This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
S_save_hek_flags should honour the "free" flag.
[perl5.git] / hv.c
CommitLineData
a0d0e21e 1/* hv.c
79072805 2 *
4bb101f2
JH
3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
4 * 2000, 2001, 2002, 2003, 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 *
a0d0e21e
LW
9 */
10
11/*
12 * "I sit beside the fire and think of all that I have seen." --Bilbo
79072805
LW
13 */
14
d5afce77
RB
15/*
16=head1 Hash Manipulation Functions
17*/
18
79072805 19#include "EXTERN.h"
864dbfa3 20#define PERL_IN_HV_C
3d78eb94 21#define PERL_HASH_INTERNAL_ACCESS
79072805
LW
22#include "perl.h"
23
d8012aaf 24#define HV_MAX_LENGTH_BEFORE_SPLIT 14
fdcd69b6 25
76e3520e 26STATIC HE*
cea2e8a9 27S_new_he(pTHX)
4633a7c4
LW
28{
29 HE* he;
333f433b
DG
30 LOCK_SV_MUTEX;
31 if (!PL_he_root)
8aacddc1 32 more_he();
333f433b
DG
33 he = PL_he_root;
34 PL_he_root = HeNEXT(he);
35 UNLOCK_SV_MUTEX;
36 return he;
4633a7c4
LW
37}
38
76e3520e 39STATIC void
cea2e8a9 40S_del_he(pTHX_ HE *p)
4633a7c4 41{
333f433b 42 LOCK_SV_MUTEX;
3280af22
NIS
43 HeNEXT(p) = (HE*)PL_he_root;
44 PL_he_root = p;
333f433b 45 UNLOCK_SV_MUTEX;
4633a7c4
LW
46}
47
333f433b 48STATIC void
cea2e8a9 49S_more_he(pTHX)
4633a7c4
LW
50{
51 register HE* he;
52 register HE* heend;
612f20c3
GS
53 XPV *ptr;
54 New(54, ptr, 1008/sizeof(XPV), XPV);
55 ptr->xpv_pv = (char*)PL_he_arenaroot;
56 PL_he_arenaroot = ptr;
57
58 he = (HE*)ptr;
4633a7c4 59 heend = &he[1008 / sizeof(HE) - 1];
612f20c3 60 PL_he_root = ++he;
4633a7c4 61 while (he < heend) {
8aacddc1
NIS
62 HeNEXT(he) = (HE*)(he + 1);
63 he++;
4633a7c4 64 }
fde52b5c 65 HeNEXT(he) = 0;
4633a7c4
LW
66}
67
d33b2eba
GS
68#ifdef PURIFY
69
70#define new_HE() (HE*)safemalloc(sizeof(HE))
71#define del_HE(p) safefree((char*)p)
72
73#else
74
75#define new_HE() new_he()
76#define del_HE(p) del_he(p)
77
78#endif
79
76e3520e 80STATIC HEK *
19692e8d 81S_save_hek_flags(pTHX_ const char *str, I32 len, U32 hash, int flags)
bbce6d69 82{
dcf933a4 83 int flags_masked = flags & HVhek_MASK;
bbce6d69
PP
84 char *k;
85 register HEK *hek;
1c846c1f 86
e05949c7 87 New(54, k, HEK_BASESIZE + len + 2, char);
bbce6d69 88 hek = (HEK*)k;
ff68c719 89 Copy(str, HEK_KEY(hek), len, char);
e05949c7 90 HEK_KEY(hek)[len] = 0;
ff68c719
PP
91 HEK_LEN(hek) = len;
92 HEK_HASH(hek) = hash;
dcf933a4
NC
93 HEK_FLAGS(hek) = (unsigned char)flags_masked;
94
95 if (flags & HVhek_FREEKEY)
96 Safefree(str);
bbce6d69
PP
97 return hek;
98}
99
dd28f7bb
DM
100/* free the pool of temporary HE/HEK pairs retunrned by hv_fetch_ent
101 * for tied hashes */
102
103void
104Perl_free_tied_hv_pool(pTHX)
105{
106 HE *ohe;
107 HE *he = PL_hv_fetch_ent_mh;
108 while (he) {
109 Safefree(HeKEY_hek(he));
110 ohe = he;
111 he = HeNEXT(he);
112 del_HE(ohe);
113 }
bf9cdc68 114 PL_hv_fetch_ent_mh = Nullhe;
dd28f7bb
DM
115}
116
d18c6117
GS
117#if defined(USE_ITHREADS)
118HE *
a8fc9800 119Perl_he_dup(pTHX_ HE *e, bool shared, CLONE_PARAMS* param)
d18c6117
GS
120{
121 HE *ret;
122
123 if (!e)
124 return Nullhe;
7766f137
GS
125 /* look for it in the table first */
126 ret = (HE*)ptr_table_fetch(PL_ptr_table, e);
127 if (ret)
128 return ret;
129
130 /* create anew and remember what it is */
d33b2eba 131 ret = new_HE();
7766f137
GS
132 ptr_table_store(PL_ptr_table, e, ret);
133
d2d73c3e 134 HeNEXT(ret) = he_dup(HeNEXT(e),shared, param);
dd28f7bb
DM
135 if (HeKLEN(e) == HEf_SVKEY) {
136 char *k;
137 New(54, k, HEK_BASESIZE + sizeof(SV*), char);
138 HeKEY_hek(ret) = (HEK*)k;
d2d73c3e 139 HeKEY_sv(ret) = SvREFCNT_inc(sv_dup(HeKEY_sv(e), param));
dd28f7bb 140 }
d18c6117 141 else if (shared)
19692e8d
NC
142 HeKEY_hek(ret) = share_hek_flags(HeKEY(e), HeKLEN(e), HeHASH(e),
143 HeKFLAGS(e));
d18c6117 144 else
19692e8d
NC
145 HeKEY_hek(ret) = save_hek_flags(HeKEY(e), HeKLEN(e), HeHASH(e),
146 HeKFLAGS(e));
d2d73c3e 147 HeVAL(ret) = SvREFCNT_inc(sv_dup(HeVAL(e), param));
d18c6117
GS
148 return ret;
149}
150#endif /* USE_ITHREADS */
151
1b1f1335 152static void
2393f1b9
JH
153S_hv_notallowed(pTHX_ int flags, const char *key, I32 klen,
154 const char *msg)
1b1f1335 155{
2393f1b9 156 SV *sv = sv_newmortal(), *esv = sv_newmortal();
19692e8d 157 if (!(flags & HVhek_FREEKEY)) {
1b1f1335
NIS
158 sv_setpvn(sv, key, klen);
159 }
160 else {
161 /* Need to free saved eventually assign to mortal SV */
34c3c4e3 162 /* XXX is this line an error ???: SV *sv = sv_newmortal(); */
1b1f1335
NIS
163 sv_usepvn(sv, (char *) key, klen);
164 }
19692e8d 165 if (flags & HVhek_UTF8) {
1b1f1335
NIS
166 SvUTF8_on(sv);
167 }
2393f1b9
JH
168 Perl_sv_setpvf(aTHX_ esv, "Attempt to %s a restricted hash", msg);
169 Perl_croak(aTHX_ SvPVX(esv), sv);
1b1f1335
NIS
170}
171
fde52b5c
PP
172/* (klen == HEf_SVKEY) is special for MAGICAL hv entries, meaning key slot
173 * contains an SV* */
174
34a6f7b4
NC
175#define HV_FETCH_ISSTORE 0x01
176#define HV_FETCH_ISEXISTS 0x02
177#define HV_FETCH_LVALUE 0x04
178#define HV_FETCH_JUST_SV 0x08
179
180/*
181=for apidoc hv_store
182
183Stores an SV in a hash. The hash key is specified as C<key> and C<klen> is
184the length of the key. The C<hash> parameter is the precomputed hash
185value; if it is zero then Perl will compute it. The return value will be
186NULL if the operation failed or if the value did not need to be actually
187stored within the hash (as in the case of tied hashes). Otherwise it can
188be dereferenced to get the original C<SV*>. Note that the caller is
189responsible for suitably incrementing the reference count of C<val> before
190the call, and decrementing it if the function returned NULL. Effectively
191a successful hv_store takes ownership of one reference to C<val>. This is
192usually what you want; a newly created SV has a reference count of one, so
193if all your code does is create SVs then store them in a hash, hv_store
194will own the only reference to the new SV, and your code doesn't need to do
195anything further to tidy up. hv_store is not implemented as a call to
196hv_store_ent, and does not create a temporary SV for the key, so if your
197key data is not already in SV form then use hv_store in preference to
198hv_store_ent.
199
200See L<perlguts/"Understanding the Magic of Tied Hashes and Arrays"> for more
201information on how to use this function on tied hashes.
202
203=cut
204*/
205
206SV**
207Perl_hv_store(pTHX_ HV *hv, const char *key, I32 klen_i32, SV *val, U32 hash)
208{
209 HE *hek;
210 STRLEN klen;
211 int flags;
212
213 if (klen_i32 < 0) {
214 klen = -klen_i32;
215 flags = HVhek_UTF8;
216 } else {
217 klen = klen_i32;
218 flags = 0;
219 }
220 hek = hv_fetch_common (hv, NULL, key, klen, flags,
221 (HV_FETCH_ISSTORE|HV_FETCH_JUST_SV), val, 0);
222 return hek ? &HeVAL(hek) : NULL;
223}
224
225SV**
226Perl_hv_store_flags(pTHX_ HV *hv, const char *key, I32 klen, SV *val,
227 register U32 hash, int flags)
228{
229 HE *hek = hv_fetch_common (hv, NULL, key, klen, flags,
230 (HV_FETCH_ISSTORE|HV_FETCH_JUST_SV), val, hash);
231 return hek ? &HeVAL(hek) : NULL;
232}
233
234/*
235=for apidoc hv_store_ent
236
237Stores C<val> in a hash. The hash key is specified as C<key>. The C<hash>
238parameter is the precomputed hash value; if it is zero then Perl will
239compute it. The return value is the new hash entry so created. It will be
240NULL if the operation failed or if the value did not need to be actually
241stored within the hash (as in the case of tied hashes). Otherwise the
242contents of the return value can be accessed using the C<He?> macros
243described here. Note that the caller is responsible for suitably
244incrementing the reference count of C<val> before the call, and
245decrementing it if the function returned NULL. Effectively a successful
246hv_store_ent takes ownership of one reference to C<val>. This is
247usually what you want; a newly created SV has a reference count of one, so
248if all your code does is create SVs then store them in a hash, hv_store
249will own the only reference to the new SV, and your code doesn't need to do
250anything further to tidy up. Note that hv_store_ent only reads the C<key>;
251unlike C<val> it does not take ownership of it, so maintaining the correct
252reference count on C<key> is entirely the caller's responsibility. hv_store
253is not implemented as a call to hv_store_ent, and does not create a temporary
254SV for the key, so if your key data is not already in SV form then use
255hv_store in preference to hv_store_ent.
256
257See L<perlguts/"Understanding the Magic of Tied Hashes and Arrays"> for more
258information on how to use this function on tied hashes.
259
260=cut
261*/
262
263HE *
264Perl_hv_store_ent(pTHX_ HV *hv, SV *keysv, SV *val, U32 hash)
265{
266 return hv_fetch_common(hv, keysv, NULL, 0, 0, HV_FETCH_ISSTORE, val, hash);
267}
268
269/*
270=for apidoc hv_exists
271
272Returns a boolean indicating whether the specified hash key exists. The
273C<klen> is the length of the key.
274
275=cut
276*/
277
278bool
279Perl_hv_exists(pTHX_ HV *hv, const char *key, I32 klen_i32)
280{
281 STRLEN klen;
282 int flags;
283
284 if (klen_i32 < 0) {
285 klen = -klen_i32;
286 flags = HVhek_UTF8;
287 } else {
288 klen = klen_i32;
289 flags = 0;
290 }
291 return hv_fetch_common(hv, NULL, key, klen, flags, HV_FETCH_ISEXISTS, 0, 0)
292 ? TRUE : FALSE;
293}
294
954c1994
GS
295/*
296=for apidoc hv_fetch
297
298Returns the SV which corresponds to the specified key in the hash. The
299C<klen> is the length of the key. If C<lval> is set then the fetch will be
300part of a store. Check that the return value is non-null before
d1be9408 301dereferencing it to an C<SV*>.
954c1994 302
96f1132b 303See L<perlguts/"Understanding the Magic of Tied Hashes and Arrays"> for more
954c1994
GS
304information on how to use this function on tied hashes.
305
306=cut
307*/
308
79072805 309SV**
c1fe5510 310Perl_hv_fetch(pTHX_ HV *hv, const char *key, I32 klen_i32, I32 lval)
79072805 311{
c1fe5510
NC
312 HE *hek;
313 STRLEN klen;
314 int flags;
315
316 if (klen_i32 < 0) {
317 klen = -klen_i32;
318 flags = HVhek_UTF8;
319 } else {
320 klen = klen_i32;
321 flags = 0;
322 }
323 hek = hv_fetch_common (hv, NULL, key, klen, flags,
b2c64049
NC
324 HV_FETCH_JUST_SV | (lval ? HV_FETCH_LVALUE : 0),
325 Nullsv, 0);
113738bb 326 return hek ? &HeVAL(hek) : NULL;
79072805
LW
327}
328
34a6f7b4
NC
329/*
330=for apidoc hv_exists_ent
331
332Returns a boolean indicating whether the specified hash key exists. C<hash>
333can be a valid precomputed hash value, or 0 to ask for it to be
334computed.
335
336=cut
337*/
338
339bool
340Perl_hv_exists_ent(pTHX_ HV *hv, SV *keysv, U32 hash)
341{
342 return hv_fetch_common(hv, keysv, NULL, 0, 0, HV_FETCH_ISEXISTS, 0, hash)
343 ? TRUE : FALSE;
344}
345
d1be9408 346/* returns an HE * structure with the all fields set */
fde52b5c 347/* note that hent_val will be a mortal sv for MAGICAL hashes */
954c1994
GS
348/*
349=for apidoc hv_fetch_ent
350
351Returns the hash entry which corresponds to the specified key in the hash.
352C<hash> must be a valid precomputed hash number for the given C<key>, or 0
353if you want the function to compute it. IF C<lval> is set then the fetch
354will be part of a store. Make sure the return value is non-null before
355accessing it. The return value when C<tb> is a tied hash is a pointer to a
356static location, so be sure to make a copy of the structure if you need to
1c846c1f 357store it somewhere.
954c1994 358
96f1132b 359See L<perlguts/"Understanding the Magic of Tied Hashes and Arrays"> for more
954c1994
GS
360information on how to use this function on tied hashes.
361
362=cut
363*/
364
fde52b5c 365HE *
864dbfa3 366Perl_hv_fetch_ent(pTHX_ HV *hv, SV *keysv, I32 lval, register U32 hash)
fde52b5c 367{
7f66fda2 368 return hv_fetch_common(hv, keysv, NULL, 0, 0,
b2c64049 369 (lval ? HV_FETCH_LVALUE : 0), Nullsv, hash);
113738bb
NC
370}
371
372HE *
c1fe5510 373S_hv_fetch_common(pTHX_ HV *hv, SV *keysv, const char *key, STRLEN klen,
b2c64049 374 int flags, int action, SV *val, register U32 hash)
113738bb 375{
b2c64049
NC
376 XPVHV* xhv;
377 U32 n_links;
378 HE *entry;
379 HE **oentry;
fde52b5c 380 SV *sv;
da58a35d 381 bool is_utf8;
113738bb 382 int masked_flags;
fde52b5c
PP
383
384 if (!hv)
385 return 0;
386
113738bb 387 if (keysv) {
e593d2fe
AE
388 if (flags & HVhek_FREEKEY)
389 Safefree(key);
113738bb 390 key = SvPV(keysv, klen);
c1fe5510 391 flags = 0;
113738bb
NC
392 is_utf8 = (SvUTF8(keysv) != 0);
393 } else {
c1fe5510 394 is_utf8 = ((flags & HVhek_UTF8) ? TRUE : FALSE);
113738bb 395 }
113738bb 396
b2c64049 397 xhv = (XPVHV*)SvANY(hv);
7f66fda2
NC
398 if (SvMAGICAL(hv)) {
399 if (SvRMAGICAL(hv) && !(action & (HV_FETCH_ISSTORE|HV_FETCH_ISEXISTS)))
400 {
401 if (mg_find((SV*)hv, PERL_MAGIC_tied) || SvGMAGICAL((SV*)hv)) {
402 sv = sv_newmortal();
113738bb 403
7f66fda2
NC
404 /* XXX should be able to skimp on the HE/HEK here when
405 HV_FETCH_JUST_SV is true. */
113738bb 406
7f66fda2
NC
407 if (!keysv) {
408 keysv = newSVpvn(key, klen);
409 if (is_utf8) {
410 SvUTF8_on(keysv);
411 }
412 } else {
413 keysv = newSVsv(keysv);
113738bb 414 }
7f66fda2
NC
415 mg_copy((SV*)hv, sv, (char *)keysv, HEf_SVKEY);
416
417 /* grab a fake HE/HEK pair from the pool or make a new one */
418 entry = PL_hv_fetch_ent_mh;
419 if (entry)
420 PL_hv_fetch_ent_mh = HeNEXT(entry);
421 else {
422 char *k;
423 entry = new_HE();
424 New(54, k, HEK_BASESIZE + sizeof(SV*), char);
425 HeKEY_hek(entry) = (HEK*)k;
426 }
427 HeNEXT(entry) = Nullhe;
428 HeSVKEY_set(entry, keysv);
429 HeVAL(entry) = sv;
430 sv_upgrade(sv, SVt_PVLV);
431 LvTYPE(sv) = 'T';
432 /* so we can free entry when freeing sv */
433 LvTARG(sv) = (SV*)entry;
434
435 /* XXX remove at some point? */
436 if (flags & HVhek_FREEKEY)
437 Safefree(key);
438
439 return entry;
113738bb 440 }
7f66fda2
NC
441#ifdef ENV_IS_CASELESS
442 else if (mg_find((SV*)hv, PERL_MAGIC_env)) {
443 U32 i;
444 for (i = 0; i < klen; ++i)
445 if (isLOWER(key[i])) {
086cb327
NC
446 /* Would be nice if we had a routine to do the
447 copy and upercase in a single pass through. */
448 char *nkey = strupr(savepvn(key,klen));
449 /* Note that this fetch is for nkey (the uppercased
450 key) whereas the store is for key (the original) */
451 entry = hv_fetch_common(hv, Nullsv, nkey, klen,
452 HVhek_FREEKEY, /* free nkey */
453 0 /* non-LVAL fetch */,
454 Nullsv /* no value */,
455 0 /* compute hash */);
456 if (!entry && (action & HV_FETCH_LVALUE)) {
457 /* This call will free key if necessary.
458 Do it this way to encourage compiler to tail
459 call optimise. */
460 entry = hv_fetch_common(hv, keysv, key, klen,
461 flags, HV_FETCH_ISSTORE,
462 NEWSV(61,0), hash);
463 } else {
464 if (flags & HVhek_FREEKEY)
465 Safefree(key);
466 }
467 return entry;
7f66fda2 468 }
902173a3 469 }
7f66fda2
NC
470#endif
471 } /* ISFETCH */
472 else if (SvRMAGICAL(hv) && (action & HV_FETCH_ISEXISTS)) {
473 if (mg_find((SV*)hv, PERL_MAGIC_tied) || SvGMAGICAL((SV*)hv)) {
474 SV* svret;
b2c64049
NC
475 /* I don't understand why hv_exists_ent has svret and sv,
476 whereas hv_exists only had one. */
477 svret = sv_newmortal();
478 sv = sv_newmortal();
7f66fda2
NC
479
480 if (keysv || is_utf8) {
481 if (!keysv) {
482 keysv = newSVpvn(key, klen);
483 SvUTF8_on(keysv);
484 } else {
485 keysv = newSVsv(keysv);
486 }
b2c64049
NC
487 mg_copy((SV*)hv, sv, (char *)sv_2mortal(keysv), HEf_SVKEY);
488 } else {
489 mg_copy((SV*)hv, sv, key, klen);
7f66fda2 490 }
b2c64049
NC
491 if (flags & HVhek_FREEKEY)
492 Safefree(key);
7f66fda2
NC
493 magic_existspack(svret, mg_find(sv, PERL_MAGIC_tiedelem));
494 /* This cast somewhat evil, but I'm merely using NULL/
495 not NULL to return the boolean exists.
496 And I know hv is not NULL. */
497 return SvTRUE(svret) ? (HE *)hv : NULL;
e7152ba2 498 }
7f66fda2
NC
499#ifdef ENV_IS_CASELESS
500 else if (mg_find((SV*)hv, PERL_MAGIC_env)) {
501 /* XXX This code isn't UTF8 clean. */
b2c64049
NC
502 const char *keysave = key;
503 /* Will need to free this, so set FREEKEY flag. */
504 key = savepvn(key,klen);
505 key = (const char*)strupr((char*)key);
7f66fda2
NC
506 is_utf8 = 0;
507 hash = 0;
b2c64049
NC
508
509 if (flags & HVhek_FREEKEY) {
510 Safefree(keysave);
511 }
512 flags |= HVhek_FREEKEY;
7f66fda2 513 }
902173a3 514#endif
7f66fda2 515 } /* ISEXISTS */
b2c64049
NC
516 else if (action & HV_FETCH_ISSTORE) {
517 bool needs_copy;
518 bool needs_store;
519 hv_magic_check (hv, &needs_copy, &needs_store);
520 if (needs_copy) {
521 bool save_taint = PL_tainted;
522 if (keysv || is_utf8) {
523 if (!keysv) {
524 keysv = newSVpvn(key, klen);
525 SvUTF8_on(keysv);
526 }
527 if (PL_tainting)
528 PL_tainted = SvTAINTED(keysv);
529 keysv = sv_2mortal(newSVsv(keysv));
530 mg_copy((SV*)hv, val, (char*)keysv, HEf_SVKEY);
531 } else {
532 mg_copy((SV*)hv, val, key, klen);
533 }
534
535 TAINT_IF(save_taint);
536 if (!xhv->xhv_array /* !HvARRAY(hv) */ && !needs_store) {
537 if (flags & HVhek_FREEKEY)
538 Safefree(key);
539 return Nullhe;
540 }
541#ifdef ENV_IS_CASELESS
542 else if (mg_find((SV*)hv, PERL_MAGIC_env)) {
543 /* XXX This code isn't UTF8 clean. */
544 const char *keysave = key;
545 /* Will need to free this, so set FREEKEY flag. */
546 key = savepvn(key,klen);
547 key = (const char*)strupr((char*)key);
548 is_utf8 = 0;
549 hash = 0;
550
551 if (flags & HVhek_FREEKEY) {
552 Safefree(keysave);
553 }
554 flags |= HVhek_FREEKEY;
555 }
556#endif
557 }
558 } /* ISSTORE */
7f66fda2 559 } /* SvMAGICAL */
fde52b5c 560
cbec9347 561 if (!xhv->xhv_array /* !HvARRAY(hv) */) {
b2c64049 562 if ((action & (HV_FETCH_LVALUE | HV_FETCH_ISSTORE))
fde52b5c 563#ifdef DYNAMIC_ENV_FETCH /* if it's an %ENV lookup, we may get it on the fly */
8aacddc1 564 || (SvRMAGICAL((SV*)hv) && mg_find((SV*)hv, PERL_MAGIC_env))
fde52b5c 565#endif
8aacddc1 566 )
cbec9347
JH
567 Newz(503, xhv->xhv_array /* HvARRAY(hv) */,
568 PERL_HV_ARRAY_ALLOC_BYTES(xhv->xhv_max+1 /* HvMAX(hv)+1 */),
569 char);
7f66fda2
NC
570#ifdef DYNAMIC_ENV_FETCH
571 else if (action & HV_FETCH_ISEXISTS) {
572 /* for an %ENV exists, if we do an insert it's by a recursive
573 store call, so avoid creating HvARRAY(hv) right now. */
574 }
575#endif
113738bb
NC
576 else {
577 /* XXX remove at some point? */
578 if (flags & HVhek_FREEKEY)
579 Safefree(key);
580
fde52b5c 581 return 0;
113738bb 582 }
fde52b5c
PP
583 }
584
19692e8d 585 if (is_utf8) {
7f66fda2 586 const char *keysave = key;
f9a63242 587 key = (char*)bytes_from_utf8((U8*)key, &klen, &is_utf8);
19692e8d 588 if (is_utf8)
c1fe5510
NC
589 flags |= HVhek_UTF8;
590 else
591 flags &= ~HVhek_UTF8;
7f66fda2
NC
592 if (key != keysave) {
593 if (flags & HVhek_FREEKEY)
594 Safefree(keysave);
19692e8d 595 flags |= HVhek_WASUTF8 | HVhek_FREEKEY;
7f66fda2 596 }
19692e8d 597 }
f9a63242 598
4b5190b5
NC
599 if (HvREHASH(hv)) {
600 PERL_HASH_INTERNAL(hash, key, klen);
b2c64049
NC
601 /* We don't have a pointer to the hv, so we have to replicate the
602 flag into every HEK, so that hv_iterkeysv can see it. */
603 /* And yes, you do need this even though you are not "storing" because
fdcd69b6
NC
604 you can flip the flags below if doing an lval lookup. (And that
605 was put in to give the semantics Andreas was expecting.) */
606 flags |= HVhek_REHASH;
4b5190b5 607 } else if (!hash) {
113738bb 608 if (keysv && (SvIsCOW_shared_hash(keysv))) {
46187eeb
NC
609 hash = SvUVX(keysv);
610 } else {
611 PERL_HASH(hash, key, klen);
612 }
613 }
effa1e2d 614
113738bb 615 masked_flags = (flags & HVhek_MASK);
b2c64049 616 n_links = 0;
113738bb 617
7f66fda2
NC
618#ifdef DYNAMIC_ENV_FETCH
619 if (!xhv->xhv_array /* !HvARRAY(hv) */) entry = Null(HE*);
620 else
621#endif
b2c64049 622 {
ab4af705
NC
623 /* entry = (HvARRAY(hv))[hash & (I32) HvMAX(hv)]; */
624 entry = ((HE**)xhv->xhv_array)[hash & (I32) xhv->xhv_max];
b2c64049
NC
625 }
626 for (; entry; ++n_links, entry = HeNEXT(entry)) {
fde52b5c
PP
627 if (HeHASH(entry) != hash) /* strings can't be equal */
628 continue;
eb160463 629 if (HeKLEN(entry) != (I32)klen)
fde52b5c 630 continue;
1c846c1f 631 if (HeKEY(entry) != key && memNE(HeKEY(entry),key,klen)) /* is this it? */
fde52b5c 632 continue;
113738bb 633 if ((HeKFLAGS(entry) ^ masked_flags) & HVhek_UTF8)
c3654f1a 634 continue;
b2c64049
NC
635
636 if (action & (HV_FETCH_LVALUE|HV_FETCH_ISSTORE)) {
637 if (HeKFLAGS(entry) != masked_flags) {
638 /* We match if HVhek_UTF8 bit in our flags and hash key's
639 match. But if entry was set previously with HVhek_WASUTF8
640 and key now doesn't (or vice versa) then we should change
641 the key's flag, as this is assignment. */
642 if (HvSHAREKEYS(hv)) {
643 /* Need to swap the key we have for a key with the flags we
644 need. As keys are shared we can't just write to the
645 flag, so we share the new one, unshare the old one. */
646 HEK *new_hek = share_hek_flags(key, klen, hash,
647 masked_flags);
648 unshare_hek (HeKEY_hek(entry));
649 HeKEY_hek(entry) = new_hek;
650 }
651 else
652 HeKFLAGS(entry) = masked_flags;
653 if (masked_flags & HVhek_ENABLEHVKFLAGS)
654 HvHASKFLAGS_on(hv);
655 }
656 if (HeVAL(entry) == &PL_sv_placeholder) {
657 /* yes, can store into placeholder slot */
658 if (action & HV_FETCH_LVALUE) {
659 if (SvMAGICAL(hv)) {
660 /* This preserves behaviour with the old hv_fetch
661 implementation which at this point would bail out
662 with a break; (at "if we find a placeholder, we
663 pretend we haven't found anything")
664
665 That break mean that if a placeholder were found, it
666 caused a call into hv_store, which in turn would
667 check magic, and if there is no magic end up pretty
668 much back at this point (in hv_store's code). */
669 break;
670 }
671 /* LVAL fetch which actaully needs a store. */
672 val = NEWSV(61,0);
673 xhv->xhv_placeholders--;
674 } else {
675 /* store */
676 if (val != &PL_sv_placeholder)
677 xhv->xhv_placeholders--;
678 }
679 HeVAL(entry) = val;
680 } else if (action & HV_FETCH_ISSTORE) {
681 SvREFCNT_dec(HeVAL(entry));
682 HeVAL(entry) = val;
683 }
684 } else if (HeVAL(entry) == &PL_sv_placeholder) {
685 /* if we find a placeholder, we pretend we haven't found
686 anything */
8aacddc1 687 break;
b2c64049 688 }
113738bb
NC
689 if (flags & HVhek_FREEKEY)
690 Safefree(key);
fde52b5c
PP
691 return entry;
692 }
693#ifdef DYNAMIC_ENV_FETCH /* %ENV lookup? If so, try to fetch the value now */
0ed29950
NC
694 if (!(action & HV_FETCH_ISSTORE)
695 && SvRMAGICAL((SV*)hv) && mg_find((SV*)hv, PERL_MAGIC_env)) {
a6c40364
GS
696 unsigned long len;
697 char *env = PerlEnv_ENVgetenv_len(key,&len);
698 if (env) {
699 sv = newSVpvn(env,len);
700 SvTAINTED_on(sv);
7fd3d16e 701 return hv_fetch_common(hv,keysv,key,klen,flags,HV_FETCH_ISSTORE,sv,
b2c64049 702 hash);
a6c40364 703 }
fde52b5c
PP
704 }
705#endif
7f66fda2
NC
706
707 if (!entry && SvREADONLY(hv) && !(action & HV_FETCH_ISEXISTS)) {
2393f1b9
JH
708 S_hv_notallowed(aTHX_ flags, key, klen,
709 "access disallowed key '%"SVf"' in"
710 );
1b1f1335 711 }
b2c64049
NC
712 if (!(action & (HV_FETCH_LVALUE|HV_FETCH_ISSTORE))) {
713 /* Not doing some form of store, so return failure. */
714 if (flags & HVhek_FREEKEY)
715 Safefree(key);
716 return 0;
717 }
113738bb 718 if (action & HV_FETCH_LVALUE) {
b2c64049
NC
719 val = NEWSV(61,0);
720 if (SvMAGICAL(hv)) {
721 /* At this point the old hv_fetch code would call to hv_store,
722 which in turn might do some tied magic. So we need to make that
723 magic check happen. */
724 /* gonna assign to this, so it better be there */
725 return hv_fetch_common(hv, keysv, key, klen, flags,
726 HV_FETCH_ISSTORE, val, hash);
727 /* XXX Surely that could leak if the fetch-was-store fails?
728 Just like the hv_fetch. */
113738bb
NC
729 }
730 }
731
b2c64049
NC
732 /* Welcome to hv_store... */
733
ab4af705 734 if (!xhv->xhv_array) {
b2c64049
NC
735 /* Not sure if we can get here. I think the only case of oentry being
736 NULL is for %ENV with dynamic env fetch. But that should disappear
737 with magic in the previous code. */
738 Newz(503, xhv->xhv_array /* HvARRAY(hv) */,
739 PERL_HV_ARRAY_ALLOC_BYTES(xhv->xhv_max+1 /* HvMAX(hv)+1 */),
740 char);
b2c64049
NC
741 }
742
ab4af705
NC
743 oentry = &((HE**)xhv->xhv_array)[hash & (I32) xhv->xhv_max];
744
b2c64049
NC
745 entry = new_HE();
746 /* share_hek_flags will do the free for us. This might be considered
747 bad API design. */
748 if (HvSHAREKEYS(hv))
749 HeKEY_hek(entry) = share_hek_flags(key, klen, hash, flags);
750 else /* gotta do the real thing */
751 HeKEY_hek(entry) = save_hek_flags(key, klen, hash, flags);
752 HeVAL(entry) = val;
753 HeNEXT(entry) = *oentry;
754 *oentry = entry;
755
756 if (val == &PL_sv_placeholder)
757 xhv->xhv_placeholders++;
758 if (masked_flags & HVhek_ENABLEHVKFLAGS)
759 HvHASKFLAGS_on(hv);
760
761 xhv->xhv_keys++; /* HvKEYS(hv)++ */
762 if (!n_links) { /* initial entry? */
763 xhv->xhv_fill++; /* HvFILL(hv)++ */
764 } else if ((xhv->xhv_keys > (IV)xhv->xhv_max)
765 || ((n_links > HV_MAX_LENGTH_BEFORE_SPLIT) && !HvREHASH(hv))) {
766 /* Use only the old HvKEYS(hv) > HvMAX(hv) condition to limit bucket
767 splits on a rehashed hash, as we're not going to split it again,
768 and if someone is lucky (evil) enough to get all the keys in one
769 list they could exhaust our memory as we repeatedly double the
770 number of buckets on every entry. Linear search feels a less worse
771 thing to do. */
772 hsplit(hv);
fde52b5c 773 }
b2c64049
NC
774
775 return entry;
fde52b5c
PP
776}
777
864dbfa3 778STATIC void
cea2e8a9 779S_hv_magic_check(pTHX_ HV *hv, bool *needs_copy, bool *needs_store)
d0066dc7
OT
780{
781 MAGIC *mg = SvMAGIC(hv);
782 *needs_copy = FALSE;
783 *needs_store = TRUE;
784 while (mg) {
785 if (isUPPER(mg->mg_type)) {
786 *needs_copy = TRUE;
787 switch (mg->mg_type) {
14befaf4
DM
788 case PERL_MAGIC_tied:
789 case PERL_MAGIC_sig:
d0066dc7 790 *needs_store = FALSE;
d0066dc7
OT
791 }
792 }
793 mg = mg->mg_moremagic;
794 }
795}
796
954c1994 797/*
a3bcc51e
TP
798=for apidoc hv_scalar
799
800Evaluates the hash in scalar context and returns the result. Handles magic when the hash is tied.
801
802=cut
803*/
804
805SV *
806Perl_hv_scalar(pTHX_ HV *hv)
807{
808 MAGIC *mg;
809 SV *sv;
810
811 if ((SvRMAGICAL(hv) && (mg = mg_find((SV*)hv, PERL_MAGIC_tied)))) {
812 sv = magic_scalarpack(hv, mg);
813 return sv;
814 }
815
816 sv = sv_newmortal();
817 if (HvFILL((HV*)hv))
818 Perl_sv_setpvf(aTHX_ sv, "%ld/%ld",
819 (long)HvFILL(hv), (long)HvMAX(hv) + 1);
820 else
821 sv_setiv(sv, 0);
822
823 return sv;
824}
825
826/*
954c1994
GS
827=for apidoc hv_delete
828
829Deletes a key/value pair in the hash. The value SV is removed from the
1c846c1f 830hash and returned to the caller. The C<klen> is the length of the key.
954c1994
GS
831The C<flags> value will normally be zero; if set to G_DISCARD then NULL
832will be returned.
833
834=cut
835*/
836
79072805 837SV *
cd6d36ac 838Perl_hv_delete(pTHX_ HV *hv, const char *key, I32 klen_i32, I32 flags)
79072805 839{
cd6d36ac
NC
840 STRLEN klen;
841 int k_flags = 0;
842
843 if (klen_i32 < 0) {
844 klen = -klen_i32;
845 k_flags |= HVhek_UTF8;
846 } else {
847 klen = klen_i32;
848 }
849 return hv_delete_common(hv, NULL, key, klen, k_flags, flags, 0);
fde52b5c
PP
850}
851
954c1994
GS
852/*
853=for apidoc hv_delete_ent
854
855Deletes a key/value pair in the hash. The value SV is removed from the
856hash and returned to the caller. The C<flags> value will normally be zero;
857if set to G_DISCARD then NULL will be returned. C<hash> can be a valid
858precomputed hash value, or 0 to ask for it to be computed.
859
860=cut
861*/
862
fde52b5c 863SV *
864dbfa3 864Perl_hv_delete_ent(pTHX_ HV *hv, SV *keysv, I32 flags, U32 hash)
fde52b5c 865{
cd6d36ac 866 return hv_delete_common(hv, keysv, NULL, 0, 0, flags, hash);
f1317c8d
NC
867}
868
869SV *
cd6d36ac
NC
870S_hv_delete_common(pTHX_ HV *hv, SV *keysv, const char *key, STRLEN klen,
871 int k_flags, I32 d_flags, U32 hash)
f1317c8d 872{
cbec9347 873 register XPVHV* xhv;
fde52b5c 874 register I32 i;
fde52b5c
PP
875 register HE *entry;
876 register HE **oentry;
877 SV *sv;
da58a35d 878 bool is_utf8;
7a9669ca 879 int masked_flags;
1c846c1f 880
fde52b5c
PP
881 if (!hv)
882 return Nullsv;
f1317c8d
NC
883
884 if (keysv) {
e593d2fe
AE
885 if (k_flags & HVhek_FREEKEY)
886 Safefree(key);
f1317c8d 887 key = SvPV(keysv, klen);
cd6d36ac 888 k_flags = 0;
f1317c8d
NC
889 is_utf8 = (SvUTF8(keysv) != 0);
890 } else {
cd6d36ac 891 is_utf8 = ((k_flags & HVhek_UTF8) ? TRUE : FALSE);
f1317c8d 892 }
f1317c8d 893
fde52b5c 894 if (SvRMAGICAL(hv)) {
0a0bb7c7
OT
895 bool needs_copy;
896 bool needs_store;
897 hv_magic_check (hv, &needs_copy, &needs_store);
898
f1317c8d 899 if (needs_copy) {
7a9669ca
NC
900 entry = hv_fetch_common(hv, keysv, key, klen,
901 k_flags & ~HVhek_FREEKEY, HV_FETCH_LVALUE,
b2c64049 902 Nullsv, hash);
7a9669ca 903 sv = entry ? HeVAL(entry) : NULL;
f1317c8d
NC
904 if (sv) {
905 if (SvMAGICAL(sv)) {
906 mg_clear(sv);
907 }
908 if (!needs_store) {
909 if (mg_find(sv, PERL_MAGIC_tiedelem)) {
910 /* No longer an element */
911 sv_unmagic(sv, PERL_MAGIC_tiedelem);
912 return sv;
913 }
914 return Nullsv; /* element cannot be deleted */
915 }
902173a3 916#ifdef ENV_IS_CASELESS
8167a60a
NC
917 else if (mg_find((SV*)hv, PERL_MAGIC_env)) {
918 /* XXX This code isn't UTF8 clean. */
919 keysv = sv_2mortal(newSVpvn(key,klen));
920 if (k_flags & HVhek_FREEKEY) {
921 Safefree(key);
922 }
923 key = strupr(SvPVX(keysv));
924 is_utf8 = 0;
925 k_flags = 0;
926 hash = 0;
7f66fda2 927 }
510ac311 928#endif
2fd1c6b8 929 }
2fd1c6b8 930 }
fde52b5c 931 }
cbec9347
JH
932 xhv = (XPVHV*)SvANY(hv);
933 if (!xhv->xhv_array /* !HvARRAY(hv) */)
fde52b5c
PP
934 return Nullsv;
935
19692e8d 936 if (is_utf8) {
7f66fda2
NC
937 const char *keysave = key;
938 key = (char*)bytes_from_utf8((U8*)key, &klen, &is_utf8);
cd6d36ac 939
19692e8d 940 if (is_utf8)
cd6d36ac
NC
941 k_flags |= HVhek_UTF8;
942 else
943 k_flags &= ~HVhek_UTF8;
7f66fda2
NC
944 if (key != keysave) {
945 if (k_flags & HVhek_FREEKEY) {
946 /* This shouldn't happen if our caller does what we expect,
947 but strictly the API allows it. */
948 Safefree(keysave);
949 }
950 k_flags |= HVhek_WASUTF8 | HVhek_FREEKEY;
951 }
cd6d36ac 952 HvHASKFLAGS_on((SV*)hv);
19692e8d 953 }
f9a63242 954
4b5190b5
NC
955 if (HvREHASH(hv)) {
956 PERL_HASH_INTERNAL(hash, key, klen);
957 } else if (!hash) {
7a9669ca
NC
958 if (keysv && (SvIsCOW_shared_hash(keysv))) {
959 hash = SvUVX(keysv);
960 } else {
961 PERL_HASH(hash, key, klen);
962 }
5afd6d42 963 PERL_HASH(hash, key, klen);
4b5190b5 964 }
fde52b5c 965
7a9669ca
NC
966 masked_flags = (k_flags & HVhek_MASK);
967
cbec9347
JH
968 /* oentry = &(HvARRAY(hv))[hash & (I32) HvMAX(hv)]; */
969 oentry = &((HE**)xhv->xhv_array)[hash & (I32) xhv->xhv_max];
fde52b5c
PP
970 entry = *oentry;
971 i = 1;
972 for (; entry; i=0, oentry = &HeNEXT(entry), entry = *oentry) {
973 if (HeHASH(entry) != hash) /* strings can't be equal */
974 continue;
eb160463 975 if (HeKLEN(entry) != (I32)klen)
fde52b5c 976 continue;
1c846c1f 977 if (HeKEY(entry) != key && memNE(HeKEY(entry),key,klen)) /* is this it? */
fde52b5c 978 continue;
7a9669ca 979 if ((HeKFLAGS(entry) ^ masked_flags) & HVhek_UTF8)
c3654f1a 980 continue;
19692e8d
NC
981 if (k_flags & HVhek_FREEKEY)
982 Safefree(key);
8aacddc1
NIS
983
984 /* if placeholder is here, it's already been deleted.... */
7996736c 985 if (HeVAL(entry) == &PL_sv_placeholder)
8aacddc1
NIS
986 {
987 if (SvREADONLY(hv))
988 return Nullsv; /* if still SvREADONLY, leave it deleted. */
03fed38d
MB
989
990 /* okay, really delete the placeholder. */
991 *oentry = HeNEXT(entry);
992 if (i && !*oentry)
993 xhv->xhv_fill--; /* HvFILL(hv)-- */
994 if (entry == xhv->xhv_eiter /* HvEITER(hv) */)
995 HvLAZYDEL_on(hv);
996 else
997 hv_free_ent(hv, entry);
998 xhv->xhv_keys--; /* HvKEYS(hv)-- */
574c8022 999 if (xhv->xhv_keys == 0)
19692e8d 1000 HvHASKFLAGS_off(hv);
03fed38d
MB
1001 xhv->xhv_placeholders--;
1002 return Nullsv;
8aacddc1
NIS
1003 }
1004 else if (SvREADONLY(hv) && HeVAL(entry) && SvREADONLY(HeVAL(entry))) {
2393f1b9
JH
1005 S_hv_notallowed(aTHX_ k_flags, key, klen,
1006 "delete readonly key '%"SVf"' from"
1007 );
8aacddc1
NIS
1008 }
1009
cd6d36ac 1010 if (d_flags & G_DISCARD)
fde52b5c 1011 sv = Nullsv;
94f7643d 1012 else {
79d01fbf 1013 sv = sv_2mortal(HeVAL(entry));
7996736c 1014 HeVAL(entry) = &PL_sv_placeholder;
94f7643d 1015 }
8aacddc1
NIS
1016
1017 /*
1018 * If a restricted hash, rather than really deleting the entry, put
1019 * a placeholder there. This marks the key as being "approved", so
1020 * we can still access via not-really-existing key without raising
1021 * an error.
1022 */
1023 if (SvREADONLY(hv)) {
7996736c 1024 HeVAL(entry) = &PL_sv_placeholder;
8aacddc1
NIS
1025 /* We'll be saving this slot, so the number of allocated keys
1026 * doesn't go down, but the number placeholders goes up */
1027 xhv->xhv_placeholders++; /* HvPLACEHOLDERS(hv)++ */
1028 } else {
a26e96df
NIS
1029 *oentry = HeNEXT(entry);
1030 if (i && !*oentry)
1031 xhv->xhv_fill--; /* HvFILL(hv)-- */
8aacddc1
NIS
1032 if (entry == xhv->xhv_eiter /* HvEITER(hv) */)
1033 HvLAZYDEL_on(hv);
1034 else
1035 hv_free_ent(hv, entry);
1036 xhv->xhv_keys--; /* HvKEYS(hv)-- */
574c8022 1037 if (xhv->xhv_keys == 0)
19692e8d 1038 HvHASKFLAGS_off(hv);
8aacddc1 1039 }
79072805
LW
1040 return sv;
1041 }
8aacddc1 1042 if (SvREADONLY(hv)) {
2393f1b9
JH
1043 S_hv_notallowed(aTHX_ k_flags, key, klen,
1044 "delete disallowed key '%"SVf"' from"
1045 );
8aacddc1
NIS
1046 }
1047
19692e8d 1048 if (k_flags & HVhek_FREEKEY)
f9a63242 1049 Safefree(key);
79072805 1050 return Nullsv;
79072805
LW
1051}
1052
76e3520e 1053STATIC void
cea2e8a9 1054S_hsplit(pTHX_ HV *hv)
79072805 1055{
cbec9347
JH
1056 register XPVHV* xhv = (XPVHV*)SvANY(hv);
1057 I32 oldsize = (I32) xhv->xhv_max+1; /* HvMAX(hv)+1 (sick) */
79072805
LW
1058 register I32 newsize = oldsize * 2;
1059 register I32 i;
cbec9347 1060 register char *a = xhv->xhv_array; /* HvARRAY(hv) */
72311751
GS
1061 register HE **aep;
1062 register HE **bep;
79072805
LW
1063 register HE *entry;
1064 register HE **oentry;
4b5190b5
NC
1065 int longest_chain = 0;
1066 int was_shared;
79072805 1067
3280af22 1068 PL_nomemok = TRUE;
8d6dde3e 1069#if defined(STRANGE_MALLOC) || defined(MYMALLOC)
d18c6117 1070 Renew(a, PERL_HV_ARRAY_ALLOC_BYTES(newsize), char);
422a93e5 1071 if (!a) {
4a33f861 1072 PL_nomemok = FALSE;
422a93e5
GA
1073 return;
1074 }
4633a7c4 1075#else
d18c6117 1076 New(2, a, PERL_HV_ARRAY_ALLOC_BYTES(newsize), char);
422a93e5 1077 if (!a) {
3280af22 1078 PL_nomemok = FALSE;
422a93e5
GA
1079 return;
1080 }
cbec9347 1081 Copy(xhv->xhv_array /* HvARRAY(hv) */, a, oldsize * sizeof(HE*), char);
fba3b22e 1082 if (oldsize >= 64) {
cbec9347
JH
1083 offer_nice_chunk(xhv->xhv_array /* HvARRAY(hv) */,
1084 PERL_HV_ARRAY_ALLOC_BYTES(oldsize));
4633a7c4
LW
1085 }
1086 else
cbec9347 1087 Safefree(xhv->xhv_array /* HvARRAY(hv) */);
4633a7c4
LW
1088#endif
1089
3280af22 1090 PL_nomemok = FALSE;
72311751 1091 Zero(&a[oldsize * sizeof(HE*)], (newsize-oldsize) * sizeof(HE*), char); /* zero 2nd half*/
cbec9347
JH
1092 xhv->xhv_max = --newsize; /* HvMAX(hv) = --newsize */
1093 xhv->xhv_array = a; /* HvARRAY(hv) = a */
72311751 1094 aep = (HE**)a;
79072805 1095
72311751 1096 for (i=0; i<oldsize; i++,aep++) {
4b5190b5
NC
1097 int left_length = 0;
1098 int right_length = 0;
1099
72311751 1100 if (!*aep) /* non-existent */
79072805 1101 continue;
72311751
GS
1102 bep = aep+oldsize;
1103 for (oentry = aep, entry = *aep; entry; entry = *oentry) {
eb160463 1104 if ((HeHASH(entry) & newsize) != (U32)i) {
fde52b5c 1105 *oentry = HeNEXT(entry);
72311751
GS
1106 HeNEXT(entry) = *bep;
1107 if (!*bep)
cbec9347 1108 xhv->xhv_fill++; /* HvFILL(hv)++ */
72311751 1109 *bep = entry;
4b5190b5 1110 right_length++;
79072805
LW
1111 continue;
1112 }
4b5190b5 1113 else {
fde52b5c 1114 oentry = &HeNEXT(entry);
4b5190b5
NC
1115 left_length++;
1116 }
79072805 1117 }
72311751 1118 if (!*aep) /* everything moved */
cbec9347 1119 xhv->xhv_fill--; /* HvFILL(hv)-- */
4b5190b5
NC
1120 /* I think we don't actually need to keep track of the longest length,
1121 merely flag if anything is too long. But for the moment while
1122 developing this code I'll track it. */
1123 if (left_length > longest_chain)
1124 longest_chain = left_length;
1125 if (right_length > longest_chain)
1126 longest_chain = right_length;
1127 }
1128
1129
1130 /* Pick your policy for "hashing isn't working" here: */
fdcd69b6 1131 if (longest_chain <= HV_MAX_LENGTH_BEFORE_SPLIT /* split worked? */
4b5190b5
NC
1132 || HvREHASH(hv)) {
1133 return;
79072805 1134 }
4b5190b5
NC
1135
1136 if (hv == PL_strtab) {
1137 /* Urg. Someone is doing something nasty to the string table.
1138 Can't win. */
1139 return;
1140 }
1141
1142 /* Awooga. Awooga. Pathological data. */
fdcd69b6 1143 /*PerlIO_printf(PerlIO_stderr(), "%p %d of %d with %d/%d buckets\n", hv,
4b5190b5
NC
1144 longest_chain, HvTOTALKEYS(hv), HvFILL(hv), 1+HvMAX(hv));*/
1145
1146 ++newsize;
1147 Newz(2, a, PERL_HV_ARRAY_ALLOC_BYTES(newsize), char);
1148 was_shared = HvSHAREKEYS(hv);
1149
1150 xhv->xhv_fill = 0;
1151 HvSHAREKEYS_off(hv);
1152 HvREHASH_on(hv);
1153
1154 aep = (HE **) xhv->xhv_array;
1155
1156 for (i=0; i<newsize; i++,aep++) {
1157 entry = *aep;
1158 while (entry) {
1159 /* We're going to trash this HE's next pointer when we chain it
1160 into the new hash below, so store where we go next. */
1161 HE *next = HeNEXT(entry);
1162 UV hash;
1163
1164 /* Rehash it */
1165 PERL_HASH_INTERNAL(hash, HeKEY(entry), HeKLEN(entry));
1166
1167 if (was_shared) {
1168 /* Unshare it. */
1169 HEK *new_hek
1170 = save_hek_flags(HeKEY(entry), HeKLEN(entry),
1171 hash, HeKFLAGS(entry));
1172 unshare_hek (HeKEY_hek(entry));
1173 HeKEY_hek(entry) = new_hek;
1174 } else {
1175 /* Not shared, so simply write the new hash in. */
1176 HeHASH(entry) = hash;
1177 }
1178 /*PerlIO_printf(PerlIO_stderr(), "%d ", HeKFLAGS(entry));*/
1179 HEK_REHASH_on(HeKEY_hek(entry));
1180 /*PerlIO_printf(PerlIO_stderr(), "%d\n", HeKFLAGS(entry));*/
1181
1182 /* Copy oentry to the correct new chain. */
1183 bep = ((HE**)a) + (hash & (I32) xhv->xhv_max);
1184 if (!*bep)
1185 xhv->xhv_fill++; /* HvFILL(hv)++ */
1186 HeNEXT(entry) = *bep;
1187 *bep = entry;
1188
1189 entry = next;
1190 }
1191 }
1192 Safefree (xhv->xhv_array);
1193 xhv->xhv_array = a; /* HvARRAY(hv) = a */
79072805
LW
1194}
1195
72940dca 1196void
864dbfa3 1197Perl_hv_ksplit(pTHX_ HV *hv, IV newmax)
72940dca 1198{
cbec9347
JH
1199 register XPVHV* xhv = (XPVHV*)SvANY(hv);
1200 I32 oldsize = (I32) xhv->xhv_max+1; /* HvMAX(hv)+1 (sick) */
72940dca
PP
1201 register I32 newsize;
1202 register I32 i;
1203 register I32 j;
72311751
GS
1204 register char *a;
1205 register HE **aep;
72940dca
PP
1206 register HE *entry;
1207 register HE **oentry;
1208
1209 newsize = (I32) newmax; /* possible truncation here */
1210 if (newsize != newmax || newmax <= oldsize)
1211 return;
1212 while ((newsize & (1 + ~newsize)) != newsize) {
1213 newsize &= ~(newsize & (1 + ~newsize)); /* get proper power of 2 */
1214 }
1215 if (newsize < newmax)
1216 newsize *= 2;
1217 if (newsize < newmax)
1218 return; /* overflow detection */
1219
cbec9347 1220 a = xhv->xhv_array; /* HvARRAY(hv) */
72940dca 1221 if (a) {
3280af22 1222 PL_nomemok = TRUE;
8d6dde3e 1223#if defined(STRANGE_MALLOC) || defined(MYMALLOC)
d18c6117 1224 Renew(a, PERL_HV_ARRAY_ALLOC_BYTES(newsize), char);
8aacddc1 1225 if (!a) {
4a33f861 1226 PL_nomemok = FALSE;
422a93e5
GA
1227 return;
1228 }
72940dca 1229#else
d18c6117 1230 New(2, a, PERL_HV_ARRAY_ALLOC_BYTES(newsize), char);
8aacddc1 1231 if (!a) {
3280af22 1232 PL_nomemok = FALSE;
422a93e5
GA
1233 return;
1234 }
cbec9347 1235 Copy(xhv->xhv_array /* HvARRAY(hv) */, a, oldsize * sizeof(HE*), char);
fba3b22e 1236 if (oldsize >= 64) {
cbec9347
JH
1237 offer_nice_chunk(xhv->xhv_array /* HvARRAY(hv) */,
1238 PERL_HV_ARRAY_ALLOC_BYTES(oldsize));
72940dca
PP
1239 }
1240 else
cbec9347 1241 Safefree(xhv->xhv_array /* HvARRAY(hv) */);
72940dca 1242#endif
3280af22 1243 PL_nomemok = FALSE;
72311751 1244 Zero(&a[oldsize * sizeof(HE*)], (newsize-oldsize) * sizeof(HE*), char); /* zero 2nd half*/
72940dca
PP
1245 }
1246 else {
d18c6117 1247 Newz(0, a, PERL_HV_ARRAY_ALLOC_BYTES(newsize), char);
72940dca 1248 }
cbec9347
JH
1249 xhv->xhv_max = --newsize; /* HvMAX(hv) = --newsize */
1250 xhv->xhv_array = a; /* HvARRAY(hv) = a */
1251 if (!xhv->xhv_fill /* !HvFILL(hv) */) /* skip rest if no entries */
72940dca
PP
1252 return;
1253
72311751
GS
1254 aep = (HE**)a;
1255 for (i=0; i<oldsize; i++,aep++) {
1256 if (!*aep) /* non-existent */
72940dca 1257 continue;
72311751 1258 for (oentry = aep, entry = *aep; entry; entry = *oentry) {
72940dca
PP
1259 if ((j = (HeHASH(entry) & newsize)) != i) {
1260 j -= i;
1261 *oentry = HeNEXT(entry);
72311751 1262 if (!(HeNEXT(entry) = aep[j]))
cbec9347 1263 xhv->xhv_fill++; /* HvFILL(hv)++ */
72311751 1264 aep[j] = entry;
72940dca
PP
1265 continue;
1266 }
1267 else
1268 oentry = &HeNEXT(entry);
1269 }
72311751 1270 if (!*aep) /* everything moved */
cbec9347 1271 xhv->xhv_fill--; /* HvFILL(hv)-- */
72940dca
PP
1272 }
1273}
1274
954c1994
GS
1275/*
1276=for apidoc newHV
1277
1278Creates a new HV. The reference count is set to 1.
1279
1280=cut
1281*/
1282
79072805 1283HV *
864dbfa3 1284Perl_newHV(pTHX)
79072805
LW
1285{
1286 register HV *hv;
cbec9347 1287 register XPVHV* xhv;
79072805 1288
a0d0e21e
LW
1289 hv = (HV*)NEWSV(502,0);
1290 sv_upgrade((SV *)hv, SVt_PVHV);
cbec9347 1291 xhv = (XPVHV*)SvANY(hv);
79072805
LW
1292 SvPOK_off(hv);
1293 SvNOK_off(hv);
1c846c1f 1294#ifndef NODEFAULT_SHAREKEYS
fde52b5c 1295 HvSHAREKEYS_on(hv); /* key-sharing on by default */
1c846c1f 1296#endif
4b5190b5 1297
cbec9347
JH
1298 xhv->xhv_max = 7; /* HvMAX(hv) = 7 (start with 8 buckets) */
1299 xhv->xhv_fill = 0; /* HvFILL(hv) = 0 */
1300 xhv->xhv_pmroot = 0; /* HvPMROOT(hv) = 0 */
79072805
LW
1301 (void)hv_iterinit(hv); /* so each() will start off right */
1302 return hv;
1303}
1304
b3ac6de7 1305HV *
864dbfa3 1306Perl_newHVhv(pTHX_ HV *ohv)
b3ac6de7 1307{
b56ba0bf 1308 HV *hv = newHV();
4beac62f 1309 STRLEN hv_max, hv_fill;
4beac62f
AMS
1310
1311 if (!ohv || (hv_fill = HvFILL(ohv)) == 0)
1312 return hv;
4beac62f 1313 hv_max = HvMAX(ohv);
b3ac6de7 1314
b56ba0bf
AMS
1315 if (!SvMAGICAL((SV *)ohv)) {
1316 /* It's an ordinary hash, so copy it fast. AMS 20010804 */
eb160463
GS
1317 STRLEN i;
1318 bool shared = !!HvSHAREKEYS(ohv);
b56ba0bf 1319 HE **ents, **oents = (HE **)HvARRAY(ohv);
ff875642
JH
1320 char *a;
1321 New(0, a, PERL_HV_ARRAY_ALLOC_BYTES(hv_max+1), char);
1322 ents = (HE**)a;
b56ba0bf
AMS
1323
1324 /* In each bucket... */
1325 for (i = 0; i <= hv_max; i++) {
1326 HE *prev = NULL, *ent = NULL, *oent = oents[i];
1327
1328 if (!oent) {
1329 ents[i] = NULL;
1330 continue;
1331 }
1332
1333 /* Copy the linked list of entries. */
1334 for (oent = oents[i]; oent; oent = HeNEXT(oent)) {
1335 U32 hash = HeHASH(oent);
1336 char *key = HeKEY(oent);
19692e8d
NC
1337 STRLEN len = HeKLEN(oent);
1338 int flags = HeKFLAGS(oent);
b56ba0bf
AMS
1339
1340 ent = new_HE();
45dea987 1341 HeVAL(ent) = newSVsv(HeVAL(oent));
19692e8d
NC
1342 HeKEY_hek(ent)
1343 = shared ? share_hek_flags(key, len, hash, flags)
1344 : save_hek_flags(key, len, hash, flags);
b56ba0bf
AMS
1345 if (prev)
1346 HeNEXT(prev) = ent;
1347 else
1348 ents[i] = ent;
1349 prev = ent;
1350 HeNEXT(ent) = NULL;
1351 }
1352 }
1353
1354 HvMAX(hv) = hv_max;
1355 HvFILL(hv) = hv_fill;
8aacddc1 1356 HvTOTALKEYS(hv) = HvTOTALKEYS(ohv);
b56ba0bf 1357 HvARRAY(hv) = ents;
1c846c1f 1358 }
b56ba0bf
AMS
1359 else {
1360 /* Iterate over ohv, copying keys and values one at a time. */
b3ac6de7 1361 HE *entry;
b56ba0bf
AMS
1362 I32 riter = HvRITER(ohv);
1363 HE *eiter = HvEITER(ohv);
1364
1365 /* Can we use fewer buckets? (hv_max is always 2^n-1) */
1366 while (hv_max && hv_max + 1 >= hv_fill * 2)
1367 hv_max = hv_max / 2;
1368 HvMAX(hv) = hv_max;
1369
4a76a316 1370 hv_iterinit(ohv);
e16e2ff8 1371 while ((entry = hv_iternext_flags(ohv, 0))) {
19692e8d
NC
1372 hv_store_flags(hv, HeKEY(entry), HeKLEN(entry),
1373 newSVsv(HeVAL(entry)), HeHASH(entry),
1374 HeKFLAGS(entry));
b3ac6de7 1375 }
b56ba0bf
AMS
1376 HvRITER(ohv) = riter;
1377 HvEITER(ohv) = eiter;
b3ac6de7 1378 }
1c846c1f 1379
b3ac6de7
IZ
1380 return hv;
1381}
1382
79072805 1383void
864dbfa3 1384Perl_hv_free_ent(pTHX_ HV *hv, register HE *entry)
79072805 1385{
16bdeea2
GS
1386 SV *val;
1387
68dc0745 1388 if (!entry)
79072805 1389 return;
16bdeea2 1390 val = HeVAL(entry);
257c9e5b 1391 if (val && isGV(val) && GvCVu(val) && HvNAME(hv))
3280af22 1392 PL_sub_generation++; /* may be deletion of method from stash */
16bdeea2 1393 SvREFCNT_dec(val);
68dc0745
PP
1394 if (HeKLEN(entry) == HEf_SVKEY) {
1395 SvREFCNT_dec(HeKEY_sv(entry));
8aacddc1 1396 Safefree(HeKEY_hek(entry));
44a8e56a
PP
1397 }
1398 else if (HvSHAREKEYS(hv))
68dc0745 1399 unshare_hek(HeKEY_hek(entry));
fde52b5c 1400 else
68dc0745 1401 Safefree(HeKEY_hek(entry));
d33b2eba 1402 del_HE(entry);
79072805
LW
1403}
1404
1405void
864dbfa3 1406Perl_hv_delayfree_ent(pTHX_ HV *hv, register HE *entry)
79072805 1407{
68dc0745 1408 if (!entry)
79072805 1409 return;
68dc0745 1410 if (isGV(HeVAL(entry)) && GvCVu(HeVAL(entry)) && HvNAME(hv))
3280af22 1411 PL_sub_generation++; /* may be deletion of method from stash */
68dc0745
PP
1412 sv_2mortal(HeVAL(entry)); /* free between statements */
1413 if (HeKLEN(entry) == HEf_SVKEY) {
1414 sv_2mortal(HeKEY_sv(entry));
1415 Safefree(HeKEY_hek(entry));
44a8e56a
PP
1416 }
1417 else if (HvSHAREKEYS(hv))
68dc0745 1418 unshare_hek(HeKEY_hek(entry));
fde52b5c 1419 else
68dc0745 1420 Safefree(HeKEY_hek(entry));
d33b2eba 1421 del_HE(entry);
79072805
LW
1422}
1423
954c1994
GS
1424/*
1425=for apidoc hv_clear
1426
1427Clears a hash, making it empty.
1428
1429=cut
1430*/
1431
79072805 1432void
864dbfa3 1433Perl_hv_clear(pTHX_ HV *hv)
79072805 1434{
cbec9347 1435 register XPVHV* xhv;
79072805
LW
1436 if (!hv)
1437 return;
49293501 1438
ecae49c0
NC
1439 DEBUG_A(Perl_hv_assert(aTHX_ hv));
1440
34c3c4e3
DM
1441 xhv = (XPVHV*)SvANY(hv);
1442
5f099cb0 1443 if (SvREADONLY(hv) && xhv->xhv_array != NULL) {
34c3c4e3 1444 /* restricted hash: convert all keys to placeholders */
3a676441
JH
1445 I32 i;
1446 HE* entry;
1447 for (i = 0; i <= (I32) xhv->xhv_max; i++) {
1448 entry = ((HE**)xhv->xhv_array)[i];
1449 for (; entry; entry = HeNEXT(entry)) {
1450 /* not already placeholder */
7996736c 1451 if (HeVAL(entry) != &PL_sv_placeholder) {
3a676441
JH
1452 if (HeVAL(entry) && SvREADONLY(HeVAL(entry))) {
1453 SV* keysv = hv_iterkeysv(entry);
1454 Perl_croak(aTHX_
1455 "Attempt to delete readonly key '%"SVf"' from a restricted hash",
1456 keysv);
1457 }
1458 SvREFCNT_dec(HeVAL(entry));
7996736c 1459 HeVAL(entry) = &PL_sv_placeholder;
3a676441
JH
1460 xhv->xhv_placeholders++; /* HvPLACEHOLDERS(hv)++ */
1461 }
34c3c4e3
DM
1462 }
1463 }
df8c6964 1464 goto reset;
49293501
MS
1465 }
1466
463ee0b2 1467 hfreeentries(hv);
8aacddc1 1468 xhv->xhv_placeholders = 0; /* HvPLACEHOLDERS(hv) = 0 */
cbec9347
JH
1469 if (xhv->xhv_array /* HvARRAY(hv) */)
1470 (void)memzero(xhv->xhv_array /* HvARRAY(hv) */,
1471 (xhv->xhv_max+1 /* HvMAX(hv)+1 */) * sizeof(HE*));
a0d0e21e
LW
1472
1473 if (SvRMAGICAL(hv))
1c846c1f 1474 mg_clear((SV*)hv);
574c8022 1475
19692e8d 1476 HvHASKFLAGS_off(hv);
bb443f97 1477 HvREHASH_off(hv);
df8c6964
TP
1478 reset:
1479 HvEITER(hv) = NULL;
79072805
LW
1480}
1481
3540d4ce
AB
1482/*
1483=for apidoc hv_clear_placeholders
1484
1485Clears any placeholders from a hash. If a restricted hash has any of its keys
1486marked as readonly and the key is subsequently deleted, the key is not actually
1487deleted but is marked by assigning it a value of &PL_sv_placeholder. This tags
1488it so it will be ignored by future operations such as iterating over the hash,
1489but will still allow the hash to have a value reaasigned to the key at some
1490future point. This function clears any such placeholder keys from the hash.
1491See Hash::Util::lock_keys() for an example of its use.
1492
1493=cut
1494*/
1495
1496void
1497Perl_hv_clear_placeholders(pTHX_ HV *hv)
1498{
1499 I32 items;
1500 items = (I32)HvPLACEHOLDERS(hv);
1501 if (items) {
1502 HE *entry;
1503 I32 riter = HvRITER(hv);
1504 HE *eiter = HvEITER(hv);
1505 hv_iterinit(hv);
1506 /* This may look suboptimal with the items *after* the iternext, but
1507 it's quite deliberate. We only get here with items==0 if we've
1508 just deleted the last placeholder in the hash. If we've just done
1509 that then it means that the hash is in lazy delete mode, and the
1510 HE is now only referenced in our iterator. If we just quit the loop
1511 and discarded our iterator then the HE leaks. So we do the && the
1512 other way to ensure iternext is called just one more time, which
1513 has the side effect of triggering the lazy delete. */
1514 while ((entry = hv_iternext_flags(hv, HV_ITERNEXT_WANTPLACEHOLDERS))
1515 && items) {
1516 SV *val = hv_iterval(hv, entry);
1517
1518 if (val == &PL_sv_placeholder) {
1519
1520 /* It seems that I have to go back in the front of the hash
1521 API to delete a hash, even though I have a HE structure
1522 pointing to the very entry I want to delete, and could hold
1523 onto the previous HE that points to it. And it's easier to
1524 go in with SVs as I can then specify the precomputed hash,
1525 and don't have fun and games with utf8 keys. */
1526 SV *key = hv_iterkeysv(entry);
1527
1528 hv_delete_ent (hv, key, G_DISCARD, HeHASH(entry));
1529 items--;
1530 }
1531 }
1532 HvRITER(hv) = riter;
1533 HvEITER(hv) = eiter;
1534 }
1535}
1536
76e3520e 1537STATIC void
cea2e8a9 1538S_hfreeentries(pTHX_ HV *hv)
79072805 1539{
a0d0e21e 1540 register HE **array;
68dc0745
PP
1541 register HE *entry;
1542 register HE *oentry = Null(HE*);
a0d0e21e
LW
1543 I32 riter;
1544 I32 max;
79072805
LW
1545
1546 if (!hv)
1547 return;
a0d0e21e 1548 if (!HvARRAY(hv))
79072805 1549 return;
a0d0e21e
LW
1550
1551 riter = 0;
1552 max = HvMAX(hv);
1553 array = HvARRAY(hv);
2f86008e
DM
1554 /* make everyone else think the array is empty, so that the destructors
1555 * called for freed entries can't recusively mess with us */
1556 HvARRAY(hv) = Null(HE**);
1557 HvFILL(hv) = 0;
1558 ((XPVHV*) SvANY(hv))->xhv_keys = 0;
1559
68dc0745 1560 entry = array[0];
a0d0e21e 1561 for (;;) {
68dc0745
PP
1562 if (entry) {
1563 oentry = entry;
1564 entry = HeNEXT(entry);
1565 hv_free_ent(hv, oentry);
a0d0e21e 1566 }
68dc0745 1567 if (!entry) {
a0d0e21e
LW
1568 if (++riter > max)
1569 break;
68dc0745 1570 entry = array[riter];
1c846c1f 1571 }
79072805 1572 }
2f86008e 1573 HvARRAY(hv) = array;
a0d0e21e 1574 (void)hv_iterinit(hv);
79072805
LW
1575}
1576
954c1994
GS
1577/*
1578=for apidoc hv_undef
1579
1580Undefines the hash.
1581
1582=cut
1583*/
1584
79072805 1585void
864dbfa3 1586Perl_hv_undef(pTHX_ HV *hv)
79072805 1587{
cbec9347 1588 register XPVHV* xhv;
79072805
LW
1589 if (!hv)
1590 return;
ecae49c0 1591 DEBUG_A(Perl_hv_assert(aTHX_ hv));
cbec9347 1592 xhv = (XPVHV*)SvANY(hv);
463ee0b2 1593 hfreeentries(hv);
cbec9347 1594 Safefree(xhv->xhv_array /* HvARRAY(hv) */);
85e6fe83 1595 if (HvNAME(hv)) {
7e8961ec
AB
1596 if(PL_stashcache)
1597 hv_delete(PL_stashcache, HvNAME(hv), strlen(HvNAME(hv)), G_DISCARD);
85e6fe83
LW
1598 Safefree(HvNAME(hv));
1599 HvNAME(hv) = 0;
1600 }
cbec9347
JH
1601 xhv->xhv_max = 7; /* HvMAX(hv) = 7 (it's a normal hash) */
1602 xhv->xhv_array = 0; /* HvARRAY(hv) = 0 */
8aacddc1 1603 xhv->xhv_placeholders = 0; /* HvPLACEHOLDERS(hv) = 0 */
a0d0e21e
LW
1604
1605 if (SvRMAGICAL(hv))
1c846c1f 1606 mg_clear((SV*)hv);
79072805
LW
1607}
1608
954c1994
GS
1609/*
1610=for apidoc hv_iterinit
1611
1612Prepares a starting point to traverse a hash table. Returns the number of
1613keys in the hash (i.e. the same as C<HvKEYS(tb)>). The return value is
1c846c1f 1614currently only meaningful for hashes without tie magic.
954c1994
GS
1615
1616NOTE: Before version 5.004_65, C<hv_iterinit> used to return the number of
1617hash buckets that happen to be in use. If you still need that esoteric
1618value, you can get it through the macro C<HvFILL(tb)>.
1619
e16e2ff8 1620
954c1994
GS
1621=cut
1622*/
1623
79072805 1624I32
864dbfa3 1625Perl_hv_iterinit(pTHX_ HV *hv)
79072805 1626{
cbec9347 1627 register XPVHV* xhv;
aa689395
PP
1628 HE *entry;
1629
1630 if (!hv)
cea2e8a9 1631 Perl_croak(aTHX_ "Bad hash");
cbec9347
JH
1632 xhv = (XPVHV*)SvANY(hv);
1633 entry = xhv->xhv_eiter; /* HvEITER(hv) */
72940dca
PP
1634 if (entry && HvLAZYDEL(hv)) { /* was deleted earlier? */
1635 HvLAZYDEL_off(hv);
68dc0745 1636 hv_free_ent(hv, entry);
72940dca 1637 }
cbec9347
JH
1638 xhv->xhv_riter = -1; /* HvRITER(hv) = -1 */
1639 xhv->xhv_eiter = Null(HE*); /* HvEITER(hv) = Null(HE*) */
1640 /* used to be xhv->xhv_fill before 5.004_65 */
8aacddc1 1641 return XHvTOTALKEYS(xhv);
79072805 1642}
954c1994
GS
1643/*
1644=for apidoc hv_iternext
1645
1646Returns entries from a hash iterator. See C<hv_iterinit>.
1647
fe7bca90
NC
1648You may call C<hv_delete> or C<hv_delete_ent> on the hash entry that the
1649iterator currently points to, without losing your place or invalidating your
1650iterator. Note that in this case the current entry is deleted from the hash
1651with your iterator holding the last reference to it. Your iterator is flagged
1652to free the entry on the next call to C<hv_iternext>, so you must not discard
1653your iterator immediately else the entry will leak - call C<hv_iternext> to
1654trigger the resource deallocation.
1655
954c1994
GS
1656=cut
1657*/
1658
79072805 1659HE *
864dbfa3 1660Perl_hv_iternext(pTHX_ HV *hv)
79072805 1661{
e16e2ff8
NC
1662 return hv_iternext_flags(hv, 0);
1663}
1664
1665/*
fe7bca90
NC
1666=for apidoc hv_iternext_flags
1667
1668Returns entries from a hash iterator. See C<hv_iterinit> and C<hv_iternext>.
1669The C<flags> value will normally be zero; if HV_ITERNEXT_WANTPLACEHOLDERS is
1670set the placeholders keys (for restricted hashes) will be returned in addition
1671to normal keys. By default placeholders are automatically skipped over.
7996736c
MHM
1672Currently a placeholder is implemented with a value that is
1673C<&Perl_sv_placeholder>. Note that the implementation of placeholders and
fe7bca90
NC
1674restricted hashes may change, and the implementation currently is
1675insufficiently abstracted for any change to be tidy.
e16e2ff8 1676
fe7bca90 1677=cut
e16e2ff8
NC
1678*/
1679
1680HE *
1681Perl_hv_iternext_flags(pTHX_ HV *hv, I32 flags)
1682{
cbec9347 1683 register XPVHV* xhv;
79072805 1684 register HE *entry;
a0d0e21e 1685 HE *oldentry;
463ee0b2 1686 MAGIC* mg;
79072805
LW
1687
1688 if (!hv)
cea2e8a9 1689 Perl_croak(aTHX_ "Bad hash");
cbec9347
JH
1690 xhv = (XPVHV*)SvANY(hv);
1691 oldentry = entry = xhv->xhv_eiter; /* HvEITER(hv) */
463ee0b2 1692
14befaf4 1693 if ((mg = SvTIED_mg((SV*)hv, PERL_MAGIC_tied))) {
8990e307 1694 SV *key = sv_newmortal();
cd1469e6 1695 if (entry) {
fde52b5c 1696 sv_setsv(key, HeSVKEY_force(entry));
cd1469e6
PP
1697 SvREFCNT_dec(HeSVKEY(entry)); /* get rid of previous key */
1698 }
a0d0e21e 1699 else {
ff68c719 1700 char *k;
bbce6d69 1701 HEK *hek;
ff68c719 1702
cbec9347
JH
1703 /* one HE per MAGICAL hash */
1704 xhv->xhv_eiter = entry = new_HE(); /* HvEITER(hv) = new_HE() */
4633a7c4 1705 Zero(entry, 1, HE);
ff68c719
PP
1706 Newz(54, k, HEK_BASESIZE + sizeof(SV*), char);
1707 hek = (HEK*)k;
1708 HeKEY_hek(entry) = hek;
fde52b5c 1709 HeKLEN(entry) = HEf_SVKEY;
a0d0e21e
LW
1710 }
1711 magic_nextpack((SV*) hv,mg,key);
8aacddc1 1712 if (SvOK(key)) {
cd1469e6 1713 /* force key to stay around until next time */
bbce6d69
PP
1714 HeSVKEY_set(entry, SvREFCNT_inc(key));
1715 return entry; /* beware, hent_val is not set */
8aacddc1 1716 }
fde52b5c
PP
1717 if (HeVAL(entry))
1718 SvREFCNT_dec(HeVAL(entry));
ff68c719 1719 Safefree(HeKEY_hek(entry));
d33b2eba 1720 del_HE(entry);
cbec9347 1721 xhv->xhv_eiter = Null(HE*); /* HvEITER(hv) = Null(HE*) */
463ee0b2 1722 return Null(HE*);
79072805 1723 }
f675dbe5 1724#ifdef DYNAMIC_ENV_FETCH /* set up %ENV for iteration */
cbec9347 1725 if (!entry && SvRMAGICAL((SV*)hv) && mg_find((SV*)hv, PERL_MAGIC_env))
f675dbe5
CB
1726 prime_env_iter();
1727#endif
463ee0b2 1728
cbec9347
JH
1729 if (!xhv->xhv_array /* !HvARRAY(hv) */)
1730 Newz(506, xhv->xhv_array /* HvARRAY(hv) */,
1731 PERL_HV_ARRAY_ALLOC_BYTES(xhv->xhv_max+1 /* HvMAX(hv)+1 */),
1732 char);
015a5f36 1733 /* At start of hash, entry is NULL. */
fde52b5c 1734 if (entry)
8aacddc1 1735 {
fde52b5c 1736 entry = HeNEXT(entry);
e16e2ff8
NC
1737 if (!(flags & HV_ITERNEXT_WANTPLACEHOLDERS)) {
1738 /*
1739 * Skip past any placeholders -- don't want to include them in
1740 * any iteration.
1741 */
7996736c 1742 while (entry && HeVAL(entry) == &PL_sv_placeholder) {
e16e2ff8
NC
1743 entry = HeNEXT(entry);
1744 }
8aacddc1
NIS
1745 }
1746 }
fde52b5c 1747 while (!entry) {
015a5f36
NC
1748 /* OK. Come to the end of the current list. Grab the next one. */
1749
cbec9347 1750 xhv->xhv_riter++; /* HvRITER(hv)++ */
eb160463 1751 if (xhv->xhv_riter > (I32)xhv->xhv_max /* HvRITER(hv) > HvMAX(hv) */) {
015a5f36 1752 /* There is no next one. End of the hash. */
cbec9347 1753 xhv->xhv_riter = -1; /* HvRITER(hv) = -1 */
fde52b5c 1754 break;
79072805 1755 }
cbec9347
JH
1756 /* entry = (HvARRAY(hv))[HvRITER(hv)]; */
1757 entry = ((HE**)xhv->xhv_array)[xhv->xhv_riter];
8aacddc1 1758
e16e2ff8 1759 if (!(flags & HV_ITERNEXT_WANTPLACEHOLDERS)) {
015a5f36
NC
1760 /* If we have an entry, but it's a placeholder, don't count it.
1761 Try the next. */
7996736c 1762 while (entry && HeVAL(entry) == &PL_sv_placeholder)
015a5f36
NC
1763 entry = HeNEXT(entry);
1764 }
1765 /* Will loop again if this linked list starts NULL
1766 (for HV_ITERNEXT_WANTPLACEHOLDERS)
1767 or if we run through it and find only placeholders. */
fde52b5c 1768 }
79072805 1769
72940dca
PP
1770 if (oldentry && HvLAZYDEL(hv)) { /* was deleted earlier? */
1771 HvLAZYDEL_off(hv);
68dc0745 1772 hv_free_ent(hv, oldentry);
72940dca 1773 }
a0d0e21e 1774
fdcd69b6
NC
1775 /*if (HvREHASH(hv) && entry && !HeKREHASH(entry))
1776 PerlIO_printf(PerlIO_stderr(), "Awooga %p %p\n", hv, entry);*/
1777
cbec9347 1778 xhv->xhv_eiter = entry; /* HvEITER(hv) = entry */
79072805
LW
1779 return entry;
1780}
1781
954c1994
GS
1782/*
1783=for apidoc hv_iterkey
1784
1785Returns the key from the current position of the hash iterator. See
1786C<hv_iterinit>.
1787
1788=cut
1789*/
1790
79072805 1791char *
864dbfa3 1792Perl_hv_iterkey(pTHX_ register HE *entry, I32 *retlen)
79072805 1793{
fde52b5c 1794 if (HeKLEN(entry) == HEf_SVKEY) {
fb73857a
PP
1795 STRLEN len;
1796 char *p = SvPV(HeKEY_sv(entry), len);
1797 *retlen = len;
1798 return p;
fde52b5c
PP
1799 }
1800 else {
1801 *retlen = HeKLEN(entry);
1802 return HeKEY(entry);
1803 }
1804}
1805
1806/* unlike hv_iterval(), this always returns a mortal copy of the key */
954c1994
GS
1807/*
1808=for apidoc hv_iterkeysv
1809
1810Returns the key as an C<SV*> from the current position of the hash
1811iterator. The return value will always be a mortal copy of the key. Also
1812see C<hv_iterinit>.
1813
1814=cut
1815*/
1816
fde52b5c 1817SV *
864dbfa3 1818Perl_hv_iterkeysv(pTHX_ register HE *entry)
fde52b5c 1819{
19692e8d
NC
1820 if (HeKLEN(entry) != HEf_SVKEY) {
1821 HEK *hek = HeKEY_hek(entry);
1822 int flags = HEK_FLAGS(hek);
1823 SV *sv;
1824
1825 if (flags & HVhek_WASUTF8) {
1826 /* Trouble :-)
1827 Andreas would like keys he put in as utf8 to come back as utf8
1828 */
1829 STRLEN utf8_len = HEK_LEN(hek);
2e5dfef7 1830 U8 *as_utf8 = bytes_to_utf8 ((U8*)HEK_KEY(hek), &utf8_len);
19692e8d 1831
2e5dfef7 1832 sv = newSVpvn ((char*)as_utf8, utf8_len);
19692e8d 1833 SvUTF8_on (sv);
c193270f 1834 Safefree (as_utf8); /* bytes_to_utf8() allocates a new string */
4b5190b5
NC
1835 } else if (flags & HVhek_REHASH) {
1836 /* We don't have a pointer to the hv, so we have to replicate the
1837 flag into every HEK. This hv is using custom a hasing
1838 algorithm. Hence we can't return a shared string scalar, as
1839 that would contain the (wrong) hash value, and might get passed
1840 into an hv routine with a regular hash */
1841
1842 sv = newSVpvn (HEK_KEY(hek), HEK_LEN(hek));
1843 if (HEK_UTF8(hek))
1844 SvUTF8_on (sv);
1845 } else {
19692e8d
NC
1846 sv = newSVpvn_share(HEK_KEY(hek),
1847 (HEK_UTF8(hek) ? -HEK_LEN(hek) : HEK_LEN(hek)),
1848 HEK_HASH(hek));
1849 }
1850 return sv_2mortal(sv);
1851 }
1852 return sv_mortalcopy(HeKEY_sv(entry));
79072805
LW
1853}
1854
954c1994
GS
1855/*
1856=for apidoc hv_iterval
1857
1858Returns the value from the current position of the hash iterator. See
1859C<hv_iterkey>.
1860
1861=cut
1862*/
1863
79072805 1864SV *
864dbfa3 1865Perl_hv_iterval(pTHX_ HV *hv, register HE *entry)
79072805 1866{
8990e307 1867 if (SvRMAGICAL(hv)) {
14befaf4 1868 if (mg_find((SV*)hv, PERL_MAGIC_tied)) {
8990e307 1869 SV* sv = sv_newmortal();
bbce6d69
PP
1870 if (HeKLEN(entry) == HEf_SVKEY)
1871 mg_copy((SV*)hv, sv, (char*)HeKEY_sv(entry), HEf_SVKEY);
1872 else mg_copy((SV*)hv, sv, HeKEY(entry), HeKLEN(entry));
463ee0b2
LW
1873 return sv;
1874 }
79072805 1875 }
fde52b5c 1876 return HeVAL(entry);
79072805
LW
1877}
1878
954c1994
GS
1879/*
1880=for apidoc hv_iternextsv
1881
1882Performs an C<hv_iternext>, C<hv_iterkey>, and C<hv_iterval> in one
1883operation.
1884
1885=cut
1886*/
1887
a0d0e21e 1888SV *
864dbfa3 1889Perl_hv_iternextsv(pTHX_ HV *hv, char **key, I32 *retlen)
a0d0e21e
LW
1890{
1891 HE *he;
e16e2ff8 1892 if ( (he = hv_iternext_flags(hv, 0)) == NULL)
a0d0e21e
LW
1893 return NULL;
1894 *key = hv_iterkey(he, retlen);
1895 return hv_iterval(hv, he);
1896}
1897
954c1994
GS
1898/*
1899=for apidoc hv_magic
1900
1901Adds magic to a hash. See C<sv_magic>.
1902
1903=cut
1904*/
1905
79072805 1906void
864dbfa3 1907Perl_hv_magic(pTHX_ HV *hv, GV *gv, int how)
79072805 1908{
a0d0e21e 1909 sv_magic((SV*)hv, (SV*)gv, how, Nullch, 0);
79072805 1910}
fde52b5c 1911
37d85e3a
JH
1912#if 0 /* use the macro from hv.h instead */
1913
bbce6d69 1914char*
864dbfa3 1915Perl_sharepvn(pTHX_ const char *sv, I32 len, U32 hash)
bbce6d69 1916{
ff68c719 1917 return HEK_KEY(share_hek(sv, len, hash));
bbce6d69
PP
1918}
1919
37d85e3a
JH
1920#endif
1921
bbce6d69 1922/* possibly free a shared string if no one has access to it
fde52b5c
PP
1923 * len and hash must both be valid for str.
1924 */
bbce6d69 1925void
864dbfa3 1926Perl_unsharepvn(pTHX_ const char *str, I32 len, U32 hash)
fde52b5c 1927{
19692e8d
NC
1928 unshare_hek_or_pvn (NULL, str, len, hash);
1929}
1930
1931
1932void
1933Perl_unshare_hek(pTHX_ HEK *hek)
1934{
1935 unshare_hek_or_pvn(hek, NULL, 0, 0);
1936}
1937
1938/* possibly free a shared string if no one has access to it
1939 hek if non-NULL takes priority over the other 3, else str, len and hash
1940 are used. If so, len and hash must both be valid for str.
1941 */
df132699 1942STATIC void
19692e8d
NC
1943S_unshare_hek_or_pvn(pTHX_ HEK *hek, const char *str, I32 len, U32 hash)
1944{
cbec9347 1945 register XPVHV* xhv;
fde52b5c
PP
1946 register HE *entry;
1947 register HE **oentry;
1948 register I32 i = 1;
1949 I32 found = 0;
c3654f1a 1950 bool is_utf8 = FALSE;
19692e8d 1951 int k_flags = 0;
f9a63242 1952 const char *save = str;
c3654f1a 1953
19692e8d
NC
1954 if (hek) {
1955 hash = HEK_HASH(hek);
1956 } else if (len < 0) {
1957 STRLEN tmplen = -len;
1958 is_utf8 = TRUE;
1959 /* See the note in hv_fetch(). --jhi */
1960 str = (char*)bytes_from_utf8((U8*)str, &tmplen, &is_utf8);
1961 len = tmplen;
1962 if (is_utf8)
1963 k_flags = HVhek_UTF8;
1964 if (str != save)
1965 k_flags |= HVhek_WASUTF8 | HVhek_FREEKEY;
c3654f1a 1966 }
1c846c1f 1967
fde52b5c 1968 /* what follows is the moral equivalent of:
6b88bc9c 1969 if ((Svp = hv_fetch(PL_strtab, tmpsv, FALSE, hash))) {
bbce6d69 1970 if (--*Svp == Nullsv)
6b88bc9c 1971 hv_delete(PL_strtab, str, len, G_DISCARD, hash);
bbce6d69 1972 } */
cbec9347 1973 xhv = (XPVHV*)SvANY(PL_strtab);
fde52b5c 1974 /* assert(xhv_array != 0) */
5f08fbcd 1975 LOCK_STRTAB_MUTEX;
cbec9347
JH
1976 /* oentry = &(HvARRAY(hv))[hash & (I32) HvMAX(hv)]; */
1977 oentry = &((HE**)xhv->xhv_array)[hash & (I32) xhv->xhv_max];
19692e8d
NC
1978 if (hek) {
1979 for (entry = *oentry; entry; i=0, oentry = &HeNEXT(entry), entry = *oentry) {
1980 if (HeKEY_hek(entry) != hek)
1981 continue;
1982 found = 1;
1983 break;
1984 }
1985 } else {
1986 int flags_masked = k_flags & HVhek_MASK;
1987 for (entry = *oentry; entry; i=0, oentry = &HeNEXT(entry), entry = *oentry) {
1988 if (HeHASH(entry) != hash) /* strings can't be equal */
1989 continue;
1990 if (HeKLEN(entry) != len)
1991 continue;
1992 if (HeKEY(entry) != str && memNE(HeKEY(entry),str,len)) /* is this it? */
1993 continue;
1994 if (HeKFLAGS(entry) != flags_masked)
1995 continue;
1996 found = 1;
1997 break;
1998 }
1999 }
2000
2001 if (found) {
2002 if (--HeVAL(entry) == Nullsv) {
2003 *oentry = HeNEXT(entry);
2004 if (i && !*oentry)
2005 xhv->xhv_fill--; /* HvFILL(hv)-- */
2006 Safefree(HeKEY_hek(entry));
2007 del_HE(entry);
2008 xhv->xhv_keys--; /* HvKEYS(hv)-- */
2009 }
fde52b5c 2010 }
19692e8d 2011
333f433b 2012 UNLOCK_STRTAB_MUTEX;
411caa50 2013 if (!found && ckWARN_d(WARN_INTERNAL))
19692e8d
NC
2014 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
2015 "Attempt to free non-existent shared string '%s'%s",
2016 hek ? HEK_KEY(hek) : str,
2017 (k_flags & HVhek_UTF8) ? " (utf8)" : "");
2018 if (k_flags & HVhek_FREEKEY)
2019 Safefree(str);
fde52b5c
PP
2020}
2021
bbce6d69
PP
2022/* get a (constant) string ptr from the global string table
2023 * string will get added if it is not already there.
fde52b5c
PP
2024 * len and hash must both be valid for str.
2025 */
bbce6d69 2026HEK *
864dbfa3 2027Perl_share_hek(pTHX_ const char *str, I32 len, register U32 hash)
fde52b5c 2028{
da58a35d 2029 bool is_utf8 = FALSE;
19692e8d 2030 int flags = 0;
f9a63242 2031 const char *save = str;
da58a35d
JH
2032
2033 if (len < 0) {
77caf834 2034 STRLEN tmplen = -len;
da58a35d 2035 is_utf8 = TRUE;
77caf834
JH
2036 /* See the note in hv_fetch(). --jhi */
2037 str = (char*)bytes_from_utf8((U8*)str, &tmplen, &is_utf8);
2038 len = tmplen;
19692e8d
NC
2039 /* If we were able to downgrade here, then than means that we were passed
2040 in a key which only had chars 0-255, but was utf8 encoded. */
2041 if (is_utf8)
2042 flags = HVhek_UTF8;
2043 /* If we found we were able to downgrade the string to bytes, then
2044 we should flag that it needs upgrading on keys or each. Also flag
2045 that we need share_hek_flags to free the string. */
2046 if (str != save)
2047 flags |= HVhek_WASUTF8 | HVhek_FREEKEY;
2048 }
2049
2050 return share_hek_flags (str, len, hash, flags);
2051}
2052
df132699 2053STATIC HEK *
19692e8d
NC
2054S_share_hek_flags(pTHX_ const char *str, I32 len, register U32 hash, int flags)
2055{
2056 register XPVHV* xhv;
2057 register HE *entry;
2058 register HE **oentry;
2059 register I32 i = 1;
2060 I32 found = 0;
2061 int flags_masked = flags & HVhek_MASK;
bbce6d69 2062
fde52b5c 2063 /* what follows is the moral equivalent of:
1c846c1f 2064
6b88bc9c 2065 if (!(Svp = hv_fetch(PL_strtab, str, len, FALSE)))
8aacddc1 2066 hv_store(PL_strtab, str, len, Nullsv, hash);
fdcd69b6
NC
2067
2068 Can't rehash the shared string table, so not sure if it's worth
2069 counting the number of entries in the linked list
bbce6d69 2070 */
cbec9347 2071 xhv = (XPVHV*)SvANY(PL_strtab);
fde52b5c 2072 /* assert(xhv_array != 0) */
5f08fbcd 2073 LOCK_STRTAB_MUTEX;
cbec9347
JH
2074 /* oentry = &(HvARRAY(hv))[hash & (I32) HvMAX(hv)]; */
2075 oentry = &((HE**)xhv->xhv_array)[hash & (I32) xhv->xhv_max];
bbce6d69 2076 for (entry = *oentry; entry; i=0, entry = HeNEXT(entry)) {
fde52b5c
PP
2077 if (HeHASH(entry) != hash) /* strings can't be equal */
2078 continue;
2079 if (HeKLEN(entry) != len)
2080 continue;
1c846c1f 2081 if (HeKEY(entry) != str && memNE(HeKEY(entry),str,len)) /* is this it? */
fde52b5c 2082 continue;
19692e8d 2083 if (HeKFLAGS(entry) != flags_masked)
c3654f1a 2084 continue;
fde52b5c 2085 found = 1;
fde52b5c
PP
2086 break;
2087 }
bbce6d69 2088 if (!found) {
d33b2eba 2089 entry = new_HE();
dcf933a4 2090 HeKEY_hek(entry) = save_hek_flags(str, len, hash, flags_masked);
bbce6d69
PP
2091 HeVAL(entry) = Nullsv;
2092 HeNEXT(entry) = *oentry;
2093 *oentry = entry;
cbec9347 2094 xhv->xhv_keys++; /* HvKEYS(hv)++ */
bbce6d69 2095 if (i) { /* initial entry? */
cbec9347 2096 xhv->xhv_fill++; /* HvFILL(hv)++ */
4c9cc595 2097 } else if (xhv->xhv_keys > (IV)xhv->xhv_max /* HvKEYS(hv) > HvMAX(hv) */) {
cbec9347 2098 hsplit(PL_strtab);
bbce6d69
PP
2099 }
2100 }
2101
2102 ++HeVAL(entry); /* use value slot as REFCNT */
5f08fbcd 2103 UNLOCK_STRTAB_MUTEX;
19692e8d
NC
2104
2105 if (flags & HVhek_FREEKEY)
f9a63242 2106 Safefree(str);
19692e8d 2107
ff68c719 2108 return HeKEY_hek(entry);
fde52b5c 2109}
ecae49c0
NC
2110
2111
2112/*
2113=for apidoc hv_assert
2114
2115Check that a hash is in an internally consistent state.
2116
2117=cut
2118*/
2119
2120void
2121Perl_hv_assert(pTHX_ HV *hv)
2122{
2123 HE* entry;
2124 int withflags = 0;
2125 int placeholders = 0;
2126 int real = 0;
2127 int bad = 0;
2128 I32 riter = HvRITER(hv);
2129 HE *eiter = HvEITER(hv);
2130
2131 (void)hv_iterinit(hv);
2132
2133 while ((entry = hv_iternext_flags(hv, HV_ITERNEXT_WANTPLACEHOLDERS))) {
2134 /* sanity check the values */
2135 if (HeVAL(entry) == &PL_sv_placeholder) {
2136 placeholders++;
2137 } else {
2138 real++;
2139 }
2140 /* sanity check the keys */
2141 if (HeSVKEY(entry)) {
2142 /* Don't know what to check on SV keys. */
2143 } else if (HeKUTF8(entry)) {
2144 withflags++;
2145 if (HeKWASUTF8(entry)) {
2146 PerlIO_printf(Perl_debug_log,
2147 "hash key has both WASUFT8 and UTF8: '%.*s'\n",
2148 (int) HeKLEN(entry), HeKEY(entry));
2149 bad = 1;
2150 }
2151 } else if (HeKWASUTF8(entry)) {
2152 withflags++;
2153 }
2154 }
2155 if (!SvTIED_mg((SV*)hv, PERL_MAGIC_tied)) {
2156 if (HvUSEDKEYS(hv) != real) {
2157 PerlIO_printf(Perl_debug_log, "Count %d key(s), but hash reports %d\n",
2158 (int) real, (int) HvUSEDKEYS(hv));
2159 bad = 1;
2160 }
2161 if (HvPLACEHOLDERS(hv) != placeholders) {
2162 PerlIO_printf(Perl_debug_log,
2163 "Count %d placeholder(s), but hash reports %d\n",
2164 (int) placeholders, (int) HvPLACEHOLDERS(hv));
2165 bad = 1;
2166 }
2167 }
2168 if (withflags && ! HvHASKFLAGS(hv)) {
2169 PerlIO_printf(Perl_debug_log,
2170 "Hash has HASKFLAGS off but I count %d key(s) with flags\n",
2171 withflags);
2172 bad = 1;
2173 }
2174 if (bad) {
2175 sv_dump((SV *)hv);
2176 }
2177 HvRITER(hv) = riter; /* Restore hash iterator state */
2178 HvEITER(hv) = eiter;
2179}