|
1
|
|
|
<?php |
|
2
|
|
|
|
|
3
|
|
|
use DenisBeliaev\SitemapParser\Link; |
|
4
|
|
|
use DenisBeliaev\SitemapParser\Page; |
|
5
|
|
|
|
|
6
|
|
|
require_once __DIR__ . '/config.php'; |
|
7
|
|
|
|
|
8
|
|
|
$opts = getopt('s:p:t:'); |
|
9
|
|
|
|
|
10
|
|
|
$scope = $opts['s']; |
|
11
|
|
|
$pPID = $opts['p'] ?? null; |
|
12
|
|
|
$task = $opts['t'] ?? 'check'; |
|
13
|
|
|
|
|
14
|
|
|
$pdo = new PDO(DSN); |
|
15
|
|
|
|
|
16
|
|
|
$stm = $pdo->prepare('SELECT url FROM pages WHERE scope=:scope'); |
|
17
|
|
|
$stm->execute([':scope' => $scope]); |
|
18
|
|
|
$items = $stm->fetchAll(PDO::FETCH_COLUMN); |
|
19
|
|
|
unset($stm); |
|
20
|
|
|
|
|
21
|
|
|
$ch = curl_init(); |
|
22
|
|
|
if ($task == 'check') { |
|
23
|
|
|
curl_setopt($ch, CURLOPT_NOBODY, true); |
|
24
|
|
|
} else if ($task == 'parse-links') { |
|
25
|
|
|
curl_setopt($ch, CURLOPT_RETURNTRANSFER, true); |
|
26
|
|
|
} |
|
27
|
|
|
foreach ($items as $item) { |
|
28
|
|
|
curl_setopt($ch, CURLOPT_URL, $item); |
|
29
|
|
|
$httpCode = 'ERR'; |
|
30
|
|
|
if ($content = curl_exec($ch)) { |
|
31
|
|
|
$httpCode = curl_getinfo($ch, CURLINFO_HTTP_CODE); |
|
32
|
|
|
if ($httpCode == 501 && $task == 'check') { |
|
33
|
|
|
curl_setopt($ch, CURLOPT_NOBODY, false); |
|
34
|
|
|
if ($content = curl_exec($ch)) { |
|
35
|
|
|
$httpCode = curl_getinfo($ch, CURLINFO_HTTP_CODE); |
|
36
|
|
|
} |
|
37
|
|
|
curl_setopt($ch, CURLOPT_NOBODY, true); |
|
38
|
|
|
} |
|
39
|
|
|
} |
|
40
|
|
|
|
|
41
|
|
|
$pdo->setAttribute(PDO::ATTR_TIMEOUT, 100); |
|
42
|
|
|
$stm = $pdo->prepare('UPDATE pages SET status=:status WHERE url=:url'); |
|
43
|
|
|
$stm->execute([':status' => $httpCode, ':url' => $item]); |
|
44
|
|
|
unset($stm); |
|
45
|
|
|
|
|
46
|
|
|
if ($task == 'parse-links') { |
|
47
|
|
|
$Page = new Page($content, $item); |
|
48
|
|
|
$links = $Page->links; |
|
49
|
|
|
foreach ($links as &$link) { |
|
50
|
|
|
$link = Link::normalize($link, $Page->base); |
|
51
|
|
|
} |
|
52
|
|
|
$links = array_unique(array_filter($links)); |
|
53
|
|
|
|
|
54
|
|
|
foreach ($links as $key => &$link) { |
|
55
|
|
|
$link = [$item, $Page->links[$key], $link]; |
|
56
|
|
|
} |
|
57
|
|
|
$links = array_filter($links, function ($value) use ($pdo) { |
|
58
|
|
|
$stm = $pdo->prepare('SELECT COUNT(url) FROM pages WHERE url=:url'); |
|
59
|
|
|
$stm->execute([':url' => $value[2]]); |
|
60
|
|
|
$result = $stm->fetchColumn(); |
|
61
|
|
|
return $result == 0; |
|
62
|
|
|
}); |
|
63
|
|
|
if (!empty($links)) { |
|
64
|
|
|
$parameters = []; |
|
65
|
|
|
foreach ($links as $link) { |
|
66
|
|
|
$parameters[] = $link[0]; |
|
67
|
|
|
$parameters[] = $link[1]; |
|
68
|
|
|
$parameters[] = $link[2]; |
|
69
|
|
|
} |
|
70
|
|
|
|
|
71
|
|
|
$pdo->setAttribute(PDO::ATTR_TIMEOUT, 100); |
|
72
|
|
|
$stm = $pdo->prepare('INSERT OR IGNORE INTO links (`page`, link, url) VALUES ' . implode(',', array_pad([], count($links), '(?,?,?)'))); |
|
73
|
|
|
if ($stm == false) { |
|
74
|
|
|
fwrite(STDERR, $pdo->errorInfo()[2] . PHP_EOL); |
|
75
|
|
|
exit; |
|
76
|
|
|
} |
|
77
|
|
|
$stm->execute($parameters); |
|
78
|
|
|
unset($stm); |
|
79
|
|
|
} |
|
80
|
|
|
} |
|
81
|
|
|
|
|
82
|
|
|
if ($pPID && !isRunning($pPID)) { |
|
83
|
|
|
exit(1); |
|
84
|
|
|
} |
|
85
|
|
|
} |
|
86
|
|
|
|
|
87
|
|
|
unset($pdo); |
|
88
|
|
|
curl_close($ch); |
|
89
|
|
|
|
|
90
|
|
|
function isRunning($pid) |
|
91
|
|
|
{ |
|
92
|
|
|
if (function_exists('posix_kill')) { |
|
93
|
|
|
return posix_kill($pid, 0); |
|
94
|
|
|
} |
|
95
|
|
|
exec('ps -W -p ' . $pid, $out); |
|
96
|
|
|
return count($out) > 1; |
|
97
|
|
|
} |
|
98
|
|
|
|
cimmwolf /
sitemap-parser
