CODEWITHSUNDEEP
c++c++11floating-pointprecisionnumeric-limits
WilliamKF
WilliamKF

Reputation: 43179

What is the benefit of max_digits10 being 0 on references to floating point types?

In C++11, what is the advantage of std::numeric_limits<Type>::max_digits10 returning 0 for a Type that is a reference to floating point number?

For example:

constexpr int precisionPositive(const float &floatVal)
{
  using numericType    = std::remove_reference<decltype(floatVal)>::type;
  constexpr int digits = std::numeric_limits<numericType>::max_digits10;

  return digits;
}

constexpr int precisionZero(const float &floatVal)
{
  using numericType    = decltype(floatVal);
  constexpr int digits = std::numeric_limits<numericType>::max_digits10;

  return digits;
}

The precisionPositive returns 9 and precisionZero returns 0.

Under what circumstances would the zero value helpful, as opposed to giving a compile time error?

Upvotes: 3

Views: 213

Answers (2)

Red.Wave
Red.Wave

Reputation: 4251

I don't have a proper answer for the question "why". Buy I might use this code:

C++11:

template<typename T>
    using bare_value_t=typename
        std::decay<T>::type;

C++14:

template<typename T>
    using bare_value_t=
        std::decay_t<T>;

C++20:

template<typename T>
    using bare_value_t=
        std::remove_cvref_t<T>;

And:

template<typename T>
    using limits=
        std::numeric_limits<bare_value_t<T> >;

cout<<limits<float&>::max_digits10;

Upvotes: 0

P.W
P.W

Reputation: 26800

As per the C++ standard (n4659), in the numeric_limits class template the default value for max_digits10 is 0.

21.3.4 Class template numeric_limits [numeric.limits]

namespace std {
template<class T> class numeric_limits {
public:
...  
static constexpr int max_digits10 = 0;

And it is only when the template is specialized for float, double etc, max_digits10 is given a particular value.

Consider the specialization for float.

21.3.4.2 numeric_limits specializations [numeric.special]

namespace std {
template<> class numeric_limits<float> {
public:
...
static constexpr int max_digits10 = 9;

The implementations also follow this. See GCC's Standard C++ Library header limits for example.

So, except for specializations where max_digits10 is given a particular value, the default value is 0 and therefore the value of max_digits10 for references to floating point types is also 0.

Upvotes: 2

Related Questions