Float comparison gives wrong result and precision changes

Question

I have version number 1.1, 1.2, 1.3 and I need to check version less or greater than 1.2, but while debugging I get wrong answer than expected

    float versionNumber = versinInfo.toFloat();
    static float const VERSION_NUMBER(1.2);
    if(abs((versionNumber - VERSION_NUMBER) <= 0.001))
    {
      // do operation
    }

versionNumber comes 1.10000005, I though to change check from 0.001 to 0.0000005 but it maynt be a correct fix

please suggest best method

For version control you can have two individual integers - major and minor which is a common practice. See also: [Is floating point math broken?](https://stackoverflow.com/questions/588004/is-floating-point-math-broken) — Hanjoung Lee, Mar 20 '19 at 12:22
If having two variables bothers you, save it as a int value like `102` for version 1.2. — Hanjoung Lee, Mar 20 '19 at 12:25
So is it better to convert both into two integers and check indipendly — Sijith, Mar 20 '19 at 12:26
Yes, your version system should not be based on floats anyway. — Matthieu Brucher, Mar 20 '19 at 12:33

score 2 · Answer 1 · answered Mar 20 '19 at 12:35

Version numbers are inherently integers. You already have a class, this class should have an integer value for the major version, and integer version for the minor version, and comparison operators (that's what you are trying to do).

Other schemes even have a third patch integer, a string for alpha/RC... Implement this properly in a class that uses proper semantics (i.e. a method named is_one_minor_version_away where you properly test is there is only one minor version change).

Also what happens if it's 2.9 against 3.0 in your case?

score 1 · Accepted Answer · answered Mar 20 '19 at 12:23

1

Why don't you create class that hold major version and minor version of passed versionInfo and add logic that checks there. The idea would be not to parse this as float, but to separate it by comma and take that info as int.

answered Mar 20 '19 at 12:23

Slobodan Markovic

36
3

Float comparison gives wrong result and precision changes

2 Answers2