|
1
|
|
|
<?php declare(strict_types=1); |
|
2
|
|
|
|
|
3
|
|
|
namespace Goose; |
|
4
|
|
|
|
|
5
|
|
|
use Goose\Utils\Helper; |
|
6
|
|
|
use DOMWrap\Document; |
|
7
|
|
|
|
|
8
|
|
|
/** |
|
9
|
|
|
* Crawler |
|
10
|
|
|
* |
|
11
|
|
|
* @package Goose |
|
12
|
|
|
* @license http://www.apache.org/licenses/LICENSE-2.0 Apache License 2.0 |
|
13
|
|
|
*/ |
|
14
|
|
|
class Crawler { |
|
15
|
|
|
/** @var Configuration */ |
|
16
|
|
|
protected $config; |
|
17
|
|
|
|
|
18
|
|
|
/** |
|
19
|
|
|
* @param Configuration $config |
|
20
|
|
|
*/ |
|
21
|
|
|
public function __construct(Configuration $config) { |
|
22
|
|
|
$this->config = $config; |
|
23
|
|
|
} |
|
24
|
|
|
|
|
25
|
|
|
/** |
|
26
|
|
|
* @return Configuration |
|
27
|
|
|
*/ |
|
28
|
|
|
public function config(): Configuration { |
|
29
|
|
|
return $this->config; |
|
30
|
|
|
} |
|
31
|
|
|
|
|
32
|
|
|
/** |
|
33
|
|
|
* @param string $url |
|
34
|
|
|
* @param string|null $rawHTML |
|
35
|
|
|
* |
|
36
|
|
|
* @return Article |
|
37
|
|
|
*/ |
|
38
|
|
|
public function crawl(string $url, string $rawHTML = null): ?Article { |
|
39
|
|
|
$article = new Article(); |
|
40
|
|
|
|
|
41
|
|
|
$parseCandidate = Helper::getCleanedUrl($url); |
|
42
|
|
|
|
|
43
|
|
|
$xmlInternalErrors = libxml_use_internal_errors(true); |
|
44
|
|
|
|
|
45
|
|
|
if (empty($rawHTML)) { |
|
46
|
|
|
$guzzle = new \GuzzleHttp\Client(); |
|
47
|
|
|
$response = $guzzle->get($parseCandidate->url, $this->config()->get('browser')); |
|
48
|
|
|
$article->setRawResponse($response); |
|
49
|
|
|
$rawHTML = $response->getBody()->getContents(); |
|
50
|
|
|
} |
|
51
|
|
|
|
|
52
|
|
|
// Generate document |
|
53
|
|
|
$doc = $this->getDocument($rawHTML); |
|
54
|
|
|
|
|
55
|
|
|
// Set core mutators |
|
56
|
|
|
$article->setFinalUrl($parseCandidate->url); |
|
57
|
|
|
$article->setDomain($parseCandidate->parts->host); |
|
58
|
|
|
$article->setLinkhash($parseCandidate->linkhash); |
|
59
|
|
|
$article->setRawHtml($rawHTML); |
|
60
|
|
|
$article->setDoc($doc); |
|
61
|
|
|
$article->setRawDoc(clone $doc); |
|
62
|
|
|
|
|
63
|
|
|
// Pre-extraction document cleaning |
|
64
|
|
|
$this->modules('cleaners', $article); |
|
65
|
|
|
|
|
66
|
|
|
// Extract content |
|
67
|
|
|
$this->modules('extractors', $article); |
|
68
|
|
|
|
|
69
|
|
|
// Post-extraction content formatting |
|
70
|
|
|
$this->modules('formatters', $article); |
|
71
|
|
|
|
|
72
|
|
|
libxml_use_internal_errors($xmlInternalErrors); |
|
73
|
|
|
|
|
74
|
|
|
return $article; |
|
75
|
|
|
} |
|
76
|
|
|
|
|
77
|
|
|
/** |
|
78
|
|
|
* @param string $rawHTML |
|
79
|
|
|
* |
|
80
|
|
|
* @return Document |
|
81
|
|
|
*/ |
|
82
|
|
|
private function getDocument(string $rawHTML): Document { |
|
83
|
|
|
$doc = new Document(); |
|
84
|
|
|
$doc->html($rawHTML); |
|
85
|
|
|
|
|
86
|
|
|
return $doc; |
|
87
|
|
|
} |
|
88
|
|
|
|
|
89
|
|
|
/** |
|
90
|
|
|
* @param string $category |
|
91
|
|
|
* @param Article $article |
|
92
|
|
|
* |
|
93
|
|
|
* @return self |
|
94
|
|
|
*/ |
|
95
|
|
|
public function modules(string $category, Article $article): self { |
|
96
|
|
|
$modules = $this->config->getModules($category); |
|
97
|
|
|
|
|
98
|
|
|
foreach ($modules as $module) { |
|
99
|
|
|
$obj = new $module($this->config()); |
|
100
|
|
|
$obj->run($article); |
|
101
|
|
|
} |
|
102
|
|
|
|
|
103
|
|
|
return $this; |
|
104
|
|
|
} |
|
105
|
|
|
} |
|
106
|
|
|
|
|
107
|
|
|
|
scotteh /
php-goose
