0

Does the following type cast create a new variable in memory? It is referenced in the printf function below, and it prints 1.23 as the answer.

#include<stdio.h>
int main(){
    int num;
    num = 1.23456;
    printf("num = %.2f \n", (float)num);
    *((float*) &num) = 1.23456;
    printf("num = %.2f \n", *((float*)&num));
    return 0;
}
Makogan
  • 8,208
  • 7
  • 44
  • 112
kasvith
  • 377
  • 5
  • 19
  • Given `int num;` the line `*((float*) &num) = 1.23456;` violates strict aliasing. Try that on some hardware and you're likely to get a `SIGBUS`. – Andrew Henle Jun 05 '17 at 09:00

3 Answers3

5
num = 1.23456;

That sets the value of num to 1. (The fractional part, .23456, is discarded, since according to the C language rules, an int can't hold fractional values)

*((float*) &num) = 1.23456;

Here you are taking the floating-point value 1.23456 and overwriting the memory location where num lives with that bit-pattern. I suspect that this is invoking undefined behavior; in any case this can only 'work' on computers where sizeof(float)==sizeof(int), so it's a bad idea to do this.

printf("num = %.2f \n", *((float*)&num));

Here you are pretending that the memory location where num is located is actually the location of a float, and printing out that memory as if it were a floating point value. It just so happens to do what you expected, on your machine, but you shouldn't rely on it doing so. By doing tricky casting and type-punning like this, you are effectively "going behind the compiler's back" and bypassing its type-system. The likely way this will bite you (even on computers where float and int both take up the same number of bytes) is when you enable optimizations and the compiler optimizes your code in some way that assumes that an int and a float can't occupy the same address in memory -- but your code is deliberately breaking that assumption, and thus your program ends up doing something unexpected (like crashing, or printing out the wrong value).

Is the following type cast creates a new variable in memory?

C being a fairly low-level language (by modern standards anyway), it is trying to do what your program told it to do, no more and no less. In particular, your program told it to treat a pointer-to-int as a pointer-to-float, and the C compiler trusts that you know what you are doing and lets your violate the type system by using casts. So the program isn't creating a new variable in memory so much as overwriting the bytes of memory where an existing variable (num) is located, in a questionable and unsafe way.

Jeremy Friesner
  • 70,199
  • 15
  • 131
  • 234
  • About the size of float https://stackoverflow.com/questions/2524737/fixed-size-floating-point-types – Ôrel Jun 05 '17 at 03:48
2

The main idea here is that when you cast to float pointer and write the value into memory:

*((float*) &num) = 1.23456;

The compiler writes the value in memory as if it was a float. So indeed you will have a float stored in the place where an int was supposed to be stored, something you wouldn't normally be allowed to do. So when you show it as a float:

printf("num = %.2f \n", *((float*)&num));

It works. But when you go back and show it as a int again after assigning it directly with:

printf("num = %d \n", num);

You will see a different value:

num = 1067320848

Which is the representation of a float value shown as an integer one.

Remember that internally integers and floats are stored differently, in what concerns the bits, due to the floating point representation.

Isac
  • 1,834
  • 3
  • 17
  • 24
  • 1
    *"It works."* -> *"It fives false impression of working."* Code breaks strict aliasing rules and is undefined behaviour. – user694733 Jun 05 '17 at 06:29
  • You are entirely true. It works in the sense that it shows the value he was expecting to see, but it's definitely a bad practice and breaks many rules. – Isac Jun 05 '17 at 09:21
0

When you declare a variable as an integer and then initialise it as a float, the program, in the background, will store it as an integer as 1 in this case, truncating all the decimal places. It should be declared as a float first.

For the case of the pointer, it is displayed as a float because it has been declared and initialised as a float and you are modifying(type casting) and displaying the value at it's memory location which is calling by reference. This means that the data type will change as well.

OfLambeth
  • 1
  • 3