2 Copyright (C) 2001-2006, William Joseph.
5 This file is part of GtkRadiant.
7 GtkRadiant is free software; you can redistribute it and/or modify
8 it under the terms of the GNU General Public License as published by
9 the Free Software Foundation; either version 2 of the License, or
10 (at your option) any later version.
12 GtkRadiant is distributed in the hope that it will be useful,
13 but WITHOUT ANY WARRANTY; without even the implied warranty of
14 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 GNU General Public License for more details.
17 You should have received a copy of the GNU General Public License
18 along with GtkRadiant; if not, write to the Free Software
19 Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA
22 #if !defined( INCLUDED_CONTAINER_HASHFUNC_H )
23 #define INCLUDED_CONTAINER_HASHFUNC_H
26 #include "string/string.h"
27 #include "container/array.h"
28 typedef unsigned long int ub4; /* unsigned 4-byte quantities */
29 typedef unsigned char ub1;
31 inline ub1 ub1_as_ub1_nocase( ub1 byte ){
32 return std::tolower( byte );
35 inline ub4 ub1x4_as_ub4_nocase( const ub1 bytes[4] ){
37 reinterpret_cast<ub1*>( &result )[0] = ub1_as_ub1_nocase( bytes[0] );
38 reinterpret_cast<ub1*>( &result )[1] = ub1_as_ub1_nocase( bytes[1] );
39 reinterpret_cast<ub1*>( &result )[2] = ub1_as_ub1_nocase( bytes[2] );
40 reinterpret_cast<ub1*>( &result )[3] = ub1_as_ub1_nocase( bytes[3] );
44 class ub1_default_traits
47 static ub1 as_ub1( ub1 byte ){
52 class ub1_nocase_traits
55 static ub1 as_ub1( ub1 byte ){
56 return ub1_as_ub1_nocase( byte );
60 class ub1x4_default_traits
63 static ub4 as_ub4( const ub1 bytes[4] ){
64 return *reinterpret_cast<const ub4*>( bytes );
68 class ub1x4_nocase_traits
71 static ub4 as_ub4( const ub1 bytes[4] ){
72 return ub1x4_as_ub4_nocase( bytes );
76 class ub4_default_traits
79 static ub4 as_ub4( ub4 i ){
84 class ub4_nocase_traits
87 static ub4 as_ub4( ub4 i ){
88 return ub1x4_as_ub4_nocase( reinterpret_cast<const ub1*>( &i ) );
93 // By Bob Jenkins, 1996. bob_jenkins@burtleburtle.net. You may use this
94 // code any way you wish, private, educational, or commercial. It's free.
96 #define hashsize( n ) ( (ub4)1 << ( n ) )
97 #define hashmask( n ) ( hashsize( n ) - 1 )
100 --------------------------------------------------------------------
101 mix -- mix 3 32-bit values reversibly.
102 For every delta with one or two bit set, and the deltas of all three
103 high bits or all three low bits, whether the original value of a,b,c
104 is almost all zero or is uniformly distributed,
105 * If mix() is run forward or backward, at least 32 bits in a,b,c
106 have at least 1/4 probability of changing.
107 * If mix() is run forward, every bit of c will change between 1/3 and
108 2/3 of the time. (Well, 22/100 and 78/100 for some 2-bit deltas.)
109 mix() was built out of 36 single-cycle latency instructions in a
110 structure that could supported 2x parallelism, like so:
118 Unfortunately, superscalar Pentiums and Sparcs can't take advantage
119 of that parallelism. They've also turned some of those single-cycle
120 latency instructions into multi-cycle latency instructions. Still,
121 this is the fastest good hash I could find. There were about 2^^68
122 to choose from. I only looked at a billion or so.
123 --------------------------------------------------------------------
125 #define mix( a,b,c ) \
127 a -= b; a -= c; a ^= ( c >> 13 ); \
128 b -= c; b -= a; b ^= ( a << 8 ); \
129 c -= a; c -= b; c ^= ( b >> 13 ); \
130 a -= b; a -= c; a ^= ( c >> 12 ); \
131 b -= c; b -= a; b ^= ( a << 16 ); \
132 c -= a; c -= b; c ^= ( b >> 5 ); \
133 a -= b; a -= c; a ^= ( c >> 3 ); \
134 b -= c; b -= a; b ^= ( a << 10 ); \
135 c -= a; c -= b; c ^= ( b >> 15 ); \
138 /* same, but slower, works on systems that might have 8 byte ub4's */
139 #define mix2( a,b,c ) \
141 a -= b; a -= c; a ^= ( c >> 13 ); \
142 b -= c; b -= a; b ^= ( a << 8 ); \
143 c -= a; c -= b; c ^= ( ( b & 0xffffffff ) >> 13 ); \
144 a -= b; a -= c; a ^= ( ( c & 0xffffffff ) >> 12 ); \
145 b -= c; b -= a; b = ( b ^ ( a << 16 ) ) & 0xffffffff; \
146 c -= a; c -= b; c = ( c ^ ( b >> 5 ) ) & 0xffffffff; \
147 a -= b; a -= c; a = ( a ^ ( c >> 3 ) ) & 0xffffffff; \
148 b -= c; b -= a; b = ( b ^ ( a << 10 ) ) & 0xffffffff; \
149 c -= a; c -= b; c = ( c ^ ( b >> 15 ) ) & 0xffffffff; \
153 --------------------------------------------------------------------
154 hash() -- hash a variable-length key into a 32-bit value
155 k : the key (the unaligned variable-length array of bytes)
156 len : the length of the key, counting by bytes
157 level : can be any 4-byte value
158 Returns a 32-bit value. Every bit of the key affects every bit of
159 the return value. Every 1-bit and 2-bit delta achieves avalanche.
160 About 36+6len instructions.
162 The best hash table sizes are powers of 2. There is no need to do
163 mod a prime (mod is sooo slow!). If you need less than 32 bits,
164 use a bitmask. For example, if you need only 10 bits, do
165 h = (h & hashmask(10));
166 In which case, the hash table should have hashsize(10) elements.
168 If you are hashing n strings (ub1 **)k, do it like this:
169 for (i=0, h=0; i<n; ++i) h = hash( k[i], len[i], h);
171 See http://burlteburtle.net/bob/hash/evahash.html
172 Use for hash table lookup, or anything where one collision in 2^32 is
173 acceptable. Do NOT use for cryptographic purposes.
174 --------------------------------------------------------------------
177 template<typename UB1Traits, typename UB4x1Traits>
179 const ub1 *k, /* the key */
180 ub4 length, /* the length of the key */
181 ub4 initval, /* the previous hash, or an arbitrary value */
182 const UB1Traits& ub1traits,
183 const UB4x1Traits& ub4x1traits
185 register ub4 a,b,c,len;
187 /* Set up the internal state */
189 a = b = 0x9e3779b9; /* the golden ratio; an arbitrary value */
190 c = initval; /* the previous hash value */
192 /*---------------------------------------- handle most of the key */
195 a += ( k[0] + ( ( ub4 ) UB1Traits::as_ub1( k[1] ) << 8 ) + ( ( ub4 ) UB1Traits::as_ub1( k[2] ) << 16 ) + ( ( ub4 ) UB1Traits::as_ub1( k[3] ) << 24 ) );
196 b += ( k[4] + ( ( ub4 ) UB1Traits::as_ub1( k[5] ) << 8 ) + ( ( ub4 ) UB1Traits::as_ub1( k[6] ) << 16 ) + ( ( ub4 ) UB1Traits::as_ub1( k[7] ) << 24 ) );
197 c += ( k[8] + ( ( ub4 ) UB1Traits::as_ub1( k[9] ) << 8 ) + ( ( ub4 ) UB1Traits::as_ub1( k[10] ) << 16 ) + ( ( ub4 ) UB1Traits::as_ub1( k[11] ) << 24 ) );
202 /*------------------------------------- handle the last 11 bytes */
204 switch ( len ) /* all the case statements fall through */
206 case 11: c += ( ( ub4 ) UB1Traits::as_ub1( k[10] ) << 24 );
207 case 10: c += ( ( ub4 ) UB1Traits::as_ub1( k[9] ) << 16 );
208 case 9: c += ( ( ub4 ) UB1Traits::as_ub1( k[8] ) << 8 );
209 /* the first byte of c is reserved for the length */
210 case 8: b += ( ( ub4 ) UB1Traits::as_ub1( k[7] ) << 24 );
211 case 7: b += ( ( ub4 ) UB1Traits::as_ub1( k[6] ) << 16 );
212 case 6: b += ( ( ub4 ) UB1Traits::as_ub1( k[5] ) << 8 );
213 case 5: b += UB1Traits::as_ub1( k[4] );
214 case 4: a += ( ( ub4 ) UB1Traits::as_ub1( k[3] ) << 24 );
215 case 3: a += ( ( ub4 ) UB1Traits::as_ub1( k[2] ) << 16 );
216 case 2: a += ( ( ub4 ) UB1Traits::as_ub1( k[1] ) << 8 );
217 case 1: a += UB1Traits::as_ub1( k[0] );
218 /* case 0: nothing left to add */
221 /*-------------------------------------------- report the result */
226 --------------------------------------------------------------------
227 This works on all machines. hash2() is identical to hash() on
228 little-endian machines, except that the length has to be measured
229 in ub4s instead of bytes. It is much faster than hash(). It
231 -- that the key be an array of ub4's, and
232 -- that all your machines have the same endianness, and
233 -- that the length be the number of ub4's in the key
234 --------------------------------------------------------------------
236 template<typename UB4Traits>
238 const ub4 *k, /* the key */
239 ub4 length, /* the length of the key, in ub4s */
240 ub4 initval, /* the previous hash, or an arbitrary value */
241 const UB4Traits& ub4traits
243 register ub4 a,b,c,len;
245 /* Set up the internal state */
247 a = b = 0x9e3779b9; /* the golden ratio; an arbitrary value */
248 c = initval; /* the previous hash value */
250 /*---------------------------------------- handle most of the key */
253 a += UB4Traits::as_ub4( k[0] );
254 b += UB4Traits::as_ub4( k[1] );
255 c += UB4Traits::as_ub4( k[2] );
260 /*-------------------------------------- handle the last 2 ub4's */
262 switch ( len ) /* all the case statements fall through */
264 /* c is reserved for the length */
265 case 2: b += UB4Traits::as_ub4( k[1] );
266 case 1: a += UB4Traits::as_ub4( k[0] );
267 /* case 0: nothing left to add */
270 /*-------------------------------------------- report the result */
276 inline hash_t hash_ub1( const ub1* key, std::size_t len, hash_t previous = 0 ){
277 return hash( key, ub4( len ), previous, ub1_default_traits(), ub1x4_default_traits() );
280 inline hash_t hash_ub1_nocase( const ub1* key, std::size_t len, hash_t previous = 0 ){
281 return hash( key, ub4( len ), previous, ub1_nocase_traits(), ub1x4_nocase_traits() );
284 template<typename UB4Traits>
285 inline hash_t hash_ub4( const ub4* key, std::size_t len, const UB4Traits& traits, hash_t previous = 0 ){
286 return hash2( key,ub4( len ), previous, traits );
289 inline ub4 hash_combine( ub4 left, ub4 right ){
290 return hash_ub1( reinterpret_cast<const ub1*>( &left ), 4, right );
293 template<typename POD>
294 inline hash_t pod_hash( const POD& pod ){
295 return hash_ub1( reinterpret_cast<const ub1*>( &pod ), sizeof( POD ) );
298 inline hash_t string_hash( const char* string, hash_t previous = 0 ){
299 return hash_ub1( reinterpret_cast<const ub1*>( string ), string_length( string ), previous );
302 inline hash_t string_hash_nocase( const char* string, hash_t previous = 0 ){
303 return hash_ub1_nocase( reinterpret_cast<const ub1*>( string ), string_length( string ), previous );
308 typedef hash_t hash_type;
309 hash_type operator()( const char* string ) const {
310 return string_hash( string );
316 typedef hash_t hash_type;
317 hash_type operator()( const CopiedString& string ) const {
318 return string_hash( string.c_str() );
322 struct HashStringNoCase
324 typedef hash_t hash_type;
325 hash_type operator()( const CopiedString& string ) const {
326 return string_hash_nocase( string.c_str() );
330 /// \brief Length of a string in ub4.
331 /// "wibble" (6) gives 2,
332 /// "and" (3) gives 1,
333 /// "bleh" (4) gives 2
334 inline std::size_t string_length_ub4( const char* string ){
335 return ( ( string_length( string ) >> 2 ) + 1 ) << 2;
338 /// \brief Hashable key type that stores a string as an array of ub4 - making hashing faster.
339 /// Also caches the 32-bit result of the hash to speed up comparison of keys.
340 template<typename UB4Traits = ub4_default_traits>
346 void copy( const HashKey& other ){
347 std::copy( other.m_key.begin(), other.m_key.end(), m_key.begin() );
348 m_hash = other.m_hash;
350 void copy( const char* string ){
351 strncpy( reinterpret_cast<char*>( m_key.data() ), string, m_key.size() );
352 for ( Array<ub4>::iterator i = m_key.begin(); i != m_key.end(); ++i )
354 *i = UB4Traits::as_ub4( *i );
356 m_hash = hash_ub4( m_key.data(), m_key.size(), ub4_default_traits() );
358 bool equal( const HashKey& other ) const {
359 return m_hash == other.m_hash && m_key.size() == other.m_key.size()
360 && std::equal( m_key.begin(), m_key.end(), other.m_key.begin() );
364 HashKey( const HashKey& other ) : m_key( other.m_key.size() ){
367 HashKey( const char* string ) : m_key( string_length_ub4( string ) ){
370 HashKey& operator=( const char* string ){
371 m_key.resize( string_length_ub4( string ) );
375 bool operator==( const HashKey& other ) const {
376 return equal( other );
378 bool operator!=( const HashKey& other ) const {
379 return !equal( other );
381 hash_t hash() const {
385 const char* c_str() const {
386 return reinterpret_cast<const char*>( m_key.data() );
391 /// \brief Hash function to use with HashKey.
394 typedef hash_t hash_type;
395 hash_type operator()( const HashKey<ub4_default_traits>& key ) const {