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