Chapter 15. Boost.Unordered

Example 15.1. Using boost::unordered_set

boost::unordered_set can be replaced with std::unordered_set in Example 15.1. boost::unordered_set doesn’t differ from std::unordered_set.

Example 15.2. Using boost::unordered_map

Example 15.3. User-defined type with Boost.Unordered

In elements of type animal are stored in a container of type boost::unordered_set. Because the hash function of boost::unordered_set doesn’t know the class animal, hash values can’t be automatically calculated for elements of this type. That’s why a hash function must be defined – otherwise the example can’t be compiled.

The name of the hash function to define is hash_value(). It must expect as its sole parameter an object of the type the hash value should be calculated for. The type of the return value of hash_value() must be std::size_t.

Usually, the definition of hash_value() is rather simple: Hash values are created by accessing the member variables of an object one after another. This is done with the function boost::hash_combine(), which is provided by Boost.Hash and defined in boost/functional/hash.hpp. You don’t have to include this header file if you use Boost.Unordered because all containers from this library access Boost.Hash to calculate hash values.

In addition to defining hash_value(), you need to make sure two objects can be compared using ==. That’s why the operator operator== is overloaded for animal in Example 15.3.

The hash containers from the C++11 standard library use a hash function from the header file functional. The hash containers from Boost.Unordered expect the hash function hash_value(). Whether you use Boost.Hash within hash_value() doesn’t matter. Boost.Hash makes sense because functions like make it easier to calculate hash values from multiple member variables step by step. However, this is only an implementation detail of hash_value(). Apart from the different hash functions used, the hash containers from Boost.Unordered and the standard library are basically equivalent.