Comparing a floating point to one or zero with Wfloat-equal option

Question

I'm somewhat familiar with the issues around performing floating point equality comparisons in code.

Currently the code base I'm compiling on (GCC, Clang) have the following option enabled: -Wfloat-equal

And in the code base there's the following example comparison:

template <typename FloatType>
void foo(FloatType v) {
  if (v == FloatType(1)) {
   ...
  }
  else if (v == FloatType(0)) {
   ....
  }
}

The foo function is invoked as follows:

double d = 123.98;
float f = 123.98f;

foo(d);
foo(f);

Given the special case of 1 and 0 which each have exact representations in floating points (double, float) and where the code is clearly after an exact equality and not something that is close by some minor difference -

Is there a way to rewrite the code such that it will not raise the associated Wfloat-equal diagnostic and will also be portable and support both float and double types?

Answer 1

How about using std::equal_to. As far as GCC is concerned it will disable a whole bunch of checks when processing the system headers including the float-equal

template <typename FloatType>
void foo(FloatType v) {
  if (std::equal_to<FloatType>()(v,FloatType(1))) {
   ...
  }
  else if (std::equal_to<FloatType>()(v,FloatType(0))) {
   ....
  }
}