|
1
|
|
|
<?php |
|
2
|
|
|
|
|
3
|
|
|
namespace PiedWeb\SeoPocketCrawler; |
|
4
|
|
|
|
|
5
|
|
|
use PiedWeb\UrlHarvester\Harvest; |
|
6
|
|
|
use PiedWeb\UrlHarvester\Indexable; |
|
7
|
|
|
use PiedWeb\UrlHarvester\Link; |
|
8
|
|
|
|
|
9
|
|
|
class CrawlerUrl |
|
10
|
|
|
{ |
|
11
|
|
|
/** @var mixed */ |
|
12
|
|
|
protected $harvest; |
|
13
|
|
|
/** @var Url */ |
|
14
|
|
|
protected $url; |
|
15
|
|
|
/** @var CrawlerConfig */ |
|
16
|
|
|
protected $config; |
|
17
|
|
|
|
|
18
|
|
|
/** @var array internal links from the current Url */ |
|
19
|
|
|
protected $links = []; |
|
20
|
|
|
|
|
21
|
|
|
public function __construct(Url $url, CrawlerConfig $config) |
|
22
|
|
|
{ |
|
23
|
|
|
$this->url = $url; |
|
24
|
|
|
$this->config = $config; |
|
25
|
|
|
|
|
26
|
|
|
$this->harvest(); |
|
27
|
|
|
} |
|
28
|
|
|
|
|
29
|
|
|
protected function harvest() |
|
30
|
|
|
{ |
|
31
|
|
|
if ($this->isNetworkError()) { |
|
32
|
|
|
return null; |
|
33
|
|
|
} |
|
34
|
|
|
|
|
35
|
|
|
if ($this->isRedirection()) { |
|
36
|
|
|
return null; |
|
37
|
|
|
} |
|
38
|
|
|
|
|
39
|
|
|
$this->defaultHarvesting(); |
|
40
|
|
|
} |
|
41
|
|
|
|
|
42
|
|
|
/** |
|
43
|
|
|
* permit to easily extend and change what is harvested, for example adding : |
|
44
|
|
|
* $this->harvestBreadcrumb(); |
|
45
|
|
|
* $this->url->setKws(','.implode(',', array_keys($this->getHarvester()->getKws())).','); // Slow ~20% |
|
46
|
|
|
* $this->url->setRatioTextCode($this->getHarvester()->getRatioTxtCode()); // Slow ~30% |
|
47
|
|
|
* $this->url->setH1($this->getHarvester()->getUniqueTag('h1') ?? '');. |
|
48
|
|
|
*/ |
|
49
|
|
|
protected function defaultHarvesting() |
|
50
|
|
|
{ |
|
51
|
|
|
$this->url->setIndexable($this->getHarvester()->indexable()); // slow ~30% |
|
52
|
|
|
|
|
53
|
|
|
$this->url->setMimeType((string) $this->getHarvester()->getResponse()->getMimeType()); |
|
54
|
|
|
|
|
55
|
|
|
$this->harvestLinks(); |
|
56
|
|
|
|
|
57
|
|
|
// Old way: $this->getHarvester()->getTextAnalysis()->getWordNumber(); |
|
58
|
|
|
$this->url->setWordCount($this->getHarvester()->getWordCount()); |
|
59
|
|
|
|
|
60
|
|
|
$this->url->setLoadTime($this->getHarvester()->getResponse()->getInfo('total_time')); |
|
61
|
|
|
|
|
62
|
|
|
$this->url->setSize($this->getHarvester()->getResponse()->getInfo('size_download')); |
|
63
|
|
|
|
|
64
|
|
|
$this->url->setTitle($this->getHarvester()->getUniqueTag('head title') ?? ''); |
|
65
|
|
|
} |
|
66
|
|
|
|
|
67
|
|
|
protected function isNetworkError() |
|
68
|
|
|
{ |
|
69
|
|
|
if (! $this->getHarvester() instanceof Harvest) { |
|
70
|
|
|
$this->url->setIndexable( |
|
71
|
|
|
42 != $this->getHarvester() ? Indexable::NOT_INDEXABLE_NETWORK_ERROR : Indexable::NOT_INDEXABLE_TOO_BIG |
|
72
|
|
|
); |
|
73
|
|
|
$this->config->getRecorder()->cache($this->getHarvester(), $this->url); |
|
74
|
|
|
|
|
75
|
|
|
return true; |
|
76
|
|
|
} |
|
77
|
|
|
|
|
78
|
|
|
$this->config->getRecorder()->cache($this->getHarvester(), $this->url); |
|
79
|
|
|
|
|
80
|
|
|
return false; |
|
81
|
|
|
} |
|
82
|
|
|
|
|
83
|
|
|
protected function isRedirection() |
|
84
|
|
|
{ |
|
85
|
|
|
if ($redir = $this->getHarvester()->getRedirectionLink()) { |
|
86
|
|
|
if ($redir->isInternalLink()) { // add to $links to permits to update counter & co |
|
87
|
|
|
$this->links[] = $redir; |
|
88
|
|
|
} |
|
89
|
|
|
$this->url->setIndexable(Indexable::NOT_INDEXABLE_3XX); |
|
90
|
|
|
|
|
91
|
|
|
return true; |
|
92
|
|
|
} |
|
93
|
|
|
|
|
94
|
|
|
return false; |
|
95
|
|
|
} |
|
96
|
|
|
|
|
97
|
|
|
protected function harvestLinks() |
|
98
|
|
|
{ |
|
99
|
|
|
$this->config->getRecorder()->recordOutboundLink($this->url, $this->getHarvester()->getLinks()); // ~10% |
|
100
|
|
|
$this->url->links = count($this->getHarvester()->getLinks()); |
|
101
|
|
|
$this->url->links_duplicate = $this->getHarvester()->getNbrDuplicateLinks(); |
|
102
|
|
|
$this->url->links_internal = count($this->getHarvester()->getLinks(Link::LINK_INTERNAL)); |
|
103
|
|
|
$this->url->links_self = count($this->getHarvester()->getLinks(Link::LINK_SELF)); |
|
104
|
|
|
$this->url->links_sub = count($this->getHarvester()->getLinks(Link::LINK_SUB)); |
|
105
|
|
|
$this->url->links_external = count($this->getHarvester()->getLinks(Link::LINK_EXTERNAL)); |
|
106
|
|
|
$this->links = $this->getHarvester()->getLinks(Link::LINK_INTERNAL); |
|
107
|
|
|
} |
|
108
|
|
|
|
|
109
|
|
|
/** |
|
110
|
|
|
* @return int|Harvest int correspond to curl error |
|
111
|
|
|
*/ |
|
112
|
|
|
public function getHarvester() |
|
113
|
|
|
{ |
|
114
|
|
|
if (null !== $this->harvest) { |
|
115
|
|
|
return $this->harvest; |
|
116
|
|
|
} |
|
117
|
|
|
|
|
118
|
|
|
$this->harvest = Harvest::fromUrl( |
|
119
|
|
|
$this->config->getBase().$this->url->getUri(), |
|
120
|
|
|
$this->config->getUserAgent(), |
|
121
|
|
|
'en,en-US;q=0.5', |
|
122
|
|
|
$this->config->getRequestCached() |
|
123
|
|
|
); |
|
124
|
|
|
|
|
125
|
|
|
if ($this->harvest instanceof Harvest && null !== $this->config->getRobotsTxtCached()) { |
|
126
|
|
|
$this->harvest->setRobotsTxt($this->config->getRobotsTxtCached()); |
|
127
|
|
|
} |
|
128
|
|
|
|
|
129
|
|
|
return $this->getHarvester(); |
|
130
|
|
|
} |
|
131
|
|
|
|
|
132
|
|
|
public function getLinks() |
|
133
|
|
|
{ |
|
134
|
|
|
return $this->links; |
|
135
|
|
|
} |
|
136
|
|
|
|
|
137
|
|
|
/* |
|
138
|
|
|
protected function harvestBreadcrumb() |
|
139
|
|
|
{ |
|
140
|
|
|
$breadcrumb = $this->getHarvester()->getBreadCrumb(); |
|
141
|
|
|
if (is_array($breadcrumb)) { |
|
142
|
|
|
$this->url->setBreadcrumbLevel(count($breadcrumb)); |
|
143
|
|
|
$this->url->setBreadcrumbFirst(isset($breadcrumb[1]) ? $breadcrumb[1]->getCleanName() : ''); |
|
144
|
|
|
$this->url->setBreadcrumbText($this->getHarvester()->getBreadCrumb('//')); |
|
145
|
|
|
} |
|
146
|
|
|
}/**/ |
|
147
|
|
|
} |
|
148
|
|
|
|
PiedWeb /
SeoPocketCrawler
