MariadbDialect needs an INSTANCE property.

[SledgeHammer01]

MariaDbDialect extends MySqlDialect, but does not implement it's own public static INSTANCE that returns a MariaDbDialect instance. This makes MariaDbDialect.INSTANCE indistinguishable from MySqlDialect.INSTANCE using instanceof.

We can create it manually, i.e.

new MariaDbDialect(MySqlDialect.MYSQL_IDENTIFIER_PROCESSING)

But it would be nice for MariaDbDialect to have that out of the box. If using a starter, then it would be problematic to create the dialect "the right way".

[schauder]

I'm not convinced a singleton instance makes sense for the JdbcMariaDbDialect or related variants, since the IdentifierProcessing for MariaDb is highly dynamic:

From https://mariadb.com/kb/en/identifier-names :

Identifiers may be quoted using the backtick character - `. Quoting is optional for identifiers that don't contain special characters, or for identifiers that are not reserved words. If the ANSI_QUOTES SQL_MODE flag is set, double quotes (") can also be used to quote identifiers. If the MSSQL flag is set, square brackets ([ and ]) can be used for quoting.

From https://mariadb.com/kb/en/identifier-case-sensitivity/ :

Whether objects are case-sensitive or not is partly determined by the underlying operating system. Unix-based systems are case-sensitive, Windows is not, while Mac OS X is usually case-insensitive by default, but devices can be configured as case-sensitive using Disk Utility.

Regarding

If using a starter, then it would be problematic to create the dialect "the right way".

The "correct" is to use the DialectResolver which uses JDBC metadata to determine the correct IdentifierProcessing.

Is there something else I'm missing?

[SledgeHammer01]

Aren't most of those things true for MySQL too? MySQL and MariaDB are pretty much the same code base. MySQL has a singleton instance.

[schauder]

That is correct, but would be more of an argument for removing that INSTANCE