dict2.cpp revision 1472:c18cbe5936b8
1/* 2 * Copyright (c) 1998, 2009, Oracle and/or its affiliates. All rights reserved. 3 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER. 4 * 5 * This code is free software; you can redistribute it and/or modify it 6 * under the terms of the GNU General Public License version 2 only, as 7 * published by the Free Software Foundation. 8 * 9 * This code is distributed in the hope that it will be useful, but WITHOUT 10 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or 11 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License 12 * version 2 for more details (a copy is included in the LICENSE file that 13 * accompanied this code). 14 * 15 * You should have received a copy of the GNU General Public License version 16 * 2 along with this work; if not, write to the Free Software Foundation, 17 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA. 18 * 19 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA 20 * or visit www.oracle.com if you need additional information or have any 21 * questions. 22 * 23 */ 24 25// Dictionaries - An Abstract Data Type 26 27#include "adlc.hpp" 28 29// #include "dict.hpp" 30 31 32//------------------------------data----------------------------------------- 33// String hash tables 34#define MAXID 20 35static char initflag = 0; // True after 1st initialization 36static char shft[MAXID] = {1,2,3,4,5,6,7,1,2,3,4,5,6,7,1,2,3,4,5,6}; 37static short xsum[MAXID + 1]; 38 39//------------------------------bucket--------------------------------------- 40class bucket { 41public: 42 int _cnt, _max; // Size of bucket 43 const void **_keyvals; // Array of keys and values 44}; 45 46//------------------------------Dict----------------------------------------- 47// The dictionary is kept has a hash table. The hash table is a even power 48// of two, for nice modulo operations. Each bucket in the hash table points 49// to a linear list of key-value pairs; each key & value is just a (void *). 50// The list starts with a count. A hash lookup finds the list head, then a 51// simple linear scan finds the key. If the table gets too full, it's 52// doubled in size; the total amount of EXTRA times all hash functions are 53// computed for the doubling is no more than the current size - thus the 54// doubling in size costs no more than a constant factor in speed. 55Dict::Dict(CmpKey initcmp, Hash inithash) : _hash(inithash), _cmp(initcmp), _arena(NULL) { 56 init(); 57} 58 59Dict::Dict(CmpKey initcmp, Hash inithash, Arena *arena) : _hash(inithash), _cmp(initcmp), _arena(arena) { 60 init(); 61} 62 63void Dict::init() { 64 int i; 65 66 // Precompute table of null character hashes 67 if( !initflag ) { // Not initializated yet? 68 xsum[0] = (1<<shft[0])+1; // Initialize 69 for( i = 1; i < MAXID + 1; i++) { 70 xsum[i] = (1<<shft[i])+1+xsum[i-1]; 71 } 72 initflag = 1; // Never again 73 } 74 75 _size = 16; // Size is a power of 2 76 _cnt = 0; // Dictionary is empty 77 _bin = (bucket*)_arena->Amalloc_4(sizeof(bucket)*_size); 78 memset(_bin,0,sizeof(bucket)*_size); 79} 80 81//------------------------------~Dict------------------------------------------ 82// Delete an existing dictionary. 83Dict::~Dict() { 84} 85 86//------------------------------Clear---------------------------------------- 87// Zap to empty; ready for re-use 88void Dict::Clear() { 89 _cnt = 0; // Empty contents 90 for( int i=0; i<_size; i++ ) 91 _bin[i]._cnt = 0; // Empty buckets, but leave allocated 92 // Leave _size & _bin alone, under the assumption that dictionary will 93 // grow to this size again. 94} 95 96//------------------------------doubhash--------------------------------------- 97// Double hash table size. If can't do so, just suffer. If can, then run 98// thru old hash table, moving things to new table. Note that since hash 99// table doubled, exactly 1 new bit is exposed in the mask - so everything 100// in the old table ends up on 1 of two lists in the new table; a hi and a 101// lo list depending on the value of the bit. 102void Dict::doubhash(void) { 103 int oldsize = _size; 104 _size <<= 1; // Double in size 105 _bin = (bucket*)_arena->Arealloc( _bin, sizeof(bucket)*oldsize, sizeof(bucket)*_size ); 106 memset( &_bin[oldsize], 0, oldsize*sizeof(bucket) ); 107 // Rehash things to spread into new table 108 for( int i=0; i < oldsize; i++) { // For complete OLD table do 109 bucket *b = &_bin[i]; // Handy shortcut for _bin[i] 110 if( !b->_keyvals ) continue; // Skip empties fast 111 112 bucket *nb = &_bin[i+oldsize]; // New bucket shortcut 113 int j = b->_max; // Trim new bucket to nearest power of 2 114 while( j > b->_cnt ) j >>= 1; // above old bucket _cnt 115 if( !j ) j = 1; // Handle zero-sized buckets 116 nb->_max = j<<1; 117 // Allocate worst case space for key-value pairs 118 nb->_keyvals = (const void**)_arena->Amalloc_4( sizeof(void *)*nb->_max*2 ); 119 int nbcnt = 0; 120 121 for( j=0; j<b->_cnt; j++ ) { // Rehash all keys in this bucket 122 const void *key = b->_keyvals[j+j]; 123 if( (_hash( key ) & (_size-1)) != i ) { // Moving to hi bucket? 124 nb->_keyvals[nbcnt+nbcnt] = key; 125 nb->_keyvals[nbcnt+nbcnt+1] = b->_keyvals[j+j+1]; 126 nb->_cnt = nbcnt = nbcnt+1; 127 b->_cnt--; // Remove key/value from lo bucket 128 b->_keyvals[j+j ] = b->_keyvals[b->_cnt+b->_cnt ]; 129 b->_keyvals[j+j+1] = b->_keyvals[b->_cnt+b->_cnt+1]; 130 j--; // Hash compacted element also 131 } 132 } // End of for all key-value pairs in bucket 133 } // End of for all buckets 134 135 136} 137 138//------------------------------Dict----------------------------------------- 139// Deep copy a dictionary. 140Dict::Dict( const Dict &d ) : _size(d._size), _cnt(d._cnt), _hash(d._hash),_cmp(d._cmp), _arena(d._arena) { 141 _bin = (bucket*)_arena->Amalloc_4(sizeof(bucket)*_size); 142 memcpy( _bin, d._bin, sizeof(bucket)*_size ); 143 for( int i=0; i<_size; i++ ) { 144 if( !_bin[i]._keyvals ) continue; 145 _bin[i]._keyvals=(const void**)_arena->Amalloc_4( sizeof(void *)*_bin[i]._max*2); 146 memcpy( _bin[i]._keyvals, d._bin[i]._keyvals,_bin[i]._cnt*2*sizeof(void*)); 147 } 148} 149 150//------------------------------Dict----------------------------------------- 151// Deep copy a dictionary. 152Dict &Dict::operator =( const Dict &d ) { 153 if( _size < d._size ) { // If must have more buckets 154 _arena = d._arena; 155 _bin = (bucket*)_arena->Arealloc( _bin, sizeof(bucket)*_size, sizeof(bucket)*d._size ); 156 memset( &_bin[_size], 0, (d._size-_size)*sizeof(bucket) ); 157 _size = d._size; 158 } 159 for( int i=0; i<_size; i++ ) // All buckets are empty 160 _bin[i]._cnt = 0; // But leave bucket allocations alone 161 _cnt = d._cnt; 162 *(Hash*)(&_hash) = d._hash; 163 *(CmpKey*)(&_cmp) = d._cmp; 164 for(int k=0; k<_size; k++ ) { 165 bucket *b = &d._bin[k]; // Shortcut to source bucket 166 for( int j=0; j<b->_cnt; j++ ) 167 Insert( b->_keyvals[j+j], b->_keyvals[j+j+1] ); 168 } 169 return *this; 170} 171 172//------------------------------Insert--------------------------------------- 173// Insert or replace a key/value pair in the given dictionary. If the 174// dictionary is too full, it's size is doubled. The prior value being 175// replaced is returned (NULL if this is a 1st insertion of that key). If 176// an old value is found, it's swapped with the prior key-value pair on the 177// list. This moves a commonly searched-for value towards the list head. 178const void *Dict::Insert(const void *key, const void *val) { 179 int hash = _hash( key ); // Get hash key 180 int i = hash & (_size-1); // Get hash key, corrected for size 181 bucket *b = &_bin[i]; // Handy shortcut 182 for( int j=0; j<b->_cnt; j++ ) 183 if( !_cmp(key,b->_keyvals[j+j]) ) { 184 const void *prior = b->_keyvals[j+j+1]; 185 b->_keyvals[j+j ] = key; // Insert current key-value 186 b->_keyvals[j+j+1] = val; 187 return prior; // Return prior 188 } 189 190 if( ++_cnt > _size ) { // Hash table is full 191 doubhash(); // Grow whole table if too full 192 i = hash & (_size-1); // Rehash 193 b = &_bin[i]; // Handy shortcut 194 } 195 if( b->_cnt == b->_max ) { // Must grow bucket? 196 if( !b->_keyvals ) { 197 b->_max = 2; // Initial bucket size 198 b->_keyvals = (const void**)_arena->Amalloc_4( sizeof(void *)*b->_max*2 ); 199 } else { 200 b->_keyvals = (const void**)_arena->Arealloc( b->_keyvals, sizeof(void *)*b->_max*2, sizeof(void *)*b->_max*4 ); 201 b->_max <<= 1; // Double bucket 202 } 203 } 204 b->_keyvals[b->_cnt+b->_cnt ] = key; 205 b->_keyvals[b->_cnt+b->_cnt+1] = val; 206 b->_cnt++; 207 return NULL; // Nothing found prior 208} 209 210//------------------------------Delete--------------------------------------- 211// Find & remove a value from dictionary. Return old value. 212const void *Dict::Delete(void *key) { 213 int i = _hash( key ) & (_size-1); // Get hash key, corrected for size 214 bucket *b = &_bin[i]; // Handy shortcut 215 for( int j=0; j<b->_cnt; j++ ) 216 if( !_cmp(key,b->_keyvals[j+j]) ) { 217 const void *prior = b->_keyvals[j+j+1]; 218 b->_cnt--; // Remove key/value from lo bucket 219 b->_keyvals[j+j ] = b->_keyvals[b->_cnt+b->_cnt ]; 220 b->_keyvals[j+j+1] = b->_keyvals[b->_cnt+b->_cnt+1]; 221 _cnt--; // One less thing in table 222 return prior; 223 } 224 return NULL; 225} 226 227//------------------------------FindDict------------------------------------- 228// Find a key-value pair in the given dictionary. If not found, return NULL. 229// If found, move key-value pair towards head of list. 230const void *Dict::operator [](const void *key) const { 231 int i = _hash( key ) & (_size-1); // Get hash key, corrected for size 232 bucket *b = &_bin[i]; // Handy shortcut 233 for( int j=0; j<b->_cnt; j++ ) 234 if( !_cmp(key,b->_keyvals[j+j]) ) 235 return b->_keyvals[j+j+1]; 236 return NULL; 237} 238 239//------------------------------CmpDict-------------------------------------- 240// CmpDict compares two dictionaries; they must have the same keys (their 241// keys must match using CmpKey) and they must have the same values (pointer 242// comparison). If so 1 is returned, if not 0 is returned. 243int Dict::operator ==(const Dict &d2) const { 244 if( _cnt != d2._cnt ) return 0; 245 if( _hash != d2._hash ) return 0; 246 if( _cmp != d2._cmp ) return 0; 247 for( int i=0; i < _size; i++) { // For complete hash table do 248 bucket *b = &_bin[i]; // Handy shortcut 249 if( b->_cnt != d2._bin[i]._cnt ) return 0; 250 if( memcmp(b->_keyvals, d2._bin[i]._keyvals, b->_cnt*2*sizeof(void*) ) ) 251 return 0; // Key-value pairs must match 252 } 253 return 1; // All match, is OK 254} 255 256 257//------------------------------print---------------------------------------- 258static void printvoid(const void* x) { printf("%p", x); } 259void Dict::print() { 260 print(printvoid, printvoid); 261} 262void Dict::print(PrintKeyOrValue print_key, PrintKeyOrValue print_value) { 263 for( int i=0; i < _size; i++) { // For complete hash table do 264 bucket *b = &_bin[i]; // Handy shortcut 265 for( int j=0; j<b->_cnt; j++ ) { 266 print_key( b->_keyvals[j+j ]); 267 printf(" -> "); 268 print_value(b->_keyvals[j+j+1]); 269 printf("\n"); 270 } 271 } 272} 273 274//------------------------------Hashing Functions---------------------------- 275// Convert string to hash key. This algorithm implements a universal hash 276// function with the multipliers frozen (ok, so it's not universal). The 277// multipliers (and allowable characters) are all odd, so the resultant sum 278// is odd - guaranteed not divisible by any power of two, so the hash tables 279// can be any power of two with good results. Also, I choose multipliers 280// that have only 2 bits set (the low is always set to be odd) so 281// multiplication requires only shifts and adds. Characters are required to 282// be in the range 0-127 (I double & add 1 to force oddness). Keys are 283// limited to MAXID characters in length. Experimental evidence on 150K of 284// C text shows excellent spreading of values for any size hash table. 285int hashstr(const void *t) { 286 register char c, k = 0; 287 register int sum = 0; 288 register const char *s = (const char *)t; 289 290 while( ((c = s[k]) != '\0') && (k < MAXID-1) ) { // Get characters till nul 291 c = (c<<1)+1; // Characters are always odd! 292 sum += c + (c<<shft[k++]); // Universal hash function 293 } 294 assert( k < (MAXID + 1), "Exceeded maximum name length"); 295 return (int)((sum+xsum[k]) >> 1); // Hash key, un-modulo'd table size 296} 297 298//------------------------------hashptr-------------------------------------- 299// Slimey cheap hash function; no guaranteed performance. Better than the 300// default for pointers, especially on MS-DOS machines. 301int hashptr(const void *key) { 302#ifdef __TURBOC__ 303 return (int)((intptr_t)key >> 16); 304#else // __TURBOC__ 305 return (int)((intptr_t)key >> 2); 306#endif 307} 308 309// Slimey cheap hash function; no guaranteed performance. 310int hashkey(const void *key) { 311 return (int)((intptr_t)key); 312} 313 314//------------------------------Key Comparator Functions--------------------- 315int cmpstr(const void *k1, const void *k2) { 316 return strcmp((const char *)k1,(const char *)k2); 317} 318 319// Cheap key comparator. 320int cmpkey(const void *key1, const void *key2) { 321 if (key1 == key2) return 0; 322 intptr_t delta = (intptr_t)key1 - (intptr_t)key2; 323 if (delta > 0) return 1; 324 return -1; 325} 326 327//============================================================================= 328//------------------------------reset------------------------------------------ 329// Create an iterator and initialize the first variables. 330void DictI::reset( const Dict *dict ) { 331 _d = dict; // The dictionary 332 _i = (int)-1; // Before the first bin 333 _j = 0; // Nothing left in the current bin 334 ++(*this); // Step to first real value 335} 336 337//------------------------------next------------------------------------------- 338// Find the next key-value pair in the dictionary, or return a NULL key and 339// value. 340void DictI::operator ++(void) { 341 if( _j-- ) { // Still working in current bin? 342 _key = _d->_bin[_i]._keyvals[_j+_j]; 343 _value = _d->_bin[_i]._keyvals[_j+_j+1]; 344 return; 345 } 346 347 while( ++_i < _d->_size ) { // Else scan for non-zero bucket 348 _j = _d->_bin[_i]._cnt; 349 if( !_j ) continue; 350 _j--; 351 _key = _d->_bin[_i]._keyvals[_j+_j]; 352 _value = _d->_bin[_i]._keyvals[_j+_j+1]; 353 return; 354 } 355 _key = _value = NULL; 356} 357