[TASK] Dynamically limit livesearch to searchFields used in subtypes

Previously, the bodytext field, for example, was only searched
if the CType was text, textpic or textmedia. This condition is
unknown to practically all developers which has been spoken to.
The only exception so far seems to be the `EXT:mask` extension,
which already dynamically extends this condition.

It was introduced with TYPO3 v4.6 with #26829 and was pretty
much untouched ever since.

With the introduction of the TCA Schema API, we can handle the
issue much more generically for all tables with subtypes. We
can thus make the search much more precise and only search in
fields that actually exist in the respective subtype like the
CType for example.

The purpose of this search restriction is to ensure that, for
example, the content in the bodytext field is no longer found
if it was filled as a text element and then changed to a plugin.

Resolves: #104539
Related: #26829
Releases: main, 13.4
Change-Id: I490a0b25568ae1919c825553215123d73a959b44
Reviewed-on: https://review.typo3.org/c/Packages/TYPO3.CMS/+/85489
Tested-by: Stefan Bürk <stefan@buerk.tech>
Tested-by: Benni Mack <benni@typo3.org>
Reviewed-by: Stefan Bürk <stefan@buerk.tech>
Reviewed-by: Benni Mack <benni@typo3.org>
Tested-by: core-ci <typo3@b13.com>

Author: kevin-appelt

typo3/sysext/backend/Classes/RecordList/DatabaseRecordList.php

@@ -30,6 +30,7 @@
 use TYPO3\CMS\Backend\RecordList\Event\ModifyRecordListTableActionsEvent;
 use TYPO3\CMS\Backend\Routing\PreviewUriBuilder;
 use TYPO3\CMS\Backend\Routing\UriBuilder;
+use TYPO3\CMS\Backend\Search\LiveSearch\DatabaseRecordProvider;
 use TYPO3\CMS\Backend\Template\Components\Buttons\ButtonInterface;
 use TYPO3\CMS\Backend\Template\Components\Buttons\GenericButton;
 use TYPO3\CMS\Backend\Tree\Repository\PageTreeRepository;
@@ -56,6 +57,7 @@
 use TYPO3\CMS\Core\Schema\Field\DateTimeFieldType;
 use TYPO3\CMS\Core\Schema\Field\NumberFieldType;
 use TYPO3\CMS\Core\Schema\SearchableSchemaFieldsCollector;
+use TYPO3\CMS\Core\Schema\TcaSchemaFactory;
 use TYPO3\CMS\Core\Service\DependencyOrderingService;
 use TYPO3\CMS\Core\Site\Entity\SiteLanguage;
 use TYPO3\CMS\Core\Type\Bitmask\Permission;
@@ -412,6 +414,7 @@ public function __construct(
         protected readonly BackendViewFactory $backendViewFactory,
         protected readonly ModuleProvider $moduleProvider,
         protected readonly SearchableSchemaFieldsCollector $searchableSchemaFieldsCollector,
+        protected readonly TcaSchemaFactory $tcaSchemaFactory,
     ) {
         $this->calcPerms = new Permission();
         $this->spaceIcon = '<span class="btn btn-default disabled" aria-hidden="true">' . $this->iconFactory->getIcon('empty-empty', IconSize::SMALL)->render() . '</span>';
@@ -2599,37 +2602,59 @@ protected function makeSearchString(string $table, int $currentPid, QueryBuilder
         }
 
         $searchableFields = $this->searchableSchemaFieldsCollector->getFields($table);
+        [$subSchemaDivisorFieldName, $fieldsSubSchemaTypes] = $this->getSchemaFieldSubSchemaTypes($table);
         // Get fields from ctrl section of TCA first
         if (MathUtility::canBeInterpretedAsInteger($this->searchString)) {
             $constraints[] = $expressionBuilder->eq('uid', (int)$this->searchString);
             foreach ($searchableFields as $field) {
                 $fieldConfig = $field->getConfiguration();
+                $searchConstraint = null;
                 if ($field instanceof NumberFieldType || $field instanceof DateTimeFieldType) {
                     if (!isset($fieldConfig['search']['pidonly'])
                         || ($fieldConfig['search']['pidonly'] && $currentPid > 0)
                     ) {
-                        $constraints[] = $expressionBuilder->and(
+                        $searchConstraint = $expressionBuilder->and(
                             $expressionBuilder->eq($field->getName(), (int)$this->searchString),
                             $expressionBuilder->eq($tablePidField, $currentPid)
                         );
+                    } else {
+                        continue;
                     }
                 } else {
-                    $constraints[] = $expressionBuilder->like(
+                    $searchConstraint = $expressionBuilder->like(
                         $field->getName(),
                         $queryBuilder->quote('%' . $this->searchString . '%')
                     );
                 }
+
+                // If this table has subtypes (e.g. tt_content.CType), we want to ensure that only CType that contain
+                // e.g. "bodytext" in their list of fields, to search through them. This is important when a field
+                // is filled but its type has been changed.
+                if ($subSchemaDivisorFieldName !== ''
+                    && isset($fieldsSubSchemaTypes[$field->getName()])
+                    && $fieldsSubSchemaTypes[$field->getName()] !== []
+                ) {
+                    // Using `IN()` with a string-value quoted list is fine for all database systems, even when
+                    // used on integer-typed fields and no additional work required here to mitigate something.
+                    $searchConstraint = $queryBuilder->expr()->and(
+                        $searchConstraint,
+                        $queryBuilder->expr()->in(
+                            $subSchemaDivisorFieldName,
+                            $queryBuilder->quoteArrayBasedValueListToStringList($fieldsSubSchemaTypes[$field->getName()])
+                        ),
+                    );
+                }
+
+                $constraints[] = $searchConstraint;
             }
         } elseif ($searchableFields->count() > 0) {
             $like = $queryBuilder->quote('%' . $queryBuilder->escapeLikeWildcards($this->searchString) . '%');
             foreach ($searchableFields as $field) {
                 $fieldConfig = $field->getConfiguration();
-                $searchConstraint = $expressionBuilder->and(
-                    $expressionBuilder->comparison(
-                        'LOWER(' . $queryBuilder->castFieldToTextType($field->getName()) . ')',
-                        'LIKE',
-                        'LOWER(' . $like . ')'
-                    )
+                $searchConstraint = $expressionBuilder->comparison(
+                    'LOWER(' . $queryBuilder->castFieldToTextType($field->getName()) . ')',
+                    'LIKE',
+                    'LOWER(' . $like . ')'
                 );
                 if (is_array($fieldConfig['search'] ?? null)) {
                     $searchConfig = $fieldConfig['search'];
@@ -2638,14 +2663,37 @@ protected function makeSearchString(string $table, int $currentPid, QueryBuilder
                         $searchConstraint = $expressionBuilder->and($expressionBuilder->like($field->getName(), $like));
                     }
                     if (($searchConfig['pidonly'] ?? false) && $currentPid > 0) {
-                        $searchConstraint = $searchConstraint->with($expressionBuilder->eq($tablePidField, (int)$currentPid));
+                        $searchConstraint = $expressionBuilder->and(
+                            $searchConstraint,
+                            $expressionBuilder->eq($tablePidField, (int)$currentPid),
+                        );
                     }
                     if ($searchConfig['andWhere'] ?? false) {
-                        $searchConstraint = $searchConstraint->with(
+                        $searchConstraint = $expressionBuilder->and(
+                            $searchConstraint,
                             QueryHelper::quoteDatabaseIdentifiers($queryBuilder->getConnection(), QueryHelper::stripLogicalOperatorPrefix($fieldConfig['search']['andWhere']))
                         );
                     }
                 }
+
+                // If this table has subtypes (e.g. tt_content.CType), we want to ensure that only CType that contain
+                // e.g. "bodytext" in their list of fields, to search through them. This is important when a field
+                // is filled but its type has been changed.
+                if ($subSchemaDivisorFieldName !== ''
+                    && isset($fieldsSubSchemaTypes[$field->getName()])
+                    && $fieldsSubSchemaTypes[$field->getName()] !== []
+                ) {
+                    // Using `IN()` with a string-value quoted list is fine for all database systems, even when
+                    // used on integer-typed fields and no additional work required here to mitigate something.
+                    $searchConstraint = $queryBuilder->expr()->and(
+                        $searchConstraint,
+                        $queryBuilder->expr()->in(
+                            $subSchemaDivisorFieldName,
+                            $queryBuilder->quoteArrayBasedValueListToStringList($fieldsSubSchemaTypes[$field->getName()])
+                        ),
+                    );
+                }
+
                 $constraints[] = $searchConstraint;
             }
         }
@@ -3420,4 +3468,42 @@ protected function addDividerToCellGroup(array &$cells): void
             $this->addActionToCellGroup($cells, '<hr class="dropdown-divider">', 'divider');
         }
     }
+
+    /**
+     * Returns table subschema divisor field name and a list of fields not included in all subSchemas along with
+     * the list of subSchemas they are included.
+     *
+     * @param string $tableName
+     * @return array{0: string, 1: array<string, list<string>>}
+     * @todo Consider to move this to {@see SearchableSchemaFieldsCollector}, a dedicated trait or a shared place to
+     *       mitigate code duplication (and maintenance in different places).
+     *       - {@see PageRecordProvider::getSchemaFieldSubSchemaTypes()}
+     *       - {@see DatabaseRecordProvider::getSchemaFieldSubSchemaTypes()}
+     */
+    protected function getSchemaFieldSubSchemaTypes(string $tableName): array
+    {
+        $result = [
+            0 => '',
+            1 => [],
+        ];
+        if (!$this->tcaSchemaFactory->has($tableName)) {
+            return $result;
+        }
+        $schema = $this->tcaSchemaFactory->get($tableName);
+        if ($schema->getSubSchemaDivisorField() === null) {
+            return $result;
+        }
+        $result[0] = $schema->getSubSchemaDivisorField()->getName();
+        foreach ($schema->getSubSchemata() as $recordType => $subSchemata) {
+            foreach ($subSchemata->getFields() as $fieldInSubschema => $fieldConfig) {
+                $result[1][$fieldInSubschema] ??= [];
+                $result[1][$fieldInSubschema][] = $recordType;
+            }
+        }
+        // Remove all fields which are contained in all sub-schemas, determined by
+        // comparing each field types count with table types count.
+        $subSchemaCount = count($schema->getSubSchemata());
+        $result[1] = array_filter($result[1], static fn($value) => count($value) < $subSchemaCount);
+        return $result;
+    }
 }

typo3/sysext/backend/Classes/Search/LiveSearch/DatabaseRecordProvider.php

@@ -44,6 +44,7 @@
 use TYPO3\CMS\Core\Schema\Field\DateTimeFieldType;
 use TYPO3\CMS\Core\Schema\Field\NumberFieldType;
 use TYPO3\CMS\Core\Schema\SearchableSchemaFieldsCollector;
+use TYPO3\CMS\Core\Schema\TcaSchemaFactory;
 use TYPO3\CMS\Core\Type\Bitmask\Permission;
 use TYPO3\CMS\Core\Utility\GeneralUtility;
 use TYPO3\CMS\Core\Utility\MathUtility;
@@ -68,6 +69,7 @@ public function __construct(
         protected readonly UriBuilder $uriBuilder,
         protected readonly QueryParser $queryParser,
         protected readonly SearchableSchemaFieldsCollector $searchableSchemaFieldsCollector,
+        protected readonly TcaSchemaFactory $tcaSchemaFactory,
     ) {
         $this->languageService = $this->languageServiceFactory->createFromUserPreferences($this->getBackendUser());
         $this->userPermissions = $this->getBackendUser()->getPagePermsClause(Permission::PAGE_SHOW);
@@ -343,6 +345,7 @@ protected function buildConstraintsForTable(string $queryString, QueryBuilder $q
         $platform = $queryBuilder->getConnection()->getDatabasePlatform();
         $isPostgres = $platform instanceof DoctrinePostgreSQLPlatform;
         $fieldsToSearchWithin = $this->searchableSchemaFieldsCollector->getFields($tableName);
+        [$subSchemaDivisorFieldName, $fieldsSubSchemaTypes] = $this->getSchemaFieldSubSchemaTypes($tableName);
         $constraints = [];
         // If the search string is a simple integer, assemble an equality comparison
         if (MathUtility::canBeInterpretedAsInteger($queryString)) {
@@ -358,19 +361,39 @@ protected function buildConstraintsForTable(string $queryString, QueryBuilder $q
             foreach ($fieldsToSearchWithin as $fieldName => $field) {
                 // Assemble the search condition only if the field is an integer
                 if ($field instanceof NumberFieldType || $field instanceof DateTimeFieldType) {
-                    $constraints[] = $queryBuilder->expr()->eq(
+                    $searchConstraint = $queryBuilder->expr()->eq(
                         $fieldName,
                         $queryBuilder->createNamedParameter($queryString, Connection::PARAM_INT)
                     );
                 } else {
                     // Otherwise assemble a like condition
-                    $constraints[] = $queryBuilder->expr()->like(
+                    $searchConstraint = $queryBuilder->expr()->like(
                         $fieldName,
                         $queryBuilder->createNamedParameter(
                             '%' . $queryBuilder->escapeLikeWildcards($queryString) . '%'
                         )
                     );
                 }
+
+                // If this table has subtypes (e.g. tt_content.CType), we want to ensure that only CType that contain
+                // e.g. "bodytext" in their list of fields, to search through them. This is important when a field
+                // is filled but its type has been changed.
+                if ($subSchemaDivisorFieldName !== ''
+                    && isset($fieldsSubSchemaTypes[$fieldName])
+                    && $fieldsSubSchemaTypes[$fieldName] !== []
+                ) {
+                    // Using `IN()` with a string-value quoted list is fine for all database systems, even when
+                    // used on integer-typed fields and no additional work required here to mitigate something.
+                    $searchConstraint = $queryBuilder->expr()->and(
+                        $searchConstraint,
+                        $queryBuilder->expr()->in(
+                            $subSchemaDivisorFieldName,
+                            $queryBuilder->quoteArrayBasedValueListToStringList($fieldsSubSchemaTypes[$fieldName])
+                        ),
+                    );
+                }
+
+                $constraints[] = $searchConstraint;
             }
         } else {
             $like = '%' . $queryBuilder->escapeLikeWildcards($queryString) . '%';
@@ -413,6 +436,24 @@ protected function buildConstraintsForTable(string $queryString, QueryBuilder $q
                     }
                 }
 
+                // If this table has subtypes (e.g. tt_content.CType), we want to ensure that only CType that contain
+                // e.g. "bodytext" in their list of fields, to search through them. This is important when a field
+                // is filled but its type has been changed.
+                if ($subSchemaDivisorFieldName !== ''
+                    && isset($fieldsSubSchemaTypes[$fieldName])
+                    && $fieldsSubSchemaTypes[$fieldName] !== []
+                ) {
+                    // Using `IN()` with a string-value quoted list is fine for all database systems, even when
+                    // used on integer-typed fields and no additional work required here to mitigate something.
+                    $searchConstraint = $queryBuilder->expr()->and(
+                        $searchConstraint,
+                        $queryBuilder->expr()->in(
+                            $subSchemaDivisorFieldName,
+                            $queryBuilder->quoteArrayBasedValueListToStringList($fieldsSubSchemaTypes[$fieldName])
+                        ),
+                    );
+                }
+
                 $constraints[] = $searchConstraint;
             }
         }
@@ -478,6 +519,44 @@ protected function getEditLink(string $tableName, array $row): string
         return $editLink;
     }
 
+    /**
+     * Returns table subschema divisor field name and a list of fields not included in all subSchemas along with
+     * the list of subSchemas they are included.
+     *
+     * @param string $tableName
+     * @return array{0: string, 1: array<string, list<string>>}
+     * @todo Consider to move this to {@see SearchableSchemaFieldsCollector}, a dedicated trait or a shared place to
+     *       mitigate code duplication (and maintenance in different places).
+     *       - {@see PageRecordProvider::getSchemaFieldSubSchemaTypes()}
+     *       - {@see DatabaseRecordList::getSchemaFieldSubSchemaTypes()}
+     */
+    protected function getSchemaFieldSubSchemaTypes(string $tableName): array
+    {
+        $result = [
+            0 => '',
+            1 => [],
+        ];
+        if (!$this->tcaSchemaFactory->has($tableName)) {
+            return $result;
+        }
+        $schema = $this->tcaSchemaFactory->get($tableName);
+        if ($schema->getSubSchemaDivisorField() === null) {
+            return $result;
+        }
+        $result[0] = $schema->getSubSchemaDivisorField()->getName();
+        foreach ($schema->getSubSchemata() as $recordType => $subSchemata) {
+            foreach ($subSchemata->getFields() as $fieldInSubschema => $fieldConfig) {
+                $result[1][$fieldInSubschema] ??= [];
+                $result[1][$fieldInSubschema][] = $recordType;
+            }
+        }
+        // Remove all fields which are contained in all sub-schemas, determined by
+        // comparing each field types count with table types count.
+        $subSchemaCount = count($schema->getSubSchemata());
+        $result[1] = array_filter($result[1], static fn($value) => count($value) < $subSchemaCount);
+        return $result;
+    }
+
     protected function getBackendUser(): BackendUserAuthentication
     {
         return $GLOBALS['BE_USER'];