Nice and fast.

cxxmph/mph_bits.h

@@ -20,10 +20,6 @@ class dynamic_2bitset {
   dynamic_2bitset() : size_(0), fill_(false)  {}
   dynamic_2bitset(uint32_t size, bool fill = false)
       : size_(size), fill_(fill), data_(ceil(size / 4.0), ones()*fill) {
-   if (data_.size()) fprintf(stderr, "creating %p size %d\n", &data_[0], data_.size());
-  }
-  ~dynamic_2bitset() {
-    if (data_.size()) fprintf(stderr, "Deleting %p size %d\n", &data_[0], data_.size());
   }
 
   const uint8_t operator[](uint32_t i) const { return get(i); }
@@ -53,19 +49,13 @@ class dynamic_2bitset {
   uint32_t size() const { return size_; }
   static const uint8_t vmask[];
   const std::vector<uint8_t>& data() const { return data_; }
-// private:
+ private:
   uint32_t size_;
   bool fill_;
   std::vector<uint8_t> data_;
   const uint8_t ones() { return std::numeric_limits<uint8_t>::max(); }
 };
 
-static void set_2bit_value(uint8_t *d, uint32_t i, uint8_t v) {
-  d[(i >> 2)] &= ((v << ((i & 3) << 1)) | dynamic_2bitset::vmask[i & 3]);
-}
-static uint32_t get_2bit_value(const uint8_t* d, uint32_t i) {
-  return (d[(i >> 2)] >> (((i & 3) << 1)) & 3);
-}
 static uint32_t nextpoweroftwo(uint32_t k) {
   if (k == 0) return 1;
   k--;

cxxmph/mph_index.cc

@@ -37,14 +37,12 @@ static uint8_t kBdzLookupIndex[] =
 
 namespace cxxmph {
 
-const uint8_t MPHIndex::valuemask[] = { 0xfc, 0xf3, 0xcf, 0x3f};
-
 MPHIndex::~MPHIndex() {
   clear();
 }
 
 void MPHIndex::clear() {
-  if (!deserialized_) delete [] ranktable_;
+  delete [] ranktable_;
   ranktable_ = NULL;
   ranktable_size_ = 0;
   // TODO(davi) implement me
@@ -162,7 +160,7 @@ void MPHIndex::Ranking() {
   uint32_t size = k_ >> 2U;
   ranktable_size_ = static_cast<uint32_t>(
       ceil(n_ / static_cast<double>(k_)));
-  if (!deserialized_) delete [] ranktable_;
+  delete [] ranktable_;
   ranktable_ = NULL;
   uint32_t* ranktable = new uint32_t[ranktable_size_];
   memset(ranktable, 0, ranktable_size_*sizeof(uint32_t));
@@ -191,11 +189,11 @@ uint32_t MPHIndex::Rank(uint32_t vertex) const {
   beg_idx_v = beg_idx_b << 2;
   // cerr << "beg_idx_v: " << beg_idx_v << endl;
   // cerr << "base rank: " << base_rank << endl;
-  cerr << "G: ";
+  // cerr << "G: ";
-  for (unsigned int i = 0; i < n_; ++i) {
+  // for (unsigned int i = 0; i < n_; ++i) {
-    cerr << static_cast<uint32_t>(g_[i]) << " ";
+  //  cerr << static_cast<uint32_t>(g_[i]) << " ";
-  }
+  // }
-  cerr << endl;
+  // cerr << endl;
   while (beg_idx_v < vertex) {
     if (g_[beg_idx_v] != kUnassigned) ++base_rank;
     ++beg_idx_v;
@@ -204,14 +202,4 @@ uint32_t MPHIndex::Rank(uint32_t vertex) const {
   return base_rank;
 }
 
-uint32_t MPHIndex::serialize_bytes_needed() const {
-  return 0;
-}
-void MPHIndex::serialize(char* memory) const {
-}
-
-bool MPHIndex::deserialize(const char* serialized_memory) {
-  return true;
-}
-
 }  // namespace cxxmph

cxxmph/mph_index.h

@@ -45,8 +45,7 @@ class MPHIndex {
  public:
   MPHIndex(double c = 1.23, uint8_t b = 7) :
       c_(c), b_(b), m_(0), n_(0), k_(0), r_(1),
-      ranktable_(NULL), ranktable_size_(0),
+      ranktable_(NULL), ranktable_size_(0) { }
-      deserialized_(false) { }
   ~MPHIndex();
 
   template <class SeededHashFcn, class ForwardIterator>
@@ -69,13 +68,6 @@ class MPHIndex {
   template <class SeededHashFcn, class Key>  // must agree with Reset
   void hash_vector(const Key& x, uint32_t* h) const;
 
-  // Serialization for mmap usage - not tested well, ping me if you care. 
-  // Serialized tables are not guaranteed to work across versions or different
-  // endianness (although they could easily be made to be).
-  uint32_t serialize_bytes_needed() const;
-  void serialize(char *memory) const;
-  bool deserialize(const char* serialized_memory);
-
  private:
   template <class SeededHashFcn, class ForwardIterator>
   bool Mapping(ForwardIterator begin, ForwardIterator end,
@@ -111,10 +103,6 @@ class MPHIndex {
   // The selected hash seed triplet for finding the edges in the minimal
   // perfect hash function graph.
   uint32_t hash_seed_[3];
-
-  bool deserialized_;
-
-  static const uint8_t valuemask[];
 };
 
 // Template method needs to go in the header file.
@@ -153,10 +141,8 @@ bool MPHIndex::Reset(
   }
   if (iterations == 0) return false;
   Assigning(edges, queue);
-  fprintf(stderr, "Assignment finished\n");
   std::vector<TriGraph::Edge>().swap(edges);
   Ranking();
-  deserialized_ = false;
   return true;
 }
 

cxxmph/mph_map_test.cc

@@ -16,7 +16,6 @@ int main(int argc, char** argv) {
     b.insert(make_pair(i, i));
   }
   b.rehash(b.size());
-  fprintf(stderr, "Insertion finished\n");
   for (int i = 0; i < 1000000; ++i) {
     auto it = b.find(i % num_keys);
     if (it == b.end()) {