[3.x] Add collation to getAlterDbCharacterSet method of MySQLi and MySQL (PDO) drivers

[richard67]

Pull Request for Issue joomla/joomla-cms#45112

Summary of Changes

This pull request (PR) changes the getAlterDbCharacterSet method of the MySQLi and MySQL (PDO) drivers to add a COLLATE clause to the ALTER DATABASE ... CHARACTER SET ... statement.

This is consistent with the method getCreateDatabaseQuery of the same drivers, see

• https://github.com/joomla-framework/database/blob/3.x-dev/src/Mysqli/MysqliDriver.php#L478-L488
• https://github.com/joomla-framework/database/blob/3.x-dev/src/Mysql/MysqlDriver.php#L382-L392

Both methods are made to be used by the CMS installation and not for specifying any possible character set. They either use utf8mb4or utf8 depending on utf8mb4 support of the database connection.

The default character set and collation of the database are only used as default values when a CREATE TABLE statement does not specify character set and collation.

The CMS does that always, so the database defaults do not have any impact on the CMS tables.

However, if you install a 3rd party extension which does not specify character set and collation in the CREATE TABLE statements in its installation SQL script, then these tables are created with a default collation which seems to depend on the MySQL or MariaDB server version (on older ones e.g. utf8mb4_general_ci, on newer ones e.g. utf8mb4_0900_ai_ci).

This can lead to "Illegal mix of collations" (or so) errors when joining text columns of these tables with text columns of core tables, or when using them in a UNION clause of a query.

Request for Comments

When this PR here is merged and the change is pulled into the CMS, tables of 3rd party extensions which do not specify character set and collation in their CREATE TABLE statements will be created with collation utf8mb4_unicode_ci (or utf8_unicode_ci if the database doesn't support utf8mb4).

If this is considered a b/c break then I can move this PR to 4.x-dev.

More to be done in a future PR

Since MySQL version 8.0.0 and MariaDB version 10.6.1, utf8 is replaced by utf8mb3, utf8 is a an alias for utf8mb3, and that alias is deprecated and might be removed in future.

Testing Instructions

See joomla/joomla-cms#45112 .

Documentation Changes Required

None.