|
1
|
|
|
<?php |
|
2
|
|
|
|
|
3
|
|
|
declare(strict_types=1); |
|
4
|
|
|
|
|
5
|
|
|
namespace Radowoj\Crawla; |
|
6
|
|
|
|
|
7
|
|
|
use Radowoj\Crawla\Link\Collection as LinkCollection; |
|
8
|
|
|
use Radowoj\Crawla\Link\CollectionInterface; |
|
9
|
|
|
use Symfony\Component\DomCrawler\Crawler as DomCrawler; |
|
10
|
|
|
use Symfony\Contracts\HttpClient\HttpClientInterface; |
|
11
|
|
|
|
|
12
|
|
|
final class Crawler implements CrawlerInterface |
|
13
|
|
|
{ |
|
14
|
|
|
public const DEPTH_ONLY_TARGET = 0; |
|
15
|
|
|
public const DEPTH_DEFAULT = 2; |
|
16
|
|
|
public const DEPTH_INFINITE = -100; |
|
17
|
|
|
|
|
18
|
|
|
private string $linkSelector = 'a'; |
|
19
|
|
|
/** @var callable|null */ |
|
20
|
|
|
private $urlValidatorCallback = null; |
|
21
|
|
|
/** @var callable|null */ |
|
22
|
|
|
private $pageVisitedCallback = null; |
|
23
|
|
|
private int $maxDepth = self::DEPTH_INFINITE; |
|
24
|
|
|
|
|
25
|
15 |
|
public function __construct( |
|
26
|
|
|
private string $baseUrl, |
|
27
|
|
|
private HttpClientInterface $client, |
|
28
|
|
|
private CollectionInterface $linksVisited = new LinkCollection(), |
|
29
|
|
|
private CollectionInterface $linksQueued = new LinkCollection(), |
|
30
|
|
|
private CollectionInterface $linksTooDeep = new LinkCollection(), |
|
31
|
|
|
) { |
|
32
|
15 |
|
} |
|
33
|
|
|
|
|
34
|
2 |
|
public function getLinkSelector(): string |
|
35
|
|
|
{ |
|
36
|
2 |
|
return $this->linkSelector; |
|
37
|
|
|
} |
|
38
|
|
|
|
|
39
|
1 |
|
public function setLinkSelector(string $linkSelector): CrawlerInterface |
|
40
|
|
|
{ |
|
41
|
1 |
|
$this->linkSelector = $linkSelector; |
|
42
|
|
|
|
|
43
|
1 |
|
return $this; |
|
44
|
|
|
} |
|
45
|
|
|
|
|
46
|
1 |
|
public function setVisited(CollectionInterface $linksVisited): CrawlerInterface |
|
47
|
|
|
{ |
|
48
|
1 |
|
$this->linksVisited = $linksVisited; |
|
49
|
|
|
|
|
50
|
1 |
|
return $this; |
|
51
|
|
|
} |
|
52
|
|
|
|
|
53
|
1 |
|
public function setQueued(CollectionInterface $linksQueued): CrawlerInterface |
|
54
|
|
|
{ |
|
55
|
1 |
|
$this->linksQueued = $linksQueued; |
|
56
|
|
|
|
|
57
|
1 |
|
return $this; |
|
58
|
|
|
} |
|
59
|
|
|
|
|
60
|
8 |
|
public function getVisited(): CollectionInterface |
|
61
|
|
|
{ |
|
62
|
8 |
|
return $this->linksVisited; |
|
63
|
|
|
} |
|
64
|
|
|
|
|
65
|
8 |
|
public function getQueued(): CollectionInterface |
|
66
|
|
|
{ |
|
67
|
8 |
|
return $this->linksQueued; |
|
68
|
|
|
} |
|
69
|
|
|
|
|
70
|
2 |
|
public function getTooDeep(): CollectionInterface |
|
71
|
|
|
{ |
|
72
|
2 |
|
return $this->linksTooDeep; |
|
73
|
|
|
} |
|
74
|
|
|
|
|
75
|
4 |
|
public function setUrlValidatorCallback(callable $urlValidatorCallback): CrawlerInterface |
|
76
|
|
|
{ |
|
77
|
4 |
|
$this->urlValidatorCallback = $urlValidatorCallback; |
|
78
|
|
|
|
|
79
|
4 |
|
return $this; |
|
80
|
|
|
} |
|
81
|
|
|
|
|
82
|
2 |
|
public function setPageVisitedCallback(callable $pageVisitedCallback): CrawlerInterface |
|
83
|
|
|
{ |
|
84
|
2 |
|
$this->pageVisitedCallback = $pageVisitedCallback; |
|
85
|
|
|
|
|
86
|
2 |
|
return $this; |
|
87
|
|
|
} |
|
88
|
|
|
|
|
89
|
6 |
|
public function crawl(int $maxDepth = self::DEPTH_DEFAULT): void |
|
90
|
|
|
{ |
|
91
|
6 |
|
$this->maxDepth = $maxDepth; |
|
92
|
6 |
|
$this->getQueued()->appendUrlsAtDepth([$this->baseUrl], 0); |
|
93
|
6 |
|
$this->crawlPages(); |
|
94
|
|
|
} |
|
95
|
|
|
|
|
96
|
6 |
|
private function crawlPages(): void |
|
97
|
|
|
{ |
|
98
|
6 |
|
while ($link = $this->getQueued()->shift()) { |
|
99
|
6 |
|
if (self::DEPTH_INFINITE !== $this->maxDepth && $link->getDepth() > $this->maxDepth) { |
|
100
|
1 |
|
$this->getTooDeep()->push($link); |
|
101
|
1 |
|
continue; |
|
102
|
|
|
} |
|
103
|
|
|
|
|
104
|
5 |
|
$response = $this->client->request('GET', $link->getUrl()); |
|
105
|
5 |
|
if (200 !== $response->getStatusCode()) { |
|
106
|
3 |
|
continue; |
|
107
|
|
|
} |
|
108
|
|
|
|
|
109
|
4 |
|
$this->getVisited()->push($link); |
|
110
|
|
|
|
|
111
|
4 |
|
$domCrawler = new DomCrawler( |
|
112
|
4 |
|
(string) $response->getContent(), |
|
113
|
4 |
|
$link->getUrl() |
|
114
|
4 |
|
); |
|
115
|
|
|
|
|
116
|
4 |
|
if (\is_callable($this->pageVisitedCallback)) { |
|
117
|
1 |
|
\call_user_func($this->pageVisitedCallback, $domCrawler); |
|
|
|
|
|
|
118
|
|
|
} |
|
119
|
|
|
|
|
120
|
4 |
|
$urls = $this->getUrls($domCrawler); |
|
121
|
4 |
|
$urls = $this->filterUrls($urls); |
|
122
|
4 |
|
$this->queueUrls($link->getDepth() + 1, $urls); |
|
123
|
|
|
} |
|
124
|
|
|
} |
|
125
|
|
|
|
|
126
|
1 |
|
private function isWithinBaseUrl($url): bool |
|
|
|
|
|
|
127
|
|
|
{ |
|
128
|
1 |
|
return 0 === mb_strpos($url, $this->baseUrl); |
|
129
|
|
|
} |
|
130
|
|
|
|
|
131
|
4 |
|
private function getUrls(DomCrawler $domCrawler): array |
|
132
|
|
|
{ |
|
133
|
4 |
|
$links = $domCrawler->filter($this->linkSelector)->links(); |
|
134
|
|
|
|
|
135
|
4 |
|
$urls = array_map(function ($link) { |
|
136
|
4 |
|
$url = $link->getUri(); |
|
137
|
4 |
|
$url = explode('#', $url); |
|
138
|
|
|
|
|
139
|
4 |
|
return $url[0]; |
|
140
|
4 |
|
}, $links); |
|
141
|
|
|
|
|
142
|
4 |
|
return array_unique($urls); |
|
143
|
|
|
} |
|
144
|
|
|
|
|
145
|
4 |
|
private function filterUrls(array $urls): array |
|
146
|
|
|
{ |
|
147
|
4 |
|
$urlConstraintCallback = \is_callable($this->urlValidatorCallback) |
|
148
|
3 |
|
? $this->urlValidatorCallback |
|
149
|
1 |
|
: $this->isWithinBaseUrl(...); |
|
150
|
|
|
|
|
151
|
4 |
|
return array_filter($urls, $urlConstraintCallback); |
|
152
|
|
|
} |
|
153
|
|
|
|
|
154
|
4 |
|
private function queueUrls(int $depth, array $urls): void |
|
155
|
|
|
{ |
|
156
|
4 |
|
$this->getQueued()->appendUrlsAtDepth( |
|
157
|
4 |
|
array_diff( |
|
158
|
4 |
|
$urls, |
|
159
|
4 |
|
$this->getQueued()->all(), |
|
160
|
4 |
|
$this->getVisited()->all() |
|
161
|
4 |
|
), |
|
162
|
4 |
|
$depth |
|
163
|
4 |
|
); |
|
164
|
|
|
} |
|
165
|
|
|
} |
|
166
|
|
|
|
radowoj /
crawla
Loading history...