This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Unify 5.008 and 5.009's B::Concise 0.58
[perl5.git] / hv.c
CommitLineData
a0d0e21e 1/* hv.c
79072805 2 *
4bb101f2 3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
b5f8cc5c 4 * 2000, 2001, 2002, 2003, 2004, 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 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 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 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 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 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 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 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 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 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 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 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 875 register HE *entry;
876 register HE **oentry;
877 SV *sv;
da58a35d 878 bool is_utf8;
7a9669ca 879 int masked_flags;
1c846c1f 880
fde52b5c 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 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 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{
cbec9347
JH
1041 register XPVHV* xhv = (XPVHV*)SvANY(hv);
1042 I32 oldsize = (I32) xhv->xhv_max+1; /* HvMAX(hv)+1 (sick) */
79072805
LW
1043 register I32 newsize = oldsize * 2;
1044 register I32 i;
cbec9347 1045 register char *a = xhv->xhv_array; /* HvARRAY(hv) */
72311751
GS
1046 register HE **aep;
1047 register HE **bep;
79072805
LW
1048 register HE *entry;
1049 register HE **oentry;
4b5190b5
NC
1050 int longest_chain = 0;
1051 int was_shared;
79072805 1052
18026298
NC
1053 /*PerlIO_printf(PerlIO_stderr(), "hsplit called for %p which had %d\n",
1054 hv, (int) oldsize);*/
1055
1056 if (HvPLACEHOLDERS(hv) && !SvREADONLY(hv)) {
1057 /* Can make this clear any placeholders first for non-restricted hashes,
1058 even though Storable rebuilds restricted hashes by putting in all the
1059 placeholders (first) before turning on the readonly flag, because
1060 Storable always pre-splits the hash. */
1061 hv_clear_placeholders(hv);
1062 }
1063
3280af22 1064 PL_nomemok = TRUE;
8d6dde3e 1065#if defined(STRANGE_MALLOC) || defined(MYMALLOC)
d18c6117 1066 Renew(a, PERL_HV_ARRAY_ALLOC_BYTES(newsize), char);
422a93e5 1067 if (!a) {
4a33f861 1068 PL_nomemok = FALSE;
422a93e5
GA
1069 return;
1070 }
4633a7c4 1071#else
d18c6117 1072 New(2, a, PERL_HV_ARRAY_ALLOC_BYTES(newsize), char);
422a93e5 1073 if (!a) {
3280af22 1074 PL_nomemok = FALSE;
422a93e5
GA
1075 return;
1076 }
cbec9347 1077 Copy(xhv->xhv_array /* HvARRAY(hv) */, a, oldsize * sizeof(HE*), char);
fba3b22e 1078 if (oldsize >= 64) {
cbec9347
JH
1079 offer_nice_chunk(xhv->xhv_array /* HvARRAY(hv) */,
1080 PERL_HV_ARRAY_ALLOC_BYTES(oldsize));
4633a7c4
LW
1081 }
1082 else
cbec9347 1083 Safefree(xhv->xhv_array /* HvARRAY(hv) */);
4633a7c4
LW
1084#endif
1085
3280af22 1086 PL_nomemok = FALSE;
72311751 1087 Zero(&a[oldsize * sizeof(HE*)], (newsize-oldsize) * sizeof(HE*), char); /* zero 2nd half*/
cbec9347
JH
1088 xhv->xhv_max = --newsize; /* HvMAX(hv) = --newsize */
1089 xhv->xhv_array = a; /* HvARRAY(hv) = a */
72311751 1090 aep = (HE**)a;
79072805 1091
72311751 1092 for (i=0; i<oldsize; i++,aep++) {
4b5190b5
NC
1093 int left_length = 0;
1094 int right_length = 0;
1095
72311751 1096 if (!*aep) /* non-existent */
79072805 1097 continue;
72311751
GS
1098 bep = aep+oldsize;
1099 for (oentry = aep, entry = *aep; entry; entry = *oentry) {
eb160463 1100 if ((HeHASH(entry) & newsize) != (U32)i) {
fde52b5c 1101 *oentry = HeNEXT(entry);
72311751
GS
1102 HeNEXT(entry) = *bep;
1103 if (!*bep)
cbec9347 1104 xhv->xhv_fill++; /* HvFILL(hv)++ */
72311751 1105 *bep = entry;
4b5190b5 1106 right_length++;
79072805
LW
1107 continue;
1108 }
4b5190b5 1109 else {
fde52b5c 1110 oentry = &HeNEXT(entry);
4b5190b5
NC
1111 left_length++;
1112 }
79072805 1113 }
72311751 1114 if (!*aep) /* everything moved */
cbec9347 1115 xhv->xhv_fill--; /* HvFILL(hv)-- */
4b5190b5
NC
1116 /* I think we don't actually need to keep track of the longest length,
1117 merely flag if anything is too long. But for the moment while
1118 developing this code I'll track it. */
1119 if (left_length > longest_chain)
1120 longest_chain = left_length;
1121 if (right_length > longest_chain)
1122 longest_chain = right_length;
1123 }
1124
1125
1126 /* Pick your policy for "hashing isn't working" here: */
fdcd69b6 1127 if (longest_chain <= HV_MAX_LENGTH_BEFORE_SPLIT /* split worked? */
4b5190b5
NC
1128 || HvREHASH(hv)) {
1129 return;
79072805 1130 }
4b5190b5
NC
1131
1132 if (hv == PL_strtab) {
1133 /* Urg. Someone is doing something nasty to the string table.
1134 Can't win. */
1135 return;
1136 }
1137
1138 /* Awooga. Awooga. Pathological data. */
fdcd69b6 1139 /*PerlIO_printf(PerlIO_stderr(), "%p %d of %d with %d/%d buckets\n", hv,
4b5190b5
NC
1140 longest_chain, HvTOTALKEYS(hv), HvFILL(hv), 1+HvMAX(hv));*/
1141
1142 ++newsize;
1143 Newz(2, a, PERL_HV_ARRAY_ALLOC_BYTES(newsize), char);
1144 was_shared = HvSHAREKEYS(hv);
1145
1146 xhv->xhv_fill = 0;
1147 HvSHAREKEYS_off(hv);
1148 HvREHASH_on(hv);
1149
1150 aep = (HE **) xhv->xhv_array;
1151
1152 for (i=0; i<newsize; i++,aep++) {
1153 entry = *aep;
1154 while (entry) {
1155 /* We're going to trash this HE's next pointer when we chain it
1156 into the new hash below, so store where we go next. */
1157 HE *next = HeNEXT(entry);
1158 UV hash;
1159
1160 /* Rehash it */
1161 PERL_HASH_INTERNAL(hash, HeKEY(entry), HeKLEN(entry));
1162
1163 if (was_shared) {
1164 /* Unshare it. */
1165 HEK *new_hek
1166 = save_hek_flags(HeKEY(entry), HeKLEN(entry),
1167 hash, HeKFLAGS(entry));
1168 unshare_hek (HeKEY_hek(entry));
1169 HeKEY_hek(entry) = new_hek;
1170 } else {
1171 /* Not shared, so simply write the new hash in. */
1172 HeHASH(entry) = hash;
1173 }
1174 /*PerlIO_printf(PerlIO_stderr(), "%d ", HeKFLAGS(entry));*/
1175 HEK_REHASH_on(HeKEY_hek(entry));
1176 /*PerlIO_printf(PerlIO_stderr(), "%d\n", HeKFLAGS(entry));*/
1177
1178 /* Copy oentry to the correct new chain. */
1179 bep = ((HE**)a) + (hash & (I32) xhv->xhv_max);
1180 if (!*bep)
1181 xhv->xhv_fill++; /* HvFILL(hv)++ */
1182 HeNEXT(entry) = *bep;
1183 *bep = entry;
1184
1185 entry = next;
1186 }
1187 }
1188 Safefree (xhv->xhv_array);
1189 xhv->xhv_array = a; /* HvARRAY(hv) = a */
79072805
LW
1190}
1191
72940dca 1192void
864dbfa3 1193Perl_hv_ksplit(pTHX_ HV *hv, IV newmax)
72940dca 1194{
cbec9347
JH
1195 register XPVHV* xhv = (XPVHV*)SvANY(hv);
1196 I32 oldsize = (I32) xhv->xhv_max+1; /* HvMAX(hv)+1 (sick) */
72940dca 1197 register I32 newsize;
1198 register I32 i;
1199 register I32 j;
72311751
GS
1200 register char *a;
1201 register HE **aep;
72940dca 1202 register HE *entry;
1203 register HE **oentry;
1204
1205 newsize = (I32) newmax; /* possible truncation here */
1206 if (newsize != newmax || newmax <= oldsize)
1207 return;
1208 while ((newsize & (1 + ~newsize)) != newsize) {
1209 newsize &= ~(newsize & (1 + ~newsize)); /* get proper power of 2 */
1210 }
1211 if (newsize < newmax)
1212 newsize *= 2;
1213 if (newsize < newmax)
1214 return; /* overflow detection */
1215
cbec9347 1216 a = xhv->xhv_array; /* HvARRAY(hv) */
72940dca 1217 if (a) {
3280af22 1218 PL_nomemok = TRUE;
8d6dde3e 1219#if defined(STRANGE_MALLOC) || defined(MYMALLOC)
d18c6117 1220 Renew(a, PERL_HV_ARRAY_ALLOC_BYTES(newsize), char);
8aacddc1 1221 if (!a) {
4a33f861 1222 PL_nomemok = FALSE;
422a93e5
GA
1223 return;
1224 }
72940dca 1225#else
d18c6117 1226 New(2, a, PERL_HV_ARRAY_ALLOC_BYTES(newsize), char);
8aacddc1 1227 if (!a) {
3280af22 1228 PL_nomemok = FALSE;
422a93e5
GA
1229 return;
1230 }
cbec9347 1231 Copy(xhv->xhv_array /* HvARRAY(hv) */, a, oldsize * sizeof(HE*), char);
fba3b22e 1232 if (oldsize >= 64) {
cbec9347
JH
1233 offer_nice_chunk(xhv->xhv_array /* HvARRAY(hv) */,
1234 PERL_HV_ARRAY_ALLOC_BYTES(oldsize));
72940dca 1235 }
1236 else
cbec9347 1237 Safefree(xhv->xhv_array /* HvARRAY(hv) */);
72940dca 1238#endif
3280af22 1239 PL_nomemok = FALSE;
72311751 1240 Zero(&a[oldsize * sizeof(HE*)], (newsize-oldsize) * sizeof(HE*), char); /* zero 2nd half*/
72940dca 1241 }
1242 else {
d18c6117 1243 Newz(0, a, PERL_HV_ARRAY_ALLOC_BYTES(newsize), char);
72940dca 1244 }
cbec9347
JH
1245 xhv->xhv_max = --newsize; /* HvMAX(hv) = --newsize */
1246 xhv->xhv_array = a; /* HvARRAY(hv) = a */
1247 if (!xhv->xhv_fill /* !HvFILL(hv) */) /* skip rest if no entries */
72940dca 1248 return;
1249
72311751
GS
1250 aep = (HE**)a;
1251 for (i=0; i<oldsize; i++,aep++) {
1252 if (!*aep) /* non-existent */
72940dca 1253 continue;
72311751 1254 for (oentry = aep, entry = *aep; entry; entry = *oentry) {
72940dca 1255 if ((j = (HeHASH(entry) & newsize)) != i) {
1256 j -= i;
1257 *oentry = HeNEXT(entry);
72311751 1258 if (!(HeNEXT(entry) = aep[j]))
cbec9347 1259 xhv->xhv_fill++; /* HvFILL(hv)++ */
72311751 1260 aep[j] = entry;
72940dca 1261 continue;
1262 }
1263 else
1264 oentry = &HeNEXT(entry);
1265 }
72311751 1266 if (!*aep) /* everything moved */
cbec9347 1267 xhv->xhv_fill--; /* HvFILL(hv)-- */
72940dca 1268 }
1269}
1270
954c1994
GS
1271/*
1272=for apidoc newHV
1273
1274Creates a new HV. The reference count is set to 1.
1275
1276=cut
1277*/
1278
79072805 1279HV *
864dbfa3 1280Perl_newHV(pTHX)
79072805
LW
1281{
1282 register HV *hv;
cbec9347 1283 register XPVHV* xhv;
79072805 1284
a0d0e21e
LW
1285 hv = (HV*)NEWSV(502,0);
1286 sv_upgrade((SV *)hv, SVt_PVHV);
cbec9347 1287 xhv = (XPVHV*)SvANY(hv);
79072805
LW
1288 SvPOK_off(hv);
1289 SvNOK_off(hv);
1c846c1f 1290#ifndef NODEFAULT_SHAREKEYS
fde52b5c 1291 HvSHAREKEYS_on(hv); /* key-sharing on by default */
1c846c1f 1292#endif
4b5190b5 1293
cbec9347
JH
1294 xhv->xhv_max = 7; /* HvMAX(hv) = 7 (start with 8 buckets) */
1295 xhv->xhv_fill = 0; /* HvFILL(hv) = 0 */
1296 xhv->xhv_pmroot = 0; /* HvPMROOT(hv) = 0 */
79072805
LW
1297 (void)hv_iterinit(hv); /* so each() will start off right */
1298 return hv;
1299}
1300
b3ac6de7 1301HV *
864dbfa3 1302Perl_newHVhv(pTHX_ HV *ohv)
b3ac6de7 1303{
b56ba0bf 1304 HV *hv = newHV();
4beac62f 1305 STRLEN hv_max, hv_fill;
4beac62f
AMS
1306
1307 if (!ohv || (hv_fill = HvFILL(ohv)) == 0)
1308 return hv;
4beac62f 1309 hv_max = HvMAX(ohv);
b3ac6de7 1310
b56ba0bf
AMS
1311 if (!SvMAGICAL((SV *)ohv)) {
1312 /* It's an ordinary hash, so copy it fast. AMS 20010804 */
eb160463
GS
1313 STRLEN i;
1314 bool shared = !!HvSHAREKEYS(ohv);
b56ba0bf 1315 HE **ents, **oents = (HE **)HvARRAY(ohv);
ff875642
JH
1316 char *a;
1317 New(0, a, PERL_HV_ARRAY_ALLOC_BYTES(hv_max+1), char);
1318 ents = (HE**)a;
b56ba0bf
AMS
1319
1320 /* In each bucket... */
1321 for (i = 0; i <= hv_max; i++) {
1322 HE *prev = NULL, *ent = NULL, *oent = oents[i];
1323
1324 if (!oent) {
1325 ents[i] = NULL;
1326 continue;
1327 }
1328
1329 /* Copy the linked list of entries. */
1330 for (oent = oents[i]; oent; oent = HeNEXT(oent)) {
1331 U32 hash = HeHASH(oent);
1332 char *key = HeKEY(oent);
19692e8d
NC
1333 STRLEN len = HeKLEN(oent);
1334 int flags = HeKFLAGS(oent);
b56ba0bf
AMS
1335
1336 ent = new_HE();
45dea987 1337 HeVAL(ent) = newSVsv(HeVAL(oent));
19692e8d
NC
1338 HeKEY_hek(ent)
1339 = shared ? share_hek_flags(key, len, hash, flags)
1340 : save_hek_flags(key, len, hash, flags);
b56ba0bf
AMS
1341 if (prev)
1342 HeNEXT(prev) = ent;
1343 else
1344 ents[i] = ent;
1345 prev = ent;
1346 HeNEXT(ent) = NULL;
1347 }
1348 }
1349
1350 HvMAX(hv) = hv_max;
1351 HvFILL(hv) = hv_fill;
8aacddc1 1352 HvTOTALKEYS(hv) = HvTOTALKEYS(ohv);
b56ba0bf 1353 HvARRAY(hv) = ents;
1c846c1f 1354 }
b56ba0bf
AMS
1355 else {
1356 /* Iterate over ohv, copying keys and values one at a time. */
b3ac6de7 1357 HE *entry;
b56ba0bf
AMS
1358 I32 riter = HvRITER(ohv);
1359 HE *eiter = HvEITER(ohv);
1360
1361 /* Can we use fewer buckets? (hv_max is always 2^n-1) */
1362 while (hv_max && hv_max + 1 >= hv_fill * 2)
1363 hv_max = hv_max / 2;
1364 HvMAX(hv) = hv_max;
1365
4a76a316 1366 hv_iterinit(ohv);
e16e2ff8 1367 while ((entry = hv_iternext_flags(ohv, 0))) {
19692e8d
NC
1368 hv_store_flags(hv, HeKEY(entry), HeKLEN(entry),
1369 newSVsv(HeVAL(entry)), HeHASH(entry),
1370 HeKFLAGS(entry));
b3ac6de7 1371 }
b56ba0bf
AMS
1372 HvRITER(ohv) = riter;
1373 HvEITER(ohv) = eiter;
b3ac6de7 1374 }
1c846c1f 1375
b3ac6de7
IZ
1376 return hv;
1377}
1378
79072805 1379void
864dbfa3 1380Perl_hv_free_ent(pTHX_ HV *hv, register HE *entry)
79072805 1381{
16bdeea2
GS
1382 SV *val;
1383
68dc0745 1384 if (!entry)
79072805 1385 return;
16bdeea2 1386 val = HeVAL(entry);
257c9e5b 1387 if (val && isGV(val) && GvCVu(val) && HvNAME(hv))
3280af22 1388 PL_sub_generation++; /* may be deletion of method from stash */
16bdeea2 1389 SvREFCNT_dec(val);
68dc0745 1390 if (HeKLEN(entry) == HEf_SVKEY) {
1391 SvREFCNT_dec(HeKEY_sv(entry));
8aacddc1 1392 Safefree(HeKEY_hek(entry));
44a8e56a 1393 }
1394 else if (HvSHAREKEYS(hv))
68dc0745 1395 unshare_hek(HeKEY_hek(entry));
fde52b5c 1396 else
68dc0745 1397 Safefree(HeKEY_hek(entry));
d33b2eba 1398 del_HE(entry);
79072805
LW
1399}
1400
1401void
864dbfa3 1402Perl_hv_delayfree_ent(pTHX_ HV *hv, register HE *entry)
79072805 1403{
68dc0745 1404 if (!entry)
79072805 1405 return;
68dc0745 1406 if (isGV(HeVAL(entry)) && GvCVu(HeVAL(entry)) && HvNAME(hv))
3280af22 1407 PL_sub_generation++; /* may be deletion of method from stash */
68dc0745 1408 sv_2mortal(HeVAL(entry)); /* free between statements */
1409 if (HeKLEN(entry) == HEf_SVKEY) {
1410 sv_2mortal(HeKEY_sv(entry));
1411 Safefree(HeKEY_hek(entry));
44a8e56a 1412 }
1413 else if (HvSHAREKEYS(hv))
68dc0745 1414 unshare_hek(HeKEY_hek(entry));
fde52b5c 1415 else
68dc0745 1416 Safefree(HeKEY_hek(entry));
d33b2eba 1417 del_HE(entry);
79072805
LW
1418}
1419
954c1994
GS
1420/*
1421=for apidoc hv_clear
1422
1423Clears a hash, making it empty.
1424
1425=cut
1426*/
1427
79072805 1428void
864dbfa3 1429Perl_hv_clear(pTHX_ HV *hv)
79072805 1430{
cbec9347 1431 register XPVHV* xhv;
79072805
LW
1432 if (!hv)
1433 return;
49293501 1434
ecae49c0
NC
1435 DEBUG_A(Perl_hv_assert(aTHX_ hv));
1436
34c3c4e3
DM
1437 xhv = (XPVHV*)SvANY(hv);
1438
5f099cb0 1439 if (SvREADONLY(hv) && xhv->xhv_array != NULL) {
34c3c4e3 1440 /* restricted hash: convert all keys to placeholders */
3a676441
JH
1441 I32 i;
1442 HE* entry;
1443 for (i = 0; i <= (I32) xhv->xhv_max; i++) {
1444 entry = ((HE**)xhv->xhv_array)[i];
1445 for (; entry; entry = HeNEXT(entry)) {
1446 /* not already placeholder */
7996736c 1447 if (HeVAL(entry) != &PL_sv_placeholder) {
3a676441
JH
1448 if (HeVAL(entry) && SvREADONLY(HeVAL(entry))) {
1449 SV* keysv = hv_iterkeysv(entry);
1450 Perl_croak(aTHX_
1451 "Attempt to delete readonly key '%"SVf"' from a restricted hash",
1452 keysv);
1453 }
1454 SvREFCNT_dec(HeVAL(entry));
7996736c 1455 HeVAL(entry) = &PL_sv_placeholder;
3a676441
JH
1456 xhv->xhv_placeholders++; /* HvPLACEHOLDERS(hv)++ */
1457 }
34c3c4e3
DM
1458 }
1459 }
df8c6964 1460 goto reset;
49293501
MS
1461 }
1462
463ee0b2 1463 hfreeentries(hv);
8aacddc1 1464 xhv->xhv_placeholders = 0; /* HvPLACEHOLDERS(hv) = 0 */
cbec9347
JH
1465 if (xhv->xhv_array /* HvARRAY(hv) */)
1466 (void)memzero(xhv->xhv_array /* HvARRAY(hv) */,
1467 (xhv->xhv_max+1 /* HvMAX(hv)+1 */) * sizeof(HE*));
a0d0e21e
LW
1468
1469 if (SvRMAGICAL(hv))
1c846c1f 1470 mg_clear((SV*)hv);
574c8022 1471
19692e8d 1472 HvHASKFLAGS_off(hv);
bb443f97 1473 HvREHASH_off(hv);
df8c6964
TP
1474 reset:
1475 HvEITER(hv) = NULL;
79072805
LW
1476}
1477
3540d4ce
AB
1478/*
1479=for apidoc hv_clear_placeholders
1480
1481Clears any placeholders from a hash. If a restricted hash has any of its keys
1482marked as readonly and the key is subsequently deleted, the key is not actually
1483deleted but is marked by assigning it a value of &PL_sv_placeholder. This tags
1484it so it will be ignored by future operations such as iterating over the hash,
1485but will still allow the hash to have a value reaasigned to the key at some
1486future point. This function clears any such placeholder keys from the hash.
1487See Hash::Util::lock_keys() for an example of its use.
1488
1489=cut
1490*/
1491
1492void
1493Perl_hv_clear_placeholders(pTHX_ HV *hv)
1494{
d3677389
NC
1495 I32 items = (I32)HvPLACEHOLDERS(hv);
1496 I32 i = HvMAX(hv);
1497
1498 if (items == 0)
1499 return;
1500
1501 do {
1502 /* Loop down the linked list heads */
1503 int first = 1;
1504 HE **oentry = &(HvARRAY(hv))[i];
1505 HE *entry = *oentry;
1506
1507 if (!entry)
1508 continue;
1509
213ce8b3 1510 for (; entry; entry = *oentry) {
d3677389
NC
1511 if (HeVAL(entry) == &PL_sv_placeholder) {
1512 *oentry = HeNEXT(entry);
1513 if (first && !*oentry)
1514 HvFILL(hv)--; /* This linked list is now empty. */
1515 if (HvEITER(hv))
1516 HvLAZYDEL_on(hv);
1517 else
1518 hv_free_ent(hv, entry);
1519
1520 if (--items == 0) {
1521 /* Finished. */
1522 HvTOTALKEYS(hv) -= HvPLACEHOLDERS(hv);
1523 if (HvKEYS(hv) == 0)
1524 HvHASKFLAGS_off(hv);
1525 HvPLACEHOLDERS(hv) = 0;
1526 return;
1527 }
213ce8b3
NC
1528 } else {
1529 oentry = &HeNEXT(entry);
1530 first = 0;
d3677389
NC
1531 }
1532 }
1533 } while (--i >= 0);
1534 /* You can't get here, hence assertion should always fail. */
1535 assert (items == 0);
1536 assert (0);
3540d4ce
AB
1537}
1538
76e3520e 1539STATIC void
cea2e8a9 1540S_hfreeentries(pTHX_ HV *hv)
79072805 1541{
a0d0e21e 1542 register HE **array;
68dc0745 1543 register HE *entry;
1544 register HE *oentry = Null(HE*);
a0d0e21e
LW
1545 I32 riter;
1546 I32 max;
79072805
LW
1547
1548 if (!hv)
1549 return;
a0d0e21e 1550 if (!HvARRAY(hv))
79072805 1551 return;
a0d0e21e
LW
1552
1553 riter = 0;
1554 max = HvMAX(hv);
1555 array = HvARRAY(hv);
2f86008e
DM
1556 /* make everyone else think the array is empty, so that the destructors
1557 * called for freed entries can't recusively mess with us */
1558 HvARRAY(hv) = Null(HE**);
1559 HvFILL(hv) = 0;
1560 ((XPVHV*) SvANY(hv))->xhv_keys = 0;
1561
68dc0745 1562 entry = array[0];
a0d0e21e 1563 for (;;) {
68dc0745 1564 if (entry) {
1565 oentry = entry;
1566 entry = HeNEXT(entry);
1567 hv_free_ent(hv, oentry);
a0d0e21e 1568 }
68dc0745 1569 if (!entry) {
a0d0e21e
LW
1570 if (++riter > max)
1571 break;
68dc0745 1572 entry = array[riter];
1c846c1f 1573 }
79072805 1574 }
2f86008e 1575 HvARRAY(hv) = array;
a0d0e21e 1576 (void)hv_iterinit(hv);
79072805
LW
1577}
1578
954c1994
GS
1579/*
1580=for apidoc hv_undef
1581
1582Undefines the hash.
1583
1584=cut
1585*/
1586
79072805 1587void
864dbfa3 1588Perl_hv_undef(pTHX_ HV *hv)
79072805 1589{
cbec9347 1590 register XPVHV* xhv;
79072805
LW
1591 if (!hv)
1592 return;
ecae49c0 1593 DEBUG_A(Perl_hv_assert(aTHX_ hv));
cbec9347 1594 xhv = (XPVHV*)SvANY(hv);
463ee0b2 1595 hfreeentries(hv);
cbec9347 1596 Safefree(xhv->xhv_array /* HvARRAY(hv) */);
85e6fe83 1597 if (HvNAME(hv)) {
7e8961ec
AB
1598 if(PL_stashcache)
1599 hv_delete(PL_stashcache, HvNAME(hv), strlen(HvNAME(hv)), G_DISCARD);
85e6fe83
LW
1600 Safefree(HvNAME(hv));
1601 HvNAME(hv) = 0;
1602 }
cbec9347
JH
1603 xhv->xhv_max = 7; /* HvMAX(hv) = 7 (it's a normal hash) */
1604 xhv->xhv_array = 0; /* HvARRAY(hv) = 0 */
8aacddc1 1605 xhv->xhv_placeholders = 0; /* HvPLACEHOLDERS(hv) = 0 */
a0d0e21e
LW
1606
1607 if (SvRMAGICAL(hv))
1c846c1f 1608 mg_clear((SV*)hv);
79072805
LW
1609}
1610
954c1994
GS
1611/*
1612=for apidoc hv_iterinit
1613
1614Prepares a starting point to traverse a hash table. Returns the number of
1615keys in the hash (i.e. the same as C<HvKEYS(tb)>). The return value is
1c846c1f 1616currently only meaningful for hashes without tie magic.
954c1994
GS
1617
1618NOTE: Before version 5.004_65, C<hv_iterinit> used to return the number of
1619hash buckets that happen to be in use. If you still need that esoteric
1620value, you can get it through the macro C<HvFILL(tb)>.
1621
e16e2ff8 1622
954c1994
GS
1623=cut
1624*/
1625
79072805 1626I32
864dbfa3 1627Perl_hv_iterinit(pTHX_ HV *hv)
79072805 1628{
cbec9347 1629 register XPVHV* xhv;
aa689395 1630 HE *entry;
1631
1632 if (!hv)
cea2e8a9 1633 Perl_croak(aTHX_ "Bad hash");
cbec9347
JH
1634 xhv = (XPVHV*)SvANY(hv);
1635 entry = xhv->xhv_eiter; /* HvEITER(hv) */
72940dca 1636 if (entry && HvLAZYDEL(hv)) { /* was deleted earlier? */
1637 HvLAZYDEL_off(hv);
68dc0745 1638 hv_free_ent(hv, entry);
72940dca 1639 }
cbec9347
JH
1640 xhv->xhv_riter = -1; /* HvRITER(hv) = -1 */
1641 xhv->xhv_eiter = Null(HE*); /* HvEITER(hv) = Null(HE*) */
1642 /* used to be xhv->xhv_fill before 5.004_65 */
8aacddc1 1643 return XHvTOTALKEYS(xhv);
79072805 1644}
954c1994
GS
1645/*
1646=for apidoc hv_iternext
1647
1648Returns entries from a hash iterator. See C<hv_iterinit>.
1649
fe7bca90
NC
1650You may call C<hv_delete> or C<hv_delete_ent> on the hash entry that the
1651iterator currently points to, without losing your place or invalidating your
1652iterator. Note that in this case the current entry is deleted from the hash
1653with your iterator holding the last reference to it. Your iterator is flagged
1654to free the entry on the next call to C<hv_iternext>, so you must not discard
1655your iterator immediately else the entry will leak - call C<hv_iternext> to
1656trigger the resource deallocation.
1657
954c1994
GS
1658=cut
1659*/
1660
79072805 1661HE *
864dbfa3 1662Perl_hv_iternext(pTHX_ HV *hv)
79072805 1663{
e16e2ff8
NC
1664 return hv_iternext_flags(hv, 0);
1665}
1666
1667/*
fe7bca90
NC
1668=for apidoc hv_iternext_flags
1669
1670Returns entries from a hash iterator. See C<hv_iterinit> and C<hv_iternext>.
1671The C<flags> value will normally be zero; if HV_ITERNEXT_WANTPLACEHOLDERS is
1672set the placeholders keys (for restricted hashes) will be returned in addition
1673to normal keys. By default placeholders are automatically skipped over.
7996736c
MHM
1674Currently a placeholder is implemented with a value that is
1675C<&Perl_sv_placeholder>. Note that the implementation of placeholders and
fe7bca90
NC
1676restricted hashes may change, and the implementation currently is
1677insufficiently abstracted for any change to be tidy.
e16e2ff8 1678
fe7bca90 1679=cut
e16e2ff8
NC
1680*/
1681
1682HE *
1683Perl_hv_iternext_flags(pTHX_ HV *hv, I32 flags)
1684{
cbec9347 1685 register XPVHV* xhv;
79072805 1686 register HE *entry;
a0d0e21e 1687 HE *oldentry;
463ee0b2 1688 MAGIC* mg;
79072805
LW
1689
1690 if (!hv)
cea2e8a9 1691 Perl_croak(aTHX_ "Bad hash");
cbec9347
JH
1692 xhv = (XPVHV*)SvANY(hv);
1693 oldentry = entry = xhv->xhv_eiter; /* HvEITER(hv) */
463ee0b2 1694
14befaf4 1695 if ((mg = SvTIED_mg((SV*)hv, PERL_MAGIC_tied))) {
8990e307 1696 SV *key = sv_newmortal();
cd1469e6 1697 if (entry) {
fde52b5c 1698 sv_setsv(key, HeSVKEY_force(entry));
cd1469e6 1699 SvREFCNT_dec(HeSVKEY(entry)); /* get rid of previous key */
1700 }
a0d0e21e 1701 else {
ff68c719 1702 char *k;
bbce6d69 1703 HEK *hek;
ff68c719 1704
cbec9347
JH
1705 /* one HE per MAGICAL hash */
1706 xhv->xhv_eiter = entry = new_HE(); /* HvEITER(hv) = new_HE() */
4633a7c4 1707 Zero(entry, 1, HE);
ff68c719 1708 Newz(54, k, HEK_BASESIZE + sizeof(SV*), char);
1709 hek = (HEK*)k;
1710 HeKEY_hek(entry) = hek;
fde52b5c 1711 HeKLEN(entry) = HEf_SVKEY;
a0d0e21e
LW
1712 }
1713 magic_nextpack((SV*) hv,mg,key);
8aacddc1 1714 if (SvOK(key)) {
cd1469e6 1715 /* force key to stay around until next time */
bbce6d69 1716 HeSVKEY_set(entry, SvREFCNT_inc(key));
1717 return entry; /* beware, hent_val is not set */
8aacddc1 1718 }
fde52b5c 1719 if (HeVAL(entry))
1720 SvREFCNT_dec(HeVAL(entry));
ff68c719 1721 Safefree(HeKEY_hek(entry));
d33b2eba 1722 del_HE(entry);
cbec9347 1723 xhv->xhv_eiter = Null(HE*); /* HvEITER(hv) = Null(HE*) */
463ee0b2 1724 return Null(HE*);
79072805 1725 }
f675dbe5 1726#ifdef DYNAMIC_ENV_FETCH /* set up %ENV for iteration */
cbec9347 1727 if (!entry && SvRMAGICAL((SV*)hv) && mg_find((SV*)hv, PERL_MAGIC_env))
f675dbe5
CB
1728 prime_env_iter();
1729#endif
463ee0b2 1730
cbec9347
JH
1731 if (!xhv->xhv_array /* !HvARRAY(hv) */)
1732 Newz(506, xhv->xhv_array /* HvARRAY(hv) */,
1733 PERL_HV_ARRAY_ALLOC_BYTES(xhv->xhv_max+1 /* HvMAX(hv)+1 */),
1734 char);
015a5f36 1735 /* At start of hash, entry is NULL. */
fde52b5c 1736 if (entry)
8aacddc1 1737 {
fde52b5c 1738 entry = HeNEXT(entry);
e16e2ff8
NC
1739 if (!(flags & HV_ITERNEXT_WANTPLACEHOLDERS)) {
1740 /*
1741 * Skip past any placeholders -- don't want to include them in
1742 * any iteration.
1743 */
7996736c 1744 while (entry && HeVAL(entry) == &PL_sv_placeholder) {
e16e2ff8
NC
1745 entry = HeNEXT(entry);
1746 }
8aacddc1
NIS
1747 }
1748 }
fde52b5c 1749 while (!entry) {
015a5f36
NC
1750 /* OK. Come to the end of the current list. Grab the next one. */
1751
cbec9347 1752 xhv->xhv_riter++; /* HvRITER(hv)++ */
eb160463 1753 if (xhv->xhv_riter > (I32)xhv->xhv_max /* HvRITER(hv) > HvMAX(hv) */) {
015a5f36 1754 /* There is no next one. End of the hash. */
cbec9347 1755 xhv->xhv_riter = -1; /* HvRITER(hv) = -1 */
fde52b5c 1756 break;
79072805 1757 }
cbec9347
JH
1758 /* entry = (HvARRAY(hv))[HvRITER(hv)]; */
1759 entry = ((HE**)xhv->xhv_array)[xhv->xhv_riter];
8aacddc1 1760
e16e2ff8 1761 if (!(flags & HV_ITERNEXT_WANTPLACEHOLDERS)) {
015a5f36
NC
1762 /* If we have an entry, but it's a placeholder, don't count it.
1763 Try the next. */
7996736c 1764 while (entry && HeVAL(entry) == &PL_sv_placeholder)
015a5f36
NC
1765 entry = HeNEXT(entry);
1766 }
1767 /* Will loop again if this linked list starts NULL
1768 (for HV_ITERNEXT_WANTPLACEHOLDERS)
1769 or if we run through it and find only placeholders. */
fde52b5c 1770 }
79072805 1771
72940dca 1772 if (oldentry && HvLAZYDEL(hv)) { /* was deleted earlier? */
1773 HvLAZYDEL_off(hv);
68dc0745 1774 hv_free_ent(hv, oldentry);
72940dca 1775 }
a0d0e21e 1776
fdcd69b6
NC
1777 /*if (HvREHASH(hv) && entry && !HeKREHASH(entry))
1778 PerlIO_printf(PerlIO_stderr(), "Awooga %p %p\n", hv, entry);*/
1779
cbec9347 1780 xhv->xhv_eiter = entry; /* HvEITER(hv) = entry */
79072805
LW
1781 return entry;
1782}
1783
954c1994
GS
1784/*
1785=for apidoc hv_iterkey
1786
1787Returns the key from the current position of the hash iterator. See
1788C<hv_iterinit>.
1789
1790=cut
1791*/
1792
79072805 1793char *
864dbfa3 1794Perl_hv_iterkey(pTHX_ register HE *entry, I32 *retlen)
79072805 1795{
fde52b5c 1796 if (HeKLEN(entry) == HEf_SVKEY) {
fb73857a 1797 STRLEN len;
1798 char *p = SvPV(HeKEY_sv(entry), len);
1799 *retlen = len;
1800 return p;
fde52b5c 1801 }
1802 else {
1803 *retlen = HeKLEN(entry);
1804 return HeKEY(entry);
1805 }
1806}
1807
1808/* unlike hv_iterval(), this always returns a mortal copy of the key */
954c1994
GS
1809/*
1810=for apidoc hv_iterkeysv
1811
1812Returns the key as an C<SV*> from the current position of the hash
1813iterator. The return value will always be a mortal copy of the key. Also
1814see C<hv_iterinit>.
1815
1816=cut
1817*/
1818
fde52b5c 1819SV *
864dbfa3 1820Perl_hv_iterkeysv(pTHX_ register HE *entry)
fde52b5c 1821{
19692e8d
NC
1822 if (HeKLEN(entry) != HEf_SVKEY) {
1823 HEK *hek = HeKEY_hek(entry);
1824 int flags = HEK_FLAGS(hek);
1825 SV *sv;
1826
1827 if (flags & HVhek_WASUTF8) {
1828 /* Trouble :-)
1829 Andreas would like keys he put in as utf8 to come back as utf8
1830 */
1831 STRLEN utf8_len = HEK_LEN(hek);
2e5dfef7 1832 U8 *as_utf8 = bytes_to_utf8 ((U8*)HEK_KEY(hek), &utf8_len);
19692e8d 1833
2e5dfef7 1834 sv = newSVpvn ((char*)as_utf8, utf8_len);
19692e8d 1835 SvUTF8_on (sv);
c193270f 1836 Safefree (as_utf8); /* bytes_to_utf8() allocates a new string */
4b5190b5
NC
1837 } else if (flags & HVhek_REHASH) {
1838 /* We don't have a pointer to the hv, so we have to replicate the
1839 flag into every HEK. This hv is using custom a hasing
1840 algorithm. Hence we can't return a shared string scalar, as
1841 that would contain the (wrong) hash value, and might get passed
1842 into an hv routine with a regular hash */
1843
1844 sv = newSVpvn (HEK_KEY(hek), HEK_LEN(hek));
1845 if (HEK_UTF8(hek))
1846 SvUTF8_on (sv);
1847 } else {
19692e8d
NC
1848 sv = newSVpvn_share(HEK_KEY(hek),
1849 (HEK_UTF8(hek) ? -HEK_LEN(hek) : HEK_LEN(hek)),
1850 HEK_HASH(hek));
1851 }
1852 return sv_2mortal(sv);
1853 }
1854 return sv_mortalcopy(HeKEY_sv(entry));
79072805
LW
1855}
1856
954c1994
GS
1857/*
1858=for apidoc hv_iterval
1859
1860Returns the value from the current position of the hash iterator. See
1861C<hv_iterkey>.
1862
1863=cut
1864*/
1865
79072805 1866SV *
864dbfa3 1867Perl_hv_iterval(pTHX_ HV *hv, register HE *entry)
79072805 1868{
8990e307 1869 if (SvRMAGICAL(hv)) {
14befaf4 1870 if (mg_find((SV*)hv, PERL_MAGIC_tied)) {
8990e307 1871 SV* sv = sv_newmortal();
bbce6d69 1872 if (HeKLEN(entry) == HEf_SVKEY)
1873 mg_copy((SV*)hv, sv, (char*)HeKEY_sv(entry), HEf_SVKEY);
1874 else mg_copy((SV*)hv, sv, HeKEY(entry), HeKLEN(entry));
463ee0b2
LW
1875 return sv;
1876 }
79072805 1877 }
fde52b5c 1878 return HeVAL(entry);
79072805
LW
1879}
1880
954c1994
GS
1881/*
1882=for apidoc hv_iternextsv
1883
1884Performs an C<hv_iternext>, C<hv_iterkey>, and C<hv_iterval> in one
1885operation.
1886
1887=cut
1888*/
1889
a0d0e21e 1890SV *
864dbfa3 1891Perl_hv_iternextsv(pTHX_ HV *hv, char **key, I32 *retlen)
a0d0e21e
LW
1892{
1893 HE *he;
e16e2ff8 1894 if ( (he = hv_iternext_flags(hv, 0)) == NULL)
a0d0e21e
LW
1895 return NULL;
1896 *key = hv_iterkey(he, retlen);
1897 return hv_iterval(hv, he);
1898}
1899
954c1994
GS
1900/*
1901=for apidoc hv_magic
1902
1903Adds magic to a hash. See C<sv_magic>.
1904
1905=cut
1906*/
1907
79072805 1908void
864dbfa3 1909Perl_hv_magic(pTHX_ HV *hv, GV *gv, int how)
79072805 1910{
a0d0e21e 1911 sv_magic((SV*)hv, (SV*)gv, how, Nullch, 0);
79072805 1912}
fde52b5c 1913
37d85e3a
JH
1914#if 0 /* use the macro from hv.h instead */
1915
bbce6d69 1916char*
864dbfa3 1917Perl_sharepvn(pTHX_ const char *sv, I32 len, U32 hash)
bbce6d69 1918{
ff68c719 1919 return HEK_KEY(share_hek(sv, len, hash));
bbce6d69 1920}
1921
37d85e3a
JH
1922#endif
1923
bbce6d69 1924/* possibly free a shared string if no one has access to it
fde52b5c 1925 * len and hash must both be valid for str.
1926 */
bbce6d69 1927void
864dbfa3 1928Perl_unsharepvn(pTHX_ const char *str, I32 len, U32 hash)
fde52b5c 1929{
19692e8d
NC
1930 unshare_hek_or_pvn (NULL, str, len, hash);
1931}
1932
1933
1934void
1935Perl_unshare_hek(pTHX_ HEK *hek)
1936{
1937 unshare_hek_or_pvn(hek, NULL, 0, 0);
1938}
1939
1940/* possibly free a shared string if no one has access to it
1941 hek if non-NULL takes priority over the other 3, else str, len and hash
1942 are used. If so, len and hash must both be valid for str.
1943 */
df132699 1944STATIC void
19692e8d
NC
1945S_unshare_hek_or_pvn(pTHX_ HEK *hek, const char *str, I32 len, U32 hash)
1946{
cbec9347 1947 register XPVHV* xhv;
fde52b5c 1948 register HE *entry;
1949 register HE **oentry;
1950 register I32 i = 1;
1951 I32 found = 0;
c3654f1a 1952 bool is_utf8 = FALSE;
19692e8d 1953 int k_flags = 0;
f9a63242 1954 const char *save = str;
c3654f1a 1955
19692e8d
NC
1956 if (hek) {
1957 hash = HEK_HASH(hek);
1958 } else if (len < 0) {
1959 STRLEN tmplen = -len;
1960 is_utf8 = TRUE;
1961 /* See the note in hv_fetch(). --jhi */
1962 str = (char*)bytes_from_utf8((U8*)str, &tmplen, &is_utf8);
1963 len = tmplen;
1964 if (is_utf8)
1965 k_flags = HVhek_UTF8;
1966 if (str != save)
1967 k_flags |= HVhek_WASUTF8 | HVhek_FREEKEY;
c3654f1a 1968 }
1c846c1f 1969
fde52b5c 1970 /* what follows is the moral equivalent of:
6b88bc9c 1971 if ((Svp = hv_fetch(PL_strtab, tmpsv, FALSE, hash))) {
bbce6d69 1972 if (--*Svp == Nullsv)
6b88bc9c 1973 hv_delete(PL_strtab, str, len, G_DISCARD, hash);
bbce6d69 1974 } */
cbec9347 1975 xhv = (XPVHV*)SvANY(PL_strtab);
fde52b5c 1976 /* assert(xhv_array != 0) */
5f08fbcd 1977 LOCK_STRTAB_MUTEX;
cbec9347
JH
1978 /* oentry = &(HvARRAY(hv))[hash & (I32) HvMAX(hv)]; */
1979 oentry = &((HE**)xhv->xhv_array)[hash & (I32) xhv->xhv_max];
19692e8d
NC
1980 if (hek) {
1981 for (entry = *oentry; entry; i=0, oentry = &HeNEXT(entry), entry = *oentry) {
1982 if (HeKEY_hek(entry) != hek)
1983 continue;
1984 found = 1;
1985 break;
1986 }
1987 } else {
1988 int flags_masked = k_flags & HVhek_MASK;
1989 for (entry = *oentry; entry; i=0, oentry = &HeNEXT(entry), entry = *oentry) {
1990 if (HeHASH(entry) != hash) /* strings can't be equal */
1991 continue;
1992 if (HeKLEN(entry) != len)
1993 continue;
1994 if (HeKEY(entry) != str && memNE(HeKEY(entry),str,len)) /* is this it? */
1995 continue;
1996 if (HeKFLAGS(entry) != flags_masked)
1997 continue;
1998 found = 1;
1999 break;
2000 }
2001 }
2002
2003 if (found) {
2004 if (--HeVAL(entry) == Nullsv) {
2005 *oentry = HeNEXT(entry);
2006 if (i && !*oentry)
2007 xhv->xhv_fill--; /* HvFILL(hv)-- */
2008 Safefree(HeKEY_hek(entry));
2009 del_HE(entry);
2010 xhv->xhv_keys--; /* HvKEYS(hv)-- */
2011 }
fde52b5c 2012 }
19692e8d 2013
333f433b 2014 UNLOCK_STRTAB_MUTEX;
411caa50 2015 if (!found && ckWARN_d(WARN_INTERNAL))
19692e8d
NC
2016 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
2017 "Attempt to free non-existent shared string '%s'%s",
2018 hek ? HEK_KEY(hek) : str,
2019 (k_flags & HVhek_UTF8) ? " (utf8)" : "");
2020 if (k_flags & HVhek_FREEKEY)
2021 Safefree(str);
fde52b5c 2022}
2023
bbce6d69 2024/* get a (constant) string ptr from the global string table
2025 * string will get added if it is not already there.
fde52b5c 2026 * len and hash must both be valid for str.
2027 */
bbce6d69 2028HEK *
864dbfa3 2029Perl_share_hek(pTHX_ const char *str, I32 len, register U32 hash)
fde52b5c 2030{
da58a35d 2031 bool is_utf8 = FALSE;
19692e8d 2032 int flags = 0;
f9a63242 2033 const char *save = str;
da58a35d
JH
2034
2035 if (len < 0) {
77caf834 2036 STRLEN tmplen = -len;
da58a35d 2037 is_utf8 = TRUE;
77caf834
JH
2038 /* See the note in hv_fetch(). --jhi */
2039 str = (char*)bytes_from_utf8((U8*)str, &tmplen, &is_utf8);
2040 len = tmplen;
19692e8d
NC
2041 /* If we were able to downgrade here, then than means that we were passed
2042 in a key which only had chars 0-255, but was utf8 encoded. */
2043 if (is_utf8)
2044 flags = HVhek_UTF8;
2045 /* If we found we were able to downgrade the string to bytes, then
2046 we should flag that it needs upgrading on keys or each. Also flag
2047 that we need share_hek_flags to free the string. */
2048 if (str != save)
2049 flags |= HVhek_WASUTF8 | HVhek_FREEKEY;
2050 }
2051
2052 return share_hek_flags (str, len, hash, flags);
2053}
2054
df132699 2055STATIC HEK *
19692e8d
NC
2056S_share_hek_flags(pTHX_ const char *str, I32 len, register U32 hash, int flags)
2057{
2058 register XPVHV* xhv;
2059 register HE *entry;
2060 register HE **oentry;
2061 register I32 i = 1;
2062 I32 found = 0;
2063 int flags_masked = flags & HVhek_MASK;
bbce6d69 2064
fde52b5c 2065 /* what follows is the moral equivalent of:
1c846c1f 2066
6b88bc9c 2067 if (!(Svp = hv_fetch(PL_strtab, str, len, FALSE)))
8aacddc1 2068 hv_store(PL_strtab, str, len, Nullsv, hash);
fdcd69b6
NC
2069
2070 Can't rehash the shared string table, so not sure if it's worth
2071 counting the number of entries in the linked list
bbce6d69 2072 */
cbec9347 2073 xhv = (XPVHV*)SvANY(PL_strtab);
fde52b5c 2074 /* assert(xhv_array != 0) */
5f08fbcd 2075 LOCK_STRTAB_MUTEX;
cbec9347
JH
2076 /* oentry = &(HvARRAY(hv))[hash & (I32) HvMAX(hv)]; */
2077 oentry = &((HE**)xhv->xhv_array)[hash & (I32) xhv->xhv_max];
bbce6d69 2078 for (entry = *oentry; entry; i=0, entry = HeNEXT(entry)) {
fde52b5c 2079 if (HeHASH(entry) != hash) /* strings can't be equal */
2080 continue;
2081 if (HeKLEN(entry) != len)
2082 continue;
1c846c1f 2083 if (HeKEY(entry) != str && memNE(HeKEY(entry),str,len)) /* is this it? */
fde52b5c 2084 continue;
19692e8d 2085 if (HeKFLAGS(entry) != flags_masked)
c3654f1a 2086 continue;
fde52b5c 2087 found = 1;
fde52b5c 2088 break;
2089 }
bbce6d69 2090 if (!found) {
d33b2eba 2091 entry = new_HE();
dcf933a4 2092 HeKEY_hek(entry) = save_hek_flags(str, len, hash, flags_masked);
bbce6d69 2093 HeVAL(entry) = Nullsv;
2094 HeNEXT(entry) = *oentry;
2095 *oentry = entry;
cbec9347 2096 xhv->xhv_keys++; /* HvKEYS(hv)++ */
bbce6d69 2097 if (i) { /* initial entry? */
cbec9347 2098 xhv->xhv_fill++; /* HvFILL(hv)++ */
4c9cc595 2099 } else if (xhv->xhv_keys > (IV)xhv->xhv_max /* HvKEYS(hv) > HvMAX(hv) */) {
cbec9347 2100 hsplit(PL_strtab);
bbce6d69 2101 }
2102 }
2103
2104 ++HeVAL(entry); /* use value slot as REFCNT */
5f08fbcd 2105 UNLOCK_STRTAB_MUTEX;
19692e8d
NC
2106
2107 if (flags & HVhek_FREEKEY)
f9a63242 2108 Safefree(str);
19692e8d 2109
ff68c719 2110 return HeKEY_hek(entry);
fde52b5c 2111}
ecae49c0
NC
2112
2113
2114/*
2115=for apidoc hv_assert
2116
2117Check that a hash is in an internally consistent state.
2118
2119=cut
2120*/
2121
2122void
2123Perl_hv_assert(pTHX_ HV *hv)
2124{
2125 HE* entry;
2126 int withflags = 0;
2127 int placeholders = 0;
2128 int real = 0;
2129 int bad = 0;
2130 I32 riter = HvRITER(hv);
2131 HE *eiter = HvEITER(hv);
2132
2133 (void)hv_iterinit(hv);
2134
2135 while ((entry = hv_iternext_flags(hv, HV_ITERNEXT_WANTPLACEHOLDERS))) {
2136 /* sanity check the values */
2137 if (HeVAL(entry) == &PL_sv_placeholder) {
2138 placeholders++;
2139 } else {
2140 real++;
2141 }
2142 /* sanity check the keys */
2143 if (HeSVKEY(entry)) {
2144 /* Don't know what to check on SV keys. */
2145 } else if (HeKUTF8(entry)) {
2146 withflags++;
2147 if (HeKWASUTF8(entry)) {
2148 PerlIO_printf(Perl_debug_log,
2149 "hash key has both WASUFT8 and UTF8: '%.*s'\n",
2150 (int) HeKLEN(entry), HeKEY(entry));
2151 bad = 1;
2152 }
2153 } else if (HeKWASUTF8(entry)) {
2154 withflags++;
2155 }
2156 }
2157 if (!SvTIED_mg((SV*)hv, PERL_MAGIC_tied)) {
2158 if (HvUSEDKEYS(hv) != real) {
2159 PerlIO_printf(Perl_debug_log, "Count %d key(s), but hash reports %d\n",
2160 (int) real, (int) HvUSEDKEYS(hv));
2161 bad = 1;
2162 }
2163 if (HvPLACEHOLDERS(hv) != placeholders) {
2164 PerlIO_printf(Perl_debug_log,
2165 "Count %d placeholder(s), but hash reports %d\n",
2166 (int) placeholders, (int) HvPLACEHOLDERS(hv));
2167 bad = 1;
2168 }
2169 }
2170 if (withflags && ! HvHASKFLAGS(hv)) {
2171 PerlIO_printf(Perl_debug_log,
2172 "Hash has HASKFLAGS off but I count %d key(s) with flags\n",
2173 withflags);
2174 bad = 1;
2175 }
2176 if (bad) {
2177 sv_dump((SV *)hv);
2178 }
2179 HvRITER(hv) = riter; /* Restore hash iterator state */
2180 HvEITER(hv) = eiter;
2181}