Re: Key strategies for hash_map

  And you have another problem. Is the key:value relationship 1:1 ?
Otherwise there there might not be such think as "the key" for a
value but "the keys" for a value.

  If it's 1:1 then may be you really should bind key and value together
in one class node. And hold pointers (or shared_ptr) to O in two
hash_maps. All this could be covered in something like
my_bi_map . But may be then you're very close to boost::multimap then

Don't know.

Something like this for GNU

<snip>
template<class K1,class K2,class HashFn1=std::hash<K1>,class
HashFn2=std::hash<K2> >
class my_bi_map {
public:
   typedef std::pair<K1,K2> node;
private:
   typedef __gnu_cxx::hash_map<K1,node *,HashFn1> map1_t;
   typedef __gnu_cxx::hash_map<K2,node *,HashFn2> map2_t;
   map1_t map1;
   map2_t map2;
public:
   my_bi_map() {
   }

   // We had ownership
   ~my_bi_map() {
     while( ! map1.empty() ) {
       delete *(map1.begin()).second;
       map1.erase(map1.begin());
     }
   }
  // add one node
   bool add(const node & n) {
     if( (map1.find(n.first) == map1.end() &&
          map2.find(n.second) != map2.end() )
      || (map1.find(n.first) != map1.end() &&
          map2.find(n.second) == map2.end() ) ) {
       std::cerr << "Arrg 1:1 criteria violated" << std::endl;
       return false;
     }

     node * nn = new node(n);
     map1[n.first] = nn;
     map2[n.second] = nn;

     return true;
   }

   bool del1(const K1 & k) {
     ...
   }

   bool del1(const K2 & k) {
     ...
   }

   const node * get1(const K1 & k) {
     ...
   }

   const node * get2(const K2 & 2) {
     ...
   }

};
</snip>

O.

--
      [ See http://www.gotw.ca/resources/clcm.htm for info about ]
      [ comp.lang.c++.moderated. First time posters: Do this! ]