Why can't I use classes with non-trivial constructors in C unions?

Understanding Union Restrictions

In C , a union represents a memory region shared by multiple data members. However, the compiler restricts the use of classes with non-trivial constructors, copy constructors, or destructors within unions.

The Problem

When attempting to declare a union with a string member, as shown below:

union U
{
   int i;
   float f;
   string s;
};

the compiler raises an error indicating that U::s has a copy constructor.

Why the Restriction?

This restriction exists because having a class with non-trivial operations in a union introduces memory management challenges. In a typical struct, each member is allocated its own memory. In a union, however, all members share the same memory address.

Consider the example below:

union U
{
  string x;
  vector<int> y;
};

U u;  // <--

If u were a struct, u.x and u.y would be initialized to an empty string and vector, respectively. However, in a union, u.x and u.y share the same address. Thus, initializing one of them would overwrite the data of the other.

Alternatives: Tagged Unions and Extensibility

To address this issue, C 0x introduced "tagged unions." These unions enable the use of a member variable to indicate which member is active, allowing for more flexible construction and destruction of class members within the union.

Alternatively, external libraries such as Boost.Variant and Boost.Any provide similar functionality while encapsulating the complexity of memory management.

The above is the detailed content of Why can't I use classes with non-trivial constructors in C unions?. For more information, please follow other related articles on the PHP Chinese website!