How to Make `std::unordered_map` Work Without Defining a Custom Hash Function?

Generic Hash for Tuples in unordered_map/unordered_set

Q: Why doesn't std::unordered_map, string> work out of the box?

In standard C , to use tuples as keys in associative containers like unordered_map or unordered_set, you need to define a custom hash function.

Q: Can this be automated for C 0x tuples without using variadic templates?

Yes, utilizing the following code:

namespace std{
    namespace
    {
        template <class T>
        inline void hash_combine(std::size_t&amp; seed, T const&amp; v)
        {
            seed ^= std::hash<T>()(v) + 0x9e3779b9 + (seed<<6) + (seed>>2);
        }

        template <class Tuple, size_t Index = std::tuple_size<Tuple>::value - 1>
        struct HashValueImpl
        {
          static void apply(size_t&amp; seed, Tuple const&amp; tuple)
          {
            HashValueImpl<Tuple, Index-1>::apply(seed, tuple);
            hash_combine(seed, std::get<Index>(tuple));
          }
        };

        template <class Tuple>
        struct HashValueImpl<Tuple,0>
        {
          static void apply(size_t&amp; seed, Tuple const&amp; tuple)
          {
            hash_combine(seed, std::get<0>(tuple));
          }
        };
    }

    template <typename ... TT>
    struct hash<std::tuple<TT...>> 
    {
        size_t
        operator()(std::tuple<TT...> const&amp; tt) const
        {                                              
            size_t seed = 0;                             
            HashValueImpl<std::tuple<TT...> >::apply(seed, tt);    
            return seed;                                 
        }                                              
    };
}

Q: Is there a simpler solution?

Standard Non-Conforming Solution (ADL Enabled):

#include 
namespace std{
    namespace
    {
        template <class T>
        inline void hash_combine(std::size_t&amp; seed, T const&amp; v)
        {
            seed ^= std::hash<T>()(v) + 0x9e3779b9 + (seed<<6) + (seed>>2);
        }

        template <class Tuple, size_t Index = std::tuple_size<Tuple>::value - 1>
        struct HashValueImpl
        {
          static void apply(size_t&amp; seed, Tuple const&amp; tuple)
          {
            HashValueImpl<Tuple, Index-1>::apply(seed, tuple);
            hash_combine(seed, std::get<Index>(tuple));
          }
        };

        template <class Tuple>
        struct HashValueImpl<Tuple,0>
        {
          static void apply(size_t&amp; seed, Tuple const&amp; tuple)
          {
            hash_combine(seed, std::get<0>(tuple));
          }
        };
    }

    template <typename ... TT>
    struct hash<std::tuple<TT...>> 
    {
        size_t
        operator()(std::tuple<TT...> const&amp; tt) const
        {                                              
            size_t seed = 0;                             
            HashValueImpl<std::tuple<TT...> >::apply(seed, tt);    
            return seed;                                 
        }                                              
    };
}

Standard Conforming Solution (No ADL):

To achieve strict standard conformance, you must move the above code into a separate namespace (e.g., hash_tuple) and modify the syntax to specify the custom hash function explicitly.

namespace hash_tuple{

// Forward non-tuple types to std::hash
template <typename TT>
struct hash
{
    size_t
    operator()(TT const&amp; tt) const
    {                                              
        return std::hash<TT>()(tt);                                 
    }                                              
};
}

Replace hash_combine and HashValueImpl from the non-conforming solution with their hash_tuple counterparts. Finally, use the following syntax:

unordered_set<tuple<double, int>, hash_tuple::hash<tuple<double, int>>> test2;

The above is the detailed content of How to Make `std::unordered_map` Work Without Defining a Custom Hash Function?. For more information, please follow other related articles on the PHP Chinese website!