1 /*
   2  * Copyright (c) 1997, 2017, 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 #include "precompiled.hpp"
  26 #include "libadt/dict.hpp"
  27 
  28 // Dictionaries - An Abstract Data Type
  29 
  30 // %%%%% includes not needed with AVM framework - Ungar
  31 
  32 #include <assert.h>
  33 
  34 //------------------------------data-----------------------------------------
  35 // String hash tables
  36 #define MAXID 20
  37 static uint8_t initflag = 0;       // True after 1st initialization
  38 static const char shft[MAXID] = {1,2,3,4,5,6,7,1,2,3,4,5,6,7,1,2,3,4,5,6};
  39 static short xsum[MAXID];
  40 
  41 //------------------------------bucket---------------------------------------
  42 class bucket : public ResourceObj {
  43 public:
  44   uint _cnt, _max;              // Size of bucket
  45   void **_keyvals;              // Array of keys and values
  46 };
  47 
  48 //------------------------------Dict-----------------------------------------
  49 // The dictionary is kept has a hash table.  The hash table is a even power
  50 // of two, for nice modulo operations.  Each bucket in the hash table points
  51 // to a linear list of key-value pairs; each key & value is just a (void *).
  52 // The list starts with a count.  A hash lookup finds the list head, then a
  53 // simple linear scan finds the key.  If the table gets too full, it's
  54 // doubled in size; the total amount of EXTRA times all hash functions are
  55 // computed for the doubling is no more than the current size - thus the
  56 // doubling in size costs no more than a constant factor in speed.
  57 Dict::Dict(CmpKey initcmp, Hash inithash) : _hash(inithash), _cmp(initcmp),
  58   _arena(Thread::current()->resource_area()) {
  59   int i;
  60 
  61   // Precompute table of null character hashes
  62   if( !initflag ) {             // Not initializated yet?
  63     xsum[0] = (1<<shft[0])+1;   // Initialize
  64     for(i=1; i<MAXID; i++) {
  65       xsum[i] = (1<<shft[i])+1+xsum[i-1];
  66     }
  67     initflag = 1;               // Never again
  68   }
  69 
  70   _size = 16;                   // Size is a power of 2
  71   _cnt = 0;                     // Dictionary is empty
  72   _bin = (bucket*)_arena->Amalloc_4(sizeof(bucket)*_size);
  73   memset((void*)_bin,0,sizeof(bucket)*_size);
  74 }
  75 
  76 Dict::Dict(CmpKey initcmp, Hash inithash, Arena *arena, int size)
  77 : _hash(inithash), _cmp(initcmp), _arena(arena) {
  78   int i;
  79 
  80   // Precompute table of null character hashes
  81   if( !initflag ) {             // Not initializated yet?
  82     xsum[0] = (1<<shft[0])+1;   // Initialize
  83     for(i=1; i<MAXID; i++) {
  84       xsum[i] = (1<<shft[i])+1+xsum[i-1];
  85     }
  86     initflag = 1;               // Never again
  87   }
  88 
  89   i=16;
  90   while( i < size ) i <<= 1;
  91   _size = i;                    // Size is a power of 2
  92   _cnt = 0;                     // Dictionary is empty
  93   _bin = (bucket*)_arena->Amalloc_4(sizeof(bucket)*_size);
  94   memset((void*)_bin,0,sizeof(bucket)*_size);
  95 }
  96 
  97 //------------------------------~Dict------------------------------------------
  98 // Delete an existing dictionary.
  99 Dict::~Dict() {
 100   /*
 101   tty->print("~Dict %d/%d: ",_cnt,_size);
 102   for( uint i=0; i < _size; i++) // For complete new table do
 103     tty->print("%d ",_bin[i]._cnt);
 104   tty->print("\n");*/
 105   /*for( uint i=0; i<_size; i++ ) {
 106     FREE_FAST( _bin[i]._keyvals );
 107     } */
 108 }
 109 
 110 //------------------------------Clear----------------------------------------
 111 // Zap to empty; ready for re-use
 112 void Dict::Clear() {
 113   _cnt = 0;                     // Empty contents
 114   for( uint i=0; i<_size; i++ )
 115     _bin[i]._cnt = 0;           // Empty buckets, but leave allocated
 116   // Leave _size & _bin alone, under the assumption that dictionary will
 117   // grow to this size again.
 118 }
 119 
 120 //------------------------------doubhash---------------------------------------
 121 // Double hash table size.  If can't do so, just suffer.  If can, then run
 122 // thru old hash table, moving things to new table.  Note that since hash
 123 // table doubled, exactly 1 new bit is exposed in the mask - so everything
 124 // in the old table ends up on 1 of two lists in the new table; a hi and a
 125 // lo list depending on the value of the bit.
 126 void Dict::doubhash(void) {
 127   uint oldsize = _size;
 128   _size <<= 1;                  // Double in size
 129   _bin = (bucket*)_arena->Arealloc(_bin, sizeof(bucket) * oldsize, sizeof(bucket) * _size);
 130   memset((void*)(&_bin[oldsize]), 0, oldsize * sizeof(bucket));
 131   // Rehash things to spread into new table
 132   for (uint i = 0; i < oldsize; i++) { // For complete OLD table do
 133     bucket *b = &_bin[i];              // Handy shortcut for _bin[i]
 134     if (!b->_keyvals) continue;        // Skip empties fast
 135 
 136     bucket *nb = &_bin[i+oldsize];     // New bucket shortcut
 137     uint j = b->_max;                  // Trim new bucket to nearest power of 2
 138     while (j > b->_cnt) { j >>= 1; }   // above old bucket _cnt
 139     if (!j) { j = 1; }                 // Handle zero-sized buckets
 140     nb->_max = j << 1;
 141     // Allocate worst case space for key-value pairs
 142     nb->_keyvals = (void**)_arena->Amalloc_4(sizeof(void *) * nb->_max * 2);
 143     uint nbcnt = 0;
 144 
 145     for (j = 0; j < b->_cnt; ) {           // Rehash all keys in this bucket
 146       void *key = b->_keyvals[j + j];
 147       if ((_hash(key) & (_size-1)) != i) { // Moving to hi bucket?
 148         nb->_keyvals[nbcnt + nbcnt] = key;
 149         nb->_keyvals[nbcnt + nbcnt + 1] = b->_keyvals[j + j + 1];
 150         nb->_cnt = nbcnt = nbcnt + 1;
 151         b->_cnt--;                         // Remove key/value from lo bucket
 152         b->_keyvals[j + j] = b->_keyvals[b->_cnt + b->_cnt];
 153         b->_keyvals[j + j + 1] = b->_keyvals[b->_cnt + b->_cnt + 1];
 154         // Don't increment j, hash compacted element also.
 155       } else {
 156         j++; // Iterate.
 157       }
 158     } // End of for all key-value pairs in bucket
 159   } // End of for all buckets
 160 }
 161 
 162 //------------------------------Dict-----------------------------------------
 163 // Deep copy a dictionary.
 164 Dict::Dict( const Dict &d ) : _size(d._size), _cnt(d._cnt), _hash(d._hash),_cmp(d._cmp), _arena(d._arena) {
 165   _bin = (bucket*)_arena->Amalloc_4(sizeof(bucket)*_size);
 166   memcpy( (void*)_bin, (void*)d._bin, sizeof(bucket)*_size );
 167   for( uint i=0; i<_size; i++ ) {
 168     if( !_bin[i]._keyvals ) continue;
 169     _bin[i]._keyvals=(void**)_arena->Amalloc_4( sizeof(void *)*_bin[i]._max*2);
 170     memcpy( _bin[i]._keyvals, d._bin[i]._keyvals,_bin[i]._cnt*2*sizeof(void*));
 171   }
 172 }
 173 
 174 //------------------------------Dict-----------------------------------------
 175 // Deep copy a dictionary.
 176 Dict &Dict::operator =( const Dict &d ) {
 177   if( _size < d._size ) {       // If must have more buckets
 178     _arena = d._arena;
 179     _bin = (bucket*)_arena->Arealloc( _bin, sizeof(bucket)*_size, sizeof(bucket)*d._size );
 180     memset( (void*)(&_bin[_size]), 0, (d._size-_size)*sizeof(bucket) );
 181     _size = d._size;
 182   }
 183   uint i;
 184   for( i=0; i<_size; i++ ) // All buckets are empty
 185     _bin[i]._cnt = 0;           // But leave bucket allocations alone
 186   _cnt = d._cnt;
 187   *(Hash*)(&_hash) = d._hash;
 188   *(CmpKey*)(&_cmp) = d._cmp;
 189   for( i=0; i<_size; i++ ) {
 190     bucket *b = &d._bin[i];     // Shortcut to source bucket
 191     for( uint j=0; j<b->_cnt; j++ )
 192       Insert( b->_keyvals[j+j], b->_keyvals[j+j+1] );
 193   }
 194   return *this;
 195 }
 196 
 197 //------------------------------Insert----------------------------------------
 198 // Insert or replace a key/value pair in the given dictionary.  If the
 199 // dictionary is too full, it's size is doubled.  The prior value being
 200 // replaced is returned (NULL if this is a 1st insertion of that key).  If
 201 // an old value is found, it's swapped with the prior key-value pair on the
 202 // list.  This moves a commonly searched-for value towards the list head.
 203 void *Dict::Insert(void *key, void *val, bool replace) {
 204   uint hash = _hash( key );     // Get hash key
 205   uint i = hash & (_size-1);    // Get hash key, corrected for size
 206   bucket *b = &_bin[i];         // Handy shortcut
 207   for( uint j=0; j<b->_cnt; j++ ) {
 208     if( !_cmp(key,b->_keyvals[j+j]) ) {
 209       if (!replace) {
 210         return b->_keyvals[j+j+1];
 211       } else {
 212         void *prior = b->_keyvals[j+j+1];
 213         b->_keyvals[j+j  ] = key;       // Insert current key-value
 214         b->_keyvals[j+j+1] = val;
 215         return prior;           // Return prior
 216       }
 217     }
 218   }
 219   if( ++_cnt > _size ) {        // Hash table is full
 220     doubhash();                 // Grow whole table if too full
 221     i = hash & (_size-1);       // Rehash
 222     b = &_bin[i];               // Handy shortcut
 223   }
 224   if( b->_cnt == b->_max ) {    // Must grow bucket?
 225     if( !b->_keyvals ) {
 226       b->_max = 2;              // Initial bucket size
 227       b->_keyvals = (void**)_arena->Amalloc_4(sizeof(void*) * b->_max * 2);
 228     } else {
 229       b->_keyvals = (void**)_arena->Arealloc(b->_keyvals, sizeof(void*) * b->_max * 2, sizeof(void*) * b->_max * 4);
 230       b->_max <<= 1;            // Double bucket
 231     }
 232   }
 233   b->_keyvals[b->_cnt+b->_cnt  ] = key;
 234   b->_keyvals[b->_cnt+b->_cnt+1] = val;
 235   b->_cnt++;
 236   return NULL;                  // Nothing found prior
 237 }
 238 
 239 //------------------------------Delete---------------------------------------
 240 // Find & remove a value from dictionary. Return old value.
 241 void *Dict::Delete(void *key) {
 242   uint i = _hash( key ) & (_size-1);    // Get hash key, corrected for size
 243   bucket *b = &_bin[i];         // Handy shortcut
 244   for( uint j=0; j<b->_cnt; j++ )
 245     if( !_cmp(key,b->_keyvals[j+j]) ) {
 246       void *prior = b->_keyvals[j+j+1];
 247       b->_cnt--;                // Remove key/value from lo bucket
 248       b->_keyvals[j+j  ] = b->_keyvals[b->_cnt+b->_cnt  ];
 249       b->_keyvals[j+j+1] = b->_keyvals[b->_cnt+b->_cnt+1];
 250       _cnt--;                   // One less thing in table
 251       return prior;
 252     }
 253   return NULL;
 254 }
 255 
 256 //------------------------------FindDict-------------------------------------
 257 // Find a key-value pair in the given dictionary.  If not found, return NULL.
 258 // If found, move key-value pair towards head of list.
 259 void *Dict::operator [](const void *key) const {
 260   uint i = _hash( key ) & (_size-1);    // Get hash key, corrected for size
 261   bucket *b = &_bin[i];         // Handy shortcut
 262   for( uint j=0; j<b->_cnt; j++ )
 263     if( !_cmp(key,b->_keyvals[j+j]) )
 264       return b->_keyvals[j+j+1];
 265   return NULL;
 266 }
 267 
 268 //------------------------------CmpDict--------------------------------------
 269 // CmpDict compares two dictionaries; they must have the same keys (their
 270 // keys must match using CmpKey) and they must have the same values (pointer
 271 // comparison).  If so 1 is returned, if not 0 is returned.
 272 int32_t Dict::operator ==(const Dict &d2) const {
 273   if( _cnt != d2._cnt ) return 0;
 274   if( _hash != d2._hash ) return 0;
 275   if( _cmp != d2._cmp ) return 0;
 276   for( uint i=0; i < _size; i++) {      // For complete hash table do
 277     bucket *b = &_bin[i];       // Handy shortcut
 278     if( b->_cnt != d2._bin[i]._cnt ) return 0;
 279     if( memcmp(b->_keyvals, d2._bin[i]._keyvals, b->_cnt*2*sizeof(void*) ) )
 280       return 0;                 // Key-value pairs must match
 281   }
 282   return 1;                     // All match, is OK
 283 }
 284 
 285 //------------------------------print------------------------------------------
 286 // Handier print routine
 287 void Dict::print() {
 288   DictI i(this); // Moved definition in iterator here because of g++.
 289   tty->print("Dict@" INTPTR_FORMAT "[%d] = {", p2i(this), _cnt);
 290   for( ; i.test(); ++i ) {
 291     tty->print("(" INTPTR_FORMAT "," INTPTR_FORMAT "),", p2i(i._key), p2i(i._value));
 292   }
 293   tty->print_cr("}");
 294 }
 295 
 296 //------------------------------Hashing Functions----------------------------
 297 // Convert string to hash key.  This algorithm implements a universal hash
 298 // function with the multipliers frozen (ok, so it's not universal).  The
 299 // multipliers (and allowable characters) are all odd, so the resultant sum
 300 // is odd - guaranteed not divisible by any power of two, so the hash tables
 301 // can be any power of two with good results.  Also, I choose multipliers
 302 // that have only 2 bits set (the low is always set to be odd) so
 303 // multiplication requires only shifts and adds.  Characters are required to
 304 // be in the range 0-127 (I double & add 1 to force oddness).  Keys are
 305 // limited to MAXID characters in length.  Experimental evidence on 150K of
 306 // C text shows excellent spreading of values for any size hash table.
 307 int hashstr(const void *t) {
 308   char c, k = 0;
 309   int32_t sum = 0;
 310   const char *s = (const char *)t;
 311 
 312   while( ((c = *s++) != '\0') && (k < MAXID-1) ) { // Get characters till null or MAXID-1
 313     c = (c<<1)+1;               // Characters are always odd!
 314     sum += c + (c<<shft[k++]);  // Universal hash function
 315   }
 316   return (int)((sum+xsum[k]) >> 1); // Hash key, un-modulo'd table size
 317 }
 318 
 319 //------------------------------hashptr--------------------------------------
 320 // Slimey cheap hash function; no guaranteed performance.  Better than the
 321 // default for pointers, especially on MS-DOS machines.
 322 int hashptr(const void *key) {
 323   return ((intptr_t)key >> 2);
 324 }
 325 
 326 // Slimey cheap hash function; no guaranteed performance.
 327 int hashkey(const void *key) {
 328   return (intptr_t)key;
 329 }
 330 
 331 //------------------------------Key Comparator Functions---------------------
 332 int32_t cmpstr(const void *k1, const void *k2) {
 333   return strcmp((const char *)k1,(const char *)k2);
 334 }
 335 
 336 // Cheap key comparator.
 337 int32_t cmpkey(const void *key1, const void *key2) {
 338   if (key1 == key2) return 0;
 339   intptr_t delta = (intptr_t)key1 - (intptr_t)key2;
 340   if (delta > 0) return 1;
 341   return -1;
 342 }
 343 
 344 //=============================================================================
 345 //------------------------------reset------------------------------------------
 346 // Create an iterator and initialize the first variables.
 347 void DictI::reset( const Dict *dict ) {
 348   _d = dict;                    // The dictionary
 349   _i = (uint)-1;                // Before the first bin
 350   _j = 0;                       // Nothing left in the current bin
 351   ++(*this);                    // Step to first real value
 352 }
 353 
 354 //------------------------------next-------------------------------------------
 355 // Find the next key-value pair in the dictionary, or return a NULL key and
 356 // value.
 357 void DictI::operator ++(void) {
 358   if( _j-- ) {                  // Still working in current bin?
 359     _key   = _d->_bin[_i]._keyvals[_j+_j];
 360     _value = _d->_bin[_i]._keyvals[_j+_j+1];
 361     return;
 362   }
 363 
 364   while( ++_i < _d->_size ) {   // Else scan for non-zero bucket
 365     _j = _d->_bin[_i]._cnt;
 366     if( !_j ) continue;
 367     _j--;
 368     _key   = _d->_bin[_i]._keyvals[_j+_j];
 369     _value = _d->_bin[_i]._keyvals[_j+_j+1];
 370     return;
 371   }
 372   _key = _value = NULL;
 373 }