Ad-hoc polymorphism

From HaskellWiki
Revision as of 18:26, 30 April 2012 by Benmachine (talk | contribs)

Jump to: navigation, search

A value is polymorphic if, depending on its context, it can assume more than one type. If the value is specified separately for each type it can have, this is called ad-hoc polymorphism, also known as overloading. Ad-hoc polymorphic values may be used at a limited number of pre-specified types.

For instance, in C++, the addition operator may be used on any numeric type, like int or float, and may be applied to new types provided you write a definition for it using those types as arguments. The range of possible types of + in C++ is limited to those built-in and those for which the programmer has provided a definition. Similarly, in C, the operator + may operate on int or float or pointers. However, the programmer cannot supply their own definitions, and only the built-in types can be used.

In Haskell, type classes are used as a mechanism for "principled overloading", i.e. ad-hoc polymorphism. For instance, the following function:

show :: (Show a) => a -> String

may adopt any of the following types:

Int -> String
Float -> String
(Maybe Int, String, Either Char Float) -> String

The types which show can adopt are precisely those which have an instance of the standard Show class defined. Each instance gives a definition of the show function to be used when the types match. Unlike in C++ function overloading, Haskell also allows overloaded values, like maxBound, that adopt a different value based on their type (e.g. maxBound :: Bool is equal to True, while maxBound :: Int may be 2147483647).

Contrast ad-hoc polymorphism with parametric polymorphism: e.g. the function length :: [a] -> Int may be applied to lists of any type, but has only one definition which it uses regardless of which type it adopts.