|
1
|
|
|
<?php |
|
2
|
|
|
|
|
3
|
|
|
namespace VDB\Spider\Discoverer; |
|
4
|
|
|
|
|
5
|
|
|
use VDB\Spider\Resource; |
|
6
|
|
|
use VDB\Spider\Filter\PreFetchFilterInterface; |
|
7
|
|
|
use VDB\Spider\Uri\DiscoveredUri; |
|
8
|
|
|
|
|
9
|
|
|
class DiscovererSet |
|
10
|
|
|
{ |
|
11
|
|
|
/** |
|
12
|
|
|
* @var Discoverer[] |
|
13
|
|
|
*/ |
|
14
|
|
|
private $discoverers = array(); |
|
15
|
|
|
|
|
16
|
|
|
/** @var Filter[] */ |
|
17
|
|
|
private $filters = array(); |
|
18
|
|
|
|
|
19
|
|
|
/** |
|
20
|
|
|
* @var int maximum crawl depth |
|
21
|
|
|
*/ |
|
22
|
|
|
public $maxDepth = 3; |
|
23
|
|
|
|
|
24
|
|
|
/** |
|
25
|
|
|
* @var array the list of already visited URIs with the depth they were discovered on as value |
|
26
|
|
|
*/ |
|
27
|
|
|
private $alreadySeenUris = array(); |
|
28
|
|
|
|
|
29
|
|
|
public function __construct(array $discoverers = array()) |
|
30
|
|
|
{ |
|
31
|
|
|
foreach ($discoverers as $alias => $discoverer) { |
|
32
|
|
|
$this->set($discoverer, is_int($alias) ? null : $alias); |
|
33
|
|
|
} |
|
34
|
|
|
} |
|
35
|
|
|
|
|
36
|
|
|
/** |
|
37
|
|
|
* @param DiscoveredUri $uri |
|
38
|
|
|
* |
|
39
|
|
|
* Mark an Uri as already seen. |
|
40
|
|
|
* |
|
41
|
|
|
* If it already exists, it is not overwritten, since we want to keep the |
|
42
|
|
|
* first depth it was found at. |
|
43
|
|
|
*/ |
|
44
|
|
|
private function markSeen(DiscoveredUri $uri) |
|
45
|
|
|
{ |
|
46
|
|
|
$uriString = $uri->normalize()->toString(); |
|
47
|
|
|
if (!array_key_exists($uriString, $this->alreadySeenUris)) { |
|
48
|
|
|
$this->alreadySeenUris[$uriString] = $uri->getDepthFound(); |
|
49
|
|
|
} |
|
50
|
|
|
} |
|
51
|
|
|
|
|
52
|
|
|
/** |
|
53
|
|
|
* @return bool Returns true if this URI was found at max depth |
|
54
|
|
|
*/ |
|
55
|
|
|
private function isAtMaxDepth(DiscoveredUri $uri) |
|
56
|
|
|
{ |
|
57
|
|
|
return $uri->getDepthFound() === $this->maxDepth; |
|
58
|
|
|
} |
|
59
|
|
|
|
|
60
|
|
|
/** |
|
61
|
|
|
* @param Resource $resource |
|
62
|
|
|
* @return UriInterface[] |
|
63
|
|
|
*/ |
|
64
|
|
|
public function discover(Resource $resource) |
|
65
|
|
|
{ |
|
66
|
|
|
$this->markSeen($resource->getUri()); |
|
67
|
|
|
|
|
68
|
|
|
if ($this->isAtMaxDepth($resource->getUri())) { |
|
69
|
|
|
return []; |
|
70
|
|
|
} |
|
71
|
|
|
|
|
72
|
|
|
$discoveredUris = []; |
|
73
|
|
|
|
|
74
|
|
|
foreach ($this->discoverers as $discoverer) { |
|
75
|
|
|
$discoveredUris = array_merge($discoveredUris, $discoverer->discover($resource)); |
|
76
|
|
|
} |
|
77
|
|
|
|
|
78
|
|
|
$this->normalize($discoveredUris); |
|
79
|
|
|
$this->removeDuplicates($discoveredUris); |
|
80
|
|
|
$this->filterAlreadySeen($discoveredUris); |
|
81
|
|
|
$this->filter($discoveredUris); |
|
82
|
|
|
|
|
83
|
|
|
foreach ($discoveredUris as $uri) { |
|
84
|
|
|
$uri->setDepthFound($resource->getUri()->getDepthFound() + 1); |
|
85
|
|
|
$this->markSeen($uri); |
|
86
|
|
|
} |
|
87
|
|
|
|
|
88
|
|
|
return $discoveredUris; |
|
89
|
|
|
} |
|
90
|
|
|
|
|
91
|
|
|
/** |
|
92
|
|
|
* Sets a discoverer. |
|
93
|
|
|
* |
|
94
|
|
|
* @param discovererInterface $discoverer The discoverer instance |
|
95
|
|
|
* @param string|null $alias An alias |
|
96
|
|
|
*/ |
|
97
|
|
|
public function set(DiscovererInterface $discoverer, $alias = null) |
|
98
|
|
|
{ |
|
99
|
|
|
$this->discoverers[$discoverer->getName()] = $discoverer; |
|
100
|
|
|
if (null !== $alias) { |
|
101
|
|
|
$this->discoverers[$alias] = $discoverer; |
|
102
|
|
|
} |
|
103
|
|
|
} |
|
104
|
|
|
|
|
105
|
|
|
/** |
|
106
|
|
|
* @param PreFetchFilterInterface $filter |
|
107
|
|
|
*/ |
|
108
|
|
|
public function addFilter(PreFetchFilterInterface $filter) |
|
109
|
|
|
{ |
|
110
|
|
|
$this->filters[] = $filter; |
|
111
|
|
|
} |
|
112
|
|
|
|
|
113
|
|
|
/** |
|
114
|
|
|
* @param UriInterface[] $discoveredUris |
|
115
|
|
|
*/ |
|
116
|
|
|
private function normalize(array &$discoveredUris) |
|
117
|
|
|
{ |
|
118
|
|
|
foreach ($discoveredUris as &$uri) { |
|
119
|
|
|
$uri->normalize(); |
|
120
|
|
|
} |
|
121
|
|
|
} |
|
122
|
|
|
|
|
123
|
|
|
/** |
|
124
|
|
|
* @param UriInterface[] $discoveredUris |
|
125
|
|
|
*/ |
|
126
|
|
|
private function filterAlreadySeen(array &$discoveredUris) |
|
127
|
|
|
{ |
|
128
|
|
|
foreach ($discoveredUris as $k => &$uri) { |
|
129
|
|
|
if (array_key_exists($uri->toString(), $this->alreadySeenUris)) { |
|
130
|
|
|
unset($discoveredUris[$k]); |
|
131
|
|
|
} |
|
132
|
|
|
} |
|
133
|
|
|
} |
|
134
|
|
|
|
|
135
|
|
|
/** |
|
136
|
|
|
* @param UriInterface[] $discoveredUris |
|
137
|
|
|
*/ |
|
138
|
|
|
private function filter(array &$discoveredUris) |
|
139
|
|
|
{ |
|
140
|
|
|
foreach ($discoveredUris as $k => &$uri) { |
|
141
|
|
|
foreach ($this->filters as $filter) { |
|
142
|
|
|
if ($filter->match($uri)) { |
|
143
|
|
|
unset($discoveredUris[$k]); |
|
144
|
|
|
} |
|
145
|
|
|
} |
|
146
|
|
|
} |
|
147
|
|
|
} |
|
148
|
|
|
|
|
149
|
|
|
/** |
|
150
|
|
|
* @param UriInterface[] $discoveredUris |
|
151
|
|
|
*/ |
|
152
|
|
|
private function removeDuplicates(array &$discoveredUris) |
|
153
|
|
|
{ |
|
154
|
|
|
// make sure there are no duplicates in the list |
|
155
|
|
|
$tmp = array(); |
|
156
|
|
|
/** @var Uri $uri */ |
|
157
|
|
|
foreach ($discoveredUris as $k => $uri) { |
|
158
|
|
|
$tmp[$k] = $uri->toString(); |
|
159
|
|
|
} |
|
160
|
|
|
|
|
161
|
|
|
// Find duplicates in temporary array |
|
162
|
|
|
$tmp = array_unique($tmp); |
|
163
|
|
|
|
|
164
|
|
|
// Remove the duplicates from original array |
|
165
|
|
|
foreach ($discoveredUris as $k => $uri) { |
|
166
|
|
|
if (!array_key_exists($k, $tmp)) { |
|
167
|
|
|
unset($discoveredUris[$k]); |
|
168
|
|
|
} |
|
169
|
|
|
} |
|
170
|
|
|
} |
|
171
|
|
|
} |
|
172
|
|
|
|
mvdbos /
php-spider
