Code Coverage |
||||||||||
Lines |
Functions and Methods |
Classes and Traits |
||||||||
| Total | |
87.86% |
123 / 140 |
|
69.23% |
9 / 13 |
CRAP | |
0.00% |
0 / 1 |
| NameTableStore | |
87.86% |
123 / 140 |
|
69.23% |
9 / 13 |
31.61 | |
0.00% |
0 / 1 |
| __construct | |
100.00% |
10 / 10 |
|
100.00% |
1 / 1 |
1 | |||
| getDBConnection | |
100.00% |
1 / 1 |
|
100.00% |
1 / 1 |
1 | |||
| getCacheKey | |
100.00% |
5 / 5 |
|
100.00% |
1 / 1 |
1 | |||
| normalizeName | |
100.00% |
3 / 3 |
|
100.00% |
1 / 1 |
2 | |||
| acquireId | |
66.67% |
10 / 15 |
|
0.00% |
0 / 1 |
3.33 | |||
| reloadMap | |
100.00% |
6 / 6 |
|
100.00% |
1 / 1 |
1 | |||
| getId | |
100.00% |
6 / 6 |
|
100.00% |
1 / 1 |
2 | |||
| getName | |
77.42% |
24 / 31 |
|
0.00% |
0 / 1 |
8.74 | |||
| getMap | |
100.00% |
1 / 1 |
|
100.00% |
1 / 1 |
1 | |||
| getTableFromCachesOrReplica | |
100.00% |
13 / 13 |
|
100.00% |
1 / 1 |
2 | |||
| loadTable | |
100.00% |
12 / 12 |
|
100.00% |
1 / 1 |
2 | |||
| store | |
86.67% |
26 / 30 |
|
0.00% |
0 / 1 |
3.02 | |||
| getFieldsToStore | |
85.71% |
6 / 7 |
|
0.00% |
0 / 1 |
3.03 | |||
| 1 | <?php |
| 2 | /** |
| 3 | * @license GPL-2.0-or-later |
| 4 | * @file |
| 5 | */ |
| 6 | |
| 7 | namespace MediaWiki\Storage; |
| 8 | |
| 9 | use Psr\Log\LoggerInterface; |
| 10 | use Wikimedia\Assert\Assert; |
| 11 | use Wikimedia\ObjectCache\BagOStuff; |
| 12 | use Wikimedia\ObjectCache\WANObjectCache; |
| 13 | use Wikimedia\Rdbms\Database; |
| 14 | use Wikimedia\Rdbms\IDatabase; |
| 15 | use Wikimedia\Rdbms\ILoadBalancer; |
| 16 | use Wikimedia\Rdbms\IReadableDatabase; |
| 17 | |
| 18 | /** |
| 19 | * @since 1.31 |
| 20 | * @author Addshore |
| 21 | */ |
| 22 | class NameTableStore { |
| 23 | |
| 24 | /** @var ILoadBalancer */ |
| 25 | private $loadBalancer; |
| 26 | |
| 27 | /** @var WANObjectCache */ |
| 28 | private $cache; |
| 29 | |
| 30 | /** @var LoggerInterface */ |
| 31 | private $logger; |
| 32 | |
| 33 | /** @var array<int,string>|null */ |
| 34 | private $tableCache = null; |
| 35 | |
| 36 | /** @var bool|string */ |
| 37 | private $domain; |
| 38 | |
| 39 | /** @var int */ |
| 40 | private $cacheTTL; |
| 41 | |
| 42 | /** @var string */ |
| 43 | private $table; |
| 44 | /** @var string */ |
| 45 | private $idField; |
| 46 | /** @var string */ |
| 47 | private $nameField; |
| 48 | /** @var null|callable */ |
| 49 | private $normalizationCallback; |
| 50 | /** @var null|callable */ |
| 51 | private $insertCallback; |
| 52 | |
| 53 | /** |
| 54 | * @param ILoadBalancer $dbLoadBalancer A load balancer for acquiring database connections |
| 55 | * @param WANObjectCache $cache A cache manager for caching data. This can be the local |
| 56 | * wiki's default instance even if $dbDomain refers to a different wiki, since |
| 57 | * makeGlobalKey() is used to constructed a key that allows cached names from |
| 58 | * the same database to be re-used between wikis. For example, enwiki and frwiki will |
| 59 | * use the same cache keys for names from the wikidatawiki database, regardless |
| 60 | * of the cache's default key space. |
| 61 | * @param LoggerInterface $logger |
| 62 | * @param string $table |
| 63 | * @param string $idField |
| 64 | * @param string $nameField |
| 65 | * @param callable|null $normalizationCallback Normalization to be applied to names before being |
| 66 | * saved or queried. This should be a callback that accepts and returns a single string. |
| 67 | * @param bool|string $dbDomain Database domain ID. Use false for the local database domain. |
| 68 | * @param callable|null $insertCallback Callback to change insert fields accordingly. |
| 69 | * This parameter was introduced in 1.32 |
| 70 | */ |
| 71 | public function __construct( |
| 72 | ILoadBalancer $dbLoadBalancer, |
| 73 | WANObjectCache $cache, |
| 74 | LoggerInterface $logger, |
| 75 | $table, |
| 76 | $idField, |
| 77 | $nameField, |
| 78 | ?callable $normalizationCallback = null, |
| 79 | $dbDomain = false, |
| 80 | ?callable $insertCallback = null |
| 81 | ) { |
| 82 | $this->loadBalancer = $dbLoadBalancer; |
| 83 | $this->cache = $cache; |
| 84 | $this->logger = $logger; |
| 85 | $this->table = $table; |
| 86 | $this->idField = $idField; |
| 87 | $this->nameField = $nameField; |
| 88 | $this->normalizationCallback = $normalizationCallback; |
| 89 | $this->domain = $dbDomain; |
| 90 | $this->cacheTTL = BagOStuff::TTL_MONTH; |
| 91 | $this->insertCallback = $insertCallback; |
| 92 | } |
| 93 | |
| 94 | /** |
| 95 | * @param int $index A database index, like DB_PRIMARY or DB_REPLICA |
| 96 | * @param int $flags Database connection flags |
| 97 | * @return IDatabase |
| 98 | */ |
| 99 | private function getDBConnection( $index, $flags = 0 ) { |
| 100 | return $this->loadBalancer->getConnection( $index, [], $this->domain, $flags ); |
| 101 | } |
| 102 | |
| 103 | /** |
| 104 | * Gets the cache key for names. |
| 105 | * |
| 106 | * The cache key is constructed based on the wiki ID passed to the constructor, and allows |
| 107 | * sharing of name tables cached for a specific database between wikis. |
| 108 | * |
| 109 | * @return string |
| 110 | */ |
| 111 | private function getCacheKey() { |
| 112 | return $this->cache->makeGlobalKey( |
| 113 | 'NameTableSqlStore', |
| 114 | $this->table, |
| 115 | $this->loadBalancer->resolveDomainID( $this->domain ) |
| 116 | ); |
| 117 | } |
| 118 | |
| 119 | /** |
| 120 | * @param string $name |
| 121 | * @return string |
| 122 | */ |
| 123 | private function normalizeName( $name ) { |
| 124 | if ( $this->normalizationCallback === null ) { |
| 125 | return $name; |
| 126 | } |
| 127 | return ( $this->normalizationCallback )( $name ); |
| 128 | } |
| 129 | |
| 130 | /** |
| 131 | * Acquire the id of the given name. |
| 132 | * This creates a row in the table if it doesn't already exist. |
| 133 | * |
| 134 | * @note If called within an atomic section, there is a chance for the acquired ID to be |
| 135 | * lost on rollback. There is no guarantee that an ID returned by this method is valid |
| 136 | * outside the transaction in which it was produced. This means that calling code should |
| 137 | * not retain the return value beyond the scope of a transaction, but rather call acquireId() |
| 138 | * again after the transaction is complete. In some rare cases, this may produce an ID |
| 139 | * different from the first call. |
| 140 | * |
| 141 | * @param string $name |
| 142 | * @throws NameTableAccessException |
| 143 | * @return int |
| 144 | */ |
| 145 | public function acquireId( string $name ) { |
| 146 | $name = $this->normalizeName( $name ); |
| 147 | |
| 148 | $table = $this->getTableFromCachesOrReplica(); |
| 149 | $searchResult = array_search( $name, $table, true ); |
| 150 | if ( $searchResult === false ) { |
| 151 | $id = $this->store( $name ); |
| 152 | |
| 153 | if ( isset( $table[$id] ) ) { |
| 154 | // This can happen when a name is assigned an ID within a transaction due to |
| 155 | // CONN_TRX_AUTOCOMMIT being unable to use a separate connection (e.g. SQLite). |
| 156 | // The right thing to do in this case is to discard the old value. According to |
| 157 | // the contract of acquireId, the caller should not have used it outside the |
| 158 | // transaction, so it should not be persisted anywhere after the rollback. |
| 159 | $m = "Got ID $id for '$name' from insert" |
| 160 | . " into '{$this->table}', but ID $id was previously associated with" |
| 161 | . " the name '{$table[$id]}'. Overriding the old value, which presumably" |
| 162 | . " has been removed from the database due to a transaction rollback."; |
| 163 | $this->logger->warning( $m ); |
| 164 | } |
| 165 | |
| 166 | $table[$id] = $name; |
| 167 | $searchResult = $id; |
| 168 | |
| 169 | $this->tableCache = $table; |
| 170 | } |
| 171 | |
| 172 | return $searchResult; |
| 173 | } |
| 174 | |
| 175 | /** |
| 176 | * Reloads the name table from the primary database, and purges the WAN cache entry. |
| 177 | * |
| 178 | * @note This should only be called in situations where the local cache has been detected |
| 179 | * to be out of sync with the database. There should be no reason to call this method |
| 180 | * from outside the NameTableStore during normal operation. This method may however be |
| 181 | * useful in unit tests. |
| 182 | * |
| 183 | * @param int $connFlags ILoadBalancer::CONN_XXX flags. Optional. |
| 184 | * |
| 185 | * @return string[] The freshly reloaded name map |
| 186 | */ |
| 187 | public function reloadMap( $connFlags = 0 ) { |
| 188 | $dbw = $this->getDBConnection( DB_PRIMARY, $connFlags ); |
| 189 | $this->tableCache = $this->loadTable( $dbw ); |
| 190 | $dbw->onTransactionPreCommitOrIdle( function () { |
| 191 | $this->cache->delete( $this->getCacheKey() ); |
| 192 | }, __METHOD__ ); |
| 193 | |
| 194 | return $this->tableCache; |
| 195 | } |
| 196 | |
| 197 | /** |
| 198 | * Get the id of the given name. |
| 199 | * If the name doesn't exist this will throw. |
| 200 | * This should be used in cases where we believe the name already exists or want to check for |
| 201 | * existence. |
| 202 | * |
| 203 | * @param string $name |
| 204 | * @throws NameTableAccessException The name does not exist |
| 205 | * @return int Id |
| 206 | */ |
| 207 | public function getId( string $name ) { |
| 208 | $name = $this->normalizeName( $name ); |
| 209 | |
| 210 | $table = $this->getTableFromCachesOrReplica(); |
| 211 | $searchResult = array_search( $name, $table, true ); |
| 212 | |
| 213 | if ( $searchResult !== false ) { |
| 214 | return $searchResult; |
| 215 | } |
| 216 | |
| 217 | throw NameTableAccessException::newFromDetails( $this->table, 'name', $name ); |
| 218 | } |
| 219 | |
| 220 | /** |
| 221 | * Get the name of the given id. |
| 222 | * If the id doesn't exist this will throw. |
| 223 | * This should be used in cases where we believe the id already exists. |
| 224 | * |
| 225 | * Note: Calls to this method will result in a primary DB select for non existing IDs. |
| 226 | * |
| 227 | * @param int $id |
| 228 | * @throws NameTableAccessException The id does not exist |
| 229 | * @return string name |
| 230 | */ |
| 231 | public function getName( int $id ) { |
| 232 | $table = $this->getTableFromCachesOrReplica(); |
| 233 | if ( array_key_exists( $id, $table ) ) { |
| 234 | return $table[$id]; |
| 235 | } |
| 236 | $fname = __METHOD__; |
| 237 | |
| 238 | $table = $this->cache->getWithSetCallback( |
| 239 | $this->getCacheKey(), |
| 240 | $this->cacheTTL, |
| 241 | function ( $oldValue, &$ttl, &$setOpts ) use ( $id, $fname ) { |
| 242 | // Check if cached value is up-to-date enough to have $id |
| 243 | if ( is_array( $oldValue ) && array_key_exists( $id, $oldValue ) ) { |
| 244 | // Completely leave the cache key alone |
| 245 | $ttl = WANObjectCache::TTL_UNCACHEABLE; |
| 246 | // Use the old value |
| 247 | return $oldValue; |
| 248 | } |
| 249 | // Regenerate from replica DB, and primary DB if needed |
| 250 | foreach ( [ DB_REPLICA, DB_PRIMARY ] as $source ) { |
| 251 | // Log a fallback to primary |
| 252 | if ( $source === DB_PRIMARY ) { |
| 253 | $this->logger->info( |
| 254 | $fname . ' falling back to primary select from ' . |
| 255 | $this->table . ' with id ' . $id |
| 256 | ); |
| 257 | } |
| 258 | $db = $this->getDBConnection( $source ); |
| 259 | $cacheSetOpts = Database::getCacheSetOptions( $db ); |
| 260 | $table = $this->loadTable( $db ); |
| 261 | if ( array_key_exists( $id, $table ) ) { |
| 262 | break; // found it |
| 263 | } |
| 264 | } |
| 265 | // Use the value from last source checked |
| 266 | $setOpts += $cacheSetOpts; |
| 267 | |
| 268 | return $table; |
| 269 | }, |
| 270 | [ 'minAsOf' => INF ] // force callback run |
| 271 | ); |
| 272 | |
| 273 | $this->tableCache = $table; |
| 274 | |
| 275 | if ( array_key_exists( $id, $table ) ) { |
| 276 | return $table[$id]; |
| 277 | } |
| 278 | |
| 279 | throw NameTableAccessException::newFromDetails( $this->table, 'id', $id ); |
| 280 | } |
| 281 | |
| 282 | /** |
| 283 | * Get the whole table, in no particular order as a map of ids to names. |
| 284 | * This method could be subject to DB or cache lag. |
| 285 | * |
| 286 | * @return string[] keys are the name ids, values are the names themselves |
| 287 | * Example: [ 1 => 'foo', 3 => 'bar' ] |
| 288 | */ |
| 289 | public function getMap() { |
| 290 | return $this->getTableFromCachesOrReplica(); |
| 291 | } |
| 292 | |
| 293 | /** |
| 294 | * @return array<int,string> |
| 295 | */ |
| 296 | private function getTableFromCachesOrReplica() { |
| 297 | if ( $this->tableCache !== null ) { |
| 298 | return $this->tableCache; |
| 299 | } |
| 300 | |
| 301 | $table = $this->cache->getWithSetCallback( |
| 302 | $this->getCacheKey(), |
| 303 | $this->cacheTTL, |
| 304 | function ( $oldValue, &$ttl, &$setOpts ) { |
| 305 | $dbr = $this->getDBConnection( DB_REPLICA ); |
| 306 | $setOpts += Database::getCacheSetOptions( $dbr ); |
| 307 | return $this->loadTable( $dbr ); |
| 308 | } |
| 309 | ); |
| 310 | |
| 311 | $this->tableCache = $table; |
| 312 | |
| 313 | return $table; |
| 314 | } |
| 315 | |
| 316 | /** |
| 317 | * Gets the table from the db |
| 318 | * |
| 319 | * @param IReadableDatabase $db |
| 320 | * @return array<int,string> |
| 321 | */ |
| 322 | private function loadTable( IReadableDatabase $db ) { |
| 323 | $result = $db->newSelectQueryBuilder() |
| 324 | ->select( [ |
| 325 | 'id' => $this->idField, |
| 326 | 'name' => $this->nameField |
| 327 | ] ) |
| 328 | ->from( $this->table ) |
| 329 | ->orderBy( 'id' ) |
| 330 | ->caller( __METHOD__ )->fetchResultSet(); |
| 331 | |
| 332 | $assocArray = []; |
| 333 | foreach ( $result as $row ) { |
| 334 | $assocArray[(int)$row->id] = $row->name; |
| 335 | } |
| 336 | |
| 337 | return $assocArray; |
| 338 | } |
| 339 | |
| 340 | /** |
| 341 | * Stores the given name in the DB, returning the ID when an insert occurs. |
| 342 | * |
| 343 | * @param string $name |
| 344 | * @return int The new or colliding ID |
| 345 | */ |
| 346 | private function store( string $name ) { |
| 347 | Assert::parameter( $name !== '', '$name', 'should not be an empty string' ); |
| 348 | // Note: this is only called internally so normalization of $name has already occurred. |
| 349 | |
| 350 | $dbw = $this->getDBConnection( DB_PRIMARY, ILoadBalancer::CONN_TRX_AUTOCOMMIT ); |
| 351 | |
| 352 | $dbw->newInsertQueryBuilder() |
| 353 | ->insertInto( $this->table ) |
| 354 | ->ignore() |
| 355 | ->row( $this->getFieldsToStore( $name ) ) |
| 356 | ->caller( __METHOD__ )->execute(); |
| 357 | |
| 358 | if ( $dbw->affectedRows() > 0 ) { |
| 359 | $id = $dbw->insertId(); |
| 360 | // As store returned an ID we know we inserted so delete from WAN cache |
| 361 | $dbw->onTransactionPreCommitOrIdle( |
| 362 | function () { |
| 363 | $this->cache->delete( $this->getCacheKey() ); |
| 364 | }, |
| 365 | __METHOD__ |
| 366 | ); |
| 367 | |
| 368 | return $id; |
| 369 | } |
| 370 | |
| 371 | $this->logger->info( |
| 372 | 'Tried to insert name into table ' . $this->table . ', but value already existed.' |
| 373 | ); |
| 374 | |
| 375 | // Note that in MySQL, even if this method somehow runs in a transaction, a plain |
| 376 | // (non-locking) SELECT will see the new row created by the other transaction, even |
| 377 | // with REPEATABLE-READ. This is due to how "consistent reads" works: the latest |
| 378 | // version of rows become visible to the snapshot after the transaction sees those |
| 379 | // rows as either matching an update query or conflicting with an insert query. |
| 380 | $id = $dbw->newSelectQueryBuilder() |
| 381 | ->select( [ 'id' => $this->idField ] ) |
| 382 | ->from( $this->table ) |
| 383 | ->where( [ $this->nameField => $name ] ) |
| 384 | ->caller( __METHOD__ )->fetchField(); |
| 385 | |
| 386 | if ( $id === false ) { |
| 387 | // Insert failed due to IGNORE flag, but DB_PRIMARY didn't give us the data |
| 388 | $m = "No insert possible but primary DB didn't give us a record for " . |
| 389 | "'{$name}' in '{$this->table}'"; |
| 390 | $this->logger->error( $m ); |
| 391 | throw new NameTableAccessException( $m ); |
| 392 | } |
| 393 | |
| 394 | return (int)$id; |
| 395 | } |
| 396 | |
| 397 | /** |
| 398 | * @param string $name |
| 399 | * @param int|null $id |
| 400 | * @return array |
| 401 | */ |
| 402 | private function getFieldsToStore( $name, $id = null ) { |
| 403 | $fields = []; |
| 404 | |
| 405 | $fields[$this->nameField] = $name; |
| 406 | |
| 407 | if ( $id !== null ) { |
| 408 | $fields[$this->idField] = $id; |
| 409 | } |
| 410 | |
| 411 | if ( $this->insertCallback !== null ) { |
| 412 | $fields = ( $this->insertCallback )( $fields ); |
| 413 | } |
| 414 | return $fields; |
| 415 | } |
| 416 | |
| 417 | } |