1
|
|
|
<?php |
2
|
|
|
|
3
|
|
|
declare( strict_types = 1 ); |
4
|
|
|
|
5
|
|
|
namespace Wikibase\Client\Usage; |
6
|
|
|
|
7
|
|
|
use MediaWiki\Cache\LinkBatchFactory; |
8
|
|
|
use TitleFactory; |
9
|
|
|
use Traversable; |
10
|
|
|
use Wikibase\DataModel\Entity\EntityId; |
11
|
|
|
use Wikibase\DataModel\Entity\ItemId; |
12
|
|
|
use Wikibase\Lib\Store\SiteLinkLookup; |
13
|
|
|
|
14
|
|
|
/** |
15
|
|
|
* A {@link UsageLookup} which decorates an inner lookup |
16
|
|
|
* and adds an implicit usage on a linked item’s description. |
17
|
|
|
* |
18
|
|
|
* An implicit usage is different from an ordinary, explicit usage |
19
|
|
|
* in that it is never recorded by a {@link UsageTracker}: |
20
|
|
|
* it is not tracked when the page is parsed and actually uses a part of an entity, |
21
|
|
|
* but rather synthesized by this class based on hard-coded knowledge |
22
|
|
|
* about where else an entity’s data is used in relation to a page. |
23
|
|
|
* However, implicit usages otherwise look exactly like explicit usages: |
24
|
|
|
* for a user of the {@link UsageLookup} interface, it is not (yet?) possible |
25
|
|
|
* to determine whether a usage from the lookup is explicit or implicit. |
26
|
|
|
* |
27
|
|
|
* This class implements one kind of implicit usage: |
28
|
|
|
* if a client page is linked to an item, it has an implicit usage |
29
|
|
|
* on that item’s description in the client wiki’s content language. |
30
|
|
|
* This is because the description is used, for example, |
31
|
|
|
* as part of the search result for the page (typically on mobile), |
32
|
|
|
* even if it is never used in the page itself. |
33
|
|
|
* |
34
|
|
|
* @see @ref md_docs_topics_usagetracking for virtual usage, |
35
|
|
|
* a similar but separate concept. |
36
|
|
|
* |
37
|
|
|
* @license GPL-2.0-or-later |
38
|
|
|
*/ |
39
|
|
|
class ImplicitDescriptionUsageLookup implements UsageLookup { |
40
|
|
|
|
41
|
|
|
/** @var UsageLookup */ |
42
|
|
|
private $usageLookup; |
43
|
|
|
|
44
|
|
|
/** @var TitleFactory */ |
45
|
|
|
private $titleFactory; |
46
|
|
|
|
47
|
|
|
/** @var LinkBatchFactory */ |
48
|
|
|
private $linkBatchFactory; |
49
|
|
|
|
50
|
|
|
/** @var string */ |
51
|
|
|
private $globalSiteId; |
52
|
|
|
|
53
|
|
|
/** @var SiteLinkLookup */ |
54
|
|
|
private $siteLinkLookup; |
55
|
|
|
|
56
|
|
|
/** |
57
|
|
|
* @param UsageLookup $usageLookup The underlying/inner lookup. |
58
|
|
|
* @param TitleFactory $titleFactory |
59
|
|
|
* @param LinkBatchFactory $linkBatchFactory |
60
|
|
|
* @param string $globalSiteId The global site ID of the client wiki. |
61
|
|
|
* @param SiteLinkLookup $siteLinkLookup |
62
|
|
|
*/ |
63
|
|
|
public function __construct( |
64
|
|
|
UsageLookup $usageLookup, |
65
|
|
|
TitleFactory $titleFactory, |
66
|
|
|
LinkBatchFactory $linkBatchFactory, |
67
|
|
|
string $globalSiteId, |
68
|
|
|
SiteLinkLookup $siteLinkLookup |
69
|
|
|
) { |
70
|
|
|
$this->usageLookup = $usageLookup; |
71
|
|
|
$this->titleFactory = $titleFactory; |
72
|
|
|
$this->linkBatchFactory = $linkBatchFactory; |
73
|
|
|
$this->globalSiteId = $globalSiteId; |
74
|
|
|
$this->siteLinkLookup = $siteLinkLookup; |
75
|
|
|
} |
76
|
|
|
|
77
|
|
|
public function getUsagesForPage( $pageId ): array { |
78
|
|
|
$usages = $this->usageLookup->getUsagesForPage( $pageId ); |
79
|
|
|
$title = $this->titleFactory->newFromID( $pageId ); |
80
|
|
|
if ( !$title ) { |
81
|
|
|
return $usages; |
82
|
|
|
} |
83
|
|
|
$entityId = $this->siteLinkLookup->getItemIdForLink( |
84
|
|
|
$this->globalSiteId, |
85
|
|
|
$title->getPrefixedText() |
86
|
|
|
); |
87
|
|
|
if ( !$entityId ) { |
88
|
|
|
return $usages; |
89
|
|
|
} |
90
|
|
|
$contentLanguage = $title->getPageLanguage()->getCode(); |
91
|
|
|
|
92
|
|
|
foreach ( $usages as $usage ) { |
93
|
|
|
if ( |
94
|
|
|
$usage->getAspect() === EntityUsage::DESCRIPTION_USAGE && |
95
|
|
|
$usage->getModifier() === $contentLanguage && |
96
|
|
|
$usage->getEntityId()->equals( $entityId ) |
97
|
|
|
) { |
98
|
|
|
// there already is an explicit usage, nothing to do |
99
|
|
|
return $usages; |
100
|
|
|
} |
101
|
|
|
} |
102
|
|
|
|
103
|
|
|
// there is no explicit usage, add the implicit one |
104
|
|
|
$usages[] = new EntityUsage( |
105
|
|
|
$entityId, |
106
|
|
|
EntityUsage::DESCRIPTION_USAGE, |
107
|
|
|
$contentLanguage |
108
|
|
|
); |
109
|
|
|
return $usages; |
110
|
|
|
} |
111
|
|
|
|
112
|
|
|
public function getPagesUsing( array $entityIds, array $aspects = [] ): Traversable { |
113
|
|
|
if ( !$this->aspectsMatchImplicitUsage( $aspects ) ) { |
114
|
|
|
// Caller is not interested in implicit usage, |
115
|
|
|
// no need to add anything |
116
|
|
|
return yield from $this->usageLookup->getPagesUsing( $entityIds, $aspects ); |
117
|
|
|
} |
118
|
|
|
|
119
|
|
|
// Find the implicit usages that we’ll add – one per page / item ID / content language |
120
|
|
|
[ $itemIdsByPageId, $contentLanguagesByPageId ] = $this->findImplicitUsages( $entityIds ); |
|
|
|
|
121
|
|
|
// Filter them according to the aspects |
122
|
|
|
[ $itemIdsByPageId, $contentLanguagesByPageId ] = $this->filterImplicitUsages( |
123
|
|
|
$itemIdsByPageId, $contentLanguagesByPageId, $aspects ); |
124
|
|
|
|
125
|
|
|
// Now decorate the inner lookup’s usages with them |
126
|
|
|
foreach ( $this->usageLookup->getPagesUsing( $entityIds, $aspects ) as $pageEntityUsages ) { |
127
|
|
|
/** @var PageEntityUsages $pageEntityUsages */ |
128
|
|
|
'@phan-var PageEntityUsages $pageEntityUsages'; |
129
|
|
|
$pageId = $pageEntityUsages->getPageId(); |
130
|
|
|
if ( isset( $itemIdsByPageId[$pageId] ) ) { |
131
|
|
|
// if equivalent usages already exist then addUsages() is a no-op |
132
|
|
|
$pageEntityUsages->addUsages( [ |
133
|
|
|
new EntityUsage( |
134
|
|
|
$itemIdsByPageId[$pageId], |
135
|
|
|
EntityUsage::DESCRIPTION_USAGE, |
136
|
|
|
$contentLanguagesByPageId[$pageId] |
137
|
|
|
), |
138
|
|
|
] ); |
139
|
|
|
unset( $itemIdsByPageId[$pageId] ); |
140
|
|
|
} |
141
|
|
|
yield $pageEntityUsages; |
142
|
|
|
} |
143
|
|
|
|
144
|
|
|
// And yield any remaining pages that the inner lookup didn’t return at all |
145
|
|
|
foreach ( $itemIdsByPageId as $pageId => $itemId ) { |
146
|
|
|
yield new PageEntityUsages( $pageId, [ |
147
|
|
|
new EntityUsage( |
148
|
|
|
$itemId, |
149
|
|
|
EntityUsage::DESCRIPTION_USAGE, |
150
|
|
|
$contentLanguagesByPageId[$pageId] |
151
|
|
|
), |
152
|
|
|
] ); |
153
|
|
|
} |
154
|
|
|
} |
155
|
|
|
|
156
|
|
|
/** |
157
|
|
|
* Whether the given aspects potentially match an implicit usage. |
158
|
|
|
* |
159
|
|
|
* @param string[] $aspects |
160
|
|
|
* @return bool |
161
|
|
|
*/ |
162
|
|
|
private function aspectsMatchImplicitUsage( array $aspects ): bool { |
163
|
|
|
if ( $aspects === [] ) { |
164
|
|
|
return true; |
165
|
|
|
} |
166
|
|
|
|
167
|
|
|
foreach ( $aspects as $aspectKey ) { |
168
|
|
|
if ( EntityUsage::stripModifier( $aspectKey ) === EntityUsage::DESCRIPTION_USAGE ) { |
169
|
|
|
// The implicit usage is on the description in the *page* content language, |
170
|
|
|
// not the wiki content language, so any description aspect matches. |
171
|
|
|
// (We’ll later filter based on the modifier and content language, |
172
|
|
|
// see filterImplicitUsages().) |
173
|
|
|
return true; |
174
|
|
|
} |
175
|
|
|
} |
176
|
|
|
return false; |
177
|
|
|
} |
178
|
|
|
|
179
|
|
|
/** |
180
|
|
|
* Find the implicit usages on the given entity IDs. |
181
|
|
|
* |
182
|
|
|
* Returns two arrays, both keyed by page ID: |
183
|
|
|
* the item ID linked to that page and the content language of the page. |
184
|
|
|
* |
185
|
|
|
* @param EntityId[] $entityIds |
186
|
|
|
* @return array [ ItemId[] $itemIdsByPageId, string[] $contentLanguagesByPageId ] |
187
|
|
|
*/ |
188
|
|
|
private function findImplicitUsages( array $entityIds ): array { |
189
|
|
|
$numericItemIds = []; |
190
|
|
|
foreach ( $entityIds as $entityId ) { |
191
|
|
|
if ( $entityId instanceof ItemId ) { |
192
|
|
|
$numericItemIds[] = $entityId->getNumericId(); |
193
|
|
|
} |
194
|
|
|
} |
195
|
|
|
|
196
|
|
|
// each link is an array [ string $siteId, string $pageName, int $itemId ] |
197
|
|
|
$links = $this->siteLinkLookup->getLinks( $numericItemIds, [ $this->globalSiteId ] ); |
198
|
|
|
// preload the titles in bulk (page ID and language) |
199
|
|
|
$titles = array_map( [ $this->titleFactory, 'newFromDBkey' ], array_column( $links, 1 ) ); |
200
|
|
|
$this->linkBatchFactory->newLinkBatch( $titles )->execute(); |
201
|
|
|
|
202
|
|
|
$itemIdsByPageId = []; |
203
|
|
|
foreach ( $links as [ $siteId, $pageName, $itemId ] ) { |
204
|
|
|
// note: this creates a new Title and looks up its page ID in the link cache; |
205
|
|
|
// this is simpler than finding the right existing Title in the $titles we have |
206
|
|
|
// (the $pageName is probably not exactly in DB key form) |
207
|
|
|
$pageId = $this->titleFactory->newFromDBkey( $pageName )->getArticleID(); |
|
|
|
|
208
|
|
|
if ( $pageId ) { |
209
|
|
|
$itemIdsByPageId[$pageId] = ItemId::newFromNumber( $itemId ); |
|
|
|
|
210
|
|
|
} |
211
|
|
|
} |
212
|
|
|
|
213
|
|
|
$contentLanguagesByPageId = []; |
214
|
|
|
foreach ( $titles as $title ) { |
215
|
|
|
$pageId = $title->getArticleID(); |
216
|
|
|
if ( $pageId ) { |
217
|
|
|
$contentLanguagesByPageId[$pageId] = $title->getPageLanguage()->getCode(); |
218
|
|
|
} |
219
|
|
|
} |
220
|
|
|
|
221
|
|
|
return [ $itemIdsByPageId, $contentLanguagesByPageId ]; |
222
|
|
|
} |
223
|
|
|
|
224
|
|
|
/** |
225
|
|
|
* Filter the implicit usages by the given aspects. |
226
|
|
|
* |
227
|
|
|
* Takes two arrays as returned by {@link findImplicitUsages} |
228
|
|
|
* and returns similar arrays, but filtered if necessary. |
229
|
|
|
* |
230
|
|
|
* @param ItemId[] $itemIdsByPageId |
231
|
|
|
* @param string[] $contentLanguagesByPageId |
232
|
|
|
* @param string[] $aspects |
233
|
|
|
* @return array [ ItemId[] $itemIdsByPageId, string[] $contentLanguagesByPageId ] |
234
|
|
|
*/ |
235
|
|
|
private function filterImplicitUsages( |
236
|
|
|
array $itemIdsByPageId, |
237
|
|
|
array $contentLanguagesByPageId, |
238
|
|
|
array $aspects |
239
|
|
|
): array { |
240
|
|
|
if ( $aspects === [] ) { |
241
|
|
|
// caller is interested in all usages, don’t filter |
242
|
|
|
return [ $itemIdsByPageId, $contentLanguagesByPageId ]; |
243
|
|
|
} |
244
|
|
|
|
245
|
|
|
$relevantLanguages = []; |
246
|
|
|
foreach ( $aspects as $aspectKey ) { |
247
|
|
|
[ $aspect, $modifier ] = EntityUsage::splitAspectKey( $aspectKey ); |
|
|
|
|
248
|
|
|
if ( $aspect !== EntityUsage::DESCRIPTION_USAGE ) { |
|
|
|
|
249
|
|
|
continue; |
250
|
|
|
} |
251
|
|
|
if ( $modifier === null ) { |
252
|
|
|
// caller is interested in all description usages, don’t filter |
253
|
|
|
return [ $itemIdsByPageId, $contentLanguagesByPageId ]; |
254
|
|
|
} |
255
|
|
|
$relevantLanguages[] = $modifier; |
256
|
|
|
} |
257
|
|
|
|
258
|
|
|
// caller is only interested in some description usages, filter |
259
|
|
|
foreach ( $contentLanguagesByPageId as $pageId => $contentLanguage ) { |
260
|
|
|
if ( !in_array( $contentLanguage, $relevantLanguages, /* strict */ true ) ) { |
261
|
|
|
unset( $itemIdsByPageId[$pageId] ); |
262
|
|
|
unset( $contentLanguagesByPageId[$pageId] ); |
263
|
|
|
} |
264
|
|
|
} |
265
|
|
|
return [ $itemIdsByPageId, $contentLanguagesByPageId ]; |
266
|
|
|
} |
267
|
|
|
|
268
|
|
|
public function getUnusedEntities( array $entityIds ): array { |
269
|
|
|
// If a page is linked to an item, it has at least a sitelink usage on it; |
270
|
|
|
// therefore, the implicit usage can never make a difference for |
271
|
|
|
// whether an entity is used or unused. |
272
|
|
|
return $this->usageLookup->getUnusedEntities( $entityIds ); |
273
|
|
|
} |
274
|
|
|
|
275
|
|
|
} |
276
|
|
|
|
This check marks access to variables or properties that have not been declared yet. While PHP has no explicit notion of declaring a variable, accessing it before a value is assigned to it is most likely a bug.