For MobiLink synchronization, including direct row handling, your data source must have unique primary keys that are not updated. In a non-relational data source such as a spreadsheet or text file, this means that one column must contain unique, unchanging values that identify the row.

See Unique primary keys.

When using direct row handling, the column names of tables are only available if the MobiLink client is configured to send column names. Alternatively, you can use column indexes to access row information.

To use column names, see:

• SQL Anywhere remotes: SendColumnNames (scn) extended option
• UltraLite remotes: Send Column Names synchronization parameter

If possible, set up your direct row handling application like a timestamp-based SQL application; maintain a last_modified column and download data based on it. This method avoids unforeseen problems that could occur if you use a different download methodology.

See Implementing timestamp-based downloads.

You cannot commit transactions with the MobiLink consolidated database. However, you can commit transactions with your direct row handling data source. When setting up transaction management, keep the following tips in mind:

To handle errors, ensure you employ appropriate transaction management, as described above. In addition, your Java or .NET code that handles rows must send any exception that occurs to the MobiLink server. If an error occurs before the MobiLink server or your application has committed changes, MobiLink rollbacks the transaction and maintains a consistent state with your application.

For direct row handling, MobiLink creates one class instance per database connection. The class instance is not destroyed at the end of a synchronization: it is destroyed when the database connection is closed. Class level variables retain values from previous synchronizations.