|
1
|
|
|
<?php |
|
2
|
|
|
|
|
3
|
|
|
namespace Spatie\Crawler\Handlers; |
|
4
|
|
|
|
|
5
|
|
|
use Spatie\Crawler\Crawler; |
|
6
|
|
|
use Spatie\Crawler\CrawlUrl; |
|
7
|
|
|
use Spatie\Crawler\LinkAdder; |
|
8
|
|
|
use Spatie\Crawler\CrawlerRobots; |
|
9
|
|
|
use Psr\Http\Message\UriInterface; |
|
10
|
|
|
use Spatie\Crawler\CrawlSubdomains; |
|
11
|
|
|
use Psr\Http\Message\StreamInterface; |
|
12
|
|
|
use Psr\Http\Message\ResponseInterface; |
|
13
|
|
|
use function GuzzleHttp\Psr7\stream_for; |
|
14
|
|
|
|
|
15
|
|
|
class CrawlRequestFulfilled |
|
16
|
|
|
{ |
|
17
|
|
|
/** @var \Spatie\Crawler\Crawler */ |
|
18
|
|
|
protected $crawler; |
|
19
|
|
|
|
|
20
|
|
|
/** @var \Spatie\Crawler\LinkAdder */ |
|
21
|
|
|
protected $linkAdder; |
|
22
|
|
|
|
|
23
|
|
|
public function __construct(Crawler $crawler) |
|
24
|
|
|
{ |
|
25
|
|
|
$this->crawler = $crawler; |
|
26
|
|
|
|
|
27
|
|
|
$this->linkAdder = new LinkAdder($this->crawler); |
|
28
|
|
|
} |
|
29
|
|
|
|
|
30
|
|
|
public function __invoke(ResponseInterface $response, $index) |
|
31
|
|
|
{ |
|
32
|
|
|
$robots = new CrawlerRobots($response, $this->crawler->mustRespectRobots()); |
|
33
|
|
|
|
|
34
|
|
|
if (! $robots->mayIndex()) { |
|
35
|
|
|
return; |
|
36
|
|
|
} |
|
37
|
|
|
|
|
38
|
|
|
$crawlUrl = $this->crawler->getCrawlQueue()->getUrlById($index); |
|
39
|
|
|
|
|
40
|
|
|
if ($this->crawler->mayExecuteJavaScript()) { |
|
41
|
|
|
$html = $this->getBodyAfterExecutingJavaScript($crawlUrl->url); |
|
42
|
|
|
|
|
43
|
|
|
$response = $response->withBody(stream_for($html)); |
|
44
|
|
|
} |
|
45
|
|
|
|
|
46
|
|
|
$this->handleCrawled($response, $crawlUrl); |
|
47
|
|
|
|
|
48
|
|
|
if (! $this->crawler->getCrawlProfile() instanceof CrawlSubdomains) { |
|
49
|
|
|
if ($crawlUrl->url->getHost() !== $this->crawler->getBaseUrl()->getHost()) { |
|
50
|
|
|
return; |
|
51
|
|
|
} |
|
52
|
|
|
} |
|
53
|
|
|
|
|
54
|
|
|
if (! $robots->mayFollow()) { |
|
55
|
|
|
return; |
|
56
|
|
|
} |
|
57
|
|
|
|
|
58
|
|
|
$body = $this->convertBodyToString($response->getBody(), $this->crawler->getMaximumResponseSize()); |
|
59
|
|
|
|
|
60
|
|
|
$this->linkAdder->addFromHtml($body, $crawlUrl->url); |
|
61
|
|
|
} |
|
62
|
|
|
|
|
63
|
|
|
protected function handleCrawled(ResponseInterface $response, CrawlUrl $crawlUrl) |
|
64
|
|
|
{ |
|
65
|
|
|
$this->crawler->getCrawlObservers()->crawled($crawlUrl, $response); |
|
66
|
|
|
} |
|
67
|
|
|
|
|
68
|
|
|
protected function convertBodyToString(StreamInterface $bodyStream, $readMaximumBytes = 1024 * 1024 * 2): string |
|
69
|
|
|
{ |
|
70
|
|
|
$bodyStream->rewind(); |
|
71
|
|
|
|
|
72
|
|
|
$body = $bodyStream->read($readMaximumBytes); |
|
73
|
|
|
|
|
74
|
|
|
return $body; |
|
75
|
|
|
} |
|
76
|
|
|
|
|
77
|
|
|
protected function getBodyAfterExecutingJavaScript(UriInterface $url): string |
|
78
|
|
|
{ |
|
79
|
|
|
$browsershot = $this->crawler->getBrowsershot(); |
|
80
|
|
|
|
|
81
|
|
|
$html = $browsershot->setUrl((string) $url)->bodyHtml(); |
|
82
|
|
|
|
|
83
|
|
|
return html_entity_decode($html); |
|
84
|
|
|
} |
|
85
|
|
|
} |
|
86
|
|
|
|
spatie /
crawler
