⟦2130c46d0⟧

TextFile

// This may look like C code, but it is really -*- C++ -*-
/* 
Copyright (C) 1988 Free Software Foundation
    written by Doug Lea (dl@rocky.oswego.edu)
     based on code by Doug Schmidt

This file is part of GNU CC.

GNU CC is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY.  No author or distributor
accepts responsibility to anyone for the consequences of using it
or for whether it serves any particular purpose or works at all,
unless he says so in writing.  Refer to the GNU CC General Public
License for full details.

Everyone is granted permission to copy, modify and redistribute
GNU CC, but only under the conditions described in the
GNU CC General Public License.   A copy of this license is
supposed to have been given to you along with GNU CC so you
can know your rights and responsibilities.  It should be in a
file named COPYING.  Among other things, the copyright notice
and this notice must be preserved on all copies.  
*/

#include <stream.h>
#include "<T>.VOHSet.h"


/* codes for status fields */
#define EMPTYCELL   0
#define VALIDCELL   1
#define DELETEDCELL 2


<T>VOHSet::<T>VOHSet(int sz = DEFAULT_INITIAL_CAPACITY)
{
// The size of the hash table is always the smallest power of 2 >= the size
// indicated by the user.  This allows several optimizations, including
// the use of actual double hashing and elimination of the mod instruction.

  size = 1;
  while (size < sz) size <<= 1;
  tab = new <T>[size];
  status = new char[size];
  for (int i = 0; i < size; ++i) status[i] = EMPTYCELL;
  count = cnt = 0;
}

<T>VOHSet::<T>VOHSet(<T>VOHSet& a)
{
  tab = new <T>[size = a.size];
  status = new char[size];
  for (int i = 0; i < size; ++i) status[i] = EMPTYCELL;
  count = cnt = 0;
  for (Pix p = a.first(); p; a.next(p)) add(a(p));
}

Pix <T>VOHSet::seek(<T&> key)
{
// Uses ordered double hashing to perform a search of the table.
// This greatly speeds up the average-case time for an unsuccessful search.

  unsigned hashval = <T>HASH(key);

  // We can avoid the mod operation since size is a power of 2.
  unsigned h = hashval & (size - 1);

  // The increment must be odd, since all odd numbers are relatively
  // prime to a power of 2!!
  unsigned inc = ((((hashval / size) << 1) + 1) & (size - 1));
  
  // There is always at least 1 empty cell, so this loop is guaranteed to halt!
  while (status[h] != EMPTYCELL)
  {
    int cmp = <T>CMP (key, tab[h]);
    if (cmp == 0)
    {
      if (status[h] == VALIDCELL)
        return Pix(&tab[h]);
      else
        return 0;
    }
    else if (cmp > 0)
      return 0;
    else
      h = ((h + inc) & (size - 1));
  }
  return 0;
}

// This adds an item if it doesn't already exist.  By performing the initial
// comparison we assure that the table always contains at least 1 empty
// spot.  This speeds up later searching by a constant factor.
// The insertion algorithm uses ordered double hashing.  See Standish's
// 1980 ``Data Structure's Techniques'' book for details.

Pix <T>VOHSet::add(<T&> x)
{
  if (size <= cnt+1) 
    resize();
  
  unsigned hashval = <T>HASH(x);
  unsigned h = hashval & (size - 1);

  if (status[h] != VALIDCELL)   // save some work if possible
  {
    if (status[h] == EMPTYCELL)
      cnt++;
    count++;
    tab[h] = x; 
    status[h] = VALIDCELL; 
    return Pix(&tab[h]);
  }
  int cmp = <T>CMP(x, tab[h]);
  if (cmp == 0)
    return Pix(&tab[h]);

  <T> item = x;
  Pix mypix = 0;
  unsigned inc = ((((hashval / size) << 1) + 1) & (size - 1));

  for (;;)
  {
    if (cmp < 0)
    {
      <T> temp = tab[h];
      tab[h] = item;
      item = temp;
      if (mypix == 0) mypix = Pix(&tab[h]);
      inc = ((((<T>HASH(item) / size) << 1) + 1) & (size - 1));
      h = ((h + inc) & (size - 1));
      cmp = <T>CMP(item, tab[h]);
    }
    else
      h = ((h + inc) & (size - 1));
    if (status[h] != VALIDCELL)
    {
      if (status[h] == EMPTYCELL)
        cnt++;
      count++;
      tab[h] = item; 
      status[h] = VALIDCELL; 
      return (mypix == 0)? Pix(&tab[h]) : mypix;
    }
  }
}


void <T>VOHSet::del(<T&> key)
{
// This performs a deletion by marking the item's status field.
// Note that we only decrease the count, *not* the cnt, since this
// would cause trouble for subsequent steps in the algorithm.  See
// Reingold and Hanson's ``Data Structure's'' book for a justification
// of this approach.

  unsigned hashval = <T>HASH(key);
  unsigned h   = hashval & (size - 1);
  unsigned inc = ((((hashval / size) << 1) + 1) & (size - 1));
  
  while (status[h] != EMPTYCELL)
  {
    int cmp = <T>CMP(key, tab[h]);
    if (cmp < 0)
      h = ((h + inc) & (size - 1));
    else if (status[h] == VALIDCELL && cmp == 0)
    {
      status[h] = DELETEDCELL;
      count--;
      return;
    }
    else 
      return;
  }
}

void <T>VOHSet::clear()
{
  for (int i = 0; i < size; ++i) status[i] = EMPTYCELL;
  count = cnt = 0;
}

void <T>VOHSet::resize(int newsize = 0)
{
  if (newsize <= count)
    newsize = count;
  int s = 1;
  while (s <= newsize) s <<= 1;
  newsize = s;

  <T>* oldtab = tab;
  char* oldstatus = status;
  int oldsize = size;
  tab = new <T>[size = newsize];
  status = new char[size];
  for (int i = 0; i < size; ++i) status[i] = EMPTYCELL;
  count = cnt = 0;
  
  for (i = 0; i < oldsize; ++i) if (oldstatus[i] == VALIDCELL) add(oldtab[i]);
  delete [oldsize] oldtab;
  delete oldstatus;
}

Pix <T>VOHSet::first()
{
  for (int pos = 0; pos < size; ++pos)
    if (status[pos] == VALIDCELL) return Pix(&tab[pos]);
  return 0;
}

void <T>VOHSet::next(Pix& i)
{
  if (i == 0) return;
  int pos = ((unsigned)i - (unsigned)tab) / sizeof(<T>) + 1;
  for (; pos < size; ++pos)
    if (status[pos] == VALIDCELL)
    {
      i = Pix(&tab[pos]);
      return;
    }
  i = 0;
}


int <T>VOHSet:: operator == (<T>VOHSet& b)
{
  if (count != b.count)
    return 0;
  else
  {
    for (int i = 0; i < size; ++i)
      if (status[i] == VALIDCELL && b.seek(tab[i]) == 0)
          return 0;
    for (i = 0; i < b.size; ++i)
      if (b.status[i] == VALIDCELL && seek(b.tab[i]) == 0)
          return 0;
    return 1;
  }
}

int <T>VOHSet:: operator != (<T>VOHSet& b)
{
  return !(*this == b);
}

int <T>VOHSet::operator <= (<T>VOHSet& b)
{
  if (count > b.count)
    return 0;
  else
  {
    for (int i = 0; i < size; ++i)
      if (status[i] == VALIDCELL && b.seek(tab[i]) == 0)
          return 0;
    return 1;
  }
}

void <T>VOHSet::operator |= (<T>VOHSet& b)
{
  if (&b == this || b.count == 0)
    return;
  for (int i = 0; i < b.size; ++i)
    if (b.status[i] == VALIDCELL) add(b.tab[i]);
}

void <T>VOHSet::operator &= (<T>VOHSet& b)
{
  if (&b == this || count == 0)
    return;
  for (int i = 0; i < size; ++i)
  {
    if (status[i] == VALIDCELL && b.seek(tab[i]) == 0)
    {
      status[i] = DELETEDCELL;
      --count;
    }
  }
}

void <T>VOHSet::operator -= (<T>VOHSet& b)
{
  for (int i = 0; i < size; ++i)
  {
    if (status[i] == VALIDCELL && b.seek(tab[i]) != 0)
    {
      status[i] = DELETEDCELL;
      --count;
    }
  }
}

int <T>VOHSet::OK()
{
  int v = tab != 0;
  v &= status != 0;
  int n = 0;
  for (int i = 0; i < size; ++i) 
  {
    if (status[i] == VALIDCELL) ++n;
    else if (status[i] != DELETEDCELL && status[i] != EMPTYCELL)
      v = 0;
  }
  v &= n == count;
  if (!v) error("invariant failure");
  return v;
}
DataMuseum.dk

DKUUG/EUUG Conference tapes

⟦2130c46d0⟧ TextFile

Derivation

TextFile
