|
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\Infrastructure; |
|
12
|
|
|
|
|
13
|
|
|
use App\Application\InfrastructurePorts\HttpClientInterface; |
|
14
|
|
|
use App\Application\InfrastructurePorts\PageListForAppInterface; |
|
15
|
|
|
use App\Domain\Exceptions\ConfigException; |
|
16
|
|
|
use App\Domain\InfrastructurePorts\PageListInterface; |
|
17
|
|
|
use Exception; |
|
18
|
|
|
use GuzzleHttp\Psr7\Response; |
|
19
|
|
|
use InvalidArgumentException; |
|
20
|
|
|
use Throwable; |
|
21
|
|
|
|
|
22
|
|
|
/** |
|
23
|
|
|
* https://www.mediawiki.org/wiki/Help:CirrusSearch |
|
24
|
|
|
* https://fr.wikipedia.org/w/api.php?action=help&modules=query%2Bsearch |
|
25
|
|
|
* https://www.mediawiki.org/wiki/Help:CirrusSearch#Insource |
|
26
|
|
|
* raw https://fr.wikipedia.org/w/api.php?action=query&list=search&srsearch=%22https://books.google%22%20insource:/\%3Cref\%3Ehttps\:\/\/books\.google/&formatversion=2&format=json |
|
27
|
|
|
* Dirty. |
|
28
|
|
|
* Class CirrusSearch |
|
29
|
|
|
*/ |
|
30
|
|
|
class CirrusSearch implements PageListInterface, PageListForAppInterface |
|
31
|
|
|
{ |
|
32
|
|
|
public const OPTION_CONTINUE = 'continue'; |
|
33
|
|
|
public const OPTION_REVERSE = 'reverse'; |
|
34
|
|
|
public const SRSORT_NONE = 'none'; |
|
35
|
|
|
public const SRSORT_RANDOM = 'random'; |
|
36
|
|
|
public const SRSORT_LAST_EDIT_DESC = 'last_edit_desc'; |
|
37
|
|
|
public const SRQIPROFILE_POPULAR_INCLINKS_PV = 'popular_inclinks_pv'; // nombre de vues de la page :) |
|
38
|
|
|
public const SRQIPROFILE_DEFAULT = 'engine_autoselect'; |
|
39
|
|
|
|
|
40
|
|
|
protected const BASE_URL = 'https://fr.wikipedia.org/w/api.php'; // todo move config |
|
41
|
|
|
protected const CONTINUE_OFFSET_FILENAME = __DIR__ . '/../../resources/cirrusSearch-{HASH}.txt'; // todo move config |
|
42
|
|
|
|
|
43
|
|
|
protected array $requestParams = []; |
|
44
|
|
|
protected array $defaultParams |
|
45
|
|
|
= [ |
|
46
|
|
|
'action' => 'query', |
|
47
|
|
|
'list' => 'search', |
|
48
|
|
|
'formatversion' => '2', |
|
49
|
|
|
'format' => 'json', |
|
50
|
|
|
'srnamespace' => 0, |
|
51
|
|
|
'srlimit' => '10', // default 10, max 500 |
|
52
|
|
|
'srprop' => 'size|wordcount|timestamp', // default 'size|wordcount|timestamp|snippet' |
|
53
|
|
|
]; |
|
54
|
|
|
protected readonly HttpClientInterface $client; |
|
55
|
|
|
|
|
56
|
|
|
/** |
|
57
|
|
|
* $options : "continue" => true for continue search |
|
58
|
|
|
*/ |
|
59
|
|
|
public function __construct(protected readonly array $params, protected ?array $options = []) |
|
60
|
|
|
{ |
|
61
|
|
|
$this->client = ServiceFactory::getHttpClient(); |
|
|
|
|
|
|
62
|
|
|
} |
|
63
|
|
|
|
|
64
|
|
|
/** |
|
65
|
|
|
* @return array |
|
66
|
|
|
* @throws ConfigException |
|
67
|
|
|
*/ |
|
68
|
|
|
public function getPageTitles(): array |
|
69
|
|
|
{ |
|
70
|
|
|
$arrayResp = $this->httpRequest(); |
|
71
|
|
|
|
|
72
|
|
|
if ($this->options[self::OPTION_CONTINUE] ?? false) { |
|
73
|
|
|
$continueOffset = 0; |
|
74
|
|
|
if (!empty($arrayResp['continue']['sroffset'])) { |
|
75
|
|
|
$continueOffset = (int)$arrayResp['continue']['sroffset']; |
|
76
|
|
|
} |
|
77
|
|
|
$this->saveOffsetInFile($continueOffset); |
|
78
|
|
|
} |
|
79
|
|
|
|
|
80
|
|
|
if (!isset($arrayResp['query']) || empty($arrayResp['query']['search'])) { |
|
81
|
|
|
return []; |
|
82
|
|
|
} |
|
83
|
|
|
$results = $arrayResp['query']['search']; |
|
84
|
|
|
|
|
85
|
|
|
$titles = []; |
|
86
|
|
|
foreach ($results as $res) { |
|
87
|
|
|
if (!empty($res['title'])) { |
|
88
|
|
|
$titles[] = trim((string)$res['title']); // trim utile ? |
|
89
|
|
|
} |
|
90
|
|
|
} |
|
91
|
|
|
|
|
92
|
|
|
if (isset($this->options[self::OPTION_REVERSE]) && $this->options[self::OPTION_REVERSE] === true) { |
|
93
|
|
|
krsort($titles); |
|
94
|
|
|
} |
|
95
|
|
|
|
|
96
|
|
|
return $titles; |
|
97
|
|
|
} |
|
98
|
|
|
|
|
99
|
|
|
/** |
|
100
|
|
|
* todo Wiki API ? |
|
101
|
|
|
* @throws ConfigException |
|
102
|
|
|
* @throws Exception |
|
103
|
|
|
*/ |
|
104
|
|
|
protected function httpRequest(): array |
|
105
|
|
|
{ |
|
106
|
|
|
$url = $this->getURL(); |
|
107
|
|
|
if ($url === '' || $url === '0') { |
|
108
|
|
|
throw new ConfigException('CirrusSearch null URL'); |
|
109
|
|
|
} |
|
110
|
|
|
|
|
111
|
|
|
// improve with curl options ? |
|
112
|
|
|
$response = $this->client->get($url); |
|
113
|
|
|
/** |
|
114
|
|
|
* @var $response Response |
|
115
|
|
|
*/ |
|
116
|
|
|
if ($response->getStatusCode() !== 200) { |
|
117
|
|
|
throw new Exception( |
|
118
|
|
|
'CirrusSearch error : ' . $response->getStatusCode() . ' ' . $response->getReasonPhrase() |
|
119
|
|
|
); |
|
120
|
|
|
} |
|
121
|
|
|
$json = $response->getBody()->getContents(); |
|
122
|
|
|
if (empty($json)) { |
|
123
|
|
|
return []; |
|
124
|
|
|
} |
|
125
|
|
|
try { |
|
126
|
|
|
$array = json_decode((string)$json, true, 512, JSON_THROW_ON_ERROR); |
|
127
|
|
|
} catch (Throwable $e) { |
|
128
|
|
|
throw new Exception($e->getMessage(), $e->getCode(), $e); |
|
129
|
|
|
} |
|
130
|
|
|
|
|
131
|
|
|
return $array; |
|
132
|
|
|
} |
|
133
|
|
|
|
|
134
|
|
|
protected function getURL(): string |
|
135
|
|
|
{ |
|
136
|
|
|
if (empty($this->params['srsearch'])) { |
|
137
|
|
|
throw new InvalidArgumentException('No "srsearch" argument in params.'); |
|
138
|
|
|
} |
|
139
|
|
|
|
|
140
|
|
|
$this->requestParams = array_merge($this->defaultParams, $this->params); |
|
141
|
|
|
if ($this->options[self::OPTION_CONTINUE] ?? false) { |
|
142
|
|
|
$this->requestParams['sroffset'] = $this->getOffsetFromFile($this->requestParams); |
|
143
|
|
|
//echo sprintf("Extract offset %s from file \n", $this->requestParams['sroffset']); |
|
144
|
|
|
} |
|
145
|
|
|
// RFC3986 : space => %20 |
|
146
|
|
|
$query = http_build_query($this->requestParams, 'bla', '&', PHP_QUERY_RFC3986); |
|
147
|
|
|
|
|
148
|
|
|
return self::BASE_URL . '?' . $query; |
|
149
|
|
|
} |
|
150
|
|
|
|
|
151
|
|
|
protected function getOffsetFromFile(array $allParams): int |
|
152
|
|
|
{ |
|
153
|
|
|
$hash = $this->hashSearchParams($allParams); |
|
154
|
|
|
$file = str_replace('{HASH}', $hash, self::CONTINUE_OFFSET_FILENAME); |
|
155
|
|
|
if (!file_exists($file)) { |
|
156
|
|
|
return 0; |
|
157
|
|
|
} |
|
158
|
|
|
|
|
159
|
|
|
return (int)trim(file_get_contents($file)); |
|
160
|
|
|
} |
|
161
|
|
|
|
|
162
|
|
|
protected function hashSearchParams(array $params): string |
|
163
|
|
|
{ |
|
164
|
|
|
if (empty($params)) { |
|
165
|
|
|
throw new InvalidArgumentException('No search argument in params.'); |
|
166
|
|
|
} |
|
167
|
|
|
if (isset($params['sroffset'])) { |
|
168
|
|
|
unset($params['sroffset']); |
|
169
|
|
|
} |
|
170
|
|
|
|
|
171
|
|
|
return md5(implode('', $params)); |
|
172
|
|
|
} |
|
173
|
|
|
|
|
174
|
|
|
protected function saveOffsetInFile(int $continueOffset = 0): void |
|
175
|
|
|
{ |
|
176
|
|
|
$hash = $this->hashSearchParams($this->requestParams); |
|
177
|
|
|
$offsetFilename = str_replace('{HASH}', $hash, self::CONTINUE_OFFSET_FILENAME); |
|
178
|
|
|
|
|
179
|
|
|
if ($continueOffset === 0 && file_exists($offsetFilename)) { |
|
180
|
|
|
@unlink($offsetFilename); |
|
|
|
|
|
|
181
|
|
|
} else { |
|
182
|
|
|
file_put_contents($offsetFilename, $continueOffset); |
|
183
|
|
|
} |
|
184
|
|
|
} |
|
185
|
|
|
} |
|
186
|
|
|
|
Dispositif /
Wikibot
Loading history...