1
|
|
|
<?php |
2
|
|
|
|
3
|
|
|
namespace whm\Smoke\Extensions\SmokeResponseRetriever\Retriever\Crawler; |
4
|
|
|
|
5
|
|
|
use phm\HttpWebdriverClient\Http\Client\HttpClient; |
6
|
|
|
use PhmLabs\Components\Init\Init; |
7
|
|
|
use Psr\Http\Message\ResponseInterface; |
8
|
|
|
use Psr\Http\Message\UriInterface; |
9
|
|
|
use whm\Crawler\Crawler as whmCrawler; |
10
|
|
|
use whm\Crawler\PageContainer\PageContainer; |
11
|
|
|
use whm\Crawler\PageContainer\PatternAwareContainer; |
12
|
|
|
use whm\Html\Uri; |
13
|
|
|
use whm\Smoke\Extensions\SmokeResponseRetriever\Retriever\CrawlingRetriever; |
14
|
|
|
use whm\Smoke\Extensions\SmokeResponseRetriever\RetrieverException; |
15
|
|
|
use whm\Smoke\Scanner\SessionContainer; |
16
|
|
|
|
17
|
|
|
class Crawler implements CrawlingRetriever |
18
|
|
|
{ |
19
|
|
|
private $startPage; |
20
|
|
|
private $httpClient; |
21
|
|
|
private $parallelRequests; |
22
|
|
|
|
23
|
|
|
private $started = false; |
24
|
|
|
private $filters; |
25
|
|
|
|
26
|
|
|
/** |
27
|
|
|
* @var PageContainer |
28
|
|
|
*/ |
29
|
|
|
private $pageContainer; |
30
|
|
|
|
31
|
|
|
/** |
32
|
|
|
* @var \whm\Crawler\Crawler |
33
|
|
|
*/ |
34
|
|
|
private $crawler; |
35
|
|
|
|
36
|
|
|
public function init(array $filters, $pageContainer, $startPage = null, $parallelRequests = 5) |
37
|
|
|
{ |
38
|
|
|
$this->filters = Init::initializeAll($filters); |
39
|
|
|
if (!is_null($startPage)) { |
40
|
|
|
$this->startPage = new Uri($startPage); |
41
|
|
|
} |
42
|
|
|
$this->initPageContainer($pageContainer); |
43
|
|
|
$this->parallelRequests = $parallelRequests; |
44
|
|
|
} |
45
|
|
|
|
46
|
|
|
public function addPage(UriInterface $uri) |
47
|
|
|
{ |
48
|
|
|
$this->pageContainer->push($uri, true); |
49
|
|
|
} |
50
|
|
|
|
51
|
|
|
private function initPageContainer($pageContainerArray) |
52
|
|
|
{ |
53
|
|
|
$this->pageContainer = Init::initialize($pageContainerArray); |
54
|
|
|
|
55
|
|
|
// @todo this should be done inside a factory |
56
|
|
|
if ($this->pageContainer instanceof PatternAwareContainer) { |
57
|
|
|
if (array_key_exists('parameters', $pageContainerArray) && array_key_exists('pattern', $pageContainerArray['parameters'])) { |
58
|
|
|
foreach ($pageContainerArray['parameters']['pattern'] as $name => $pattern) { |
59
|
|
|
$this->pageContainer->registerPattern($name, $pattern); |
60
|
|
|
} |
61
|
|
|
} |
62
|
|
|
} |
63
|
|
|
} |
64
|
|
|
|
65
|
|
|
public function getStartPage() |
66
|
|
|
{ |
67
|
|
|
return $this->startPage; |
68
|
|
|
} |
69
|
|
|
|
70
|
|
|
public function setStartPage(UriInterface $startPage) |
71
|
|
|
{ |
72
|
|
|
$this->startPage = $startPage; |
73
|
|
|
} |
74
|
|
|
|
75
|
|
|
public function setHttpClient(HttpClient $httpClient) |
76
|
|
|
{ |
77
|
|
|
$this->httpClient = $httpClient; |
78
|
|
|
} |
79
|
|
|
|
80
|
|
|
/** |
81
|
|
|
* @return ResponseInterface |
82
|
|
|
*/ |
83
|
|
|
public function next() |
84
|
|
|
{ |
85
|
|
|
if (!$this->started) { |
86
|
|
|
$this->started = true; |
87
|
|
|
|
88
|
|
|
if (is_null($this->startPage)) { |
89
|
|
|
throw new \RuntimeException('The crawler you are using needs a start page to work, but it is not defined. '); |
90
|
|
|
} |
91
|
|
|
|
92
|
|
|
$this->crawler = new whmCrawler($this->httpClient, $this->pageContainer, $this->startPage, $this->parallelRequests); |
93
|
|
|
|
94
|
|
|
foreach ($this->filters as $filter) { |
95
|
|
|
$this->crawler->addFilter($filter); |
96
|
|
|
} |
97
|
|
|
} |
98
|
|
|
|
99
|
|
|
return $this->crawler->next(); |
100
|
|
|
} |
101
|
|
|
|
102
|
|
|
public function getComingFrom(UriInterface $uri) |
103
|
|
|
{ |
104
|
|
|
return $this->crawler->getComingFrom($uri); |
105
|
|
|
} |
106
|
|
|
|
107
|
|
|
public function getOriginUri(UriInterface $uri) |
108
|
|
|
{ |
109
|
|
|
return $uri; |
110
|
|
|
} |
111
|
|
|
|
112
|
|
|
public function setSessionContainer(SessionContainer $sessionContainer) |
113
|
|
|
{ |
114
|
|
|
} |
115
|
|
|
|
116
|
|
|
public function getOccuredExceptions() |
117
|
|
|
{ |
118
|
|
|
return []; |
119
|
|
|
} |
120
|
|
|
} |
121
|
|
|
|