The NULL problem

Databases support the notion of NULL - a column with an unspecified value. The ADO.NET containers ( DataReader, DataSet ) support the transport of these untyped objects ( as DBNull ). However the .NET Framework native types ( ie. String, Integer ) do not support them. The new Business Logic Layer transfers data from DataReaders to Custom Business Objects whose properties are comprised of native data types. This results in errors when the DataReader item contains NULL; therefore, a solution is required to deal with this scenario.

Option #1 - Use a translation service

The concept here is that NULL values are translated into application friendly values ( and vice versa ). This translation must occur when data is read from the database as well as when data is sent to the database. The application friendly values should be values which would not normally be used within the application ( ie. for dates - Date.MinValue, for Integers - Integer.MinValue, etc... ). The current DotNetNuke implementation uses this technique in some cases ( representing NULL as -1 for ID values ). The problem is that the translation service must be handled in all application layers - resulting in poorly abstracted design, overly complicated code, and extra processing.

Option #2 - Use Nullable Types

The System.Data.SqlTypes namespace contains SQL Server data types for most native types which also support the INullable interface. This means that these data types support DBNull. However the problem with this namespace is that it is database specific and the types are not Serializable. An alternative may be an open source project called Nullable Types ( http://nullabletypes.sourceforge.net/ ) which contains database independent types that implement INullable as well as ISerializable. This may be a cleaner option ( DBNull handled consistently across all application layers ) except for the fact it adds a project dependency to another open source project. More research is required.

Option #3 - Do not allow NULLs in database tables

Another philosophy is to avoid the NULL issue altogether by not allowing NULLs in your database. This approach relies on the application/database always populating fields with valid data. However, this is not always possible or preferable when working with relational databases ( issues include how to deal with optional foreign key constraints, uninitialized dates, etc... ).

Considering the sheer number of applications which have to deal with this dilemma, you would think there would be a documented best practice solution. However, the deeper you dig into this problem, the more complications become exposed.