|
1
|
|
|
<?php |
|
2
|
|
|
|
|
3
|
|
|
namespace Lbc\Crawler; |
|
4
|
|
|
|
|
5
|
|
|
use Lbc\Filter\CitySanitizer; |
|
6
|
|
|
use Lbc\Filter\CpSanitizer; |
|
7
|
|
|
use Lbc\Filter\DefaultSanitizer; |
|
8
|
|
|
use Lbc\Filter\KeySanitizer; |
|
9
|
|
|
use Lbc\Parser\AdUrlParser; |
|
10
|
|
|
use League\Uri\Schemes\Http; |
|
11
|
|
|
use Symfony\Component\DomCrawler\Crawler; |
|
12
|
|
|
|
|
13
|
|
|
/** |
|
14
|
|
|
* Class AdCrawler |
|
15
|
|
|
* @package Lbc\Crawler |
|
16
|
|
|
*/ |
|
17
|
|
|
class AdCrawler extends CrawlerAbstract |
|
18
|
|
|
{ |
|
19
|
|
|
/** |
|
20
|
|
|
* @var AdUrlParser |
|
21
|
|
|
*/ |
|
22
|
|
|
protected $url; |
|
23
|
|
|
|
|
24
|
|
|
/** |
|
25
|
|
|
* @param $url |
|
26
|
|
|
* @return AdUrlParser |
|
27
|
|
|
*/ |
|
28
|
14 |
|
protected function setUrlParser($url) |
|
29
|
|
|
{ |
|
30
|
14 |
|
$this->url = new AdUrlParser($url); |
|
31
|
14 |
|
} |
|
32
|
|
|
|
|
33
|
|
|
/** |
|
34
|
|
|
* Return a full ad information |
|
35
|
|
|
* |
|
36
|
|
|
* @return array |
|
37
|
|
|
*/ |
|
38
|
6 |
|
public function getAll() |
|
39
|
|
|
{ |
|
40
|
6 |
|
return array_merge( |
|
41
|
|
|
[ |
|
42
|
6 |
|
'id' => $this->getUrlParser()->getId(), |
|
43
|
6 |
|
'category' => $this->getUrlParser()->getCategory(), |
|
44
|
6 |
|
], |
|
45
|
6 |
|
$this->getPictures(), |
|
46
|
6 |
|
$this->getProperties(), |
|
47
|
6 |
|
$this->getDescription() |
|
48
|
6 |
|
); |
|
49
|
|
|
} |
|
50
|
|
|
|
|
51
|
|
|
/** |
|
52
|
|
|
* Return an array with the Thumbs pictures url |
|
53
|
|
|
* |
|
54
|
|
|
* @param Crawler $node |
|
55
|
|
|
* @return array |
|
56
|
|
|
*/ |
|
57
|
8 |
|
public function getPictures(Crawler $node = null) |
|
58
|
|
|
{ |
|
59
|
8 |
|
$node = $node ?: $this->node; |
|
60
|
|
|
|
|
61
|
|
|
$images = [ |
|
62
|
8 |
|
'images_thumbs' => [], |
|
63
|
8 |
|
'images' => [], |
|
64
|
8 |
|
]; |
|
65
|
|
|
|
|
66
|
|
|
$node |
|
67
|
8 |
|
->filter('.adview_main script') |
|
68
|
|
|
->each(function (Crawler $crawler) use (&$images) { |
|
69
|
8 |
|
if (preg_match_all( |
|
70
|
8 |
|
'#//img.+.leboncoin.fr/.*\.jpg#', |
|
71
|
8 |
|
$crawler->html(), |
|
72
|
|
|
$matches |
|
73
|
8 |
|
)) { |
|
74
|
8 |
|
foreach ($matches[0] as $image) { |
|
75
|
8 |
|
if (preg_match('/thumb/', $image)) { |
|
76
|
8 |
|
array_push( |
|
77
|
8 |
|
$images['images_thumbs'], |
|
78
|
8 |
|
(string)Http::createFromString($image) |
|
79
|
8 |
|
->withScheme($this->sheme) |
|
80
|
8 |
|
); |
|
81
|
|
|
|
|
82
|
8 |
|
continue; |
|
83
|
|
|
} |
|
84
|
|
|
|
|
85
|
8 |
|
array_push( |
|
86
|
8 |
|
$images['images'], |
|
87
|
8 |
|
(string)Http::createFromString($image) |
|
88
|
8 |
|
->withScheme($this->sheme) |
|
89
|
8 |
|
); |
|
90
|
8 |
|
} |
|
91
|
8 |
|
} |
|
92
|
8 |
|
}); |
|
93
|
|
|
|
|
94
|
8 |
|
return $images; |
|
95
|
|
|
} |
|
96
|
|
|
|
|
97
|
|
|
/** |
|
98
|
|
|
* Return the common information (price, cp, city) |
|
99
|
|
|
* |
|
100
|
|
|
* @param Crawler $node |
|
101
|
|
|
* |
|
102
|
|
|
* @return array |
|
103
|
|
|
*/ |
|
104
|
8 |
|
public function getProperties(Crawler $node = null) |
|
105
|
|
|
{ |
|
106
|
8 |
|
$node = $node ?: $this->node; |
|
107
|
|
|
|
|
108
|
|
|
$properties = [ |
|
109
|
8 |
|
'titre' => DefaultSanitizer::clean( |
|
110
|
8 |
|
$node->filter('h1')->text() |
|
111
|
8 |
|
), |
|
112
|
|
|
'created_at' => $node |
|
113
|
8 |
|
->filter('*[itemprop=availabilityStarts]') |
|
114
|
8 |
|
->first() |
|
115
|
8 |
|
->attr('content'), |
|
116
|
8 |
|
'is_pro' => ($node->filter('.ispro')->count()), |
|
117
|
8 |
|
]; |
|
118
|
|
|
|
|
119
|
8 |
|
$node->filter('h2') |
|
120
|
8 |
|
->each(function (Crawler $crawler) use (&$properties) { |
|
121
|
8 |
|
$properties = array_merge( |
|
122
|
8 |
|
$properties, |
|
123
|
8 |
|
$this->sanitize( |
|
124
|
8 |
|
$crawler->filter('.property')->text(), |
|
125
|
8 |
|
$crawler->filter('.value')->text() |
|
126
|
8 |
|
) |
|
127
|
8 |
|
); |
|
128
|
8 |
|
}); |
|
129
|
|
|
|
|
130
|
8 |
|
return ['properties' => $properties]; |
|
131
|
|
|
} |
|
132
|
|
|
|
|
133
|
|
|
/** |
|
134
|
|
|
* Return the description |
|
135
|
|
|
* |
|
136
|
|
|
* @param Crawler $node |
|
137
|
|
|
* @return string |
|
138
|
|
|
*/ |
|
139
|
8 |
|
public function getDescription(Crawler $node = null) |
|
140
|
|
|
{ |
|
141
|
8 |
|
$node = $node ?: $this->node; |
|
142
|
|
|
|
|
143
|
|
|
return [ |
|
144
|
8 |
|
'description' => $this->getFieldValue( |
|
145
|
8 |
|
$node->filter("p[itemprop=description]"), |
|
146
|
|
|
null |
|
147
|
8 |
|
) |
|
148
|
8 |
|
]; |
|
149
|
|
|
} |
|
150
|
|
|
|
|
151
|
|
|
/** |
|
152
|
|
|
* Transform the properties name into a snake_case string and sanitize |
|
153
|
|
|
* the value |
|
154
|
|
|
* |
|
155
|
|
|
* @param string $key |
|
156
|
|
|
* @param string $value |
|
157
|
|
|
* @return string |
|
158
|
|
|
*/ |
|
159
|
8 |
|
private function sanitize($key, $value) |
|
160
|
|
|
{ |
|
161
|
8 |
|
$key = KeySanitizer::clean($key); |
|
162
|
|
|
|
|
163
|
8 |
|
if ($key == 'ville') { |
|
164
|
|
|
return [ |
|
165
|
8 |
|
'ville' => CitySanitizer::clean($value), |
|
166
|
8 |
|
'cp' => CpSanitizer::clean($value), |
|
167
|
8 |
|
]; |
|
168
|
|
|
} |
|
169
|
|
|
|
|
170
|
8 |
|
$filterName = 'Lbc\\Filter\\' . ucfirst($key) . 'Sanitizer'; |
|
171
|
|
|
|
|
172
|
8 |
|
if (!class_exists($filterName)) { |
|
173
|
8 |
|
$filterName = 'Lbc\\Filter\\DefaultSanitizer'; |
|
174
|
8 |
|
} |
|
175
|
|
|
|
|
176
|
8 |
|
return [$key => call_user_func("$filterName::clean", $value)]; |
|
177
|
|
|
} |
|
178
|
|
|
} |
|
179
|
|
|
|
rfussien /
leboncoin-crawler
