Re:In Erlang/OTP 27, +0.0 will no longer be exactly equal to...
by jacquesm on Wednesday May 10, 2023 @06:46AM
(#35885718)
> Can we really not have equality operators that do a comparison with a 1% tolerance or something as a sensible default equivalence instead of blatantly wrong bitwise comparison?
Then it wouldn't really be an equality comparison any more. For instance according to your rule 1.0 and 1.01 are the same number, but they clearly are not. But if you want to do this regularly you could create a special type and overload the equality operator (if your language supports it, otherwise you'll have to call the function directly) that points to 'boolean about_equal(float f1, float f2, float epsilon)' or something equivalent where epsilon is the amount that you would allow f1 and f2 to deviate from each other to still call them equal. And you could define epsilon as ((abs(f1) + abs(f2)) / 100) .
The bitwise comparison isn't 'blatantly wrong', it's the expected behavior in just about every programming language. I'm not aware of any language that has a native float operator that does this but as you can see you can usually simply add one yourself if you really need it. But this need rarely comes up and usually indicates that you are doing something wrong.
Then it wouldn't really be an equality comparison any more. For instance according to your rule 1.0 and 1.01 are the same number, but they clearly are not. But if you want to do this regularly you could create a special type and overload the equality operator (if your language supports it, otherwise you'll have to call the function directly) that points to 'boolean about_equal(float f1, float f2, float epsilon)' or something equivalent where epsilon is the amount that you would allow f1 and f2 to deviate from each other to still call them equal. And you could define epsilon as ((abs(f1) + abs(f2)) / 100) .
The bitwise comparison isn't 'blatantly wrong', it's the expected behavior in just about every programming language. I'm not aware of any language that has a native float operator that does this but as you can see you can usually simply add one yourself if you really need it. But this need rarely comes up and usually indicates that you are doing something wrong.