Union

Question

#include <stdio.h>

int main() {
union f {
int a;
double fl;
};
union f U;
U.a = 5;
printf("%f", U.fl);
return 0;
}

Why does it output 0.00000?

Answer 1

Why didn't nobody mention that `int` and `float` are incompatible types and their bitfield get interpreted differently and that's why the output became zero?

32-bit FP field
+#+########+####################+
| S | Exponent | Fraction |
+#+########+####################+

32-bit Integer field
+#+#############################+
| S | Magnitude |
+#+#############################+

_____
https://en.wikipedia.org/wiki/Single-precision_floating-point_format
https://www.dummies.com/programming/c/the-real-difference-between-integers-and-floating-point-values/

Answer 2

"they will still be 0 because of the interpretation"

It produces different `representation` in different implementations. For example, setting the

U.fl = 154543.5005;

in VC++, and `printf("%d", U.a);` yields

17179869

Which is the decimal representation of the binary value of `154543.5005` ¹

Conversly, setting the

U.a = 1128;

and printf("%f", U.fl); yields

-92559592117444873187641221127133740829952104046278057318678528.000000

Which is the FP representation of the binary value of `1128` ² (The above actual value might be 0.000000000000000000000000001)

_____
¹ Verification: https://rextester.com/DTTXK12756
² Verification: https://rextester.com/YECDT81643

Answer 3

Ok Ketan Lalcheta. Understand it now.

Answer 4

A union allows to store different data types in the same memory location.
It is like a structure because it has members. However, a union variable uses the same memory location for all its member's and only one member at a time can occupy the memory location.

When assignment is performed, the union memory location will be used for that member until another member assignment is performed.

Trying to access a member that isn't occupying the memory location gives unexpected results.

https://www.sololearn.com/learn/C/2944/

Answer 5

ShortCode in addition to what Sekiro has said, just try below:

printf("%d",U.fl); and it would print assigned value of a i.e. 5. FL and a refers same location and fl is having value same as a until we overwrite value... just change %d instead of %f

Answer 6

Yup! As I said, it has an atomic value like

0.0000000000000000001

and the increment from the above value to the next value ( 0.0000000000000000002) cannot be achieved by unit step increment.

FP numbers spectrum

-Inf |~~~~~|~~~~|~~~|~~|~||~|~~|~~~|~~~~|~~~~~| +Inf

Integer's

-Inf |~~~~~|~~~~~|~~~~~||~~~~~|~~~~~|~~~~~| +Inf

_______
http://mirror.informatimago.com/next/developer.apple.com/documentation/mac/PPCNumerics/PPCNumerics-18.html

Answer 7

Glad to see you again Gordie
Last night I was messing with intrinsics again!
I just wanted to know if it's possible to accumulate 2 int64_t in the lower and higher part of an __m128i and then add them up as one __m128i, and finally retrieve the big integer byte-by-byte to the output?

Answer 8

What's the problems with float?
int -> char (able)
int -> float (0.0000)

Answer 9

So C++ Soldier (Babak), even if they have binary form, they will still be 0 because of the interpretation?

Answer 10

So, C++ Soldier (Babak) its not 0?
https://code.sololearn.com/csBvoJdHLcpC/?ref=app
I see all the fp higher than 0 except 0.