispatcher = new EventDispatcher(); |
|
93
|
|
|
} |
|
94
|
|
|
return $this->dispatcher; |
|
95
|
|
|
} |
|
96
|
|
|
|
|
97
|
|
|
|
|
98
|
|
|
/** |
|
99
|
|
|
* @param UriInterface $uri |
|
|
|
|
|
|
100
|
|
|
* @return Resource|false |
|
101
|
|
|
*/ |
|
102
|
|
|
protected function fetchResource(DiscoveredUri $uri) |
|
103
|
|
|
{ |
|
104
|
|
|
$this->dispatch(SpiderEvents::SPIDER_CRAWL_PRE_REQUEST, new GenericEvent($this, array('uri' => $uri))); |
|
105
|
|
|
|
|
106
|
|
|
try { |
|
107
|
|
|
$resource = $this->getRequestHandler()->request($uri); |
|
108
|
|
|
|
|
109
|
|
|
$this->dispatch(SpiderEvents::SPIDER_CRAWL_POST_REQUEST, new GenericEvent($this, array('uri' => $uri))); // necessary until we have 'finally' |
|
110
|
|
|
|
|
111
|
|
|
if ($this->matchesPostfetchFilter($resource)) { |
|
112
|
|
|
return false; |
|
113
|
|
|
} |
|
114
|
|
|
|
|
115
|
|
|
return $resource; |
|
116
|
|
|
} catch (\Exception $e) { |
|
117
|
|
|
$this->dispatch( |
|
118
|
|
|
SpiderEvents::SPIDER_CRAWL_ERROR_REQUEST, |
|
119
|
|
|
new GenericEvent($this, array('uri' => $uri, 'message' => $e->getMessage())) |
|
120
|
|
|
); |
|
121
|
|
|
|
|
122
|
|
|
$this->dispatch(SpiderEvents::SPIDER_CRAWL_POST_REQUEST, new GenericEvent($this, array('uri' => $uri))); // necessary until we have 'finally' |
|
123
|
|
|
|
|
124
|
|
|
return false; |
|
125
|
|
|
} |
|
126
|
|
|
} |
|
127
|
|
|
|
|
128
|
|
|
/** |
|
129
|
|
|
* @param Resource $resource |
|
130
|
|
|
* @return bool |
|
131
|
|
|
*/ |
|
132
|
|
|
private function matchesPostfetchFilter(Resource $resource) |
|
133
|
|
|
{ |
|
134
|
|
|
foreach ($this->postFetchFilters as $filter) { |
|
135
|
|
|
if ($filter->match($resource)) { |
|
136
|
|
|
$this->dispatch( |
|
137
|
|
|
SpiderEvents::SPIDER_CRAWL_FILTER_POSTFETCH, |
|
138
|
|
|
new GenericEvent($this, array('uri' => $resource->getUri())) |
|
139
|
|
|
); |
|
140
|
|
|
return true; |
|
141
|
|
|
} |
|
142
|
|
|
} |
|
143
|
|
|
return false; |
|
144
|
|
|
} |
|
145
|
|
|
|
|
146
|
|
|
/** |
|
147
|
|
|
* @param PersistenceHandlerInterface $persistenceHandler |
|
148
|
|
|
*/ |
|
149
|
|
|
public function setPersistenceHandler(PersistenceHandlerInterface $persistenceHandler) |
|
150
|
|
|
{ |
|
151
|
|
|
$this->persistenceHandler = $persistenceHandler; |
|
152
|
|
|
} |
|
153
|
|
|
|
|
154
|
|
|
/** |
|
155
|
|
|
* @return PersistenceHandlerInterface |
|
156
|
|
|
*/ |
|
157
|
|
|
public function getPersistenceHandler() |
|
158
|
|
|
{ |
|
159
|
|
|
if (!$this->persistenceHandler) { |
|
160
|
|
|
$this->persistenceHandler = new MemoryPersistenceHandler(); |
|
161
|
|
|
} |
|
162
|
|
|
|
|
163
|
|
|
return $this->persistenceHandler; |
|
164
|
|
|
} |
|
165
|
|
|
|
|
166
|
|
|
/** |
|
167
|
|
|
* @param RequestHandlerInterface $requestHandler |
|
168
|
|
|
*/ |
|
169
|
|
|
public function setRequestHandler(RequestHandlerInterface $requestHandler) |
|
170
|
|
|
{ |
|
171
|
|
|
$this->requestHandler = $requestHandler; |
|
172
|
|
|
} |
|
173
|
|
|
|
|
174
|
|
|
/** |
|
175
|
|
|
* @return RequestHandlerInterface |
|
176
|
|
|
*/ |
|
177
|
|
|
public function getRequestHandler() |
|
178
|
|
|
{ |
|
179
|
|
|
if (!$this->requestHandler) { |
|
180
|
|
|
$this->requestHandler = new GuzzleRequestHandler(); |
|
181
|
|
|
} |
|
182
|
|
|
|
|
183
|
|
|
return $this->requestHandler; |
|
184
|
|
|
} |
|
185
|
|
|
|
|
186
|
|
|
} |
|
187
|
|
|
|
mvdbos /
php-spider
This check looks for
@paramannotations where the type inferred by our type inference engine differs from the declared type.It makes a suggestion as to what type it considers more descriptive.
Most often this is a case of a parameter that can be null in addition to its declared types.