Comparisons between integers and floats are inexact

[HertzDevil]

Int::Primitive#==(Float::Primitive) and Float::Primitive#==(Int::Primitive), as well as the other relational operators, all convert the integer argument to the same type as the float argument before doing the comparison. This leads to some odd results when the conversion is inexact:

UInt64::MAX == 2.0 ** 64          # => true
UInt128::MAX == Float32::INFINITY # => true

This is problematic because #hash is exact:

{UInt64::MAX => 'a', 2.0 ** 64 => 'b'}[UInt64::MAX] # => 'b'

The inexact pairs of types are:

• Int32, Float32
• UInt32, Float32
• Int64, Float32 | Float64
• UInt64, Float32 | Float64
• Int128, Float32 | Float64
• UInt128, Float32 | Float64

Python handles this correctly, which they call a nightmare. Ruby handles equality here and comparisons here.