This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Upgrade to CPAN-1.86
[perl5.git] / hv.h
CommitLineData
a0d0e21e 1/* hv.h
79072805 2 *
4bb101f2 3 * Copyright (C) 1991, 1992, 1993, 1996, 1997, 1998, 1999,
1d325971 4 * 2000, 2001, 2002, 2005, by Larry Wall and others
79072805
LW
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 *
79072805
LW
9 */
10
5cbe4eec 11/* entry in hash value chain */
79072805 12struct he {
6f51351d
NC
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
15 body arenas */
5cbe4eec
MF
16 HE *hent_next; /* next entry in chain */
17 HEK *hent_hek; /* hash key */
de616631
NC
18 union {
19 SV *hent_val; /* scalar value that was hashed */
20 Size_t hent_refcount; /* references for this shared hash key */
21 } he_valu;
bbce6d69
PP
22};
23
5cbe4eec 24/* hash key -- defined separately for use as shared pointer */
ff68c719 25struct hek {
5cbe4eec
MF
26 U32 hek_hash; /* hash of key */
27 I32 hek_len; /* length of hash key */
28 char hek_key[1]; /* variable-length hash key */
e05949c7 29 /* the hash-key is \0-terminated */
1e54db1a
JH
30 /* after the \0 there is a byte for flags, such as whether the key
31 is UTF-8 */
79072805
LW
32};
33
cbae3960
NC
34struct shared_he {
35 struct he shared_he_he;
36 struct hek shared_he_hek;
37};
bfcb3514
NC
38
39/* Subject to change.
40 Don't access this directly.
41*/
42struct xpvhv_aux {
7423f6db 43 HEK *xhv_name; /* name, if a symbol table */
86f55936 44 AV *xhv_backreferences; /* back references for weak references */
bfcb3514
NC
45 HE *xhv_eiter; /* current entry of iterator */
46 I32 xhv_riter; /* current root of iterator */
47};
48
5cbe4eec 49/* hash structure: */
6ee623d5 50/* This structure must match the beginning of struct xpvmg in sv.h. */
79072805 51struct xpvhv {
311a25d9 52 NV xnv_nv; /* numeric value, if any */
463ee0b2
LW
53 STRLEN xhv_fill; /* how full xhv_array currently is */
54 STRLEN xhv_max; /* subscript of last element of xhv_array */
e4305a63 55 union {
311a25d9
NC
56 IV xivu_iv; /* integer value or pv offset */
57 UV xivu_uv;
58 void * xivu_p1;
59 } xiv_u;
79072805
LW
60 MAGIC* xmg_magic; /* magic for scalar array */
61 HV* xmg_stash; /* class package */
79072805
LW
62};
63
311a25d9 64#define xhv_keys xiv_u.xivu_iv
e4305a63 65
c5f87f51 66#if 0
59813432 67typedef struct xpvhv xpvhv_allocated;
9f1501b2
NC
68#else
69typedef struct {
70 STRLEN xhv_fill; /* how full xhv_array currently is */
71 STRLEN xhv_max; /* subscript of last element of xhv_array */
72 union {
311a25d9
NC
73 IV xivu_iv; /* integer value or pv offset */
74 UV xivu_uv;
75 void * xivu_p1;
76 } xiv_u;
9f1501b2
NC
77 MAGIC* xmg_magic; /* magic for scalar array */
78 HV* xmg_stash; /* class package */
79} xpvhv_allocated;
80#endif
e4305a63 81
5cbe4eec 82/* hash a key */
5afd6d42
JH
83/* FYI: This is the "One-at-a-Time" algorithm by Bob Jenkins
84 * from requirements by Colin Plumb.
85 * (http://burtleburtle.net/bob/hash/doobs.html) */
86/* The use of a temporary pointer and the casting games
87 * is needed to serve the dual purposes of
88 * (a) the hashed data being interpreted as "unsigned char" (new since 5.8,
df805c3e 89 * a "char" can be either signed or unsigned, depending on the compiler)
5afd6d42 90 * (b) catering for old code that uses a "char"
830b38bd 91 *
504f80c1
JH
92 * The "hash seed" feature was added in Perl 5.8.1 to perturb the results
93 * to avoid "algorithmic complexity attacks".
830b38bd
JH
94 *
95 * If USE_HASH_SEED is defined, hash randomisation is done by default
96 * If USE_HASH_SEED_EXPLICIT is defined, hash randomisation is done
97 * only if the environment variable PERL_HASH_SEED is set.
98 * For maximal control, one can define PERL_HASH_SEED.
90ce55a1 99 * (see also perl.c:perl_parse()).
5afd6d42 100 */
830b38bd
JH
101#ifndef PERL_HASH_SEED
102# if defined(USE_HASH_SEED) || defined(USE_HASH_SEED_EXPLICIT)
103# define PERL_HASH_SEED PL_hash_seed
104# else
105# define PERL_HASH_SEED 0
106# endif
504f80c1 107#endif
bf6bd887
PP
108#define PERL_HASH(hash,str,len) \
109 STMT_START { \
0bd48802 110 register const char * const s_PeRlHaSh_tmp = str; \
5afd6d42 111 register const unsigned char *s_PeRlHaSh = (const unsigned char *)s_PeRlHaSh_tmp; \
bf6bd887 112 register I32 i_PeRlHaSh = len; \
504f80c1 113 register U32 hash_PeRlHaSh = PERL_HASH_SEED; \
a6fe520e
MF
114 while (i_PeRlHaSh--) { \
115 hash_PeRlHaSh += *s_PeRlHaSh++; \
116 hash_PeRlHaSh += (hash_PeRlHaSh << 10); \
117 hash_PeRlHaSh ^= (hash_PeRlHaSh >> 6); \
118 } \
119 hash_PeRlHaSh += (hash_PeRlHaSh << 3); \
120 hash_PeRlHaSh ^= (hash_PeRlHaSh >> 11); \
4fee5c71 121 (hash) = (hash_PeRlHaSh + (hash_PeRlHaSh << 15)); \
bf6bd887
PP
122 } STMT_END
123
3d78eb94
NC
124/* Only hv.c and mod_perl should be doing this. */
125#ifdef PERL_HASH_INTERNAL_ACCESS
4b5190b5
NC
126#define PERL_HASH_INTERNAL(hash,str,len) \
127 STMT_START { \
0bd48802 128 register const char * const s_PeRlHaSh_tmp = str; \
4b5190b5
NC
129 register const unsigned char *s_PeRlHaSh = (const unsigned char *)s_PeRlHaSh_tmp; \
130 register I32 i_PeRlHaSh = len; \
008fb0c0 131 register U32 hash_PeRlHaSh = PL_rehash_seed; \
4b5190b5
NC
132 while (i_PeRlHaSh--) { \
133 hash_PeRlHaSh += *s_PeRlHaSh++; \
134 hash_PeRlHaSh += (hash_PeRlHaSh << 10); \
135 hash_PeRlHaSh ^= (hash_PeRlHaSh >> 6); \
136 } \
137 hash_PeRlHaSh += (hash_PeRlHaSh << 3); \
138 hash_PeRlHaSh ^= (hash_PeRlHaSh >> 11); \
139 (hash) = (hash_PeRlHaSh + (hash_PeRlHaSh << 15)); \
140 } STMT_END
141#endif
142
954c1994 143/*
ccfc67b7
JH
144=head1 Hash Manipulation Functions
145
954c1994
GS
146=for apidoc AmU||HEf_SVKEY
147This flag, used in the length slot of hash entries and magic structures,
d1be9408 148specifies the structure contains an C<SV*> pointer where a C<char*> pointer
954c1994
GS
149is to be expected. (For information only--not to be used).
150
ccfc67b7
JH
151=head1 Handy Values
152
954c1994
GS
153=for apidoc AmU||Nullhv
154Null HV pointer.
155
ccfc67b7
JH
156=head1 Hash Manipulation Functions
157
954c1994 158=for apidoc Am|char*|HvNAME|HV* stash
9282b5fd
SH
159Returns the package name of a stash, or NULL if C<stash> isn't a stash.
160See C<SvSTASH>, C<CvSTASH>.
954c1994
GS
161
162=for apidoc Am|void*|HeKEY|HE* he
163Returns the actual pointer stored in the key slot of the hash entry. The
164pointer may be either C<char*> or C<SV*>, depending on the value of
165C<HeKLEN()>. Can be assigned to. The C<HePV()> or C<HeSVKEY()> macros are
166usually preferable for finding the value of a key.
167
168=for apidoc Am|STRLEN|HeKLEN|HE* he
169If this is negative, and amounts to C<HEf_SVKEY>, it indicates the entry
170holds an C<SV*> key. Otherwise, holds the actual length of the key. Can
171be assigned to. The C<HePV()> macro is usually preferable for finding key
172lengths.
173
174=for apidoc Am|SV*|HeVAL|HE* he
175Returns the value slot (type C<SV*>) stored in the hash entry.
176
177=for apidoc Am|U32|HeHASH|HE* he
178Returns the computed hash stored in the hash entry.
179
180=for apidoc Am|char*|HePV|HE* he|STRLEN len
181Returns the key slot of the hash entry as a C<char*> value, doing any
182necessary dereferencing of possibly C<SV*> keys. The length of the string
183is placed in C<len> (this is a macro, so do I<not> use C<&len>). If you do
184not care about what the length of the key is, you may use the global
185variable C<PL_na>, though this is rather less efficient than using a local
186variable. Remember though, that hash keys in perl are free to contain
187embedded nulls, so using C<strlen()> or similar is not a good way to find
188the length of hash keys. This is very similar to the C<SvPV()> macro
189described elsewhere in this document.
190
191=for apidoc Am|SV*|HeSVKEY|HE* he
a0714e2c 192Returns the key as an C<SV*>, or C<NULL> if the hash entry does not
954c1994
GS
193contain an C<SV*> key.
194
195=for apidoc Am|SV*|HeSVKEY_force|HE* he
196Returns the key as an C<SV*>. Will create and return a temporary mortal
197C<SV*> if the hash entry contains only a C<char*> key.
198
199=for apidoc Am|SV*|HeSVKEY_set|HE* he|SV* sv
200Sets the key to a given C<SV*>, taking care to set the appropriate flags to
201indicate the presence of an C<SV*> key, and returns the same
202C<SV*>.
203
204=cut
205*/
bf6bd887 206
bf5b86ae 207/* these hash entry flags ride on hent_klen (for use only in magic/tied HVs) */
d1be9408 208#define HEf_SVKEY -2 /* hent_key is an SV* */
bf6bd887
PP
209
210
79072805 211#define Nullhv Null(HV*)
43e6e717 212#define HvARRAY(hv) ((hv)->sv_u.svu_hash)
79072805 213#define HvFILL(hv) ((XPVHV*) SvANY(hv))->xhv_fill
463ee0b2 214#define HvMAX(hv) ((XPVHV*) SvANY(hv))->xhv_max
b79f7545
NC
215/* This quite intentionally does no flag checking first. That's your
216 responsibility. */
217#define HvAUX(hv) ((struct xpvhv_aux*)&(HvARRAY(hv)[HvMAX(hv)+1]))
bfcb3514
NC
218#define HvRITER(hv) (*Perl_hv_riter_p(aTHX_ (HV*)(hv)))
219#define HvEITER(hv) (*Perl_hv_eiter_p(aTHX_ (HV*)(hv)))
220#define HvRITER_set(hv,r) Perl_hv_riter_set(aTHX_ (HV*)(hv), r)
221#define HvEITER_set(hv,e) Perl_hv_eiter_set(aTHX_ (HV*)(hv), e)
b79f7545
NC
222#define HvRITER_get(hv) (SvOOK(hv) ? HvAUX(hv)->xhv_riter : -1)
223#define HvEITER_get(hv) (SvOOK(hv) ? HvAUX(hv)->xhv_eiter : 0)
7423f6db 224#define HvNAME(hv) HvNAME_get(hv)
bfcb3514
NC
225/* FIXME - all of these should use a UTF8 aware API, which should also involve
226 getting the length. */
9b9d0b15 227/* This macro may go away without notice. */
b79f7545
NC
228#define HvNAME_HEK(hv) (SvOOK(hv) ? HvAUX(hv)->xhv_name : 0)
229#define HvNAME_get(hv) ((SvOOK(hv) && (HvAUX(hv)->xhv_name)) \
230 ? HEK_KEY(HvAUX(hv)->xhv_name) : 0)
231#define HvNAMELEN_get(hv) ((SvOOK(hv) && (HvAUX(hv)->xhv_name)) \
232 ? HEK_LEN(HvAUX(hv)->xhv_name) : 0)
a0d0e21e 233
8aacddc1
NIS
234/* the number of keys (including any placeholers) */
235#define XHvTOTALKEYS(xhv) ((xhv)->xhv_keys)
236
8aacddc1
NIS
237/*
238 * HvKEYS gets the number of keys that actually exist(), and is provided
239 * for backwards compatibility with old XS code. The core uses HvUSEDKEYS
240 * (keys, excluding placeholdes) and HvTOTALKEYS (including placeholders)
241 */
ca732855
NC
242#define HvKEYS(hv) HvUSEDKEYS(hv)
243#define HvUSEDKEYS(hv) (HvTOTALKEYS(hv) - HvPLACEHOLDERS_get(hv))
8aacddc1 244#define HvTOTALKEYS(hv) XHvTOTALKEYS((XPVHV*) SvANY(hv))
ca732855
NC
245#define HvPLACEHOLDERS(hv) (*Perl_hv_placeholders_p(aTHX_ (HV*)hv))
246#define HvPLACEHOLDERS_get(hv) (SvMAGIC(hv) ? Perl_hv_placeholders_get(aTHX_ (HV*)hv) : 0)
247#define HvPLACEHOLDERS_set(hv,p) Perl_hv_placeholders_set(aTHX_ (HV*)hv, p)
8aacddc1 248
bf6bd887
PP
249#define HvSHAREKEYS(hv) (SvFLAGS(hv) & SVphv_SHAREKEYS)
250#define HvSHAREKEYS_on(hv) (SvFLAGS(hv) |= SVphv_SHAREKEYS)
251#define HvSHAREKEYS_off(hv) (SvFLAGS(hv) &= ~SVphv_SHAREKEYS)
252
19692e8d
NC
253/* This is an optimisation flag. It won't be set if all hash keys have a 0
254 * flag. Currently the only flags relate to utf8.
255 * Hence it won't be set if all keys are 8 bit only. It will be set if any key
256 * is utf8 (including 8 bit keys that were entered as utf8, and need upgrading
257 * when retrieved during iteration. It may still be set when there are no longer
258 * any utf8 keys.
fdcd69b6 259 * See HVhek_ENABLEHVKFLAGS for the trigger.
19692e8d
NC
260 */
261#define HvHASKFLAGS(hv) (SvFLAGS(hv) & SVphv_HASKFLAGS)
262#define HvHASKFLAGS_on(hv) (SvFLAGS(hv) |= SVphv_HASKFLAGS)
263#define HvHASKFLAGS_off(hv) (SvFLAGS(hv) &= ~SVphv_HASKFLAGS)
574c8022 264
bf5b86ae
GS
265#define HvLAZYDEL(hv) (SvFLAGS(hv) & SVphv_LAZYDEL)
266#define HvLAZYDEL_on(hv) (SvFLAGS(hv) |= SVphv_LAZYDEL)
267#define HvLAZYDEL_off(hv) (SvFLAGS(hv) &= ~SVphv_LAZYDEL)
268
4b5190b5
NC
269#define HvREHASH(hv) (SvFLAGS(hv) & SVphv_REHASH)
270#define HvREHASH_on(hv) (SvFLAGS(hv) |= SVphv_REHASH)
271#define HvREHASH_off(hv) (SvFLAGS(hv) &= ~SVphv_REHASH)
272
a0d0e21e
LW
273/* Maybe amagical: */
274/* #define HV_AMAGICmb(hv) (SvFLAGS(hv) & (SVpgv_badAM | SVpgv_AM)) */
275
276#define HV_AMAGIC(hv) (SvFLAGS(hv) & SVpgv_AM)
277#define HV_AMAGIC_on(hv) (SvFLAGS(hv) |= SVpgv_AM)
278#define HV_AMAGIC_off(hv) (SvFLAGS(hv) &= ~SVpgv_AM)
279
280/*
281#define HV_AMAGICbad(hv) (SvFLAGS(hv) & SVpgv_badAM)
282#define HV_badAMAGIC_on(hv) (SvFLAGS(hv) |= SVpgv_badAM)
283#define HV_badAMAGIC_off(hv) (SvFLAGS(hv) &= ~SVpgv_badAM)
284*/
bf6bd887
PP
285
286#define Nullhe Null(HE*)
287#define HeNEXT(he) (he)->hent_next
ff68c719
PP
288#define HeKEY_hek(he) (he)->hent_hek
289#define HeKEY(he) HEK_KEY(HeKEY_hek(he))
bbce6d69 290#define HeKEY_sv(he) (*(SV**)HeKEY(he))
ff68c719 291#define HeKLEN(he) HEK_LEN(HeKEY_hek(he))
da58a35d 292#define HeKUTF8(he) HEK_UTF8(HeKEY_hek(he))
19692e8d 293#define HeKWASUTF8(he) HEK_WASUTF8(HeKEY_hek(he))
4b5190b5 294#define HeKREHASH(he) HEK_REHASH(HeKEY_hek(he))
da58a35d 295#define HeKLEN_UTF8(he) (HeKUTF8(he) ? -HeKLEN(he) : HeKLEN(he))
19692e8d 296#define HeKFLAGS(he) HEK_FLAGS(HeKEY_hek(he))
de616631 297#define HeVAL(he) (he)->he_valu.hent_val
ff68c719 298#define HeHASH(he) HEK_HASH(HeKEY_hek(he))
1e422769
PP
299#define HePV(he,lp) ((HeKLEN(he) == HEf_SVKEY) ? \
300 SvPV(HeKEY_sv(he),lp) : \
301 (((lp = HeKLEN(he)) >= 0) ? \
bd61b366 302 HeKEY(he) : NULL))
1e422769 303
bbce6d69
PP
304#define HeSVKEY(he) ((HeKEY(he) && \
305 HeKLEN(he) == HEf_SVKEY) ? \
a0714e2c 306 HeKEY_sv(he) : NULL)
bbce6d69
PP
307
308#define HeSVKEY_force(he) (HeKEY(he) ? \
309 ((HeKLEN(he) == HEf_SVKEY) ? \
310 HeKEY_sv(he) : \
79cb57f6 311 sv_2mortal(newSVpvn(HeKEY(he), \
bbce6d69 312 HeKLEN(he)))) : \
3280af22 313 &PL_sv_undef)
1e422769 314#define HeSVKEY_set(he,sv) ((HeKLEN(he) = HEf_SVKEY), (HeKEY_sv(he) = sv))
bbce6d69 315
ff68c719 316#define Nullhek Null(HEK*)
71be2cbc 317#define HEK_BASESIZE STRUCT_OFFSET(HEK, hek_key[0])
ff68c719
PP
318#define HEK_HASH(hek) (hek)->hek_hash
319#define HEK_LEN(hek) (hek)->hek_len
320#define HEK_KEY(hek) (hek)->hek_key
19692e8d
NC
321#define HEK_FLAGS(hek) (*((unsigned char *)(HEK_KEY(hek))+HEK_LEN(hek)+1))
322
323#define HVhek_UTF8 0x01 /* Key is utf8 encoded. */
324#define HVhek_WASUTF8 0x02 /* Key is bytes here, but was supplied as utf8. */
4b5190b5 325#define HVhek_REHASH 0x04 /* This key is in an hv using a custom HASH . */
19692e8d 326#define HVhek_FREEKEY 0x100 /* Internal flag to say key is malloc()ed. */
e16e2ff8
NC
327#define HVhek_PLACEHOLD 0x200 /* Internal flag to create placeholder.
328 * (may change, but Storable is a core module) */
19692e8d
NC
329#define HVhek_MASK 0xFF
330
fdcd69b6
NC
331/* Which flags enable HvHASKFLAGS? Somewhat a hack on a hack, as
332 HVhek_REHASH is only needed because the rehash flag has to be duplicated
333 into all keys as hv_iternext has no access to the hash flags. At this
334 point Storable's tests get upset, because sometimes hashes are "keyed"
335 and sometimes not, depending on the order of data insertion, and whether
336 it triggered rehashing. So currently HVhek_REHAS is exempt.
337*/
338
339#define HVhek_ENABLEHVKFLAGS (HVhek_MASK - HVhek_REHASH)
340
19692e8d
NC
341#define HEK_UTF8(hek) (HEK_FLAGS(hek) & HVhek_UTF8)
342#define HEK_UTF8_on(hek) (HEK_FLAGS(hek) |= HVhek_UTF8)
343#define HEK_UTF8_off(hek) (HEK_FLAGS(hek) &= ~HVhek_UTF8)
344#define HEK_WASUTF8(hek) (HEK_FLAGS(hek) & HVhek_WASUTF8)
345#define HEK_WASUTF8_on(hek) (HEK_FLAGS(hek) |= HVhek_WASUTF8)
346#define HEK_WASUTF8_off(hek) (HEK_FLAGS(hek) &= ~HVhek_WASUTF8)
4b5190b5
NC
347#define HEK_REHASH(hek) (HEK_FLAGS(hek) & HVhek_REHASH)
348#define HEK_REHASH_on(hek) (HEK_FLAGS(hek) |= HVhek_REHASH)
d18c6117 349
5cbe4eec 350/* calculate HV array allocation */
36768cf4
NC
351#ifndef PERL_USE_LARGE_HV_ALLOC
352/* Default to allocating the correct size - default to assuming that malloc()
353 is not broken and is efficient at allocating blocks sized at powers-of-two.
354*/
d18c6117
GS
355# define PERL_HV_ARRAY_ALLOC_BYTES(size) ((size) * sizeof(HE*))
356#else
357# define MALLOC_OVERHEAD 16
358# define PERL_HV_ARRAY_ALLOC_BYTES(size) \
359 (((size) < 64) \
360 ? (size) * sizeof(HE*) \
361 : (size) * sizeof(HE*) * 2 - MALLOC_OVERHEAD)
362#endif
37d85e3a 363
e16e2ff8
NC
364/* Flags for hv_iternext_flags. */
365#define HV_ITERNEXT_WANTPLACEHOLDERS 0x01 /* Don't skip placeholders. */
366
7a7b9979 367#define hv_iternext(hv) hv_iternext_flags(hv, 0)
bd61b366 368#define hv_magic(hv, gv, how) sv_magic((SV*)(hv), (SV*)(gv), how, NULL, 0)
7a7b9979 369
37d85e3a
JH
370/* available as a function in hv.c */
371#define Perl_sharepvn(sv, len, hash) HEK_KEY(share_hek(sv, len, hash))
372#define sharepvn(sv, len, hash) Perl_sharepvn(sv, len, hash)
bfcb3514 373
d1db91c6
NC
374#define share_hek_hek(hek) \
375 (++(((struct shared_he *)(((char *)hek) \
376 - STRUCT_OFFSET(struct shared_he, \
377 shared_he_hek))) \
de616631 378 ->shared_he_he.he_valu.hent_refcount), \
d1db91c6
NC
379 hek)
380
bfcb3514
NC
381/*
382 * Local variables:
383 * c-indentation-style: bsd
384 * c-basic-offset: 4
385 * indent-tabs-mode: t
386 * End:
387 *
388 * ex: set ts=8 sts=4 sw=4 noet:
389 */