3 * Copyright (C) 1991, 1992, 1993, 1996, 1997, 1998, 1999,
4 * 2000, 2001, 2002, 2005, by Larry Wall and others
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.
11 /* entry in hash value chain */
13 /* Keep hent_next first in this structure, because sv_free_arenas take
14 advantage of this to share code between the he arenas and the SV
16 HE *hent_next; /* next entry in chain */
17 HEK *hent_hek; /* hash key */
19 SV *hent_val; /* scalar value that was hashed */
20 Size_t hent_refcount; /* references for this shared hash key */
24 /* hash key -- defined separately for use as shared pointer */
26 U32 hek_hash; /* hash of key */
27 I32 hek_len; /* length of hash key */
28 char hek_key[1]; /* variable-length hash key */
29 /* the hash-key is \0-terminated */
30 /* after the \0 there is a byte for flags, such as whether the key
35 struct he shared_he_he;
36 struct hek shared_he_hek;
40 Don't access this directly.
43 HEK *xhv_name; /* name, if a symbol table */
44 AV *xhv_backreferences; /* back references for weak references */
45 HE *xhv_eiter; /* current entry of iterator */
46 I32 xhv_riter; /* current root of iterator */
50 /* This structure must match the beginning of struct xpvmg in sv.h. */
52 NV xnv_nv; /* numeric value, if any */
53 STRLEN xhv_fill; /* how full xhv_array currently is */
54 STRLEN xhv_max; /* subscript of last element of xhv_array */
56 IV xivu_iv; /* integer value or pv offset */
61 MAGIC* xmg_magic; /* linked list of magicalness */
62 HV* xmg_ourstash; /* Stash for our (when SvPAD_OUR is true) */
64 HV* xmg_stash; /* class package */
67 #define xhv_keys xiv_u.xivu_iv
70 typedef struct xpvhv xpvhv_allocated;
73 STRLEN xhv_fill; /* how full xhv_array currently is */
74 STRLEN xhv_max; /* subscript of last element of xhv_array */
76 IV xivu_iv; /* integer value or pv offset */
81 MAGIC* xmg_magic; /* linked list of magicalness */
82 HV* xmg_ourstash; /* Stash for our (when SvPAD_OUR is true) */
84 HV* xmg_stash; /* class package */
89 /* FYI: This is the "One-at-a-Time" algorithm by Bob Jenkins
90 * from requirements by Colin Plumb.
91 * (http://burtleburtle.net/bob/hash/doobs.html) */
92 /* The use of a temporary pointer and the casting games
93 * is needed to serve the dual purposes of
94 * (a) the hashed data being interpreted as "unsigned char" (new since 5.8,
95 * a "char" can be either signed or unsigned, depending on the compiler)
96 * (b) catering for old code that uses a "char"
98 * The "hash seed" feature was added in Perl 5.8.1 to perturb the results
99 * to avoid "algorithmic complexity attacks".
101 * If USE_HASH_SEED is defined, hash randomisation is done by default
102 * If USE_HASH_SEED_EXPLICIT is defined, hash randomisation is done
103 * only if the environment variable PERL_HASH_SEED is set.
104 * For maximal control, one can define PERL_HASH_SEED.
105 * (see also perl.c:perl_parse()).
107 #ifndef PERL_HASH_SEED
108 # if defined(USE_HASH_SEED) || defined(USE_HASH_SEED_EXPLICIT)
109 # define PERL_HASH_SEED PL_hash_seed
111 # define PERL_HASH_SEED 0
114 #define PERL_HASH(hash,str,len) \
116 register const char * const s_PeRlHaSh_tmp = str; \
117 register const unsigned char *s_PeRlHaSh = (const unsigned char *)s_PeRlHaSh_tmp; \
118 register I32 i_PeRlHaSh = len; \
119 register U32 hash_PeRlHaSh = PERL_HASH_SEED; \
120 while (i_PeRlHaSh--) { \
121 hash_PeRlHaSh += *s_PeRlHaSh++; \
122 hash_PeRlHaSh += (hash_PeRlHaSh << 10); \
123 hash_PeRlHaSh ^= (hash_PeRlHaSh >> 6); \
125 hash_PeRlHaSh += (hash_PeRlHaSh << 3); \
126 hash_PeRlHaSh ^= (hash_PeRlHaSh >> 11); \
127 (hash) = (hash_PeRlHaSh + (hash_PeRlHaSh << 15)); \
130 /* Only hv.c and mod_perl should be doing this. */
131 #ifdef PERL_HASH_INTERNAL_ACCESS
132 #define PERL_HASH_INTERNAL(hash,str,len) \
134 register const char * const s_PeRlHaSh_tmp = str; \
135 register const unsigned char *s_PeRlHaSh = (const unsigned char *)s_PeRlHaSh_tmp; \
136 register I32 i_PeRlHaSh = len; \
137 register U32 hash_PeRlHaSh = PL_rehash_seed; \
138 while (i_PeRlHaSh--) { \
139 hash_PeRlHaSh += *s_PeRlHaSh++; \
140 hash_PeRlHaSh += (hash_PeRlHaSh << 10); \
141 hash_PeRlHaSh ^= (hash_PeRlHaSh >> 6); \
143 hash_PeRlHaSh += (hash_PeRlHaSh << 3); \
144 hash_PeRlHaSh ^= (hash_PeRlHaSh >> 11); \
145 (hash) = (hash_PeRlHaSh + (hash_PeRlHaSh << 15)); \
150 =head1 Hash Manipulation Functions
152 =for apidoc AmU||HEf_SVKEY
153 This flag, used in the length slot of hash entries and magic structures,
154 specifies the structure contains an C<SV*> pointer where a C<char*> pointer
155 is to be expected. (For information only--not to be used).
159 =for apidoc AmU||Nullhv
162 =head1 Hash Manipulation Functions
164 =for apidoc Am|char*|HvNAME|HV* stash
165 Returns the package name of a stash, or NULL if C<stash> isn't a stash.
166 See C<SvSTASH>, C<CvSTASH>.
168 =for apidoc Am|void*|HeKEY|HE* he
169 Returns the actual pointer stored in the key slot of the hash entry. The
170 pointer may be either C<char*> or C<SV*>, depending on the value of
171 C<HeKLEN()>. Can be assigned to. The C<HePV()> or C<HeSVKEY()> macros are
172 usually preferable for finding the value of a key.
174 =for apidoc Am|STRLEN|HeKLEN|HE* he
175 If this is negative, and amounts to C<HEf_SVKEY>, it indicates the entry
176 holds an C<SV*> key. Otherwise, holds the actual length of the key. Can
177 be assigned to. The C<HePV()> macro is usually preferable for finding key
180 =for apidoc Am|SV*|HeVAL|HE* he
181 Returns the value slot (type C<SV*>) stored in the hash entry.
183 =for apidoc Am|U32|HeHASH|HE* he
184 Returns the computed hash stored in the hash entry.
186 =for apidoc Am|char*|HePV|HE* he|STRLEN len
187 Returns the key slot of the hash entry as a C<char*> value, doing any
188 necessary dereferencing of possibly C<SV*> keys. The length of the string
189 is placed in C<len> (this is a macro, so do I<not> use C<&len>). If you do
190 not care about what the length of the key is, you may use the global
191 variable C<PL_na>, though this is rather less efficient than using a local
192 variable. Remember though, that hash keys in perl are free to contain
193 embedded nulls, so using C<strlen()> or similar is not a good way to find
194 the length of hash keys. This is very similar to the C<SvPV()> macro
195 described elsewhere in this document.
197 =for apidoc Am|SV*|HeSVKEY|HE* he
198 Returns the key as an C<SV*>, or C<NULL> if the hash entry does not
199 contain an C<SV*> key.
201 =for apidoc Am|SV*|HeSVKEY_force|HE* he
202 Returns the key as an C<SV*>. Will create and return a temporary mortal
203 C<SV*> if the hash entry contains only a C<char*> key.
205 =for apidoc Am|SV*|HeSVKEY_set|HE* he|SV* sv
206 Sets the key to a given C<SV*>, taking care to set the appropriate flags to
207 indicate the presence of an C<SV*> key, and returns the same
213 /* these hash entry flags ride on hent_klen (for use only in magic/tied HVs) */
214 #define HEf_SVKEY -2 /* hent_key is an SV* */
217 #define Nullhv Null(HV*)
218 #define HvARRAY(hv) ((hv)->sv_u.svu_hash)
219 #define HvFILL(hv) ((XPVHV*) SvANY(hv))->xhv_fill
220 #define HvMAX(hv) ((XPVHV*) SvANY(hv))->xhv_max
221 /* This quite intentionally does no flag checking first. That's your
223 #define HvAUX(hv) ((struct xpvhv_aux*)&(HvARRAY(hv)[HvMAX(hv)+1]))
224 #define HvRITER(hv) (*Perl_hv_riter_p(aTHX_ (HV*)(hv)))
225 #define HvEITER(hv) (*Perl_hv_eiter_p(aTHX_ (HV*)(hv)))
226 #define HvRITER_set(hv,r) Perl_hv_riter_set(aTHX_ (HV*)(hv), r)
227 #define HvEITER_set(hv,e) Perl_hv_eiter_set(aTHX_ (HV*)(hv), e)
228 #define HvRITER_get(hv) (SvOOK(hv) ? HvAUX(hv)->xhv_riter : -1)
229 #define HvEITER_get(hv) (SvOOK(hv) ? HvAUX(hv)->xhv_eiter : 0)
230 #define HvNAME(hv) HvNAME_get(hv)
231 /* FIXME - all of these should use a UTF8 aware API, which should also involve
232 getting the length. */
233 /* This macro may go away without notice. */
234 #define HvNAME_HEK(hv) (SvOOK(hv) ? HvAUX(hv)->xhv_name : 0)
235 #define HvNAME_get(hv) ((SvOOK(hv) && (HvAUX(hv)->xhv_name)) \
236 ? HEK_KEY(HvAUX(hv)->xhv_name) : 0)
237 #define HvNAMELEN_get(hv) ((SvOOK(hv) && (HvAUX(hv)->xhv_name)) \
238 ? HEK_LEN(HvAUX(hv)->xhv_name) : 0)
240 /* the number of keys (including any placeholers) */
241 #define XHvTOTALKEYS(xhv) ((xhv)->xhv_keys)
244 * HvKEYS gets the number of keys that actually exist(), and is provided
245 * for backwards compatibility with old XS code. The core uses HvUSEDKEYS
246 * (keys, excluding placeholdes) and HvTOTALKEYS (including placeholders)
248 #define HvKEYS(hv) HvUSEDKEYS(hv)
249 #define HvUSEDKEYS(hv) (HvTOTALKEYS(hv) - HvPLACEHOLDERS_get(hv))
250 #define HvTOTALKEYS(hv) XHvTOTALKEYS((XPVHV*) SvANY(hv))
251 #define HvPLACEHOLDERS(hv) (*Perl_hv_placeholders_p(aTHX_ (HV*)hv))
252 #define HvPLACEHOLDERS_get(hv) (SvMAGIC(hv) ? Perl_hv_placeholders_get(aTHX_ (HV*)hv) : 0)
253 #define HvPLACEHOLDERS_set(hv,p) Perl_hv_placeholders_set(aTHX_ (HV*)hv, p)
255 #define HvSHAREKEYS(hv) (SvFLAGS(hv) & SVphv_SHAREKEYS)
256 #define HvSHAREKEYS_on(hv) (SvFLAGS(hv) |= SVphv_SHAREKEYS)
257 #define HvSHAREKEYS_off(hv) (SvFLAGS(hv) &= ~SVphv_SHAREKEYS)
259 /* This is an optimisation flag. It won't be set if all hash keys have a 0
260 * flag. Currently the only flags relate to utf8.
261 * Hence it won't be set if all keys are 8 bit only. It will be set if any key
262 * is utf8 (including 8 bit keys that were entered as utf8, and need upgrading
263 * when retrieved during iteration. It may still be set when there are no longer
265 * See HVhek_ENABLEHVKFLAGS for the trigger.
267 #define HvHASKFLAGS(hv) (SvFLAGS(hv) & SVphv_HASKFLAGS)
268 #define HvHASKFLAGS_on(hv) (SvFLAGS(hv) |= SVphv_HASKFLAGS)
269 #define HvHASKFLAGS_off(hv) (SvFLAGS(hv) &= ~SVphv_HASKFLAGS)
271 #define HvLAZYDEL(hv) (SvFLAGS(hv) & SVphv_LAZYDEL)
272 #define HvLAZYDEL_on(hv) (SvFLAGS(hv) |= SVphv_LAZYDEL)
273 #define HvLAZYDEL_off(hv) (SvFLAGS(hv) &= ~SVphv_LAZYDEL)
275 #define HvREHASH(hv) (SvFLAGS(hv) & SVphv_REHASH)
276 #define HvREHASH_on(hv) (SvFLAGS(hv) |= SVphv_REHASH)
277 #define HvREHASH_off(hv) (SvFLAGS(hv) &= ~SVphv_REHASH)
279 /* Maybe amagical: */
280 /* #define HV_AMAGICmb(hv) (SvFLAGS(hv) & (SVpgv_badAM | SVpgv_AM)) */
282 #define HV_AMAGIC(hv) (SvFLAGS(hv) & SVpgv_AM)
283 #define HV_AMAGIC_on(hv) (SvFLAGS(hv) |= SVpgv_AM)
284 #define HV_AMAGIC_off(hv) (SvFLAGS(hv) &= ~SVpgv_AM)
287 #define HV_AMAGICbad(hv) (SvFLAGS(hv) & SVpgv_badAM)
288 #define HV_badAMAGIC_on(hv) (SvFLAGS(hv) |= SVpgv_badAM)
289 #define HV_badAMAGIC_off(hv) (SvFLAGS(hv) &= ~SVpgv_badAM)
292 #define Nullhe Null(HE*)
293 #define HeNEXT(he) (he)->hent_next
294 #define HeKEY_hek(he) (he)->hent_hek
295 #define HeKEY(he) HEK_KEY(HeKEY_hek(he))
296 #define HeKEY_sv(he) (*(SV**)HeKEY(he))
297 #define HeKLEN(he) HEK_LEN(HeKEY_hek(he))
298 #define HeKUTF8(he) HEK_UTF8(HeKEY_hek(he))
299 #define HeKWASUTF8(he) HEK_WASUTF8(HeKEY_hek(he))
300 #define HeKREHASH(he) HEK_REHASH(HeKEY_hek(he))
301 #define HeKLEN_UTF8(he) (HeKUTF8(he) ? -HeKLEN(he) : HeKLEN(he))
302 #define HeKFLAGS(he) HEK_FLAGS(HeKEY_hek(he))
303 #define HeVAL(he) (he)->he_valu.hent_val
304 #define HeHASH(he) HEK_HASH(HeKEY_hek(he))
305 #define HePV(he,lp) ((HeKLEN(he) == HEf_SVKEY) ? \
306 SvPV(HeKEY_sv(he),lp) : \
307 (((lp = HeKLEN(he)) >= 0) ? \
310 #define HeSVKEY(he) ((HeKEY(he) && \
311 HeKLEN(he) == HEf_SVKEY) ? \
314 #define HeSVKEY_force(he) (HeKEY(he) ? \
315 ((HeKLEN(he) == HEf_SVKEY) ? \
317 sv_2mortal(newSVpvn(HeKEY(he), \
320 #define HeSVKEY_set(he,sv) ((HeKLEN(he) = HEf_SVKEY), (HeKEY_sv(he) = sv))
322 #define Nullhek Null(HEK*)
323 #define HEK_BASESIZE STRUCT_OFFSET(HEK, hek_key[0])
324 #define HEK_HASH(hek) (hek)->hek_hash
325 #define HEK_LEN(hek) (hek)->hek_len
326 #define HEK_KEY(hek) (hek)->hek_key
327 #define HEK_FLAGS(hek) (*((unsigned char *)(HEK_KEY(hek))+HEK_LEN(hek)+1))
329 #define HVhek_UTF8 0x01 /* Key is utf8 encoded. */
330 #define HVhek_WASUTF8 0x02 /* Key is bytes here, but was supplied as utf8. */
331 #define HVhek_REHASH 0x04 /* This key is in an hv using a custom HASH . */
332 #define HVhek_FREEKEY 0x100 /* Internal flag to say key is malloc()ed. */
333 #define HVhek_PLACEHOLD 0x200 /* Internal flag to create placeholder.
334 * (may change, but Storable is a core module) */
335 #define HVhek_MASK 0xFF
337 /* Which flags enable HvHASKFLAGS? Somewhat a hack on a hack, as
338 HVhek_REHASH is only needed because the rehash flag has to be duplicated
339 into all keys as hv_iternext has no access to the hash flags. At this
340 point Storable's tests get upset, because sometimes hashes are "keyed"
341 and sometimes not, depending on the order of data insertion, and whether
342 it triggered rehashing. So currently HVhek_REHAS is exempt.
345 #define HVhek_ENABLEHVKFLAGS (HVhek_MASK - HVhek_REHASH)
347 #define HEK_UTF8(hek) (HEK_FLAGS(hek) & HVhek_UTF8)
348 #define HEK_UTF8_on(hek) (HEK_FLAGS(hek) |= HVhek_UTF8)
349 #define HEK_UTF8_off(hek) (HEK_FLAGS(hek) &= ~HVhek_UTF8)
350 #define HEK_WASUTF8(hek) (HEK_FLAGS(hek) & HVhek_WASUTF8)
351 #define HEK_WASUTF8_on(hek) (HEK_FLAGS(hek) |= HVhek_WASUTF8)
352 #define HEK_WASUTF8_off(hek) (HEK_FLAGS(hek) &= ~HVhek_WASUTF8)
353 #define HEK_REHASH(hek) (HEK_FLAGS(hek) & HVhek_REHASH)
354 #define HEK_REHASH_on(hek) (HEK_FLAGS(hek) |= HVhek_REHASH)
356 /* calculate HV array allocation */
357 #ifndef PERL_USE_LARGE_HV_ALLOC
358 /* Default to allocating the correct size - default to assuming that malloc()
359 is not broken and is efficient at allocating blocks sized at powers-of-two.
361 # define PERL_HV_ARRAY_ALLOC_BYTES(size) ((size) * sizeof(HE*))
363 # define MALLOC_OVERHEAD 16
364 # define PERL_HV_ARRAY_ALLOC_BYTES(size) \
366 ? (size) * sizeof(HE*) \
367 : (size) * sizeof(HE*) * 2 - MALLOC_OVERHEAD)
370 /* Flags for hv_iternext_flags. */
371 #define HV_ITERNEXT_WANTPLACEHOLDERS 0x01 /* Don't skip placeholders. */
373 #define hv_iternext(hv) hv_iternext_flags(hv, 0)
374 #define hv_magic(hv, gv, how) sv_magic((SV*)(hv), (SV*)(gv), how, NULL, 0)
376 /* available as a function in hv.c */
377 #define Perl_sharepvn(sv, len, hash) HEK_KEY(share_hek(sv, len, hash))
378 #define sharepvn(sv, len, hash) Perl_sharepvn(sv, len, hash)
380 #define share_hek_hek(hek) \
381 (++(((struct shared_he *)(((char *)hek) \
382 - STRUCT_OFFSET(struct shared_he, \
384 ->shared_he_he.he_valu.hent_refcount), \
389 * c-indentation-style: bsd
391 * indent-tabs-mode: t
394 * ex: set ts=8 sts=4 sw=4 noet: