|
1
|
|
|
<?php |
|
2
|
|
|
|
|
3
|
|
|
namespace PiedWeb\SeoPocketCrawler; |
|
4
|
|
|
|
|
5
|
|
|
use PiedWeb\UrlHarvester\Harvest; |
|
6
|
|
|
use PiedWeb\UrlHarvester\Link; |
|
7
|
|
|
|
|
8
|
|
|
class Recorder |
|
9
|
|
|
{ |
|
10
|
|
|
const LINKS_DIR = '/links'; |
|
11
|
|
|
const CACHE_DIR = '/cache'; |
|
12
|
|
|
|
|
13
|
|
|
const CACHE_NONE = 0; |
|
14
|
|
|
const CACHE_ID = 2; |
|
15
|
|
|
const CACHE_URI = 1; |
|
16
|
|
|
|
|
17
|
|
|
protected $folder; |
|
18
|
|
|
protected $cacheMethod; |
|
19
|
9 |
|
|
|
20
|
|
|
public function __construct($folder, $cacheMethod = self::CACHE_ID) |
|
21
|
9 |
|
{ |
|
22
|
9 |
|
$this->folder = $folder; |
|
23
|
|
|
$this->cacheMethod = $cacheMethod; |
|
24
|
9 |
|
|
|
25
|
3 |
|
if (!file_exists($folder)) { |
|
26
|
3 |
|
mkdir($folder); |
|
27
|
3 |
|
} |
|
28
|
|
|
|
|
29
|
9 |
|
if (!file_exists($folder.Recorder::LINKS_DIR)) { |
|
30
|
|
|
mkdir($folder.Recorder::LINKS_DIR); |
|
31
|
6 |
|
$this->initLinksIndex(); |
|
32
|
|
|
} |
|
33
|
6 |
|
|
|
34
|
|
|
if (!file_exists($folder.Recorder::CACHE_DIR)) { |
|
35
|
|
|
mkdir($folder.Recorder::CACHE_DIR); |
|
36
|
|
|
} |
|
37
|
6 |
|
} |
|
38
|
6 |
|
|
|
39
|
6 |
|
public function cache(Harvest $harvest, Url $url) |
|
40
|
6 |
|
{ |
|
41
|
6 |
|
if (Recorder::CACHE_NONE === $this->cacheMethod || !$this->mustWeCache($harvest)) { |
|
42
|
|
|
return; |
|
43
|
|
|
} |
|
44
|
6 |
|
|
|
45
|
|
|
$filePath = $this->getCacheFilePath($url); |
|
46
|
2 |
|
if (!file_exists($filePath)) { |
|
47
|
|
|
file_put_contents( |
|
48
|
7 |
|
$filePath, |
|
49
|
|
|
$harvest->getResponse()->getHeaders(false).PHP_EOL.PHP_EOL.$harvest->getResponse()->getContent() |
|
50
|
7 |
|
); |
|
51
|
3 |
|
|
|
52
|
|
|
return file_put_contents($filePath.'---info', json_encode($harvest->getResponse()->getInfo())); |
|
53
|
4 |
|
} |
|
54
|
|
|
} |
|
55
|
|
|
|
|
56
|
|
|
public function getCacheFilePath(Url $url) |
|
57
|
3 |
|
{ |
|
58
|
|
|
if (Recorder::CACHE_URI === $this->cacheMethod) { |
|
59
|
3 |
|
return $this->getCacheFilePathWithUrlAsFilename($url); |
|
60
|
3 |
|
} else { |
|
61
|
3 |
|
return $this->getCacheFilePathWithIdAsFilename($url); |
|
62
|
|
|
} |
|
63
|
3 |
|
} |
|
64
|
3 |
|
|
|
65
|
3 |
|
protected function getCacheFilePathWithUrlAsFilename(Url $url) |
|
66
|
3 |
|
{ |
|
67
|
|
|
$url = trim($url->uri, '/').'/'; |
|
68
|
3 |
|
$urlPart = explode('/', $url); |
|
69
|
3 |
|
$folder = $this->folder.Recorder::CACHE_DIR; |
|
70
|
|
|
|
|
71
|
|
|
$urlPartLenght = count($urlPart); |
|
72
|
|
|
for ($i = 0; $i < $urlPartLenght; ++$i) { |
|
73
|
|
|
if ($i == $urlPartLenght - 1) { |
|
74
|
|
|
return $folder.'/'.(empty($urlPart[$i]) ? 'index.html' : $urlPart[$i]); |
|
75
|
|
|
} else { |
|
76
|
4 |
|
$folder .= '/'.$urlPart[$i]; |
|
77
|
|
|
if (!file_exists($folder) || !is_dir($folder)) { |
|
78
|
4 |
|
mkdir($folder); |
|
79
|
|
|
} |
|
80
|
|
|
} |
|
81
|
6 |
|
} |
|
82
|
|
|
} |
|
83
|
6 |
|
|
|
84
|
|
|
protected function getCacheFilePathWithIdAsFilename(Url $url) |
|
85
|
|
|
{ |
|
86
|
9 |
|
return $this->folder.Recorder::CACHE_DIR.'/'.(string) $url->id; |
|
87
|
|
|
} |
|
88
|
9 |
|
|
|
89
|
|
|
protected function mustWeCache(Harvest $harvest) |
|
90
|
9 |
|
{ |
|
91
|
9 |
|
return false !== strpos($harvest->getResponse()->getContentType(), 'text/html'); |
|
92
|
9 |
|
} |
|
93
|
|
|
|
|
94
|
9 |
|
public function record(array $urls) |
|
95
|
9 |
|
{ |
|
96
|
|
|
$dataCsv = fopen($this->folder.'/data.csv', 'w'); |
|
97
|
|
|
$indexCsv = fopen($this->folder.'/index.csv', 'w'); |
|
98
|
9 |
|
|
|
99
|
|
|
if (false !== $dataCsv && false !== $indexCsv) { |
|
100
|
9 |
|
$header = array_keys(get_object_vars(array_values($urls)[0])); |
|
101
|
|
|
fputcsv($dataCsv, $header); |
|
102
|
|
|
fputcsv($indexCsv, ['id', 'uri']); |
|
103
|
|
|
|
|
104
|
|
|
foreach ($urls as $url) { |
|
105
|
|
|
fputcsv($dataCsv, get_object_vars($url)); |
|
106
|
6 |
|
fputcsv($indexCsv, [$url->id, $url->uri]); |
|
107
|
|
|
} |
|
108
|
6 |
|
|
|
109
|
6 |
|
fclose($dataCsv); |
|
110
|
6 |
|
|
|
111
|
6 |
|
return true; |
|
112
|
|
|
} |
|
113
|
6 |
|
|
|
114
|
|
|
return false; |
|
115
|
6 |
|
} |
|
116
|
|
|
|
|
117
|
|
|
public function recordInboundLink(Link $link, Url $from, Url $to) |
|
118
|
6 |
|
{ |
|
119
|
6 |
|
file_put_contents( |
|
120
|
6 |
|
$this->folder.Recorder::LINKS_DIR.'/To_'.(string) $to->id.'_'.((int) $link->mayFollow()), |
|
121
|
6 |
|
$this->inboundLinkToStr($link).PHP_EOL, // can use ->relativize to get only /uri |
|
122
|
|
|
FILE_APPEND |
|
123
|
|
|
); |
|
124
|
|
|
} |
|
125
|
|
|
|
|
126
|
|
|
protected function inboundLinkToStr(Link $link) |
|
127
|
|
|
{ |
|
128
|
|
|
return $link->getParentUrl().';'.$link->getAnchor().';'.((int) $link->mayFollow()).';'.$link->getType(); |
|
129
|
|
|
} |
|
130
|
|
|
|
|
131
|
|
|
public function recordOutboundLink(Url $from, array $links) |
|
132
|
|
|
{ |
|
133
|
|
|
$links = array_map(function (Link $link) { |
|
134
|
|
|
return $link->getUrl().';'.$link->getAnchor().';'.((int) $link->mayFollow()).';'.$link->getType(); |
|
135
|
|
|
}, $links); |
|
136
|
|
|
file_put_contents($this->folder.Recorder::LINKS_DIR.'/From_'.(string) $from->id, implode(PHP_EOL, $links)); |
|
137
|
|
|
} |
|
138
|
|
|
|
|
139
|
|
|
protected function initLinksIndex() |
|
140
|
|
|
{ |
|
141
|
|
|
if (!file_exists($this->folder.Recorder::LINKS_DIR.'/Index.csv')) { |
|
142
|
|
|
file_put_contents($this->folder.Recorder::LINKS_DIR.'/Index.csv', 'From,To'.PHP_EOL); |
|
143
|
|
|
} |
|
144
|
|
|
} |
|
145
|
|
|
|
|
146
|
|
|
public static function removeBase(string $base, string $url) |
|
147
|
|
|
{ |
|
148
|
|
|
return (0 === strpos($url, $base)) ? substr_replace($url, '', 0, strlen($base)) : null; |
|
149
|
|
|
} |
|
150
|
|
|
|
|
151
|
|
|
public function recordLinksIndex(string $base, Url $from, $urls, array $links) |
|
152
|
|
|
{ |
|
153
|
|
|
$everAdded = []; |
|
154
|
|
|
$content = ''; |
|
155
|
|
|
|
|
156
|
|
|
foreach ($links as $link) { |
|
157
|
|
|
$content .= $from->getId(); |
|
158
|
|
|
if (in_array($link->getUrl(), $everAdded)) { // like Google, we sould not add duplicate link, |
|
159
|
|
|
// so we say the juice is lost -1 |
|
160
|
|
|
$content .= ',-1'.PHP_EOL; |
|
161
|
|
|
} else { |
|
162
|
|
|
$everAdded[] = $link->getUrl(); |
|
163
|
|
|
$relative = self::removeBase($base, $link->getPageUrl()); |
|
164
|
|
|
$content .= ','.(isset($urls[$relative]) ? $urls[$relative]->getId() : 0).PHP_EOL; // 0 = external |
|
165
|
|
|
} |
|
166
|
|
|
} |
|
167
|
|
|
file_put_contents($this->folder.Recorder::LINKS_DIR.'/Index.csv', $content, FILE_APPEND); |
|
168
|
|
|
} |
|
169
|
|
|
} |
|
170
|
|
|
|
PiedWeb /
SeoPocketCrawler
