To determine the ranges of numeric types with std::numeric_limits [18.2.1], different syntax have to be used depending on numeric type. Specifically, numeric_limits<T>::min() for integral types returns the minimum finite value, whereas for floating point types it returns the minimum positive normalized value. The difference in semantics makes client code unnecessarily complex and error prone.

boost::numeric::bounds<> provides a consistent interface for retrieving the maximum finite value, the minimum finite value and the minimum positive normalized value (0 for integral types) for numeric types. The selection of implementation is performed at compile time, so there is no runtime overhead.

template<class N>
struct bounds
{
    static N lowest  () { return implementation_defined; }
    static N highest () { return implementation_defined; }
    static N smallest() { return implementation_defined; }
};

Members

Returns the minimum finite value, equivalent to numeric_limits<T>::min() when T is an integral type, and to -numeric_limits<T>::max() when T is a floating point type.

Returns the maximum finite value, equivalent to numeric_limits<T>::max().

Returns the smallest positive normalized value for floating point types with denormalization, or returns 0 for integral types.

The following example demonstrates the use of numeric::bounds<> and the equivalent code using numeric_limits:

#include <iostream>

#include <boost/numeric/conversion/bounds.hpp>
#include <boost/limits.hpp>

int main() {

    std::cout << "numeric::bounds versus numeric_limits example.\n";

    std::cout << "The maximum value for float:\n";
    std::cout << boost::numeric::bounds<float>::highest() << "\n";
    std::cout << std::numeric_limits<float>::max() << "\n";

    std::cout << "The minimum value for float:\n";
    std::cout << boost::numeric::bounds<float>::lowest() << "\n";
    std::cout << -std::numeric_limits<float>::max() << "\n";

    std::cout << "The smallest positive value for float:\n";
    std::cout << boost::numeric::bounds<float>::smallest() << "\n";
    std::cout << std::numeric_limits<float>::min() << "\n";

    return 0;
}