[TASK] Avoid costly database schema queries and class state

Some places in our code base need to deal with database schema
information, which are quite costly doing them over and over again.

This change streamlines occurrences to use the internal cached
schema information (`Connection->getSchemaInformation()`) to
replace the occurrences and centralizing them to one place. With
that, instance state can be removed in two cases and in one case
the uncached usage replaced by the cached schema information.

`SchemaInformation->listTableColumns()` is added to mimic the
`AbstractSchemaManager->listTableColumns()` method and return
same structure `array<string, Column>` instead of `Column[]`,
considerable not a feature and doable as task because the
complete class is marked as `@internal`.

Note that `SchemaInformation` uses only a code cache right now
and will get a second level (`runtime`) cache in a dedicated
change eliminating the need to read it from PHP files over and
over again.

Resolves: #107390
Releases: main
Change-Id: Ia7e60a0691eebf91a2fc284c17b7f424efd406c6
Reviewed-on: https://review.typo3.org/c/Packages/TYPO3.CMS/+/90562
Tested-by: Stefan Froemken <froemken@gmail.com>
Tested-by: Benni Mack <benni@typo3.org>
Reviewed-by: Stefan Froemken <froemken@gmail.com>
Tested-by: Stefan Bürk <stefan@buerk.tech>
Tested-by: core-ci <typo3@b13.com>
Reviewed-by: Stefan Bürk <stefan@buerk.tech>
Reviewed-by: Benni Mack <benni@typo3.org>

Author: sbuerk

typo3/sysext/core/Classes/Database/Schema/SchemaInformation.php

@@ -18,6 +18,7 @@
 namespace TYPO3\CMS\Core\Database\Schema;
 
 use Doctrine\DBAL\Connection;
+use Doctrine\DBAL\Schema\Column;
 use Doctrine\DBAL\Schema\Schema;
 use Doctrine\DBAL\Schema\Table;
 use TYPO3\CMS\Core\Cache\Frontend\FrontendInterface;
@@ -27,6 +28,8 @@
  * specific schema related information. This should only be used in context where no changes are expected to happen.
  *
  * @internal This class is only for internal core usage and is not part of the public core API.
+ * @todo Add `runtime` cache as a second layer cache to reduce filesystem operation due to the used filesystem cache.
+ * @todo Investigate if filesystem cache layer can be removed AFTER runtime level cache has been added.
  */
 final class SchemaInformation
 {
@@ -65,6 +68,19 @@ public function listTableNames(): array
         return $tableNames;
     }
 
+    /**
+     * @return array<string, Column>
+     */
+    public function listTableColumns(string $table): array
+    {
+        $columnsIndexedByName = [];
+        $columns = $this->introspectTable($table)->getColumns();
+        array_walk($columns, static function (Column $column) use (&$columnsIndexedByName): void {
+            $columnsIndexedByName[$column->getName()] = $column;
+        });
+        return $columnsIndexedByName;
+    }
+
     /**
      * Similar to doctrine DBAL/AbstractSchemaManager, but with a cache-layer.
      * This is used core internally to auto-add types, for instance in Connection::insert().

typo3/sysext/core/Classes/Resource/Index/MetaDataRepository.php

@@ -45,13 +45,6 @@ class MetaDataRepository implements SingletonInterface
      */
     protected $tableName = 'sys_file_metadata';
 
-    /**
-     * Internal storage for database table fields
-     *
-     * @var array
-     */
-    protected $tableFields = [];
-
     public function __construct(
         protected readonly EventDispatcherInterface $eventDispatcher,
     ) {}
@@ -228,13 +221,9 @@ public function removeByFileUid($fileUid)
      */
     protected function getTableFields(): array
     {
-        if (empty($this->tableFields)) {
-            $this->tableFields = GeneralUtility::makeInstance(ConnectionPool::class)
-                ->getConnectionForTable($this->tableName)
-                ->createSchemaManager()
-                ->listTableColumns($this->tableName);
-        }
-
-        return $this->tableFields;
+        return GeneralUtility::makeInstance(ConnectionPool::class)
+            ->getConnectionForTable($this->tableName)
+            ->getSchemaInformation()
+            ->listTableColumns($this->tableName);
     }
 }

typo3/sysext/core/Classes/Resource/ProcessedFileRepository.php

@@ -39,14 +39,6 @@ class ProcessedFileRepository implements LoggerAwareInterface, SingletonInterfac
 {
     use LoggerAwareTrait;
 
-    /**
-     * As determining the table columns is a costly operation this is done only once
-     * during runtime and cached afterward.
-     *
-     * @see cleanUnavailableColumns()
-     */
-    protected array $tableColumns = [];
-
     public function __construct(
         protected readonly ResourceFactory $factory,
         protected readonly TaskTypeRegistry $taskTypeRegistry
@@ -329,15 +321,10 @@ protected function createDomainObject(array $databaseRow): ProcessedFile
      */
     protected function cleanUnavailableColumns(array $data): array
     {
-        // As determining the table columns is a costly operation this is done only once during runtime and cached then
-        if ($this->tableColumns === []) {
-            $this->tableColumns = GeneralUtility::makeInstance(ConnectionPool::class)
-                ->getConnectionForTable('sys_file_processedfile')
-                ->createSchemaManager()
-                ->listTableColumns('sys_file_processedfile');
-        }
-
-        return array_intersect_key($data, $this->tableColumns);
+        return array_intersect_key($data, GeneralUtility::makeInstance(ConnectionPool::class)
+            ->getConnectionForTable('sys_file_processedfile')
+            ->getSchemaInformation()
+            ->listTableColumns('sys_file_processedfile'));
     }
 
     /**

typo3/sysext/lowlevel/Classes/Controller/DatabaseIntegrityController.php

@@ -2711,7 +2711,7 @@ protected function search(ServerRequestInterface $request): string
             // Get fields list
             $connection = GeneralUtility::makeInstance(ConnectionPool::class)->getConnectionForTable($table);
             $identifierQuoteCharacter = $this->platformHelper->getIdentifierQuoteCharacter($connection->getDatabasePlatform());
-            $tableColumns = $connection->createSchemaManager()->listTableColumns($table);
+            $tableColumns = $connection->getSchemaInformation()->listTableColumns($table);
             $normalizedTableColumns = [];
             $fields = [];
             foreach ($tableColumns as $column) {