|
1
|
|
|
<?php |
|
2
|
|
|
/** |
|
3
|
|
|
* This file is part of dispositif/wikibot application (@github) |
|
4
|
|
|
* 2019/2020 © Philippe M. <[email protected]> |
|
5
|
|
|
* For the full copyright and MIT license information, please view the license file. |
|
6
|
|
|
*/ |
|
7
|
|
|
|
|
8
|
|
|
declare(strict_types=1); |
|
9
|
|
|
|
|
10
|
|
|
|
|
11
|
|
|
namespace App\Domain; |
|
12
|
|
|
|
|
13
|
|
|
use App\Application\Http\ExternHttpClient; |
|
14
|
|
|
use App\Infrastructure\TagParser; |
|
15
|
|
|
use Exception; |
|
16
|
|
|
use Psr\Log\LoggerInterface; |
|
17
|
|
|
|
|
18
|
|
|
/** |
|
19
|
|
|
* Représente une page web d'un Lien Externe (hors wiki) |
|
20
|
|
|
* Class ExternPage |
|
21
|
|
|
* |
|
22
|
|
|
* @package App\Domain |
|
23
|
|
|
*/ |
|
24
|
|
|
class ExternPage |
|
25
|
|
|
{ |
|
26
|
|
|
/** |
|
27
|
|
|
* @var string |
|
28
|
|
|
*/ |
|
29
|
|
|
private $url; |
|
30
|
|
|
|
|
31
|
|
|
/** |
|
32
|
|
|
* @var string |
|
33
|
|
|
*/ |
|
34
|
|
|
private $html; |
|
35
|
|
|
|
|
36
|
|
|
/** |
|
37
|
|
|
* @var LoggerInterface|null |
|
38
|
|
|
*/ |
|
39
|
|
|
private $log; |
|
40
|
|
|
|
|
41
|
|
|
/** |
|
42
|
|
|
* ExternPage constructor. |
|
43
|
|
|
* |
|
44
|
|
|
* @param string $url |
|
45
|
|
|
* @param string $html |
|
46
|
|
|
* @param LoggerInterface|null $log |
|
47
|
|
|
* |
|
48
|
|
|
* @throws Exception |
|
49
|
|
|
*/ |
|
50
|
5 |
|
public function __construct(string $url, string $html, ?LoggerInterface $log = null) |
|
51
|
|
|
{ |
|
52
|
5 |
|
if (!ExternHttpClient::isWebURL($url)) { |
|
53
|
|
|
throw new Exception('string is not an URL '.$url); |
|
54
|
|
|
} |
|
55
|
5 |
|
$this->url = $url; |
|
56
|
5 |
|
$this->html = $html; |
|
57
|
5 |
|
$this->log = $log; |
|
58
|
5 |
|
} |
|
59
|
|
|
|
|
60
|
|
|
/** |
|
61
|
|
|
* @return string |
|
62
|
|
|
*/ |
|
63
|
|
|
public function getUrl(): string |
|
64
|
|
|
{ |
|
65
|
|
|
return $this->url; |
|
66
|
|
|
} |
|
67
|
|
|
|
|
68
|
|
|
/** |
|
69
|
|
|
* @return array |
|
70
|
|
|
* @throws Exception |
|
71
|
|
|
*/ |
|
72
|
5 |
|
public function getData(): array |
|
73
|
|
|
{ |
|
74
|
5 |
|
$ld = $this->parseLdJson($this->html); |
|
75
|
5 |
|
$meta = $this->parseMetaTags($this->html); |
|
76
|
|
|
|
|
77
|
5 |
|
$meta['html-lang'] = $this->parseHtmlLang($this->html); // <html lang="en"> |
|
78
|
|
|
|
|
79
|
|
|
return ['JSON-LD' => $ld, 'meta' => $meta]; |
|
80
|
|
|
} |
|
81
|
|
|
|
|
82
|
|
|
/** |
|
83
|
|
|
* extract LD-JSON metadata from <script type="application/ld+json">. |
|
84
|
|
|
* |
|
85
|
|
|
* @param string $html |
|
86
|
|
|
* |
|
87
|
|
|
* @return array |
|
88
|
|
|
* @throws Exception |
|
89
|
5 |
|
* @throws Exception |
|
90
|
|
|
*/ |
|
91
|
5 |
|
private function parseLdJson(string $html): array |
|
92
|
5 |
|
{ |
|
93
|
5 |
|
$parser = new TagParser(); |
|
94
|
|
|
$results = $parser->importHtml($html)->xpathResults( |
|
95
|
|
|
'//script[@type="application/ld+json"]' |
|
96
|
5 |
|
); |
|
97
|
3 |
|
|
|
98
|
|
|
foreach ($results as $result) { |
|
99
|
3 |
|
$json = trim($result); |
|
100
|
|
|
// filtrage empty value (todo?) |
|
101
|
|
|
if (0 === strlen($json)) { |
|
102
|
3 |
|
continue; |
|
103
|
3 |
|
} |
|
104
|
|
|
$data = json_decode($json, true); |
|
105
|
|
|
if (!is_array($data) |
|
106
|
|
|
|| (isset($data['@type']) && is_string($data['@type']) && preg_match('#Breadcrumb#i', $data['@type'])) |
|
107
|
3 |
|
) { |
|
108
|
|
|
continue; |
|
109
|
|
|
} |
|
110
|
2 |
|
|
|
111
|
|
|
return $data; |
|
112
|
|
|
} |
|
113
|
|
|
|
|
114
|
|
|
return []; |
|
115
|
|
|
} |
|
116
|
|
|
|
|
117
|
|
|
/** |
|
118
|
|
|
* todo move? /refac/delete? |
|
119
|
|
|
* |
|
120
|
5 |
|
* @param string $str |
|
121
|
|
|
* |
|
122
|
5 |
|
* @return array |
|
123
|
|
|
*/ |
|
124
|
|
|
private function parseMetaTags(string $str): array |
|
125
|
|
|
{ |
|
126
|
|
|
$pattern = ' |
|
127
|
|
|
~<\s*meta\s |
|
128
|
|
|
# using lookahead to capture type to $1 |
|
129
|
|
|
(?=[^>]*? |
|
130
|
|
|
\b(?:name|property|http-equiv)\s*=\s* |
|
131
|
|
|
(?|"\s*([^"]*?)\s*"|\'\s*([^\']*?)\s*\'| |
|
132
|
|
|
([^"\'>]*?)(?=\s*/?\s*>|\s\w+\s*=)) |
|
133
|
|
|
) |
|
134
|
|
|
# capture content to $2 |
|
135
|
|
|
[^>]*?\bcontent\s*=\s* |
|
136
|
|
|
(?|"\s*([^"]*?)\s*"|\'\s*([^\']*?)\s*\'| |
|
137
|
5 |
|
([^"\'>]*?)(?=\s*/?\s*>|\s\w+\s*=)) |
|
138
|
5 |
|
[^>]*> |
|
139
|
|
|
~ix'; |
|
140
|
5 |
|
|
|
141
|
|
|
if (preg_match_all($pattern, $str, $out)) { |
|
142
|
|
|
$combine = array_combine($out[1], $out[2]); |
|
143
|
|
|
|
|
144
|
|
|
return $combine ? $combine : []; |
|
145
|
|
|
} |
|
146
|
|
|
|
|
147
|
|
|
return []; |
|
148
|
|
|
} |
|
149
|
|
|
|
|
150
|
|
|
/** |
|
151
|
|
|
* test.com => test.com |
|
152
|
|
|
* bla.test.com => test.com |
|
153
|
|
|
* test.co.uk => test.co.uk (national commercial subdomain) |
|
154
|
|
|
* site.google.com => site.google.com (blog) |
|
155
|
|
|
* |
|
156
|
|
|
* @return string |
|
157
|
|
|
* @throws Exception |
|
158
|
|
|
*/ |
|
159
|
|
|
public function getPrettyDomainName(): string |
|
160
|
|
|
{ |
|
161
|
|
|
$subDomain = $this->getSubDomain(); |
|
162
|
|
|
|
|
163
|
|
|
if (!strpos($subDomain, '.co.uk') && !strpos($subDomain, '.co.ma') && !strpos($subDomain, '.co.kr') |
|
164
|
|
|
&& !strpos($subDomain, 'site.google.') |
|
165
|
|
|
) { |
|
166
|
|
|
// bla.test.com => Test.com |
|
167
|
|
|
if (preg_match('#\w+\.\w+$#', $subDomain, $matches)) { |
|
168
|
|
|
return $matches[0]; |
|
169
|
|
|
} |
|
170
|
|
|
} |
|
171
|
|
|
|
|
172
|
|
|
return $subDomain; |
|
173
|
|
|
} |
|
174
|
|
|
|
|
175
|
|
|
/** |
|
176
|
|
|
* @return string|null |
|
177
|
|
|
* @throws Exception |
|
178
|
|
|
*/ |
|
179
|
|
|
public function getSubDomain(): string |
|
180
|
|
|
{ |
|
181
|
|
|
try { |
|
182
|
|
|
return ExternDomains::extractSubDomain($this->url); |
|
183
|
|
|
} catch (Exception $e) { |
|
184
|
|
|
if ($this->log) { |
|
185
|
|
|
$this->log->warning('ExternDomains::extractSubDomain NULL '.$this->url); |
|
186
|
|
|
} |
|
187
|
|
|
throw new Exception('ExternDomains::extractSubDomain NULL'); |
|
188
|
|
|
} |
|
189
|
|
|
} |
|
190
|
|
|
|
|
191
|
|
|
/** |
|
192
|
|
|
* Extract language from <html lang="en-us"> tag. |
|
193
|
|
|
* |
|
194
|
|
|
* @param string $html |
|
195
|
|
|
*/ |
|
196
|
|
|
private function parseHtmlLang(string $html): ?string |
|
197
|
|
|
{ |
|
198
|
|
|
if (preg_match('#<html(?: [^>]+)? lang="([A-Z-]{2,15})"(?: [^>]+)?>#i', $html, $matches)) { |
|
199
|
|
|
// 'en-us' => 'en' // todo move in Language |
|
200
|
|
|
$lang = preg_replace('#^([a-z]+)-[a-z]+$#i', '$1', $matches[1]); |
|
201
|
|
|
|
|
202
|
|
|
return $lang; |
|
203
|
|
|
} |
|
204
|
|
|
|
|
205
|
|
|
return null; |
|
206
|
|
|
} |
|
207
|
|
|
} |
|
208
|
|
|
|
Dispositif /
Wikibot
