|
1
|
|
|
<?php |
|
2
|
|
|
|
|
3
|
|
|
namespace TreeHouse\IoBundle\Bridge\WorkerBundle\Executor; |
|
4
|
|
|
|
|
5
|
|
|
use Doctrine\Common\Persistence\ManagerRegistry; |
|
6
|
|
|
use Psr\Log\LoggerInterface; |
|
7
|
|
|
use Symfony\Component\OptionsResolver\Exception\InvalidArgumentException; |
|
8
|
|
|
use Symfony\Component\OptionsResolver\Options; |
|
9
|
|
|
use Symfony\Component\OptionsResolver\OptionsResolver; |
|
10
|
|
|
use TreeHouse\IoBundle\Entity\Scraper as ScraperEntity; |
|
11
|
|
|
use TreeHouse\IoBundle\Scrape\Exception\CrawlException; |
|
12
|
|
|
use TreeHouse\IoBundle\Scrape\Exception\RateLimitException; |
|
13
|
|
|
use TreeHouse\IoBundle\Scrape\ScraperFactory; |
|
14
|
|
|
use TreeHouse\WorkerBundle\Exception\RescheduleException; |
|
15
|
|
|
use TreeHouse\WorkerBundle\Executor\AbstractExecutor; |
|
16
|
|
|
|
|
17
|
|
|
class ScrapeUrlExecutor extends AbstractExecutor |
|
18
|
|
|
{ |
|
19
|
|
|
const NAME = 'scrape.url'; |
|
20
|
|
|
|
|
21
|
|
|
/** |
|
22
|
|
|
* @var ManagerRegistry |
|
23
|
|
|
*/ |
|
24
|
|
|
protected $doctrine; |
|
25
|
|
|
|
|
26
|
|
|
/** |
|
27
|
|
|
* @var ScraperFactory |
|
28
|
|
|
*/ |
|
29
|
|
|
protected $factory; |
|
30
|
|
|
|
|
31
|
|
|
/** |
|
32
|
|
|
* @var LoggerInterface |
|
33
|
|
|
*/ |
|
34
|
|
|
protected $logger; |
|
35
|
|
|
|
|
36
|
|
|
/** |
|
37
|
|
|
* @param ManagerRegistry $doctrine |
|
38
|
|
|
* @param ScraperFactory $factory |
|
39
|
|
|
* @param LoggerInterface $logger |
|
40
|
|
|
*/ |
|
41
|
|
|
public function __construct(ManagerRegistry $doctrine, ScraperFactory $factory, LoggerInterface $logger) |
|
42
|
|
|
{ |
|
43
|
|
|
$this->doctrine = $doctrine; |
|
44
|
|
|
$this->factory = $factory; |
|
45
|
|
|
$this->logger = $logger; |
|
46
|
|
|
} |
|
47
|
|
|
|
|
48
|
|
|
/** |
|
49
|
|
|
* @inheritdoc |
|
50
|
|
|
*/ |
|
51
|
|
|
public function getName() |
|
52
|
|
|
{ |
|
53
|
|
|
return self::NAME; |
|
54
|
|
|
} |
|
55
|
|
|
|
|
56
|
|
|
/** |
|
57
|
|
|
* @inheritdoc |
|
58
|
|
|
*/ |
|
59
|
|
|
public function configurePayload(OptionsResolver $resolver) |
|
60
|
|
|
{ |
|
61
|
|
|
$resolver->setRequired(0); |
|
62
|
|
|
$resolver->setAllowedTypes(0, 'numeric'); |
|
63
|
|
|
$resolver->setNormalizer(0, function (Options $options, $value) { |
|
64
|
|
|
if (null === $scraper = $this->findScraper($value)) { |
|
65
|
|
|
throw new InvalidArgumentException(sprintf('Could not find scraper with id %d', $value)); |
|
66
|
|
|
} |
|
67
|
|
|
|
|
68
|
|
|
return $scraper; |
|
69
|
|
|
}); |
|
70
|
|
|
|
|
71
|
|
|
$resolver->setRequired(1); |
|
72
|
|
|
$resolver->setAllowedTypes(1, 'string'); |
|
73
|
|
|
} |
|
74
|
|
|
|
|
75
|
|
|
/** |
|
76
|
|
|
* @inheritdoc |
|
77
|
|
|
*/ |
|
78
|
|
|
public function execute(array $payload) |
|
79
|
|
|
{ |
|
80
|
|
|
/** @var ScraperEntity $entity */ |
|
81
|
|
|
/** @var string $url */ |
|
82
|
|
|
list($entity, $url) = $payload; |
|
83
|
|
|
|
|
84
|
|
|
$scraper = $this->factory->createScraper($entity); |
|
85
|
|
|
$scraper->setAsync(true); |
|
86
|
|
|
|
|
87
|
|
|
try { |
|
88
|
|
|
$scraper->scrape($entity, $url); |
|
89
|
|
|
|
|
90
|
|
|
return true; |
|
91
|
|
|
} catch (RateLimitException $e) { |
|
92
|
|
|
$re = new RescheduleException(); |
|
93
|
|
|
|
|
94
|
|
|
if ($date = $e->getRetryDate()) { |
|
95
|
|
|
$re->setRescheduleDate($date); |
|
96
|
|
|
} |
|
97
|
|
|
|
|
98
|
|
|
throw $re; |
|
99
|
|
|
} catch (CrawlException $e) { |
|
100
|
|
|
$this->logger->error($e->getMessage(), ['url' => $e->getUrl()]); |
|
101
|
|
|
|
|
102
|
|
|
return false; |
|
103
|
|
|
} |
|
104
|
|
|
} |
|
105
|
|
|
|
|
106
|
|
|
/** |
|
107
|
|
|
* @param int $scraperId |
|
108
|
|
|
* |
|
109
|
|
|
* @return ScraperEntity |
|
110
|
|
|
*/ |
|
111
|
|
|
protected function findScraper($scraperId) |
|
112
|
|
|
{ |
|
113
|
|
|
return $this->doctrine->getRepository('TreeHouseIoBundle:Scraper')->find($scraperId); |
|
114
|
|
|
} |
|
115
|
|
|
} |
|
116
|
|
|
|