|
1
|
|
|
<?php |
|
2
|
|
|
/** |
|
3
|
|
|
* @author Matthijs van den Bos <[email protected]> |
|
4
|
|
|
* @copyright 2013 Matthijs van den Bos |
|
5
|
|
|
*/ |
|
6
|
|
|
|
|
7
|
|
|
namespace VDB\Spider\QueueManager; |
|
8
|
|
|
|
|
9
|
|
|
use VDB\Uri\UriInterface; |
|
10
|
|
|
use VDB\Spider\Exception\QueueException; |
|
11
|
|
|
use Symfony\Component\EventDispatcher\Event; |
|
12
|
|
|
use Symfony\Component\EventDispatcher\EventDispatcher; |
|
13
|
|
|
use Symfony\Component\EventDispatcher\EventDispatcherInterface; |
|
14
|
|
|
use Symfony\Component\EventDispatcher\GenericEvent; |
|
15
|
|
|
use VDB\Spider\Event\SpiderEvents; |
|
16
|
|
|
|
|
17
|
|
|
/** |
|
18
|
|
|
* TODO: make this an Iterable |
|
19
|
|
|
*/ |
|
20
|
|
|
class InMemoryQueueManager implements QueueManager |
|
21
|
|
|
{ |
|
22
|
|
|
/** @var int The maximum depth for the crawl */ |
|
23
|
|
|
public $maxDepth = 3; |
|
24
|
|
|
|
|
25
|
|
|
/** @var int The maximum size of the process queue for this spider. 0 means infinite */ |
|
26
|
|
|
public $maxQueueSize = 0; |
|
27
|
|
|
|
|
28
|
|
|
/** @var int the amount of times a Resource was enqueued */ |
|
29
|
|
|
private $currentQueueSize = 0; |
|
30
|
|
|
|
|
31
|
|
|
/** @var UriInterface[] the list of URIs to process */ |
|
32
|
|
|
private $traversalQueue = array(); |
|
33
|
|
|
|
|
34
|
|
|
/** @var int The traversal algorithm to use. Choose from the class constants |
|
35
|
|
|
*/ |
|
36
|
|
|
private $traversalAlgorithm = self::ALGORITHM_DEPTH_FIRST; |
|
37
|
|
|
|
|
38
|
|
|
/** @var EventDispatcherInterface */ |
|
39
|
|
|
private $dispatcher; |
|
40
|
|
|
|
|
41
|
|
|
/** |
|
42
|
|
|
* @param int $traversalAlgorithm Choose from the class constants |
|
43
|
|
|
* TODO: This should be extracted to a Strategy pattern |
|
44
|
|
|
*/ |
|
45
|
|
|
public function setTraversalAlgorithm($traversalAlgorithm) |
|
46
|
|
|
{ |
|
47
|
|
|
$this->traversalAlgorithm = $traversalAlgorithm; |
|
48
|
|
|
} |
|
49
|
|
|
|
|
50
|
|
|
/** |
|
51
|
|
|
* @return int |
|
52
|
|
|
*/ |
|
53
|
|
|
public function getTraversalAlgorithm() |
|
54
|
|
|
{ |
|
55
|
|
|
return $this->traversalAlgorithm; |
|
56
|
|
|
} |
|
57
|
|
|
|
|
58
|
|
|
/** |
|
59
|
|
|
* @param EventDispatcherInterface $eventDispatcher |
|
60
|
|
|
* @return $this |
|
61
|
|
|
*/ |
|
62
|
|
|
public function setDispatcher(EventDispatcherInterface $eventDispatcher) |
|
63
|
|
|
{ |
|
64
|
|
|
$this->dispatcher = $eventDispatcher; |
|
65
|
|
|
|
|
66
|
|
|
return $this; |
|
67
|
|
|
} |
|
68
|
|
|
|
|
69
|
|
|
/** |
|
70
|
|
|
* @return EventDispatcherInterface |
|
71
|
|
|
*/ |
|
72
|
|
|
public function getDispatcher() |
|
73
|
|
|
{ |
|
74
|
|
|
if (!$this->dispatcher) { |
|
75
|
|
|
$this->dispatcher = new EventDispatcher(); |
|
76
|
|
|
} |
|
77
|
|
|
return $this->dispatcher; |
|
78
|
|
|
} |
|
79
|
|
|
|
|
80
|
|
|
/** |
|
81
|
|
|
* @param UriInterface |
|
82
|
|
|
*/ |
|
83
|
|
|
public function addUri(UriInterface $uri) |
|
84
|
|
|
{ |
|
85
|
|
|
if ($this->maxQueueSize != 0 && $this->currentQueueSize >= $this->maxQueueSize) { |
|
86
|
|
|
throw new QueueException('Maximum Queue Size of ' . $this->maxQueueSize . ' reached'); |
|
87
|
|
|
} |
|
88
|
|
|
|
|
89
|
|
|
$this->currentQueueSize++; |
|
90
|
|
|
array_push($this->traversalQueue, $uri); |
|
91
|
|
|
|
|
92
|
|
|
$this->getDispatcher()->dispatch( |
|
93
|
|
|
SpiderEvents::SPIDER_CRAWL_POST_ENQUEUE, |
|
94
|
|
|
new GenericEvent($this, array('uri' => $uri)) |
|
95
|
|
|
); |
|
96
|
|
|
} |
|
97
|
|
|
|
|
98
|
|
|
public function next() |
|
99
|
|
|
{ |
|
100
|
|
|
if ($this->traversalAlgorithm === static::ALGORITHM_DEPTH_FIRST) { |
|
101
|
|
|
return array_pop($this->traversalQueue); |
|
102
|
|
|
} elseif ($this->traversalAlgorithm === static::ALGORITHM_BREADTH_FIRST) { |
|
103
|
|
|
return array_shift($this->traversalQueue); |
|
104
|
|
|
} else { |
|
105
|
|
|
throw new \LogicException('No search algorithm set'); |
|
106
|
|
|
} |
|
107
|
|
|
} |
|
108
|
|
|
} |
|
109
|
|
|
|
mvdbos /
php-spider
