Unions provide different views of looking at the same memory location.
Take an example of a union declared in C.

union xyz
{
	float x;
	int y[2];
	char z[4];
} u1;

The variable u1 when allocated in memory has typically 4 bytes.
These 4 bytes can be accessed in three ways. Making the memory
type polymorphic.

If accessed via u1.x the memory is treated by compiler as a float.
If accessed via field y say u1.y[1] , the last two bytes of the
memory are treated an integer.Similarly through array field z
the same memory can be viewed as an array of characters. 

This can be very serious in compilers which are strict in type
checking . You get an official way of breaking the strict type
checking rules.

One more example will drive the point home ( concocted java code)


class PQR { ... }

union dangerous {
PQR modifiablehande; // object reference to class PQR 
int culprit;
} u1; 

// creating object in heap and making "modifiablehandle" point to it 
u1.modifiablehandle = new PQR( ); 
u1.culprit = u1.culprit + 5 ; // got the point ?( pointer arithmetic
?) 

Now you can modify the content of the object handle . I don't
think James Gosling can tolerate this . So you got the point. 

By Prof. Rajesh Patkar

back