| 1 | /* hv.h |
| 2 | * |
| 3 | * Copyright (C) 1991, 1992, 1993, 1996, 1997, 1998, 1999, |
| 4 | * 2000, 2001, 2002, 2005, by Larry Wall and others |
| 5 | * |
| 6 | * You may distribute under the terms of either the GNU General Public |
| 7 | * License or the Artistic License, as specified in the README file. |
| 8 | * |
| 9 | */ |
| 10 | |
| 11 | /* typedefs to eliminate some typing */ |
| 12 | typedef struct he HE; |
| 13 | typedef struct hek HEK; |
| 14 | |
| 15 | /* entry in hash value chain */ |
| 16 | struct he { |
| 17 | HE *hent_next; /* next entry in chain */ |
| 18 | HEK *hent_hek; /* hash key */ |
| 19 | SV *hent_val; /* scalar value that was hashed */ |
| 20 | }; |
| 21 | |
| 22 | /* hash key -- defined separately for use as shared pointer */ |
| 23 | struct hek { |
| 24 | U32 hek_hash; /* hash of key */ |
| 25 | I32 hek_len; /* length of hash key */ |
| 26 | char hek_key[1]; /* variable-length hash key */ |
| 27 | /* the hash-key is \0-terminated */ |
| 28 | /* after the \0 there is a byte for flags, such as whether the key |
| 29 | is UTF-8 */ |
| 30 | }; |
| 31 | |
| 32 | /* hash structure: */ |
| 33 | /* This structure must match the beginning of struct xpvmg in sv.h. */ |
| 34 | struct xpvhv { |
| 35 | char * xhv_array; /* pointer to malloced string */ |
| 36 | STRLEN xhv_fill; /* how full xhv_array currently is */ |
| 37 | STRLEN xhv_max; /* subscript of last element of xhv_array */ |
| 38 | IV xhv_keys; /* how many elements in the array */ |
| 39 | NV xnv_nv; /* numeric value, if any */ |
| 40 | #define xhv_placeholders xnv_nv |
| 41 | MAGIC* xmg_magic; /* magic for scalar array */ |
| 42 | HV* xmg_stash; /* class package */ |
| 43 | |
| 44 | I32 xhv_riter; /* current root of iterator */ |
| 45 | HE *xhv_eiter; /* current entry of iterator */ |
| 46 | PMOP *xhv_pmroot; /* list of pm's for this package */ |
| 47 | char *xhv_name; /* name, if a symbol table */ |
| 48 | }; |
| 49 | |
| 50 | /* hash a key */ |
| 51 | /* FYI: This is the "One-at-a-Time" algorithm by Bob Jenkins |
| 52 | * from requirements by Colin Plumb. |
| 53 | * (http://burtleburtle.net/bob/hash/doobs.html) */ |
| 54 | /* The use of a temporary pointer and the casting games |
| 55 | * is needed to serve the dual purposes of |
| 56 | * (a) the hashed data being interpreted as "unsigned char" (new since 5.8, |
| 57 | * a "char" can be either signed or unsigned, depending on the compiler) |
| 58 | * (b) catering for old code that uses a "char" |
| 59 | * |
| 60 | * The "hash seed" feature was added in Perl 5.8.1 to perturb the results |
| 61 | * to avoid "algorithmic complexity attacks". |
| 62 | * |
| 63 | * If USE_HASH_SEED is defined, hash randomisation is done by default |
| 64 | * If USE_HASH_SEED_EXPLICIT is defined, hash randomisation is done |
| 65 | * only if the environment variable PERL_HASH_SEED is set. |
| 66 | * For maximal control, one can define PERL_HASH_SEED. |
| 67 | * (see also perl.c:perl_parse()). |
| 68 | */ |
| 69 | #ifndef PERL_HASH_SEED |
| 70 | # if defined(USE_HASH_SEED) || defined(USE_HASH_SEED_EXPLICIT) |
| 71 | # define PERL_HASH_SEED PL_hash_seed |
| 72 | # else |
| 73 | # define PERL_HASH_SEED 0 |
| 74 | # endif |
| 75 | #endif |
| 76 | #define PERL_HASH(hash,str,len) \ |
| 77 | STMT_START { \ |
| 78 | register const char *s_PeRlHaSh_tmp = str; \ |
| 79 | register const unsigned char *s_PeRlHaSh = (const unsigned char *)s_PeRlHaSh_tmp; \ |
| 80 | register I32 i_PeRlHaSh = len; \ |
| 81 | register U32 hash_PeRlHaSh = PERL_HASH_SEED; \ |
| 82 | while (i_PeRlHaSh--) { \ |
| 83 | hash_PeRlHaSh += *s_PeRlHaSh++; \ |
| 84 | hash_PeRlHaSh += (hash_PeRlHaSh << 10); \ |
| 85 | hash_PeRlHaSh ^= (hash_PeRlHaSh >> 6); \ |
| 86 | } \ |
| 87 | hash_PeRlHaSh += (hash_PeRlHaSh << 3); \ |
| 88 | hash_PeRlHaSh ^= (hash_PeRlHaSh >> 11); \ |
| 89 | (hash) = (hash_PeRlHaSh + (hash_PeRlHaSh << 15)); \ |
| 90 | } STMT_END |
| 91 | |
| 92 | /* Only hv.c and mod_perl should be doing this. */ |
| 93 | #ifdef PERL_HASH_INTERNAL_ACCESS |
| 94 | #define PERL_HASH_INTERNAL(hash,str,len) \ |
| 95 | STMT_START { \ |
| 96 | register const char *s_PeRlHaSh_tmp = str; \ |
| 97 | register const unsigned char *s_PeRlHaSh = (const unsigned char *)s_PeRlHaSh_tmp; \ |
| 98 | register I32 i_PeRlHaSh = len; \ |
| 99 | register U32 hash_PeRlHaSh = PL_rehash_seed; \ |
| 100 | while (i_PeRlHaSh--) { \ |
| 101 | hash_PeRlHaSh += *s_PeRlHaSh++; \ |
| 102 | hash_PeRlHaSh += (hash_PeRlHaSh << 10); \ |
| 103 | hash_PeRlHaSh ^= (hash_PeRlHaSh >> 6); \ |
| 104 | } \ |
| 105 | hash_PeRlHaSh += (hash_PeRlHaSh << 3); \ |
| 106 | hash_PeRlHaSh ^= (hash_PeRlHaSh >> 11); \ |
| 107 | (hash) = (hash_PeRlHaSh + (hash_PeRlHaSh << 15)); \ |
| 108 | } STMT_END |
| 109 | #endif |
| 110 | |
| 111 | /* |
| 112 | =head1 Hash Manipulation Functions |
| 113 | |
| 114 | =for apidoc AmU||HEf_SVKEY |
| 115 | This flag, used in the length slot of hash entries and magic structures, |
| 116 | specifies the structure contains an C<SV*> pointer where a C<char*> pointer |
| 117 | is to be expected. (For information only--not to be used). |
| 118 | |
| 119 | =head1 Handy Values |
| 120 | |
| 121 | =for apidoc AmU||Nullhv |
| 122 | Null HV pointer. |
| 123 | |
| 124 | =head1 Hash Manipulation Functions |
| 125 | |
| 126 | =for apidoc Am|char*|HvNAME|HV* stash |
| 127 | Returns the package name of a stash, or NULL if C<stash> isn't a stash. |
| 128 | See C<SvSTASH>, C<CvSTASH>. |
| 129 | |
| 130 | =for apidoc Am|void*|HeKEY|HE* he |
| 131 | Returns the actual pointer stored in the key slot of the hash entry. The |
| 132 | pointer may be either C<char*> or C<SV*>, depending on the value of |
| 133 | C<HeKLEN()>. Can be assigned to. The C<HePV()> or C<HeSVKEY()> macros are |
| 134 | usually preferable for finding the value of a key. |
| 135 | |
| 136 | =for apidoc Am|STRLEN|HeKLEN|HE* he |
| 137 | If this is negative, and amounts to C<HEf_SVKEY>, it indicates the entry |
| 138 | holds an C<SV*> key. Otherwise, holds the actual length of the key. Can |
| 139 | be assigned to. The C<HePV()> macro is usually preferable for finding key |
| 140 | lengths. |
| 141 | |
| 142 | =for apidoc Am|SV*|HeVAL|HE* he |
| 143 | Returns the value slot (type C<SV*>) stored in the hash entry. |
| 144 | |
| 145 | =for apidoc Am|U32|HeHASH|HE* he |
| 146 | Returns the computed hash stored in the hash entry. |
| 147 | |
| 148 | =for apidoc Am|char*|HePV|HE* he|STRLEN len |
| 149 | Returns the key slot of the hash entry as a C<char*> value, doing any |
| 150 | necessary dereferencing of possibly C<SV*> keys. The length of the string |
| 151 | is placed in C<len> (this is a macro, so do I<not> use C<&len>). If you do |
| 152 | not care about what the length of the key is, you may use the global |
| 153 | variable C<PL_na>, though this is rather less efficient than using a local |
| 154 | variable. Remember though, that hash keys in perl are free to contain |
| 155 | embedded nulls, so using C<strlen()> or similar is not a good way to find |
| 156 | the length of hash keys. This is very similar to the C<SvPV()> macro |
| 157 | described elsewhere in this document. |
| 158 | |
| 159 | =for apidoc Am|SV*|HeSVKEY|HE* he |
| 160 | Returns the key as an C<SV*>, or C<Nullsv> if the hash entry does not |
| 161 | contain an C<SV*> key. |
| 162 | |
| 163 | =for apidoc Am|SV*|HeSVKEY_force|HE* he |
| 164 | Returns the key as an C<SV*>. Will create and return a temporary mortal |
| 165 | C<SV*> if the hash entry contains only a C<char*> key. |
| 166 | |
| 167 | =for apidoc Am|SV*|HeSVKEY_set|HE* he|SV* sv |
| 168 | Sets the key to a given C<SV*>, taking care to set the appropriate flags to |
| 169 | indicate the presence of an C<SV*> key, and returns the same |
| 170 | C<SV*>. |
| 171 | |
| 172 | =cut |
| 173 | */ |
| 174 | |
| 175 | /* these hash entry flags ride on hent_klen (for use only in magic/tied HVs) */ |
| 176 | #define HEf_SVKEY -2 /* hent_key is an SV* */ |
| 177 | |
| 178 | |
| 179 | #define Nullhv Null(HV*) |
| 180 | #define HvARRAY(hv) (*(HE***)&((XPVHV*) SvANY(hv))->xhv_array) |
| 181 | #define HvFILL(hv) ((XPVHV*) SvANY(hv))->xhv_fill |
| 182 | #define HvMAX(hv) ((XPVHV*) SvANY(hv))->xhv_max |
| 183 | #define HvRITER(hv) ((XPVHV*) SvANY(hv))->xhv_riter |
| 184 | #define HvRITER_get(hv) (0 + ((XPVHV*) SvANY(hv))->xhv_riter) |
| 185 | #define HvRITER_set(hv,r) (HvRITER(hv) = (r)) |
| 186 | #define HvEITER(hv) ((XPVHV*) SvANY(hv))->xhv_eiter |
| 187 | #define HvEITER_get(hv) (0 + ((XPVHV*) SvANY(hv))->xhv_eiter) |
| 188 | #define HvEITER_set(hv,e) (HvEITER(hv) = (e)) |
| 189 | #define HvPMROOT(hv) ((XPVHV*) SvANY(hv))->xhv_pmroot |
| 190 | #define HvNAME(hv) ((XPVHV*) SvANY(hv))->xhv_name |
| 191 | /* FIXME - all of these should use a UTF8 aware API, which should also involve |
| 192 | getting the length. */ |
| 193 | #define HvNAME_get(hv) (0 + ((XPVHV*) SvANY(hv))->xhv_name) |
| 194 | #define hv_name_set(hv,name,length,flags) \ |
| 195 | (HvNAME((hv)) = (name) ? savepvn(name, length) : 0) |
| 196 | |
| 197 | /* the number of keys (including any placeholers) */ |
| 198 | #define XHvTOTALKEYS(xhv) ((xhv)->xhv_keys) |
| 199 | |
| 200 | /* The number of placeholders in the enumerated-keys hash */ |
| 201 | #define XHvPLACEHOLDERS(xhv) ((xhv)->xhv_placeholders) |
| 202 | |
| 203 | /* the number of keys that exist() (i.e. excluding placeholders) */ |
| 204 | #define XHvUSEDKEYS(xhv) (XHvTOTALKEYS(xhv) - (IV)XHvPLACEHOLDERS(xhv)) |
| 205 | |
| 206 | /* |
| 207 | * HvKEYS gets the number of keys that actually exist(), and is provided |
| 208 | * for backwards compatibility with old XS code. The core uses HvUSEDKEYS |
| 209 | * (keys, excluding placeholdes) and HvTOTALKEYS (including placeholders) |
| 210 | */ |
| 211 | #define HvKEYS(hv) XHvUSEDKEYS((XPVHV*) SvANY(hv)) |
| 212 | #define HvUSEDKEYS(hv) XHvUSEDKEYS((XPVHV*) SvANY(hv)) |
| 213 | #define HvTOTALKEYS(hv) XHvTOTALKEYS((XPVHV*) SvANY(hv)) |
| 214 | #define HvPLACEHOLDERS(hv) (XHvPLACEHOLDERS((XPVHV*) SvANY(hv))) |
| 215 | #define HvPLACEHOLDERS_get(hv) (0 + XHvPLACEHOLDERS((XPVHV*) SvANY(hv))) |
| 216 | #define HvPLACEHOLDERS_set(hv, p) \ |
| 217 | (XHvPLACEHOLDERS((XPVHV*) SvANY(hv)) = (p)) |
| 218 | |
| 219 | #define HvSHAREKEYS(hv) (SvFLAGS(hv) & SVphv_SHAREKEYS) |
| 220 | #define HvSHAREKEYS_on(hv) (SvFLAGS(hv) |= SVphv_SHAREKEYS) |
| 221 | #define HvSHAREKEYS_off(hv) (SvFLAGS(hv) &= ~SVphv_SHAREKEYS) |
| 222 | |
| 223 | /* This is an optimisation flag. It won't be set if all hash keys have a 0 |
| 224 | * flag. Currently the only flags relate to utf8. |
| 225 | * Hence it won't be set if all keys are 8 bit only. It will be set if any key |
| 226 | * is utf8 (including 8 bit keys that were entered as utf8, and need upgrading |
| 227 | * when retrieved during iteration. It may still be set when there are no longer |
| 228 | * any utf8 keys. |
| 229 | * See HVhek_ENABLEHVKFLAGS for the trigger. |
| 230 | */ |
| 231 | #define HvHASKFLAGS(hv) (SvFLAGS(hv) & SVphv_HASKFLAGS) |
| 232 | #define HvHASKFLAGS_on(hv) (SvFLAGS(hv) |= SVphv_HASKFLAGS) |
| 233 | #define HvHASKFLAGS_off(hv) (SvFLAGS(hv) &= ~SVphv_HASKFLAGS) |
| 234 | |
| 235 | #define HvLAZYDEL(hv) (SvFLAGS(hv) & SVphv_LAZYDEL) |
| 236 | #define HvLAZYDEL_on(hv) (SvFLAGS(hv) |= SVphv_LAZYDEL) |
| 237 | #define HvLAZYDEL_off(hv) (SvFLAGS(hv) &= ~SVphv_LAZYDEL) |
| 238 | |
| 239 | #define HvREHASH(hv) (SvFLAGS(hv) & SVphv_REHASH) |
| 240 | #define HvREHASH_on(hv) (SvFLAGS(hv) |= SVphv_REHASH) |
| 241 | #define HvREHASH_off(hv) (SvFLAGS(hv) &= ~SVphv_REHASH) |
| 242 | |
| 243 | /* Maybe amagical: */ |
| 244 | /* #define HV_AMAGICmb(hv) (SvFLAGS(hv) & (SVpgv_badAM | SVpgv_AM)) */ |
| 245 | |
| 246 | #define HV_AMAGIC(hv) (SvFLAGS(hv) & SVpgv_AM) |
| 247 | #define HV_AMAGIC_on(hv) (SvFLAGS(hv) |= SVpgv_AM) |
| 248 | #define HV_AMAGIC_off(hv) (SvFLAGS(hv) &= ~SVpgv_AM) |
| 249 | |
| 250 | /* |
| 251 | #define HV_AMAGICbad(hv) (SvFLAGS(hv) & SVpgv_badAM) |
| 252 | #define HV_badAMAGIC_on(hv) (SvFLAGS(hv) |= SVpgv_badAM) |
| 253 | #define HV_badAMAGIC_off(hv) (SvFLAGS(hv) &= ~SVpgv_badAM) |
| 254 | */ |
| 255 | |
| 256 | #define Nullhe Null(HE*) |
| 257 | #define HeNEXT(he) (he)->hent_next |
| 258 | #define HeKEY_hek(he) (he)->hent_hek |
| 259 | #define HeKEY(he) HEK_KEY(HeKEY_hek(he)) |
| 260 | #define HeKEY_sv(he) (*(SV**)HeKEY(he)) |
| 261 | #define HeKLEN(he) HEK_LEN(HeKEY_hek(he)) |
| 262 | #define HeKUTF8(he) HEK_UTF8(HeKEY_hek(he)) |
| 263 | #define HeKWASUTF8(he) HEK_WASUTF8(HeKEY_hek(he)) |
| 264 | #define HeKREHASH(he) HEK_REHASH(HeKEY_hek(he)) |
| 265 | #define HeKLEN_UTF8(he) (HeKUTF8(he) ? -HeKLEN(he) : HeKLEN(he)) |
| 266 | #define HeKFLAGS(he) HEK_FLAGS(HeKEY_hek(he)) |
| 267 | #define HeVAL(he) (he)->hent_val |
| 268 | #define HeHASH(he) HEK_HASH(HeKEY_hek(he)) |
| 269 | #define HePV(he,lp) ((HeKLEN(he) == HEf_SVKEY) ? \ |
| 270 | SvPV(HeKEY_sv(he),lp) : \ |
| 271 | (((lp = HeKLEN(he)) >= 0) ? \ |
| 272 | HeKEY(he) : Nullch)) |
| 273 | |
| 274 | #define HeSVKEY(he) ((HeKEY(he) && \ |
| 275 | HeKLEN(he) == HEf_SVKEY) ? \ |
| 276 | HeKEY_sv(he) : Nullsv) |
| 277 | |
| 278 | #define HeSVKEY_force(he) (HeKEY(he) ? \ |
| 279 | ((HeKLEN(he) == HEf_SVKEY) ? \ |
| 280 | HeKEY_sv(he) : \ |
| 281 | sv_2mortal(newSVpvn(HeKEY(he), \ |
| 282 | HeKLEN(he)))) : \ |
| 283 | &PL_sv_undef) |
| 284 | #define HeSVKEY_set(he,sv) ((HeKLEN(he) = HEf_SVKEY), (HeKEY_sv(he) = sv)) |
| 285 | |
| 286 | #define Nullhek Null(HEK*) |
| 287 | #define HEK_BASESIZE STRUCT_OFFSET(HEK, hek_key[0]) |
| 288 | #define HEK_HASH(hek) (hek)->hek_hash |
| 289 | #define HEK_LEN(hek) (hek)->hek_len |
| 290 | #define HEK_KEY(hek) (hek)->hek_key |
| 291 | #define HEK_FLAGS(hek) (*((unsigned char *)(HEK_KEY(hek))+HEK_LEN(hek)+1)) |
| 292 | |
| 293 | #define HVhek_UTF8 0x01 /* Key is utf8 encoded. */ |
| 294 | #define HVhek_WASUTF8 0x02 /* Key is bytes here, but was supplied as utf8. */ |
| 295 | #define HVhek_REHASH 0x04 /* This key is in an hv using a custom HASH . */ |
| 296 | #define HVhek_FREEKEY 0x100 /* Internal flag to say key is malloc()ed. */ |
| 297 | #define HVhek_PLACEHOLD 0x200 /* Internal flag to create placeholder. |
| 298 | * (may change, but Storable is a core module) */ |
| 299 | #define HVhek_MASK 0xFF |
| 300 | |
| 301 | /* Which flags enable HvHASKFLAGS? Somewhat a hack on a hack, as |
| 302 | HVhek_REHASH is only needed because the rehash flag has to be duplicated |
| 303 | into all keys as hv_iternext has no access to the hash flags. At this |
| 304 | point Storable's tests get upset, because sometimes hashes are "keyed" |
| 305 | and sometimes not, depending on the order of data insertion, and whether |
| 306 | it triggered rehashing. So currently HVhek_REHAS is exempt. |
| 307 | */ |
| 308 | |
| 309 | #define HVhek_ENABLEHVKFLAGS (HVhek_MASK - HVhek_REHASH) |
| 310 | |
| 311 | #define HEK_UTF8(hek) (HEK_FLAGS(hek) & HVhek_UTF8) |
| 312 | #define HEK_UTF8_on(hek) (HEK_FLAGS(hek) |= HVhek_UTF8) |
| 313 | #define HEK_UTF8_off(hek) (HEK_FLAGS(hek) &= ~HVhek_UTF8) |
| 314 | #define HEK_WASUTF8(hek) (HEK_FLAGS(hek) & HVhek_WASUTF8) |
| 315 | #define HEK_WASUTF8_on(hek) (HEK_FLAGS(hek) |= HVhek_WASUTF8) |
| 316 | #define HEK_WASUTF8_off(hek) (HEK_FLAGS(hek) &= ~HVhek_WASUTF8) |
| 317 | #define HEK_REHASH(hek) (HEK_FLAGS(hek) & HVhek_REHASH) |
| 318 | #define HEK_REHASH_on(hek) (HEK_FLAGS(hek) |= HVhek_REHASH) |
| 319 | |
| 320 | /* calculate HV array allocation */ |
| 321 | #ifndef PERL_USE_LARGE_HV_ALLOC |
| 322 | /* Default to allocating the correct size - default to assuming that malloc() |
| 323 | is not broken and is efficient at allocating blocks sized at powers-of-two. |
| 324 | */ |
| 325 | # define PERL_HV_ARRAY_ALLOC_BYTES(size) ((size) * sizeof(HE*)) |
| 326 | #else |
| 327 | # define MALLOC_OVERHEAD 16 |
| 328 | # define PERL_HV_ARRAY_ALLOC_BYTES(size) \ |
| 329 | (((size) < 64) \ |
| 330 | ? (size) * sizeof(HE*) \ |
| 331 | : (size) * sizeof(HE*) * 2 - MALLOC_OVERHEAD) |
| 332 | #endif |
| 333 | |
| 334 | /* Flags for hv_iternext_flags. */ |
| 335 | #define HV_ITERNEXT_WANTPLACEHOLDERS 0x01 /* Don't skip placeholders. */ |
| 336 | |
| 337 | /* available as a function in hv.c */ |
| 338 | #define Perl_sharepvn(sv, len, hash) HEK_KEY(share_hek(sv, len, hash)) |
| 339 | #define sharepvn(sv, len, hash) Perl_sharepvn(sv, len, hash) |
| 340 | |
| 341 | /* |
| 342 | * Local variables: |
| 343 | * c-indentation-style: bsd |
| 344 | * c-basic-offset: 4 |
| 345 | * indent-tabs-mode: t |
| 346 | * End: |
| 347 | * |
| 348 | * ex: set ts=8 sts=4 sw=4 noet: |
| 349 | */ |