|
1
|
|
|
<?php |
|
2
|
|
|
|
|
3
|
|
|
/** |
|
4
|
|
|
* Entity. |
|
5
|
|
|
*/ |
|
6
|
|
|
|
|
7
|
|
|
namespace PiedWeb\UrlHarvester; |
|
8
|
|
|
|
|
9
|
|
|
use Symfony\Component\DomCrawler\Crawler as DomCrawler; |
|
10
|
|
|
|
|
11
|
|
|
class Link |
|
12
|
|
|
{ |
|
13
|
|
|
/** @var Url */ |
|
14
|
|
|
protected $url; |
|
15
|
|
|
|
|
16
|
|
|
/** @var string cache */ |
|
17
|
|
|
protected $anchor; |
|
18
|
|
|
|
|
19
|
|
|
/** @var \DomElement */ |
|
20
|
|
|
protected $element; |
|
21
|
|
|
|
|
22
|
|
|
/** @var Harvest */ |
|
23
|
|
|
protected $parentDoc; |
|
24
|
|
|
|
|
25
|
|
|
/** @var int */ |
|
26
|
|
|
protected $wrapper; |
|
27
|
|
|
|
|
28
|
|
|
/** @var int */ |
|
29
|
|
|
protected $type; |
|
30
|
|
|
|
|
31
|
|
|
// Ce serait dans une liste, dans une phrase... |
|
32
|
|
|
protected $context; |
|
33
|
|
|
|
|
34
|
|
|
// wrapper related |
|
35
|
|
|
const LINK_A = 1; |
|
36
|
|
|
const LINK_SRC = 4; |
|
37
|
|
|
const LINK_3XX = 2; |
|
38
|
|
|
const LINK_301 = 3; |
|
39
|
|
|
|
|
40
|
|
|
// type related |
|
41
|
|
|
const LINK_SELF = 1; |
|
42
|
|
|
const LINK_INTERNAL = 2; |
|
43
|
|
|
const LINK_SUB = 3; |
|
44
|
|
|
const LINK_EXTERNAL = 4; |
|
45
|
|
|
|
|
46
|
|
|
/** |
|
47
|
|
|
* Add trailing slash for domain. Eg: https://piedweb.com => https://piedweb.com/ and '/test ' = '/test'. |
|
48
|
|
|
*/ |
|
49
|
36 |
|
public static function normalizeUrl(string $url): string |
|
50
|
|
|
{ |
|
51
|
36 |
|
$url = trim($url); |
|
52
|
|
|
|
|
53
|
36 |
|
if ('' == preg_replace('@(.*\://?([^\/]+))@', '', $url)) { |
|
54
|
12 |
|
$url .= '/'; |
|
55
|
|
|
} |
|
56
|
|
|
|
|
57
|
36 |
|
return $url; |
|
58
|
|
|
} |
|
59
|
|
|
|
|
60
|
18 |
|
protected static function getWrapperFromElement(\DomElement $element): ?int |
|
61
|
|
|
{ |
|
62
|
18 |
|
if ('a' == $element->tagName && $element->getAttribute('href')) { |
|
63
|
18 |
|
return self::LINK_A; |
|
64
|
|
|
} |
|
65
|
|
|
|
|
66
|
6 |
|
if ($element->getAttribute('src')) { |
|
67
|
6 |
|
return self::LINK_SRC; |
|
68
|
|
|
} |
|
69
|
|
|
|
|
70
|
6 |
|
return null; |
|
71
|
|
|
} |
|
72
|
|
|
|
|
73
|
|
|
/** |
|
74
|
|
|
* Always submit absoute Url ! |
|
75
|
|
|
*/ |
|
76
|
24 |
|
public function __construct(string $url, Harvest $parent, \DOMElement $element = null, int $wrapper = null) |
|
77
|
|
|
{ |
|
78
|
24 |
|
$this->url = new Url(self::normalizeUrl($url)); |
|
79
|
|
|
|
|
80
|
24 |
|
$this->parentDoc = $parent; |
|
81
|
|
|
|
|
82
|
24 |
|
if (null !== $element) { |
|
83
|
18 |
|
$this->setAnchor($element); |
|
84
|
|
|
} |
|
85
|
|
|
|
|
86
|
24 |
|
$this->element = $element; |
|
87
|
|
|
|
|
88
|
24 |
|
$this->wrapper = $wrapper ?? (null !== $element ? self::getWrapperFromElement($element) : null); |
|
89
|
24 |
|
} |
|
90
|
|
|
|
|
91
|
3 |
|
public static function createRedirection(string $url, Harvest $parent, int $redirType = null): self |
|
92
|
|
|
{ |
|
93
|
3 |
|
return new self($url, $parent, null, $redirType ?? self::LINK_3XX); |
|
94
|
|
|
} |
|
95
|
|
|
|
|
96
|
|
|
public function getWrapper(): ?int |
|
97
|
|
|
{ |
|
98
|
|
|
return $this->wrapper; |
|
99
|
|
|
} |
|
100
|
|
|
|
|
101
|
18 |
|
protected function setAnchor(\DomElement $element) |
|
102
|
|
|
{ |
|
103
|
|
|
// Get classic text anchor |
|
104
|
18 |
|
$this->anchor = $element->textContent; |
|
105
|
|
|
|
|
106
|
|
|
// If get nothing, then maybe we can get an alternative text (eg: img) |
|
107
|
18 |
|
if (empty($this->anchor)) { |
|
108
|
6 |
|
$alt = (new DomCrawler($element))->filter('*[alt]'); |
|
109
|
6 |
|
if ($alt->count() > 0) { |
|
110
|
|
|
$this->anchor = $alt->eq(0)->attr('alt') ?? ''; |
|
111
|
|
|
} |
|
112
|
|
|
} |
|
113
|
|
|
|
|
114
|
|
|
// Limit to 100 characters |
|
115
|
|
|
// Totally subjective |
|
116
|
18 |
|
$this->anchor = substr(Helper::clean($this->anchor), 0, 99); |
|
117
|
|
|
|
|
118
|
18 |
|
return $this; |
|
119
|
|
|
} |
|
120
|
|
|
|
|
121
|
6 |
|
public function getUrl(): Url |
|
122
|
|
|
{ |
|
123
|
6 |
|
return $this->url; |
|
124
|
|
|
} |
|
125
|
|
|
|
|
126
|
3 |
|
public function getPageUrl(): string |
|
127
|
|
|
{ |
|
128
|
3 |
|
return $this->url->getDocumentUrl(); //return preg_replace('/(\#.*)/si', '', $this->url->get()); |
|
129
|
|
|
} |
|
130
|
|
|
|
|
131
|
9 |
|
public function getParentUrl(): Url |
|
132
|
|
|
{ |
|
133
|
9 |
|
return $this->parentDoc->getUrl(); |
|
134
|
|
|
} |
|
135
|
|
|
|
|
136
|
9 |
|
public function getAnchor() |
|
137
|
|
|
{ |
|
138
|
9 |
|
return $this->anchor; |
|
139
|
|
|
} |
|
140
|
|
|
|
|
141
|
3 |
|
public function getElement() |
|
142
|
|
|
{ |
|
143
|
3 |
|
return $this->element; |
|
144
|
|
|
} |
|
145
|
|
|
|
|
146
|
|
|
/** |
|
147
|
|
|
* @return bool |
|
148
|
|
|
*/ |
|
149
|
9 |
|
public function mayFollow() |
|
150
|
|
|
{ |
|
151
|
|
|
// check meta robots and headers |
|
152
|
9 |
|
if (null !== $this->parentDoc && !$this->parentDoc->mayFollow()) { |
|
153
|
|
|
return false; |
|
154
|
|
|
} |
|
155
|
|
|
|
|
156
|
|
|
// check "wrapper" rel |
|
157
|
9 |
|
if (null !== $this->element && null !== $this->element->getAttribute('rel')) { |
|
158
|
6 |
|
if (preg_match('(nofollow|sponsored|ugc)', $this->element->getAttribute('rel'))) { |
|
159
|
3 |
|
return false; |
|
160
|
|
|
} |
|
161
|
|
|
} |
|
162
|
|
|
|
|
163
|
6 |
|
return true; |
|
164
|
|
|
} |
|
165
|
|
|
|
|
166
|
|
|
/** |
|
167
|
|
|
* @return string |
|
168
|
|
|
*/ |
|
169
|
|
|
public function getRelAttribute(): ?string |
|
170
|
|
|
{ |
|
171
|
|
|
return null !== $this->element ? $this->element->getAttribute('rel') : null; |
|
172
|
|
|
} |
|
173
|
|
|
|
|
174
|
9 |
|
public function isInternalLink(): bool |
|
175
|
|
|
{ |
|
176
|
9 |
|
return $this->url->getOrigin() == $this->getParentUrl()->getOrigin(); |
|
177
|
|
|
} |
|
178
|
|
|
|
|
179
|
9 |
|
public function isSubLink(): bool |
|
180
|
|
|
{ |
|
181
|
9 |
|
return !$this->isInternalLink() |
|
182
|
9 |
|
&& $this->url->getRegistrableDomain() == $this->getParentUrl()->getRegistrableDomain(); |
|
183
|
|
|
//&& strtolower(substr($this->getHost(), -strlen($this->parentDomain))) === $this->parentDomain; |
|
184
|
|
|
} |
|
185
|
|
|
|
|
186
|
9 |
|
public function isSelfLink(): bool |
|
187
|
|
|
{ |
|
188
|
9 |
|
return $this->isInternalLink() |
|
189
|
9 |
|
&& $this->url->getDocumentUrl() == $this->getParentUrl()->getDocumentUrl(); |
|
190
|
|
|
} |
|
191
|
|
|
|
|
192
|
6 |
|
public function getType() |
|
193
|
|
|
{ |
|
194
|
6 |
|
if ($this->isSelfLink()) { |
|
195
|
6 |
|
return self::LINK_SELF; |
|
196
|
|
|
} |
|
197
|
|
|
|
|
198
|
6 |
|
if ($this->isInternalLink()) { |
|
199
|
6 |
|
return self::LINK_INTERNAL; |
|
200
|
|
|
} |
|
201
|
|
|
|
|
202
|
6 |
|
if ($this->isSubLink()) { |
|
203
|
3 |
|
return self::LINK_SUB; |
|
204
|
|
|
} |
|
205
|
|
|
|
|
206
|
6 |
|
return self::LINK_EXTERNAL; |
|
207
|
|
|
} |
|
208
|
|
|
|
|
209
|
|
|
// todo useless ?! |
|
210
|
|
|
public function getAbsoluteInternalLink() |
|
211
|
|
|
{ |
|
212
|
|
|
if ($this->isInternalLink()) { |
|
213
|
|
|
return substr($this->url, strlen($this->getParentUrl()->getOrigin())); |
|
214
|
|
|
} |
|
215
|
|
|
} |
|
216
|
|
|
} |
|
217
|
|
|
|
PiedWeb /
UrlHarvester
