C Data Types: Are int x=3 & float y=3.0 Equal?

[fraz ali]

Hello everyone!
In C language,does two different variables with different data types but equal values equate each other?
For example,are these two statements equal:

int x=3;
float y=3.0;

?
thanks in advance..

 

[Samy_A]

When you compare an int x with a float y (a tricky thing to do), as in

if (x==y)

the int is implicitly converted to float before the comparison is done.

 

[Vanadium 50]

the int is implicitly converted to float before the comparison is done.

One should be very careful about testing two floats for strict equality. If a and b are equal, a/5 and b*0.2 might not be.

 

[Samy_A]

One should be very careful about testing two floats for strict equality. If a and b are equal, a/5 and b*0.2 might not be.

Absolutely.
Sometimes I had to test for a==b, and it only worked if I tested for abs(a-b) < some sufficiently small number. Don't know if there is a more elegant way to solve this.

 

[Dave Ritche]

But will they be equal?

#include<conio.h>
#include<stdio.h>
int x=3;
float y=3.0;
if (x==y)
{
printf("X is equal to Y");
}
else
{
printf("not equal");
}
getch();
}

What will this program return?

 

[fraz ali]

When you compare an int x with a float y (a tricky thing to do), as in

if (x==y)

the int is implicitly converted to float before the comparison is done.

thanks

 

[jim mcnamara]

Equality of floating point datatypes (IEEE-754) is a pain in the butt. Most folks handle this incorrectly when they start to program.

And yes, what you are talking about is called a 'promotion' - FWIW I do not know why - it looks more like a 'demotion' to me. Doubles and floats are promoted to integers, so as explained above, you run the risk of getting incorrect answers.

You need to investigate the concept of an ULP and probably the idea behind the fenv.h "functions": fesetround and fegetround. Promoting implictly rounds values.
This is a technical paper on the subject:
https://docs.oracle.com/cd/E19957-01/806-3568/ncg_goldberg.html
A less mathematically daunting article:
http://www.cygnus-software.com/papers/comparingfloats/comparingfloats.htm

This whole subject is one the main reasons COBOL is still prominent in financial code - you lose precision with floating point. This gives auditors nightmares.

Floating point specs were updated in IEEE-754-2008, but the changes not seem to be available on most commodity x86 implementations. Fujitsu SPARC 10 cpus have it. Some graphics cards( GPUS ) support it through add-ons like CUDA.

 

[Samy_A]

But will they be equal?
<code>
#include<conio.h>
#include<stdio.h>
int x=3;
float y=3.0;
if (x==y)
{
printf("X is equal to Y");
}
else
{
printf("not equal");
}
getch();
}
</code>
What will this program return?

My C++ compiler returns "X is equal to Y".
I don't know whether this is compiler dependent.

 

[Mark44]

Hello everyone!
In C language,does two different variables with different data types but equal values equate each other?
For example,are these two statements equal:

int x=3;
float y=3.0;

The values stored in x and y do not have equal values, based on the bit patterns that are stored in the two different memory locations.

And yes, what you are talking about is called a 'promotion' - FWIW I do not know why - it looks more like a 'demotion' to me. Doubles and floats are promoted to integers, so as explained above, you run the risk of getting incorrect answers.

Jim, this is not true. Assuming you are talking about the comparison if (x == y), since the value in y is float, the the int value of x is promoted to a float as well, and then the comparison can proceed. doubles and floats are never "promoted" to ints, which would likely cause loss of precision, but in circumstances such as the one below, floating point values can be demoted to integral values.

int num = 3.6;

In the declaration/definition above, the double value 3.6 is truncated to 3 to initialize the variable, num.

 

[jim mcnamara]

@Mark44 Yup - you are correct - I have it backwards. Thanks for the correction.

 

[Vanadium 50]

What will this program return?

It's implementation dependent. My understanding is that if your implementation follows the IEEE-754 standards, it will return "equal". If you replace float y = 3.0 with float y = 3, it should always return equal: it is comparing "the float that 3 gets promoted to" with "the float that 3 gets promoted to", and such should always be true. That number may or may not be the same as the float 3.0. (IEEE says it is)

Many profs will say "never test floats for strict equality". I think a better way to think about it is "testing floats for strict equality is almost never what you want to do". If you test two floats for equality, you are asking if the two numbers are close enough to each other as to have the same binary representation. So it's not equality, it's a range - and if it's a range, shouldn't the programmer decide what that range should be, and not leave it to the compiler?