|
1
|
|
|
<?php |
|
2
|
|
|
namespace CViniciusSDias\GoogleCrawler; |
|
3
|
|
|
|
|
4
|
|
|
use CViniciusSDias\GoogleCrawler\Exception\InvalidGoogleHtmlException; |
|
5
|
|
|
use CViniciusSDias\GoogleCrawler\Exception\InvalidResultException; |
|
6
|
|
|
use CViniciusSDias\GoogleCrawler\Proxy\{ |
|
7
|
|
|
GoogleProxyInterface, NoProxy |
|
8
|
|
|
}; |
|
9
|
|
|
use Symfony\Component\DomCrawler\Crawler as DomCrawler; |
|
10
|
|
|
use Symfony\Component\DomCrawler\Link; |
|
11
|
|
|
use DOMElement; |
|
12
|
|
|
|
|
13
|
|
|
/** |
|
14
|
|
|
* Google Crawler |
|
15
|
|
|
* |
|
16
|
|
|
* @package CViniciusSDias\GoogleCrawler |
|
17
|
|
|
* @author Vinicius Dias |
|
18
|
|
|
*/ |
|
19
|
|
|
class Crawler |
|
20
|
|
|
{ |
|
21
|
|
|
/** @var GoogleProxyInterface $proxy */ |
|
22
|
|
|
protected $proxy; |
|
23
|
|
|
/** @var SearchTermInterface $searchTerm */ |
|
24
|
|
|
private $searchTerm; |
|
25
|
|
|
/** @var string $countrySpecificSuffix */ |
|
26
|
|
|
private $googleDomain; |
|
27
|
|
|
/** @var string $countryCode */ |
|
28
|
|
|
private $countryCode; |
|
29
|
|
|
|
|
30
|
8 |
|
public function __construct( |
|
31
|
|
|
SearchTermInterface $searchTerm, |
|
32
|
|
|
GoogleProxyInterface $proxy = null, |
|
33
|
|
|
string $googleDomain = 'google.com', |
|
34
|
|
|
string $countryCode = '' |
|
35
|
|
|
) { |
|
36
|
8 |
|
$this->proxy = is_null($proxy) ? new NoProxy() : $proxy; |
|
|
|
|
|
|
37
|
8 |
|
$this->searchTerm = $searchTerm; |
|
38
|
|
|
|
|
39
|
8 |
|
if (stripos($googleDomain, 'google.') === false || stripos($googleDomain, 'http') === 0) { |
|
40
|
2 |
|
throw new \InvalidArgumentException('Invalid google domain'); |
|
41
|
|
|
} |
|
42
|
6 |
|
$this->googleDomain = $googleDomain; |
|
43
|
|
|
|
|
44
|
6 |
|
$this->countryCode = strtoupper($countryCode); |
|
45
|
6 |
|
} |
|
46
|
|
|
|
|
47
|
|
|
/** |
|
48
|
|
|
* Returns the 100 first found results for the specified search term |
|
49
|
|
|
* |
|
50
|
|
|
* @return ResultList |
|
51
|
|
|
* @throws \GuzzleHttp\Exception\ServerException If the proxy was overused |
|
52
|
|
|
* @throws \GuzzleHttp\Exception\ConnectException If the proxy is unavailable or $countrySpecificSuffix is invalid |
|
53
|
|
|
*/ |
|
54
|
2 |
|
public function getResults(): ResultList |
|
55
|
|
|
{ |
|
56
|
2 |
|
$googleUrl = $this->getGoogleUrl(); |
|
57
|
2 |
|
$response = $this->proxy->getHttpResponse($googleUrl); |
|
58
|
1 |
|
$stringResponse = (string) $response->getBody(); |
|
59
|
1 |
|
$domCrawler = new DomCrawler($stringResponse); |
|
60
|
1 |
|
$googleResultList = $domCrawler->filterXPath('//div[@class="ZINbbc xpd O9g5cc uUPGi"]'); |
|
61
|
1 |
|
if ($googleResultList->count() === 0) { |
|
62
|
1 |
|
throw new InvalidGoogleHtmlException('No parseable element found'); |
|
63
|
|
|
} |
|
64
|
|
|
|
|
65
|
|
|
$resultList = new ResultList($googleResultList->count()); |
|
66
|
|
|
|
|
67
|
|
|
foreach ($googleResultList as $googleResultElement) { |
|
68
|
|
|
try { |
|
69
|
|
|
$parsedResult = $this->parseDomElement($googleResultElement); |
|
70
|
|
|
$resultList->addResult($parsedResult); |
|
71
|
|
|
} catch (InvalidResultException $exception) { |
|
72
|
|
|
error_log( |
|
73
|
|
|
'Error parsing the following result: ' . print_r($googleResultElement, true), |
|
74
|
|
|
3, |
|
75
|
|
|
__DIR__ . '/../var/log/crawler-errors.log' |
|
76
|
|
|
); |
|
77
|
|
|
} |
|
78
|
|
|
} |
|
79
|
|
|
|
|
80
|
|
|
return $resultList; |
|
81
|
|
|
} |
|
82
|
|
|
|
|
83
|
|
|
/** |
|
84
|
|
|
* If $resultLink is a valid link, this method assembles the Result and adds it to $googleResults |
|
85
|
|
|
* |
|
86
|
|
|
* @param Link $resultLink |
|
87
|
|
|
* @param DOMElement $descriptionElement |
|
88
|
|
|
* @return Result |
|
89
|
|
|
* @throws InvalidResultException |
|
90
|
|
|
*/ |
|
91
|
|
|
private function createResult(Link $resultLink, DOMElement $descriptionElement): Result |
|
92
|
|
|
{ |
|
93
|
|
|
$description = $descriptionElement->nodeValue |
|
94
|
|
|
?? 'A description for this result isn\'t available due to the robots.txt file.'; |
|
95
|
|
|
|
|
96
|
|
|
$googleResult = new Result(); |
|
97
|
|
|
$googleResult |
|
98
|
|
|
->setTitle($resultLink->getNode()->nodeValue) |
|
99
|
|
|
->setUrl($this->parseUrl($resultLink->getUri())) |
|
100
|
|
|
->setDescription($description); |
|
101
|
|
|
|
|
102
|
|
|
return $googleResult; |
|
103
|
|
|
} |
|
104
|
|
|
|
|
105
|
|
|
/** |
|
106
|
|
|
* Parses the URL using the parser provided by $proxy |
|
107
|
|
|
* |
|
108
|
|
|
* @param string $url |
|
109
|
|
|
* @return string |
|
110
|
|
|
* @throws InvalidResultException |
|
111
|
|
|
*/ |
|
112
|
|
|
private function parseUrl(string $url): string |
|
113
|
|
|
{ |
|
114
|
|
|
return $this->proxy->parseUrl($url); |
|
115
|
|
|
} |
|
116
|
|
|
|
|
117
|
|
|
/** |
|
118
|
|
|
* Assembles the Google URL using the previously informed data |
|
119
|
|
|
*/ |
|
120
|
6 |
|
private function getGoogleUrl(): string |
|
121
|
|
|
{ |
|
122
|
6 |
|
$domain = $this->googleDomain; |
|
123
|
6 |
|
$url = "https://$domain/search?q={$this->searchTerm}&num=100"; |
|
124
|
6 |
|
if (!empty($this->countryCode)) { |
|
125
|
2 |
|
$url .= "&gl={$this->countryCode}"; |
|
126
|
|
|
} |
|
127
|
|
|
|
|
128
|
6 |
|
return $url; |
|
129
|
|
|
} |
|
130
|
|
|
|
|
131
|
|
|
private function isImageSuggestion(DomCrawler $resultCrawler) |
|
132
|
|
|
{ |
|
133
|
|
|
$resultCount = $resultCrawler |
|
134
|
|
|
->filterXpath('//img') |
|
135
|
|
|
->count(); |
|
136
|
|
|
|
|
137
|
|
|
return $resultCount > 0; |
|
138
|
|
|
} |
|
139
|
|
|
|
|
140
|
|
|
private function parseDomElement(DOMElement $result): Result |
|
141
|
|
|
{ |
|
142
|
|
|
$resultCrawler = new DomCrawler($result); |
|
143
|
|
|
$linkElement = $resultCrawler->filterXPath('//a')->getNode(0); |
|
144
|
|
|
if (is_null($linkElement)) { |
|
145
|
|
|
throw new InvalidResultException('Link element not found'); |
|
146
|
|
|
} |
|
147
|
|
|
|
|
148
|
|
|
$resultLink = new Link($linkElement, 'http://google.com/'); |
|
149
|
|
|
$descriptionElement = $resultCrawler->filterXPath('//div[@class="BNeawe s3v9rd AP7Wnd"]//div[@class="BNeawe s3v9rd AP7Wnd"]')->getNode(0); |
|
150
|
|
|
|
|
151
|
|
|
if (is_null($descriptionElement)) { |
|
152
|
|
|
throw new InvalidResultException('Description element not found'); |
|
153
|
|
|
} |
|
154
|
|
|
|
|
155
|
|
|
if ($this->isImageSuggestion($resultCrawler)) { |
|
156
|
|
|
throw new InvalidResultException('Result is an image suggestion'); |
|
157
|
|
|
} |
|
158
|
|
|
|
|
159
|
|
|
if (strpos($resultLink->getUri(), 'http://google.com') === false) { |
|
160
|
|
|
throw new InvalidResultException('Result is a google suggestion'); |
|
161
|
|
|
} |
|
162
|
|
|
|
|
163
|
|
|
$googleResult = $this->createResult($resultLink, $descriptionElement); |
|
164
|
|
|
return $googleResult; |
|
165
|
|
|
} |
|
166
|
|
|
} |
|
167
|
|
|
|
CViniciusSDias /
google-crawler
Our type inference engine has found an assignment to a property that is incompatible with the declared type of that property.
Either this assignment is in error or the assigned type should be added to the documentation/type hint for that property..