Perfect hash working, but it is slower.

cxxmph/hollow_iterator.h

@@ -17,7 +17,7 @@ struct hollow_iterator_base
   typedef typename iterator::pointer pointer;
 
   hollow_iterator_base(container* c, presence* p, iterator it)
-      : c_(c), p_(p), it_(it) { find_present(); }
+      : c_(c), p_(p), it_(it) { if (c_) find_present(); }
   self_reference operator++() {
     ++it_; find_present();
   }
@@ -44,6 +44,7 @@ struct hollow_iterator : public hollow_iterator_base<
     container_type, std::vector<bool>, typename container_type::iterator> {
   typedef hollow_iterator_base<
       container_type, std::vector<bool>, typename container_type::iterator> parent_class;
+  hollow_iterator() : parent_class(NULL, NULL, typename container_type::iterator())  { }
   hollow_iterator(typename parent_class::container* c,
                   typename parent_class::presence* p,
                   typename parent_class::iterator it)
@@ -58,6 +59,7 @@ struct hollow_const_iterator : public hollow_iterator_base<
   typedef hollow_const_iterator<container_type> self_type;
   typedef hollow_iterator<container_type> non_const_type;
   hollow_const_iterator(non_const_type rhs) : parent_class(rhs.c_, rhs.p_, typename container_type::const_iterator(rhs.it_)) { }
+  hollow_const_iterator() : parent_class(NULL, NULL, typename container_type::iterator())  { }
   hollow_const_iterator(const typename parent_class::container* c,
                         const typename parent_class::presence* p,
                         typename parent_class::iterator it)

cxxmph/hollow_iterator_test.cc

@@ -31,5 +31,8 @@ int main(int argc, char** argv) {
   auto it1 = hollow_iterator<vector<int>>(&v, &p, v.begin());
   auto it2 = hollow_const_iterator<vector<int>>(&v, &p, v.begin());
   if (it1 != it2) exit(-1);
+
+  hollow_iterator<vector<int>> default_constructed;
+  default_constructed = hollow_iterator<vector<int>>(&v, &p, v.begin());
 }
 

cxxmph/mph_map.h

@@ -7,6 +7,7 @@
 // and should not be used if performance is a concern. In fact, you should only
 // use it for educational purposes.
 
+#include <iostream>
 #include <algorithm>
 #include <unordered_map>
 #include <vector>
@@ -71,9 +72,8 @@ class mph_map {
   data_type& operator[](const key_type &k);
   const data_type& operator[](const key_type &k) const;
 
-  size_type bucket_count() const { return index_.perfect_hash_size() + slack_.bucket_count(); }
+  size_type bucket_count() const { return index_.size() + slack_.bucket_count(); }
-  // FIXME: not sure if this has the semantics I want
+  void rehash(size_type nbuckets /*ignored*/); 
-  void rehash(size_type nbuckets /*ignored*/) { pack(); }
 
  protected:  // mimicking STL implementation
   EqualKey equal_;
@@ -131,6 +131,7 @@ MPH_MAP_METHOD_DECL(insert_return_type, insert)(const value_type& x) {
   }
   values_.push_back(x);
   present_.push_back(true);
+  ++size_;
   slack_.insert(make_pair(x.first, values_.size() - 1));
   if (should_pack) pack();
   it = find(x.first);
@@ -143,10 +144,12 @@ MPH_MAP_METHOD_DECL(void_type, pack)() {
       make_iterator_first(begin()),
       make_iterator_first(end()), size_);
   assert(success);
-  std::vector<value_type> new_values(index_.size());
+  std::vector<value_type> new_values(index_.perfect_hash_size());
-  std::vector<bool> new_present(index_.size(), false);
+  new_values.reserve(new_values.size() * 2);
-  for (iterator it(begin()), it_end(end()); it != it_end; ++it) {
+  std::vector<bool> new_present(index_.perfect_hash_size(), false);
-    size_type id = index_.index(it->first);
+  new_present.reserve(new_present.size() * 2);
+  for (iterator it = begin(), it_end = end(); it != it_end; ++it) {
+    size_type id = index_.perfect_hash(it->first);
     assert(id < new_values.size());
     new_values[id] = *it;
     new_present[id] = true;
@@ -168,11 +171,13 @@ MPH_MAP_METHOD_DECL(void_type, clear)() {
   present_.clear();
   slack_.clear();
   index_.clear();
+  size_ = 0;
 }
 
 MPH_MAP_METHOD_DECL(void_type, erase)(iterator pos) {
   present_[pos - begin] = false;
   *pos = value_type();
+  --size_;
 }
 MPH_MAP_METHOD_DECL(void_type, erase)(const key_type& k) {
   iterator it = find(k);
@@ -214,6 +219,13 @@ MPH_MAP_METHOD_DECL(my_int32_t, index)(const key_type& k) const {
 MPH_MAP_METHOD_DECL(data_type&, operator[])(const key_type& k) {
   return insert(make_pair(k, data_type())).first->second;
 }
+MPH_MAP_METHOD_DECL(void_type, rehash)(size_type nbuckets) {
+  pack();
+  vector<value_type>(values_.begin(), values_.end()).swap(values_);
+  vector<bool>(present_.begin(), present_.end()).swap(present_);
+  slack_type().swap(slack_);
+}
+
 
 }  // namespace cxxmph
 

cxxmph/mph_map_test.cc

@@ -11,21 +11,25 @@ using cxxmph::mph_map;
 
 int main(int argc, char** argv) {
   mph_map<int64_t, int64_t> b;
-  for (int i = 0; i < 100*1000; ++i) {
+  int32_t num_keys = 1000*10;
+  for (int i = 0; i < num_keys; ++i) {
     b.insert(make_pair(i, i));
   }
-  for (int i = 0; i < 1000*1000; ++i) {
+  for (int i = 0; i < num_keys; ++i) {
-    b.find(i);
+    auto it = b.find(i);
+    if (it->first != it->second || it->first != i) {
+      std::cerr << "Found " << it->first << " looking for " << i << std::endl;
+      exit(-1);
+    }
   }
-  /*
   mph_map<string, int> h;
   h.insert(std::make_pair("-1",-1));
   mph_map<string, int>::const_iterator it;
   for (it = h.begin(); it != h.end(); ++it) {
-    std::cerr << it->first << " -> " << it->second << std::endl;
+    if (it->second != -1) exit(-1);
   }
-  std::cerr << "Search -1 gives " << h.find("-1")->second << std::endl;
+  int32_t num_valid = 100;
-  for (int i = 0; i < 100; ++i) {
+  for (int i = 0; i < num_valid; ++i) {
      char buf[10];    
      snprintf(buf, 10, "%d", i);
      h.insert(std::make_pair(buf, i));
@@ -34,18 +38,17 @@ int main(int argc, char** argv) {
     for (int i = 1000; i > 0; --i) {
        char buf[10];    
        snprintf(buf, 10, "%d", i - 1);
-       h.find(buf);
+       auto it = h.find(buf);
-       std::cerr << "Search " << i - 1 << " gives " << h.find(buf)->second << std::endl;
+       if (i < num_valid && it->second != i - 1) exit(-1);
     }
   }
   for (int j = 0; j < 100; ++j) {
     for (int i = 1000; i > 0; --i) {
        char buf[10];    
-       snprintf(buf, 10, "%d", i*100 - 1);
+       int key = i*100 - 1;
-       h.find(buf);
+       snprintf(buf, 10, "%d", key);
-       std::cerr << "Search " << i*100 - 1 << " gives " << h.find(buf)->second << std::endl;
+       auto it = h.find(buf);
+       if (key < num_valid && it->second != key) exit(-1);
     }
   }
-  */
-
 }