This is a live mirror of the Perl 5 development currently hosted at https://github.com/perl/perl5
Explode if anyone attempts to sv_upgrade PL_mess_sv.
[perl5.git] / hv.c
CommitLineData
a0d0e21e 1/* hv.c
79072805 2 *
4bb101f2 3 * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999,
af3babe4 4 * 2000, 2001, 2002, 2003, 2004, 2005, by Larry Wall and others
79072805
LW
5 *
6 * You may distribute under the terms of either the GNU General Public
7 * License or the Artistic License, as specified in the README file.
8 *
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
166f8a29
DM
17
18A HV structure represents a Perl hash. It consists mainly of an array
19of pointers, each of which points to a linked list of HE structures. The
20array is indexed by the hash function of the key, so each linked list
21represents all the hash entries with the same hash value. Each HE contains
22a pointer to the actual value, plus a pointer to a HEK structure which
23holds the key and hash value.
24
25=cut
26
d5afce77
RB
27*/
28
79072805 29#include "EXTERN.h"
864dbfa3 30#define PERL_IN_HV_C
3d78eb94 31#define PERL_HASH_INTERNAL_ACCESS
79072805
LW
32#include "perl.h"
33
d8012aaf 34#define HV_MAX_LENGTH_BEFORE_SPLIT 14
fdcd69b6 35
76e3520e 36STATIC HE*
cea2e8a9 37S_new_he(pTHX)
4633a7c4
LW
38{
39 HE* he;
333f433b
DG
40 LOCK_SV_MUTEX;
41 if (!PL_he_root)
8aacddc1 42 more_he();
333f433b
DG
43 he = PL_he_root;
44 PL_he_root = HeNEXT(he);
45 UNLOCK_SV_MUTEX;
46 return he;
4633a7c4
LW
47}
48
76e3520e 49STATIC void
cea2e8a9 50S_del_he(pTHX_ HE *p)
4633a7c4 51{
333f433b 52 LOCK_SV_MUTEX;
3280af22
NIS
53 HeNEXT(p) = (HE*)PL_he_root;
54 PL_he_root = p;
333f433b 55 UNLOCK_SV_MUTEX;
4633a7c4
LW
56}
57
333f433b 58STATIC void
cea2e8a9 59S_more_he(pTHX)
4633a7c4
LW
60{
61 register HE* he;
62 register HE* heend;
612f20c3
GS
63 XPV *ptr;
64 New(54, ptr, 1008/sizeof(XPV), XPV);
65 ptr->xpv_pv = (char*)PL_he_arenaroot;
66 PL_he_arenaroot = ptr;
67
68 he = (HE*)ptr;
4633a7c4 69 heend = &he[1008 / sizeof(HE) - 1];
612f20c3 70 PL_he_root = ++he;
4633a7c4 71 while (he < heend) {
8aacddc1
NIS
72 HeNEXT(he) = (HE*)(he + 1);
73 he++;
4633a7c4 74 }
fde52b5c 75 HeNEXT(he) = 0;
4633a7c4
LW
76}
77
d33b2eba
GS
78#ifdef PURIFY
79
80#define new_HE() (HE*)safemalloc(sizeof(HE))
81#define del_HE(p) safefree((char*)p)
82
83#else
84
85#define new_HE() new_he()
86#define del_HE(p) del_he(p)
87
88#endif
89
76e3520e 90STATIC HEK *
19692e8d 91S_save_hek_flags(pTHX_ const char *str, I32 len, U32 hash, int flags)
bbce6d69 92{
35a4481c 93 const int flags_masked = flags & HVhek_MASK;
bbce6d69 94 char *k;
95 register HEK *hek;
1c846c1f 96
e05949c7 97 New(54, k, HEK_BASESIZE + len + 2, char);
bbce6d69 98 hek = (HEK*)k;
ff68c719 99 Copy(str, HEK_KEY(hek), len, char);
e05949c7 100 HEK_KEY(hek)[len] = 0;
ff68c719 101 HEK_LEN(hek) = len;
102 HEK_HASH(hek) = hash;
dcf933a4
NC
103 HEK_FLAGS(hek) = (unsigned char)flags_masked;
104
105 if (flags & HVhek_FREEKEY)
106 Safefree(str);
bbce6d69 107 return hek;
108}
109
dd28f7bb
DM
110/* free the pool of temporary HE/HEK pairs retunrned by hv_fetch_ent
111 * for tied hashes */
112
113void
114Perl_free_tied_hv_pool(pTHX)
115{
116 HE *ohe;
117 HE *he = PL_hv_fetch_ent_mh;
118 while (he) {
119 Safefree(HeKEY_hek(he));
120 ohe = he;
121 he = HeNEXT(he);
122 del_HE(ohe);
123 }
bf9cdc68 124 PL_hv_fetch_ent_mh = Nullhe;
dd28f7bb
DM
125}
126
d18c6117
GS
127#if defined(USE_ITHREADS)
128HE *
a8fc9800 129Perl_he_dup(pTHX_ HE *e, bool shared, CLONE_PARAMS* param)
d18c6117
GS
130{
131 HE *ret;
132
133 if (!e)
134 return Nullhe;
7766f137
GS
135 /* look for it in the table first */
136 ret = (HE*)ptr_table_fetch(PL_ptr_table, e);
137 if (ret)
138 return ret;
139
140 /* create anew and remember what it is */
d33b2eba 141 ret = new_HE();
7766f137
GS
142 ptr_table_store(PL_ptr_table, e, ret);
143
d2d73c3e 144 HeNEXT(ret) = he_dup(HeNEXT(e),shared, param);
dd28f7bb
DM
145 if (HeKLEN(e) == HEf_SVKEY) {
146 char *k;
147 New(54, k, HEK_BASESIZE + sizeof(SV*), char);
148 HeKEY_hek(ret) = (HEK*)k;
d2d73c3e 149 HeKEY_sv(ret) = SvREFCNT_inc(sv_dup(HeKEY_sv(e), param));
dd28f7bb 150 }
d18c6117 151 else if (shared)
19692e8d
NC
152 HeKEY_hek(ret) = share_hek_flags(HeKEY(e), HeKLEN(e), HeHASH(e),
153 HeKFLAGS(e));
d18c6117 154 else
19692e8d
NC
155 HeKEY_hek(ret) = save_hek_flags(HeKEY(e), HeKLEN(e), HeHASH(e),
156 HeKFLAGS(e));
d2d73c3e 157 HeVAL(ret) = SvREFCNT_inc(sv_dup(HeVAL(e), param));
d18c6117
GS
158 return ret;
159}
160#endif /* USE_ITHREADS */
161
1b1f1335 162static void
2393f1b9
JH
163S_hv_notallowed(pTHX_ int flags, const char *key, I32 klen,
164 const char *msg)
1b1f1335 165{
2393f1b9 166 SV *sv = sv_newmortal(), *esv = sv_newmortal();
19692e8d 167 if (!(flags & HVhek_FREEKEY)) {
1b1f1335
NIS
168 sv_setpvn(sv, key, klen);
169 }
170 else {
171 /* Need to free saved eventually assign to mortal SV */
34c3c4e3 172 /* XXX is this line an error ???: SV *sv = sv_newmortal(); */
1b1f1335
NIS
173 sv_usepvn(sv, (char *) key, klen);
174 }
19692e8d 175 if (flags & HVhek_UTF8) {
1b1f1335
NIS
176 SvUTF8_on(sv);
177 }
2393f1b9
JH
178 Perl_sv_setpvf(aTHX_ esv, "Attempt to %s a restricted hash", msg);
179 Perl_croak(aTHX_ SvPVX(esv), sv);
1b1f1335
NIS
180}
181
fde52b5c 182/* (klen == HEf_SVKEY) is special for MAGICAL hv entries, meaning key slot
183 * contains an SV* */
184
34a6f7b4
NC
185#define HV_FETCH_ISSTORE 0x01
186#define HV_FETCH_ISEXISTS 0x02
187#define HV_FETCH_LVALUE 0x04
188#define HV_FETCH_JUST_SV 0x08
189
190/*
191=for apidoc hv_store
192
193Stores an SV in a hash. The hash key is specified as C<key> and C<klen> is
194the length of the key. The C<hash> parameter is the precomputed hash
195value; if it is zero then Perl will compute it. The return value will be
196NULL if the operation failed or if the value did not need to be actually
197stored within the hash (as in the case of tied hashes). Otherwise it can
198be dereferenced to get the original C<SV*>. Note that the caller is
199responsible for suitably incrementing the reference count of C<val> before
200the call, and decrementing it if the function returned NULL. Effectively
201a successful hv_store takes ownership of one reference to C<val>. This is
202usually what you want; a newly created SV has a reference count of one, so
203if all your code does is create SVs then store them in a hash, hv_store
204will own the only reference to the new SV, and your code doesn't need to do
205anything further to tidy up. hv_store is not implemented as a call to
206hv_store_ent, and does not create a temporary SV for the key, so if your
207key data is not already in SV form then use hv_store in preference to
208hv_store_ent.
209
210See L<perlguts/"Understanding the Magic of Tied Hashes and Arrays"> for more
211information on how to use this function on tied hashes.
212
213=cut
214*/
215
216SV**
217Perl_hv_store(pTHX_ HV *hv, const char *key, I32 klen_i32, SV *val, U32 hash)
218{
219 HE *hek;
220 STRLEN klen;
221 int flags;
222
223 if (klen_i32 < 0) {
224 klen = -klen_i32;
225 flags = HVhek_UTF8;
226 } else {
227 klen = klen_i32;
228 flags = 0;
229 }
230 hek = hv_fetch_common (hv, NULL, key, klen, flags,
52d01cc2 231 (HV_FETCH_ISSTORE|HV_FETCH_JUST_SV), val, hash);
34a6f7b4
NC
232 return hek ? &HeVAL(hek) : NULL;
233}
234
235SV**
236Perl_hv_store_flags(pTHX_ HV *hv, const char *key, I32 klen, SV *val,
237 register U32 hash, int flags)
238{
239 HE *hek = hv_fetch_common (hv, NULL, key, klen, flags,
240 (HV_FETCH_ISSTORE|HV_FETCH_JUST_SV), val, hash);
241 return hek ? &HeVAL(hek) : NULL;
242}
243
244/*
245=for apidoc hv_store_ent
246
247Stores C<val> in a hash. The hash key is specified as C<key>. The C<hash>
248parameter is the precomputed hash value; if it is zero then Perl will
249compute it. The return value is the new hash entry so created. It will be
250NULL if the operation failed or if the value did not need to be actually
251stored within the hash (as in the case of tied hashes). Otherwise the
252contents of the return value can be accessed using the C<He?> macros
253described here. Note that the caller is responsible for suitably
254incrementing the reference count of C<val> before the call, and
255decrementing it if the function returned NULL. Effectively a successful
256hv_store_ent takes ownership of one reference to C<val>. This is
257usually what you want; a newly created SV has a reference count of one, so
258if all your code does is create SVs then store them in a hash, hv_store
259will own the only reference to the new SV, and your code doesn't need to do
260anything further to tidy up. Note that hv_store_ent only reads the C<key>;
261unlike C<val> it does not take ownership of it, so maintaining the correct
262reference count on C<key> is entirely the caller's responsibility. hv_store
263is not implemented as a call to hv_store_ent, and does not create a temporary
264SV for the key, so if your key data is not already in SV form then use
265hv_store in preference to hv_store_ent.
266
267See L<perlguts/"Understanding the Magic of Tied Hashes and Arrays"> for more
268information on how to use this function on tied hashes.
269
270=cut
271*/
272
273HE *
274Perl_hv_store_ent(pTHX_ HV *hv, SV *keysv, SV *val, U32 hash)
275{
276 return hv_fetch_common(hv, keysv, NULL, 0, 0, HV_FETCH_ISSTORE, val, hash);
277}
278
279/*
280=for apidoc hv_exists
281
282Returns a boolean indicating whether the specified hash key exists. The
283C<klen> is the length of the key.
284
285=cut
286*/
287
288bool
289Perl_hv_exists(pTHX_ HV *hv, const char *key, I32 klen_i32)
290{
291 STRLEN klen;
292 int flags;
293
294 if (klen_i32 < 0) {
295 klen = -klen_i32;
296 flags = HVhek_UTF8;
297 } else {
298 klen = klen_i32;
299 flags = 0;
300 }
301 return hv_fetch_common(hv, NULL, key, klen, flags, HV_FETCH_ISEXISTS, 0, 0)
302 ? TRUE : FALSE;
303}
304
954c1994
GS
305/*
306=for apidoc hv_fetch
307
308Returns the SV which corresponds to the specified key in the hash. The
309C<klen> is the length of the key. If C<lval> is set then the fetch will be
310part of a store. Check that the return value is non-null before
d1be9408 311dereferencing it to an C<SV*>.
954c1994 312
96f1132b 313See L<perlguts/"Understanding the Magic of Tied Hashes and Arrays"> for more
954c1994
GS
314information on how to use this function on tied hashes.
315
316=cut
317*/
318
79072805 319SV**
c1fe5510 320Perl_hv_fetch(pTHX_ HV *hv, const char *key, I32 klen_i32, I32 lval)
79072805 321{
c1fe5510
NC
322 HE *hek;
323 STRLEN klen;
324 int flags;
325
326 if (klen_i32 < 0) {
327 klen = -klen_i32;
328 flags = HVhek_UTF8;
329 } else {
330 klen = klen_i32;
331 flags = 0;
332 }
333 hek = hv_fetch_common (hv, NULL, key, klen, flags,
b2c64049
NC
334 HV_FETCH_JUST_SV | (lval ? HV_FETCH_LVALUE : 0),
335 Nullsv, 0);
113738bb 336 return hek ? &HeVAL(hek) : NULL;
79072805
LW
337}
338
34a6f7b4
NC
339/*
340=for apidoc hv_exists_ent
341
342Returns a boolean indicating whether the specified hash key exists. C<hash>
343can be a valid precomputed hash value, or 0 to ask for it to be
344computed.
345
346=cut
347*/
348
349bool
350Perl_hv_exists_ent(pTHX_ HV *hv, SV *keysv, U32 hash)
351{
352 return hv_fetch_common(hv, keysv, NULL, 0, 0, HV_FETCH_ISEXISTS, 0, hash)
353 ? TRUE : FALSE;
354}
355
d1be9408 356/* returns an HE * structure with the all fields set */
fde52b5c 357/* note that hent_val will be a mortal sv for MAGICAL hashes */
954c1994
GS
358/*
359=for apidoc hv_fetch_ent
360
361Returns the hash entry which corresponds to the specified key in the hash.
362C<hash> must be a valid precomputed hash number for the given C<key>, or 0
363if you want the function to compute it. IF C<lval> is set then the fetch
364will be part of a store. Make sure the return value is non-null before
365accessing it. The return value when C<tb> is a tied hash is a pointer to a
366static location, so be sure to make a copy of the structure if you need to
1c846c1f 367store it somewhere.
954c1994 368
96f1132b 369See L<perlguts/"Understanding the Magic of Tied Hashes and Arrays"> for more
954c1994
GS
370information on how to use this function on tied hashes.
371
372=cut
373*/
374
fde52b5c 375HE *
864dbfa3 376Perl_hv_fetch_ent(pTHX_ HV *hv, SV *keysv, I32 lval, register U32 hash)
fde52b5c 377{
7f66fda2 378 return hv_fetch_common(hv, keysv, NULL, 0, 0,
b2c64049 379 (lval ? HV_FETCH_LVALUE : 0), Nullsv, hash);
113738bb
NC
380}
381
8f8d40ab 382STATIC HE *
c1fe5510 383S_hv_fetch_common(pTHX_ HV *hv, SV *keysv, const char *key, STRLEN klen,
b2c64049 384 int flags, int action, SV *val, register U32 hash)
113738bb 385{
27da23d5 386 dVAR;
b2c64049
NC
387 XPVHV* xhv;
388 U32 n_links;
389 HE *entry;
390 HE **oentry;
fde52b5c 391 SV *sv;
da58a35d 392 bool is_utf8;
113738bb 393 int masked_flags;
fde52b5c 394
395 if (!hv)
396 return 0;
397
113738bb 398 if (keysv) {
e593d2fe
AE
399 if (flags & HVhek_FREEKEY)
400 Safefree(key);
113738bb 401 key = SvPV(keysv, klen);
c1fe5510 402 flags = 0;
113738bb
NC
403 is_utf8 = (SvUTF8(keysv) != 0);
404 } else {
c1fe5510 405 is_utf8 = ((flags & HVhek_UTF8) ? TRUE : FALSE);
113738bb 406 }
113738bb 407
b2c64049 408 xhv = (XPVHV*)SvANY(hv);
7f66fda2
NC
409 if (SvMAGICAL(hv)) {
410 if (SvRMAGICAL(hv) && !(action & (HV_FETCH_ISSTORE|HV_FETCH_ISEXISTS)))
411 {
412 if (mg_find((SV*)hv, PERL_MAGIC_tied) || SvGMAGICAL((SV*)hv)) {
413 sv = sv_newmortal();
113738bb 414
7f66fda2
NC
415 /* XXX should be able to skimp on the HE/HEK here when
416 HV_FETCH_JUST_SV is true. */
113738bb 417
7f66fda2
NC
418 if (!keysv) {
419 keysv = newSVpvn(key, klen);
420 if (is_utf8) {
421 SvUTF8_on(keysv);
422 }
423 } else {
424 keysv = newSVsv(keysv);
113738bb 425 }
7f66fda2
NC
426 mg_copy((SV*)hv, sv, (char *)keysv, HEf_SVKEY);
427
428 /* grab a fake HE/HEK pair from the pool or make a new one */
429 entry = PL_hv_fetch_ent_mh;
430 if (entry)
431 PL_hv_fetch_ent_mh = HeNEXT(entry);
432 else {
433 char *k;
434 entry = new_HE();
435 New(54, k, HEK_BASESIZE + sizeof(SV*), char);
436 HeKEY_hek(entry) = (HEK*)k;
437 }
438 HeNEXT(entry) = Nullhe;
439 HeSVKEY_set(entry, keysv);
440 HeVAL(entry) = sv;
441 sv_upgrade(sv, SVt_PVLV);
442 LvTYPE(sv) = 'T';
443 /* so we can free entry when freeing sv */
444 LvTARG(sv) = (SV*)entry;
445
446 /* XXX remove at some point? */
447 if (flags & HVhek_FREEKEY)
448 Safefree(key);
449
450 return entry;
113738bb 451 }
7f66fda2
NC
452#ifdef ENV_IS_CASELESS
453 else if (mg_find((SV*)hv, PERL_MAGIC_env)) {
454 U32 i;
455 for (i = 0; i < klen; ++i)
456 if (isLOWER(key[i])) {
086cb327
NC
457 /* Would be nice if we had a routine to do the
458 copy and upercase in a single pass through. */
e1ec3a88 459 const char *nkey = strupr(savepvn(key,klen));
086cb327
NC
460 /* Note that this fetch is for nkey (the uppercased
461 key) whereas the store is for key (the original) */
462 entry = hv_fetch_common(hv, Nullsv, nkey, klen,
463 HVhek_FREEKEY, /* free nkey */
464 0 /* non-LVAL fetch */,
465 Nullsv /* no value */,
466 0 /* compute hash */);
467 if (!entry && (action & HV_FETCH_LVALUE)) {
468 /* This call will free key if necessary.
469 Do it this way to encourage compiler to tail
470 call optimise. */
471 entry = hv_fetch_common(hv, keysv, key, klen,
472 flags, HV_FETCH_ISSTORE,
473 NEWSV(61,0), hash);
474 } else {
475 if (flags & HVhek_FREEKEY)
476 Safefree(key);
477 }
478 return entry;
7f66fda2 479 }
902173a3 480 }
7f66fda2
NC
481#endif
482 } /* ISFETCH */
483 else if (SvRMAGICAL(hv) && (action & HV_FETCH_ISEXISTS)) {
484 if (mg_find((SV*)hv, PERL_MAGIC_tied) || SvGMAGICAL((SV*)hv)) {
485 SV* svret;
b2c64049
NC
486 /* I don't understand why hv_exists_ent has svret and sv,
487 whereas hv_exists only had one. */
488 svret = sv_newmortal();
489 sv = sv_newmortal();
7f66fda2
NC
490
491 if (keysv || is_utf8) {
492 if (!keysv) {
493 keysv = newSVpvn(key, klen);
494 SvUTF8_on(keysv);
495 } else {
496 keysv = newSVsv(keysv);
497 }
b2c64049
NC
498 mg_copy((SV*)hv, sv, (char *)sv_2mortal(keysv), HEf_SVKEY);
499 } else {
500 mg_copy((SV*)hv, sv, key, klen);
7f66fda2 501 }
b2c64049
NC
502 if (flags & HVhek_FREEKEY)
503 Safefree(key);
7f66fda2
NC
504 magic_existspack(svret, mg_find(sv, PERL_MAGIC_tiedelem));
505 /* This cast somewhat evil, but I'm merely using NULL/
506 not NULL to return the boolean exists.
507 And I know hv is not NULL. */
508 return SvTRUE(svret) ? (HE *)hv : NULL;
e7152ba2 509 }
7f66fda2
NC
510#ifdef ENV_IS_CASELESS
511 else if (mg_find((SV*)hv, PERL_MAGIC_env)) {
512 /* XXX This code isn't UTF8 clean. */
b2c64049
NC
513 const char *keysave = key;
514 /* Will need to free this, so set FREEKEY flag. */
515 key = savepvn(key,klen);
516 key = (const char*)strupr((char*)key);
7f66fda2
NC
517 is_utf8 = 0;
518 hash = 0;
8b4f7dd5 519 keysv = 0;
b2c64049
NC
520
521 if (flags & HVhek_FREEKEY) {
522 Safefree(keysave);
523 }
524 flags |= HVhek_FREEKEY;
7f66fda2 525 }
902173a3 526#endif
7f66fda2 527 } /* ISEXISTS */
b2c64049
NC
528 else if (action & HV_FETCH_ISSTORE) {
529 bool needs_copy;
530 bool needs_store;
531 hv_magic_check (hv, &needs_copy, &needs_store);
532 if (needs_copy) {
533 bool save_taint = PL_tainted;
534 if (keysv || is_utf8) {
535 if (!keysv) {
536 keysv = newSVpvn(key, klen);
537 SvUTF8_on(keysv);
538 }
539 if (PL_tainting)
540 PL_tainted = SvTAINTED(keysv);
541 keysv = sv_2mortal(newSVsv(keysv));
542 mg_copy((SV*)hv, val, (char*)keysv, HEf_SVKEY);
543 } else {
544 mg_copy((SV*)hv, val, key, klen);
545 }
546
547 TAINT_IF(save_taint);
548 if (!xhv->xhv_array /* !HvARRAY(hv) */ && !needs_store) {
549 if (flags & HVhek_FREEKEY)
550 Safefree(key);
551 return Nullhe;
552 }
553#ifdef ENV_IS_CASELESS
554 else if (mg_find((SV*)hv, PERL_MAGIC_env)) {
555 /* XXX This code isn't UTF8 clean. */
556 const char *keysave = key;
557 /* Will need to free this, so set FREEKEY flag. */
558 key = savepvn(key,klen);
559 key = (const char*)strupr((char*)key);
560 is_utf8 = 0;
561 hash = 0;
8b4f7dd5 562 keysv = 0;
b2c64049
NC
563
564 if (flags & HVhek_FREEKEY) {
565 Safefree(keysave);
566 }
567 flags |= HVhek_FREEKEY;
568 }
569#endif
570 }
571 } /* ISSTORE */
7f66fda2 572 } /* SvMAGICAL */
fde52b5c 573
cbec9347 574 if (!xhv->xhv_array /* !HvARRAY(hv) */) {
b2c64049 575 if ((action & (HV_FETCH_LVALUE | HV_FETCH_ISSTORE))
fde52b5c 576#ifdef DYNAMIC_ENV_FETCH /* if it's an %ENV lookup, we may get it on the fly */
8aacddc1 577 || (SvRMAGICAL((SV*)hv) && mg_find((SV*)hv, PERL_MAGIC_env))
fde52b5c 578#endif
8aacddc1 579 )
cbec9347
JH
580 Newz(503, xhv->xhv_array /* HvARRAY(hv) */,
581 PERL_HV_ARRAY_ALLOC_BYTES(xhv->xhv_max+1 /* HvMAX(hv)+1 */),
582 char);
7f66fda2
NC
583#ifdef DYNAMIC_ENV_FETCH
584 else if (action & HV_FETCH_ISEXISTS) {
585 /* for an %ENV exists, if we do an insert it's by a recursive
586 store call, so avoid creating HvARRAY(hv) right now. */
587 }
588#endif
113738bb
NC
589 else {
590 /* XXX remove at some point? */
591 if (flags & HVhek_FREEKEY)
592 Safefree(key);
593
fde52b5c 594 return 0;
113738bb 595 }
fde52b5c 596 }
597
19692e8d 598 if (is_utf8) {
7f66fda2 599 const char *keysave = key;
f9a63242 600 key = (char*)bytes_from_utf8((U8*)key, &klen, &is_utf8);
19692e8d 601 if (is_utf8)
c1fe5510
NC
602 flags |= HVhek_UTF8;
603 else
604 flags &= ~HVhek_UTF8;
7f66fda2
NC
605 if (key != keysave) {
606 if (flags & HVhek_FREEKEY)
607 Safefree(keysave);
19692e8d 608 flags |= HVhek_WASUTF8 | HVhek_FREEKEY;
7f66fda2 609 }
19692e8d 610 }
f9a63242 611
4b5190b5
NC
612 if (HvREHASH(hv)) {
613 PERL_HASH_INTERNAL(hash, key, klen);
b2c64049
NC
614 /* We don't have a pointer to the hv, so we have to replicate the
615 flag into every HEK, so that hv_iterkeysv can see it. */
616 /* And yes, you do need this even though you are not "storing" because
fdcd69b6
NC
617 you can flip the flags below if doing an lval lookup. (And that
618 was put in to give the semantics Andreas was expecting.) */
619 flags |= HVhek_REHASH;
4b5190b5 620 } else if (!hash) {
113738bb 621 if (keysv && (SvIsCOW_shared_hash(keysv))) {
46187eeb
NC
622 hash = SvUVX(keysv);
623 } else {
624 PERL_HASH(hash, key, klen);
625 }
626 }
effa1e2d 627
113738bb 628 masked_flags = (flags & HVhek_MASK);
b2c64049 629 n_links = 0;
113738bb 630
7f66fda2
NC
631#ifdef DYNAMIC_ENV_FETCH
632 if (!xhv->xhv_array /* !HvARRAY(hv) */) entry = Null(HE*);
633 else
634#endif
b2c64049 635 {
ab4af705
NC
636 /* entry = (HvARRAY(hv))[hash & (I32) HvMAX(hv)]; */
637 entry = ((HE**)xhv->xhv_array)[hash & (I32) xhv->xhv_max];
b2c64049
NC
638 }
639 for (; entry; ++n_links, entry = HeNEXT(entry)) {
fde52b5c 640 if (HeHASH(entry) != hash) /* strings can't be equal */
641 continue;
eb160463 642 if (HeKLEN(entry) != (I32)klen)
fde52b5c 643 continue;
1c846c1f 644 if (HeKEY(entry) != key && memNE(HeKEY(entry),key,klen)) /* is this it? */
fde52b5c 645 continue;
113738bb 646 if ((HeKFLAGS(entry) ^ masked_flags) & HVhek_UTF8)
c3654f1a 647 continue;
b2c64049
NC
648
649 if (action & (HV_FETCH_LVALUE|HV_FETCH_ISSTORE)) {
650 if (HeKFLAGS(entry) != masked_flags) {
651 /* We match if HVhek_UTF8 bit in our flags and hash key's
652 match. But if entry was set previously with HVhek_WASUTF8
653 and key now doesn't (or vice versa) then we should change
654 the key's flag, as this is assignment. */
655 if (HvSHAREKEYS(hv)) {
656 /* Need to swap the key we have for a key with the flags we
657 need. As keys are shared we can't just write to the
658 flag, so we share the new one, unshare the old one. */
659 HEK *new_hek = share_hek_flags(key, klen, hash,
660 masked_flags);
661 unshare_hek (HeKEY_hek(entry));
662 HeKEY_hek(entry) = new_hek;
663 }
664 else
665 HeKFLAGS(entry) = masked_flags;
666 if (masked_flags & HVhek_ENABLEHVKFLAGS)
667 HvHASKFLAGS_on(hv);
668 }
669 if (HeVAL(entry) == &PL_sv_placeholder) {
670 /* yes, can store into placeholder slot */
671 if (action & HV_FETCH_LVALUE) {
672 if (SvMAGICAL(hv)) {
673 /* This preserves behaviour with the old hv_fetch
674 implementation which at this point would bail out
675 with a break; (at "if we find a placeholder, we
676 pretend we haven't found anything")
677
678 That break mean that if a placeholder were found, it
679 caused a call into hv_store, which in turn would
680 check magic, and if there is no magic end up pretty
681 much back at this point (in hv_store's code). */
682 break;
683 }
684 /* LVAL fetch which actaully needs a store. */
685 val = NEWSV(61,0);
686 xhv->xhv_placeholders--;
687 } else {
688 /* store */
689 if (val != &PL_sv_placeholder)
690 xhv->xhv_placeholders--;
691 }
692 HeVAL(entry) = val;
693 } else if (action & HV_FETCH_ISSTORE) {
694 SvREFCNT_dec(HeVAL(entry));
695 HeVAL(entry) = val;
696 }
27bcc0a7 697 } else if (HeVAL(entry) == &PL_sv_placeholder) {
b2c64049
NC
698 /* if we find a placeholder, we pretend we haven't found
699 anything */
8aacddc1 700 break;
b2c64049 701 }
113738bb
NC
702 if (flags & HVhek_FREEKEY)
703 Safefree(key);
fde52b5c 704 return entry;
705 }
706#ifdef DYNAMIC_ENV_FETCH /* %ENV lookup? If so, try to fetch the value now */
0ed29950
NC
707 if (!(action & HV_FETCH_ISSTORE)
708 && SvRMAGICAL((SV*)hv) && mg_find((SV*)hv, PERL_MAGIC_env)) {
a6c40364
GS
709 unsigned long len;
710 char *env = PerlEnv_ENVgetenv_len(key,&len);
711 if (env) {
712 sv = newSVpvn(env,len);
713 SvTAINTED_on(sv);
7fd3d16e 714 return hv_fetch_common(hv,keysv,key,klen,flags,HV_FETCH_ISSTORE,sv,
b2c64049 715 hash);
a6c40364 716 }
fde52b5c 717 }
718#endif
7f66fda2
NC
719
720 if (!entry && SvREADONLY(hv) && !(action & HV_FETCH_ISEXISTS)) {
2393f1b9
JH
721 S_hv_notallowed(aTHX_ flags, key, klen,
722 "access disallowed key '%"SVf"' in"
723 );
1b1f1335 724 }
b2c64049
NC
725 if (!(action & (HV_FETCH_LVALUE|HV_FETCH_ISSTORE))) {
726 /* Not doing some form of store, so return failure. */
727 if (flags & HVhek_FREEKEY)
728 Safefree(key);
729 return 0;
730 }
113738bb 731 if (action & HV_FETCH_LVALUE) {
b2c64049
NC
732 val = NEWSV(61,0);
733 if (SvMAGICAL(hv)) {
734 /* At this point the old hv_fetch code would call to hv_store,
735 which in turn might do some tied magic. So we need to make that
736 magic check happen. */
737 /* gonna assign to this, so it better be there */
738 return hv_fetch_common(hv, keysv, key, klen, flags,
739 HV_FETCH_ISSTORE, val, hash);
740 /* XXX Surely that could leak if the fetch-was-store fails?
741 Just like the hv_fetch. */
113738bb
NC
742 }
743 }
744
b2c64049
NC
745 /* Welcome to hv_store... */
746
ab4af705 747 if (!xhv->xhv_array) {
b2c64049
NC
748 /* Not sure if we can get here. I think the only case of oentry being
749 NULL is for %ENV with dynamic env fetch. But that should disappear
750 with magic in the previous code. */
751 Newz(503, xhv->xhv_array /* HvARRAY(hv) */,
752 PERL_HV_ARRAY_ALLOC_BYTES(xhv->xhv_max+1 /* HvMAX(hv)+1 */),
753 char);
b2c64049
NC
754 }
755
ab4af705
NC
756 oentry = &((HE**)xhv->xhv_array)[hash & (I32) xhv->xhv_max];
757
b2c64049
NC
758 entry = new_HE();
759 /* share_hek_flags will do the free for us. This might be considered
760 bad API design. */
761 if (HvSHAREKEYS(hv))
762 HeKEY_hek(entry) = share_hek_flags(key, klen, hash, flags);
763 else /* gotta do the real thing */
764 HeKEY_hek(entry) = save_hek_flags(key, klen, hash, flags);
765 HeVAL(entry) = val;
766 HeNEXT(entry) = *oentry;
767 *oentry = entry;
768
769 if (val == &PL_sv_placeholder)
770 xhv->xhv_placeholders++;
771 if (masked_flags & HVhek_ENABLEHVKFLAGS)
772 HvHASKFLAGS_on(hv);
773
774 xhv->xhv_keys++; /* HvKEYS(hv)++ */
775 if (!n_links) { /* initial entry? */
776 xhv->xhv_fill++; /* HvFILL(hv)++ */
777 } else if ((xhv->xhv_keys > (IV)xhv->xhv_max)
778 || ((n_links > HV_MAX_LENGTH_BEFORE_SPLIT) && !HvREHASH(hv))) {
779 /* Use only the old HvKEYS(hv) > HvMAX(hv) condition to limit bucket
780 splits on a rehashed hash, as we're not going to split it again,
781 and if someone is lucky (evil) enough to get all the keys in one
782 list they could exhaust our memory as we repeatedly double the
783 number of buckets on every entry. Linear search feels a less worse
784 thing to do. */
785 hsplit(hv);
fde52b5c 786 }
b2c64049
NC
787
788 return entry;
fde52b5c 789}
790
864dbfa3 791STATIC void
cea2e8a9 792S_hv_magic_check(pTHX_ HV *hv, bool *needs_copy, bool *needs_store)
d0066dc7
OT
793{
794 MAGIC *mg = SvMAGIC(hv);
795 *needs_copy = FALSE;
796 *needs_store = TRUE;
797 while (mg) {
798 if (isUPPER(mg->mg_type)) {
799 *needs_copy = TRUE;
800 switch (mg->mg_type) {
14befaf4
DM
801 case PERL_MAGIC_tied:
802 case PERL_MAGIC_sig:
d0066dc7 803 *needs_store = FALSE;
d0066dc7
OT
804 }
805 }
806 mg = mg->mg_moremagic;
807 }
808}
809
954c1994 810/*
a3bcc51e
TP
811=for apidoc hv_scalar
812
813Evaluates the hash in scalar context and returns the result. Handles magic when the hash is tied.
814
815=cut
816*/
817
818SV *
819Perl_hv_scalar(pTHX_ HV *hv)
820{
821 MAGIC *mg;
822 SV *sv;
823
824 if ((SvRMAGICAL(hv) && (mg = mg_find((SV*)hv, PERL_MAGIC_tied)))) {
825 sv = magic_scalarpack(hv, mg);
826 return sv;
827 }
828
829 sv = sv_newmortal();
830 if (HvFILL((HV*)hv))
831 Perl_sv_setpvf(aTHX_ sv, "%ld/%ld",
832 (long)HvFILL(hv), (long)HvMAX(hv) + 1);
833 else
834 sv_setiv(sv, 0);
835
836 return sv;
837}
838
839/*
954c1994
GS
840=for apidoc hv_delete
841
842Deletes a key/value pair in the hash. The value SV is removed from the
1c846c1f 843hash and returned to the caller. The C<klen> is the length of the key.
954c1994
GS
844The C<flags> value will normally be zero; if set to G_DISCARD then NULL
845will be returned.
846
847=cut
848*/
849
79072805 850SV *
cd6d36ac 851Perl_hv_delete(pTHX_ HV *hv, const char *key, I32 klen_i32, I32 flags)
79072805 852{
cd6d36ac
NC
853 STRLEN klen;
854 int k_flags = 0;
855
856 if (klen_i32 < 0) {
857 klen = -klen_i32;
858 k_flags |= HVhek_UTF8;
859 } else {
860 klen = klen_i32;
861 }
862 return hv_delete_common(hv, NULL, key, klen, k_flags, flags, 0);
fde52b5c 863}
864
954c1994
GS
865/*
866=for apidoc hv_delete_ent
867
868Deletes a key/value pair in the hash. The value SV is removed from the
869hash and returned to the caller. The C<flags> value will normally be zero;
870if set to G_DISCARD then NULL will be returned. C<hash> can be a valid
871precomputed hash value, or 0 to ask for it to be computed.
872
873=cut
874*/
875
fde52b5c 876SV *
864dbfa3 877Perl_hv_delete_ent(pTHX_ HV *hv, SV *keysv, I32 flags, U32 hash)
fde52b5c 878{
cd6d36ac 879 return hv_delete_common(hv, keysv, NULL, 0, 0, flags, hash);
f1317c8d
NC
880}
881
8f8d40ab 882STATIC SV *
cd6d36ac
NC
883S_hv_delete_common(pTHX_ HV *hv, SV *keysv, const char *key, STRLEN klen,
884 int k_flags, I32 d_flags, U32 hash)
f1317c8d 885{
27da23d5 886 dVAR;
cbec9347 887 register XPVHV* xhv;
fde52b5c 888 register I32 i;
fde52b5c 889 register HE *entry;
890 register HE **oentry;
891 SV *sv;
da58a35d 892 bool is_utf8;
7a9669ca 893 int masked_flags;
1c846c1f 894
fde52b5c 895 if (!hv)
896 return Nullsv;
f1317c8d
NC
897
898 if (keysv) {
e593d2fe
AE
899 if (k_flags & HVhek_FREEKEY)
900 Safefree(key);
f1317c8d 901 key = SvPV(keysv, klen);
cd6d36ac 902 k_flags = 0;
f1317c8d
NC
903 is_utf8 = (SvUTF8(keysv) != 0);
904 } else {
cd6d36ac 905 is_utf8 = ((k_flags & HVhek_UTF8) ? TRUE : FALSE);
f1317c8d 906 }
f1317c8d 907
fde52b5c 908 if (SvRMAGICAL(hv)) {
0a0bb7c7
OT
909 bool needs_copy;
910 bool needs_store;
911 hv_magic_check (hv, &needs_copy, &needs_store);
912
f1317c8d 913 if (needs_copy) {
7a9669ca
NC
914 entry = hv_fetch_common(hv, keysv, key, klen,
915 k_flags & ~HVhek_FREEKEY, HV_FETCH_LVALUE,
b2c64049 916 Nullsv, hash);
7a9669ca 917 sv = entry ? HeVAL(entry) : NULL;
f1317c8d
NC
918 if (sv) {
919 if (SvMAGICAL(sv)) {
920 mg_clear(sv);
921 }
922 if (!needs_store) {
923 if (mg_find(sv, PERL_MAGIC_tiedelem)) {
924 /* No longer an element */
925 sv_unmagic(sv, PERL_MAGIC_tiedelem);
926 return sv;
927 }
928 return Nullsv; /* element cannot be deleted */
929 }
902173a3 930#ifdef ENV_IS_CASELESS
8167a60a
NC
931 else if (mg_find((SV*)hv, PERL_MAGIC_env)) {
932 /* XXX This code isn't UTF8 clean. */
933 keysv = sv_2mortal(newSVpvn(key,klen));
934 if (k_flags & HVhek_FREEKEY) {
935 Safefree(key);
936 }
937 key = strupr(SvPVX(keysv));
938 is_utf8 = 0;
939 k_flags = 0;
940 hash = 0;
7f66fda2 941 }
510ac311 942#endif
2fd1c6b8 943 }
2fd1c6b8 944 }
fde52b5c 945 }
cbec9347
JH
946 xhv = (XPVHV*)SvANY(hv);
947 if (!xhv->xhv_array /* !HvARRAY(hv) */)
fde52b5c 948 return Nullsv;
949
19692e8d 950 if (is_utf8) {
7f66fda2
NC
951 const char *keysave = key;
952 key = (char*)bytes_from_utf8((U8*)key, &klen, &is_utf8);
cd6d36ac 953
19692e8d 954 if (is_utf8)
cd6d36ac
NC
955 k_flags |= HVhek_UTF8;
956 else
957 k_flags &= ~HVhek_UTF8;
7f66fda2
NC
958 if (key != keysave) {
959 if (k_flags & HVhek_FREEKEY) {
960 /* This shouldn't happen if our caller does what we expect,
961 but strictly the API allows it. */
962 Safefree(keysave);
963 }
964 k_flags |= HVhek_WASUTF8 | HVhek_FREEKEY;
965 }
cd6d36ac 966 HvHASKFLAGS_on((SV*)hv);
19692e8d 967 }
f9a63242 968
4b5190b5
NC
969 if (HvREHASH(hv)) {
970 PERL_HASH_INTERNAL(hash, key, klen);
971 } else if (!hash) {
7a9669ca
NC
972 if (keysv && (SvIsCOW_shared_hash(keysv))) {
973 hash = SvUVX(keysv);
974 } else {
975 PERL_HASH(hash, key, klen);
976 }
4b5190b5 977 }
fde52b5c 978
7a9669ca
NC
979 masked_flags = (k_flags & HVhek_MASK);
980
cbec9347
JH
981 /* oentry = &(HvARRAY(hv))[hash & (I32) HvMAX(hv)]; */
982 oentry = &((HE**)xhv->xhv_array)[hash & (I32) xhv->xhv_max];
fde52b5c 983 entry = *oentry;
984 i = 1;
985 for (; entry; i=0, oentry = &HeNEXT(entry), entry = *oentry) {
986 if (HeHASH(entry) != hash) /* strings can't be equal */
987 continue;
eb160463 988 if (HeKLEN(entry) != (I32)klen)
fde52b5c 989 continue;
1c846c1f 990 if (HeKEY(entry) != key && memNE(HeKEY(entry),key,klen)) /* is this it? */
fde52b5c 991 continue;
7a9669ca 992 if ((HeKFLAGS(entry) ^ masked_flags) & HVhek_UTF8)
c3654f1a 993 continue;
8aacddc1
NIS
994
995 /* if placeholder is here, it's already been deleted.... */
7996736c 996 if (HeVAL(entry) == &PL_sv_placeholder)
8aacddc1 997 {
b84d0860
NC
998 if (k_flags & HVhek_FREEKEY)
999 Safefree(key);
1000 return Nullsv;
8aacddc1
NIS
1001 }
1002 else if (SvREADONLY(hv) && HeVAL(entry) && SvREADONLY(HeVAL(entry))) {
2393f1b9
JH
1003 S_hv_notallowed(aTHX_ k_flags, key, klen,
1004 "delete readonly key '%"SVf"' from"
1005 );
8aacddc1 1006 }
b84d0860
NC
1007 if (k_flags & HVhek_FREEKEY)
1008 Safefree(key);
8aacddc1 1009
cd6d36ac 1010 if (d_flags & G_DISCARD)
fde52b5c 1011 sv = Nullsv;
94f7643d 1012 else {
79d01fbf 1013 sv = sv_2mortal(HeVAL(entry));
7996736c 1014 HeVAL(entry) = &PL_sv_placeholder;
94f7643d 1015 }
8aacddc1
NIS
1016
1017 /*
1018 * If a restricted hash, rather than really deleting the entry, put
1019 * a placeholder there. This marks the key as being "approved", so
1020 * we can still access via not-really-existing key without raising
1021 * an error.
1022 */
1023 if (SvREADONLY(hv)) {
754604c4 1024 SvREFCNT_dec(HeVAL(entry));
7996736c 1025 HeVAL(entry) = &PL_sv_placeholder;
8aacddc1
NIS
1026 /* We'll be saving this slot, so the number of allocated keys
1027 * doesn't go down, but the number placeholders goes up */
1028 xhv->xhv_placeholders++; /* HvPLACEHOLDERS(hv)++ */
1029 } else {
a26e96df
NIS
1030 *oentry = HeNEXT(entry);
1031 if (i && !*oentry)
1032 xhv->xhv_fill--; /* HvFILL(hv)-- */
8aacddc1
NIS
1033 if (entry == xhv->xhv_eiter /* HvEITER(hv) */)
1034 HvLAZYDEL_on(hv);
1035 else
1036 hv_free_ent(hv, entry);
1037 xhv->xhv_keys--; /* HvKEYS(hv)-- */
574c8022 1038 if (xhv->xhv_keys == 0)
19692e8d 1039 HvHASKFLAGS_off(hv);
8aacddc1 1040 }
79072805
LW
1041 return sv;
1042 }
8aacddc1 1043 if (SvREADONLY(hv)) {
2393f1b9
JH
1044 S_hv_notallowed(aTHX_ k_flags, key, klen,
1045 "delete disallowed key '%"SVf"' from"
1046 );
8aacddc1
NIS
1047 }
1048
19692e8d 1049 if (k_flags & HVhek_FREEKEY)
f9a63242 1050 Safefree(key);
79072805 1051 return Nullsv;
79072805
LW
1052}
1053
76e3520e 1054STATIC void
cea2e8a9 1055S_hsplit(pTHX_ HV *hv)
79072805 1056{
cbec9347
JH
1057 register XPVHV* xhv = (XPVHV*)SvANY(hv);
1058 I32 oldsize = (I32) xhv->xhv_max+1; /* HvMAX(hv)+1 (sick) */
79072805
LW
1059 register I32 newsize = oldsize * 2;
1060 register I32 i;
cbec9347 1061 register char *a = xhv->xhv_array; /* HvARRAY(hv) */
72311751
GS
1062 register HE **aep;
1063 register HE **bep;
79072805
LW
1064 register HE *entry;
1065 register HE **oentry;
4b5190b5
NC
1066 int longest_chain = 0;
1067 int was_shared;
79072805 1068
18026298
NC
1069 /*PerlIO_printf(PerlIO_stderr(), "hsplit called for %p which had %d\n",
1070 hv, (int) oldsize);*/
1071
1072 if (HvPLACEHOLDERS(hv) && !SvREADONLY(hv)) {
1073 /* Can make this clear any placeholders first for non-restricted hashes,
1074 even though Storable rebuilds restricted hashes by putting in all the
1075 placeholders (first) before turning on the readonly flag, because
1076 Storable always pre-splits the hash. */
1077 hv_clear_placeholders(hv);
1078 }
1079
3280af22 1080 PL_nomemok = TRUE;
8d6dde3e 1081#if defined(STRANGE_MALLOC) || defined(MYMALLOC)
d18c6117 1082 Renew(a, PERL_HV_ARRAY_ALLOC_BYTES(newsize), char);
422a93e5 1083 if (!a) {
4a33f861 1084 PL_nomemok = FALSE;
422a93e5
GA
1085 return;
1086 }
4633a7c4 1087#else
d18c6117 1088 New(2, a, PERL_HV_ARRAY_ALLOC_BYTES(newsize), char);
422a93e5 1089 if (!a) {
3280af22 1090 PL_nomemok = FALSE;
422a93e5
GA
1091 return;
1092 }
cbec9347 1093 Copy(xhv->xhv_array /* HvARRAY(hv) */, a, oldsize * sizeof(HE*), char);
fba3b22e 1094 if (oldsize >= 64) {
cbec9347
JH
1095 offer_nice_chunk(xhv->xhv_array /* HvARRAY(hv) */,
1096 PERL_HV_ARRAY_ALLOC_BYTES(oldsize));
4633a7c4
LW
1097 }
1098 else
cbec9347 1099 Safefree(xhv->xhv_array /* HvARRAY(hv) */);
4633a7c4
LW
1100#endif
1101
3280af22 1102 PL_nomemok = FALSE;
72311751 1103 Zero(&a[oldsize * sizeof(HE*)], (newsize-oldsize) * sizeof(HE*), char); /* zero 2nd half*/
cbec9347
JH
1104 xhv->xhv_max = --newsize; /* HvMAX(hv) = --newsize */
1105 xhv->xhv_array = a; /* HvARRAY(hv) = a */
72311751 1106 aep = (HE**)a;
79072805 1107
72311751 1108 for (i=0; i<oldsize; i++,aep++) {
4b5190b5
NC
1109 int left_length = 0;
1110 int right_length = 0;
1111
72311751 1112 if (!*aep) /* non-existent */
79072805 1113 continue;
72311751
GS
1114 bep = aep+oldsize;
1115 for (oentry = aep, entry = *aep; entry; entry = *oentry) {
eb160463 1116 if ((HeHASH(entry) & newsize) != (U32)i) {
fde52b5c 1117 *oentry = HeNEXT(entry);
72311751
GS
1118 HeNEXT(entry) = *bep;
1119 if (!*bep)
cbec9347 1120 xhv->xhv_fill++; /* HvFILL(hv)++ */
72311751 1121 *bep = entry;
4b5190b5 1122 right_length++;
79072805
LW
1123 continue;
1124 }
4b5190b5 1125 else {
fde52b5c 1126 oentry = &HeNEXT(entry);
4b5190b5
NC
1127 left_length++;
1128 }
79072805 1129 }
72311751 1130 if (!*aep) /* everything moved */
cbec9347 1131 xhv->xhv_fill--; /* HvFILL(hv)-- */
4b5190b5
NC
1132 /* I think we don't actually need to keep track of the longest length,
1133 merely flag if anything is too long. But for the moment while
1134 developing this code I'll track it. */
1135 if (left_length > longest_chain)
1136 longest_chain = left_length;
1137 if (right_length > longest_chain)
1138 longest_chain = right_length;
1139 }
1140
1141
1142 /* Pick your policy for "hashing isn't working" here: */
fdcd69b6 1143 if (longest_chain <= HV_MAX_LENGTH_BEFORE_SPLIT /* split worked? */
4b5190b5
NC
1144 || HvREHASH(hv)) {
1145 return;
79072805 1146 }
4b5190b5
NC
1147
1148 if (hv == PL_strtab) {
1149 /* Urg. Someone is doing something nasty to the string table.
1150 Can't win. */
1151 return;
1152 }
1153
1154 /* Awooga. Awooga. Pathological data. */
fdcd69b6 1155 /*PerlIO_printf(PerlIO_stderr(), "%p %d of %d with %d/%d buckets\n", hv,
4b5190b5
NC
1156 longest_chain, HvTOTALKEYS(hv), HvFILL(hv), 1+HvMAX(hv));*/
1157
1158 ++newsize;
1159 Newz(2, a, PERL_HV_ARRAY_ALLOC_BYTES(newsize), char);
1160 was_shared = HvSHAREKEYS(hv);
1161
1162 xhv->xhv_fill = 0;
1163 HvSHAREKEYS_off(hv);
1164 HvREHASH_on(hv);
1165
1166 aep = (HE **) xhv->xhv_array;
1167
1168 for (i=0; i<newsize; i++,aep++) {
1169 entry = *aep;
1170 while (entry) {
1171 /* We're going to trash this HE's next pointer when we chain it
1172 into the new hash below, so store where we go next. */
1173 HE *next = HeNEXT(entry);
1174 UV hash;
1175
1176 /* Rehash it */
1177 PERL_HASH_INTERNAL(hash, HeKEY(entry), HeKLEN(entry));
1178
1179 if (was_shared) {
1180 /* Unshare it. */
1181 HEK *new_hek
1182 = save_hek_flags(HeKEY(entry), HeKLEN(entry),
1183 hash, HeKFLAGS(entry));
1184 unshare_hek (HeKEY_hek(entry));
1185 HeKEY_hek(entry) = new_hek;
1186 } else {
1187 /* Not shared, so simply write the new hash in. */
1188 HeHASH(entry) = hash;
1189 }
1190 /*PerlIO_printf(PerlIO_stderr(), "%d ", HeKFLAGS(entry));*/
1191 HEK_REHASH_on(HeKEY_hek(entry));
1192 /*PerlIO_printf(PerlIO_stderr(), "%d\n", HeKFLAGS(entry));*/
1193
1194 /* Copy oentry to the correct new chain. */
1195 bep = ((HE**)a) + (hash & (I32) xhv->xhv_max);
1196 if (!*bep)
1197 xhv->xhv_fill++; /* HvFILL(hv)++ */
1198 HeNEXT(entry) = *bep;
1199 *bep = entry;
1200
1201 entry = next;
1202 }
1203 }
1204 Safefree (xhv->xhv_array);
1205 xhv->xhv_array = a; /* HvARRAY(hv) = a */
79072805
LW
1206}
1207
72940dca 1208void
864dbfa3 1209Perl_hv_ksplit(pTHX_ HV *hv, IV newmax)
72940dca 1210{
cbec9347
JH
1211 register XPVHV* xhv = (XPVHV*)SvANY(hv);
1212 I32 oldsize = (I32) xhv->xhv_max+1; /* HvMAX(hv)+1 (sick) */
72940dca 1213 register I32 newsize;
1214 register I32 i;
1215 register I32 j;
72311751
GS
1216 register char *a;
1217 register HE **aep;
72940dca 1218 register HE *entry;
1219 register HE **oentry;
1220
1221 newsize = (I32) newmax; /* possible truncation here */
1222 if (newsize != newmax || newmax <= oldsize)
1223 return;
1224 while ((newsize & (1 + ~newsize)) != newsize) {
1225 newsize &= ~(newsize & (1 + ~newsize)); /* get proper power of 2 */
1226 }
1227 if (newsize < newmax)
1228 newsize *= 2;
1229 if (newsize < newmax)
1230 return; /* overflow detection */
1231
cbec9347 1232 a = xhv->xhv_array; /* HvARRAY(hv) */
72940dca 1233 if (a) {
3280af22 1234 PL_nomemok = TRUE;
8d6dde3e 1235#if defined(STRANGE_MALLOC) || defined(MYMALLOC)
d18c6117 1236 Renew(a, PERL_HV_ARRAY_ALLOC_BYTES(newsize), char);
8aacddc1 1237 if (!a) {
4a33f861 1238 PL_nomemok = FALSE;
422a93e5
GA
1239 return;
1240 }
72940dca 1241#else
d18c6117 1242 New(2, a, PERL_HV_ARRAY_ALLOC_BYTES(newsize), char);
8aacddc1 1243 if (!a) {
3280af22 1244 PL_nomemok = FALSE;
422a93e5
GA
1245 return;
1246 }
cbec9347 1247 Copy(xhv->xhv_array /* HvARRAY(hv) */, a, oldsize * sizeof(HE*), char);
fba3b22e 1248 if (oldsize >= 64) {
cbec9347
JH
1249 offer_nice_chunk(xhv->xhv_array /* HvARRAY(hv) */,
1250 PERL_HV_ARRAY_ALLOC_BYTES(oldsize));
72940dca 1251 }
1252 else
cbec9347 1253 Safefree(xhv->xhv_array /* HvARRAY(hv) */);
72940dca 1254#endif
3280af22 1255 PL_nomemok = FALSE;
72311751 1256 Zero(&a[oldsize * sizeof(HE*)], (newsize-oldsize) * sizeof(HE*), char); /* zero 2nd half*/
72940dca 1257 }
1258 else {
d18c6117 1259 Newz(0, a, PERL_HV_ARRAY_ALLOC_BYTES(newsize), char);
72940dca 1260 }
cbec9347
JH
1261 xhv->xhv_max = --newsize; /* HvMAX(hv) = --newsize */
1262 xhv->xhv_array = a; /* HvARRAY(hv) = a */
1263 if (!xhv->xhv_fill /* !HvFILL(hv) */) /* skip rest if no entries */
72940dca 1264 return;
1265
72311751
GS
1266 aep = (HE**)a;
1267 for (i=0; i<oldsize; i++,aep++) {
1268 if (!*aep) /* non-existent */
72940dca 1269 continue;
72311751 1270 for (oentry = aep, entry = *aep; entry; entry = *oentry) {
72940dca 1271 if ((j = (HeHASH(entry) & newsize)) != i) {
1272 j -= i;
1273 *oentry = HeNEXT(entry);
72311751 1274 if (!(HeNEXT(entry) = aep[j]))
cbec9347 1275 xhv->xhv_fill++; /* HvFILL(hv)++ */
72311751 1276 aep[j] = entry;
72940dca 1277 continue;
1278 }
1279 else
1280 oentry = &HeNEXT(entry);
1281 }
72311751 1282 if (!*aep) /* everything moved */
cbec9347 1283 xhv->xhv_fill--; /* HvFILL(hv)-- */
72940dca 1284 }
1285}
1286
954c1994
GS
1287/*
1288=for apidoc newHV
1289
1290Creates a new HV. The reference count is set to 1.
1291
1292=cut
1293*/
1294
79072805 1295HV *
864dbfa3 1296Perl_newHV(pTHX)
79072805
LW
1297{
1298 register HV *hv;
cbec9347 1299 register XPVHV* xhv;
79072805 1300
a0d0e21e
LW
1301 hv = (HV*)NEWSV(502,0);
1302 sv_upgrade((SV *)hv, SVt_PVHV);
cbec9347 1303 xhv = (XPVHV*)SvANY(hv);
79072805
LW
1304 SvPOK_off(hv);
1305 SvNOK_off(hv);
1c846c1f 1306#ifndef NODEFAULT_SHAREKEYS
fde52b5c 1307 HvSHAREKEYS_on(hv); /* key-sharing on by default */
1c846c1f 1308#endif
4b5190b5 1309
cbec9347
JH
1310 xhv->xhv_max = 7; /* HvMAX(hv) = 7 (start with 8 buckets) */
1311 xhv->xhv_fill = 0; /* HvFILL(hv) = 0 */
1312 xhv->xhv_pmroot = 0; /* HvPMROOT(hv) = 0 */
79072805
LW
1313 (void)hv_iterinit(hv); /* so each() will start off right */
1314 return hv;
1315}
1316
b3ac6de7 1317HV *
864dbfa3 1318Perl_newHVhv(pTHX_ HV *ohv)
b3ac6de7 1319{
b56ba0bf 1320 HV *hv = newHV();
4beac62f 1321 STRLEN hv_max, hv_fill;
4beac62f
AMS
1322
1323 if (!ohv || (hv_fill = HvFILL(ohv)) == 0)
1324 return hv;
4beac62f 1325 hv_max = HvMAX(ohv);
b3ac6de7 1326
b56ba0bf
AMS
1327 if (!SvMAGICAL((SV *)ohv)) {
1328 /* It's an ordinary hash, so copy it fast. AMS 20010804 */
eb160463
GS
1329 STRLEN i;
1330 bool shared = !!HvSHAREKEYS(ohv);
b56ba0bf 1331 HE **ents, **oents = (HE **)HvARRAY(ohv);
ff875642
JH
1332 char *a;
1333 New(0, a, PERL_HV_ARRAY_ALLOC_BYTES(hv_max+1), char);
1334 ents = (HE**)a;
b56ba0bf
AMS
1335
1336 /* In each bucket... */
1337 for (i = 0; i <= hv_max; i++) {
1338 HE *prev = NULL, *ent = NULL, *oent = oents[i];
1339
1340 if (!oent) {
1341 ents[i] = NULL;
1342 continue;
1343 }
1344
1345 /* Copy the linked list of entries. */
1346 for (oent = oents[i]; oent; oent = HeNEXT(oent)) {
1347 U32 hash = HeHASH(oent);
1348 char *key = HeKEY(oent);
19692e8d
NC
1349 STRLEN len = HeKLEN(oent);
1350 int flags = HeKFLAGS(oent);
b56ba0bf
AMS
1351
1352 ent = new_HE();
45dea987 1353 HeVAL(ent) = newSVsv(HeVAL(oent));
19692e8d
NC
1354 HeKEY_hek(ent)
1355 = shared ? share_hek_flags(key, len, hash, flags)
1356 : save_hek_flags(key, len, hash, flags);
b56ba0bf
AMS
1357 if (prev)
1358 HeNEXT(prev) = ent;
1359 else
1360 ents[i] = ent;
1361 prev = ent;
1362 HeNEXT(ent) = NULL;
1363 }
1364 }
1365
1366 HvMAX(hv) = hv_max;
1367 HvFILL(hv) = hv_fill;
8aacddc1 1368 HvTOTALKEYS(hv) = HvTOTALKEYS(ohv);
b56ba0bf 1369 HvARRAY(hv) = ents;
1c846c1f 1370 }
b56ba0bf
AMS
1371 else {
1372 /* Iterate over ohv, copying keys and values one at a time. */
b3ac6de7 1373 HE *entry;
b56ba0bf
AMS
1374 I32 riter = HvRITER(ohv);
1375 HE *eiter = HvEITER(ohv);
1376
1377 /* Can we use fewer buckets? (hv_max is always 2^n-1) */
1378 while (hv_max && hv_max + 1 >= hv_fill * 2)
1379 hv_max = hv_max / 2;
1380 HvMAX(hv) = hv_max;
1381
4a76a316 1382 hv_iterinit(ohv);
e16e2ff8 1383 while ((entry = hv_iternext_flags(ohv, 0))) {
19692e8d
NC
1384 hv_store_flags(hv, HeKEY(entry), HeKLEN(entry),
1385 newSVsv(HeVAL(entry)), HeHASH(entry),
1386 HeKFLAGS(entry));
b3ac6de7 1387 }
b56ba0bf
AMS
1388 HvRITER(ohv) = riter;
1389 HvEITER(ohv) = eiter;
b3ac6de7 1390 }
1c846c1f 1391
b3ac6de7
IZ
1392 return hv;
1393}
1394
79072805 1395void
864dbfa3 1396Perl_hv_free_ent(pTHX_ HV *hv, register HE *entry)
79072805 1397{
16bdeea2
GS
1398 SV *val;
1399
68dc0745 1400 if (!entry)
79072805 1401 return;
16bdeea2 1402 val = HeVAL(entry);
257c9e5b 1403 if (val && isGV(val) && GvCVu(val) && HvNAME(hv))
3280af22 1404 PL_sub_generation++; /* may be deletion of method from stash */
16bdeea2 1405 SvREFCNT_dec(val);
68dc0745 1406 if (HeKLEN(entry) == HEf_SVKEY) {
1407 SvREFCNT_dec(HeKEY_sv(entry));
8aacddc1 1408 Safefree(HeKEY_hek(entry));
44a8e56a 1409 }
1410 else if (HvSHAREKEYS(hv))
68dc0745 1411 unshare_hek(HeKEY_hek(entry));
fde52b5c 1412 else
68dc0745 1413 Safefree(HeKEY_hek(entry));
d33b2eba 1414 del_HE(entry);
79072805
LW
1415}
1416
1417void
864dbfa3 1418Perl_hv_delayfree_ent(pTHX_ HV *hv, register HE *entry)
79072805 1419{
68dc0745 1420 if (!entry)
79072805 1421 return;
68dc0745 1422 if (isGV(HeVAL(entry)) && GvCVu(HeVAL(entry)) && HvNAME(hv))
3280af22 1423 PL_sub_generation++; /* may be deletion of method from stash */
68dc0745 1424 sv_2mortal(HeVAL(entry)); /* free between statements */
1425 if (HeKLEN(entry) == HEf_SVKEY) {
1426 sv_2mortal(HeKEY_sv(entry));
1427 Safefree(HeKEY_hek(entry));
44a8e56a 1428 }
1429 else if (HvSHAREKEYS(hv))
68dc0745 1430 unshare_hek(HeKEY_hek(entry));
fde52b5c 1431 else
68dc0745 1432 Safefree(HeKEY_hek(entry));
d33b2eba 1433 del_HE(entry);
79072805
LW
1434}
1435
954c1994
GS
1436/*
1437=for apidoc hv_clear
1438
1439Clears a hash, making it empty.
1440
1441=cut
1442*/
1443
79072805 1444void
864dbfa3 1445Perl_hv_clear(pTHX_ HV *hv)
79072805 1446{
27da23d5 1447 dVAR;
cbec9347 1448 register XPVHV* xhv;
79072805
LW
1449 if (!hv)
1450 return;
49293501 1451
ecae49c0
NC
1452 DEBUG_A(Perl_hv_assert(aTHX_ hv));
1453
34c3c4e3
DM
1454 xhv = (XPVHV*)SvANY(hv);
1455
5f099cb0 1456 if (SvREADONLY(hv) && xhv->xhv_array != NULL) {
34c3c4e3 1457 /* restricted hash: convert all keys to placeholders */
3a676441
JH
1458 I32 i;
1459 HE* entry;
1460 for (i = 0; i <= (I32) xhv->xhv_max; i++) {
1461 entry = ((HE**)xhv->xhv_array)[i];
1462 for (; entry; entry = HeNEXT(entry)) {
1463 /* not already placeholder */
7996736c 1464 if (HeVAL(entry) != &PL_sv_placeholder) {
3a676441
JH
1465 if (HeVAL(entry) && SvREADONLY(HeVAL(entry))) {
1466 SV* keysv = hv_iterkeysv(entry);
1467 Perl_croak(aTHX_
1468 "Attempt to delete readonly key '%"SVf"' from a restricted hash",
1469 keysv);
1470 }
1471 SvREFCNT_dec(HeVAL(entry));
7996736c 1472 HeVAL(entry) = &PL_sv_placeholder;
3a676441
JH
1473 xhv->xhv_placeholders++; /* HvPLACEHOLDERS(hv)++ */
1474 }
34c3c4e3
DM
1475 }
1476 }
df8c6964 1477 goto reset;
49293501
MS
1478 }
1479
463ee0b2 1480 hfreeentries(hv);
8aacddc1 1481 xhv->xhv_placeholders = 0; /* HvPLACEHOLDERS(hv) = 0 */
cbec9347
JH
1482 if (xhv->xhv_array /* HvARRAY(hv) */)
1483 (void)memzero(xhv->xhv_array /* HvARRAY(hv) */,
1484 (xhv->xhv_max+1 /* HvMAX(hv)+1 */) * sizeof(HE*));
a0d0e21e
LW
1485
1486 if (SvRMAGICAL(hv))
1c846c1f 1487 mg_clear((SV*)hv);
574c8022 1488
19692e8d 1489 HvHASKFLAGS_off(hv);
bb443f97 1490 HvREHASH_off(hv);
df8c6964
TP
1491 reset:
1492 HvEITER(hv) = NULL;
79072805
LW
1493}
1494
3540d4ce
AB
1495/*
1496=for apidoc hv_clear_placeholders
1497
1498Clears any placeholders from a hash. If a restricted hash has any of its keys
1499marked as readonly and the key is subsequently deleted, the key is not actually
1500deleted but is marked by assigning it a value of &PL_sv_placeholder. This tags
1501it so it will be ignored by future operations such as iterating over the hash,
4cdaeff7 1502but will still allow the hash to have a value reassigned to the key at some
3540d4ce
AB
1503future point. This function clears any such placeholder keys from the hash.
1504See Hash::Util::lock_keys() for an example of its use.
1505
1506=cut
1507*/
1508
1509void
1510Perl_hv_clear_placeholders(pTHX_ HV *hv)
1511{
27da23d5 1512 dVAR;
d3677389
NC
1513 I32 items = (I32)HvPLACEHOLDERS(hv);
1514 I32 i = HvMAX(hv);
1515
1516 if (items == 0)
1517 return;
1518
1519 do {
1520 /* Loop down the linked list heads */
1521 int first = 1;
1522 HE **oentry = &(HvARRAY(hv))[i];
1523 HE *entry = *oentry;
1524
1525 if (!entry)
1526 continue;
1527
213ce8b3 1528 for (; entry; entry = *oentry) {
d3677389
NC
1529 if (HeVAL(entry) == &PL_sv_placeholder) {
1530 *oentry = HeNEXT(entry);
1531 if (first && !*oentry)
1532 HvFILL(hv)--; /* This linked list is now empty. */
1533 if (HvEITER(hv))
1534 HvLAZYDEL_on(hv);
1535 else
1536 hv_free_ent(hv, entry);
1537
1538 if (--items == 0) {
1539 /* Finished. */
fe2774ed 1540 HvTOTALKEYS(hv) -= (IV)HvPLACEHOLDERS(hv);
d3677389
NC
1541 if (HvKEYS(hv) == 0)
1542 HvHASKFLAGS_off(hv);
1543 HvPLACEHOLDERS(hv) = 0;
1544 return;
1545 }
213ce8b3
NC
1546 } else {
1547 oentry = &HeNEXT(entry);
1548 first = 0;
d3677389
NC
1549 }
1550 }
1551 } while (--i >= 0);
1552 /* You can't get here, hence assertion should always fail. */
1553 assert (items == 0);
1554 assert (0);
3540d4ce
AB
1555}
1556
76e3520e 1557STATIC void
cea2e8a9 1558S_hfreeentries(pTHX_ HV *hv)
79072805 1559{
a0d0e21e 1560 register HE **array;
68dc0745 1561 register HE *entry;
1562 register HE *oentry = Null(HE*);
a0d0e21e
LW
1563 I32 riter;
1564 I32 max;
79072805
LW
1565
1566 if (!hv)
1567 return;
a0d0e21e 1568 if (!HvARRAY(hv))
79072805 1569 return;
a0d0e21e
LW
1570
1571 riter = 0;
1572 max = HvMAX(hv);
1573 array = HvARRAY(hv);
2f86008e
DM
1574 /* make everyone else think the array is empty, so that the destructors
1575 * called for freed entries can't recusively mess with us */
1576 HvARRAY(hv) = Null(HE**);
1577 HvFILL(hv) = 0;
1578 ((XPVHV*) SvANY(hv))->xhv_keys = 0;
1579
68dc0745 1580 entry = array[0];
a0d0e21e 1581 for (;;) {
68dc0745 1582 if (entry) {
1583 oentry = entry;
1584 entry = HeNEXT(entry);
1585 hv_free_ent(hv, oentry);
a0d0e21e 1586 }
68dc0745 1587 if (!entry) {
a0d0e21e
LW
1588 if (++riter > max)
1589 break;
68dc0745 1590 entry = array[riter];
1c846c1f 1591 }
79072805 1592 }
2f86008e 1593 HvARRAY(hv) = array;
a0d0e21e 1594 (void)hv_iterinit(hv);
79072805
LW
1595}
1596
954c1994
GS
1597/*
1598=for apidoc hv_undef
1599
1600Undefines the hash.
1601
1602=cut
1603*/
1604
79072805 1605void
864dbfa3 1606Perl_hv_undef(pTHX_ HV *hv)
79072805 1607{
cbec9347 1608 register XPVHV* xhv;
79072805
LW
1609 if (!hv)
1610 return;
ecae49c0 1611 DEBUG_A(Perl_hv_assert(aTHX_ hv));
cbec9347 1612 xhv = (XPVHV*)SvANY(hv);
463ee0b2 1613 hfreeentries(hv);
cbec9347 1614 Safefree(xhv->xhv_array /* HvARRAY(hv) */);
85e6fe83 1615 if (HvNAME(hv)) {
7e8961ec
AB
1616 if(PL_stashcache)
1617 hv_delete(PL_stashcache, HvNAME(hv), strlen(HvNAME(hv)), G_DISCARD);
85e6fe83
LW
1618 Safefree(HvNAME(hv));
1619 HvNAME(hv) = 0;
1620 }
cbec9347
JH
1621 xhv->xhv_max = 7; /* HvMAX(hv) = 7 (it's a normal hash) */
1622 xhv->xhv_array = 0; /* HvARRAY(hv) = 0 */
8aacddc1 1623 xhv->xhv_placeholders = 0; /* HvPLACEHOLDERS(hv) = 0 */
a0d0e21e
LW
1624
1625 if (SvRMAGICAL(hv))
1c846c1f 1626 mg_clear((SV*)hv);
79072805
LW
1627}
1628
954c1994
GS
1629/*
1630=for apidoc hv_iterinit
1631
1632Prepares a starting point to traverse a hash table. Returns the number of
1633keys in the hash (i.e. the same as C<HvKEYS(tb)>). The return value is
1c846c1f 1634currently only meaningful for hashes without tie magic.
954c1994
GS
1635
1636NOTE: Before version 5.004_65, C<hv_iterinit> used to return the number of
1637hash buckets that happen to be in use. If you still need that esoteric
1638value, you can get it through the macro C<HvFILL(tb)>.
1639
e16e2ff8 1640
954c1994
GS
1641=cut
1642*/
1643
79072805 1644I32
864dbfa3 1645Perl_hv_iterinit(pTHX_ HV *hv)
79072805 1646{
cbec9347 1647 register XPVHV* xhv;
aa689395 1648 HE *entry;
1649
1650 if (!hv)
cea2e8a9 1651 Perl_croak(aTHX_ "Bad hash");
cbec9347
JH
1652 xhv = (XPVHV*)SvANY(hv);
1653 entry = xhv->xhv_eiter; /* HvEITER(hv) */
72940dca 1654 if (entry && HvLAZYDEL(hv)) { /* was deleted earlier? */
1655 HvLAZYDEL_off(hv);
68dc0745 1656 hv_free_ent(hv, entry);
72940dca 1657 }
cbec9347
JH
1658 xhv->xhv_riter = -1; /* HvRITER(hv) = -1 */
1659 xhv->xhv_eiter = Null(HE*); /* HvEITER(hv) = Null(HE*) */
1660 /* used to be xhv->xhv_fill before 5.004_65 */
8aacddc1 1661 return XHvTOTALKEYS(xhv);
79072805 1662}
954c1994
GS
1663/*
1664=for apidoc hv_iternext
1665
1666Returns entries from a hash iterator. See C<hv_iterinit>.
1667
fe7bca90
NC
1668You may call C<hv_delete> or C<hv_delete_ent> on the hash entry that the
1669iterator currently points to, without losing your place or invalidating your
1670iterator. Note that in this case the current entry is deleted from the hash
1671with your iterator holding the last reference to it. Your iterator is flagged
1672to free the entry on the next call to C<hv_iternext>, so you must not discard
1673your iterator immediately else the entry will leak - call C<hv_iternext> to
1674trigger the resource deallocation.
1675
954c1994
GS
1676=cut
1677*/
1678
79072805 1679HE *
864dbfa3 1680Perl_hv_iternext(pTHX_ HV *hv)
79072805 1681{
e16e2ff8
NC
1682 return hv_iternext_flags(hv, 0);
1683}
1684
1685/*
fe7bca90
NC
1686=for apidoc hv_iternext_flags
1687
1688Returns entries from a hash iterator. See C<hv_iterinit> and C<hv_iternext>.
1689The C<flags> value will normally be zero; if HV_ITERNEXT_WANTPLACEHOLDERS is
1690set the placeholders keys (for restricted hashes) will be returned in addition
1691to normal keys. By default placeholders are automatically skipped over.
7996736c
MHM
1692Currently a placeholder is implemented with a value that is
1693C<&Perl_sv_placeholder>. Note that the implementation of placeholders and
fe7bca90
NC
1694restricted hashes may change, and the implementation currently is
1695insufficiently abstracted for any change to be tidy.
e16e2ff8 1696
fe7bca90 1697=cut
e16e2ff8
NC
1698*/
1699
1700HE *
1701Perl_hv_iternext_flags(pTHX_ HV *hv, I32 flags)
1702{
27da23d5 1703 dVAR;
cbec9347 1704 register XPVHV* xhv;
79072805 1705 register HE *entry;
a0d0e21e 1706 HE *oldentry;
463ee0b2 1707 MAGIC* mg;
79072805
LW
1708
1709 if (!hv)
cea2e8a9 1710 Perl_croak(aTHX_ "Bad hash");
cbec9347
JH
1711 xhv = (XPVHV*)SvANY(hv);
1712 oldentry = entry = xhv->xhv_eiter; /* HvEITER(hv) */
463ee0b2 1713
14befaf4 1714 if ((mg = SvTIED_mg((SV*)hv, PERL_MAGIC_tied))) {
8990e307 1715 SV *key = sv_newmortal();
cd1469e6 1716 if (entry) {
fde52b5c 1717 sv_setsv(key, HeSVKEY_force(entry));
cd1469e6 1718 SvREFCNT_dec(HeSVKEY(entry)); /* get rid of previous key */
1719 }
a0d0e21e 1720 else {
ff68c719 1721 char *k;
bbce6d69 1722 HEK *hek;
ff68c719 1723
cbec9347
JH
1724 /* one HE per MAGICAL hash */
1725 xhv->xhv_eiter = entry = new_HE(); /* HvEITER(hv) = new_HE() */
4633a7c4 1726 Zero(entry, 1, HE);
ff68c719 1727 Newz(54, k, HEK_BASESIZE + sizeof(SV*), char);
1728 hek = (HEK*)k;
1729 HeKEY_hek(entry) = hek;
fde52b5c 1730 HeKLEN(entry) = HEf_SVKEY;
a0d0e21e
LW
1731 }
1732 magic_nextpack((SV*) hv,mg,key);
8aacddc1 1733 if (SvOK(key)) {
cd1469e6 1734 /* force key to stay around until next time */
bbce6d69 1735 HeSVKEY_set(entry, SvREFCNT_inc(key));
1736 return entry; /* beware, hent_val is not set */
8aacddc1 1737 }
fde52b5c 1738 if (HeVAL(entry))
1739 SvREFCNT_dec(HeVAL(entry));
ff68c719 1740 Safefree(HeKEY_hek(entry));
d33b2eba 1741 del_HE(entry);
cbec9347 1742 xhv->xhv_eiter = Null(HE*); /* HvEITER(hv) = Null(HE*) */
463ee0b2 1743 return Null(HE*);
79072805 1744 }
f675dbe5 1745#ifdef DYNAMIC_ENV_FETCH /* set up %ENV for iteration */
cbec9347 1746 if (!entry && SvRMAGICAL((SV*)hv) && mg_find((SV*)hv, PERL_MAGIC_env))
f675dbe5
CB
1747 prime_env_iter();
1748#endif
463ee0b2 1749
cbec9347
JH
1750 if (!xhv->xhv_array /* !HvARRAY(hv) */)
1751 Newz(506, xhv->xhv_array /* HvARRAY(hv) */,
1752 PERL_HV_ARRAY_ALLOC_BYTES(xhv->xhv_max+1 /* HvMAX(hv)+1 */),
1753 char);
015a5f36 1754 /* At start of hash, entry is NULL. */
fde52b5c 1755 if (entry)
8aacddc1 1756 {
fde52b5c 1757 entry = HeNEXT(entry);
e16e2ff8
NC
1758 if (!(flags & HV_ITERNEXT_WANTPLACEHOLDERS)) {
1759 /*
1760 * Skip past any placeholders -- don't want to include them in
1761 * any iteration.
1762 */
7996736c 1763 while (entry && HeVAL(entry) == &PL_sv_placeholder) {
e16e2ff8
NC
1764 entry = HeNEXT(entry);
1765 }
8aacddc1
NIS
1766 }
1767 }
fde52b5c 1768 while (!entry) {
015a5f36
NC
1769 /* OK. Come to the end of the current list. Grab the next one. */
1770
cbec9347 1771 xhv->xhv_riter++; /* HvRITER(hv)++ */
eb160463 1772 if (xhv->xhv_riter > (I32)xhv->xhv_max /* HvRITER(hv) > HvMAX(hv) */) {
015a5f36 1773 /* There is no next one. End of the hash. */
cbec9347 1774 xhv->xhv_riter = -1; /* HvRITER(hv) = -1 */
fde52b5c 1775 break;
79072805 1776 }
cbec9347
JH
1777 /* entry = (HvARRAY(hv))[HvRITER(hv)]; */
1778 entry = ((HE**)xhv->xhv_array)[xhv->xhv_riter];
8aacddc1 1779
e16e2ff8 1780 if (!(flags & HV_ITERNEXT_WANTPLACEHOLDERS)) {
015a5f36
NC
1781 /* If we have an entry, but it's a placeholder, don't count it.
1782 Try the next. */
7996736c 1783 while (entry && HeVAL(entry) == &PL_sv_placeholder)
015a5f36
NC
1784 entry = HeNEXT(entry);
1785 }
1786 /* Will loop again if this linked list starts NULL
1787 (for HV_ITERNEXT_WANTPLACEHOLDERS)
1788 or if we run through it and find only placeholders. */
fde52b5c 1789 }
79072805 1790
72940dca 1791 if (oldentry && HvLAZYDEL(hv)) { /* was deleted earlier? */
1792 HvLAZYDEL_off(hv);
68dc0745 1793 hv_free_ent(hv, oldentry);
72940dca 1794 }
a0d0e21e 1795
fdcd69b6
NC
1796 /*if (HvREHASH(hv) && entry && !HeKREHASH(entry))
1797 PerlIO_printf(PerlIO_stderr(), "Awooga %p %p\n", hv, entry);*/
1798
cbec9347 1799 xhv->xhv_eiter = entry; /* HvEITER(hv) = entry */
79072805
LW
1800 return entry;
1801}
1802
954c1994
GS
1803/*
1804=for apidoc hv_iterkey
1805
1806Returns the key from the current position of the hash iterator. See
1807C<hv_iterinit>.
1808
1809=cut
1810*/
1811
79072805 1812char *
864dbfa3 1813Perl_hv_iterkey(pTHX_ register HE *entry, I32 *retlen)
79072805 1814{
fde52b5c 1815 if (HeKLEN(entry) == HEf_SVKEY) {
fb73857a 1816 STRLEN len;
1817 char *p = SvPV(HeKEY_sv(entry), len);
1818 *retlen = len;
1819 return p;
fde52b5c 1820 }
1821 else {
1822 *retlen = HeKLEN(entry);
1823 return HeKEY(entry);
1824 }
1825}
1826
1827/* unlike hv_iterval(), this always returns a mortal copy of the key */
954c1994
GS
1828/*
1829=for apidoc hv_iterkeysv
1830
1831Returns the key as an C<SV*> from the current position of the hash
1832iterator. The return value will always be a mortal copy of the key. Also
1833see C<hv_iterinit>.
1834
1835=cut
1836*/
1837
fde52b5c 1838SV *
864dbfa3 1839Perl_hv_iterkeysv(pTHX_ register HE *entry)
fde52b5c 1840{
19692e8d
NC
1841 if (HeKLEN(entry) != HEf_SVKEY) {
1842 HEK *hek = HeKEY_hek(entry);
1843 int flags = HEK_FLAGS(hek);
1844 SV *sv;
1845
1846 if (flags & HVhek_WASUTF8) {
1847 /* Trouble :-)
1848 Andreas would like keys he put in as utf8 to come back as utf8
1849 */
1850 STRLEN utf8_len = HEK_LEN(hek);
2e5dfef7 1851 U8 *as_utf8 = bytes_to_utf8 ((U8*)HEK_KEY(hek), &utf8_len);
19692e8d 1852
2e5dfef7 1853 sv = newSVpvn ((char*)as_utf8, utf8_len);
19692e8d 1854 SvUTF8_on (sv);
c193270f 1855 Safefree (as_utf8); /* bytes_to_utf8() allocates a new string */
4b5190b5
NC
1856 } else if (flags & HVhek_REHASH) {
1857 /* We don't have a pointer to the hv, so we have to replicate the
1858 flag into every HEK. This hv is using custom a hasing
1859 algorithm. Hence we can't return a shared string scalar, as
1860 that would contain the (wrong) hash value, and might get passed
1861 into an hv routine with a regular hash */
1862
1863 sv = newSVpvn (HEK_KEY(hek), HEK_LEN(hek));
1864 if (HEK_UTF8(hek))
1865 SvUTF8_on (sv);
1866 } else {
19692e8d
NC
1867 sv = newSVpvn_share(HEK_KEY(hek),
1868 (HEK_UTF8(hek) ? -HEK_LEN(hek) : HEK_LEN(hek)),
1869 HEK_HASH(hek));
1870 }
1871 return sv_2mortal(sv);
1872 }
1873 return sv_mortalcopy(HeKEY_sv(entry));
79072805
LW
1874}
1875
954c1994
GS
1876/*
1877=for apidoc hv_iterval
1878
1879Returns the value from the current position of the hash iterator. See
1880C<hv_iterkey>.
1881
1882=cut
1883*/
1884
79072805 1885SV *
864dbfa3 1886Perl_hv_iterval(pTHX_ HV *hv, register HE *entry)
79072805 1887{
8990e307 1888 if (SvRMAGICAL(hv)) {
14befaf4 1889 if (mg_find((SV*)hv, PERL_MAGIC_tied)) {
8990e307 1890 SV* sv = sv_newmortal();
bbce6d69 1891 if (HeKLEN(entry) == HEf_SVKEY)
1892 mg_copy((SV*)hv, sv, (char*)HeKEY_sv(entry), HEf_SVKEY);
1893 else mg_copy((SV*)hv, sv, HeKEY(entry), HeKLEN(entry));
463ee0b2
LW
1894 return sv;
1895 }
79072805 1896 }
fde52b5c 1897 return HeVAL(entry);
79072805
LW
1898}
1899
954c1994
GS
1900/*
1901=for apidoc hv_iternextsv
1902
1903Performs an C<hv_iternext>, C<hv_iterkey>, and C<hv_iterval> in one
1904operation.
1905
1906=cut
1907*/
1908
a0d0e21e 1909SV *
864dbfa3 1910Perl_hv_iternextsv(pTHX_ HV *hv, char **key, I32 *retlen)
a0d0e21e
LW
1911{
1912 HE *he;
e16e2ff8 1913 if ( (he = hv_iternext_flags(hv, 0)) == NULL)
a0d0e21e
LW
1914 return NULL;
1915 *key = hv_iterkey(he, retlen);
1916 return hv_iterval(hv, he);
1917}
1918
954c1994
GS
1919/*
1920=for apidoc hv_magic
1921
1922Adds magic to a hash. See C<sv_magic>.
1923
1924=cut
1925*/
1926
79072805 1927void
864dbfa3 1928Perl_hv_magic(pTHX_ HV *hv, GV *gv, int how)
79072805 1929{
a0d0e21e 1930 sv_magic((SV*)hv, (SV*)gv, how, Nullch, 0);
79072805 1931}
fde52b5c 1932
37d85e3a
JH
1933#if 0 /* use the macro from hv.h instead */
1934
bbce6d69 1935char*
864dbfa3 1936Perl_sharepvn(pTHX_ const char *sv, I32 len, U32 hash)
bbce6d69 1937{
ff68c719 1938 return HEK_KEY(share_hek(sv, len, hash));
bbce6d69 1939}
1940
37d85e3a
JH
1941#endif
1942
bbce6d69 1943/* possibly free a shared string if no one has access to it
fde52b5c 1944 * len and hash must both be valid for str.
1945 */
bbce6d69 1946void
864dbfa3 1947Perl_unsharepvn(pTHX_ const char *str, I32 len, U32 hash)
fde52b5c 1948{
19692e8d
NC
1949 unshare_hek_or_pvn (NULL, str, len, hash);
1950}
1951
1952
1953void
1954Perl_unshare_hek(pTHX_ HEK *hek)
1955{
1956 unshare_hek_or_pvn(hek, NULL, 0, 0);
1957}
1958
1959/* possibly free a shared string if no one has access to it
1960 hek if non-NULL takes priority over the other 3, else str, len and hash
1961 are used. If so, len and hash must both be valid for str.
1962 */
df132699 1963STATIC void
19692e8d
NC
1964S_unshare_hek_or_pvn(pTHX_ HEK *hek, const char *str, I32 len, U32 hash)
1965{
cbec9347 1966 register XPVHV* xhv;
fde52b5c 1967 register HE *entry;
1968 register HE **oentry;
1969 register I32 i = 1;
1970 I32 found = 0;
c3654f1a 1971 bool is_utf8 = FALSE;
19692e8d 1972 int k_flags = 0;
f9a63242 1973 const char *save = str;
c3654f1a 1974
19692e8d
NC
1975 if (hek) {
1976 hash = HEK_HASH(hek);
1977 } else if (len < 0) {
1978 STRLEN tmplen = -len;
1979 is_utf8 = TRUE;
1980 /* See the note in hv_fetch(). --jhi */
1981 str = (char*)bytes_from_utf8((U8*)str, &tmplen, &is_utf8);
1982 len = tmplen;
1983 if (is_utf8)
1984 k_flags = HVhek_UTF8;
1985 if (str != save)
1986 k_flags |= HVhek_WASUTF8 | HVhek_FREEKEY;
c3654f1a 1987 }
1c846c1f 1988
fde52b5c 1989 /* what follows is the moral equivalent of:
6b88bc9c 1990 if ((Svp = hv_fetch(PL_strtab, tmpsv, FALSE, hash))) {
bbce6d69 1991 if (--*Svp == Nullsv)
6b88bc9c 1992 hv_delete(PL_strtab, str, len, G_DISCARD, hash);
bbce6d69 1993 } */
cbec9347 1994 xhv = (XPVHV*)SvANY(PL_strtab);
fde52b5c 1995 /* assert(xhv_array != 0) */
5f08fbcd 1996 LOCK_STRTAB_MUTEX;
cbec9347
JH
1997 /* oentry = &(HvARRAY(hv))[hash & (I32) HvMAX(hv)]; */
1998 oentry = &((HE**)xhv->xhv_array)[hash & (I32) xhv->xhv_max];
19692e8d
NC
1999 if (hek) {
2000 for (entry = *oentry; entry; i=0, oentry = &HeNEXT(entry), entry = *oentry) {
2001 if (HeKEY_hek(entry) != hek)
2002 continue;
2003 found = 1;
2004 break;
2005 }
2006 } else {
35a4481c 2007 const int flags_masked = k_flags & HVhek_MASK;
19692e8d
NC
2008 for (entry = *oentry; entry; i=0, oentry = &HeNEXT(entry), entry = *oentry) {
2009 if (HeHASH(entry) != hash) /* strings can't be equal */
2010 continue;
2011 if (HeKLEN(entry) != len)
2012 continue;
2013 if (HeKEY(entry) != str && memNE(HeKEY(entry),str,len)) /* is this it? */
2014 continue;
2015 if (HeKFLAGS(entry) != flags_masked)
2016 continue;
2017 found = 1;
2018 break;
2019 }
2020 }
2021
2022 if (found) {
2023 if (--HeVAL(entry) == Nullsv) {
2024 *oentry = HeNEXT(entry);
2025 if (i && !*oentry)
2026 xhv->xhv_fill--; /* HvFILL(hv)-- */
2027 Safefree(HeKEY_hek(entry));
2028 del_HE(entry);
2029 xhv->xhv_keys--; /* HvKEYS(hv)-- */
2030 }
fde52b5c 2031 }
19692e8d 2032
333f433b 2033 UNLOCK_STRTAB_MUTEX;
411caa50 2034 if (!found && ckWARN_d(WARN_INTERNAL))
19692e8d 2035 Perl_warner(aTHX_ packWARN(WARN_INTERNAL),
472d47bc
SB
2036 "Attempt to free non-existent shared string '%s'%s"
2037 pTHX__FORMAT,
19692e8d 2038 hek ? HEK_KEY(hek) : str,
472d47bc 2039 ((k_flags & HVhek_UTF8) ? " (utf8)" : "") pTHX__VALUE);
19692e8d
NC
2040 if (k_flags & HVhek_FREEKEY)
2041 Safefree(str);
fde52b5c 2042}
2043
bbce6d69 2044/* get a (constant) string ptr from the global string table
2045 * string will get added if it is not already there.
fde52b5c 2046 * len and hash must both be valid for str.
2047 */
bbce6d69 2048HEK *
864dbfa3 2049Perl_share_hek(pTHX_ const char *str, I32 len, register U32 hash)
fde52b5c 2050{
da58a35d 2051 bool is_utf8 = FALSE;
19692e8d 2052 int flags = 0;
f9a63242 2053 const char *save = str;
da58a35d
JH
2054
2055 if (len < 0) {
77caf834 2056 STRLEN tmplen = -len;
da58a35d 2057 is_utf8 = TRUE;
77caf834
JH
2058 /* See the note in hv_fetch(). --jhi */
2059 str = (char*)bytes_from_utf8((U8*)str, &tmplen, &is_utf8);
2060 len = tmplen;
19692e8d
NC
2061 /* If we were able to downgrade here, then than means that we were passed
2062 in a key which only had chars 0-255, but was utf8 encoded. */
2063 if (is_utf8)
2064 flags = HVhek_UTF8;
2065 /* If we found we were able to downgrade the string to bytes, then
2066 we should flag that it needs upgrading on keys or each. Also flag
2067 that we need share_hek_flags to free the string. */
2068 if (str != save)
2069 flags |= HVhek_WASUTF8 | HVhek_FREEKEY;
2070 }
2071
2072 return share_hek_flags (str, len, hash, flags);
2073}
2074
df132699 2075STATIC HEK *
19692e8d
NC
2076S_share_hek_flags(pTHX_ const char *str, I32 len, register U32 hash, int flags)
2077{
2078 register XPVHV* xhv;
2079 register HE *entry;
2080 register HE **oentry;
2081 register I32 i = 1;
2082 I32 found = 0;
35a4481c 2083 const int flags_masked = flags & HVhek_MASK;
bbce6d69 2084
fde52b5c 2085 /* what follows is the moral equivalent of:
1c846c1f 2086
6b88bc9c 2087 if (!(Svp = hv_fetch(PL_strtab, str, len, FALSE)))
8aacddc1 2088 hv_store(PL_strtab, str, len, Nullsv, hash);
fdcd69b6
NC
2089
2090 Can't rehash the shared string table, so not sure if it's worth
2091 counting the number of entries in the linked list
bbce6d69 2092 */
cbec9347 2093 xhv = (XPVHV*)SvANY(PL_strtab);
fde52b5c 2094 /* assert(xhv_array != 0) */
5f08fbcd 2095 LOCK_STRTAB_MUTEX;
cbec9347
JH
2096 /* oentry = &(HvARRAY(hv))[hash & (I32) HvMAX(hv)]; */
2097 oentry = &((HE**)xhv->xhv_array)[hash & (I32) xhv->xhv_max];
bbce6d69 2098 for (entry = *oentry; entry; i=0, entry = HeNEXT(entry)) {
fde52b5c 2099 if (HeHASH(entry) != hash) /* strings can't be equal */
2100 continue;
2101 if (HeKLEN(entry) != len)
2102 continue;
1c846c1f 2103 if (HeKEY(entry) != str && memNE(HeKEY(entry),str,len)) /* is this it? */
fde52b5c 2104 continue;
19692e8d 2105 if (HeKFLAGS(entry) != flags_masked)
c3654f1a 2106 continue;
fde52b5c 2107 found = 1;
fde52b5c 2108 break;
2109 }
bbce6d69 2110 if (!found) {
d33b2eba 2111 entry = new_HE();
dcf933a4 2112 HeKEY_hek(entry) = save_hek_flags(str, len, hash, flags_masked);
bbce6d69 2113 HeVAL(entry) = Nullsv;
2114 HeNEXT(entry) = *oentry;
2115 *oentry = entry;
cbec9347 2116 xhv->xhv_keys++; /* HvKEYS(hv)++ */
bbce6d69 2117 if (i) { /* initial entry? */
cbec9347 2118 xhv->xhv_fill++; /* HvFILL(hv)++ */
4c9cc595 2119 } else if (xhv->xhv_keys > (IV)xhv->xhv_max /* HvKEYS(hv) > HvMAX(hv) */) {
cbec9347 2120 hsplit(PL_strtab);
bbce6d69 2121 }
2122 }
2123
2124 ++HeVAL(entry); /* use value slot as REFCNT */
5f08fbcd 2125 UNLOCK_STRTAB_MUTEX;
19692e8d
NC
2126
2127 if (flags & HVhek_FREEKEY)
f9a63242 2128 Safefree(str);
19692e8d 2129
ff68c719 2130 return HeKEY_hek(entry);
fde52b5c 2131}
ecae49c0
NC
2132
2133
2134/*
2135=for apidoc hv_assert
2136
2137Check that a hash is in an internally consistent state.
2138
2139=cut
2140*/
2141
2142void
2143Perl_hv_assert(pTHX_ HV *hv)
2144{
27da23d5 2145 dVAR;
ecae49c0
NC
2146 HE* entry;
2147 int withflags = 0;
2148 int placeholders = 0;
2149 int real = 0;
2150 int bad = 0;
2151 I32 riter = HvRITER(hv);
2152 HE *eiter = HvEITER(hv);
2153
2154 (void)hv_iterinit(hv);
2155
2156 while ((entry = hv_iternext_flags(hv, HV_ITERNEXT_WANTPLACEHOLDERS))) {
2157 /* sanity check the values */
2158 if (HeVAL(entry) == &PL_sv_placeholder) {
2159 placeholders++;
2160 } else {
2161 real++;
2162 }
2163 /* sanity check the keys */
2164 if (HeSVKEY(entry)) {
2165 /* Don't know what to check on SV keys. */
2166 } else if (HeKUTF8(entry)) {
2167 withflags++;
2168 if (HeKWASUTF8(entry)) {
2169 PerlIO_printf(Perl_debug_log,
2170 "hash key has both WASUFT8 and UTF8: '%.*s'\n",
2171 (int) HeKLEN(entry), HeKEY(entry));
2172 bad = 1;
2173 }
2174 } else if (HeKWASUTF8(entry)) {
2175 withflags++;
2176 }
2177 }
2178 if (!SvTIED_mg((SV*)hv, PERL_MAGIC_tied)) {
2179 if (HvUSEDKEYS(hv) != real) {
2180 PerlIO_printf(Perl_debug_log, "Count %d key(s), but hash reports %d\n",
2181 (int) real, (int) HvUSEDKEYS(hv));
2182 bad = 1;
2183 }
2184 if (HvPLACEHOLDERS(hv) != placeholders) {
2185 PerlIO_printf(Perl_debug_log,
2186 "Count %d placeholder(s), but hash reports %d\n",
2187 (int) placeholders, (int) HvPLACEHOLDERS(hv));
2188 bad = 1;
2189 }
2190 }
2191 if (withflags && ! HvHASKFLAGS(hv)) {
2192 PerlIO_printf(Perl_debug_log,
2193 "Hash has HASKFLAGS off but I count %d key(s) with flags\n",
2194 withflags);
2195 bad = 1;
2196 }
2197 if (bad) {
2198 sv_dump((SV *)hv);
2199 }
2200 HvRITER(hv) = riter; /* Restore hash iterator state */
2201 HvEITER(hv) = eiter;
2202}
af3babe4
NC
2203
2204/*
2205 * Local variables:
2206 * c-indentation-style: bsd
2207 * c-basic-offset: 4
2208 * indent-tabs-mode: t
2209 * End:
2210 *
2211 * vim: shiftwidth=4:
2212*/