|
1
|
|
|
<?php |
|
2
|
|
|
/* |
|
3
|
|
|
* This file is part of dispositif/wikibot application (@github) |
|
4
|
|
|
* 2019-2023 © Philippe M./Irønie <[email protected]> |
|
5
|
|
|
* For the full copyright and MIT license information, view the license file. |
|
6
|
|
|
*/ |
|
7
|
|
|
|
|
8
|
|
|
declare(strict_types=1); |
|
9
|
|
|
|
|
10
|
|
|
|
|
11
|
|
|
namespace App\Domain\Publisher; |
|
12
|
|
|
|
|
13
|
|
|
|
|
14
|
|
|
use Exception; |
|
15
|
|
|
|
|
16
|
|
|
/** |
|
17
|
|
|
* Parsing/mapping Open Graph and Dublin Core meta data, and HTML meta tags |
|
18
|
|
|
* Currently only used by ExternMapper (mixed with JSON-LD mapping) for ExternRefWorker. |
|
19
|
|
|
*/ |
|
20
|
|
|
class OpenGraphMapper implements MapperInterface |
|
21
|
|
|
{ |
|
22
|
|
|
use ExternConverterTrait; |
|
23
|
|
|
|
|
24
|
|
|
/** |
|
25
|
|
|
* Allowing use of HTML <title> or <h1> to predict web page title ? |
|
26
|
|
|
*/ |
|
27
|
|
|
protected $htmlTitleAllowed = true; |
|
28
|
|
|
|
|
29
|
|
|
protected $titleFromHtmlState = false; |
|
30
|
|
|
|
|
31
|
5 |
|
/** |
|
32
|
|
|
* @param array|null $options |
|
33
|
|
|
*/ |
|
34
|
5 |
|
public function __construct(?array $options = []) |
|
35
|
5 |
|
{ |
|
36
|
5 |
|
if (!empty($options['htmlTitleAllowed']) && is_bool($options['htmlTitleAllowed'])) { |
|
37
|
5 |
|
$this->htmlTitleAllowed = $options['htmlTitleAllowed']; |
|
38
|
5 |
|
} |
|
39
|
5 |
|
} |
|
40
|
5 |
|
|
|
41
|
|
|
/** |
|
42
|
5 |
|
* Mapping from Open Graph and Dublin Core meta tags |
|
43
|
5 |
|
* https://ogp.me/ |
|
44
|
5 |
|
* https://www.dublincore.org/schemas/ |
|
45
|
|
|
* todo extract DC ? |
|
46
|
5 |
|
* |
|
47
|
5 |
|
* @param array $meta |
|
48
|
5 |
|
* |
|
49
|
5 |
|
* @return array |
|
50
|
|
|
* @throws Exception |
|
51
|
5 |
|
*/ |
|
52
|
|
|
public function process($meta): array |
|
53
|
|
|
{ |
|
54
|
5 |
|
$seoSanitizer = new SeoSanitizer(); |
|
55
|
5 |
|
|
|
56
|
5 |
|
return [ |
|
57
|
5 |
|
'DATA-TYPE' => 'Open Graph/Dublin Core', |
|
58
|
|
|
'DATA-ARTICLE' => $this->isAnArticle($meta['og:type'] ?? ''), |
|
59
|
5 |
|
'site' => $this->clean($meta['og:site_name'] ?? null), |
|
60
|
5 |
|
'titre' => $seoSanitizer->cleanSEOTitle( |
|
61
|
5 |
|
$meta['prettyDomainName'], |
|
62
|
5 |
|
$this->predictBestTitle($meta) |
|
63
|
|
|
), |
|
64
|
|
|
'url' => $meta['og:url'] ?? $meta['URL'] ?? $meta['html-url'] ?? null, |
|
65
|
|
|
'langue' => $this->convertLangue( |
|
66
|
5 |
|
$meta['og:locale'] ?? $meta['DC.language'] ?? $meta['citation_language'] ?? $meta['lang'] ?? |
|
67
|
5 |
|
$meta['language'] ?? |
|
68
|
5 |
|
$meta['content-language'] ?? $meta['Content-Language'] ?? $meta['html-lang'] ?? null |
|
69
|
5 |
|
), |
|
70
|
5 |
|
'consulté le' => date('d-m-Y'), |
|
71
|
5 |
|
'auteur' => $this->cleanAuthor( |
|
72
|
5 |
|
$meta['og:article:author'] ?? |
|
73
|
5 |
|
$meta['article:author'] ?? $meta['citation_author'] ?? $meta['article:author_name'] ?? null |
|
74
|
|
|
), |
|
75
|
|
|
'format' => $this->convertOGtype2format($meta['og:type'] ?? null), |
|
76
|
|
|
'date' => $this->convertDate( |
|
77
|
|
|
$meta['og:article:published_time'] ?? $meta['article:published_time'] ?? |
|
78
|
|
|
$meta['DC.date'] ?? $meta['citation_date'] ?? $meta['citation_publication_date'] ?? null |
|
79
|
|
|
), |
|
80
|
|
|
'accès url' => $this->convertURLaccess($meta), |
|
81
|
|
|
|
|
82
|
|
|
// DUBLIN CORE ONLY |
|
83
|
|
|
'périodique' => $this->clean($meta['DC.isPartOf'] ?? $meta['citation_journal_title'] ?? null), |
|
84
|
|
|
'et al.' => $this->authorsEtAl( |
|
85
|
|
|
$meta['citation_authors'] ?? $meta['DC.Contributor'] ?? null, |
|
86
|
|
|
true |
|
87
|
|
|
), |
|
88
|
|
|
'auteur1' => $this->wikifyPressAgency( |
|
89
|
|
|
$this->cleanAuthor( |
|
90
|
|
|
$this->authorsEtAl($meta['citation_authors'] ?? $meta['DC.Contributor'] ?? $meta['Author'] ?? null) |
|
91
|
|
|
) |
|
92
|
|
|
), |
|
93
|
|
|
'volume' => $meta['citation_volume'] ?? null, |
|
94
|
|
|
'numéro' => $meta['citation_issue'] ?? null, |
|
95
|
|
|
'page' => $this->convertDCpage($meta), |
|
96
|
|
|
'doi' => $meta['citation_doi'] ?? $meta['DOI'] ?? null, |
|
97
|
|
|
'éditeur' => $meta['DC.publisher'] ?? $meta['dc.publisher'] ?? null, // Persée dégeulasse todo? |
|
98
|
|
|
'pmid' => $meta['citation_pmid'] ?? null, |
|
99
|
|
|
'issn' => $meta["citation_issn"] ?? null, |
|
100
|
|
|
'isbn' => $meta["citation_isbn"] ?? null, |
|
101
|
|
|
// "prism.eIssn" => "2262-7197" |
|
102
|
|
|
]; |
|
103
|
|
|
} |
|
104
|
|
|
|
|
105
|
|
|
public function isTitleFromHtmlState(): bool |
|
106
|
|
|
{ |
|
107
|
|
|
return $this->titleFromHtmlState; |
|
108
|
|
|
} |
|
109
|
|
|
|
|
110
|
|
|
/** |
|
111
|
|
|
* Todo extraire cette logique to ExternConverterTrait or ExternPageTitlePredictor ? |
|
112
|
|
|
*/ |
|
113
|
|
|
private function predictBestTitle(array $meta): ?string |
|
114
|
|
|
{ |
|
115
|
|
|
// Mode "pas de titre html" |
|
116
|
|
|
if (!$this->htmlTitleAllowed) { |
|
117
|
|
|
return $this->getBestTitleFromMetadata($meta); |
|
118
|
|
|
} |
|
119
|
|
|
|
|
120
|
|
|
if (null === $this->getBestTitleFromMetadata($meta) |
|
121
|
|
|
&& !empty($meta['html-title']) |
|
122
|
|
|
) { |
|
123
|
|
|
$this->titleFromHtmlState = true; |
|
124
|
|
|
} |
|
125
|
|
|
|
|
126
|
|
|
// Responsibility ?!! sanitize title here conflicts with ExternMapper:postprocess() |
|
127
|
|
|
return $this->chooseBestTitle( |
|
128
|
|
|
$this->getBestTitleFromMetadata($meta), |
|
129
|
|
|
$meta['html-title'], |
|
130
|
|
|
$meta['html-h1'] |
|
131
|
|
|
); |
|
132
|
|
|
} |
|
133
|
|
|
|
|
134
|
|
|
/** |
|
135
|
|
|
* Choose page's title from OpenGrap or Dublin core. |
|
136
|
|
|
*/ |
|
137
|
|
|
private function getBestTitleFromMetadata(array $meta): ?string |
|
138
|
|
|
{ |
|
139
|
|
|
if (!empty($meta['og:title'])) { |
|
140
|
|
|
return $meta['og:title']; |
|
141
|
|
|
} |
|
142
|
|
|
if (!empty($meta['twitter:title'])) { |
|
143
|
|
|
return $meta['twitter:title']; |
|
144
|
|
|
} |
|
145
|
|
|
if (!empty($meta['DC.title'])) { |
|
146
|
|
|
return $meta['DC.title']; |
|
147
|
|
|
} |
|
148
|
|
|
|
|
149
|
|
|
return null; |
|
150
|
|
|
} |
|
151
|
|
|
|
|
152
|
|
|
/** |
|
153
|
|
|
* Choose best title from meta-title, html-title and html-h1. |
|
154
|
|
|
* Title is sanitized in ExternMapper::postprocess() |
|
155
|
|
|
*/ |
|
156
|
|
|
public function chooseBestTitle(?string $metaTitle, ?string $htmlTitle, ?string $htmlH1): ?string |
|
157
|
|
|
{ |
|
158
|
|
|
// clean all titles |
|
159
|
|
|
$metaTitle = $this->clean($metaTitle); |
|
160
|
|
|
$htmlTitle = $this->clean($htmlTitle); |
|
161
|
|
|
$htmlH1 = $this->clean($htmlH1); |
|
162
|
|
|
|
|
163
|
|
|
// check if htmlh1 included in htmltitle, if yes use htmlh1 |
|
164
|
|
|
if (!empty($metaTitle)) { |
|
165
|
|
|
return $metaTitle; |
|
166
|
|
|
} |
|
167
|
|
|
if (!empty($htmlH1) && !empty($htmlTitle) && strpos($htmlTitle, $htmlH1) !== false) { |
|
168
|
|
|
return $htmlH1; |
|
169
|
|
|
} |
|
170
|
|
|
|
|
171
|
|
|
return $htmlTitle ?? $htmlH1 ?? null; |
|
172
|
|
|
} |
|
173
|
|
|
} |
|
174
|
|
|
|
Dispositif /
Wikibot
