|
1
|
|
|
<?php |
|
2
|
|
|
|
|
3
|
|
|
namespace PiedWeb\SeoPocketCrawler; |
|
4
|
|
|
|
|
5
|
|
|
use League\Csv\Reader; |
|
6
|
|
|
|
|
7
|
|
|
/** |
|
8
|
|
|
* Page Rank Calculator. |
|
9
|
|
|
*/ |
|
10
|
|
|
class SimplePageRankCalculator |
|
11
|
|
|
{ |
|
12
|
|
|
/** |
|
13
|
|
|
* @var CrawlerConfig |
|
14
|
|
|
*/ |
|
15
|
|
|
protected $config; |
|
16
|
|
|
|
|
17
|
|
|
/** |
|
18
|
|
|
* @var int |
|
19
|
|
|
*/ |
|
20
|
|
|
protected $pagesNbr; |
|
21
|
|
|
|
|
22
|
|
|
/** |
|
23
|
|
|
* @var array [id, pagerank] |
|
24
|
|
|
*/ |
|
25
|
|
|
protected $results; |
|
26
|
|
|
|
|
27
|
|
|
protected $maxIteration = 10000; |
|
28
|
|
|
|
|
29
|
|
|
/** |
|
30
|
|
|
* @var array |
|
31
|
|
|
*/ |
|
32
|
|
|
protected $linksTo = []; |
|
33
|
|
|
protected $nbrLinksFrom = []; |
|
34
|
|
|
|
|
35
|
|
|
protected $dampingFactor = 0.85; |
|
36
|
|
|
|
|
37
|
|
|
public function __construct(string $id, ?string $dataDirectory = null) |
|
38
|
|
|
{ |
|
39
|
|
|
$this->config = CrawlerConfig::loadFrom($id, $dataDirectory); |
|
40
|
|
|
|
|
41
|
|
|
$this->initLinksIndex(); |
|
42
|
|
|
|
|
43
|
|
|
$this->results = array_fill_keys(array_reverse(array_keys($this->linksTo)), null); |
|
44
|
|
|
|
|
45
|
|
|
for ($iteration = 0; $iteration < $this->maxIteration; ++$iteration) { |
|
46
|
|
|
$this->calcul(); |
|
47
|
|
|
} |
|
48
|
|
|
} |
|
49
|
|
|
|
|
50
|
|
|
public function record() |
|
51
|
|
|
{ |
|
52
|
|
|
// merge it with previous data harvested |
|
53
|
|
|
$data = $this->config->getDataFromPreviousCrawl(); |
|
54
|
|
|
$urls = $data['urls']; |
|
55
|
|
|
|
|
56
|
|
|
foreach ($urls as $k => $url) { |
|
57
|
|
|
if (isset($this->results[$url->id])) { |
|
58
|
|
|
$urls[$k]->pagerank = $this->results[$url->id]; |
|
59
|
|
|
} |
|
60
|
|
|
} |
|
61
|
|
|
|
|
62
|
|
|
(new Recorder($this->config->getDataFolder(), $this->config->getCacheMethod()))->record($urls); |
|
63
|
|
|
|
|
64
|
|
|
// return data filepath |
|
65
|
|
|
return realpath($this->config->getDataFolder()).'/data.csv'; |
|
66
|
|
|
} |
|
67
|
|
|
|
|
68
|
|
|
protected function calcul() |
|
69
|
|
|
{ |
|
70
|
|
|
$ids = array_keys($this->results); |
|
71
|
|
|
foreach ($ids as $id) { |
|
72
|
|
|
$sumPR = 0; |
|
73
|
|
|
foreach ($this->getLinksTo($id) as $link) { |
|
74
|
|
|
$sumPR = $sumPR + $this->results[$link] / $this->getNbrLinksFrom($link); |
|
75
|
|
|
} |
|
76
|
|
|
|
|
77
|
|
|
$this->results[$id] = $this->dampingFactor * $sumPR + (1 - $this->dampingFactor) / $this->getPagesNbr(); |
|
78
|
|
|
} |
|
79
|
|
|
} |
|
80
|
|
|
|
|
81
|
|
|
protected function getPagesNbr() |
|
82
|
|
|
{ |
|
83
|
|
|
if (null !== $this->pagesNbr) { |
|
84
|
|
|
return $this->pagesNbr; |
|
85
|
|
|
} |
|
86
|
|
|
|
|
87
|
|
|
return $this->pagesNbr = count($this->linksTo); |
|
88
|
|
|
} |
|
89
|
|
|
|
|
90
|
|
|
protected function getLinksTo(int $id): ?array |
|
91
|
|
|
{ |
|
92
|
|
|
return $this->linksTo[$id]; |
|
93
|
|
|
} |
|
94
|
|
|
|
|
95
|
|
|
protected function getNbrLinksFrom(int $id): ?int |
|
96
|
|
|
{ |
|
97
|
|
|
return $this->nbrLinksFrom[$id]; |
|
98
|
|
|
} |
|
99
|
|
|
|
|
100
|
|
|
protected function initLinksIndex() |
|
101
|
|
|
{ |
|
102
|
|
|
$csv = Reader::createFromPath($this->config->getDataFolder().Recorder::LINKS_DIR.'/Index.csv', 'r'); |
|
103
|
|
|
$csv->setHeaderOffset(0); |
|
104
|
|
|
|
|
105
|
|
|
$records = $csv->getRecords(); |
|
106
|
|
|
foreach ($records as $r) { |
|
107
|
|
|
if (! isset($this->linksTo[$r['To']])) { |
|
108
|
|
|
$this->linksTo[$r['To']] = []; |
|
109
|
|
|
} |
|
110
|
|
|
$this->linksTo[$r['To']][] = $r['From']; |
|
111
|
|
|
|
|
112
|
|
|
$this->nbrLinksFrom[$r['From']] = ($this->nbrLinksFrom[$r['From']] ?? 0) + 1; |
|
113
|
|
|
} |
|
114
|
|
|
} |
|
115
|
|
|
} |
|
116
|
|
|
|
PiedWeb /
SeoPocketCrawler
