In a past life I developed an MFC app that was a configuration editor for a complex search engine.  The engine's configuration, expressed as a large graph of objects, was serialized to persistent storage as a collection of XML files.

Reading in the files into the data model was easy enough, but I soon found myself having to deal with missing (i.e. optional) elements and attributes.  I resorted to creating wrapper classes for these potentially absent entities (expressed as bools, ints and CStrings).  The wrapper classes had an extra bit that indicated whether the object had a value - i.e. was defined.  Something along the lines of this:

  1   class DefinedBool
  2   {
  3     protected:
  4       // The object's "defined" state.
  5       bool  m_bDefined;
  6   
  7       // The object's value.
  8       bool m_bValue;
  9   
 10       ...
 11   };

If I'd developed the app several years later, I'd have used .NET, and specifically .NET 2.0, which brings with it the concept of a "nullable" value type.  The entire DefinedBool class shown above can be replaced by the following declaration:

  bool? _someBoolAttribute;

The question mark following the data type indicates that the variable can can have a null value.  This declaration is actually shorthand for a nullable instance of a bool which would be defined as:

  System.Nullable<bool> _someBoolAttribute;

We work with _someBoolAttribute by using its HasValue and Value properties, like so:

  1   // Print value of someBoolAttribute
  2   if (_someBoolAttribute.HasValue)
  3       System.Console.Writeline ("someBoolAttribute is {0}",
  4                                  someBoolAttribute.Value.ToString());
  5   else
  6       System.Console.Writeline ("someBoolAttribute undefined");

And there you have it - a simple and elegant feature that allows us to identify a value type's potentially undefined state.