Christopher Howard
One of the things that I love about Haskell is the syntax for creating user defined types. E.g.:
Data QueryResult = NoResult | DatabaseError String | Results [String]
I can prototype an entire program around these self-made types and it is a lot of fun. Out of intellect curiosity, though: what would be the equivalent construction in C or C++? (Say, for the type defined above).
I think, the closest and cleanest translation is a set of four classes, one QueryResult superclass with subclasses NoResult, DatabaseError and Results. This also gives you something resembling Haskell's value constructors, although not for free. Finally it helps with separation of concerns, when you write accessor functions. You don't get pattern matching, but you can get combinators like 'maybe' for Maybe or 'either' for Either. Unions work, too, but they are not nearly as clean. In general, avoid unions in C++, even though they seem to be a natural way to express ADTs. They really are not. Class hierarchies are the way to go. Greets, Ertugrul -- nightmare = unsafePerformIO (getWrongWife >>= sex) http://ertes.de/