|
1
|
|
|
<?php |
|
2
|
|
|
|
|
3
|
|
|
namespace PiedWeb\SeoPocketCrawler; |
|
4
|
|
|
|
|
5
|
|
|
use PiedWeb\UrlHarvester\Harvest; |
|
6
|
|
|
use PiedWeb\UrlHarvester\Indexable; |
|
7
|
|
|
use Spatie\Robots\RobotsTxt; |
|
8
|
|
|
|
|
9
|
|
|
class Crawler |
|
10
|
|
|
{ |
|
11
|
|
|
protected $userAgent; |
|
12
|
|
|
protected $project; |
|
13
|
|
|
protected $ignore; |
|
14
|
|
|
protected $limit; |
|
15
|
|
|
protected $recorder; |
|
16
|
|
|
|
|
17
|
|
|
protected $currentClick = 0; |
|
18
|
|
|
|
|
19
|
|
|
protected $counter = 0; |
|
20
|
|
|
|
|
21
|
|
|
protected $base; |
|
22
|
|
|
protected $urls = []; |
|
23
|
|
|
|
|
24
|
3 |
|
public function __construct(string $startUrl, string $ignore, int $limit, string $userAgent) |
|
25
|
|
|
{ |
|
26
|
3 |
|
$this->urls[$startUrl] = null; |
|
27
|
3 |
|
$this->base = Harvest::getDomainAndSchemeFrom($startUrl); |
|
28
|
3 |
|
$this->project = preg_replace("([^\w\s\d\-_~,;\[\]\(\).])", '', $startUrl); |
|
29
|
3 |
|
$this->ignore = new RobotsTxt($ignore); |
|
30
|
3 |
|
$this->userAgent = $userAgent; |
|
31
|
3 |
|
$this->limit = $limit; |
|
32
|
|
|
|
|
33
|
3 |
|
$this->initRecorderAndCache(); |
|
34
|
3 |
|
} |
|
35
|
|
|
|
|
36
|
3 |
|
public function getDataFolder() |
|
37
|
|
|
{ |
|
38
|
3 |
|
return __DIR__.'/../data/'.$this->project; |
|
39
|
|
|
} |
|
40
|
|
|
|
|
41
|
3 |
|
public function getCacheFolder() |
|
42
|
|
|
{ |
|
43
|
3 |
|
return __DIR__.'/../cache/'.$this->project; |
|
44
|
|
|
} |
|
45
|
|
|
|
|
46
|
3 |
|
protected function initRecorderAndCache() |
|
47
|
|
|
{ |
|
48
|
3 |
|
$this->recorder = new Recorder($this->getDataFolder()); |
|
49
|
|
|
|
|
50
|
3 |
|
exec('rm -rf '.$this->getDataFolder()); |
|
51
|
3 |
|
exec('rm -rf '.$this->getCacheFolder()); |
|
52
|
|
|
|
|
53
|
3 |
|
if (!file_exists($this->getDataFolder())) { |
|
54
|
3 |
|
mkdir($this->getDataFolder()); |
|
55
|
3 |
|
mkdir($this->getDataFolder().'/links'); |
|
56
|
3 |
|
mkdir($this->getCacheFolder()); |
|
57
|
|
|
} |
|
58
|
3 |
|
} |
|
59
|
|
|
|
|
60
|
3 |
|
public function crawl(bool $debug = false) |
|
61
|
|
|
{ |
|
62
|
3 |
|
$nothingUpdated = true; |
|
63
|
|
|
|
|
64
|
3 |
|
if ($debug) { |
|
65
|
3 |
|
echo PHP_EOL.PHP_EOL.'// -----'.PHP_EOL.'// '.$this->counter.' crawled / ' |
|
66
|
3 |
|
.count($this->urls).' found '.PHP_EOL.'// -----'.PHP_EOL; |
|
67
|
|
|
} |
|
68
|
|
|
|
|
69
|
3 |
|
foreach ($this->urls as $urlToParse => $url) { |
|
70
|
3 |
|
if (null !== $url && (false === $url->can_be_crawled || true === $url->can_be_crawled)) { // déjà crawlé |
|
71
|
|
|
continue; |
|
72
|
|
|
} |
|
73
|
|
|
|
|
74
|
3 |
|
if ($debug) { |
|
75
|
3 |
|
echo ' '.$urlToParse.PHP_EOL; |
|
76
|
|
|
} |
|
77
|
|
|
|
|
78
|
3 |
|
$nothingUpdated = false; |
|
79
|
3 |
|
++$this->counter; |
|
80
|
|
|
|
|
81
|
3 |
|
$this->harvest($urlToParse); |
|
82
|
|
|
} |
|
83
|
|
|
|
|
84
|
3 |
|
++$this->currentClick; |
|
85
|
|
|
|
|
86
|
3 |
|
$record = $nothingUpdated || $this->currentClick >= $this->limit; |
|
87
|
|
|
|
|
88
|
3 |
|
return $record ? $this->recorder->record($this->urls) : $this->crawl($debug); |
|
89
|
|
|
} |
|
90
|
|
|
|
|
91
|
2 |
|
protected function cache($harvest) |
|
92
|
|
|
{ |
|
93
|
2 |
|
if (false === strpos($harvest->getResponse()->getContentType(), 'text/html')) { |
|
94
|
|
|
return; |
|
95
|
|
|
} |
|
96
|
|
|
|
|
97
|
2 |
|
$url = ltrim($harvest->getAbsoluteInternalLink($harvest->getResponse()->getEffectiveUrl()), '/'); |
|
98
|
2 |
|
$urlPart = explode('/', $url); |
|
99
|
2 |
|
$folder = $this->getCacheFolder(); |
|
100
|
|
|
|
|
101
|
2 |
|
$urlPartLenght = count($urlPart); |
|
102
|
2 |
|
for ($i = 0; $i < $urlPartLenght; ++$i) { |
|
103
|
2 |
|
if ($i == count($urlPart) - 1) { |
|
104
|
2 |
|
$filename = empty($urlPart[$i]) ? 'index.html' : $urlPart[$i]; |
|
105
|
2 |
|
file_put_contents($folder.'/'.$filename, $harvest->getResponse()->getContent()); |
|
106
|
|
|
} else { |
|
107
|
|
|
$folder .= '/'.$urlPart[$i]; |
|
108
|
|
|
if (!file_exists($folder)) { |
|
109
|
|
|
mkdir($folder); |
|
110
|
|
|
} |
|
111
|
|
|
} |
|
112
|
|
|
} |
|
113
|
2 |
|
} |
|
114
|
|
|
|
|
115
|
3 |
|
protected function harvest(string $urlToParse) |
|
116
|
|
|
{ |
|
117
|
3 |
|
$url = $this->urls[$urlToParse] = $this->urls[$urlToParse] ?? new Url($urlToParse, $this->currentClick); |
|
118
|
|
|
|
|
119
|
3 |
|
$url->updated_at = date('Ymd'); |
|
120
|
3 |
|
$url->can_be_crawled = $this->ignore->allows($urlToParse, $this->userAgent); |
|
121
|
|
|
|
|
122
|
3 |
|
if (false === $url->can_be_crawled) { |
|
123
|
|
|
return; |
|
124
|
|
|
} |
|
125
|
|
|
|
|
126
|
3 |
|
$harvest = Harvest::fromUrl($urlToParse, $this->userAgent); |
|
127
|
|
|
|
|
128
|
3 |
|
if (!$harvest instanceof Harvest) { |
|
129
|
1 |
|
$url->indexable = Indexable::NOT_INDEXABLE_NETWORK_ERROR; |
|
130
|
|
|
|
|
131
|
1 |
|
return; |
|
132
|
|
|
} |
|
133
|
|
|
|
|
134
|
2 |
|
$url->indexable = $harvest->isIndexable(); |
|
135
|
|
|
|
|
136
|
2 |
|
if (Indexable::NOT_INDEXABLE_3XX === $url->indexable) { |
|
137
|
|
|
$redir = $harvest->getRedirection(); |
|
138
|
|
|
if (false !== $redir) { |
|
139
|
|
|
$links = Harvest::LINK_INTERNAL === $harvest->getType($redir) ? [$redir] : []; |
|
|
|
|
|
|
140
|
|
|
} |
|
141
|
|
|
} else { |
|
142
|
2 |
|
$this->cache($harvest); |
|
143
|
|
|
|
|
144
|
2 |
|
$this->recorder->recordOutboundLink($url, $harvest->getLinks()); |
|
145
|
|
|
|
|
146
|
2 |
|
$url->links = count($harvest->getLinks()); |
|
147
|
2 |
|
$url->links_duplicate = $harvest->getNbrDuplicateLinks(); |
|
148
|
2 |
|
$url->links_internal = count($harvest->getLinks(Harvest::LINK_INTERNAL)); |
|
149
|
2 |
|
$url->links_self = count($harvest->getLinks(Harvest::LINK_SELF)); |
|
150
|
2 |
|
$url->links_sub = count($harvest->getLinks(Harvest::LINK_SUB)); |
|
151
|
2 |
|
$url->links_external = count($harvest->getLinks(Harvest::LINK_EXTERNAL)); |
|
152
|
|
|
|
|
153
|
2 |
|
$url->ratio_text_code = $harvest->getRatioTxtCode(); |
|
154
|
2 |
|
$url->load_time = $harvest->getResponse()->getInfo('total_time'); |
|
155
|
2 |
|
$url->size = $harvest->getResponse()->getInfo('size_download'); |
|
156
|
|
|
|
|
157
|
2 |
|
$breadcrumb = $harvest->getBreadCrumb(); |
|
158
|
2 |
|
if (is_array($breadcrumb)) { |
|
159
|
|
|
$url->breadcrumb_level = count($breadcrumb); |
|
160
|
|
|
$url->breadcrumb_fisrt = isset($breadcrumb[1]) ? $breadcrumb[1]->getCleanName() : ''; |
|
|
|
|
|
|
161
|
|
|
$url->breadcrumb_text = $harvest->getBreadCrumb('//'); |
|
162
|
|
|
} |
|
163
|
|
|
|
|
164
|
2 |
|
$url->title = $harvest->getUniqueTag('head title') ?? ''; |
|
165
|
2 |
|
$url->kws = ','.implode(',', $harvest->getKws()).','; |
|
166
|
2 |
|
$url->h1 = $harvest->getUniqueTag('h1') ?? ''; |
|
167
|
|
|
} |
|
168
|
|
|
|
|
169
|
2 |
|
foreach ($harvest->getLinks(Harvest::LINK_INTERNAL) as $link) { |
|
170
|
2 |
|
$linkUrl = $link->getPageUrl(); |
|
171
|
2 |
|
$this->urls[$linkUrl] = $this->urls[$linkUrl] ?? new Url($linkUrl, ($this->currentClick + 1)); |
|
172
|
2 |
|
$this->recorder->recordInboundLink($url, $this->urls[$linkUrl]); |
|
173
|
2 |
|
++$this->urls[$linkUrl]->inboundlinks; |
|
174
|
|
|
} |
|
175
|
2 |
|
} |
|
176
|
|
|
} |
|
177
|
|
|
|
PiedWeb /
SeoPocketCrawler
Loading history...