3 * Copyright (C) 1991, 1992, 1993, 1996, 1997, 1998, 1999,
4 * 2000, 2001, 2002, 2003, 2005, 2006, 2007, 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. */
53 NV xnv_nv; /* numeric value, if any */
58 } xpad_cop_seq; /* used by pad.c for cop_sequence */
60 U32 xbm_previous; /* how many characters in string before rare? */
62 U8 xbm_rare; /* rarest character in string */
63 } xbm_s; /* fields from PVBM */
65 STRLEN xhv_fill; /* how full xhv_array currently is */
66 STRLEN xhv_max; /* subscript of last element of xhv_array */
68 IV xivu_iv; /* integer value or pv offset */
75 MAGIC* xmg_magic; /* linked list of magicalness */
76 HV* xmg_ourstash; /* Stash for our (when SvPAD_OUR is true) */
78 HV* xmg_stash; /* class package */
81 #define xhv_keys xiv_u.xivu_iv
84 STRLEN xhv_fill; /* how full xhv_array currently is */
85 STRLEN xhv_max; /* subscript of last element of xhv_array */
87 IV xivu_iv; /* integer value or pv offset */
94 MAGIC* xmg_magic; /* linked list of magicalness */
95 HV* xmg_ourstash; /* Stash for our (when SvPAD_OUR is true) */
97 HV* xmg_stash; /* class package */
101 /* FYI: This is the "One-at-a-Time" algorithm by Bob Jenkins
102 * from requirements by Colin Plumb.
103 * (http://burtleburtle.net/bob/hash/doobs.html) */
104 /* The use of a temporary pointer and the casting games
105 * is needed to serve the dual purposes of
106 * (a) the hashed data being interpreted as "unsigned char" (new since 5.8,
107 * a "char" can be either signed or unsigned, depending on the compiler)
108 * (b) catering for old code that uses a "char"
110 * The "hash seed" feature was added in Perl 5.8.1 to perturb the results
111 * to avoid "algorithmic complexity attacks".
113 * If USE_HASH_SEED is defined, hash randomisation is done by default
114 * If USE_HASH_SEED_EXPLICIT is defined, hash randomisation is done
115 * only if the environment variable PERL_HASH_SEED is set.
116 * For maximal control, one can define PERL_HASH_SEED.
117 * (see also perl.c:perl_parse()).
119 #ifndef PERL_HASH_SEED
120 # if defined(USE_HASH_SEED) || defined(USE_HASH_SEED_EXPLICIT)
121 # define PERL_HASH_SEED PL_hash_seed
123 # define PERL_HASH_SEED 0
126 #define PERL_HASH(hash,str,len) \
128 register const char * const s_PeRlHaSh_tmp = str; \
129 register const unsigned char *s_PeRlHaSh = (const unsigned char *)s_PeRlHaSh_tmp; \
130 register I32 i_PeRlHaSh = len; \
131 register U32 hash_PeRlHaSh = PERL_HASH_SEED; \
132 while (i_PeRlHaSh--) { \
133 hash_PeRlHaSh += *s_PeRlHaSh++; \
134 hash_PeRlHaSh += (hash_PeRlHaSh << 10); \
135 hash_PeRlHaSh ^= (hash_PeRlHaSh >> 6); \
137 hash_PeRlHaSh += (hash_PeRlHaSh << 3); \
138 hash_PeRlHaSh ^= (hash_PeRlHaSh >> 11); \
139 (hash) = (hash_PeRlHaSh + (hash_PeRlHaSh << 15)); \
142 /* Only hv.c and mod_perl should be doing this. */
143 #ifdef PERL_HASH_INTERNAL_ACCESS
144 #define PERL_HASH_INTERNAL(hash,str,len) \
146 register const char * const s_PeRlHaSh_tmp = str; \
147 register const unsigned char *s_PeRlHaSh = (const unsigned char *)s_PeRlHaSh_tmp; \
148 register I32 i_PeRlHaSh = len; \
149 register U32 hash_PeRlHaSh = PL_rehash_seed; \
150 while (i_PeRlHaSh--) { \
151 hash_PeRlHaSh += *s_PeRlHaSh++; \
152 hash_PeRlHaSh += (hash_PeRlHaSh << 10); \
153 hash_PeRlHaSh ^= (hash_PeRlHaSh >> 6); \
155 hash_PeRlHaSh += (hash_PeRlHaSh << 3); \
156 hash_PeRlHaSh ^= (hash_PeRlHaSh >> 11); \
157 (hash) = (hash_PeRlHaSh + (hash_PeRlHaSh << 15)); \
162 =head1 Hash Manipulation Functions
164 =for apidoc AmU||HEf_SVKEY
165 This flag, used in the length slot of hash entries and magic structures,
166 specifies the structure contains an C<SV*> pointer where a C<char*> pointer
167 is to be expected. (For information only--not to be used).
171 =for apidoc AmU||Nullhv
174 =head1 Hash Manipulation Functions
176 =for apidoc Am|char*|HvNAME|HV* stash
177 Returns the package name of a stash, or NULL if C<stash> isn't a stash.
178 See C<SvSTASH>, C<CvSTASH>.
180 =for apidoc Am|void*|HeKEY|HE* he
181 Returns the actual pointer stored in the key slot of the hash entry. The
182 pointer may be either C<char*> or C<SV*>, depending on the value of
183 C<HeKLEN()>. Can be assigned to. The C<HePV()> or C<HeSVKEY()> macros are
184 usually preferable for finding the value of a key.
186 =for apidoc Am|STRLEN|HeKLEN|HE* he
187 If this is negative, and amounts to C<HEf_SVKEY>, it indicates the entry
188 holds an C<SV*> key. Otherwise, holds the actual length of the key. Can
189 be assigned to. The C<HePV()> macro is usually preferable for finding key
192 =for apidoc Am|SV*|HeVAL|HE* he
193 Returns the value slot (type C<SV*>) stored in the hash entry.
195 =for apidoc Am|U32|HeHASH|HE* he
196 Returns the computed hash stored in the hash entry.
198 =for apidoc Am|char*|HePV|HE* he|STRLEN len
199 Returns the key slot of the hash entry as a C<char*> value, doing any
200 necessary dereferencing of possibly C<SV*> keys. The length of the string
201 is placed in C<len> (this is a macro, so do I<not> use C<&len>). If you do
202 not care about what the length of the key is, you may use the global
203 variable C<PL_na>, though this is rather less efficient than using a local
204 variable. Remember though, that hash keys in perl are free to contain
205 embedded nulls, so using C<strlen()> or similar is not a good way to find
206 the length of hash keys. This is very similar to the C<SvPV()> macro
207 described elsewhere in this document.
209 =for apidoc Am|SV*|HeSVKEY|HE* he
210 Returns the key as an C<SV*>, or C<NULL> if the hash entry does not
211 contain an C<SV*> key.
213 =for apidoc Am|SV*|HeSVKEY_force|HE* he
214 Returns the key as an C<SV*>. Will create and return a temporary mortal
215 C<SV*> if the hash entry contains only a C<char*> key.
217 =for apidoc Am|SV*|HeSVKEY_set|HE* he|SV* sv
218 Sets the key to a given C<SV*>, taking care to set the appropriate flags to
219 indicate the presence of an C<SV*> key, and returns the same
225 /* these hash entry flags ride on hent_klen (for use only in magic/tied HVs) */
226 #define HEf_SVKEY -2 /* hent_key is an SV* */
229 #define Nullhv Null(HV*)
230 #define HvARRAY(hv) ((hv)->sv_u.svu_hash)
231 #define HvFILL(hv) ((XPVHV*) SvANY(hv))->xhv_fill
232 #define HvMAX(hv) ((XPVHV*) SvANY(hv))->xhv_max
233 /* This quite intentionally does no flag checking first. That's your
235 #define HvAUX(hv) ((struct xpvhv_aux*)&(HvARRAY(hv)[HvMAX(hv)+1]))
236 #define HvRITER(hv) (*Perl_hv_riter_p(aTHX_ (HV*)(hv)))
237 #define HvEITER(hv) (*Perl_hv_eiter_p(aTHX_ (HV*)(hv)))
238 #define HvRITER_set(hv,r) Perl_hv_riter_set(aTHX_ (HV*)(hv), r)
239 #define HvEITER_set(hv,e) Perl_hv_eiter_set(aTHX_ (HV*)(hv), e)
240 #define HvRITER_get(hv) (SvOOK(hv) ? HvAUX(hv)->xhv_riter : -1)
241 #define HvEITER_get(hv) (SvOOK(hv) ? HvAUX(hv)->xhv_eiter : 0)
242 #define HvNAME(hv) HvNAME_get(hv)
243 /* FIXME - all of these should use a UTF8 aware API, which should also involve
244 getting the length. */
245 /* This macro may go away without notice. */
246 #define HvNAME_HEK(hv) (SvOOK(hv) ? HvAUX(hv)->xhv_name : 0)
247 #define HvNAME_get(hv) ((SvOOK(hv) && (HvAUX(hv)->xhv_name)) \
248 ? HEK_KEY(HvAUX(hv)->xhv_name) : 0)
249 #define HvNAMELEN_get(hv) ((SvOOK(hv) && (HvAUX(hv)->xhv_name)) \
250 ? HEK_LEN(HvAUX(hv)->xhv_name) : 0)
252 /* the number of keys (including any placeholers) */
253 #define XHvTOTALKEYS(xhv) ((xhv)->xhv_keys)
256 * HvKEYS gets the number of keys that actually exist(), and is provided
257 * for backwards compatibility with old XS code. The core uses HvUSEDKEYS
258 * (keys, excluding placeholdes) and HvTOTALKEYS (including placeholders)
260 #define HvKEYS(hv) HvUSEDKEYS(hv)
261 #define HvUSEDKEYS(hv) (HvTOTALKEYS(hv) - HvPLACEHOLDERS_get(hv))
262 #define HvTOTALKEYS(hv) XHvTOTALKEYS((XPVHV*) SvANY(hv))
263 #define HvPLACEHOLDERS(hv) (*Perl_hv_placeholders_p(aTHX_ (HV*)hv))
264 #define HvPLACEHOLDERS_get(hv) (SvMAGIC(hv) ? Perl_hv_placeholders_get(aTHX_ (HV*)hv) : 0)
265 #define HvPLACEHOLDERS_set(hv,p) Perl_hv_placeholders_set(aTHX_ (HV*)hv, p)
267 #define HvSHAREKEYS(hv) (SvFLAGS(hv) & SVphv_SHAREKEYS)
268 #define HvSHAREKEYS_on(hv) (SvFLAGS(hv) |= SVphv_SHAREKEYS)
269 #define HvSHAREKEYS_off(hv) (SvFLAGS(hv) &= ~SVphv_SHAREKEYS)
271 /* This is an optimisation flag. It won't be set if all hash keys have a 0
272 * flag. Currently the only flags relate to utf8.
273 * Hence it won't be set if all keys are 8 bit only. It will be set if any key
274 * is utf8 (including 8 bit keys that were entered as utf8, and need upgrading
275 * when retrieved during iteration. It may still be set when there are no longer
277 * See HVhek_ENABLEHVKFLAGS for the trigger.
279 #define HvHASKFLAGS(hv) (SvFLAGS(hv) & SVphv_HASKFLAGS)
280 #define HvHASKFLAGS_on(hv) (SvFLAGS(hv) |= SVphv_HASKFLAGS)
281 #define HvHASKFLAGS_off(hv) (SvFLAGS(hv) &= ~SVphv_HASKFLAGS)
283 #define HvLAZYDEL(hv) (SvFLAGS(hv) & SVphv_LAZYDEL)
284 #define HvLAZYDEL_on(hv) (SvFLAGS(hv) |= SVphv_LAZYDEL)
285 #define HvLAZYDEL_off(hv) (SvFLAGS(hv) &= ~SVphv_LAZYDEL)
287 #define HvREHASH(hv) (SvFLAGS(hv) & SVphv_REHASH)
288 #define HvREHASH_on(hv) (SvFLAGS(hv) |= SVphv_REHASH)
289 #define HvREHASH_off(hv) (SvFLAGS(hv) &= ~SVphv_REHASH)
291 #define Nullhe Null(HE*)
292 #define HeNEXT(he) (he)->hent_next
293 #define HeKEY_hek(he) (he)->hent_hek
294 #define HeKEY(he) HEK_KEY(HeKEY_hek(he))
295 #define HeKEY_sv(he) (*(SV**)HeKEY(he))
296 #define HeKLEN(he) HEK_LEN(HeKEY_hek(he))
297 #define HeKUTF8(he) HEK_UTF8(HeKEY_hek(he))
298 #define HeKWASUTF8(he) HEK_WASUTF8(HeKEY_hek(he))
299 #define HeKREHASH(he) HEK_REHASH(HeKEY_hek(he))
300 #define HeKLEN_UTF8(he) (HeKUTF8(he) ? -HeKLEN(he) : HeKLEN(he))
301 #define HeKFLAGS(he) HEK_FLAGS(HeKEY_hek(he))
302 #define HeVAL(he) (he)->he_valu.hent_val
303 #define HeHASH(he) HEK_HASH(HeKEY_hek(he))
304 #define HePV(he,lp) ((HeKLEN(he) == HEf_SVKEY) ? \
305 SvPV(HeKEY_sv(he),lp) : \
306 (((lp = HeKLEN(he)) >= 0) ? \
309 #define HeSVKEY(he) ((HeKEY(he) && \
310 HeKLEN(he) == HEf_SVKEY) ? \
313 #define HeSVKEY_force(he) (HeKEY(he) ? \
314 ((HeKLEN(he) == HEf_SVKEY) ? \
316 sv_2mortal(newSVpvn(HeKEY(he), \
319 #define HeSVKEY_set(he,sv) ((HeKLEN(he) = HEf_SVKEY), (HeKEY_sv(he) = sv))
321 #define Nullhek Null(HEK*)
322 #define HEK_BASESIZE STRUCT_OFFSET(HEK, hek_key[0])
323 #define HEK_HASH(hek) (hek)->hek_hash
324 #define HEK_LEN(hek) (hek)->hek_len
325 #define HEK_KEY(hek) (hek)->hek_key
326 #define HEK_FLAGS(hek) (*((unsigned char *)(HEK_KEY(hek))+HEK_LEN(hek)+1))
328 #define HVhek_UTF8 0x01 /* Key is utf8 encoded. */
329 #define HVhek_WASUTF8 0x02 /* Key is bytes here, but was supplied as utf8. */
330 #define HVhek_REHASH 0x04 /* This key is in an hv using a custom HASH . */
331 #define HVhek_UNSHARED 0x08 /* This key isn't a shared hash key. */
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_REHASH is exempt.
346 #define HVhek_ENABLEHVKFLAGS (HVhek_MASK & ~(HVhek_REHASH|HVhek_UNSHARED))
348 #define HEK_UTF8(hek) (HEK_FLAGS(hek) & HVhek_UTF8)
349 #define HEK_UTF8_on(hek) (HEK_FLAGS(hek) |= HVhek_UTF8)
350 #define HEK_UTF8_off(hek) (HEK_FLAGS(hek) &= ~HVhek_UTF8)
351 #define HEK_WASUTF8(hek) (HEK_FLAGS(hek) & HVhek_WASUTF8)
352 #define HEK_WASUTF8_on(hek) (HEK_FLAGS(hek) |= HVhek_WASUTF8)
353 #define HEK_WASUTF8_off(hek) (HEK_FLAGS(hek) &= ~HVhek_WASUTF8)
354 #define HEK_REHASH(hek) (HEK_FLAGS(hek) & HVhek_REHASH)
355 #define HEK_REHASH_on(hek) (HEK_FLAGS(hek) |= HVhek_REHASH)
357 /* calculate HV array allocation */
358 #ifndef PERL_USE_LARGE_HV_ALLOC
359 /* Default to allocating the correct size - default to assuming that malloc()
360 is not broken and is efficient at allocating blocks sized at powers-of-two.
362 # define PERL_HV_ARRAY_ALLOC_BYTES(size) ((size) * sizeof(HE*))
364 # define MALLOC_OVERHEAD 16
365 # define PERL_HV_ARRAY_ALLOC_BYTES(size) \
367 ? (size) * sizeof(HE*) \
368 : (size) * sizeof(HE*) * 2 - MALLOC_OVERHEAD)
371 /* Flags for hv_iternext_flags. */
372 #define HV_ITERNEXT_WANTPLACEHOLDERS 0x01 /* Don't skip placeholders. */
374 #define hv_iternext(hv) hv_iternext_flags(hv, 0)
375 #define hv_magic(hv, gv, how) sv_magic((SV*)(hv), (SV*)(gv), how, NULL, 0)
377 /* available as a function in hv.c */
378 #define Perl_sharepvn(sv, len, hash) HEK_KEY(share_hek(sv, len, hash))
379 #define sharepvn(sv, len, hash) Perl_sharepvn(sv, len, hash)
381 #define share_hek_hek(hek) \
382 (++(((struct shared_he *)(((char *)hek) \
383 - STRUCT_OFFSET(struct shared_he, \
385 ->shared_he_he.he_valu.hent_refcount), \
388 /* This refcounted he structure is used for storing the hints used for lexical
389 pragmas. Without threads, it's basically struct he + refcount.
390 With threads, life gets more complex as the structure needs to be shared
391 between threads (because it hangs from OPs, which are shared), hence the
392 alternate definition and mutex. */
396 /* Gosh. This really isn't a good name any longer. */
397 struct refcounted_he {
398 struct refcounted_he *refcounted_he_next; /* next entry in chain */
400 U32 refcounted_he_hash;
401 U32 refcounted_he_keylen;
403 HEK *refcounted_he_hek; /* hint key */
405 U32 refcounted_he_refcnt; /* reference count */
407 IV refcounted_he_u_iv;
408 UV refcounted_he_u_uv;
409 STRLEN refcounted_he_u_len;
411 /* First byte is flags. Then NUL-terminated value. Then for ithreads,
412 non-NUL terminated key. */
413 char refcounted_he_data[1];
416 /* Flag bits are HVhek_UTF8, HVhek_WASUTF8, then */
417 #define HVrhek_undef 0x00 /* Value is undef. */
418 #define HVrhek_PV 0x10 /* Value is a string. */
419 #define HVrhek_IV 0x20 /* Value is IV/UV. */
420 #define HVrhek_delete 0x30 /* Value is placeholder - signifies delete. */
421 #define HVrhek_typemask 0x30
422 #define HVrhek_UTF8 0x40 /* string value is utf8. */
423 #define HVrhek_UV 0x40 /* integer value is UV. */
426 # define HINTS_REFCNT_LOCK MUTEX_LOCK(&PL_hints_mutex)
427 # define HINTS_REFCNT_UNLOCK MUTEX_UNLOCK(&PL_hints_mutex)
429 # define HINTS_REFCNT_LOCK NOOP
430 # define HINTS_REFCNT_UNLOCK NOOP
435 # define HINTS_REFCNT_INIT MUTEX_INIT(&PL_hints_mutex)
436 # define HINTS_REFCNT_TERM MUTEX_DESTROY(&PL_hints_mutex)
438 # define HINTS_REFCNT_INIT NOOP
439 # define HINTS_REFCNT_TERM NOOP
444 * c-indentation-style: bsd
446 * indent-tabs-mode: t
449 * ex: set ts=8 sts=4 sw=4 noet: