c++ doubles and floats, how to get around the ugly truth

Asked Aug 13 '18 at 21:26

Active Aug 24 '18 at 23:22

Viewed 71 times

double f1 = 0.00000001;
double f2 =  7.00000001;
double f3 = -7.00000000;
double sum = f2 + f3;
bool shouldBeTrue = sum == f1;

shouldBeTrue is false using VS C++ 2015/2017

float f1 = 0.00001;
float f2 = 7.00001;
float f3 = -7.00000;
float sum = f2 + f3;
bool shouldBeTrue = sum == f1;

shouldBeTrue is false using VS C++ 2015/2017

how to implement this in c++ so I can store the values written in the code and not the approximations? any new types or library available? (e.g. like decimal type in c#)

edited Aug 24 '18 at 23:22

asked Aug 13 '18 at 21:26

ovi

Use hexadecimal floating point, instead of decimal approximations which will be converted to hexadecimal floating point. – Eljay Aug 13 '18 at 22:21
You can't store the values written in the code. They are not valid float/double values. – user253751 Aug 14 '18 at 00:33
Just change the name of that Boolean to `mightBeTrue`. – Pete Becker Aug 14 '18 at 14:55
@immibis, that is not true, c# does it, c++ should be able to do it too. – ovi Aug 15 '18 at 06:28
@Pete Beker, good one! :) – ovi Aug 15 '18 at 06:28
C# doesn't do it. – user253751 Aug 15 '18 at 22:26
@immibis please try before you speak, c# has a way to do it decimal f1 = 0.00000001M; decimal f2 = 7.00000001M; decimal f3 = -7.00000000M; decimal sum = f2 + f3; bool shouldBeTrue = sum == f1; – ovi Aug 16 '18 at 23:27
@Ovi-WanKenobi decimal isn't double or float... – user253751 Aug 16 '18 at 23:48
@immibis, I think you are missing the point, c# has a way for me as a dev to get exactly the same value I write/expect, whereas c++ seems it doesn't, and I am asking for a similar way in c++, a library, etc. – ovi Aug 23 '18 at 19:32
So you are asking how to get a decimal type in C++. Just say that... – user253751 Aug 23 '18 at 22:55
yes, I added a note in the initial description. thanks – ovi Aug 24 '18 at 23:22

c++ doubles and floats, how to get around the ugly truth

0 Answers0