|
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
|
|
|
namespace App\Domain\Publisher; |
|
11
|
|
|
|
|
12
|
|
|
use App\Domain\Utils\ArrayProcessTrait; |
|
13
|
|
|
use App\Domain\Utils\TextUtil; |
|
14
|
|
|
use Psr\Log\LoggerInterface; |
|
15
|
|
|
|
|
16
|
|
|
/** |
|
17
|
|
|
* Generic mapper for press/revue article on web. |
|
18
|
|
|
* Using JSON-LD and meta tags to obtain {article} data. |
|
19
|
|
|
* Generic mapper for web pages URL to wiki-template references. |
|
20
|
|
|
* Converting to {article}, {lien web} or {lien brisé} |
|
21
|
|
|
* Using JSON-LD, Open Graph and Dublin Core meta extracted from HTML. |
|
22
|
|
|
*/ |
|
23
|
|
|
class ExternMapper implements MapperInterface |
|
24
|
|
|
{ |
|
25
|
|
|
use ArrayProcessTrait; |
|
26
|
|
|
|
|
27
|
|
|
// Added if title extracted from HTML <title> or <h1> |
|
28
|
|
|
final public const TITLE_TO_VERIFY_COMMENT = '<!-- Vérifiez ce titre -->'; |
|
29
|
|
|
// if title extracted from json-ld, or anything else |
|
30
|
|
|
final public const TITLE_VERY_MAX_LENGTH = 150; |
|
31
|
|
|
// if title extracted from HTML <title> or <h1> is too long, it's probably SEO |
|
32
|
|
|
final public const TITLE_HTML_MAX_LENGTH = 80; |
|
33
|
|
|
// if title contains too many all-caps words, it's probably SEO |
|
34
|
5 |
|
final public const TITLE_MAX_ALLCAPS = 2; |
|
35
|
|
|
// if site name extracted for meta data is too long, it's probably SEO |
|
36
|
5 |
|
final public const SITE_MAX_LENGTH = 40; |
|
37
|
5 |
|
final public const SITE_MAX_ALLCAPS = 1; |
|
38
|
|
|
private bool $titleFromHtmlState = false; |
|
39
|
5 |
|
|
|
40
|
|
|
/** |
|
41
|
5 |
|
* @param mixed[] $options |
|
42
|
|
|
*/ |
|
43
|
5 |
|
public function __construct(private readonly LoggerInterface $log, private readonly ?array $options = []) |
|
44
|
|
|
{ |
|
45
|
|
|
} |
|
46
|
5 |
|
|
|
47
|
|
|
public function process($data): array |
|
48
|
5 |
|
{ |
|
49
|
5 |
|
$parsedData = $this->processMapping($data); |
|
50
|
5 |
|
|
|
51
|
3 |
|
return ($parsedData === []) ? [] : $this->postProcess($parsedData); |
|
52
|
|
|
} |
|
53
|
5 |
|
|
|
54
|
5 |
|
protected function processMapping($data): array |
|
55
|
|
|
{ |
|
56
|
|
|
$mapJson = []; |
|
57
|
|
|
$mapMeta = []; |
|
58
|
5 |
|
$this->titleFromHtmlState = false; |
|
59
|
3 |
|
|
|
60
|
2 |
|
if (!empty($data['JSON-LD'])) { |
|
61
|
2 |
|
$mapJson = $this->processJsonLDMapping($data['JSON-LD']); |
|
62
|
|
|
} |
|
63
|
|
|
if (!empty($data['meta'])) { |
|
64
|
3 |
|
$openGraphMapper = new OpenGraphMapper($this->options); // todo inject/extract to reduce instanciationS ? |
|
65
|
|
|
$mapMeta = $openGraphMapper->process($data['meta']); |
|
66
|
|
|
$this->titleFromHtmlState = $openGraphMapper->isTitleFromHtmlState(); |
|
67
|
2 |
|
} |
|
68
|
|
|
|
|
69
|
|
|
// langue absente JSON-LD mais array_merge risqué (doublon) |
|
70
|
|
|
if ($mapJson !== []) { |
|
71
|
|
|
if (!isset($mapJson['langue']) && isset($mapMeta['langue'])) { |
|
72
|
|
|
$mapJson['langue'] = $mapMeta['langue']; |
|
73
|
|
|
$mapJson['DATA-TYPE'] = 'JSON-LD+META'; |
|
74
|
|
|
} |
|
75
|
|
|
// récupère "accès url" de OpenGraph (prévaut sur JSON:'isAccessibleForFree' |
|
76
|
|
|
if (isset($mapMeta['accès url'])) { |
|
77
|
3 |
|
$mapJson['accès url'] = $mapMeta['accès url']; |
|
78
|
|
|
$mapJson['DATA-TYPE'] = 'JSON-LD+META'; |
|
79
|
3 |
|
} |
|
80
|
2 |
|
|
|
81
|
|
|
return $mapJson; |
|
82
|
|
|
} |
|
83
|
1 |
|
|
|
84
|
1 |
|
return $mapMeta; |
|
85
|
1 |
|
} |
|
86
|
|
|
|
|
87
|
|
|
/** |
|
88
|
|
|
* todo move to mapper ? |
|
89
|
|
|
*/ |
|
90
|
|
|
private function processJsonLDMapping(array $LDdata): array |
|
91
|
|
|
{ |
|
92
|
3 |
|
if ($this->checkJSONLD($LDdata)) { |
|
93
|
|
|
return (new JsonLDMapper())->process($LDdata); |
|
94
|
3 |
|
} |
|
95
|
|
|
// gestion des multiples objets comme Figaro |
|
96
|
|
|
foreach ($LDdata as $dat) { |
|
97
|
|
|
if (is_array($dat) && $this->checkJSONLD($dat)) { |
|
98
|
|
|
return (new JsonLDMapper())->process($dat); |
|
99
|
|
|
} |
|
100
|
|
|
} |
|
101
|
|
|
|
|
102
|
|
|
return []; |
|
103
|
|
|
} |
|
104
|
|
|
|
|
105
|
5 |
|
protected function checkJSONLD(array $jsonLD): bool |
|
106
|
|
|
{ |
|
107
|
5 |
|
return isset($jsonLD['headline']) && isset($jsonLD['@type']); |
|
108
|
5 |
|
} |
|
109
|
3 |
|
|
|
110
|
|
|
/** |
|
111
|
|
|
* Data sanitization. |
|
112
|
|
|
* todo complexity/conditions |
|
113
|
5 |
|
* todo Config parameter for post-process |
|
114
|
|
|
*/ |
|
115
|
|
|
protected function postProcess(array $data): array |
|
116
|
|
|
{ |
|
117
|
5 |
|
$data = $this->deleteEmptyValueArray($data); |
|
118
|
|
|
if (isset($data['langue']) && 'fr' === $data['langue']) { |
|
119
|
|
|
unset($data['langue']); |
|
120
|
|
|
} |
|
121
|
|
|
|
|
122
|
|
|
// Ça m'énerve ! Gallica met "vidéo" pour livre numérisé |
|
123
|
|
|
if (isset($data['site']) && $data['site'] === 'Gallica') { |
|
124
|
|
|
unset($data['format']); |
|
125
|
|
|
} |
|
126
|
|
|
if (isset($data['site']) && TextUtil::countAllCapsWords($data['site']) > self::SITE_MAX_ALLCAPS) { |
|
127
|
|
|
$this->log->debug('lowercase site name'); |
|
128
|
|
|
$data['site'] = TextUtil::mb_ucfirst(mb_strtolower((string) $data['site'])); |
|
129
|
|
|
} |
|
130
|
|
|
// SEO : cut site name if too long if no domain.name and no wiki link |
|
131
|
|
|
if ( |
|
132
|
|
|
isset($data['site']) |
|
133
|
|
|
&& false === mb_strpos((string) $data['site'], '.') |
|
134
|
|
|
&& false === mb_strpos((string) $data['site'], '[[')) { |
|
135
|
|
|
$data['site'] = TextUtil::cutTextOnSpace($data['site'], self::SITE_MAX_LENGTH); |
|
136
|
|
|
} |
|
137
|
|
|
|
|
138
|
|
|
// lowercase title if too many ALLCAPS words |
|
139
|
|
|
if (isset($data['titre']) && TextUtil::countAllCapsWords($data['titre']) > self::TITLE_MAX_ALLCAPS) { |
|
140
|
|
|
$this->log->debug('lowercase title'); |
|
141
|
|
|
$data['titre'] = TextUtil::mb_ucfirst(mb_strtolower((string) $data['titre'])); |
|
142
|
|
|
} |
|
143
|
|
|
|
|
144
|
|
|
// title has 150 chars max, or is cut with "…" at the end |
|
145
|
|
|
if (isset($data['titre'])) { |
|
146
|
|
|
$data['titre'] = TextUtil::cutTextOnSpace($data['titre'], self::TITLE_VERY_MAX_LENGTH); |
|
147
|
|
|
$data['titre'] = $this->addVerifyCommentIfNecessary($data['titre']); |
|
148
|
|
|
} |
|
149
|
|
|
|
|
150
|
|
|
return $data; |
|
151
|
|
|
} |
|
152
|
|
|
|
|
153
|
|
|
/** |
|
154
|
|
|
* todo Créer un modèle {titre à vérifier} ? |
|
155
|
|
|
*/ |
|
156
|
|
|
private function addVerifyCommentIfNecessary(?string $title): ?string |
|
157
|
|
|
{ |
|
158
|
|
|
if ( |
|
159
|
|
|
!empty($title) |
|
160
|
|
|
&& mb_strlen($title) >= 30 |
|
161
|
|
|
&& $this->titleFromHtmlState |
|
162
|
|
|
) { |
|
163
|
|
|
/** @noinspection PhpRedundantOptionalArgumentInspection */ |
|
164
|
|
|
$title = TextUtil::cutTextOnSpace($title, self::TITLE_HTML_MAX_LENGTH); |
|
165
|
|
|
$title .= self::TITLE_TO_VERIFY_COMMENT; |
|
166
|
|
|
} |
|
167
|
|
|
|
|
168
|
|
|
return $title; |
|
169
|
|
|
} |
|
170
|
|
|
} |
|
171
|
|
|
|
Dispositif /
Wikibot
