This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Comment to record why we can't clear placeholders in hsplit
[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
8f8d40ab 372STATIC HE *
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
8f8d40ab 869STATIC SV *
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 }
4b5190b5 963 }
fde52b5c 964
7a9669ca
NC
965 masked_flags = (k_flags & HVhek_MASK);
966
cbec9347
JH
967 /* oentry = &(HvARRAY(hv))[hash & (I32) HvMAX(hv)]; */
968 oentry = &((HE**)xhv->xhv_array)[hash & (I32) xhv->xhv_max];
fde52b5c
PP
969 entry = *oentry;
970 i = 1;
971 for (; entry; i=0, oentry = &HeNEXT(entry), entry = *oentry) {
972 if (HeHASH(entry) != hash) /* strings can't be equal */
973 continue;
eb160463 974 if (HeKLEN(entry) != (I32)klen)
fde52b5c 975 continue;
1c846c1f 976 if (HeKEY(entry) != key && memNE(HeKEY(entry),key,klen)) /* is this it? */
fde52b5c 977 continue;
7a9669ca 978 if ((HeKFLAGS(entry) ^ masked_flags) & HVhek_UTF8)
c3654f1a 979 continue;
19692e8d
NC
980 if (k_flags & HVhek_FREEKEY)
981 Safefree(key);
8aacddc1
NIS
982
983 /* if placeholder is here, it's already been deleted.... */
7996736c 984 if (HeVAL(entry) == &PL_sv_placeholder)
8aacddc1 985 {
03fed38d 986 return Nullsv;
8aacddc1
NIS
987 }
988 else if (SvREADONLY(hv) && HeVAL(entry) && SvREADONLY(HeVAL(entry))) {
2393f1b9
JH
989 S_hv_notallowed(aTHX_ k_flags, key, klen,
990 "delete readonly key '%"SVf"' from"
991 );
8aacddc1
NIS
992 }
993
cd6d36ac 994 if (d_flags & G_DISCARD)
fde52b5c 995 sv = Nullsv;
94f7643d 996 else {
79d01fbf 997 sv = sv_2mortal(HeVAL(entry));
7996736c 998 HeVAL(entry) = &PL_sv_placeholder;
94f7643d 999 }
8aacddc1
NIS
1000
1001 /*
1002 * If a restricted hash, rather than really deleting the entry, put
1003 * a placeholder there. This marks the key as being "approved", so
1004 * we can still access via not-really-existing key without raising
1005 * an error.
1006 */
1007 if (SvREADONLY(hv)) {
754604c4 1008 SvREFCNT_dec(HeVAL(entry));
7996736c 1009 HeVAL(entry) = &PL_sv_placeholder;
8aacddc1
NIS
1010 /* We'll be saving this slot, so the number of allocated keys
1011 * doesn't go down, but the number placeholders goes up */
1012 xhv->xhv_placeholders++; /* HvPLACEHOLDERS(hv)++ */
1013 } else {
a26e96df
NIS
1014 *oentry = HeNEXT(entry);
1015 if (i && !*oentry)
1016 xhv->xhv_fill--; /* HvFILL(hv)-- */
8aacddc1
NIS
1017 if (entry == xhv->xhv_eiter /* HvEITER(hv) */)
1018 HvLAZYDEL_on(hv);
1019 else
1020 hv_free_ent(hv, entry);
1021 xhv->xhv_keys--; /* HvKEYS(hv)-- */
574c8022 1022 if (xhv->xhv_keys == 0)
19692e8d 1023 HvHASKFLAGS_off(hv);
8aacddc1 1024 }
79072805
LW
1025 return sv;
1026 }
8aacddc1 1027 if (SvREADONLY(hv)) {
2393f1b9
JH
1028 S_hv_notallowed(aTHX_ k_flags, key, klen,
1029 "delete disallowed key '%"SVf"' from"
1030 );
8aacddc1
NIS
1031 }
1032
19692e8d 1033 if (k_flags & HVhek_FREEKEY)
f9a63242 1034 Safefree(key);
79072805 1035 return Nullsv;
79072805
LW
1036}
1037
76e3520e 1038STATIC void
cea2e8a9 1039S_hsplit(pTHX_ HV *hv)
79072805 1040{
752f6c60
NC
1041 /* Can't make this clear any placeholders first for non-restricted hashes,
1042 as Storable rebuilds restricted hashes by putting in all the
1043 placeholders (first) before turning on the readonly flag. Hence midway
1044 through restoring the hash there are placeholders which need to remain
1045 even though the hash isn't (currently) flagged as restricted. */
cbec9347
JH
1046 register XPVHV* xhv = (XPVHV*)SvANY(hv);
1047 I32 oldsize = (I32) xhv->xhv_max+1; /* HvMAX(hv)+1 (sick) */
79072805
LW
1048 register I32 newsize = oldsize * 2;
1049 register I32 i;
cbec9347 1050 register char *a = xhv->xhv_array; /* HvARRAY(hv) */
72311751
GS
1051 register HE **aep;
1052 register HE **bep;
79072805
LW
1053 register HE *entry;
1054 register HE **oentry;
4b5190b5
NC
1055 int longest_chain = 0;
1056 int was_shared;
79072805 1057
3280af22 1058 PL_nomemok = TRUE;
8d6dde3e 1059#if defined(STRANGE_MALLOC) || defined(MYMALLOC)
d18c6117 1060 Renew(a, PERL_HV_ARRAY_ALLOC_BYTES(newsize), char);
422a93e5 1061 if (!a) {
4a33f861 1062 PL_nomemok = FALSE;
422a93e5
GA
1063 return;
1064 }
4633a7c4 1065#else
d18c6117 1066 New(2, a, PERL_HV_ARRAY_ALLOC_BYTES(newsize), char);
422a93e5 1067 if (!a) {
3280af22 1068 PL_nomemok = FALSE;
422a93e5
GA
1069 return;
1070 }
cbec9347 1071 Copy(xhv->xhv_array /* HvARRAY(hv) */, a, oldsize * sizeof(HE*), char);
fba3b22e 1072 if (oldsize >= 64) {
cbec9347
JH
1073 offer_nice_chunk(xhv->xhv_array /* HvARRAY(hv) */,
1074 PERL_HV_ARRAY_ALLOC_BYTES(oldsize));
4633a7c4
LW
1075 }
1076 else
cbec9347 1077 Safefree(xhv->xhv_array /* HvARRAY(hv) */);
4633a7c4
LW
1078#endif
1079
3280af22 1080 PL_nomemok = FALSE;
72311751 1081 Zero(&a[oldsize * sizeof(HE*)], (newsize-oldsize) * sizeof(HE*), char); /* zero 2nd half*/
cbec9347
JH
1082 xhv->xhv_max = --newsize; /* HvMAX(hv) = --newsize */
1083 xhv->xhv_array = a; /* HvARRAY(hv) = a */
72311751 1084 aep = (HE**)a;
79072805 1085
72311751 1086 for (i=0; i<oldsize; i++,aep++) {
4b5190b5
NC
1087 int left_length = 0;
1088 int right_length = 0;
1089
72311751 1090 if (!*aep) /* non-existent */
79072805 1091 continue;
72311751
GS
1092 bep = aep+oldsize;
1093 for (oentry = aep, entry = *aep; entry; entry = *oentry) {
eb160463 1094 if ((HeHASH(entry) & newsize) != (U32)i) {
fde52b5c 1095 *oentry = HeNEXT(entry);
72311751
GS
1096 HeNEXT(entry) = *bep;
1097 if (!*bep)
cbec9347 1098 xhv->xhv_fill++; /* HvFILL(hv)++ */
72311751 1099 *bep = entry;
4b5190b5 1100 right_length++;
79072805
LW
1101 continue;
1102 }
4b5190b5 1103 else {
fde52b5c 1104 oentry = &HeNEXT(entry);
4b5190b5
NC
1105 left_length++;
1106 }
79072805 1107 }
72311751 1108 if (!*aep) /* everything moved */
cbec9347 1109 xhv->xhv_fill--; /* HvFILL(hv)-- */
4b5190b5
NC
1110 /* I think we don't actually need to keep track of the longest length,
1111 merely flag if anything is too long. But for the moment while
1112 developing this code I'll track it. */
1113 if (left_length > longest_chain)
1114 longest_chain = left_length;
1115 if (right_length > longest_chain)
1116 longest_chain = right_length;
1117 }
1118
1119
1120 /* Pick your policy for "hashing isn't working" here: */
fdcd69b6 1121 if (longest_chain <= HV_MAX_LENGTH_BEFORE_SPLIT /* split worked? */
4b5190b5
NC
1122 || HvREHASH(hv)) {
1123 return;
79072805 1124 }
4b5190b5
NC
1125
1126 if (hv == PL_strtab) {
1127 /* Urg. Someone is doing something nasty to the string table.
1128 Can't win. */
1129 return;
1130 }
1131
1132 /* Awooga. Awooga. Pathological data. */
fdcd69b6 1133 /*PerlIO_printf(PerlIO_stderr(), "%p %d of %d with %d/%d buckets\n", hv,
4b5190b5
NC
1134 longest_chain, HvTOTALKEYS(hv), HvFILL(hv), 1+HvMAX(hv));*/
1135
1136 ++newsize;
1137 Newz(2, a, PERL_HV_ARRAY_ALLOC_BYTES(newsize), char);
1138 was_shared = HvSHAREKEYS(hv);
1139
1140 xhv->xhv_fill = 0;
1141 HvSHAREKEYS_off(hv);
1142 HvREHASH_on(hv);
1143
1144 aep = (HE **) xhv->xhv_array;
1145
1146 for (i=0; i<newsize; i++,aep++) {
1147 entry = *aep;
1148 while (entry) {
1149 /* We're going to trash this HE's next pointer when we chain it
1150 into the new hash below, so store where we go next. */
1151 HE *next = HeNEXT(entry);
1152 UV hash;
1153
1154 /* Rehash it */
1155 PERL_HASH_INTERNAL(hash, HeKEY(entry), HeKLEN(entry));
1156
1157 if (was_shared) {
1158 /* Unshare it. */
1159 HEK *new_hek
1160 = save_hek_flags(HeKEY(entry), HeKLEN(entry),
1161 hash, HeKFLAGS(entry));
1162 unshare_hek (HeKEY_hek(entry));
1163 HeKEY_hek(entry) = new_hek;
1164 } else {
1165 /* Not shared, so simply write the new hash in. */
1166 HeHASH(entry) = hash;
1167 }
1168 /*PerlIO_printf(PerlIO_stderr(), "%d ", HeKFLAGS(entry));*/
1169 HEK_REHASH_on(HeKEY_hek(entry));
1170 /*PerlIO_printf(PerlIO_stderr(), "%d\n", HeKFLAGS(entry));*/
1171
1172 /* Copy oentry to the correct new chain. */
1173 bep = ((HE**)a) + (hash & (I32) xhv->xhv_max);
1174 if (!*bep)
1175 xhv->xhv_fill++; /* HvFILL(hv)++ */
1176 HeNEXT(entry) = *bep;
1177 *bep = entry;
1178
1179 entry = next;
1180 }
1181 }
1182 Safefree (xhv->xhv_array);
1183 xhv->xhv_array = a; /* HvARRAY(hv) = a */
79072805
LW
1184}
1185
72940dca 1186void
864dbfa3 1187Perl_hv_ksplit(pTHX_ HV *hv, IV newmax)
72940dca 1188{
cbec9347
JH
1189 register XPVHV* xhv = (XPVHV*)SvANY(hv);
1190 I32 oldsize = (I32) xhv->xhv_max+1; /* HvMAX(hv)+1 (sick) */
72940dca
PP
1191 register I32 newsize;
1192 register I32 i;
1193 register I32 j;
72311751
GS
1194 register char *a;
1195 register HE **aep;
72940dca
PP
1196 register HE *entry;
1197 register HE **oentry;
1198
1199 newsize = (I32) newmax; /* possible truncation here */
1200 if (newsize != newmax || newmax <= oldsize)
1201 return;
1202 while ((newsize & (1 + ~newsize)) != newsize) {
1203 newsize &= ~(newsize & (1 + ~newsize)); /* get proper power of 2 */
1204 }
1205 if (newsize < newmax)
1206 newsize *= 2;
1207 if (newsize < newmax)
1208 return; /* overflow detection */
1209
cbec9347 1210 a = xhv->xhv_array; /* HvARRAY(hv) */
72940dca 1211 if (a) {
3280af22 1212 PL_nomemok = TRUE;
8d6dde3e 1213#if defined(STRANGE_MALLOC) || defined(MYMALLOC)
d18c6117 1214 Renew(a, PERL_HV_ARRAY_ALLOC_BYTES(newsize), char);
8aacddc1 1215 if (!a) {
4a33f861 1216 PL_nomemok = FALSE;
422a93e5
GA
1217 return;
1218 }
72940dca 1219#else
d18c6117 1220 New(2, a, PERL_HV_ARRAY_ALLOC_BYTES(newsize), char);
8aacddc1 1221 if (!a) {
3280af22 1222 PL_nomemok = FALSE;
422a93e5
GA
1223 return;
1224 }
cbec9347 1225 Copy(xhv->xhv_array /* HvARRAY(hv) */, a, oldsize * sizeof(HE*), char);
fba3b22e 1226 if (oldsize >= 64) {
cbec9347
JH
1227 offer_nice_chunk(xhv->xhv_array /* HvARRAY(hv) */,
1228 PERL_HV_ARRAY_ALLOC_BYTES(oldsize));
72940dca
PP
1229 }
1230 else
cbec9347 1231 Safefree(xhv->xhv_array /* HvARRAY(hv) */);
72940dca 1232#endif
3280af22 1233 PL_nomemok = FALSE;
72311751 1234 Zero(&a[oldsize * sizeof(HE*)], (newsize-oldsize) * sizeof(HE*), char); /* zero 2nd half*/
72940dca
PP
1235 }
1236 else {
d18c6117 1237 Newz(0, a, PERL_HV_ARRAY_ALLOC_BYTES(newsize), char);
72940dca 1238 }
cbec9347
JH
1239 xhv->xhv_max = --newsize; /* HvMAX(hv) = --newsize */
1240 xhv->xhv_array = a; /* HvARRAY(hv) = a */
1241 if (!xhv->xhv_fill /* !HvFILL(hv) */) /* skip rest if no entries */
72940dca
PP
1242 return;
1243
72311751
GS
1244 aep = (HE**)a;
1245 for (i=0; i<oldsize; i++,aep++) {
1246 if (!*aep) /* non-existent */
72940dca 1247 continue;
72311751 1248 for (oentry = aep, entry = *aep; entry; entry = *oentry) {
72940dca
PP
1249 if ((j = (HeHASH(entry) & newsize)) != i) {
1250 j -= i;
1251 *oentry = HeNEXT(entry);
72311751 1252 if (!(HeNEXT(entry) = aep[j]))
cbec9347 1253 xhv->xhv_fill++; /* HvFILL(hv)++ */
72311751 1254 aep[j] = entry;
72940dca
PP
1255 continue;
1256 }
1257 else
1258 oentry = &HeNEXT(entry);
1259 }
72311751 1260 if (!*aep) /* everything moved */
cbec9347 1261 xhv->xhv_fill--; /* HvFILL(hv)-- */
72940dca
PP
1262 }
1263}
1264
954c1994
GS
1265/*
1266=for apidoc newHV
1267
1268Creates a new HV. The reference count is set to 1.
1269
1270=cut
1271*/
1272
79072805 1273HV *
864dbfa3 1274Perl_newHV(pTHX)
79072805
LW
1275{
1276 register HV *hv;
cbec9347 1277 register XPVHV* xhv;
79072805 1278
a0d0e21e
LW
1279 hv = (HV*)NEWSV(502,0);
1280 sv_upgrade((SV *)hv, SVt_PVHV);
cbec9347 1281 xhv = (XPVHV*)SvANY(hv);
79072805
LW
1282 SvPOK_off(hv);
1283 SvNOK_off(hv);
1c846c1f 1284#ifndef NODEFAULT_SHAREKEYS
fde52b5c 1285 HvSHAREKEYS_on(hv); /* key-sharing on by default */
1c846c1f 1286#endif
4b5190b5 1287
cbec9347
JH
1288 xhv->xhv_max = 7; /* HvMAX(hv) = 7 (start with 8 buckets) */
1289 xhv->xhv_fill = 0; /* HvFILL(hv) = 0 */
1290 xhv->xhv_pmroot = 0; /* HvPMROOT(hv) = 0 */
79072805
LW
1291 (void)hv_iterinit(hv); /* so each() will start off right */
1292 return hv;
1293}
1294
b3ac6de7 1295HV *
864dbfa3 1296Perl_newHVhv(pTHX_ HV *ohv)
b3ac6de7 1297{
b56ba0bf 1298 HV *hv = newHV();
4beac62f 1299 STRLEN hv_max, hv_fill;
4beac62f
AMS
1300
1301 if (!ohv || (hv_fill = HvFILL(ohv)) == 0)
1302 return hv;
4beac62f 1303 hv_max = HvMAX(ohv);
b3ac6de7 1304
b56ba0bf
AMS
1305 if (!SvMAGICAL((SV *)ohv)) {
1306 /* It's an ordinary hash, so copy it fast. AMS 20010804 */
eb160463
GS
1307 STRLEN i;
1308 bool shared = !!HvSHAREKEYS(ohv);
b56ba0bf 1309 HE **ents, **oents = (HE **)HvARRAY(ohv);
ff875642
JH
1310 char *a;
1311 New(0, a, PERL_HV_ARRAY_ALLOC_BYTES(hv_max+1), char);
1312 ents = (HE**)a;
b56ba0bf
AMS
1313
1314 /* In each bucket... */
1315 for (i = 0; i <= hv_max; i++) {
1316 HE *prev = NULL, *ent = NULL, *oent = oents[i];
1317
1318 if (!oent) {
1319 ents[i] = NULL;
1320 continue;
1321 }
1322
1323 /* Copy the linked list of entries. */
1324 for (oent = oents[i]; oent; oent = HeNEXT(oent)) {
1325 U32 hash = HeHASH(oent);
1326 char *key = HeKEY(oent);
19692e8d
NC
1327 STRLEN len = HeKLEN(oent);
1328 int flags = HeKFLAGS(oent);
b56ba0bf
AMS
1329
1330 ent = new_HE();
45dea987 1331 HeVAL(ent) = newSVsv(HeVAL(oent));
19692e8d
NC
1332 HeKEY_hek(ent)
1333 = shared ? share_hek_flags(key, len, hash, flags)
1334 : save_hek_flags(key, len, hash, flags);
b56ba0bf
AMS
1335 if (prev)
1336 HeNEXT(prev) = ent;
1337 else
1338 ents[i] = ent;
1339 prev = ent;
1340 HeNEXT(ent) = NULL;
1341 }
1342 }
1343
1344 HvMAX(hv) = hv_max;
1345 HvFILL(hv) = hv_fill;
8aacddc1 1346 HvTOTALKEYS(hv) = HvTOTALKEYS(ohv);
b56ba0bf 1347 HvARRAY(hv) = ents;
1c846c1f 1348 }
b56ba0bf
AMS
1349 else {
1350 /* Iterate over ohv, copying keys and values one at a time. */
b3ac6de7 1351 HE *entry;
b56ba0bf
AMS
1352 I32 riter = HvRITER(ohv);
1353 HE *eiter = HvEITER(ohv);
1354
1355 /* Can we use fewer buckets? (hv_max is always 2^n-1) */
1356 while (hv_max && hv_max + 1 >= hv_fill * 2)
1357 hv_max = hv_max / 2;
1358 HvMAX(hv) = hv_max;
1359
4a76a316 1360 hv_iterinit(ohv);
e16e2ff8 1361 while ((entry = hv_iternext_flags(ohv, 0))) {
19692e8d
NC
1362 hv_store_flags(hv, HeKEY(entry), HeKLEN(entry),
1363 newSVsv(HeVAL(entry)), HeHASH(entry),
1364 HeKFLAGS(entry));
b3ac6de7 1365 }
b56ba0bf
AMS
1366 HvRITER(ohv) = riter;
1367 HvEITER(ohv) = eiter;
b3ac6de7 1368 }
1c846c1f 1369
b3ac6de7
IZ
1370 return hv;
1371}
1372
79072805 1373void
864dbfa3 1374Perl_hv_free_ent(pTHX_ HV *hv, register HE *entry)
79072805 1375{
16bdeea2
GS
1376 SV *val;
1377
68dc0745 1378 if (!entry)
79072805 1379 return;
16bdeea2 1380 val = HeVAL(entry);
257c9e5b 1381 if (val && isGV(val) && GvCVu(val) && HvNAME(hv))
3280af22 1382 PL_sub_generation++; /* may be deletion of method from stash */
16bdeea2 1383 SvREFCNT_dec(val);
68dc0745
PP
1384 if (HeKLEN(entry) == HEf_SVKEY) {
1385 SvREFCNT_dec(HeKEY_sv(entry));
8aacddc1 1386 Safefree(HeKEY_hek(entry));
44a8e56a
PP
1387 }
1388 else if (HvSHAREKEYS(hv))
68dc0745 1389 unshare_hek(HeKEY_hek(entry));
fde52b5c 1390 else
68dc0745 1391 Safefree(HeKEY_hek(entry));
d33b2eba 1392 del_HE(entry);
79072805
LW
1393}
1394
1395void
864dbfa3 1396Perl_hv_delayfree_ent(pTHX_ HV *hv, register HE *entry)
79072805 1397{
68dc0745 1398 if (!entry)
79072805 1399 return;
68dc0745 1400 if (isGV(HeVAL(entry)) && GvCVu(HeVAL(entry)) && HvNAME(hv))
3280af22 1401 PL_sub_generation++; /* may be deletion of method from stash */
68dc0745
PP
1402 sv_2mortal(HeVAL(entry)); /* free between statements */
1403 if (HeKLEN(entry) == HEf_SVKEY) {
1404 sv_2mortal(HeKEY_sv(entry));
1405 Safefree(HeKEY_hek(entry));
44a8e56a
PP
1406 }
1407 else if (HvSHAREKEYS(hv))
68dc0745 1408 unshare_hek(HeKEY_hek(entry));
fde52b5c 1409 else
68dc0745 1410 Safefree(HeKEY_hek(entry));
d33b2eba 1411 del_HE(entry);
79072805
LW
1412}
1413
954c1994
GS
1414/*
1415=for apidoc hv_clear
1416
1417Clears a hash, making it empty.
1418
1419=cut
1420*/
1421
79072805 1422void
864dbfa3 1423Perl_hv_clear(pTHX_ HV *hv)
79072805 1424{
cbec9347 1425 register XPVHV* xhv;
79072805
LW
1426 if (!hv)
1427 return;
49293501 1428
ecae49c0
NC
1429 DEBUG_A(Perl_hv_assert(aTHX_ hv));
1430
34c3c4e3
DM
1431 xhv = (XPVHV*)SvANY(hv);
1432
5f099cb0 1433 if (SvREADONLY(hv) && xhv->xhv_array != NULL) {
34c3c4e3 1434 /* restricted hash: convert all keys to placeholders */
3a676441
JH
1435 I32 i;
1436 HE* entry;
1437 for (i = 0; i <= (I32) xhv->xhv_max; i++) {
1438 entry = ((HE**)xhv->xhv_array)[i];
1439 for (; entry; entry = HeNEXT(entry)) {
1440 /* not already placeholder */
7996736c 1441 if (HeVAL(entry) != &PL_sv_placeholder) {
3a676441
JH
1442 if (HeVAL(entry) && SvREADONLY(HeVAL(entry))) {
1443 SV* keysv = hv_iterkeysv(entry);
1444 Perl_croak(aTHX_
1445 "Attempt to delete readonly key '%"SVf"' from a restricted hash",
1446 keysv);
1447 }
1448 SvREFCNT_dec(HeVAL(entry));
7996736c 1449 HeVAL(entry) = &PL_sv_placeholder;
3a676441
JH
1450 xhv->xhv_placeholders++; /* HvPLACEHOLDERS(hv)++ */
1451 }
34c3c4e3
DM
1452 }
1453 }
df8c6964 1454 goto reset;
49293501
MS
1455 }
1456
463ee0b2 1457 hfreeentries(hv);
8aacddc1 1458 xhv->xhv_placeholders = 0; /* HvPLACEHOLDERS(hv) = 0 */
cbec9347
JH
1459 if (xhv->xhv_array /* HvARRAY(hv) */)
1460 (void)memzero(xhv->xhv_array /* HvARRAY(hv) */,
1461 (xhv->xhv_max+1 /* HvMAX(hv)+1 */) * sizeof(HE*));
a0d0e21e
LW
1462
1463 if (SvRMAGICAL(hv))
1c846c1f 1464 mg_clear((SV*)hv);
574c8022 1465
19692e8d 1466 HvHASKFLAGS_off(hv);
bb443f97 1467 HvREHASH_off(hv);
df8c6964
TP
1468 reset:
1469 HvEITER(hv) = NULL;
79072805
LW
1470}
1471
3540d4ce
AB
1472/*
1473=for apidoc hv_clear_placeholders
1474
1475Clears any placeholders from a hash. If a restricted hash has any of its keys
1476marked as readonly and the key is subsequently deleted, the key is not actually
1477deleted but is marked by assigning it a value of &PL_sv_placeholder. This tags
1478it so it will be ignored by future operations such as iterating over the hash,
1479but will still allow the hash to have a value reaasigned to the key at some
1480future point. This function clears any such placeholder keys from the hash.
1481See Hash::Util::lock_keys() for an example of its use.
1482
1483=cut
1484*/
1485
1486void
1487Perl_hv_clear_placeholders(pTHX_ HV *hv)
1488{
d3677389
NC
1489 I32 items = (I32)HvPLACEHOLDERS(hv);
1490 I32 i = HvMAX(hv);
1491
1492 if (items == 0)
1493 return;
1494
1495 do {
1496 /* Loop down the linked list heads */
1497 int first = 1;
1498 HE **oentry = &(HvARRAY(hv))[i];
1499 HE *entry = *oentry;
1500
1501 if (!entry)
1502 continue;
1503
1504 for (; entry; first=0, oentry = &HeNEXT(entry), entry = *oentry) {
1505 if (HeVAL(entry) == &PL_sv_placeholder) {
1506 *oentry = HeNEXT(entry);
1507 if (first && !*oentry)
1508 HvFILL(hv)--; /* This linked list is now empty. */
1509 if (HvEITER(hv))
1510 HvLAZYDEL_on(hv);
1511 else
1512 hv_free_ent(hv, entry);
1513
1514 if (--items == 0) {
1515 /* Finished. */
1516 HvTOTALKEYS(hv) -= HvPLACEHOLDERS(hv);
1517 if (HvKEYS(hv) == 0)
1518 HvHASKFLAGS_off(hv);
1519 HvPLACEHOLDERS(hv) = 0;
1520 return;
1521 }
1522 }
1523 }
1524 } while (--i >= 0);
1525 /* You can't get here, hence assertion should always fail. */
1526 assert (items == 0);
1527 assert (0);
3540d4ce
AB
1528}
1529
76e3520e 1530STATIC void
cea2e8a9 1531S_hfreeentries(pTHX_ HV *hv)
79072805 1532{
a0d0e21e 1533 register HE **array;
68dc0745
PP
1534 register HE *entry;
1535 register HE *oentry = Null(HE*);
a0d0e21e
LW
1536 I32 riter;
1537 I32 max;
79072805
LW
1538
1539 if (!hv)
1540 return;
a0d0e21e 1541 if (!HvARRAY(hv))
79072805 1542 return;
a0d0e21e
LW
1543
1544 riter = 0;
1545 max = HvMAX(hv);
1546 array = HvARRAY(hv);
2f86008e
DM
1547 /* make everyone else think the array is empty, so that the destructors
1548 * called for freed entries can't recusively mess with us */
1549 HvARRAY(hv) = Null(HE**);
1550 HvFILL(hv) = 0;
1551 ((XPVHV*) SvANY(hv))->xhv_keys = 0;
1552
68dc0745 1553 entry = array[0];
a0d0e21e 1554 for (;;) {
68dc0745
PP
1555 if (entry) {
1556 oentry = entry;
1557 entry = HeNEXT(entry);
1558 hv_free_ent(hv, oentry);
a0d0e21e 1559 }
68dc0745 1560 if (!entry) {
a0d0e21e
LW
1561 if (++riter > max)
1562 break;
68dc0745 1563 entry = array[riter];
1c846c1f 1564 }
79072805 1565 }
2f86008e 1566 HvARRAY(hv) = array;
a0d0e21e 1567 (void)hv_iterinit(hv);
79072805
LW
1568}
1569
954c1994
GS
1570/*
1571=for apidoc hv_undef
1572
1573Undefines the hash.
1574
1575=cut
1576*/
1577
79072805 1578void
864dbfa3 1579Perl_hv_undef(pTHX_ HV *hv)
79072805 1580{
cbec9347 1581 register XPVHV* xhv;
79072805
LW
1582 if (!hv)
1583 return;
ecae49c0 1584 DEBUG_A(Perl_hv_assert(aTHX_ hv));
cbec9347 1585 xhv = (XPVHV*)SvANY(hv);
463ee0b2 1586 hfreeentries(hv);
cbec9347 1587 Safefree(xhv->xhv_array /* HvARRAY(hv) */);
85e6fe83 1588 if (HvNAME(hv)) {
7e8961ec
AB
1589 if(PL_stashcache)
1590 hv_delete(PL_stashcache, HvNAME(hv), strlen(HvNAME(hv)), G_DISCARD);
85e6fe83
LW
1591 Safefree(HvNAME(hv));
1592 HvNAME(hv) = 0;
1593 }
cbec9347
JH
1594 xhv->xhv_max = 7; /* HvMAX(hv) = 7 (it's a normal hash) */
1595 xhv->xhv_array = 0; /* HvARRAY(hv) = 0 */
8aacddc1 1596 xhv->xhv_placeholders = 0; /* HvPLACEHOLDERS(hv) = 0 */
a0d0e21e
LW
1597
1598 if (SvRMAGICAL(hv))
1c846c1f 1599 mg_clear((SV*)hv);
79072805
LW
1600}
1601
954c1994
GS
1602/*
1603=for apidoc hv_iterinit
1604
1605Prepares a starting point to traverse a hash table. Returns the number of
1606keys in the hash (i.e. the same as C<HvKEYS(tb)>). The return value is
1c846c1f 1607currently only meaningful for hashes without tie magic.
954c1994
GS
1608
1609NOTE: Before version 5.004_65, C<hv_iterinit> used to return the number of
1610hash buckets that happen to be in use. If you still need that esoteric
1611value, you can get it through the macro C<HvFILL(tb)>.
1612
e16e2ff8 1613
954c1994
GS
1614=cut
1615*/
1616
79072805 1617I32
864dbfa3 1618Perl_hv_iterinit(pTHX_ HV *hv)
79072805 1619{
cbec9347 1620 register XPVHV* xhv;
aa689395
PP
1621 HE *entry;
1622
1623 if (!hv)
cea2e8a9 1624 Perl_croak(aTHX_ "Bad hash");
cbec9347
JH
1625 xhv = (XPVHV*)SvANY(hv);
1626 entry = xhv->xhv_eiter; /* HvEITER(hv) */
72940dca
PP
1627 if (entry && HvLAZYDEL(hv)) { /* was deleted earlier? */
1628 HvLAZYDEL_off(hv);
68dc0745 1629 hv_free_ent(hv, entry);
72940dca 1630 }
cbec9347
JH
1631 xhv->xhv_riter = -1; /* HvRITER(hv) = -1 */
1632 xhv->xhv_eiter = Null(HE*); /* HvEITER(hv) = Null(HE*) */
1633 /* used to be xhv->xhv_fill before 5.004_65 */
8aacddc1 1634 return XHvTOTALKEYS(xhv);
79072805 1635}
954c1994
GS
1636/*
1637=for apidoc hv_iternext
1638
1639Returns entries from a hash iterator. See C<hv_iterinit>.
1640
fe7bca90
NC
1641You may call C<hv_delete> or C<hv_delete_ent> on the hash entry that the
1642iterator currently points to, without losing your place or invalidating your
1643iterator. Note that in this case the current entry is deleted from the hash
1644with your iterator holding the last reference to it. Your iterator is flagged
1645to free the entry on the next call to C<hv_iternext>, so you must not discard
1646your iterator immediately else the entry will leak - call C<hv_iternext> to
1647trigger the resource deallocation.
1648
954c1994
GS
1649=cut
1650*/
1651
79072805 1652HE *
864dbfa3 1653Perl_hv_iternext(pTHX_ HV *hv)
79072805 1654{
e16e2ff8
NC
1655 return hv_iternext_flags(hv, 0);
1656}
1657
1658/*
fe7bca90
NC
1659=for apidoc hv_iternext_flags
1660
1661Returns entries from a hash iterator. See C<hv_iterinit> and C<hv_iternext>.
1662The C<flags> value will normally be zero; if HV_ITERNEXT_WANTPLACEHOLDERS is
1663set the placeholders keys (for restricted hashes) will be returned in addition
1664to normal keys. By default placeholders are automatically skipped over.
7996736c
MHM
1665Currently a placeholder is implemented with a value that is
1666C<&Perl_sv_placeholder>. Note that the implementation of placeholders and
fe7bca90
NC
1667restricted hashes may change, and the implementation currently is
1668insufficiently abstracted for any change to be tidy.
e16e2ff8 1669
fe7bca90 1670=cut
e16e2ff8
NC
1671*/
1672
1673HE *
1674Perl_hv_iternext_flags(pTHX_ HV *hv, I32 flags)
1675{
cbec9347 1676 register XPVHV* xhv;
79072805 1677 register HE *entry;
a0d0e21e 1678 HE *oldentry;
463ee0b2 1679 MAGIC* mg;
79072805
LW
1680
1681 if (!hv)
cea2e8a9 1682 Perl_croak(aTHX_ "Bad hash");
cbec9347
JH
1683 xhv = (XPVHV*)SvANY(hv);
1684 oldentry = entry = xhv->xhv_eiter; /* HvEITER(hv) */
463ee0b2 1685
14befaf4 1686 if ((mg = SvTIED_mg((SV*)hv, PERL_MAGIC_tied))) {
8990e307 1687 SV *key = sv_newmortal();
cd1469e6 1688 if (entry) {
fde52b5c 1689 sv_setsv(key, HeSVKEY_force(entry));
cd1469e6
PP
1690 SvREFCNT_dec(HeSVKEY(entry)); /* get rid of previous key */
1691 }
a0d0e21e 1692 else {
ff68c719 1693 char *k;
bbce6d69 1694 HEK *hek;
ff68c719 1695
cbec9347
JH
1696 /* one HE per MAGICAL hash */
1697 xhv->xhv_eiter = entry = new_HE(); /* HvEITER(hv) = new_HE() */
4633a7c4 1698 Zero(entry, 1, HE);
ff68c719
PP
1699 Newz(54, k, HEK_BASESIZE + sizeof(SV*), char);
1700 hek = (HEK*)k;
1701 HeKEY_hek(entry) = hek;
fde52b5c 1702 HeKLEN(entry) = HEf_SVKEY;
a0d0e21e
LW
1703 }
1704 magic_nextpack((SV*) hv,mg,key);
8aacddc1 1705 if (SvOK(key)) {
cd1469e6 1706 /* force key to stay around until next time */
bbce6d69
PP
1707 HeSVKEY_set(entry, SvREFCNT_inc(key));
1708 return entry; /* beware, hent_val is not set */
8aacddc1 1709 }
fde52b5c
PP
1710 if (HeVAL(entry))
1711 SvREFCNT_dec(HeVAL(entry));
ff68c719 1712 Safefree(HeKEY_hek(entry));
d33b2eba 1713 del_HE(entry);
cbec9347 1714 xhv->xhv_eiter = Null(HE*); /* HvEITER(hv) = Null(HE*) */
463ee0b2 1715 return Null(HE*);
79072805 1716 }
f675dbe5 1717#ifdef DYNAMIC_ENV_FETCH /* set up %ENV for iteration */
cbec9347 1718 if (!entry && SvRMAGICAL((SV*)hv) && mg_find((SV*)hv, PERL_MAGIC_env))
f675dbe5
CB
1719 prime_env_iter();
1720#endif
463ee0b2 1721
cbec9347
JH
1722 if (!xhv->xhv_array /* !HvARRAY(hv) */)
1723 Newz(506, xhv->xhv_array /* HvARRAY(hv) */,
1724 PERL_HV_ARRAY_ALLOC_BYTES(xhv->xhv_max+1 /* HvMAX(hv)+1 */),
1725 char);
015a5f36 1726 /* At start of hash, entry is NULL. */
fde52b5c 1727 if (entry)
8aacddc1 1728 {
fde52b5c 1729 entry = HeNEXT(entry);
e16e2ff8
NC
1730 if (!(flags & HV_ITERNEXT_WANTPLACEHOLDERS)) {
1731 /*
1732 * Skip past any placeholders -- don't want to include them in
1733 * any iteration.
1734 */
7996736c 1735 while (entry && HeVAL(entry) == &PL_sv_placeholder) {
e16e2ff8
NC
1736 entry = HeNEXT(entry);
1737 }
8aacddc1
NIS
1738 }
1739 }
fde52b5c 1740 while (!entry) {
015a5f36
NC
1741 /* OK. Come to the end of the current list. Grab the next one. */
1742
cbec9347 1743 xhv->xhv_riter++; /* HvRITER(hv)++ */
eb160463 1744 if (xhv->xhv_riter > (I32)xhv->xhv_max /* HvRITER(hv) > HvMAX(hv) */) {
015a5f36 1745 /* There is no next one. End of the hash. */
cbec9347 1746 xhv->xhv_riter = -1; /* HvRITER(hv) = -1 */
fde52b5c 1747 break;
79072805 1748 }
cbec9347
JH
1749 /* entry = (HvARRAY(hv))[HvRITER(hv)]; */
1750 entry = ((HE**)xhv->xhv_array)[xhv->xhv_riter];
8aacddc1 1751
e16e2ff8 1752 if (!(flags & HV_ITERNEXT_WANTPLACEHOLDERS)) {
015a5f36
NC
1753 /* If we have an entry, but it's a placeholder, don't count it.
1754 Try the next. */
7996736c 1755 while (entry && HeVAL(entry) == &PL_sv_placeholder)
015a5f36
NC
1756 entry = HeNEXT(entry);
1757 }
1758 /* Will loop again if this linked list starts NULL
1759 (for HV_ITERNEXT_WANTPLACEHOLDERS)
1760 or if we run through it and find only placeholders. */
fde52b5c 1761 }
79072805 1762
72940dca
PP
1763 if (oldentry && HvLAZYDEL(hv)) { /* was deleted earlier? */
1764 HvLAZYDEL_off(hv);
68dc0745 1765 hv_free_ent(hv, oldentry);
72940dca 1766 }
a0d0e21e 1767
fdcd69b6
NC
1768 /*if (HvREHASH(hv) && entry && !HeKREHASH(entry))
1769 PerlIO_printf(PerlIO_stderr(), "Awooga %p %p\n", hv, entry);*/
1770
cbec9347 1771 xhv->xhv_eiter = entry; /* HvEITER(hv) = entry */
79072805
LW
1772 return entry;
1773}
1774
954c1994
GS
1775/*
1776=for apidoc hv_iterkey
1777
1778Returns the key from the current position of the hash iterator. See
1779C<hv_iterinit>.
1780
1781=cut
1782*/
1783
79072805 1784char *
864dbfa3 1785Perl_hv_iterkey(pTHX_ register HE *entry, I32 *retlen)
79072805 1786{
fde52b5c 1787 if (HeKLEN(entry) == HEf_SVKEY) {
fb73857a
PP
1788 STRLEN len;
1789 char *p = SvPV(HeKEY_sv(entry), len);
1790 *retlen = len;
1791 return p;
fde52b5c
PP
1792 }
1793 else {
1794 *retlen = HeKLEN(entry);
1795 return HeKEY(entry);
1796 }
1797}
1798
1799/* unlike hv_iterval(), this always returns a mortal copy of the key */
954c1994
GS
1800/*
1801=for apidoc hv_iterkeysv
1802
1803Returns the key as an C<SV*> from the current position of the hash
1804iterator. The return value will always be a mortal copy of the key. Also
1805see C<hv_iterinit>.
1806
1807=cut
1808*/
1809
fde52b5c 1810SV *
864dbfa3 1811Perl_hv_iterkeysv(pTHX_ register HE *entry)
fde52b5c 1812{
19692e8d
NC
1813 if (HeKLEN(entry) != HEf_SVKEY) {
1814 HEK *hek = HeKEY_hek(entry);
1815 int flags = HEK_FLAGS(hek);
1816 SV *sv;
1817
1818 if (flags & HVhek_WASUTF8) {
1819 /* Trouble :-)
1820 Andreas would like keys he put in as utf8 to come back as utf8
1821 */
1822 STRLEN utf8_len = HEK_LEN(hek);
2e5dfef7 1823 U8 *as_utf8 = bytes_to_utf8 ((U8*)HEK_KEY(hek), &utf8_len);
19692e8d 1824
2e5dfef7 1825 sv = newSVpvn ((char*)as_utf8, utf8_len);
19692e8d 1826 SvUTF8_on (sv);
c193270f 1827 Safefree (as_utf8); /* bytes_to_utf8() allocates a new string */
4b5190b5
NC
1828 } else if (flags & HVhek_REHASH) {
1829 /* We don't have a pointer to the hv, so we have to replicate the
1830 flag into every HEK. This hv is using custom a hasing
1831 algorithm. Hence we can't return a shared string scalar, as
1832 that would contain the (wrong) hash value, and might get passed
1833 into an hv routine with a regular hash */
1834
1835 sv = newSVpvn (HEK_KEY(hek), HEK_LEN(hek));
1836 if (HEK_UTF8(hek))
1837 SvUTF8_on (sv);
1838 } else {
19692e8d
NC
1839 sv = newSVpvn_share(HEK_KEY(hek),
1840 (HEK_UTF8(hek) ? -HEK_LEN(hek) : HEK_LEN(hek)),
1841 HEK_HASH(hek));
1842 }
1843 return sv_2mortal(sv);
1844 }
1845 return sv_mortalcopy(HeKEY_sv(entry));
79072805
LW
1846}
1847
954c1994
GS
1848/*
1849=for apidoc hv_iterval
1850
1851Returns the value from the current position of the hash iterator. See
1852C<hv_iterkey>.
1853
1854=cut
1855*/
1856
79072805 1857SV *
864dbfa3 1858Perl_hv_iterval(pTHX_ HV *hv, register HE *entry)
79072805 1859{
8990e307 1860 if (SvRMAGICAL(hv)) {
14befaf4 1861 if (mg_find((SV*)hv, PERL_MAGIC_tied)) {
8990e307 1862 SV* sv = sv_newmortal();
bbce6d69
PP
1863 if (HeKLEN(entry) == HEf_SVKEY)
1864 mg_copy((SV*)hv, sv, (char*)HeKEY_sv(entry), HEf_SVKEY);
1865 else mg_copy((SV*)hv, sv, HeKEY(entry), HeKLEN(entry));
463ee0b2
LW
1866 return sv;
1867 }
79072805 1868 }
fde52b5c 1869 return HeVAL(entry);
79072805
LW
1870}
1871
954c1994
GS
1872/*
1873=for apidoc hv_iternextsv
1874
1875Performs an C<hv_iternext>, C<hv_iterkey>, and C<hv_iterval> in one
1876operation.
1877
1878=cut
1879*/
1880
a0d0e21e 1881SV *
864dbfa3 1882Perl_hv_iternextsv(pTHX_ HV *hv, char **key, I32 *retlen)
a0d0e21e
LW
1883{
1884 HE *he;
e16e2ff8 1885 if ( (he = hv_iternext_flags(hv, 0)) == NULL)
a0d0e21e
LW
1886 return NULL;
1887 *key = hv_iterkey(he, retlen);
1888 return hv_iterval(hv, he);
1889}
1890
954c1994
GS
1891/*
1892=for apidoc hv_magic
1893
1894Adds magic to a hash. See C<sv_magic>.
1895
1896=cut
1897*/
1898
79072805 1899void
864dbfa3 1900Perl_hv_magic(pTHX_ HV *hv, GV *gv, int how)
79072805 1901{
a0d0e21e 1902 sv_magic((SV*)hv, (SV*)gv, how, Nullch, 0);
79072805 1903}
fde52b5c 1904
37d85e3a
JH
1905#if 0 /* use the macro from hv.h instead */
1906
bbce6d69 1907char*
864dbfa3 1908Perl_sharepvn(pTHX_ const char *sv, I32 len, U32 hash)
bbce6d69 1909{
ff68c719 1910 return HEK_KEY(share_hek(sv, len, hash));
bbce6d69
PP
1911}
1912
37d85e3a
JH
1913#endif
1914
bbce6d69 1915/* possibly free a shared string if no one has access to it
fde52b5c
PP
1916 * len and hash must both be valid for str.
1917 */
bbce6d69 1918void
864dbfa3 1919Perl_unsharepvn(pTHX_ const char *str, I32 len, U32 hash)
fde52b5c 1920{
19692e8d
NC
1921 unshare_hek_or_pvn (NULL, str, len, hash);
1922}
1923
1924
1925void
1926Perl_unshare_hek(pTHX_ HEK *hek)
1927{
1928 unshare_hek_or_pvn(hek, NULL, 0, 0);
1929}
1930
1931/* possibly free a shared string if no one has access to it
1932 hek if non-NULL takes priority over the other 3, else str, len and hash
1933 are used. If so, len and hash must both be valid for str.
1934 */
df132699 1935STATIC void
19692e8d
NC
1936S_unshare_hek_or_pvn(pTHX_ HEK *hek, const char *str, I32 len, U32 hash)
1937{
cbec9347 1938 register XPVHV* xhv;
fde52b5c
PP
1939 register HE *entry;
1940 register HE **oentry;
1941 register I32 i = 1;
1942 I32 found = 0;
c3654f1a 1943 bool is_utf8 = FALSE;
19692e8d 1944 int k_flags = 0;
f9a63242 1945 const char *save = str;
c3654f1a 1946
19692e8d
NC
1947 if (hek) {
1948 hash = HEK_HASH(hek);
1949 } else if (len < 0) {
1950 STRLEN tmplen = -len;
1951 is_utf8 = TRUE;
1952 /* See the note in hv_fetch(). --jhi */
1953 str = (char*)bytes_from_utf8((U8*)str, &tmplen, &is_utf8);
1954 len = tmplen;
1955 if (is_utf8)
1956 k_flags = HVhek_UTF8;
1957 if (str != save)
1958 k_flags |= HVhek_WASUTF8 | HVhek_FREEKEY;
c3654f1a 1959 }
1c846c1f 1960
fde52b5c 1961 /* what follows is the moral equivalent of:
6b88bc9c 1962 if ((Svp = hv_fetch(PL_strtab, tmpsv, FALSE, hash))) {
bbce6d69 1963 if (--*Svp == Nullsv)
6b88bc9c 1964 hv_delete(PL_strtab, str, len, G_DISCARD, hash);
bbce6d69 1965 } */
cbec9347 1966 xhv = (XPVHV*)SvANY(PL_strtab);
fde52b5c 1967 /* assert(xhv_array != 0) */
5f08fbcd 1968 LOCK_STRTAB_MUTEX;
cbec9347
JH
1969 /* oentry = &(HvARRAY(hv))[hash & (I32) HvMAX(hv)]; */
1970 oentry = &((HE**)xhv->xhv_array)[hash & (I32) xhv->xhv_max];
19692e8d
NC
1971 if (hek) {
1972 for (entry = *oentry; entry; i=0, oentry = &HeNEXT(entry), entry = *oentry) {
1973 if (HeKEY_hek(entry) != hek)
1974 continue;
1975 found = 1;
1976 break;
1977 }
1978 } else {
1979 int flags_masked = k_flags & HVhek_MASK;
1980 for (entry = *oentry; entry; i=0, oentry = &HeNEXT(entry), entry = *oentry) {
1981 if (HeHASH(entry) != hash) /* strings can't be equal */
1982 continue;
1983 if (HeKLEN(entry) != len)
1984 continue;
1985 if (HeKEY(entry) != str && memNE(HeKEY(entry),str,len)) /* is this it? */
1986 continue;
1987 if (HeKFLAGS(entry) != flags_masked)
1988 continue;
1989 found = 1;
1990 break;
1991 }
1992 }
1993
1994 if (found) {
1995 if (--HeVAL(entry) == Nullsv) {
1996 *oentry = HeNEXT(entry);
1997 if (i && !*oentry)
1998 xhv->xhv_fill--; /* HvFILL(hv)-- */
1999 Safefree(HeKEY_hek(entry));
2000 del_HE(entry);
2001 xhv->xhv_keys--; /* HvKEYS(hv)-- */
2002 }
fde52b5c 2003 }
19692e8d 2004
333f433b 2005 UNLOCK_STRTAB_MUTEX;
411caa50 2006 if (!found && ckWARN_d(WARN_INTERNAL))
19692e8d
NC
2007 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
2008 "Attempt to free non-existent shared string '%s'%s",
2009 hek ? HEK_KEY(hek) : str,
2010 (k_flags & HVhek_UTF8) ? " (utf8)" : "");
2011 if (k_flags & HVhek_FREEKEY)
2012 Safefree(str);
fde52b5c
PP
2013}
2014
bbce6d69
PP
2015/* get a (constant) string ptr from the global string table
2016 * string will get added if it is not already there.
fde52b5c
PP
2017 * len and hash must both be valid for str.
2018 */
bbce6d69 2019HEK *
864dbfa3 2020Perl_share_hek(pTHX_ const char *str, I32 len, register U32 hash)
fde52b5c 2021{
da58a35d 2022 bool is_utf8 = FALSE;
19692e8d 2023 int flags = 0;
f9a63242 2024 const char *save = str;
da58a35d
JH
2025
2026 if (len < 0) {
77caf834 2027 STRLEN tmplen = -len;
da58a35d 2028 is_utf8 = TRUE;
77caf834
JH
2029 /* See the note in hv_fetch(). --jhi */
2030 str = (char*)bytes_from_utf8((U8*)str, &tmplen, &is_utf8);
2031 len = tmplen;
19692e8d
NC
2032 /* If we were able to downgrade here, then than means that we were passed
2033 in a key which only had chars 0-255, but was utf8 encoded. */
2034 if (is_utf8)
2035 flags = HVhek_UTF8;
2036 /* If we found we were able to downgrade the string to bytes, then
2037 we should flag that it needs upgrading on keys or each. Also flag
2038 that we need share_hek_flags to free the string. */
2039 if (str != save)
2040 flags |= HVhek_WASUTF8 | HVhek_FREEKEY;
2041 }
2042
2043 return share_hek_flags (str, len, hash, flags);
2044}
2045
df132699 2046STATIC HEK *
19692e8d
NC
2047S_share_hek_flags(pTHX_ const char *str, I32 len, register U32 hash, int flags)
2048{
2049 register XPVHV* xhv;
2050 register HE *entry;
2051 register HE **oentry;
2052 register I32 i = 1;
2053 I32 found = 0;
2054 int flags_masked = flags & HVhek_MASK;
bbce6d69 2055
fde52b5c 2056 /* what follows is the moral equivalent of:
1c846c1f 2057
6b88bc9c 2058 if (!(Svp = hv_fetch(PL_strtab, str, len, FALSE)))
8aacddc1 2059 hv_store(PL_strtab, str, len, Nullsv, hash);
fdcd69b6
NC
2060
2061 Can't rehash the shared string table, so not sure if it's worth
2062 counting the number of entries in the linked list
bbce6d69 2063 */
cbec9347 2064 xhv = (XPVHV*)SvANY(PL_strtab);
fde52b5c 2065 /* assert(xhv_array != 0) */
5f08fbcd 2066 LOCK_STRTAB_MUTEX;
cbec9347
JH
2067 /* oentry = &(HvARRAY(hv))[hash & (I32) HvMAX(hv)]; */
2068 oentry = &((HE**)xhv->xhv_array)[hash & (I32) xhv->xhv_max];
bbce6d69 2069 for (entry = *oentry; entry; i=0, entry = HeNEXT(entry)) {
fde52b5c
PP
2070 if (HeHASH(entry) != hash) /* strings can't be equal */
2071 continue;
2072 if (HeKLEN(entry) != len)
2073 continue;
1c846c1f 2074 if (HeKEY(entry) != str && memNE(HeKEY(entry),str,len)) /* is this it? */
fde52b5c 2075 continue;
19692e8d 2076 if (HeKFLAGS(entry) != flags_masked)
c3654f1a 2077 continue;
fde52b5c 2078 found = 1;
fde52b5c
PP
2079 break;
2080 }
bbce6d69 2081 if (!found) {
d33b2eba 2082 entry = new_HE();
dcf933a4 2083 HeKEY_hek(entry) = save_hek_flags(str, len, hash, flags_masked);
bbce6d69
PP
2084 HeVAL(entry) = Nullsv;
2085 HeNEXT(entry) = *oentry;
2086 *oentry = entry;
cbec9347 2087 xhv->xhv_keys++; /* HvKEYS(hv)++ */
bbce6d69 2088 if (i) { /* initial entry? */
cbec9347 2089 xhv->xhv_fill++; /* HvFILL(hv)++ */
4c9cc595 2090 } else if (xhv->xhv_keys > (IV)xhv->xhv_max /* HvKEYS(hv) > HvMAX(hv) */) {
cbec9347 2091 hsplit(PL_strtab);
bbce6d69
PP
2092 }
2093 }
2094
2095 ++HeVAL(entry); /* use value slot as REFCNT */
5f08fbcd 2096 UNLOCK_STRTAB_MUTEX;
19692e8d
NC
2097
2098 if (flags & HVhek_FREEKEY)
f9a63242 2099 Safefree(str);
19692e8d 2100
ff68c719 2101 return HeKEY_hek(entry);
fde52b5c 2102}
ecae49c0
NC
2103
2104
2105/*
2106=for apidoc hv_assert
2107
2108Check that a hash is in an internally consistent state.
2109
2110=cut
2111*/
2112
2113void
2114Perl_hv_assert(pTHX_ HV *hv)
2115{
2116 HE* entry;
2117 int withflags = 0;
2118 int placeholders = 0;
2119 int real = 0;
2120 int bad = 0;
2121 I32 riter = HvRITER(hv);
2122 HE *eiter = HvEITER(hv);
2123
2124 (void)hv_iterinit(hv);
2125
2126 while ((entry = hv_iternext_flags(hv, HV_ITERNEXT_WANTPLACEHOLDERS))) {
2127 /* sanity check the values */
2128 if (HeVAL(entry) == &PL_sv_placeholder) {
2129 placeholders++;
2130 } else {
2131 real++;
2132 }
2133 /* sanity check the keys */
2134 if (HeSVKEY(entry)) {
2135 /* Don't know what to check on SV keys. */
2136 } else if (HeKUTF8(entry)) {
2137 withflags++;
2138 if (HeKWASUTF8(entry)) {
2139 PerlIO_printf(Perl_debug_log,
2140 "hash key has both WASUFT8 and UTF8: '%.*s'\n",
2141 (int) HeKLEN(entry), HeKEY(entry));
2142 bad = 1;
2143 }
2144 } else if (HeKWASUTF8(entry)) {
2145 withflags++;
2146 }
2147 }
2148 if (!SvTIED_mg((SV*)hv, PERL_MAGIC_tied)) {
2149 if (HvUSEDKEYS(hv) != real) {
2150 PerlIO_printf(Perl_debug_log, "Count %d key(s), but hash reports %d\n",
2151 (int) real, (int) HvUSEDKEYS(hv));
2152 bad = 1;
2153 }
2154 if (HvPLACEHOLDERS(hv) != placeholders) {
2155 PerlIO_printf(Perl_debug_log,
2156 "Count %d placeholder(s), but hash reports %d\n",
2157 (int) placeholders, (int) HvPLACEHOLDERS(hv));
2158 bad = 1;
2159 }
2160 }
2161 if (withflags && ! HvHASKFLAGS(hv)) {
2162 PerlIO_printf(Perl_debug_log,
2163 "Hash has HASKFLAGS off but I count %d key(s) with flags\n",
2164 withflags);
2165 bad = 1;
2166 }
2167 if (bad) {
2168 sv_dump((SV *)hv);
2169 }
2170 HvRITER(hv) = riter; /* Restore hash iterator state */
2171 HvEITER(hv) = eiter;
2172}