Subtracting unsigned ints

[bg_ie]

Hi,

I have a function which compares two unsigned ints and returns their
difference in order to establish which is the greater of the two.

int Compare(unsigned int time_left, unsigned int time_right)
{
return (time_left - time_right);
}

usage -

if( Compare(time_left, time_right) > 0)
// the time has come.

The problem is that when time_left is 5 and time_right is 10, I get -5,
but if time_left is 5 and time_right is 0xc0000005 (a number greater
than 5) I get a positive value when I wish to get a negitive one.

I understand why I get this positive value - because the negaitve bit
is set in time_right, but how can I change my code so that I get the
desired results.

Thanks for your help,

Barry.

 

[Ian Collins]

Hi,

I have a function which compares two unsigned ints and returns their
difference in order to establish which is the greater of the two.

int Compare(unsigned int time_left, unsigned int time_right)
{
return (time_left - time_right);
}

usage -

if( Compare(time_left, time_right) > 0)
// the time has come.

The problem is that when time_left is 5 and time_right is 10, I get -5,
but if time_left is 5 and time_right is 0xc0000005 (a number greater
than 5) I get a positive value when I wish to get a negitive one.

I understand why I get this positive value - because the negaitve bit
is set in time_right, but how can I change my code so that I get the
desired results.

If the exact difference doesn't matter;

return (time_left > time_right);

or

return ((time_left > time_right) ? 1 : 0);

 

[Peter Nilsson]

Hi,

I have a function which compares two unsigned ints and returns their
difference in order to establish which is the greater of the two.

You can't subtract two unsigned ints and yield a negative number.

int Compare(unsigned int time_left, unsigned int time_right)
{
return (time_left - time_right);
}

usage -

if( Compare(time_left, time_right) > 0)
// the time has come.

The problem is that when time_left is 5 and time_right is 10, I get -5,
but if time_left is 5 and time_right is 0xc0000005 (a number greater
than 5) I get a positive value when I wish to get a negitive one.

Try...

if (time_left < time_right) return -1;
if (time_left > time_right) return +1;
return 0;

Or...

return (time_left < time_right) ? -1 : (time_left > time_right);

Or...

return (time_left > time_right) - (time_left < time_right);

I understand why I get this positive value - because the negaitve bit
is set in time_right, ...

No, time_right is an unsigned int. There is _no_ negative (sign) bit.

The problem is that arithmetic on unsigned ints is performed
modulo one more than UINT_MAX.

 

[Nelu]

Hi,

I have a function which compares two unsigned ints and returns their
difference in order to establish which is the greater of the two.

int Compare(unsigned int time_left, unsigned int time_right)
{
return (time_left - time_right);
}

usage -

if( Compare(time_left, time_right) > 0)
// the time has come.

The problem is that when time_left is 5 and time_right is 10, I get -5,
but if time_left is 5 and time_right is 0xc0000005 (a number greater
than 5) I get a positive value when I wish to get a negitive one.

I understand why I get this positive value - because the negaitve bit
is set in time_right, but how can I change my code so that I get the
desired results.

It doesn't matter what you set. It's supposed to be an unsigned int so
it gets promoted to something higher than int (assuming that int has 4
bytes) to be able to store the positive value. They probably get
promoted to long so you have a difference between two long
arguments. But, you are returning int even if the unsigned value
doesn't fit in an int.

 

[Eric Sosman]

It doesn't matter what you set. It's supposed to be an unsigned int so
it gets promoted to something higher than int (assuming that int has 4
bytes) to be able to store the positive value. They probably get
promoted to long so you have a difference between two long
arguments. But, you are returning int even if the unsigned value
doesn't fit in an int.

You were doing fine until the second sentence, which is
wrong and misleading: Wrong because unsigned int is not "promotable"
at all, and misleading because the four-byte int is a red herring
that has nothing to do with anything. The third sentence, since
its premise is a falsehood, is simply meaningless.

Your final sentence, though, correctly identifies the crux
of the problem. Others have offered solutions.

 

[pete]

It doesn't matter what you set. It's supposed to be an unsigned int so
it gets promoted to something higher than int (assuming that int has 4
bytes) to be able to store the positive value. They probably get
promoted to long so you have a difference between two long
arguments. But, you are returning int even if the unsigned value
doesn't fit in an int.

I don't understand what you mean.
There's nothing getting promoted in the shown code.
The type of (time_left - time_right) is unsigned int.
The only conversion in the shown code
is that the return value of the function,
is (time_left - time_right) converted to type int,
which is implementation defined
if (time_left - time_right) is greater than INT_MAX.

 

[Jack Klein]

Hi,

I have a function which compares two unsigned ints and returns their
difference in order to establish which is the greater of the two.

int Compare(unsigned int time_left, unsigned int time_right)
{
return (time_left - time_right);
}

usage -

if( Compare(time_left, time_right) > 0)
// the time has come.

The problem is that when time_left is 5 and time_right is 10, I get -5,
but if time_left is 5 and time_right is 0xc0000005 (a number greater
than 5) I get a positive value when I wish to get a negitive one.

I understand why I get this positive value - because the negaitve bit
is set in time_right, but how can I change my code so that I get the
desired results.

Thanks for your help,

What happens if the difference is has greater magnitude that a signed
int can hold? If one of the values is UINT_MAX and the other is 0,
the difference between them is either positive or negative UINT_MAX,
and that won't fit in a signed int on any platform you are ever likely
to see in your career.

Still assuming it fits (untested):

int Compare(unsigned int time_left, unsigned int time_right)
{
int result;

if (time_left >= time_right)
{
result = time_left - time_right;
}
else
{
result = -(int)(time_right - time_left);
}
return result;
}

 

[Nelu]

You were doing fine until the second sentence, which is
wrong and misleading: Wrong because unsigned int is not "promotable"
at all,
and misleading because the four-byte int is a red herring
that has nothing to do with anything. The third sentence, since

The 4 byte thing was a reference to the hex value and the explanation
the OP provided.
I got the promotion thing wrong. I have re-read the standard. I
should've taken another look before posting. Sorry about that.

its premise is a falsehood, is simply meaningless.

Yes it is.

 

[CBFalconer]

I have a function which compares two unsigned ints and returns
their difference in order to establish which is the greater of
the two.

int Compare(unsigned int time_left, unsigned int time_right)
{
return (time_left - time_right);
}

usage -

if( Compare(time_left, time_right) > 0)
// the time has come.

The problem is that when time_left is 5 and time_right is 10, I
get -5, but if time_left is 5 and time_right is 0xc0000005 (a
number greater than 5) I get a positive value when I wish to get
a negitive one.

I understand why I get this positive value - because the negaitve
bit is set in time_right, but how can I change my code so that I
get the desired results.

By avoiding overflows and undefined behaviour. Forget 'sign bits'.

int compare(unsigned int left, unsigned int right)
{
return (left < right) - (left > right);
}

 

[Peter Nilsson]

You were doing fine until the second sentence, which is
wrong and misleading: Wrong because unsigned int is not "promotable"
at all, ...

To clarify, although not subject to integral promotion, an unsigned int
operand is potentially subject to arithmetic conversion. However, in
the
case at hand, arithmetic conversion is not applicable since both
operands
are unsigned int.

 

[Eric Sosman]

By avoiding overflows and undefined behaviour. Forget 'sign bits'.

int compare(unsigned int left, unsigned int right)
{
return (left < right) - (left > right);
}

Aside: I think this is backwards from the O.P.'s intent.
Either exchange the operands of `-' or interchange `<' and `>'.
(Suggestion: Don't make *both* changes. ;-)