| 1 | /* Copyright (C) 1995, 1996, 1997, 2000, 2001, 2003, 2004, 2006, 2008, |
| 2 | * 2009, 2010, 2011, 2012 Free Software Foundation, Inc. |
| 3 | * |
| 4 | * This library is free software; you can redistribute it and/or |
| 5 | * modify it under the terms of the GNU Lesser General Public License |
| 6 | * as published by the Free Software Foundation; either version 3 of |
| 7 | * the License, or (at your option) any later version. |
| 8 | * |
| 9 | * This library is distributed in the hope that it will be useful, but |
| 10 | * WITHOUT ANY WARRANTY; without even the implied warranty of |
| 11 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 12 | * Lesser General Public License for more details. |
| 13 | * |
| 14 | * You should have received a copy of the GNU Lesser General Public |
| 15 | * License along with this library; if not, write to the Free Software |
| 16 | * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA |
| 17 | * 02110-1301 USA |
| 18 | */ |
| 19 | |
| 20 | |
| 21 | \f |
| 22 | #ifdef HAVE_CONFIG_H |
| 23 | # include <config.h> |
| 24 | #endif |
| 25 | |
| 26 | #ifdef HAVE_WCHAR_H |
| 27 | #include <wchar.h> |
| 28 | #endif |
| 29 | |
| 30 | #include <math.h> |
| 31 | #include <unistr.h> |
| 32 | |
| 33 | #include "libguile/_scm.h" |
| 34 | #include "libguile/chars.h" |
| 35 | #include "libguile/ports.h" |
| 36 | #include "libguile/strings.h" |
| 37 | #include "libguile/symbols.h" |
| 38 | #include "libguile/vectors.h" |
| 39 | |
| 40 | #include "libguile/validate.h" |
| 41 | #include "libguile/hash.h" |
| 42 | \f |
| 43 | |
| 44 | #ifndef floor |
| 45 | extern double floor(); |
| 46 | #endif |
| 47 | |
| 48 | |
| 49 | unsigned long |
| 50 | scm_string_hash (const unsigned char *str, size_t len) |
| 51 | { |
| 52 | /* from suggestion at: */ |
| 53 | /* http://srfi.schemers.org/srfi-13/mail-archive/msg00112.html */ |
| 54 | |
| 55 | unsigned long h = 0; |
| 56 | while (len-- > 0) |
| 57 | h = *str++ + h*37; |
| 58 | return h; |
| 59 | } |
| 60 | |
| 61 | unsigned long |
| 62 | scm_i_string_hash (SCM str) |
| 63 | { |
| 64 | size_t len = scm_i_string_length (str); |
| 65 | size_t i = 0; |
| 66 | |
| 67 | unsigned long h = 0; |
| 68 | while (len-- > 0) |
| 69 | h = (unsigned long) scm_i_string_ref (str, i++) + h * 37; |
| 70 | |
| 71 | scm_remember_upto_here_1 (str); |
| 72 | return h; |
| 73 | } |
| 74 | |
| 75 | unsigned long |
| 76 | scm_i_locale_string_hash (const char *str, size_t len) |
| 77 | { |
| 78 | #ifdef HAVE_WCHAR_H |
| 79 | mbstate_t state; |
| 80 | wchar_t c; |
| 81 | size_t byte_idx = 0, nbytes; |
| 82 | unsigned long h = 0; |
| 83 | |
| 84 | if (len == (size_t) -1) |
| 85 | len = strlen (str); |
| 86 | |
| 87 | while ((nbytes = mbrtowc (&c, str + byte_idx, len - byte_idx, &state)) > 0) |
| 88 | { |
| 89 | if (nbytes >= (size_t) -2) |
| 90 | /* Invalid input string; punt. */ |
| 91 | return scm_i_string_hash (scm_from_locale_stringn (str, len)); |
| 92 | |
| 93 | h = (unsigned long) c + h * 37; |
| 94 | byte_idx += nbytes; |
| 95 | } |
| 96 | |
| 97 | return h; |
| 98 | #else |
| 99 | return scm_i_string_hash (scm_from_locale_stringn (str, len)); |
| 100 | #endif |
| 101 | } |
| 102 | |
| 103 | unsigned long |
| 104 | scm_i_latin1_string_hash (const char *str, size_t len) |
| 105 | { |
| 106 | const scm_t_uint8 *ustr = (const scm_t_uint8 *) str; |
| 107 | size_t i = 0; |
| 108 | unsigned long h = 0; |
| 109 | |
| 110 | if (len == (size_t) -1) |
| 111 | len = strlen (str); |
| 112 | |
| 113 | for (; i < len; i++) |
| 114 | h = (unsigned long) ustr[i] + h * 37; |
| 115 | |
| 116 | return h; |
| 117 | } |
| 118 | |
| 119 | unsigned long |
| 120 | scm_i_utf8_string_hash (const char *str, size_t len) |
| 121 | { |
| 122 | const scm_t_uint8 *ustr = (const scm_t_uint8 *) str; |
| 123 | size_t byte_idx = 0; |
| 124 | unsigned long h = 0; |
| 125 | |
| 126 | if (len == (size_t) -1) |
| 127 | len = strlen (str); |
| 128 | |
| 129 | while (byte_idx < len) |
| 130 | { |
| 131 | ucs4_t c; |
| 132 | int nbytes; |
| 133 | |
| 134 | nbytes = u8_mbtouc (&c, ustr + byte_idx, len - byte_idx); |
| 135 | if (nbytes == 0) |
| 136 | break; |
| 137 | else if (nbytes < 0) |
| 138 | /* Bad UTF-8; punt. */ |
| 139 | return scm_i_string_hash (scm_from_utf8_stringn (str, len)); |
| 140 | |
| 141 | h = (unsigned long) c + h * 37; |
| 142 | byte_idx += nbytes; |
| 143 | } |
| 144 | |
| 145 | return h; |
| 146 | } |
| 147 | |
| 148 | |
| 149 | /* Dirk:FIXME:: why downcase for characters? (2x: scm_hasher, scm_ihashv) */ |
| 150 | /* Dirk:FIXME:: scm_hasher could be made static. */ |
| 151 | |
| 152 | |
| 153 | unsigned long |
| 154 | scm_hasher(SCM obj, unsigned long n, size_t d) |
| 155 | { |
| 156 | switch (SCM_ITAG3 (obj)) { |
| 157 | case scm_tc3_int_1: |
| 158 | case scm_tc3_int_2: |
| 159 | return SCM_I_INUM(obj) % n; /* SCM_INUMP(obj) */ |
| 160 | case scm_tc3_imm24: |
| 161 | if (SCM_CHARP(obj)) |
| 162 | return (unsigned)(scm_c_downcase(SCM_CHAR(obj))) % n; |
| 163 | switch (SCM_UNPACK (obj)) { |
| 164 | case SCM_EOL_BITS: |
| 165 | d = 256; |
| 166 | break; |
| 167 | case SCM_BOOL_T_BITS: |
| 168 | d = 257; |
| 169 | break; |
| 170 | case SCM_BOOL_F_BITS: |
| 171 | d = 258; |
| 172 | break; |
| 173 | case SCM_EOF_VAL_BITS: |
| 174 | d = 259; |
| 175 | break; |
| 176 | default: |
| 177 | d = 263; /* perhaps should be error */ |
| 178 | } |
| 179 | return d % n; |
| 180 | default: |
| 181 | return 263 % n; /* perhaps should be error */ |
| 182 | case scm_tc3_cons: |
| 183 | switch SCM_TYP7(obj) { |
| 184 | default: |
| 185 | return 263 % n; |
| 186 | case scm_tc7_smob: |
| 187 | return 263 % n; |
| 188 | case scm_tc7_number: |
| 189 | switch SCM_TYP16 (obj) { |
| 190 | case scm_tc16_big: |
| 191 | return scm_to_ulong (scm_modulo (obj, scm_from_ulong (n))); |
| 192 | case scm_tc16_real: |
| 193 | { |
| 194 | double r = SCM_REAL_VALUE (obj); |
| 195 | if (floor (r) == r && !isinf (r) && !isnan (r)) |
| 196 | { |
| 197 | obj = scm_inexact_to_exact (obj); |
| 198 | return scm_to_ulong (scm_modulo (obj, scm_from_ulong (n))); |
| 199 | } |
| 200 | } |
| 201 | /* Fall through */ |
| 202 | case scm_tc16_complex: |
| 203 | case scm_tc16_fraction: |
| 204 | obj = scm_number_to_string (obj, scm_from_int (10)); |
| 205 | /* Fall through */ |
| 206 | } |
| 207 | /* Fall through */ |
| 208 | case scm_tc7_string: |
| 209 | { |
| 210 | unsigned long hash = |
| 211 | scm_i_string_hash (obj) % n; |
| 212 | return hash; |
| 213 | } |
| 214 | case scm_tc7_symbol: |
| 215 | return scm_i_symbol_hash (obj) % n; |
| 216 | case scm_tc7_pointer: |
| 217 | { |
| 218 | /* Pointer objects are typically used to store addresses of heap |
| 219 | objects. On most platforms, these are at least 3-byte |
| 220 | aligned (on x86_64-*-gnu, `malloc' returns 4-byte aligned |
| 221 | addresses), so get rid of the least significant bits. */ |
| 222 | scm_t_uintptr significant_bits; |
| 223 | |
| 224 | significant_bits = (scm_t_uintptr) SCM_POINTER_VALUE (obj) >> 4UL; |
| 225 | return (size_t) significant_bits % n; |
| 226 | } |
| 227 | case scm_tcs_struct: |
| 228 | return scm_i_struct_hash (obj, n, d); |
| 229 | case scm_tc7_wvect: |
| 230 | case scm_tc7_vector: |
| 231 | { |
| 232 | size_t len = SCM_SIMPLE_VECTOR_LENGTH (obj); |
| 233 | if (len > 5) |
| 234 | { |
| 235 | size_t i = d/2; |
| 236 | unsigned long h = 1; |
| 237 | while (i--) |
| 238 | { |
| 239 | SCM elt = SCM_SIMPLE_VECTOR_REF (obj, h % len); |
| 240 | h = ((h << 8) + (scm_hasher (elt, n, 2))) % n; |
| 241 | } |
| 242 | return h; |
| 243 | } |
| 244 | else |
| 245 | { |
| 246 | size_t i = len; |
| 247 | unsigned long h = (n)-1; |
| 248 | while (i--) |
| 249 | { |
| 250 | SCM elt = SCM_SIMPLE_VECTOR_REF (obj, h % len); |
| 251 | h = ((h << 8) + (scm_hasher (elt, n, d/len))) % n; |
| 252 | } |
| 253 | return h; |
| 254 | } |
| 255 | } |
| 256 | case scm_tcs_cons_imcar: |
| 257 | case scm_tcs_cons_nimcar: |
| 258 | if (d) return (scm_hasher (SCM_CAR (obj), n, d/2) |
| 259 | + scm_hasher (SCM_CDR (obj), n, d/2)) % n; |
| 260 | else return 1; |
| 261 | case scm_tc7_port: |
| 262 | return ((SCM_RDNG & SCM_CELL_WORD_0 (obj)) ? 260 : 261) % n; |
| 263 | case scm_tc7_program: |
| 264 | return 262 % n; |
| 265 | } |
| 266 | } |
| 267 | } |
| 268 | |
| 269 | |
| 270 | \f |
| 271 | |
| 272 | |
| 273 | unsigned long |
| 274 | scm_ihashq (SCM obj, unsigned long n) |
| 275 | { |
| 276 | return (SCM_UNPACK (obj) >> 1) % n; |
| 277 | } |
| 278 | |
| 279 | |
| 280 | SCM_DEFINE (scm_hashq, "hashq", 2, 0, 0, |
| 281 | (SCM key, SCM size), |
| 282 | "Determine a hash value for @var{key} that is suitable for\n" |
| 283 | "lookups in a hashtable of size @var{size}, where @code{eq?} is\n" |
| 284 | "used as the equality predicate. The function returns an\n" |
| 285 | "integer in the range 0 to @var{size} - 1. Note that\n" |
| 286 | "@code{hashq} may use internal addresses. Thus two calls to\n" |
| 287 | "hashq where the keys are @code{eq?} are not guaranteed to\n" |
| 288 | "deliver the same value if the key object gets garbage collected\n" |
| 289 | "in between. This can happen, for example with symbols:\n" |
| 290 | "@code{(hashq 'foo n) (gc) (hashq 'foo n)} may produce two\n" |
| 291 | "different values, since @code{foo} will be garbage collected.") |
| 292 | #define FUNC_NAME s_scm_hashq |
| 293 | { |
| 294 | unsigned long sz = scm_to_unsigned_integer (size, 1, ULONG_MAX); |
| 295 | return scm_from_ulong (scm_ihashq (key, sz)); |
| 296 | } |
| 297 | #undef FUNC_NAME |
| 298 | |
| 299 | |
| 300 | \f |
| 301 | |
| 302 | |
| 303 | unsigned long |
| 304 | scm_ihashv (SCM obj, unsigned long n) |
| 305 | { |
| 306 | if (SCM_CHARP(obj)) |
| 307 | return ((unsigned long) (scm_c_downcase (SCM_CHAR (obj)))) % n; /* downcase!?!! */ |
| 308 | |
| 309 | if (SCM_NUMP(obj)) |
| 310 | return (unsigned long) scm_hasher(obj, n, 10); |
| 311 | else |
| 312 | return SCM_UNPACK (obj) % n; |
| 313 | } |
| 314 | |
| 315 | |
| 316 | SCM_DEFINE (scm_hashv, "hashv", 2, 0, 0, |
| 317 | (SCM key, SCM size), |
| 318 | "Determine a hash value for @var{key} that is suitable for\n" |
| 319 | "lookups in a hashtable of size @var{size}, where @code{eqv?} is\n" |
| 320 | "used as the equality predicate. The function returns an\n" |
| 321 | "integer in the range 0 to @var{size} - 1. Note that\n" |
| 322 | "@code{(hashv key)} may use internal addresses. Thus two calls\n" |
| 323 | "to hashv where the keys are @code{eqv?} are not guaranteed to\n" |
| 324 | "deliver the same value if the key object gets garbage collected\n" |
| 325 | "in between. This can happen, for example with symbols:\n" |
| 326 | "@code{(hashv 'foo n) (gc) (hashv 'foo n)} may produce two\n" |
| 327 | "different values, since @code{foo} will be garbage collected.") |
| 328 | #define FUNC_NAME s_scm_hashv |
| 329 | { |
| 330 | unsigned long sz = scm_to_unsigned_integer (size, 1, ULONG_MAX); |
| 331 | return scm_from_ulong (scm_ihashv (key, sz)); |
| 332 | } |
| 333 | #undef FUNC_NAME |
| 334 | |
| 335 | |
| 336 | \f |
| 337 | |
| 338 | |
| 339 | unsigned long |
| 340 | scm_ihash (SCM obj, unsigned long n) |
| 341 | { |
| 342 | return (unsigned long) scm_hasher (obj, n, 10); |
| 343 | } |
| 344 | |
| 345 | SCM_DEFINE (scm_hash, "hash", 2, 0, 0, |
| 346 | (SCM key, SCM size), |
| 347 | "Determine a hash value for @var{key} that is suitable for\n" |
| 348 | "lookups in a hashtable of size @var{size}, where @code{equal?}\n" |
| 349 | "is used as the equality predicate. The function returns an\n" |
| 350 | "integer in the range 0 to @var{size} - 1.") |
| 351 | #define FUNC_NAME s_scm_hash |
| 352 | { |
| 353 | unsigned long sz = scm_to_unsigned_integer (size, 1, ULONG_MAX); |
| 354 | return scm_from_ulong (scm_ihash (key, sz)); |
| 355 | } |
| 356 | #undef FUNC_NAME |
| 357 | |
| 358 | |
| 359 | \f |
| 360 | |
| 361 | |
| 362 | void |
| 363 | scm_init_hash () |
| 364 | { |
| 365 | #include "libguile/hash.x" |
| 366 | } |
| 367 | |
| 368 | |
| 369 | /* |
| 370 | Local Variables: |
| 371 | c-file-style: "gnu" |
| 372 | End: |
| 373 | */ |