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