In the Torque Runtime, all information about a specific Database is gathered in a so-called Database adapter class. So if you want to support a new database in the runtime, you need to provide a Database Adapter class for this database.

If you are adding support for a new RDBMS, then you will probably also want to support the database in the Torque generator. To do this, you need to create a set of Velocity templates--used by Torque to generate a SQL schema for your RDBMS--in the templates component. The recommend method for doing this is to copy an existing set of templates and adapt them to your RDBMS as needed. This is not elaborated further here.

A database adapter class is a class that implements org.apache.torque.adapter.DB and encapsulates access to a specific RDBMS implementation. Database adapter classes already found in Torque include DBOracle, DBMM, DBSybase, etc. These classes allow Torque to gain access to a wide range of databases in a uniform manner. This allows you to easily swap between databases without any modification to Torque or the application built on top of Torque.

Why is this necessary if Java already offers uniform database access in the form of JDBC? Unfortunately, underlying databases still use different SQL implementations and conventions. For example, the use of single and double quotes varies. The use of database adapter classes in Torque endeavors to overcome this problem.

To add a new database adapter class to Torque you must follow these steps:

• Create a new class DB<dbname> that implements org.apache.torque.adapter.DB (where dbname is the name of the database or database driver you wish to add to Torque). An abstract implementation, org.apache.torque.adapter.AbstractDBAdapter is provided which sets some defaults so that you only need to implement a number of methods.
• Implement getStringDelimiter(). This method has to return the character that the specific database implementation uses to indicate string literals in SQL. This will usually be a single qoute (').
• Implement generateLimits(). This method has to provide the special SQL expressions to limit the number of records returned and/or the offset into the result set. This is only needed if supportsNativeLimit() or supportsNativeOffset() return true.
• Implement getIdMethodType(). This method should return the method the database uses to generates unique Ids. Valid return values are org.apache.torque.adapter.IDMethod.AUTO_INCREMENT, org.apache.torque.adapter.IDMethod.SEQUENCE, and org.apache.torque.adapter.IDMethod.NO_ID_METHOD.
• Implement String getIDMethodSQL(Object obj). Databases that use sequences for Id generation should return the SQL to get the next Id from the sequence, where obj is the name of the sequence. Databases that use auto increment fields should return the last generated id; obj is the name of the table in this case. Databases that do not support this must return null.
• Implement lockTable(Connection con, String table). This method should place a lock on a database table, if this is possible for the database. This method is not used internally by Torque, and was only retained to provide backwards compatibility.
• Implement unlockTable(Connection con, String table). This method should release the table lock acquired by the above-mentioned method, if possible. This method is not used internally by Torque, and was only retained to provide backwards compatibility.
• Implement ignoreCase(String in). This method should implement a mechanism for case insensitive comparisons in the database. Usually this converts the string to Uppercase, for example UPPER (<in>). If such a mechanism does not exist in your database, simply return the in parameter without any modification. DO NOT return in.toUpper().
• Some databases (for example Interbase) do not support the function returned by their implementation of ignoreCase() in the ORDER BY clause of SELECT statements. If your database falls into this category you should override the default implementation of String ignoreCaseInOrderBy(String in). It is NOT required to override this method for other databases--the default implementation calls ignoreCase() and respectCase().

The adapter you wrote does not need to be compiled into Torque but it can be referenced in the configuration. Just use a new short name for the adapter and provide the class name in the configuration file. See the following example: