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\ExternLink; |
11
|
|
|
|
12
|
|
|
use App\Domain\ExternLink\Validators\RobotNoIndexValidator; |
13
|
|
|
use App\Domain\InfrastructurePorts\DeadlinkArchiverInterface; |
14
|
|
|
use App\Domain\InfrastructurePorts\ExternHttpClientInterface; |
15
|
|
|
use App\Domain\InfrastructurePorts\InternetDomainParserInterface; |
16
|
|
|
use App\Domain\Models\Summary; |
17
|
|
|
use App\Domain\Models\Wiki\AbstractWikiTemplate; |
18
|
|
|
use App\Domain\Models\Wiki\ArticleTemplate; |
19
|
|
|
use App\Domain\Models\Wiki\LienWebTemplate; |
20
|
|
|
use App\Domain\Publisher\ExternMapper; |
21
|
|
|
use App\Domain\Utils\WikiTextUtil; |
22
|
|
|
use App\Domain\WikiOptimizer\OptimizerFactory; |
23
|
|
|
use App\Domain\WikiTemplateFactory; |
24
|
|
|
use Exception; |
25
|
|
|
use Normalizer; |
26
|
|
|
use Psr\Log\LoggerInterface; |
27
|
|
|
use Psr\Log\NullLogger; |
28
|
|
|
use Throwable; |
29
|
|
|
|
30
|
|
|
/** |
31
|
|
|
* TODO refac too big (responsibility) |
32
|
|
|
*/ |
33
|
|
|
class ExternRefTransformer implements ExternRefTransformerInterface |
34
|
|
|
{ |
35
|
|
|
use SummaryExternTrait, PublisherLogicTrait; |
|
|
|
|
36
|
|
|
|
37
|
|
|
public const HTTP_REQUEST_LOOP_DELAY = 10; |
38
|
|
|
public const SKIP_DOMAIN_FILENAME = __DIR__ . '/../resources/config_skip_domain.txt'; |
39
|
|
|
public const REPLACE_404 = true; |
40
|
|
|
public const REPLACE_410 = true; |
41
|
|
|
public const CONFIG_PRESSE = __DIR__ . '/../resources/config_presse.yaml'; |
42
|
|
|
public const CONFIG_NEWSPAPER_JSON = __DIR__ . '/../resources/data_newspapers.json'; |
43
|
|
|
public const CONFIG_SCIENTIFIC_JSON = __DIR__ . '/../resources/data_scientific_domain.json'; |
44
|
|
|
public const CONFIG_SCIENTIFIC_WIKI_JSON = __DIR__ . '/../resources/data_scientific_wiki.json'; |
45
|
|
|
|
46
|
|
|
public bool $skipSiteBlacklisted = true; |
47
|
|
|
public bool $skipRobotNoIndex = true; |
48
|
|
|
public array $summaryLog = []; |
49
|
|
|
|
50
|
|
|
protected $config; |
51
|
|
|
protected string $registrableDomain; |
52
|
|
|
protected string $url; |
53
|
|
|
protected array $publisherData = []; |
54
|
|
|
protected array $skip_domain = []; |
55
|
|
|
protected ExternPage $externalPage; |
56
|
|
|
protected ?Summary $summary; |
57
|
|
|
protected ?string $originDomain; |
58
|
|
|
protected array $options = []; |
59
|
|
|
private readonly ExternHttpErrorLogic $externHttpErrorLogic; |
60
|
|
|
private readonly CheckURL $urlChecker; |
61
|
|
|
|
62
|
|
|
public function __construct( |
63
|
|
|
protected ExternMapper $mapper, |
64
|
|
|
protected ExternHttpClientInterface $httpClient, |
65
|
|
|
protected InternetDomainParserInterface $domainParser, |
66
|
|
|
protected LoggerInterface $log = new NullLogger(), |
67
|
|
|
protected ?DeadlinkArchiverInterface $deadlinkArchiver = null |
68
|
|
|
) |
69
|
|
|
{ |
70
|
|
|
$this->importConfigAndData(); |
71
|
|
|
$this->externHttpErrorLogic = new ExternHttpErrorLogic( |
|
|
|
|
72
|
|
|
new DeadLinkTransformer($deadlinkArchiver, $domainParser, null, $log), |
73
|
|
|
$log |
74
|
|
|
); |
75
|
|
|
$this->urlChecker = new CheckURL($domainParser, $log); |
|
|
|
|
76
|
|
|
} |
77
|
|
|
|
78
|
|
|
/** |
79
|
|
|
* Transform "http://bla" => "{lien web|...}}", "{article}" or "{lien brisé}". |
80
|
|
|
* |
81
|
|
|
* TODO Refac : chain of responsibility or composite pattern |
82
|
|
|
* todo refac : return data DTO ? to much responsability! |
83
|
|
|
* |
84
|
|
|
* @throws Exception |
85
|
|
|
*/ |
86
|
|
|
public function process(string $url, Summary $summary = new Summary(), array $options = []): string |
87
|
|
|
{ |
88
|
|
|
$this->url = $url; |
89
|
|
|
$this->options = $options; // used only to pass RegistrableDomain of archived deadlink |
90
|
|
|
|
91
|
|
|
if (!$this->urlChecker->isURLAuthorized($url)) { |
92
|
|
|
return $url; |
93
|
|
|
} |
94
|
|
|
$this->registrableDomain = $this->urlChecker->getRegistrableDomain($url); // hack |
95
|
|
|
if ($this->isSiteBlackListed()) { |
96
|
|
|
return $url; |
97
|
|
|
} |
98
|
|
|
|
99
|
|
|
if (!$this->validateConfigWebDomain($this->registrableDomain)) { |
100
|
|
|
return $url; |
101
|
|
|
} |
102
|
|
|
|
103
|
|
|
try { |
104
|
|
|
$url = WikiTextUtil::normalizeUrlForTemplate($url); |
105
|
|
|
$pageData = $this->extractPageDataFromUrl($url); // ['JSON-LD'] & ['meta'] !! |
106
|
|
|
} catch (Exception $exception) { |
107
|
|
|
return $this->externHttpErrorLogic->manageByHttpErrorMessage($exception->getMessage(), $url); |
108
|
|
|
} |
109
|
|
|
if ($this->emptyPageData($pageData, $url)) { |
110
|
|
|
return $url; |
111
|
|
|
} |
112
|
|
|
$noIndexValidator = new RobotNoIndexValidator($pageData, $url, $this->log); // todo inject |
113
|
|
|
if ($noIndexValidator->validate() && $this->skipRobotNoIndex) { |
114
|
|
|
// TODO ? return {lien web| titre=Titre inconnu... |note=noindex } |
115
|
|
|
// http://www.nydailynews.com/entertainment/jessica-barth-details-alleged-harvey-weinstein-encounter-article-1.3557986 |
116
|
|
|
return $url; |
117
|
|
|
} |
118
|
|
|
|
119
|
|
|
$mappedData = $this->mapper->process($pageData); // only json-ld or only meta, after postprocess |
120
|
|
|
if ($this->emptyMapData($mappedData, $url)) { |
121
|
|
|
// TODO ? return {lien web| titre=Titre inconnu... site=prettydomain ... |
122
|
|
|
return $url; |
123
|
|
|
} |
124
|
|
|
$mappedData = $this->unsetAccesLibre($mappedData); |
125
|
|
|
|
126
|
|
|
$this->addSummaryLog($mappedData, $summary); |
127
|
|
|
$this->tagAndLog($mappedData); |
128
|
|
|
|
129
|
|
|
$template = $this->chooseTemplateByData($this->registrableDomain, $mappedData); |
130
|
|
|
|
131
|
|
|
$mappedData = $this->replaceSomeData($mappedData, $template); // template specif + data + url |
132
|
|
|
|
133
|
|
|
$serialized = $this->optimizeAndSerialize($template, $mappedData); |
134
|
|
|
|
135
|
|
|
$normalized = Normalizer::normalize($serialized); // sometimes :bool |
136
|
|
|
if (!empty($normalized) && is_string($normalized)) { |
137
|
|
|
return $normalized; |
138
|
|
|
} |
139
|
|
|
if (!empty($serialized)) { |
140
|
|
|
return $serialized; |
141
|
|
|
} |
142
|
|
|
|
143
|
|
|
return $url; // error fallback |
144
|
|
|
} |
145
|
|
|
|
146
|
|
|
protected function isSiteBlackListed(): bool |
147
|
|
|
{ |
148
|
|
|
if ($this->skipSiteBlacklisted && in_array($this->registrableDomain, $this->skip_domain)) { |
149
|
|
|
$this->log->notice("Skip web site " . $this->registrableDomain); |
150
|
|
|
return true; |
151
|
|
|
} |
152
|
|
|
return false; |
153
|
|
|
} |
154
|
|
|
|
155
|
|
|
/** |
156
|
|
|
* todo move transformer |
157
|
|
|
*/ |
158
|
|
|
protected function validateConfigWebDomain(string $domain): bool |
159
|
|
|
{ |
160
|
|
|
$this->logDebugConfigWebDomain($domain); |
161
|
|
|
|
162
|
|
|
// todo move to config |
163
|
|
|
$this->config[$domain] ??= []; |
164
|
|
|
$this->config[$domain] = is_array($this->config[$domain]) ? $this->config[$domain] : []; |
165
|
|
|
|
166
|
|
|
if ($this->config[$domain] === 'deactivated' || isset($this->config[$domain]['deactivated'])) { |
167
|
|
|
$this->log->info("Domain " . $domain . " disabled\n"); |
168
|
|
|
|
169
|
|
|
return false; |
170
|
|
|
} |
171
|
|
|
|
172
|
|
|
return true; |
173
|
|
|
} |
174
|
|
|
|
175
|
|
|
protected function logDebugConfigWebDomain(string $domain): void |
176
|
|
|
{ |
177
|
|
|
if (!isset($this->config[$domain])) { |
178
|
|
|
$this->log->debug("Domain " . $domain . " non configuré"); |
179
|
|
|
} else { |
180
|
|
|
$this->log->debug("Domain " . $domain . " configuré"); |
181
|
|
|
} |
182
|
|
|
} |
183
|
|
|
|
184
|
|
|
/** |
185
|
|
|
* Stay |
186
|
|
|
* @throws Exception |
187
|
|
|
*/ |
188
|
|
|
protected function extractPageDataFromUrl(string $url): array |
189
|
|
|
{ |
190
|
|
|
sleep(self::HTTP_REQUEST_LOOP_DELAY); |
191
|
|
|
$this->externalPage = ExternPageFactory::fromURL($url, $this->domainParser, $this->httpClient, $this->log); |
192
|
|
|
$pageData = $this->externalPage->getData(); |
193
|
|
|
$this->log->debug('metaData', $pageData); |
194
|
|
|
|
195
|
|
|
return $pageData; |
196
|
|
|
} |
197
|
|
|
|
198
|
|
|
// stay |
199
|
|
|
|
200
|
|
|
protected function emptyPageData(array $pageData, string $url): bool |
201
|
|
|
{ |
202
|
|
|
if ($pageData === [] |
203
|
|
|
|| (empty($pageData['JSON-LD']) && empty($pageData['meta'])) |
204
|
|
|
) { |
205
|
|
|
$this->log->notice('No metadata : ' . $url); |
206
|
|
|
|
207
|
|
|
return true; |
208
|
|
|
} |
209
|
|
|
|
210
|
|
|
return false; |
211
|
|
|
} |
212
|
|
|
|
213
|
|
|
/** |
214
|
|
|
* check dataValide |
215
|
|
|
* Pas de skip domaine car s'agit peut-être d'un 404 ou erreur juste sur cette URL |
216
|
|
|
*/ |
217
|
|
|
protected function emptyMapData(array $mapData, string $url): bool |
218
|
|
|
{ |
219
|
|
|
if ($mapData === [] || empty($mapData['url']) || empty($mapData['titre'])) { |
220
|
|
|
$this->log->info('Mapping incomplet : ' . $url); |
221
|
|
|
|
222
|
|
|
return true; |
223
|
|
|
} |
224
|
|
|
return false; |
225
|
|
|
} |
226
|
|
|
|
227
|
|
|
// stay |
228
|
|
|
|
229
|
|
|
/** |
230
|
|
|
* Pas de 'accès url=libre' # débat février 2021 |
231
|
|
|
*/ |
232
|
|
|
protected function unsetAccesLibre(array $mapData): array |
233
|
|
|
{ |
234
|
|
|
if (isset($mapData['accès url']) && $mapData['accès url'] === 'libre') { |
235
|
|
|
unset($mapData['accès url']); |
236
|
|
|
} |
237
|
|
|
return $mapData; |
238
|
|
|
} |
239
|
|
|
|
240
|
|
|
/** |
241
|
|
|
* todo Stay ? |
242
|
|
|
* todo refac lisible |
243
|
|
|
* @throws Exception |
244
|
|
|
*/ |
245
|
|
|
protected function chooseTemplateByData(string $domain, array $mapData): AbstractWikiTemplate |
246
|
|
|
{ |
247
|
|
|
// Logique : choix template |
248
|
|
|
$this->config[$domain]['template'] ??= []; |
249
|
|
|
$mapData['DATA-ARTICLE'] ??= false; |
250
|
|
|
|
251
|
|
|
if (!empty($mapData['doi'])) { |
252
|
|
|
$templateName = 'article'; |
253
|
|
|
} |
254
|
|
|
|
255
|
|
|
if ($this->config[$domain]['template'] === 'article' |
256
|
|
|
|| ($this->config[$domain]['template'] === 'auto' && $mapData['DATA-ARTICLE']) |
257
|
|
|
|| ($mapData['DATA-ARTICLE'] && !empty($this->publisherData['newspaper'][$domain])) |
258
|
|
|
|| $this->isScientificDomain() |
259
|
|
|
) { |
260
|
|
|
$templateName = 'article'; |
261
|
|
|
} |
262
|
|
|
if (!isset($templateName) || $this->config[$domain]['template'] === 'lien web') { |
263
|
|
|
$templateName = 'lien web'; |
264
|
|
|
} |
265
|
|
|
|
266
|
|
|
// date obligatoire pour {article} |
267
|
|
|
if (!isset($mapData['date'])) { |
268
|
|
|
$templateName = 'lien web'; |
269
|
|
|
} |
270
|
|
|
|
271
|
|
|
$template = WikiTemplateFactory::create($templateName); |
272
|
|
|
$template->userSeparator = " |"; |
273
|
|
|
$this->summary->memo['count ' . $templateName] = 1 + ($this->summary->memo['count ' . $templateName] ?? 0); |
274
|
|
|
|
275
|
|
|
return $template; |
|
|
|
|
276
|
|
|
} |
277
|
|
|
|
278
|
|
|
protected function replaceSomeData(array $mapData, AbstractWikiTemplate $template): array |
279
|
|
|
{ |
280
|
|
|
$mapData = $this->replaceSitenameByConfig($mapData, $template); |
281
|
|
|
$mapData = $this->fallbackIfSitenameNull($mapData, $template); |
282
|
|
|
$mapData = $this->correctSiteViaWebarchiver($mapData); |
283
|
|
|
|
284
|
|
|
$mapData = $this->replaceURLbyOriginal($mapData); |
285
|
|
|
|
286
|
|
|
if ($template instanceof ArticleTemplate) { |
287
|
|
|
unset($mapData['site']); |
288
|
|
|
} |
289
|
|
|
unset($mapData['DATA-TYPE']); // ugly |
290
|
|
|
unset($mapData['DATA-ARTICLE']); // ugly |
291
|
|
|
unset($mapData['url-access']); |
292
|
|
|
|
293
|
|
|
return $mapData; |
294
|
|
|
} |
295
|
|
|
|
296
|
|
|
// postprocess data |
297
|
|
|
|
298
|
|
|
protected function fallbackIfSitenameNull(array $mapData, AbstractWikiTemplate $template): array |
299
|
|
|
{ |
300
|
|
|
if (empty($mapData['site']) && $template instanceof LienWebTemplate) { |
301
|
|
|
try { |
302
|
|
|
$mapData['site'] = $this->externalPage->getPrettyDomainName(); |
303
|
|
|
} catch (Throwable $e) { |
304
|
|
|
unset($e); |
305
|
|
|
} |
306
|
|
|
} |
307
|
|
|
return $mapData; |
308
|
|
|
} |
309
|
|
|
|
310
|
|
|
protected function replaceURLbyOriginal(array $mapData): array |
311
|
|
|
{ |
312
|
|
|
$mapData['url'] = $this->url; |
313
|
|
|
|
314
|
|
|
return $mapData; |
315
|
|
|
} |
316
|
|
|
|
317
|
|
|
/** |
318
|
|
|
* |
319
|
|
|
* @throws Exception |
320
|
|
|
*/ |
321
|
|
|
protected function optimizeAndSerialize(AbstractWikiTemplate $template, array $mapData): string |
322
|
|
|
{ |
323
|
|
|
$template->hydrate($mapData); |
324
|
|
|
$optimizer = OptimizerFactory::fromTemplate($template); |
325
|
|
|
$optimizer->doTasks(); |
326
|
|
|
$templateOptimized = $optimizer->getOptiTemplate(); |
327
|
|
|
|
328
|
|
|
$serialized = $templateOptimized->serialize(true); |
329
|
|
|
$this->log->info('Serialized 444: ' . $serialized . "\n"); |
330
|
|
|
return $serialized; |
331
|
|
|
} |
332
|
|
|
|
333
|
|
|
protected function correctSiteViaWebarchiver(array $mapData): array |
334
|
|
|
{ |
335
|
|
|
if (!empty($this->options['originalRegistrableDomain']) && $mapData['site']) { |
336
|
|
|
$mapData['site'] = $this->options['originalRegistrableDomain'].' via '.$mapData['site']; |
337
|
|
|
} |
338
|
|
|
|
339
|
|
|
return $mapData; |
340
|
|
|
} |
341
|
|
|
} |
342
|
|
|
|