mfbt/BloomFilter.h

   1 /* -*- Mode: C++; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 2 -*- */
   2 /* vim: set ts=8 sts=2 et sw=2 tw=80: */
   3 /* This Source Code Form is subject to the terms of the Mozilla Public
   4  * License, v. 2.0. If a copy of the MPL was not distributed with this
   5  * file, You can obtain one at http://mozilla.org/MPL/2.0/. */
   6
   7 /*
   8  * A counting Bloom filter implementation.  This allows consumers to
   9  * do fast probabilistic "is item X in set Y?" testing which will
  10  * never answer "no" when the correct answer is "yes" (but might
  11  * incorrectly answer "yes" when the correct answer is "no").
  12  */
  13
  14 #ifndef mozilla_BloomFilter_h
  15 #define mozilla_BloomFilter_h
  16
  17 #include "mozilla/Assertions.h"
  18 #include "mozilla/Likely.h"
  19
  20 #include <stdint.h>
  21 #include <string.h>
  22
  23 namespace mozilla {
  24
  25 /*
  26  * This class implements a counting Bloom filter as described at
  27  * <http://en.wikipedia.org/wiki/Bloom_filter#Counting_filters>, with
  28  * 8-bit counters.  This allows quick probabilistic answers to the
  29  * question "is object X in set Y?" where the contents of Y might not
  30  * be time-invariant.  The probabilistic nature of the test means that
  31  * sometimes the answer will be "yes" when it should be "no".  If the
  32  * answer is "no", then X is guaranteed not to be in Y.
  33  *
  34  * The filter is parametrized on KeySize, which is the size of the key
  35  * generated by each of hash functions used by the filter, in bits,
  36  * and the type of object T being added and removed.  T must implement
  37  * a |uint32_t hash() const| method which returns a uint32_t hash key
  38  * that will be used to generate the two separate hash functions for
  39  * the Bloom filter.  This hash key MUST be well-distributed for good
  40  * results!  KeySize is not allowed to be larger than 16.
  41  *
  42  * The filter uses exactly 2**KeySize bytes of memory.  From now on we
  43  * will refer to the memory used by the filter as M.
  44  *
  45  * The expected rate of incorrect "yes" answers depends on M and on
  46  * the number N of objects in set Y.  As long as N is small compared
  47  * to M, the rate of such answers is expected to be approximately
  48  * 4*(N/M)**2 for this filter.  In practice, if Y has a few hundred
  49  * elements then using a KeySize of 12 gives a reasonably low
  50  * incorrect answer rate.  A KeySize of 12 has the additional benefit
  51  * of using exactly one page for the filter in typical hardware
  52  * configurations.
  53  */
  54
  55 template <unsigned KeySize, class T>
  56 class BloomFilter {
  57   /*
  58    * A counting Bloom filter with 8-bit counters.  For now we assume
  59    * that having two hash functions is enough, but we may revisit that
  60    * decision later.
  61    *
  62    * The filter uses an array with 2**KeySize entries.
  63    *
  64    * Assuming a well-distributed hash function, a Bloom filter with
  65    * array size M containing N elements and
  66    * using k hash function has expected false positive rate exactly
  67    *
  68    * $  (1 - (1 - 1/M)^{kN})^k  $
  69    *
  70    * because each array slot has a
  71    *
  72    * $  (1 - 1/M)^{kN}  $
  73    *
  74    * chance of being 0, and the expected false positive rate is the
  75    * probability that all of the k hash functions will hit a nonzero
  76    * slot.
  77    *
  78    * For reasonable assumptions (M large, kN large, which should both
  79    * hold if we're worried about false positives) about M and kN this
  80    * becomes approximately
  81    *
  82    * $$  (1 - \exp(-kN/M))^k   $$
  83    *
  84    * For our special case of k == 2, that's $(1 - \exp(-2N/M))^2$,
  85    * or in other words
  86    *
  87    * $$    N/M = -0.5 * \ln(1 - \sqrt(r))   $$
  88    *
  89    * where r is the false positive rate.  This can be used to compute
  90    * the desired KeySize for a given load N and false positive rate r.
  91    *
  92    * If N/M is assumed small, then the false positive rate can
  93    * further be approximated as 4*N^2/M^2.  So increasing KeySize by
  94    * 1, which doubles M, reduces the false positive rate by about a
  95    * factor of 4, and a false positive rate of 1% corresponds to
  96    * about M/N == 20.
  97    *
  98    * What this means in practice is that for a few hundred keys using a
  99    * KeySize of 12 gives false positive rates on the order of 0.25-4%.
 100    *
 101    * Similarly, using a KeySize of 10 would lead to a 4% false
 102    * positive rate for N == 100 and to quite bad false positive
 103    * rates for larger N.
 104    */
 105  public:
 106   BloomFilter() {
 107     static_assert(KeySize <= kKeyShift, "KeySize too big");
 108
 109     // Should we have a custom operator new using calloc instead and
 110     // require that we're allocated via the operator?
 111     clear();
 112   }
 113
 114   /*
 115    * Clear the filter.  This should be done before reusing it, because
 116    * just removing all items doesn't clear counters that hit the upper
 117    * bound.
 118    */
 119   void clear();
 120
 121   /*
 122    * Add an item to the filter.
 123    */
 124   void add(const T* aValue);
 125
 126   /*
 127    * Remove an item from the filter.
 128    */
 129   void remove(const T* aValue);
 130
 131   /*
 132    * Check whether the filter might contain an item.  This can
 133    * sometimes return true even if the item is not in the filter,
 134    * but will never return false for items that are actually in the
 135    * filter.
 136    */
 137   bool mightContain(const T* aValue) const;
 138
 139   /*
 140    * Methods for add/remove/contain when we already have a hash computed
 141    */
 142   void add(uint32_t aHash);
 143   void remove(uint32_t aHash);
 144   bool mightContain(uint32_t aHash) const;
 145
 146  private:
 147   static const size_t kArraySize = (1 << KeySize);
 148   static const uint32_t kKeyMask = (1 << KeySize) - 1;
 149   static const uint32_t kKeyShift = 16;
 150
 151   static uint32_t hash1(uint32_t aHash) { return aHash & kKeyMask; }
 152   static uint32_t hash2(uint32_t aHash) {
 153     return (aHash >> kKeyShift) & kKeyMask;
 154   }
 155
 156   uint8_t& firstSlot(uint32_t aHash) { return mCounters[hash1(aHash)]; }
 157   uint8_t& secondSlot(uint32_t aHash) { return mCounters[hash2(aHash)]; }
 158
 159   const uint8_t& firstSlot(uint32_t aHash) const {
 160     return mCounters[hash1(aHash)];
 161   }
 162   const uint8_t& secondSlot(uint32_t aHash) const {
 163     return mCounters[hash2(aHash)];
 164   }
 165
 166   static bool full(const uint8_t& aSlot) { return aSlot == UINT8_MAX; }
 167
 168   uint8_t mCounters[kArraySize];
 169 };
 170
 171 template <unsigned KeySize, class T>
 172 inline void BloomFilter<KeySize, T>::clear() {
 173   memset(mCounters, 0, kArraySize);
 174 }
 175
 176 template <unsigned KeySize, class T>
 177 inline void BloomFilter<KeySize, T>::add(uint32_t aHash) {
 178   uint8_t& slot1 = firstSlot(aHash);
 179   if (MOZ_LIKELY(!full(slot1))) {
 180     ++slot1;
 181   }
 182   uint8_t& slot2 = secondSlot(aHash);
 183   if (MOZ_LIKELY(!full(slot2))) {
 184     ++slot2;
 185   }
 186 }
 187
 188 template <unsigned KeySize, class T>
 189 MOZ_ALWAYS_INLINE void BloomFilter<KeySize, T>::add(const T* aValue) {
 190   uint32_t hash = aValue->hash();
 191   return add(hash);
 192 }
 193
 194 template <unsigned KeySize, class T>
 195 inline void BloomFilter<KeySize, T>::remove(uint32_t aHash) {
 196   // If the slots are full, we don't know whether we bumped them to be
 197   // there when we added or not, so just leave them full.
 198   uint8_t& slot1 = firstSlot(aHash);
 199   if (MOZ_LIKELY(!full(slot1))) {
 200     --slot1;
 201   }
 202   uint8_t& slot2 = secondSlot(aHash);
 203   if (MOZ_LIKELY(!full(slot2))) {
 204     --slot2;
 205   }
 206 }
 207
 208 template <unsigned KeySize, class T>
 209 MOZ_ALWAYS_INLINE void BloomFilter<KeySize, T>::remove(const T* aValue) {
 210   uint32_t hash = aValue->hash();
 211   remove(hash);
 212 }
 213
 214 template <unsigned KeySize, class T>
 215 MOZ_ALWAYS_INLINE bool BloomFilter<KeySize, T>::mightContain(
 216     uint32_t aHash) const {
 217   // Check that all the slots for this hash contain something
 218   return firstSlot(aHash) && secondSlot(aHash);
 219 }
 220
 221 template <unsigned KeySize, class T>
 222 MOZ_ALWAYS_INLINE bool BloomFilter<KeySize, T>::mightContain(
 223     const T* aValue) const {
 224   uint32_t hash = aValue->hash();
 225   return mightContain(hash);
 226 }
 227
 228 }  // namespace mozilla
 229
 230 #endif /* mozilla_BloomFilter_h */