C++, float and double having different effects on an if statement?

[Lord Anoobis]

Homework Statement

Write a program that asks for the starting time and the number of minutes of the call,
and... pretty straightforward

Relevant Equations

None. Checked with both MS Visual Studio and DEV C++.

So I used this statement,

if (call_time > 0 && call_time <= 6.59)

where the variable was declared as float. What is strange is that it didn't execute for call_time = 6.59, but for some other values, like

if (call_time > 0 && call_time <= 2.59)

the end value of the range does not cause any issues. However, when I changed the declaration to type double, the issue disappeared entirely. Why is this happening?

 

[FactChecker]

It would help if you tell us what the exact value of call_time is in each case when the observed behavior occurs.

 

[Vanadium 50]

Checking for strict equality in floats is a bad idea. Doubles only slightly better. The various "6.59"'s in the problem may not be strictly equal.

 

[Lord Anoobis]

It would help if you tell us what the exact value of call_time is in each case when the observed behavior occurs.

I noted it for 1.59, 3.59, 5.59 and 6.59 where the given range in the problem is from 0 to 6.59.

 

[Lord Anoobis]

Checking for strict equality in floats is a bad idea. Doubles only slightly better. The various "6.59"'s in the problem may not be strictly equal.

Is this to do with the way the numbers are stored in memory? Also, how would I go about avoiding this in future?

 

[FactChecker]

I noted it for 1.59, 3.59, 5.59 and 6.59 where the given range in the problem is from 0 to 6.59.

We need more accuracy than that for the value of call_time. When you say 1.59, do you mean 1.59000 or 1.59100? What exactly is it? As a general rule, when your algorithm is depending on the equality of floats, it is best to give the program a little leeway for how it stores the fractional part. Use call_time < 2.59001 if that would not cause a problem with what you want it to do.

 

[phinds]

Is this to do with the way the numbers are stored in memory? Also, how would I go about avoiding this in future?

Yes.

I've been explaining for about 5 decades now why the test "is 1.4 + 1.6 equal to 2.0" fails in computers (early on it always failed, now it depends on the data type).

Data in computers is stored in BINARY, not decimal so what YOU see as 1.4 or 1.6, the computer may see as 1.39999998 or 1.60000000000001 for example

As @Vanadium 50 said, do NOT use exact comparisons for non-integer numbers. Do something like take the ABS of the difference and check that it is less than some small amount.

 

[Vanadium 50]

You have

call_time <= 6.59

If you replaced it with

call_time < 6.59

would it matter? If so, your code is in trouble.