Uses of Run-Time Type Information

If we have a means to obtain a typeid from a pointer, reference or class, we can provide a couple of useful operations on typeids:

• 1. Type names: Given a typeid, which is an abstract encoding of a type, we would like to obtain a textual representation of it. For example, for a C++ class A, we would like to obtain a character string "A" from its typeid.
• 2. Typeid comparison: Given two typeids, we should be able to determine if they represent the same type or if they represent types related by inheritance or unrelated types.
• 3. Pointer and reference casting: Given a pointer and a typeid, we can determine if the actual pointed object is of the type represented by that typeid. Furthermore, we could cast a given pointer p to a C++ type encoded by a given typeid t at runtime and, if the cast succeeds, return a pointer p' of type t.
• 4. Run-time object creation: Given a typeid t, we would like to create an object of type t and return it as a a pointer of type t.

Operation 1 allows us to compare types at run-time and determine their relationship (e.g. subclass to superclass). Operation 2 maps types to a textual representation which can be necessary if the user desires to be given type information in a readable way. Operation 3 gives the possibility to cast pointers and references at run-time, which is mostly used in the form of downcasts, i.e. casts from a base type to a derived type and gives the possibility to interpret a pointer in a different way. Operation 4 allows a program to postpone the decision of creating new objects until run-time and then create objects from run-time selected types.