This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Add a union in place of xnv_nv, which allows AVs and HVs to re-use
[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/* typedefs to eliminate some typing */
79072805 12typedef struct he HE;
ff68c719 13typedef struct hek HEK;
79072805 14
5cbe4eec 15/* entry in hash value chain */
79072805 16struct he {
5cbe4eec
MF
17 HE *hent_next; /* next entry in chain */
18 HEK *hent_hek; /* hash key */
19 SV *hent_val; /* scalar value that was hashed */
bbce6d69
PP
20};
21
5cbe4eec 22/* hash key -- defined separately for use as shared pointer */
ff68c719 23struct hek {
5cbe4eec
MF
24 U32 hek_hash; /* hash of key */
25 I32 hek_len; /* length of hash key */
26 char hek_key[1]; /* variable-length hash key */
e05949c7 27 /* the hash-key is \0-terminated */
1e54db1a
JH
28 /* after the \0 there is a byte for flags, such as whether the key
29 is UTF-8 */
79072805
LW
30};
31
bfcb3514
NC
32
33/* Subject to change.
34 Don't access this directly.
35*/
36struct xpvhv_aux {
37 char *xhv_name; /* name, if a symbol table */
38 HE *xhv_eiter; /* current entry of iterator */
39 I32 xhv_riter; /* current root of iterator */
40};
41
5cbe4eec 42/* hash structure: */
6ee623d5 43/* This structure must match the beginning of struct xpvmg in sv.h. */
79072805 44struct xpvhv {
463ee0b2
LW
45 STRLEN xhv_fill; /* how full xhv_array currently is */
46 STRLEN xhv_max; /* subscript of last element of xhv_array */
e4305a63
NC
47 IV for_rent;
48 union {
49 NV xnvu_nv; /* numeric value, if any */
50 struct {
51 void *xnv_p1;
52 union {
53 void *xnv_p2;
54 IV xnv_i2; /* how many elements in the array */
55 } xnv_u2;
56 } xnv_s;
57 } xnv_u;
79072805
LW
58 MAGIC* xmg_magic; /* magic for scalar array */
59 HV* xmg_stash; /* class package */
8d2f4536 60 /* list of pm's for this package is now stored in symtab magic. */
79072805
LW
61};
62
e4305a63
NC
63#define xhv_aux xnv_u.xnv_s.xnv_p1
64#define xhv_keys xnv_u.xnv_s.xnv_u2.xnv_i2
65
66
5cbe4eec 67/* hash a key */
5afd6d42
JH
68/* FYI: This is the "One-at-a-Time" algorithm by Bob Jenkins
69 * from requirements by Colin Plumb.
70 * (http://burtleburtle.net/bob/hash/doobs.html) */
71/* The use of a temporary pointer and the casting games
72 * is needed to serve the dual purposes of
73 * (a) the hashed data being interpreted as "unsigned char" (new since 5.8,
74 * a "char" can be either signed or signed, depending on the compiler)
75 * (b) catering for old code that uses a "char"
830b38bd 76 *
504f80c1
JH
77 * The "hash seed" feature was added in Perl 5.8.1 to perturb the results
78 * to avoid "algorithmic complexity attacks".
830b38bd
JH
79 *
80 * If USE_HASH_SEED is defined, hash randomisation is done by default
81 * If USE_HASH_SEED_EXPLICIT is defined, hash randomisation is done
82 * only if the environment variable PERL_HASH_SEED is set.
83 * For maximal control, one can define PERL_HASH_SEED.
90ce55a1 84 * (see also perl.c:perl_parse()).
5afd6d42 85 */
830b38bd
JH
86#ifndef PERL_HASH_SEED
87# if defined(USE_HASH_SEED) || defined(USE_HASH_SEED_EXPLICIT)
88# define PERL_HASH_SEED PL_hash_seed
89# else
90# define PERL_HASH_SEED 0
91# endif
504f80c1 92#endif
bf6bd887
PP
93#define PERL_HASH(hash,str,len) \
94 STMT_START { \
5afd6d42
JH
95 register const char *s_PeRlHaSh_tmp = str; \
96 register const unsigned char *s_PeRlHaSh = (const unsigned char *)s_PeRlHaSh_tmp; \
bf6bd887 97 register I32 i_PeRlHaSh = len; \
504f80c1 98 register U32 hash_PeRlHaSh = PERL_HASH_SEED; \
a6fe520e
MF
99 while (i_PeRlHaSh--) { \
100 hash_PeRlHaSh += *s_PeRlHaSh++; \
101 hash_PeRlHaSh += (hash_PeRlHaSh << 10); \
102 hash_PeRlHaSh ^= (hash_PeRlHaSh >> 6); \
103 } \
104 hash_PeRlHaSh += (hash_PeRlHaSh << 3); \
105 hash_PeRlHaSh ^= (hash_PeRlHaSh >> 11); \
4fee5c71 106 (hash) = (hash_PeRlHaSh + (hash_PeRlHaSh << 15)); \
bf6bd887
PP
107 } STMT_END
108
3d78eb94
NC
109/* Only hv.c and mod_perl should be doing this. */
110#ifdef PERL_HASH_INTERNAL_ACCESS
4b5190b5
NC
111#define PERL_HASH_INTERNAL(hash,str,len) \
112 STMT_START { \
113 register const char *s_PeRlHaSh_tmp = str; \
114 register const unsigned char *s_PeRlHaSh = (const unsigned char *)s_PeRlHaSh_tmp; \
115 register I32 i_PeRlHaSh = len; \
008fb0c0 116 register U32 hash_PeRlHaSh = PL_rehash_seed; \
4b5190b5
NC
117 while (i_PeRlHaSh--) { \
118 hash_PeRlHaSh += *s_PeRlHaSh++; \
119 hash_PeRlHaSh += (hash_PeRlHaSh << 10); \
120 hash_PeRlHaSh ^= (hash_PeRlHaSh >> 6); \
121 } \
122 hash_PeRlHaSh += (hash_PeRlHaSh << 3); \
123 hash_PeRlHaSh ^= (hash_PeRlHaSh >> 11); \
124 (hash) = (hash_PeRlHaSh + (hash_PeRlHaSh << 15)); \
125 } STMT_END
126#endif
127
954c1994 128/*
ccfc67b7
JH
129=head1 Hash Manipulation Functions
130
954c1994
GS
131=for apidoc AmU||HEf_SVKEY
132This flag, used in the length slot of hash entries and magic structures,
d1be9408 133specifies the structure contains an C<SV*> pointer where a C<char*> pointer
954c1994
GS
134is to be expected. (For information only--not to be used).
135
ccfc67b7
JH
136=head1 Handy Values
137
954c1994
GS
138=for apidoc AmU||Nullhv
139Null HV pointer.
140
ccfc67b7
JH
141=head1 Hash Manipulation Functions
142
954c1994
GS
143=for apidoc Am|char*|HvNAME|HV* stash
144Returns the package name of a stash. See C<SvSTASH>, C<CvSTASH>.
145
146=for apidoc Am|void*|HeKEY|HE* he
147Returns the actual pointer stored in the key slot of the hash entry. The
148pointer may be either C<char*> or C<SV*>, depending on the value of
149C<HeKLEN()>. Can be assigned to. The C<HePV()> or C<HeSVKEY()> macros are
150usually preferable for finding the value of a key.
151
152=for apidoc Am|STRLEN|HeKLEN|HE* he
153If this is negative, and amounts to C<HEf_SVKEY>, it indicates the entry
154holds an C<SV*> key. Otherwise, holds the actual length of the key. Can
155be assigned to. The C<HePV()> macro is usually preferable for finding key
156lengths.
157
158=for apidoc Am|SV*|HeVAL|HE* he
159Returns the value slot (type C<SV*>) stored in the hash entry.
160
161=for apidoc Am|U32|HeHASH|HE* he
162Returns the computed hash stored in the hash entry.
163
164=for apidoc Am|char*|HePV|HE* he|STRLEN len
165Returns the key slot of the hash entry as a C<char*> value, doing any
166necessary dereferencing of possibly C<SV*> keys. The length of the string
167is placed in C<len> (this is a macro, so do I<not> use C<&len>). If you do
168not care about what the length of the key is, you may use the global
169variable C<PL_na>, though this is rather less efficient than using a local
170variable. Remember though, that hash keys in perl are free to contain
171embedded nulls, so using C<strlen()> or similar is not a good way to find
172the length of hash keys. This is very similar to the C<SvPV()> macro
173described elsewhere in this document.
174
175=for apidoc Am|SV*|HeSVKEY|HE* he
176Returns the key as an C<SV*>, or C<Nullsv> if the hash entry does not
177contain an C<SV*> key.
178
179=for apidoc Am|SV*|HeSVKEY_force|HE* he
180Returns the key as an C<SV*>. Will create and return a temporary mortal
181C<SV*> if the hash entry contains only a C<char*> key.
182
183=for apidoc Am|SV*|HeSVKEY_set|HE* he|SV* sv
184Sets the key to a given C<SV*>, taking care to set the appropriate flags to
185indicate the presence of an C<SV*> key, and returns the same
186C<SV*>.
187
188=cut
189*/
bf6bd887 190
bf5b86ae 191/* these hash entry flags ride on hent_klen (for use only in magic/tied HVs) */
d1be9408 192#define HEf_SVKEY -2 /* hent_key is an SV* */
bf6bd887
PP
193
194
79072805 195#define Nullhv Null(HV*)
7b2c381c 196#define HvARRAY(hv) (*(HE***)&((hv)->sv_u.sv_array))
79072805 197#define HvFILL(hv) ((XPVHV*) SvANY(hv))->xhv_fill
463ee0b2 198#define HvMAX(hv) ((XPVHV*) SvANY(hv))->xhv_max
bfcb3514
NC
199#define HvRITER(hv) (*Perl_hv_riter_p(aTHX_ (HV*)(hv)))
200#define HvEITER(hv) (*Perl_hv_eiter_p(aTHX_ (HV*)(hv)))
201#define HvRITER_set(hv,r) Perl_hv_riter_set(aTHX_ (HV*)(hv), r)
202#define HvEITER_set(hv,e) Perl_hv_eiter_set(aTHX_ (HV*)(hv), e)
203#define HvRITER_get(hv) (((XPVHV *)SvANY(hv))->xhv_aux ? \
204 ((struct xpvhv_aux*)((XPVHV *)SvANY(hv))->xhv_aux)->xhv_riter : -1)
205#define HvEITER_get(hv) (((XPVHV *)SvANY(hv))->xhv_aux ? \
206 ((struct xpvhv_aux *)((XPVHV *)SvANY(hv))->xhv_aux)->xhv_eiter : 0)
207#define HvNAME(hv) (*Perl_hv_name_p(aTHX_ (HV*)hv))
208/* FIXME - all of these should use a UTF8 aware API, which should also involve
209 getting the length. */
210#define HvNAME_get(hv) (((XPVHV *)SvANY(hv))->xhv_aux ? \
211 ((struct xpvhv_aux *)((XPVHV *)SvANY(hv))->xhv_aux)->xhv_name : 0)
a0d0e21e 212
8aacddc1
NIS
213/* the number of keys (including any placeholers) */
214#define XHvTOTALKEYS(xhv) ((xhv)->xhv_keys)
215
8aacddc1
NIS
216/*
217 * HvKEYS gets the number of keys that actually exist(), and is provided
218 * for backwards compatibility with old XS code. The core uses HvUSEDKEYS
219 * (keys, excluding placeholdes) and HvTOTALKEYS (including placeholders)
220 */
ca732855
NC
221#define HvKEYS(hv) HvUSEDKEYS(hv)
222#define HvUSEDKEYS(hv) (HvTOTALKEYS(hv) - HvPLACEHOLDERS_get(hv))
8aacddc1 223#define HvTOTALKEYS(hv) XHvTOTALKEYS((XPVHV*) SvANY(hv))
ca732855
NC
224#define HvPLACEHOLDERS(hv) (*Perl_hv_placeholders_p(aTHX_ (HV*)hv))
225#define HvPLACEHOLDERS_get(hv) (SvMAGIC(hv) ? Perl_hv_placeholders_get(aTHX_ (HV*)hv) : 0)
226#define HvPLACEHOLDERS_set(hv,p) Perl_hv_placeholders_set(aTHX_ (HV*)hv, p)
8aacddc1 227
bf6bd887
PP
228#define HvSHAREKEYS(hv) (SvFLAGS(hv) & SVphv_SHAREKEYS)
229#define HvSHAREKEYS_on(hv) (SvFLAGS(hv) |= SVphv_SHAREKEYS)
230#define HvSHAREKEYS_off(hv) (SvFLAGS(hv) &= ~SVphv_SHAREKEYS)
231
19692e8d
NC
232/* This is an optimisation flag. It won't be set if all hash keys have a 0
233 * flag. Currently the only flags relate to utf8.
234 * Hence it won't be set if all keys are 8 bit only. It will be set if any key
235 * is utf8 (including 8 bit keys that were entered as utf8, and need upgrading
236 * when retrieved during iteration. It may still be set when there are no longer
237 * any utf8 keys.
fdcd69b6 238 * See HVhek_ENABLEHVKFLAGS for the trigger.
19692e8d
NC
239 */
240#define HvHASKFLAGS(hv) (SvFLAGS(hv) & SVphv_HASKFLAGS)
241#define HvHASKFLAGS_on(hv) (SvFLAGS(hv) |= SVphv_HASKFLAGS)
242#define HvHASKFLAGS_off(hv) (SvFLAGS(hv) &= ~SVphv_HASKFLAGS)
574c8022 243
bf5b86ae
GS
244#define HvLAZYDEL(hv) (SvFLAGS(hv) & SVphv_LAZYDEL)
245#define HvLAZYDEL_on(hv) (SvFLAGS(hv) |= SVphv_LAZYDEL)
246#define HvLAZYDEL_off(hv) (SvFLAGS(hv) &= ~SVphv_LAZYDEL)
247
4b5190b5
NC
248#define HvREHASH(hv) (SvFLAGS(hv) & SVphv_REHASH)
249#define HvREHASH_on(hv) (SvFLAGS(hv) |= SVphv_REHASH)
250#define HvREHASH_off(hv) (SvFLAGS(hv) &= ~SVphv_REHASH)
251
a0d0e21e
LW
252/* Maybe amagical: */
253/* #define HV_AMAGICmb(hv) (SvFLAGS(hv) & (SVpgv_badAM | SVpgv_AM)) */
254
255#define HV_AMAGIC(hv) (SvFLAGS(hv) & SVpgv_AM)
256#define HV_AMAGIC_on(hv) (SvFLAGS(hv) |= SVpgv_AM)
257#define HV_AMAGIC_off(hv) (SvFLAGS(hv) &= ~SVpgv_AM)
258
259/*
260#define HV_AMAGICbad(hv) (SvFLAGS(hv) & SVpgv_badAM)
261#define HV_badAMAGIC_on(hv) (SvFLAGS(hv) |= SVpgv_badAM)
262#define HV_badAMAGIC_off(hv) (SvFLAGS(hv) &= ~SVpgv_badAM)
263*/
bf6bd887
PP
264
265#define Nullhe Null(HE*)
266#define HeNEXT(he) (he)->hent_next
ff68c719
PP
267#define HeKEY_hek(he) (he)->hent_hek
268#define HeKEY(he) HEK_KEY(HeKEY_hek(he))
bbce6d69 269#define HeKEY_sv(he) (*(SV**)HeKEY(he))
ff68c719 270#define HeKLEN(he) HEK_LEN(HeKEY_hek(he))
da58a35d 271#define HeKUTF8(he) HEK_UTF8(HeKEY_hek(he))
19692e8d 272#define HeKWASUTF8(he) HEK_WASUTF8(HeKEY_hek(he))
4b5190b5 273#define HeKREHASH(he) HEK_REHASH(HeKEY_hek(he))
da58a35d 274#define HeKLEN_UTF8(he) (HeKUTF8(he) ? -HeKLEN(he) : HeKLEN(he))
19692e8d 275#define HeKFLAGS(he) HEK_FLAGS(HeKEY_hek(he))
bf6bd887 276#define HeVAL(he) (he)->hent_val
ff68c719 277#define HeHASH(he) HEK_HASH(HeKEY_hek(he))
1e422769
PP
278#define HePV(he,lp) ((HeKLEN(he) == HEf_SVKEY) ? \
279 SvPV(HeKEY_sv(he),lp) : \
280 (((lp = HeKLEN(he)) >= 0) ? \
281 HeKEY(he) : Nullch))
282
bbce6d69
PP
283#define HeSVKEY(he) ((HeKEY(he) && \
284 HeKLEN(he) == HEf_SVKEY) ? \
285 HeKEY_sv(he) : Nullsv)
286
287#define HeSVKEY_force(he) (HeKEY(he) ? \
288 ((HeKLEN(he) == HEf_SVKEY) ? \
289 HeKEY_sv(he) : \
79cb57f6 290 sv_2mortal(newSVpvn(HeKEY(he), \
bbce6d69 291 HeKLEN(he)))) : \
3280af22 292 &PL_sv_undef)
1e422769 293#define HeSVKEY_set(he,sv) ((HeKLEN(he) = HEf_SVKEY), (HeKEY_sv(he) = sv))
bbce6d69 294
ff68c719 295#define Nullhek Null(HEK*)
71be2cbc 296#define HEK_BASESIZE STRUCT_OFFSET(HEK, hek_key[0])
ff68c719
PP
297#define HEK_HASH(hek) (hek)->hek_hash
298#define HEK_LEN(hek) (hek)->hek_len
299#define HEK_KEY(hek) (hek)->hek_key
19692e8d
NC
300#define HEK_FLAGS(hek) (*((unsigned char *)(HEK_KEY(hek))+HEK_LEN(hek)+1))
301
302#define HVhek_UTF8 0x01 /* Key is utf8 encoded. */
303#define HVhek_WASUTF8 0x02 /* Key is bytes here, but was supplied as utf8. */
4b5190b5 304#define HVhek_REHASH 0x04 /* This key is in an hv using a custom HASH . */
19692e8d 305#define HVhek_FREEKEY 0x100 /* Internal flag to say key is malloc()ed. */
e16e2ff8
NC
306#define HVhek_PLACEHOLD 0x200 /* Internal flag to create placeholder.
307 * (may change, but Storable is a core module) */
19692e8d
NC
308#define HVhek_MASK 0xFF
309
fdcd69b6
NC
310/* Which flags enable HvHASKFLAGS? Somewhat a hack on a hack, as
311 HVhek_REHASH is only needed because the rehash flag has to be duplicated
312 into all keys as hv_iternext has no access to the hash flags. At this
313 point Storable's tests get upset, because sometimes hashes are "keyed"
314 and sometimes not, depending on the order of data insertion, and whether
315 it triggered rehashing. So currently HVhek_REHAS is exempt.
316*/
317
318#define HVhek_ENABLEHVKFLAGS (HVhek_MASK - HVhek_REHASH)
319
19692e8d
NC
320#define HEK_UTF8(hek) (HEK_FLAGS(hek) & HVhek_UTF8)
321#define HEK_UTF8_on(hek) (HEK_FLAGS(hek) |= HVhek_UTF8)
322#define HEK_UTF8_off(hek) (HEK_FLAGS(hek) &= ~HVhek_UTF8)
323#define HEK_WASUTF8(hek) (HEK_FLAGS(hek) & HVhek_WASUTF8)
324#define HEK_WASUTF8_on(hek) (HEK_FLAGS(hek) |= HVhek_WASUTF8)
325#define HEK_WASUTF8_off(hek) (HEK_FLAGS(hek) &= ~HVhek_WASUTF8)
4b5190b5
NC
326#define HEK_REHASH(hek) (HEK_FLAGS(hek) & HVhek_REHASH)
327#define HEK_REHASH_on(hek) (HEK_FLAGS(hek) |= HVhek_REHASH)
d18c6117 328
5cbe4eec 329/* calculate HV array allocation */
36768cf4
NC
330#ifndef PERL_USE_LARGE_HV_ALLOC
331/* Default to allocating the correct size - default to assuming that malloc()
332 is not broken and is efficient at allocating blocks sized at powers-of-two.
333*/
d18c6117
GS
334# define PERL_HV_ARRAY_ALLOC_BYTES(size) ((size) * sizeof(HE*))
335#else
336# define MALLOC_OVERHEAD 16
337# define PERL_HV_ARRAY_ALLOC_BYTES(size) \
338 (((size) < 64) \
339 ? (size) * sizeof(HE*) \
340 : (size) * sizeof(HE*) * 2 - MALLOC_OVERHEAD)
341#endif
37d85e3a 342
e16e2ff8
NC
343/* Flags for hv_iternext_flags. */
344#define HV_ITERNEXT_WANTPLACEHOLDERS 0x01 /* Don't skip placeholders. */
345
37d85e3a
JH
346/* available as a function in hv.c */
347#define Perl_sharepvn(sv, len, hash) HEK_KEY(share_hek(sv, len, hash))
348#define sharepvn(sv, len, hash) Perl_sharepvn(sv, len, hash)
bfcb3514
NC
349
350/*
351 * Local variables:
352 * c-indentation-style: bsd
353 * c-basic-offset: 4
354 * indent-tabs-mode: t
355 * End:
356 *
357 * ex: set ts=8 sts=4 sw=4 noet:
358 */