Code Coverage |
||||||||||
Lines |
Functions and Methods |
Classes and Traits |
||||||||
| Total | |
89.49% |
247 / 276 |
|
60.00% |
9 / 15 |
CRAP | |
0.00% |
0 / 1 |
| GenerateInvitationList | |
89.82% |
247 / 275 |
|
60.00% |
9 / 15 |
55.97 | |
0.00% |
0 / 1 |
| __construct | |
100.00% |
13 / 13 |
|
100.00% |
1 / 1 |
1 | |||
| execute | |
87.50% |
14 / 16 |
|
0.00% |
0 / 1 |
4.03 | |||
| getArticlesByWiki | |
75.00% |
21 / 28 |
|
0.00% |
0 / 1 |
11.56 | |||
| getAllRevisionsForWiki | |
98.48% |
65 / 66 |
|
0.00% |
0 / 1 |
8 | |||
| getRevisionFilterConditions | |
72.00% |
36 / 50 |
|
0.00% |
0 / 1 |
5.55 | |||
| rankUsers | |
100.00% |
10 / 10 |
|
100.00% |
1 / 1 |
2 | |||
| getDeltasByUser | |
100.00% |
16 / 16 |
|
100.00% |
1 / 1 |
5 | |||
| getUserDataByWiki | |
100.00% |
5 / 5 |
|
100.00% |
1 / 1 |
3 | |||
| getUserScore | |
100.00% |
12 / 12 |
|
100.00% |
1 / 1 |
1 | |||
| getOverallBytesScore | |
80.00% |
12 / 15 |
|
0.00% |
0 / 1 |
4.13 | |||
| getBytesScoreForPage | |
100.00% |
4 / 4 |
|
100.00% |
1 / 1 |
1 | |||
| getEditCountScore | |
100.00% |
5 / 5 |
|
100.00% |
1 / 1 |
1 | |||
| getEditCount | |
100.00% |
13 / 13 |
|
100.00% |
1 / 1 |
3 | |||
| getRecentActivityScore | |
100.00% |
6 / 6 |
|
100.00% |
1 / 1 |
1 | |||
| getDaysSinceLastEdit | |
93.75% |
15 / 16 |
|
0.00% |
0 / 1 |
4.00 | |||
| 1 | <?php |
| 2 | |
| 3 | declare( strict_types=1 ); |
| 4 | |
| 5 | namespace MediaWiki\Extension\CampaignEvents\Maintenance; |
| 6 | |
| 7 | use ChangeTags; |
| 8 | use Maintenance; |
| 9 | use MediaWiki\MainConfigNames; |
| 10 | use MediaWiki\MediaWikiServices; |
| 11 | use MediaWiki\Page\PageIdentity; |
| 12 | use MediaWiki\Storage\NameTableAccessException; |
| 13 | use MediaWiki\Utils\MWTimestamp; |
| 14 | use MediaWiki\WikiMap\WikiMap; |
| 15 | use Wikimedia\Rdbms\IReadableDatabase; |
| 16 | use Wikimedia\Rdbms\SelectQueryBuilder; |
| 17 | use Wikimedia\Timestamp\ConvertibleTimestamp; |
| 18 | |
| 19 | /** |
| 20 | * This scripts takes a list of pages as input, and outputs a list of users who've edited those pages the most. Each |
| 21 | * user is given a score from 0 to 100 based on how likely they are to be a productive participant in an event that is |
| 22 | * focused on improving the given pages. This score is based on how many edits someone made (and how big they are) to |
| 23 | * the pages in the list, their global edit count, and their recent global activity. |
| 24 | * NOTE: This script is just a demo / proof of concept. It is not based on any real-world data and the calculations |
| 25 | * are very much non-rigorous. |
| 26 | */ |
| 27 | class GenerateInvitationList extends Maintenance { |
| 28 | /** |
| 29 | * How many days to look back into the past when scanning revisions. |
| 30 | * TODO: Is 3 years OK? |
| 31 | */ |
| 32 | public const CUTOFF_DAYS = 3 * 365; |
| 33 | |
| 34 | public function __construct() { |
| 35 | parent::__construct(); |
| 36 | $this->addDescription( |
| 37 | 'Generates a list of potential event participants by looking at who contributed to a given list of pages' |
| 38 | ); |
| 39 | $this->requireExtension( 'CampaignEvents' ); |
| 40 | $this->addOption( |
| 41 | 'listfile', |
| 42 | 'Path to a file with a list of articles to get contributors for. The file should have one page per ' . |
| 43 | 'line, in the following format: `[wikiID, or empty for the local wiki]:[page title]`. All the pages ' . |
| 44 | 'must be in the mainspace.', |
| 45 | true, |
| 46 | true |
| 47 | ); |
| 48 | } |
| 49 | |
| 50 | /** |
| 51 | * @inheritDoc |
| 52 | */ |
| 53 | public function execute(): void { |
| 54 | $pagesByWiki = $this->getArticlesByWiki(); |
| 55 | $this->output( |
| 56 | "==Articles==\n" . implode( "\n", array_merge( ...array_values( $pagesByWiki ) ) ) . "\n\n" |
| 57 | ); |
| 58 | |
| 59 | $revisionsByWiki = []; |
| 60 | foreach ( $pagesByWiki as $wiki => $pages ) { |
| 61 | $revisionsByWiki[$wiki] = $this->getAllRevisionsForWiki( $wiki, $pages ); |
| 62 | } |
| 63 | $revisionsByWiki = array_filter( $revisionsByWiki ); |
| 64 | if ( !$revisionsByWiki ) { |
| 65 | $this->output( "No revisions found.\n" ); |
| 66 | return; |
| 67 | } |
| 68 | |
| 69 | $rankedUsers = $this->rankUsers( $revisionsByWiki ); |
| 70 | $out = "\n==Contributor scores==\n"; |
| 71 | foreach ( $rankedUsers as $username => $score ) { |
| 72 | $out .= "$username - $score\n"; |
| 73 | } |
| 74 | $this->output( $out . "\n\n" ); |
| 75 | } |
| 76 | |
| 77 | /** |
| 78 | * Reads a list of articles from the file passed as `listfile` to the script. |
| 79 | * |
| 80 | * @return PageIdentity[][] Map of [ wiki ID => non-empty list of articles ] |
| 81 | * @phan-return non-empty-array<string|false,non-empty-list<PageIdentity>> |
| 82 | */ |
| 83 | private function getArticlesByWiki(): array { |
| 84 | $listPath = $this->getOption( 'listfile' ); |
| 85 | // phpcs:ignore Generic.PHP.NoSilencedErrors.Discouraged |
| 86 | $rawList = @file_get_contents( $listPath ); |
| 87 | if ( $rawList === false ) { |
| 88 | $this->fatalError( "Cannot read list of articles" ); |
| 89 | } |
| 90 | |
| 91 | $curWikiID = WikiMap::getCurrentWikiId(); |
| 92 | $pageStoreFactory = MediaWikiServices::getInstance()->getPageStoreFactory(); |
| 93 | $pagesByWiki = []; |
| 94 | foreach ( explode( "\n", $rawList ) as $line ) { |
| 95 | if ( $line === '' ) { |
| 96 | continue; |
| 97 | } |
| 98 | |
| 99 | $lineParts = explode( ':', $line, 2 ); |
| 100 | if ( count( $lineParts ) !== 2 ) { |
| 101 | $this->fatalError( "Line without wiki ID: $line" ); |
| 102 | } |
| 103 | // XXX: We're using the actual wiki ID instead of WikiAwareEntity::LOCAL for the local wiki, so that PHP |
| 104 | // won't autocast it to `0` when used as array key. |
| 105 | $wikiID = $lineParts[0] === '' ? $curWikiID : $lineParts[0]; |
| 106 | $title = $lineParts[1]; |
| 107 | $pageStore = $pageStoreFactory->getPageStore( $wikiID ); |
| 108 | // Note: If $title happens to contain a namespace identifier, or really anything that cannot be parsed in |
| 109 | // the context of the current wiki, this method won't behave correctly due to T353916. There doesn't seem |
| 110 | // to be much that we can do about it. |
| 111 | $page = $pageStore->getPageByText( $title ); |
| 112 | if ( !$page ) { |
| 113 | $this->fatalError( "Invalid title: $title" ); |
| 114 | } elseif ( !$page->exists() ) { |
| 115 | $this->fatalError( "Page does not exist: $title" ); |
| 116 | } elseif ( $page->getNamespace() !== NS_MAIN ) { |
| 117 | $this->fatalError( "Page is not in the mainspace: $title" ); |
| 118 | } |
| 119 | $pagesByWiki[$wikiID] ??= []; |
| 120 | $pagesByWiki[$wikiID][] = $page; |
| 121 | } |
| 122 | |
| 123 | if ( !$pagesByWiki ) { |
| 124 | $this->fatalError( "Empty list of articles" ); |
| 125 | } |
| 126 | |
| 127 | return $pagesByWiki; |
| 128 | } |
| 129 | |
| 130 | /** |
| 131 | * @param string|false $wikiID |
| 132 | * @param PageIdentity[] $pages |
| 133 | * @return array[] List of arrays with revision data. The page is only included for debugging, and callers should |
| 134 | * not rely on its format. |
| 135 | * @phan-return list<array{username:string,userID:int,actorID:int,page:string,delta:int}> |
| 136 | */ |
| 137 | private function getAllRevisionsForWiki( $wikiID, array $pages ): array { |
| 138 | $revisionStore = MediaWikiServices::getInstance()->getRevisionStoreFactory()->getRevisionStore( $wikiID ); |
| 139 | $dbProvider = MediaWikiServices::getInstance()->getDBLoadBalancerFactory(); |
| 140 | // This script may potentially scan a lot of revisions. Although the queries can use good indexes, sending them |
| 141 | // to vslow hosts shouldn't hurt. |
| 142 | $dbr = $dbProvider->getReplicaDatabase( $wikiID, 'vslow' ); |
| 143 | |
| 144 | $pagesByID = []; |
| 145 | foreach ( $pages as $page ) { |
| 146 | $pageID = $page->getId( $wikiID ); |
| 147 | $pagesByID[$pageID] = $page; |
| 148 | } |
| 149 | asort( $pagesByID ); |
| 150 | $pageChunks = array_chunk( array_keys( $pagesByID ), 25 ); |
| 151 | $totalPageChunks = count( $pageChunks ); |
| 152 | |
| 153 | $baseWhereConds = $this->getRevisionFilterConditions( $wikiID, $dbr ); |
| 154 | |
| 155 | $batchSize = 2500; |
| 156 | $revisions = []; |
| 157 | $pageBatchIdx = 1; |
| 158 | |
| 159 | // Process the list of pages in smaller chunks, to avoid the optimizer making wrong decisions, and also to keep |
| 160 | // the queries more readable. |
| 161 | foreach ( $pageChunks as $batchPageIDs ) { |
| 162 | $lastPage = 0; |
| 163 | $lastTimestamp = $dbr->timestamp( '20000101000000' ); |
| 164 | $lastRevID = 0; |
| 165 | $innerBatchIdx = 1; |
| 166 | do { |
| 167 | $progressMsg = "Running $wikiID batch #$pageBatchIdx.$innerBatchIdx of $totalPageChunks " . |
| 168 | "from pageID=" . min( $batchPageIDs ); |
| 169 | if ( $lastRevID !== 0 ) { |
| 170 | $progressMsg .= ", ts=$lastTimestamp, rev=$lastRevID"; |
| 171 | } |
| 172 | $this->output( $progressMsg . "\n" ); |
| 173 | $revQueryBuilder = $revisionStore->newSelectQueryBuilder( $dbr ); |
| 174 | $res = $revQueryBuilder |
| 175 | ->field( 'actor_name' ) |
| 176 | // Needed for the user_is_temp check. |
| 177 | ->joinUser() |
| 178 | ->where( $baseWhereConds ) |
| 179 | ->andWhere( [ 'rev_page' => $batchPageIDs ] ) |
| 180 | ->andWhere( [ |
| 181 | $dbr->buildComparison( '>', [ |
| 182 | 'rev_page' => $lastPage, |
| 183 | 'rev_timestamp' => $lastTimestamp, |
| 184 | 'rev_id' => $lastRevID |
| 185 | ] ) |
| 186 | ] ) |
| 187 | ->orderBy( [ 'rev_page', 'rev_timestamp', 'rev_id' ], SelectQueryBuilder::SORT_ASC ) |
| 188 | ->limit( $batchSize ) |
| 189 | ->caller( __METHOD__ ) |
| 190 | ->fetchResultSet(); |
| 191 | |
| 192 | $parents = []; |
| 193 | foreach ( $res as $row ) { |
| 194 | $parentID = (int)$row->rev_parent_id; |
| 195 | if ( $parentID !== 0 ) { |
| 196 | $parents[$row->rev_id] = $parentID; |
| 197 | } |
| 198 | } |
| 199 | |
| 200 | $parentSizes = $revisionStore->getRevisionSizes( array_values( $parents ) ); |
| 201 | |
| 202 | foreach ( $res as $row ) { |
| 203 | $parentID = $parents[$row->rev_id] ?? null; |
| 204 | $parentSize = $parentID ? $parentSizes[$parentID] : 0; |
| 205 | $revisions[] = [ |
| 206 | 'username' => $row->actor_name, |
| 207 | 'userID' => $row->rev_user, |
| 208 | 'actorID' => $row->rev_actor, |
| 209 | 'page' => $pagesByID[$row->rev_page]->__toString(), |
| 210 | 'delta' => (int)$row->rev_len - $parentSize |
| 211 | ]; |
| 212 | $lastPage = (int)$row->rev_page; |
| 213 | $lastTimestamp = $row->rev_timestamp; |
| 214 | $lastRevID = (int)$row->rev_id; |
| 215 | } |
| 216 | |
| 217 | $innerBatchIdx++; |
| 218 | sleep( 1 ); |
| 219 | } while ( $res->numRows() >= $batchSize ); |
| 220 | $pageBatchIdx++; |
| 221 | } |
| 222 | |
| 223 | return $revisions; |
| 224 | } |
| 225 | |
| 226 | /** |
| 227 | * Returns an array of filters to apply to the revision query. |
| 228 | * |
| 229 | * @param string|false $wikiID |
| 230 | * @param IReadableDatabase $dbr |
| 231 | * @return array |
| 232 | */ |
| 233 | private function getRevisionFilterConditions( $wikiID, IReadableDatabase $dbr ): array { |
| 234 | $filterConditions = []; |
| 235 | |
| 236 | // Exclude all sorts of deleted revisions to avoid any chance of data leaks. |
| 237 | $filterConditions['rev_deleted'] = 0; |
| 238 | |
| 239 | // Exclude anons and temp users. |
| 240 | $filterConditions[] = $dbr->expr( 'actor_user', '!=', null ); |
| 241 | $filterConditions['user_is_temp'] = 0; |
| 242 | |
| 243 | // Exclude anything too old. |
| 244 | $startTime = (int)ConvertibleTimestamp::now( TS_UNIX ) - self::CUTOFF_DAYS * 24 * 60 * 60; |
| 245 | $filterConditions[] = $dbr->expr( 'rev_timestamp', '>=', $dbr->timestamp( $startTime ) ); |
| 246 | |
| 247 | // Exclude both edits that have been reverted, and edits that revert other edits. Neither of these is relevant, |
| 248 | // and can easily skew the deltas. |
| 249 | $nameTableStoreFactory = MediaWikiServices::getInstance()->getNameTableStoreFactory(); |
| 250 | $changeTagDefStore = $nameTableStoreFactory->getChangeTagDef( $wikiID ); |
| 251 | $revertTagIDs = []; |
| 252 | foreach ( [ ...ChangeTags::REVERT_TAGS, ChangeTags::TAG_REVERTED ] as $tagName ) { |
| 253 | try { |
| 254 | $revertTagIDs[] = $changeTagDefStore->getId( $tagName ); |
| 255 | } catch ( NameTableAccessException $e ) { |
| 256 | // There's no tag ID if no revisions have ever been tagged with this tag. |
| 257 | } |
| 258 | } |
| 259 | if ( $revertTagIDs ) { |
| 260 | $tagSubquery = $dbr->newSelectQueryBuilder() |
| 261 | ->select( '1' ) |
| 262 | ->from( 'change_tag' ) |
| 263 | ->where( [ 'ct_rev_id = rev_id', 'ct_tag_id' => $revertTagIDs ] ); |
| 264 | $filterConditions[] = 'NOT EXISTS(' . $tagSubquery->getSQL() . ')'; |
| 265 | } |
| 266 | |
| 267 | // Exclude users who have a sitewide infinite block. |
| 268 | $readOldBlockSchema = (bool)( MediaWikiServices::getInstance()->getMainConfig() |
| 269 | ->get( MainConfigNames::BlockTargetMigrationStage ) & SCHEMA_COMPAT_READ_OLD ); |
| 270 | if ( $readOldBlockSchema ) { |
| 271 | $blocksSubquery = $dbr->newSelectQueryBuilder() |
| 272 | ->select( '1' ) |
| 273 | ->from( 'ipblocks' ) |
| 274 | ->where( [ |
| 275 | $dbr->expr( 'ipb_user', '!=', 0 ), |
| 276 | 'actor_rev_user.actor_user = ipb_user', |
| 277 | 'ipb_expiry' => $dbr->getInfinity(), |
| 278 | 'ipb_sitewide' => 1, |
| 279 | ] ); |
| 280 | } else { |
| 281 | $blocksSubquery = $dbr->newSelectQueryBuilder() |
| 282 | ->select( '1' ) |
| 283 | ->from( 'block' ) |
| 284 | ->join( 'block_target', null, 'bt_id=bl_target' ) |
| 285 | ->where( [ |
| 286 | $dbr->expr( 'bt_user', '!=', null ), |
| 287 | 'actor_rev_user.actor_user = bt_user', |
| 288 | 'bl_expiry' => $dbr->getInfinity(), |
| 289 | 'bl_sitewide' => 1, |
| 290 | ] ); |
| 291 | } |
| 292 | $filterConditions[] = 'NOT EXISTS(' . $blocksSubquery->getSQL() . ')'; |
| 293 | |
| 294 | // Exclude bots. Note, this only checks whether a user is *currently* a bot, not whether |
| 295 | // they were a bot at the time the edit was made. |
| 296 | // XXX: Ideally we would use GroupPermissionLookup to list user groups with the 'bot' right, but that |
| 297 | // only works for the local wiki. |
| 298 | $botSubquery = $dbr->newSelectQueryBuilder() |
| 299 | ->select( '1' ) |
| 300 | ->from( 'user_groups' ) |
| 301 | ->where( [ |
| 302 | 'actor_rev_user.actor_user = ug_user', |
| 303 | 'ug_group' => 'bot', |
| 304 | ] ); |
| 305 | $filterConditions[] = 'NOT EXISTS(' . $botSubquery->getSQL() . ')'; |
| 306 | |
| 307 | return $filterConditions; |
| 308 | } |
| 309 | |
| 310 | /** |
| 311 | * This method takes a list of contributors along with the total number of bytes they added for each page, and |
| 312 | * returns a list of the same users, ranked by the likelihood of them being interested in the event. |
| 313 | * |
| 314 | * @param array[] $revisionsByWiki |
| 315 | * @phpcs:ignore Generic.Files.LineLength |
| 316 | * @phan-param array<string,list<array{username:string,userID:int,actorID:int,page:string,delta:int}>> $revisionsByWiki |
| 317 | * @return array<string,int> List of users along with their score, sorted from highest to lowest. |
| 318 | */ |
| 319 | private function rankUsers( array $revisionsByWiki ): array { |
| 320 | $deltasByUser = $this->getDeltasByUser( $revisionsByWiki ); |
| 321 | $userDataByWiki = $this->getUserDataByWiki( $revisionsByWiki ); |
| 322 | $this->output( "==Scoring debug info==\n" ); |
| 323 | $rankedUsers = []; |
| 324 | foreach ( $deltasByUser as $username => $byteDeltas ) { |
| 325 | // Make sure the username is a string to satisfy the type hint. PHP will have transformed it to an integer |
| 326 | // if the username is numeric (when used as array key). |
| 327 | $score = $this->getUserScore( (string)$username, $byteDeltas, $userDataByWiki[$username] ); |
| 328 | $rankedUsers[$username] = $score; |
| 329 | } |
| 330 | arsort( $rankedUsers ); |
| 331 | $this->output( "\n\n" ); |
| 332 | return $rankedUsers; |
| 333 | } |
| 334 | |
| 335 | /** |
| 336 | * @param array[] $revisionsByWiki |
| 337 | * * @phpcs:ignore Generic.Files.LineLength |
| 338 | * * @phan-param array<string,list<array{username:string,userID:int,actorID:int,page:string,delta:int}>> $revisionsByWiki |
| 339 | * @return array<string,int[]> For each user, this contains a list of deltas in bytes across all relevant pages. |
| 340 | */ |
| 341 | private function getDeltasByUser( array $revisionsByWiki ): array { |
| 342 | $listByUser = []; |
| 343 | // Flatten the list, merging revisions from all wikis. |
| 344 | $revisions = array_merge( ...array_values( $revisionsByWiki ) ); |
| 345 | foreach ( $revisions as [ 'username' => $username, 'page' => $pageKey, 'delta' => $delta ] ) { |
| 346 | $listByUser[$username] ??= []; |
| 347 | $listByUser[$username][$pageKey] ??= 0; |
| 348 | $listByUser[$username][$pageKey] += $delta; |
| 349 | } |
| 350 | |
| 351 | $deltas = []; |
| 352 | $this->output( "==Contributions==\n" ); |
| 353 | foreach ( $listByUser as $user => $userDeltas ) { |
| 354 | foreach ( $userDeltas as $pageKey => $delta ) { |
| 355 | $this->output( "$user - $pageKey - $delta\n" ); |
| 356 | } |
| 357 | // TODO: What should we do with negative totals? Large negative deltas do not necessarily indicate that a |
| 358 | // user is not interested in the article. This problem is somewhat mitigated by the exclusion of reverts, |
| 359 | // but there are still situations where a negative delta might be a good thing. For instance, if someone has |
| 360 | // moved a section of the article to a separate page. In general, the byte count itself is far from being |
| 361 | // perfect as a metric. For now, we're excluding negative deltas because some of the formulas below expect |
| 362 | // the total delta to be positive. |
| 363 | $positiveDeltas = array_filter( $userDeltas, static fn ( $x ) => $x > 0 ); |
| 364 | if ( $positiveDeltas ) { |
| 365 | $deltas[$user] = array_values( $positiveDeltas ); |
| 366 | } |
| 367 | } |
| 368 | $this->output( "\n\n" ); |
| 369 | return $deltas; |
| 370 | } |
| 371 | |
| 372 | /** |
| 373 | * Returns user identifiers (name, ID, actor ID) for each contributor, for each wiki where they made edits to |
| 374 | * articles in the worklist. This can't just use UserIdentity because that doesn't include the actor ID, which we |
| 375 | * need for other queries later (particularly in getDaysSinceLastEdit()). Alternatively we could use |
| 376 | * ActorNormalization or a join on the user table, but both seem unnecessary (and potentially slow) when we already |
| 377 | * have the actor ID available. |
| 378 | * |
| 379 | * @param array[] $revisionsByWiki |
| 380 | * @phpcs:ignore Generic.Files.LineLength |
| 381 | * @phan-param array<string,list<array{username:string,userID:int,actorID:int,page:string,delta:int}>> $revisionsByWiki |
| 382 | * @return int[][][] Indexed by username first, then wiki ID. |
| 383 | * @phan-return array<string,array<string,array{userID:int,actorID:int}>> |
| 384 | */ |
| 385 | private function getUserDataByWiki( array $revisionsByWiki ): array { |
| 386 | $userData = []; |
| 387 | foreach ( $revisionsByWiki as $wiki => $revisions ) { |
| 388 | foreach ( $revisions as [ 'username' => $username, 'userID' => $userID, 'actorID' => $actorID ] ) { |
| 389 | $userData[$username][$wiki] = [ 'userID' => $userID, 'actorID' => $actorID ]; |
| 390 | } |
| 391 | } |
| 392 | return $userData; |
| 393 | } |
| 394 | |
| 395 | /** |
| 396 | * Returns a score from 0 to 100 for a given user. |
| 397 | * |
| 398 | * @param string $username |
| 399 | * @param int[] $byteDeltas |
| 400 | * @param int[][] $userDataByWiki Map of [ wiki => [ userID: int, actorID: int ] ] |
| 401 | * @phan-param array<string,array{userID:int,actorID:int}> $userDataByWiki |
| 402 | * @return int |
| 403 | */ |
| 404 | private function getUserScore( string $username, array $byteDeltas, array $userDataByWiki ): int { |
| 405 | // Idea: Maybe check how many edits each user has for each page, and handle each edit separately. |
| 406 | // This would allow us to better handle outliers like single edits that add a lot of content. |
| 407 | // Unsure how valuable that would be though, because huge edits can represent at least two different things: |
| 408 | // 1 - Automated maintenance operation (e.g., adding archive links via IABot) |
| 409 | // 2 - Substantial additions of content (for example, but not necessarily, upon page creation) |
| 410 | // Our goal is to avoid 1, while catching 2, which might be difficult. Still, if a user has multiple edits for a |
| 411 | // given page, it's more likely that they may have a genuine interest in the article subject, as opposed to them |
| 412 | // performing some mass-maintenance operation that happened to touch a certain article. |
| 413 | |
| 414 | $bytesScore = $this->getOverallBytesScore( $username, $byteDeltas ); |
| 415 | $editCountScore = $this->getEditCountScore( $username, $userDataByWiki ); |
| 416 | $recentActivityScore = $this->getRecentActivityScore( $username, $userDataByWiki ); |
| 417 | |
| 418 | // Once we have a (0, 1) score for each criterion, we combine them to obtain an overall score. This is currently |
| 419 | // doing a weighted geometric mean. Amongst the advantages of the geometric mean is that it's conveniently |
| 420 | // sensitive to small values. In practice, this means that even a single low score (around zero) will bring the |
| 421 | // overall score down to around zero. |
| 422 | $bytesWeight = 4; |
| 423 | $editCountWeight = 1; |
| 424 | $recentActivityWeight = 5; |
| 425 | $overallScore = ( |
| 426 | ( $bytesScore ** $bytesWeight ) * |
| 427 | ( $editCountScore ** $editCountWeight ) * |
| 428 | ( $recentActivityScore ** $recentActivityWeight ) |
| 429 | ) ** ( 1 / ( $bytesWeight + $editCountWeight + $recentActivityWeight ) ); |
| 430 | return (int)round( 100 * $overallScore ); |
| 431 | } |
| 432 | |
| 433 | /** |
| 434 | * Returns a (0, 1) score based on the number and size of contributions that a single user made across all pages |
| 435 | * in the worklist. |
| 436 | * |
| 437 | * @param string $username |
| 438 | * @param int[] $deltas |
| 439 | * @return float |
| 440 | */ |
| 441 | private function getOverallBytesScore( string $username, array $deltas ): float { |
| 442 | // This function computed a (0, 1) score for each page. Then, we get the maximum of those scores and "boost" |
| 443 | // it by using the other scores. Let us indicate the overall scoring function with f(x), where x is a |
| 444 | // k-dimentional vector. Let x_m be the component in x with the maximum value. We then have f(x) = x_m * b(x), |
| 445 | // where b(x) is the boosting function, which outputs values in [ 1, 1 / x_m ]. |
| 446 | // f(x) satisfies the following conditions: |
| 447 | // * f(x) ∈ [0, 1] |
| 448 | // * f( x_1 ) = x_1 (single variable case) |
| 449 | // * f( 0, ..., 0 ) = 0, f( 1, ..., 1 ) = 1 |
| 450 | // * f(x) >= x_m, where the equality holds true iff all components (at most with the exception of x_m) are 0, |
| 451 | // or x_m = 1 |
| 452 | // Note that the case x_m = 0 is defined separately to avoid annoyances with denominators. |
| 453 | // The b(x) currently used is calculated by taking x_m and linearly amplifying it by a factor proportional to |
| 454 | // the second largest component, then iterating the process for every component. This is very empirical, and |
| 455 | // it would be better to scale the amplification based on the actual byte values, by establishing more rigorous |
| 456 | // relationship to determine how much we shuld favour a given total delta being spread across multiple pages, as |
| 457 | // opposed to being concentrated in a single page. Hopefully, this simple approach would suffice for now. |
| 458 | // The two-dimensional version of the function can be visualised in https://www.desmos.com/3d/41a47c8129 |
| 459 | rsort( $deltas ); |
| 460 | $numPages = count( $deltas ); |
| 461 | $maxBytesScore = $this->getBytesScoreForPage( $deltas[0] ); |
| 462 | $this->output( "User $username max bytes $deltas[0] with score $maxBytesScore\n" ); |
| 463 | $bytesScore = $maxBytesScore; |
| 464 | if ( $maxBytesScore !== 0.0 ) { |
| 465 | for ( $i = 1; $i < $numPages; $i++ ) { |
| 466 | $curBytesScore = $this->getBytesScoreForPage( $deltas[$i] ); |
| 467 | if ( $curBytesScore === 0.0 ) { |
| 468 | // Scores from here on no longer have any effect. |
| 469 | break; |
| 470 | } |
| 471 | $this->output( "User $username bytes score #$i: $curBytesScore\n" ); |
| 472 | $damping = 1; |
| 473 | $bytesScore *= 1 + ( $curBytesScore ** $damping ) * ( 1 / $bytesScore - 1 ); |
| 474 | } |
| 475 | } |
| 476 | $this->output( "User $username overall bytes score $bytesScore\n" ); |
| 477 | return $bytesScore; |
| 478 | } |
| 479 | |
| 480 | /** |
| 481 | * Returns a (0, 1) score based on the contributions made to a single page. |
| 482 | * |
| 483 | * @param int $delta |
| 484 | * @return float |
| 485 | */ |
| 486 | private function getBytesScoreForPage( int $delta ): float { |
| 487 | // Because we use bytes as the main metric in determining the overall score, it's important that the score |
| 488 | // function is as good as possible. This is a logistic-like model, but multiplied by a function that's really |
| 489 | // flat near 0, which acts as a sort of high-pass filter. |
| 490 | // The values for the two parameters have been computed numerically via gradient descent in order to minimize |
| 491 | // the sum of squared residuals. Then, they've been approximated to more readable values. Both the original fit |
| 492 | // and the rounded fit can be visualized in https://www.desmos.com/calculator/eu7u0kwkd6 |
| 493 | $scaledX = $delta / 1000; |
| 494 | $baseScore = 2 / ( 1 + exp( -0.42 * ( $scaledX ** 1.1 ) ) ) - 1; |
| 495 | $flatteningFactor = exp( -0.00001 / ( $scaledX ** 10 ) ); |
| 496 | return $baseScore * $flatteningFactor; |
| 497 | } |
| 498 | |
| 499 | /** |
| 500 | * Returns a (0, 1) score based on the edit count of the given user. |
| 501 | * |
| 502 | * @param string $username |
| 503 | * @param int[][] $userDataByWiki Map of [ wiki => [ userID: int, actorID: int ] ] |
| 504 | * @phan-param array<string,array{userID:int,actorID:int}> $userDataByWiki |
| 505 | * @return float |
| 506 | */ |
| 507 | private function getEditCountScore( string $username, array $userDataByWiki ): float { |
| 508 | $editCount = $this->getEditCount( $userDataByWiki ); |
| 509 | // This one uses the same base model as the bytes score, but with different parameters (approximated in the |
| 510 | // same way). The graph can be visualised in https://www.desmos.com/calculator/4mhrtnhf4i |
| 511 | $scaledEC = $editCount / 1000; |
| 512 | $editCountScore = 2 / ( 1 + exp( -3 * ( $scaledEC ** 0.66 ) ) ) - 1; |
| 513 | $this->output( "User $username edit count $editCount, score $editCountScore\n" ); |
| 514 | return $editCountScore; |
| 515 | } |
| 516 | |
| 517 | /** |
| 518 | * @param int[][] $userDataByWiki Map of [ wiki => [ userID: int, actorID: int ] ] |
| 519 | * @phan-param array<string,array{userID:int,actorID:int}> $userDataByWiki |
| 520 | * @return int |
| 521 | */ |
| 522 | private function getEditCount( array $userDataByWiki ): int { |
| 523 | // XXX: UserEditTracker is only available for the local wiki, and the global edit count is a CentralAuth thing |
| 524 | $totalEditCount = 0; |
| 525 | foreach ( $userDataByWiki as $wiki => [ 'userID' => $userID ] ) { |
| 526 | $dbr = MediaWikiServices::getInstance()->getDBLoadBalancerFactory() |
| 527 | ->getReplicaDatabase( $wiki ); |
| 528 | $curWikiEditCount = $dbr->newSelectQueryBuilder() |
| 529 | ->select( 'user_editcount' ) |
| 530 | ->from( 'user' ) |
| 531 | ->where( [ 'user_id' => $userID ] ) |
| 532 | ->caller( __METHOD__ ) |
| 533 | ->fetchField(); |
| 534 | if ( $curWikiEditCount !== null ) { |
| 535 | $totalEditCount += (int)$curWikiEditCount; |
| 536 | } |
| 537 | } |
| 538 | return $totalEditCount; |
| 539 | } |
| 540 | |
| 541 | /** |
| 542 | * Returns a (0, 1) score based on the recent activity (edits) of the given user. |
| 543 | * |
| 544 | * @param string $username |
| 545 | * @param int[][] $userDataByWiki Map of [ wiki => [ userID: int, actorID: int ] ] |
| 546 | * @phan-param array<string,array{userID:int,actorID:int}> $userDataByWiki |
| 547 | * @return float |
| 548 | */ |
| 549 | private function getRecentActivityScore( string $username, array $userDataByWiki ): float { |
| 550 | // This uses a rational function, so that it does not decay exponentially over time. See |
| 551 | // https://www.desmos.com/calculator/vzhiigbxc9 |
| 552 | // Note: we already have a hard cutoff in the revision query (self::CUTOFF_DAYS), so anything before that won't |
| 553 | // even be scored. |
| 554 | $daysSinceLastEdit = $this->getDaysSinceLastEdit( $userDataByWiki ); |
| 555 | $xMonths = $daysSinceLastEdit / 30; |
| 556 | // TODO: This may have to be scaled down. Think of what the overall score would be in function of the recent |
| 557 | // activity score assuming that the other scores are all 1. For instance, with c=7 and d=0.5 to speed up decay, |
| 558 | // and then reducing the recent activity weight. |
| 559 | $recentActivityScore = ( $xMonths ** 2 + 1.2 * $xMonths + 0.7 ) / |
| 560 | ( 0.3 * ( $xMonths ** 3 ) + $xMonths ** 2 + 1.2 * $xMonths + 0.7 ); |
| 561 | $this->output( "User $username last edit $daysSinceLastEdit days ago, score $recentActivityScore\n" ); |
| 562 | return $recentActivityScore; |
| 563 | } |
| 564 | |
| 565 | /** |
| 566 | * @param int[][] $userDataByWiki Map of [ wiki => [ userID: int, actorID: int ] ] |
| 567 | * @phan-param array<string,array{userID:int,actorID:int}> $userDataByWiki |
| 568 | * @return float |
| 569 | */ |
| 570 | private function getDaysSinceLastEdit( array $userDataByWiki ): float { |
| 571 | // XXX: UserEditTracker is only available for the local wiki, so just use its query directly. |
| 572 | $lastEditTS = 0; |
| 573 | foreach ( $userDataByWiki as $wiki => [ 'actorID' => $actorID ] ) { |
| 574 | $dbr = MediaWikiServices::getInstance()->getDBLoadBalancerFactory() |
| 575 | ->getReplicaDatabase( $wiki ); |
| 576 | $curWikiTS = $dbr->newSelectQueryBuilder() |
| 577 | ->select( 'rev_timestamp' ) |
| 578 | ->from( 'revision' ) |
| 579 | ->where( [ 'rev_actor' => $actorID ] ) |
| 580 | ->orderBy( 'rev_timestamp', SelectQueryBuilder::SORT_DESC ) |
| 581 | ->caller( __METHOD__ ) |
| 582 | ->fetchField(); |
| 583 | if ( $curWikiTS ) { |
| 584 | $lastEditTS = max( $lastEditTS, (int)MWTimestamp::convert( TS_UNIX, $curWikiTS ) ); |
| 585 | } |
| 586 | } |
| 587 | if ( $lastEditTS === 0 ) { |
| 588 | $this->fatalError( "No last edit from user who has edits?!" ); |
| 589 | } |
| 590 | return ( MWTimestamp::time() - $lastEditTS ) / ( 60 * 60 * 24 ); |
| 591 | } |
| 592 | } |
| 593 | |
| 594 | return GenerateInvitationList::class; |