|
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\Infrastructure; |
|
11
|
|
|
|
|
12
|
|
|
use App\Application\InfrastructurePorts\PageListForAppInterface; |
|
13
|
|
|
use App\Domain\InfrastructurePorts\PageListInterface; |
|
14
|
|
|
use Exception; |
|
15
|
|
|
use GuzzleHttp\Client; |
|
16
|
|
|
|
|
17
|
|
|
/** |
|
18
|
|
|
* @unused |
|
19
|
|
|
* Data import from https://wstat.fr (frwiki daily dump parsing). |
|
20
|
|
|
* https://wstat.fr/template/index.php?title=Ouvrage&query=inclusions¶m=isbn&start=50000&limit=50&format=json |
|
21
|
|
|
* Class WstatImport. |
|
22
|
|
|
*/ |
|
23
|
|
|
class WstatImport implements PageListInterface, PageListForAppInterface |
|
24
|
|
|
{ |
|
25
|
|
|
final public const MAX_IMPORT = 50000; |
|
26
|
|
|
|
|
27
|
|
|
private array $params = []; |
|
28
|
|
|
|
|
29
|
|
|
private ?int $max = 100; |
|
30
|
|
|
|
|
31
|
|
|
public function __construct(private readonly Client $client, ?array $params = null, ?int $max = 500) |
|
32
|
|
|
{ |
|
33
|
|
|
$this->max = min(self::MAX_IMPORT, $max); |
|
34
|
|
|
|
|
35
|
|
|
//example |
|
36
|
|
|
// "nom de page" : https://wstat.fr/template/index.php?title=Ouvrage&query=inclusions-title&start=105000&limit=5000 |
|
37
|
|
|
// "modèle complet" : https://wstat.fr/template/index.php?title=Ouvrage&query=inclusions&start=105000&limit=5000 |
|
38
|
|
|
if (!$params) { |
|
39
|
|
|
$params = [ |
|
40
|
|
|
'title' => 'Ouvrage', |
|
41
|
|
|
'query' => 'inclusions-title', |
|
42
|
|
|
// 'param' => 'isbn', |
|
43
|
|
|
'start' => 50000, |
|
44
|
|
|
'limit' => 5000, |
|
45
|
|
|
]; |
|
46
|
|
|
} |
|
47
|
|
|
$this->params = $params; |
|
48
|
|
|
} |
|
49
|
|
|
|
|
50
|
|
|
public function getUrl(): string |
|
51
|
|
|
{ |
|
52
|
|
|
$this->params['format'] = 'json'; |
|
53
|
|
|
// todo verify http_build_query() enc_type parameter |
|
54
|
|
|
return 'https://wstat.fr/template/index.php?'.http_build_query($this->params); |
|
55
|
|
|
} |
|
56
|
|
|
|
|
57
|
|
|
/** |
|
58
|
|
|
* @return array [ ['title' => ..., 'template' => ...] ] |
|
59
|
|
|
* @throws Exception |
|
60
|
|
|
*/ |
|
61
|
|
|
public function getData(): array |
|
62
|
|
|
{ |
|
63
|
|
|
$data = []; |
|
64
|
|
|
while (true) { |
|
65
|
|
|
$json = $this->import($this->getUrl()); |
|
66
|
|
|
$raw = json_decode($json, true, 512, JSON_THROW_ON_ERROR); |
|
67
|
|
|
if (empty($raw)) { |
|
68
|
|
|
return []; |
|
69
|
|
|
} |
|
70
|
|
|
$data = array_merge($data, $this->parsingWstatData($raw)); |
|
71
|
|
|
echo count($data)." titles\n"; |
|
72
|
|
|
if ($this->max <= 0) { |
|
73
|
|
|
break; |
|
74
|
|
|
} |
|
75
|
|
|
|
|
76
|
|
|
// next page initialisation |
|
77
|
|
|
$this->params['start'] = ((int) $this->params['start'] + $this->params['limit']); |
|
78
|
|
|
sleep(3); |
|
79
|
|
|
} |
|
80
|
|
|
|
|
81
|
|
|
return $data; |
|
82
|
|
|
} |
|
83
|
|
|
|
|
84
|
|
|
/** |
|
85
|
|
|
* Explode raw string. |
|
86
|
|
|
* |
|
87
|
|
|
* |
|
88
|
|
|
* @return array [['title' => ..., 'template' => ...]] |
|
89
|
|
|
*/ |
|
90
|
|
|
private function parsingWstatData(array $raw): array |
|
91
|
|
|
{ |
|
92
|
|
|
// Generator ? |
|
93
|
|
|
// Alexandre S. Giffard|{{Ouvrage|langue=|auteur1=|prénom... |
|
94
|
|
|
$data = []; |
|
95
|
|
|
foreach ($raw as $line) { |
|
96
|
|
|
// end of page ? |
|
97
|
|
|
if ('<!-- + -->' === $line) { |
|
98
|
|
|
continue; |
|
99
|
|
|
} |
|
100
|
|
|
$this->max -= 1; |
|
101
|
|
|
|
|
102
|
|
|
// validate and explode wstat data |
|
103
|
|
|
$pos = mb_strpos((string) $line, '|', 0); |
|
104
|
|
|
if (false === $pos || 0 === $pos) { |
|
105
|
|
|
continue; |
|
106
|
|
|
} |
|
107
|
|
|
$title = trim(mb_substr((string) $line, 0, $pos)); |
|
108
|
|
|
$template = trim(mb_substr((string) $line, $pos + 1)); |
|
109
|
|
|
$data[] = ['title' => $title, 'template' => $template]; |
|
110
|
|
|
} |
|
111
|
|
|
|
|
112
|
|
|
return $data; |
|
113
|
|
|
} |
|
114
|
|
|
|
|
115
|
|
|
/** |
|
116
|
|
|
* |
|
117
|
|
|
* @return string |
|
118
|
|
|
* @throws Exception |
|
119
|
|
|
*/ |
|
120
|
|
|
private function import(string $url) |
|
121
|
|
|
{ |
|
122
|
|
|
$response = $this->client->get($url); |
|
123
|
|
|
if (200 !== $response->getStatusCode()) { |
|
124
|
|
|
throw new Exception( |
|
125
|
|
|
sprintf('Error code: %s reason: %s', $response->getStatusCode(), $response->getReasonPhrase()) |
|
126
|
|
|
); |
|
127
|
|
|
} |
|
128
|
|
|
|
|
129
|
|
|
return $response->getBody()->getContents(); |
|
130
|
|
|
} |
|
131
|
|
|
|
|
132
|
|
|
public function getPageTitles(): array |
|
133
|
|
|
{ |
|
134
|
|
|
// TODO: Implement getPageTitles() method. |
|
135
|
|
|
return []; |
|
136
|
|
|
} |
|
137
|
|
|
} |
|
138
|
|
|
|
Dispositif /
Wikibot
