Modulus of a negative number

[Jim Hunter]

Hello all,

I have been reading K&R2, and it says that "the sign of the result for % [is]
machine-dependent for negative operands." Does this mean that the absolute
value of the result is guaranteed, but the sign isn't? Or does it mean that
(for example) -100 % 3 == -1 on some machines and 2 on others? The latter
behavior seems illogical to me, but I've been unable to find a definitive
statement about this.

Thanks in advance for satisfying my curiosity and for taking mercy on a
newsgroup newbie.

Jim

 

[Walter Roberson]

:I have been reading K&R2, and it says that "the sign of the result for % [is]
:machine-dependent for negative operands." Does this mean that the absolute
:value of the result is guaranteed, but the sign isn't?

No.

: Or does it mean that
for example) -100 % 3 == -1 on some machines and 2 on others?

Yes.

: The latter
:behavior seems illogical to me, but I've been unable to find a definitive
:statement about this.

How were you thinking it could work with the absolute value being right but
the sign possibly being wrong? A positive value for % means that many
from the beginning; a negative value for % means that many from the end.
When the result isn't exactly half-way inbetween, changing the sign gives
a dfiferent meaning.

 

[Peter Nilsson]

I have been reading K&R2, and it says that "the sign of the result for %
[is] machine-dependent for negative operands." Does this mean that the
absolute value of the result is guaranteed, but the sign isn't?

No. C89 gives an implementation two options if either operand to / (or
%)
are negative: round up, or round down (this includes round to zero
which
is _required_ by C99.) In either case, if the result is representable,
the following must hold...

a == (a/b)*b + (a%b)

Examples...

5 / -3 == -2 and 5 % -3 == -1
or 5 / -3 == -1 and 5 % -3 == 2

Or does it mean that (for example) -100 % 3 == -1 on some machines
and 2 on others?
Yes.

The latter behavior seems illogical to me,

It's sometimes useful mathematically for the modulus to always be
positive.
This, and round towards zero, are the only mechanisms that you are
likely
to encounter. C89 just happens to give an implementation 8 possible
variations.

 

[ruang]

One simple example in which the positive number is more useful, is calculating time. For instance, if the current time is 1pm, and I want to figure out the time 2 hours ago, I would do the following:
(1-2)%12 = -1%12 = 11
which is the correct answer. In this case the answer of -1 does no good at all. The reason we want the positive answer in this case is because the time scale is cyclic from 0 to 12, no negative values. I would imagine many real-life examples in which some cyclic scale goes from 0 to some positive number, in which case would prefer to use the positive answer over the negative.