Difference between Structures and Union

[jamesjohnney]

I dont want the normal ordinary answer for this question like it differes only in memory allocation and all.....
I need ..like where we use Union and where can structure.... can any one please help me in this regards...

[treuss]

You use unions only in two pretty rare cases:

1. You want the members to affect each other directly, i.e. changing the value of one member will always affect the value of another member (due to the fact that they are referring to the same memory).

2. You have two data members, and for each object you will always only use one of them. Using a union will in this case save memory, as the memory of the object will not need to be big enough for both members.

[Graham]

You use unions only in two pretty rare cases:

1. You want the members to affect each other directly, i.e. changing the value of one member will always affect the value of another member (due to the fact that they are referring to the same memory).

Which is irrelevant because reading one member having written to another is prohibited by the standard (except for one specific exemption). E.g.

union thing
{
	int i;
	char c[4];
};

thing u;

u.i = 2;
cout << u.c[0] << u.c[1] << u.c[2] << u.c[3]; // illegal

2. You have two data members, and for each object you will always only use one of them. Using a union will in this case save memory, as the memory of the object will not need to be big enough for both members.

So, only one case, then.

[googler]

Unions are used mostly in low-level C code for the two reasons mentioned by treuss. In C++ they're not used much because they're mostly incompatible with constructors and destructors. The members of a union may not have them (although the union as a whole can). So they don't work very well for creating self-managing objects with RAII and such. Unions are really only good as "passive" data structures. A C++ union doesn't strictly have to be POD, but it's pretty much all they're good for anyway.
structs in C++ are practically the same as classes, the only difference being the default access control. Since classes were added when structs already existed, I guess you can think of "class" as syntactic sugar for struct.

[Jacko123]

1) You cannot write the union as

//Wont work

union u
{
      int i;
      string str;
};

//Will work

struct s
{
      int i;
      string str;
};

A union member shall not be of a class type (or array
thereof) that has a non-trivial constructor.
2) Union in C++ cannot be used as the Base class while a structure can.

[treuss]

Which is irrelevant because reading one member having written to another is prohibited by the standard (except for one specific exemption

Interessting. What is the exemption ?

[exterminator]

Interessting. What is the exemption ?

http://www.codeguru.com/forum/showpo...59&postcount=4

[NMTop40]

There is another reason for using a union - when you want to ensure that a buffer is aligned.

Note you can do something like this:

union aligned_buffer
{
   double d;
   char data[ sizeof( double ) ];
};

You can then safely do this:

aligned_buffer buf;
double x = 1.5;
memcpy( buf.data, &x, sizeof( double ) );
double *py = static_cast< double * >( buf.data ); // may need reinterpret_cast

You cannot do double y = buf.d; safely, according to the standard, but I have not yet come across a compiler where it fails.

The reason you need d in the union is to ensure the other member, data, is aligned so that you can safely do the cast.

Note the memcpy does not require alignment. (Note if this is C then you obviously use a C-cast, but alignment issues apply in C as well as C++).

[Jacko123]

Which is irrelevant because reading one member having written to another is prohibited by the standard (except for one specific exemption). E.g.

union thing
{
	int i;
	char c[4];
};

thing u;

u.i = 2;
cout << u.c[0] << u.c[1] << u.c[2] << u.c[3]; // illegal

I'm still unclear about the above...Can you please explain about the same?
Thanks in advance.