]> www.ginac.de Git - ginac.git/commitdiff
added exhashmap<> as a replacement for map<> that uses hashing
authorChristian Bauer <Christian.Bauer@uni-mainz.de>
Wed, 20 Aug 2003 21:09:59 +0000 (21:09 +0000)
committerChristian Bauer <Christian.Bauer@uni-mainz.de>
Wed, 20 Aug 2003 21:09:59 +0000 (21:09 +0000)
ginac/ginac.h
ginac/hash_map.h [new file with mode: 0644]

index 183705359d57db4ead9ecd69ed610f85a1e08234..362661a730d9b7973327f278f60b1eae53628404 100644 (file)
 #include "inifcns.h"
 #include "fderivative.h"
 #include "operators.h"
+#include "hash_map.h"
 
-#ifndef GINAC_BASE_ONLY
 #include "idx.h"
 #include "indexed.h"
 #include "tensor.h"
 #include "color.h"
 #include "clifford.h"
-#endif // ndef GINAC_BASE_ONLY
 
 #ifdef __MAKECINT__
 #pragma link C++ nestedclass;
diff --git a/ginac/hash_map.h b/ginac/hash_map.h
new file mode 100644 (file)
index 0000000..3673f69
--- /dev/null
@@ -0,0 +1,606 @@
+/** @file hash_map.h
+ *
+ *  Replacement for map<> using hash tables. */
+
+/*
+ *  GiNaC Copyright (C) 1999-2003 Johannes Gutenberg University Mainz, Germany
+ *
+ *  This program is free software; you can redistribute it and/or modify
+ *  it under the terms of the GNU General Public License as published by
+ *  the Free Software Foundation; either version 2 of the License, or
+ *  (at your option) any later version.
+ *
+ *  This program is distributed in the hope that it will be useful,
+ *  but WITHOUT ANY WARRANTY; without even the implied warranty of
+ *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+ *  GNU General Public License for more details.
+ *
+ *  You should have received a copy of the GNU General Public License
+ *  along with this program; if not, write to the Free Software
+ *  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
+ */
+
+#ifndef __GINAC_HASH_MAP_H__
+#define __GINAC_HASH_MAP_H__
+
+#include <list>
+#include <iterator>
+#include <algorithm>
+#include <functional>
+#include <utility>
+
+
+namespace GiNaC {
+
+/*
+ *  "Hashmap Light" - buckets only contain one value, quadratic probing,
+ *  grows automatically
+ */
+
+namespace internal {
+
+// List of prime numbers shamelessly stolen from GCC STL
+enum { num_primes = 29 };
+
+static const unsigned long prime_list[num_primes] =
+{
+       31ul,        53ul,         97ul,         193ul,       389ul,
+       769ul,       1543ul,       3079ul,       6151ul,      12289ul,
+       24593ul,     49157ul,      98317ul,      196613ul,    393241ul,
+       786433ul,    1572869ul,    3145739ul,    6291469ul,   12582917ul,
+       25165843ul,  50331653ul,   100663319ul,  201326611ul, 402653189ul,
+       805306457ul, 1610612741ul, 3221225473ul, 4294967291ul
+};
+
+inline unsigned long next_prime(unsigned long n)
+{
+       const unsigned long *first = prime_list;
+       const unsigned long *last = prime_list + num_primes;
+       const unsigned long *pos = std::lower_bound(first, last, n);
+       return pos == last ? *(last - 1) : *pos;
+}
+
+} // namespace internal
+
+
+// Define default arguments
+template <typename T, template <class> class A = std::allocator>
+class exhashmap;
+
+
+/** Pair Associative Container with 'ex' objects as keys, that is implemented
+ *  with a hash table and can be used as a replacement for map<> in many cases.
+ *
+ *  Differences to map<>:
+ *   - no lower_bound()/upper_bound()
+ *   - no "insert with a hint" insert(iterator, key_type)
+ *   - no reverse iterators, no rbegin()/rend()
+ *   - no operator<()
+ *   - comparison functor is hardcoded to ex_is_less
+ *   - bucket_count() returns the number of buckets allocated in the hash table
+ *   - insert() and erase() invalidate all iterators
+ *   - average complexity of find() is constant time, worst case is O(n) */
+template <typename T, template <class> class A>
+class exhashmap {
+public:
+       static const unsigned min_num_buckets = 5; // must be prime
+
+       // Standard types
+       typedef ex key_type;
+       typedef T mapped_type;
+       typedef std::pair<key_type, T> value_type;
+       typedef ex_is_less key_compare;
+       typedef ex_is_equal key_equal;
+       typedef value_type & reference;
+       typedef const value_type & const_reference;
+       typedef value_type * pointer;
+       typedef const value_type * const_pointer;
+
+protected:
+       // Private types
+       enum bucket_state {
+               EMPTY,  ///< bucket empty (never used)
+               USED,   ///< bucket in use
+               ERASED  ///< bucket empty (element deleted), but may be part of a search chain
+       };
+       typedef std::pair<bucket_state, value_type> Bucket;
+
+public:
+       // More standard types
+       typedef A<Bucket> allocator_type;
+
+protected:
+       // More private types
+       typedef std::vector<Bucket, allocator_type> Table;
+
+       typedef typename Table::iterator table_iterator;
+       typedef typename Table::const_iterator table_const_iterator;
+
+public:
+       // Iterators
+       template <typename Pointer, typename Reference, class TableIterator>
+       class exhashmap_iterator : public std::iterator<std::forward_iterator_tag, value_type, typename Table::difference_type, Pointer, Reference> {
+       protected:
+               friend class exhashmap;
+
+       public:
+               exhashmap_iterator() {}
+               exhashmap_iterator(TableIterator t, TableIterator te)
+                : it(t), table_end(te) {}
+
+               // Allow iterator to const_iterator conversion
+               template <typename P, typename R, class TI>
+               exhashmap_iterator(const exhashmap_iterator<P, R, TI> &other)
+                : it(other.get_it_()), table_end(other.get_table_end_()) {}
+
+               typename exhashmap_iterator::reference operator*() const
+               {
+                       return it->second;
+               }
+
+               typename exhashmap_iterator::pointer operator->() const
+               {
+                       return &(it->second);
+               }
+
+               exhashmap_iterator &operator++()
+               {
+                       increment();
+                       return *this;
+               }
+
+               exhashmap_iterator operator++(int)
+               {
+                       exhashmap_iterator tmp = *this;
+                       increment();
+                       return tmp;
+               }
+
+               template <typename P, typename R, class TI>
+               bool operator==(const exhashmap_iterator<P, R, TI> &other) const
+               {
+                       return it == other.get_it_();
+               }
+
+               template <typename P, typename R, class TI>
+               bool operator!=(const exhashmap_iterator<P, R, TI> &other) const
+               {
+                       return it != other.get_it_();
+               }
+
+               // Private access function
+               TableIterator get_it_() const { return it; }
+               TableIterator get_table_end_() const { return table_end; }
+
+       protected:
+               TableIterator it;        ///< Pointer to current bucket
+               TableIterator table_end; ///< Pointer to one-past-last bucket
+
+               void increment()
+               {
+                       if (it != table_end)
+                               ++it;
+
+                       // Skip empty and erased buckets
+                       while (it != table_end && it->first != USED)
+                               ++it;
+               }
+       };
+
+       typedef exhashmap_iterator<value_type*, value_type&, table_iterator> iterator;
+       typedef exhashmap_iterator<const value_type*, const value_type&, table_const_iterator> const_iterator;
+
+       // More standard types
+       typedef typename Table::size_type size_type;
+       typedef typename Table::difference_type difference_type;
+
+       class value_compare : public std::binary_function<value_type, value_type, bool>, private key_compare {
+               friend class exhashmap;
+       public:
+               bool operator()(const value_type &lhs, const value_type &rhs) const
+               {
+                       return key_compare::operator()(lhs.first, rhs.first);
+               }
+
+               bool operator()(const key_type &lhs, const value_type &rhs) const
+               {
+                       return key_compare::operator()(lhs, rhs.first);
+               }
+
+               bool operator()(const value_type &lhs, const key_type &rhs) const
+               {
+                       return key_compare::operator()(lhs.first, rhs);
+               }
+       };
+
+protected:
+       // Private data
+       Table hashtab;         ///< Vector of buckets, each bucket is kept sorted
+       size_type num_buckets; ///< Number of buckets (= hashtab.size())
+       size_type num_entries; ///< Number of values stored in container (cached for faster operation of size())
+
+       /** Return index of key in hash table. */
+       static size_type hash_index(const key_type &x, size_type nbuckets)
+       {
+               return x.gethash() % nbuckets;
+       }
+
+       static table_iterator find_bucket(const key_type &x, table_iterator tab, size_type nbuckets);
+       static table_const_iterator find_bucket(const key_type &x, table_const_iterator tab, size_type nbuckets);
+       static table_iterator find_bucket_for_insertion(const key_type &x, table_iterator tab, size_type nbuckets);
+
+       /** Return pointer to bucket corresponding to key (or first empty bucket). */
+       table_iterator find_bucket(const key_type &x)
+       {
+               return find_bucket(x, hashtab.begin(), num_buckets);
+       }
+
+       /** Return pointer to bucket corresponding to key (or first empty bucket). */
+       table_const_iterator find_bucket(const key_type &x) const
+       {
+               return find_bucket(x, hashtab.begin(), num_buckets);
+       }
+
+       /** Return pointer to bucket corresponding to key (or first empty or erased bucket). */
+       table_iterator find_bucket_for_insertion(const key_type &x)
+       {
+               return find_bucket_for_insertion(x, hashtab.begin(), num_buckets);
+       }
+
+       /** Return number of entries above which the table will grow. */
+       size_type hwm() const
+       {
+               // Try to keep at least 25% of the buckets free
+               return num_buckets - (num_buckets >> 2);
+       }
+
+       /** Empty all buckets in the table. */
+       void empty_all_buckets()
+       {
+               for (table_iterator i = hashtab.begin(); i != hashtab.end(); ++i)
+                       i->first = EMPTY;
+       }
+
+       void grow();
+
+public:
+       // 23.3.1.1 Construct/copy/destroy
+       exhashmap() : num_buckets(min_num_buckets), num_entries(0)
+       {
+               hashtab.resize(num_buckets);
+               empty_all_buckets();
+       }
+
+       explicit exhashmap(size_type nbuckets) : num_entries(0)
+       {
+               num_buckets = internal::next_prime(nbuckets);
+               hashtab.resize(num_buckets);
+               empty_all_buckets();
+       }
+
+       template <class InputIterator>
+       exhashmap(InputIterator first, InputIterator last) : num_buckets(min_num_buckets), num_entries(0)
+       {
+               hashtab.resize(num_buckets);
+               empty_all_buckets();
+               insert(first, last);
+       }
+
+       exhashmap &operator=(const exhashmap &other)
+       {
+               exhashmap(other).swap(*this);
+               return *this;
+       }
+
+       // Iterators
+       iterator begin()
+       {
+               // Find first used bucket
+               table_iterator bucket = hashtab.begin();
+               while (bucket != hashtab.end() && bucket->first != USED)
+                       ++bucket;
+               return iterator(bucket, hashtab.end());
+       }
+
+       const_iterator begin() const
+       {
+               // Find first used bucket
+               table_const_iterator bucket = hashtab.begin();
+               while (bucket != hashtab.end() && bucket->first != USED)
+                       ++bucket;
+               return const_iterator(bucket, hashtab.end());
+       }
+
+       iterator end()
+       {
+               return iterator(hashtab.end(), hashtab.end());
+       }
+
+       const_iterator end() const
+       {
+               return const_iterator(hashtab.end(), hashtab.end());
+       }
+
+       // Capacity
+       bool empty() const
+       {
+               return num_entries == 0;
+       }
+
+       size_type size() const
+       {
+               return num_entries;
+       }
+
+       size_type max_size() const
+       {
+               return hashtab.max_size();
+       }
+
+       size_type bucket_count() const
+       {
+               return num_buckets;
+       }
+
+       // 23.3.1.2 Element access
+       T &operator[](const key_type &x)
+       {
+               return insert(value_type(x, mapped_type())).first->second;
+       }
+
+       // Modifiers
+       std::pair<iterator, bool> insert(const value_type &x);
+
+       template <class InputIterator>
+       void insert(InputIterator first, InputIterator last)
+       {
+               for (; first != last; ++first)
+                       insert(*first);
+       }
+
+       void erase(iterator position)
+       {
+               table_iterator bucket = position.get_it_();
+               bucket->first = ERASED;
+               bucket->second.first = 0;
+               --num_entries;
+       }
+
+       size_type erase(const key_type &x);
+
+       void swap(exhashmap &other)
+       {
+               hashtab.swap(other.hashtab);
+               std::swap(num_buckets, other.num_buckets);
+               std::swap(num_entries, other.num_entries);
+       }
+
+       void clear();
+
+       // Observers
+       key_compare key_comp() const
+       {
+               return key_compare();
+       }
+
+       value_compare value_comp() const
+       {
+               return value_compare();
+       }
+
+       // 23.3.1.3 Map operations
+       iterator find(const key_type &x);
+       const_iterator find(const key_type &x) const;
+
+       size_type count(const key_type &x) const
+       {
+               return find(x) == end() ? 0 : 1;
+       }
+
+       std::pair<iterator, iterator> equal_range(const key_type &x)
+       {
+               iterator i = find(x);
+               if (i == end())
+                       return std::make_pair(i, i);
+               else {
+                       iterator j = ++i;
+                       return std::make_pair(i, j);
+               }
+       }
+
+       std::pair<const_iterator, const_iterator> equal_range(const key_type &x) const
+       {
+               const_iterator i = find(x);
+               if (i == end())
+                       return std::make_pair(i, i);
+               else {
+                       const_iterator j = ++i;
+                       return std::make_pair(i, j);
+               }
+       }
+
+       friend bool operator==(const exhashmap &lhs, const exhashmap &rhs)
+       {
+               if (lhs.num_entries != rhs.num_entries || lhs.num_buckets != rhs.num_buckets)
+                       return false;
+
+               // We can't compare the tables directly as the elements may be
+               // in different order due to the collision handling. We therefore
+               // look up each value from the lhs map in the rhs map separately.
+               for (const_iterator itl = lhs.begin(); itl != lhs.end(); ++itl) {
+                       const_iterator itr = rhs.find(itl->first);
+                       if (itr == rhs.end())
+                               return false;
+                       if (itl->second != itr->second)
+                               return false;
+               }
+               return true;
+       }
+
+       friend bool operator!=(const exhashmap &lhs, const exhashmap &rhs)
+       {
+               return !(lhs == rhs);
+       }
+
+       void dump() const
+       {
+               std::clog << "num_entries = " << num_entries << std::endl;
+               std::clog << "num_buckets = " << num_buckets << std::endl;
+               size_type t = 0;
+               for (table_const_iterator it = hashtab.begin(); it != hashtab.end(); ++it, ++t) {
+                       std::clog << " bucket " << t << ": ";
+                       std::clog << (it->first == EMPTY ? "free" : (it->first == USED ? "used" : "erased")) << ", " << it->second.first << " -> " << it->second.second << std::endl;
+               }
+       }
+};
+
+/** Return pointer to bucket corresponding to key (or first empty bucket). */
+template <typename T, template <class> class A>
+inline typename exhashmap<T, A>::table_iterator exhashmap<T, A>::find_bucket(const key_type &x, table_iterator tab, size_type nbuckets)
+{
+       // Quadratic probing
+       size_type h = hash_index(x, nbuckets);
+       size_type d = 1;
+       table_iterator it = tab + h;
+       while (it->first != EMPTY && !(it->first == USED && key_equal()(it->second.first, x))) {
+               h = (h + d) % nbuckets;
+               d += 2;
+               it = tab + h;
+       }
+       return it;
+}
+
+/** Return pointer to bucket corresponding to key (or first empty bucket). */
+template <typename T, template <class> class A>
+inline typename exhashmap<T, A>::table_const_iterator exhashmap<T, A>::find_bucket(const key_type &x, table_const_iterator tab, size_type nbuckets)
+{
+       // Quadratic probing
+       size_type h = hash_index(x, nbuckets);
+       size_type d = 1;
+       table_const_iterator it = tab + h;
+       while (it->first != EMPTY && !(it->first == USED && key_equal()(it->second.first, x))) {
+               h = (h + d) % nbuckets;
+               d += 2;
+               it = tab + h;
+       }
+       return it;
+}
+
+/** Return pointer to bucket corresponding to key (or first empty or erased bucket). */
+template <typename T, template <class> class A>
+inline typename exhashmap<T, A>::table_iterator exhashmap<T, A>::find_bucket_for_insertion(const key_type &x, table_iterator tab, size_type nbuckets)
+{
+       // Quadratic probing
+       size_type h = hash_index(x, nbuckets);
+       size_type d = 1;
+       table_iterator it = tab + h;
+       while (it->first != EMPTY && !key_equal()(it->second.first, x)) {
+               h = (h + d) % nbuckets;
+               d += 2;
+               it = tab + h;
+       }
+       return it;
+}
+
+/** Grow hash table */
+template <typename T, template <class> class A>
+void exhashmap<T, A>::grow()
+{
+       // Allocate new empty hash table
+       size_type new_num_buckets = internal::next_prime(num_buckets + 1);
+       Table new_hashtab;
+       new_hashtab.resize(new_num_buckets);
+       for (table_iterator it = new_hashtab.begin(); it != new_hashtab.end(); ++it)
+               it->first = EMPTY;
+
+       // Re-insert all elements into new table
+       for (table_const_iterator it = hashtab.begin(); it != hashtab.end(); ++it) {
+               if (it->first == USED) {
+                       table_iterator bucket = find_bucket(it->second.first, new_hashtab.begin(), new_num_buckets);
+                       *bucket = *it;
+               }
+       }
+
+       // Swap with the old table
+       hashtab.swap(new_hashtab);
+       num_buckets = new_num_buckets;
+}
+
+template <typename T, template <class> class A>
+std::pair<typename exhashmap<T, A>::iterator, bool> exhashmap<T, A>::insert(const value_type &x)
+{
+       table_iterator bucket = find_bucket_for_insertion(x.first);
+       if (bucket->first == USED) {
+               // Value already in map
+               return std::make_pair(iterator(bucket, hashtab.end()), false);
+       } else {
+               // Insert new value
+               bucket->first = USED;
+               bucket->second = x;
+               ++num_entries;
+               if (num_entries >= hwm()) {
+                       grow();
+                       bucket = find_bucket(x.first);
+               }
+               return std::make_pair(iterator(bucket, hashtab.end()), true);
+       }
+}
+
+template <typename T, template <class> class A>
+typename exhashmap<T, A>::size_type exhashmap<T, A>::erase(const key_type &x)
+{
+       iterator i = find(x);
+       if (i != end()) {
+               erase(i);
+               return 1;
+       } else
+               return 0;
+}
+
+template <typename T, template <class> class A>
+typename exhashmap<T, A>::iterator exhashmap<T, A>::find(const key_type &x)
+{
+       table_iterator bucket = find_bucket(x);
+       if (bucket->first == USED)
+               return iterator(bucket, hashtab.end());
+       else
+               return end();
+}
+
+template <typename T, template <class> class A>
+typename exhashmap<T, A>::const_iterator exhashmap<T, A>::find(const key_type &x) const
+{
+       table_const_iterator bucket = find_bucket(x);
+       if (bucket->first == USED)
+               return const_iterator(bucket, hashtab.end());
+       else
+               return end();
+}
+
+template <typename T, template <class> class A>
+void exhashmap<T, A>::clear()
+{
+       for (table_iterator i = hashtab.begin(); i != hashtab.end(); ++i) {
+               i->first = EMPTY;
+               i->second.first = 0;
+               i->second.second = T();
+       }
+       num_entries = 0;
+}
+
+} // namespace GiNaC
+
+
+// Specializations of Standard Library algorithms
+namespace std {
+
+/** Specialization of std::swap() for exhashmap. */
+template <typename T, template <class> class A>
+inline void swap(GiNaC::exhashmap<T, A> &lhs, GiNaC::exhashmap<T, A> &rhs)
+{
+       lhs.swap(rhs);
+}
+
+} // namespace std
+
+#endif // ndef __GINAC_HASH_MAP_H__