Hello Gents,

I would like to suggest a new data structure that combines 'std::unordered_set' and a 'std::stack' behavior.

I will refere to it as 'std::hashed_stack'.

Important methods of 'std::hashed_stack' container are:

 -'push': to insert a key, and if it already exists nothing will happen. 

 -'pop': to extract the last key inserted.

The nodes of this data structure will have an extra data members ;let's name it "node_type::last_inserted"; which points to the bucket index of the last key inserted. (Single linked list of bucket indexes).

"node_type::last_inserted" can self point to the bucket index of its key, if this key's hash is congruent with the bucket index (hash collision)

'std::hashed_stack::end()': node will have a garbage key value, but a "node_type::last_inserted" value equal to the bucket index of the last inserted key. Thus, when popping keys, we start from ::end(), we read the bucket index of the last inserted key, we visit that bucket then grab the head of the linked list, we read its "node_type::last_inserted" value, we update our "end" node with this value, then erase that key .i.e poping it.

Testing existence of any key, will follow the "std::unordered_set" data structure behavior.(hash->traverse)

Only push & pop operations will be allowed to modify this container. 

All the usual member functions are the same as 'std::stack' with some extra, i.e:

std::hashed_stack::size()

std::hashed_stack::empty()

std::hashed_stack::contains(const T& key)

std::hashed_stack::push(const T& key)

std::hashed_stack::bucket_index(const T& key)

std::hashed_stack::pop()

std::hashed_stack::begin() /cbegin()

std::hashed_stack::end() / cend()

std::hashed_stack::first() -> key

std::hashed_stack::last() -> key

std::hashed_stack::clear()

std::hashed_stack::bucket_count()

.

.

.

...etc

Comments and discussions are welcome. 

Regards

Nadir