|
1
|
|
|
<?php |
|
2
|
|
|
|
|
3
|
|
|
namespace PiedWeb\UrlHarvester; |
|
4
|
|
|
|
|
5
|
|
|
use Spatie\Robots\RobotsHeaders; |
|
6
|
|
|
|
|
7
|
|
|
class Indexable |
|
8
|
|
|
{ |
|
9
|
|
|
// https://stackoverflow.com/questions/1880148/how-to-get-name-of-the-constant |
|
10
|
|
|
const INDEXABLE = 0; |
|
11
|
|
|
const NOT_INDEXABLE_ROBOTS = 1; |
|
12
|
|
|
const NOT_INDEXABLE_HEADER = 2; |
|
13
|
|
|
const NOT_INDEXABLE_META = 3; |
|
14
|
|
|
const NOT_INDEXABLE_CANONICAL = 4; |
|
15
|
|
|
const NOT_INDEXABLE_4XX = 5; |
|
16
|
|
|
const NOT_INDEXABLE_5XX = 6; |
|
17
|
|
|
const NOT_INDEXABLE_NETWORK_ERROR = 7; |
|
18
|
|
|
const NOT_INDEXABLE_3XX = 8; |
|
19
|
|
|
const NOT_INDEXABLE_NOT_HTML = 9; |
|
20
|
|
|
|
|
21
|
|
|
/** @var Harvest */ |
|
22
|
|
|
protected $harvest; |
|
23
|
|
|
|
|
24
|
|
|
/** @var string */ |
|
25
|
|
|
protected $isIndexableFor; |
|
26
|
|
|
|
|
27
|
9 |
|
public function __construct(Harvest $harvest, string $isIndexableFor = 'googlebot') |
|
28
|
|
|
{ |
|
29
|
9 |
|
$this->harvest = $harvest; |
|
30
|
9 |
|
$this->isIndexableFor = $isIndexableFor; |
|
31
|
9 |
|
} |
|
32
|
|
|
|
|
33
|
9 |
|
public function robotsTxtAllows() |
|
34
|
|
|
{ |
|
35
|
9 |
|
$url = $this->harvest->getResponse()->getUrl(); |
|
36
|
9 |
|
$robotsTxt = $this->harvest->getRobotsTxt(); |
|
37
|
|
|
|
|
38
|
9 |
|
return '' === $robotsTxt ? true : $robotsTxt->allows($url, $this->isIndexableFor); |
|
39
|
|
|
} |
|
40
|
|
|
|
|
41
|
9 |
|
public function metaAllows() |
|
42
|
|
|
{ |
|
43
|
9 |
|
$meta = $this->harvest->getMeta($this->isIndexableFor); |
|
44
|
9 |
|
$generic = $this->harvest->getMeta('robots'); |
|
45
|
|
|
|
|
46
|
9 |
|
return !(false !== stripos($meta, 'noindex') || false !== stripos($generic, 'noindex')); |
|
47
|
|
|
} |
|
48
|
|
|
|
|
49
|
9 |
|
public function headersAllow() |
|
50
|
|
|
{ |
|
51
|
9 |
|
$headers = explode(PHP_EOL, $this->harvest->getResponse()->getHeaders(false)); |
|
52
|
|
|
|
|
53
|
9 |
|
return RobotsHeaders::create($headers)->mayIndex($this->isIndexableFor); |
|
54
|
|
|
} |
|
55
|
|
|
|
|
56
|
|
|
/** |
|
57
|
|
|
* @return int |
|
58
|
|
|
*/ |
|
59
|
9 |
|
public static function isIndexable(Harvest $harvest, string $isIndexableFor = 'googlebot') |
|
60
|
|
|
{ |
|
61
|
9 |
|
$self = new self($harvest, $isIndexableFor); |
|
62
|
|
|
|
|
63
|
|
|
// robots |
|
64
|
9 |
|
if (!$self->robotsTxtAllows()) { |
|
65
|
3 |
|
return self::NOT_INDEXABLE_ROBOTS; |
|
66
|
|
|
} |
|
67
|
|
|
|
|
68
|
6 |
|
if (!$self->headersAllow()) { |
|
69
|
|
|
return self::NOT_INDEXABLE_HEADER; |
|
70
|
|
|
} |
|
71
|
|
|
|
|
72
|
6 |
|
if (!$self->metaAllows()) { |
|
73
|
|
|
return self::NOT_INDEXABLE_META; |
|
74
|
|
|
} |
|
75
|
|
|
|
|
76
|
|
|
// canonical |
|
77
|
6 |
|
if (!$harvest->isCanonicalCorrect()) { |
|
78
|
|
|
return self::NOT_INDEXABLE_CANONICAL; |
|
79
|
|
|
} |
|
80
|
|
|
|
|
81
|
6 |
|
$statusCode = $harvest->getResponse()->getStatusCode(); |
|
82
|
|
|
// status 4XX |
|
83
|
6 |
|
if ($statusCode < 500 && $statusCode > 399) { |
|
84
|
|
|
return self::NOT_INDEXABLE_5XX; |
|
85
|
|
|
} |
|
86
|
|
|
|
|
87
|
|
|
// status 5XX |
|
88
|
6 |
|
if ($statusCode < 600 && $statusCode > 499) { |
|
89
|
|
|
return self::NOT_INDEXABLE_5XX; |
|
90
|
|
|
} |
|
91
|
|
|
|
|
92
|
|
|
// status 3XX |
|
93
|
6 |
|
if ($statusCode < 400 && $statusCode > 299) { |
|
94
|
3 |
|
return self::NOT_INDEXABLE_3XX; |
|
95
|
|
|
} |
|
96
|
|
|
|
|
97
|
3 |
|
return self::INDEXABLE; |
|
98
|
|
|
} |
|
99
|
|
|
} |
|
100
|
|
|
|
PiedWeb /
UrlHarvester
