Commit | Line | Data |
---|---|---|
a0d0e21e | 1 | /* hv.c |
79072805 | 2 | * |
1129b882 NC |
3 | * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, 2000, |
4 | * 2001, 2002, 2003, 2004, 2005, 2006, 2007, 2008 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 | /* | |
4ac71550 TC |
12 | * I sit beside the fire and think |
13 | * of all that I have seen. | |
14 | * --Bilbo | |
15 | * | |
16 | * [p.278 of _The Lord of the Rings_, II/iii: "The Ring Goes South"] | |
79072805 LW |
17 | */ |
18 | ||
d5afce77 RB |
19 | /* |
20 | =head1 Hash Manipulation Functions | |
166f8a29 | 21 | |
db4fbf16 FC |
22 | A HV structure represents a Perl hash. It consists mainly of an array |
23 | of pointers, each of which points to a linked list of HE structures. The | |
166f8a29 | 24 | array is indexed by the hash function of the key, so each linked list |
db4fbf16 | 25 | represents all the hash entries with the same hash value. Each HE contains |
166f8a29 DM |
26 | a pointer to the actual value, plus a pointer to a HEK structure which |
27 | holds the key and hash value. | |
28 | ||
29 | =cut | |
30 | ||
d5afce77 RB |
31 | */ |
32 | ||
79072805 | 33 | #include "EXTERN.h" |
864dbfa3 | 34 | #define PERL_IN_HV_C |
3d78eb94 | 35 | #define PERL_HASH_INTERNAL_ACCESS |
79072805 LW |
36 | #include "perl.h" |
37 | ||
d8012aaf | 38 | #define HV_MAX_LENGTH_BEFORE_SPLIT 14 |
fdcd69b6 | 39 | |
d75ce684 | 40 | static const char S_strtab_error[] |
5d2b1485 NC |
41 | = "Cannot modify shared string table in hv_%s"; |
42 | ||
c941fb51 NC |
43 | #ifdef PURIFY |
44 | ||
45 | #define new_HE() (HE*)safemalloc(sizeof(HE)) | |
46 | #define del_HE(p) safefree((char*)p) | |
47 | ||
48 | #else | |
49 | ||
76e3520e | 50 | STATIC HE* |
cea2e8a9 | 51 | S_new_he(pTHX) |
4633a7c4 | 52 | { |
97aff369 | 53 | dVAR; |
4633a7c4 | 54 | HE* he; |
0bd48802 | 55 | void ** const root = &PL_body_roots[HE_SVSLOT]; |
6a93a7e5 | 56 | |
6a93a7e5 | 57 | if (!*root) |
1e30fcd5 | 58 | Perl_more_bodies(aTHX_ HE_SVSLOT, sizeof(HE), PERL_ARENA_SIZE); |
10edeb5d | 59 | he = (HE*) *root; |
ce3e5c45 | 60 | assert(he); |
6a93a7e5 | 61 | *root = HeNEXT(he); |
333f433b | 62 | return he; |
4633a7c4 LW |
63 | } |
64 | ||
c941fb51 NC |
65 | #define new_HE() new_he() |
66 | #define del_HE(p) \ | |
67 | STMT_START { \ | |
6a93a7e5 NC |
68 | HeNEXT(p) = (HE*)(PL_body_roots[HE_SVSLOT]); \ |
69 | PL_body_roots[HE_SVSLOT] = p; \ | |
c941fb51 | 70 | } STMT_END |
d33b2eba | 71 | |
d33b2eba | 72 | |
d33b2eba GS |
73 | |
74 | #endif | |
75 | ||
76e3520e | 76 | STATIC HEK * |
5f66b61c | 77 | S_save_hek_flags(const char *str, I32 len, U32 hash, int flags) |
bbce6d69 | 78 | { |
35a4481c | 79 | const int flags_masked = flags & HVhek_MASK; |
bbce6d69 | 80 | char *k; |
81 | register HEK *hek; | |
1c846c1f | 82 | |
7918f24d NC |
83 | PERL_ARGS_ASSERT_SAVE_HEK_FLAGS; |
84 | ||
a02a5408 | 85 | Newx(k, HEK_BASESIZE + len + 2, char); |
bbce6d69 | 86 | hek = (HEK*)k; |
ff68c719 | 87 | Copy(str, HEK_KEY(hek), len, char); |
e05949c7 | 88 | HEK_KEY(hek)[len] = 0; |
ff68c719 | 89 | HEK_LEN(hek) = len; |
90 | HEK_HASH(hek) = hash; | |
45e34800 | 91 | HEK_FLAGS(hek) = (unsigned char)flags_masked | HVhek_UNSHARED; |
dcf933a4 NC |
92 | |
93 | if (flags & HVhek_FREEKEY) | |
94 | Safefree(str); | |
bbce6d69 | 95 | return hek; |
96 | } | |
97 | ||
4a31713e | 98 | /* free the pool of temporary HE/HEK pairs returned by hv_fetch_ent |
dd28f7bb DM |
99 | * for tied hashes */ |
100 | ||
101 | void | |
102 | Perl_free_tied_hv_pool(pTHX) | |
103 | { | |
97aff369 | 104 | dVAR; |
dd28f7bb DM |
105 | HE *he = PL_hv_fetch_ent_mh; |
106 | while (he) { | |
9d4ba2ae | 107 | HE * const ohe = he; |
dd28f7bb | 108 | Safefree(HeKEY_hek(he)); |
dd28f7bb DM |
109 | he = HeNEXT(he); |
110 | del_HE(ohe); | |
111 | } | |
4608196e | 112 | PL_hv_fetch_ent_mh = NULL; |
dd28f7bb DM |
113 | } |
114 | ||
d18c6117 | 115 | #if defined(USE_ITHREADS) |
0bff533c NC |
116 | HEK * |
117 | Perl_hek_dup(pTHX_ HEK *source, CLONE_PARAMS* param) | |
118 | { | |
566771cc | 119 | HEK *shared; |
9d4ba2ae | 120 | |
7918f24d | 121 | PERL_ARGS_ASSERT_HEK_DUP; |
9d4ba2ae | 122 | PERL_UNUSED_ARG(param); |
0bff533c | 123 | |
566771cc NC |
124 | if (!source) |
125 | return NULL; | |
126 | ||
127 | shared = (HEK*)ptr_table_fetch(PL_ptr_table, source); | |
0bff533c NC |
128 | if (shared) { |
129 | /* We already shared this hash key. */ | |
454f1e26 | 130 | (void)share_hek_hek(shared); |
0bff533c NC |
131 | } |
132 | else { | |
658b4a4a | 133 | shared |
6e838c70 NC |
134 | = share_hek_flags(HEK_KEY(source), HEK_LEN(source), |
135 | HEK_HASH(source), HEK_FLAGS(source)); | |
658b4a4a | 136 | ptr_table_store(PL_ptr_table, source, shared); |
0bff533c | 137 | } |
658b4a4a | 138 | return shared; |
0bff533c NC |
139 | } |
140 | ||
d18c6117 | 141 | HE * |
5c4138a0 | 142 | Perl_he_dup(pTHX_ const HE *e, bool shared, CLONE_PARAMS* param) |
d18c6117 GS |
143 | { |
144 | HE *ret; | |
145 | ||
7918f24d NC |
146 | PERL_ARGS_ASSERT_HE_DUP; |
147 | ||
d18c6117 | 148 | if (!e) |
4608196e | 149 | return NULL; |
7766f137 GS |
150 | /* look for it in the table first */ |
151 | ret = (HE*)ptr_table_fetch(PL_ptr_table, e); | |
152 | if (ret) | |
153 | return ret; | |
154 | ||
155 | /* create anew and remember what it is */ | |
d33b2eba | 156 | ret = new_HE(); |
7766f137 GS |
157 | ptr_table_store(PL_ptr_table, e, ret); |
158 | ||
d2d73c3e | 159 | HeNEXT(ret) = he_dup(HeNEXT(e),shared, param); |
dd28f7bb DM |
160 | if (HeKLEN(e) == HEf_SVKEY) { |
161 | char *k; | |
ad64d0ec | 162 | Newx(k, HEK_BASESIZE + sizeof(const SV *), char); |
dd28f7bb | 163 | HeKEY_hek(ret) = (HEK*)k; |
a09252eb | 164 | HeKEY_sv(ret) = sv_dup_inc(HeKEY_sv(e), param); |
dd28f7bb | 165 | } |
c21d1a0f | 166 | else if (shared) { |
0bff533c NC |
167 | /* This is hek_dup inlined, which seems to be important for speed |
168 | reasons. */ | |
1b6737cc | 169 | HEK * const source = HeKEY_hek(e); |
658b4a4a | 170 | HEK *shared = (HEK*)ptr_table_fetch(PL_ptr_table, source); |
c21d1a0f NC |
171 | |
172 | if (shared) { | |
173 | /* We already shared this hash key. */ | |
454f1e26 | 174 | (void)share_hek_hek(shared); |
c21d1a0f NC |
175 | } |
176 | else { | |
658b4a4a | 177 | shared |
6e838c70 NC |
178 | = share_hek_flags(HEK_KEY(source), HEK_LEN(source), |
179 | HEK_HASH(source), HEK_FLAGS(source)); | |
658b4a4a | 180 | ptr_table_store(PL_ptr_table, source, shared); |
c21d1a0f | 181 | } |
658b4a4a | 182 | HeKEY_hek(ret) = shared; |
c21d1a0f | 183 | } |
d18c6117 | 184 | else |
19692e8d NC |
185 | HeKEY_hek(ret) = save_hek_flags(HeKEY(e), HeKLEN(e), HeHASH(e), |
186 | HeKFLAGS(e)); | |
a09252eb | 187 | HeVAL(ret) = sv_dup_inc(HeVAL(e), param); |
d18c6117 GS |
188 | return ret; |
189 | } | |
190 | #endif /* USE_ITHREADS */ | |
191 | ||
1b1f1335 | 192 | static void |
2393f1b9 JH |
193 | S_hv_notallowed(pTHX_ int flags, const char *key, I32 klen, |
194 | const char *msg) | |
1b1f1335 | 195 | { |
1b6737cc | 196 | SV * const sv = sv_newmortal(); |
7918f24d NC |
197 | |
198 | PERL_ARGS_ASSERT_HV_NOTALLOWED; | |
199 | ||
19692e8d | 200 | if (!(flags & HVhek_FREEKEY)) { |
1b1f1335 NIS |
201 | sv_setpvn(sv, key, klen); |
202 | } | |
203 | else { | |
204 | /* Need to free saved eventually assign to mortal SV */ | |
34c3c4e3 | 205 | /* XXX is this line an error ???: SV *sv = sv_newmortal(); */ |
1b1f1335 NIS |
206 | sv_usepvn(sv, (char *) key, klen); |
207 | } | |
19692e8d | 208 | if (flags & HVhek_UTF8) { |
1b1f1335 NIS |
209 | SvUTF8_on(sv); |
210 | } | |
be2597df | 211 | Perl_croak(aTHX_ msg, SVfARG(sv)); |
1b1f1335 NIS |
212 | } |
213 | ||
fde52b5c | 214 | /* (klen == HEf_SVKEY) is special for MAGICAL hv entries, meaning key slot |
215 | * contains an SV* */ | |
216 | ||
34a6f7b4 NC |
217 | /* |
218 | =for apidoc hv_store | |
219 | ||
a05d6c5d TC |
220 | Stores an SV in a hash. The hash key is specified as C<key> and the |
221 | absolute value of C<klen> is the length of the key. If C<klen> is | |
222 | negative the key is assumed to be in UTF-8-encoded Unicode. The | |
223 | C<hash> parameter is the precomputed hash value; if it is zero then | |
224 | Perl will compute it. | |
225 | ||
226 | The return value will be | |
34a6f7b4 NC |
227 | NULL if the operation failed or if the value did not need to be actually |
228 | stored within the hash (as in the case of tied hashes). Otherwise it can | |
229 | be dereferenced to get the original C<SV*>. Note that the caller is | |
230 | responsible for suitably incrementing the reference count of C<val> before | |
231 | the call, and decrementing it if the function returned NULL. Effectively | |
232 | a successful hv_store takes ownership of one reference to C<val>. This is | |
233 | usually what you want; a newly created SV has a reference count of one, so | |
234 | if all your code does is create SVs then store them in a hash, hv_store | |
235 | will own the only reference to the new SV, and your code doesn't need to do | |
236 | anything further to tidy up. hv_store is not implemented as a call to | |
237 | hv_store_ent, and does not create a temporary SV for the key, so if your | |
238 | key data is not already in SV form then use hv_store in preference to | |
239 | hv_store_ent. | |
240 | ||
241 | See L<perlguts/"Understanding the Magic of Tied Hashes and Arrays"> for more | |
242 | information on how to use this function on tied hashes. | |
243 | ||
34a6f7b4 NC |
244 | =for apidoc hv_store_ent |
245 | ||
246 | Stores C<val> in a hash. The hash key is specified as C<key>. The C<hash> | |
247 | parameter is the precomputed hash value; if it is zero then Perl will | |
248 | compute it. The return value is the new hash entry so created. It will be | |
249 | NULL if the operation failed or if the value did not need to be actually | |
250 | stored within the hash (as in the case of tied hashes). Otherwise the | |
251 | contents of the return value can be accessed using the C<He?> macros | |
252 | described here. Note that the caller is responsible for suitably | |
253 | incrementing the reference count of C<val> before the call, and | |
254 | decrementing it if the function returned NULL. Effectively a successful | |
255 | hv_store_ent takes ownership of one reference to C<val>. This is | |
256 | usually what you want; a newly created SV has a reference count of one, so | |
257 | if all your code does is create SVs then store them in a hash, hv_store | |
258 | will own the only reference to the new SV, and your code doesn't need to do | |
259 | anything further to tidy up. Note that hv_store_ent only reads the C<key>; | |
260 | unlike C<val> it does not take ownership of it, so maintaining the correct | |
261 | reference count on C<key> is entirely the caller's responsibility. hv_store | |
262 | is not implemented as a call to hv_store_ent, and does not create a temporary | |
263 | SV for the key, so if your key data is not already in SV form then use | |
264 | hv_store in preference to hv_store_ent. | |
265 | ||
266 | See L<perlguts/"Understanding the Magic of Tied Hashes and Arrays"> for more | |
267 | information on how to use this function on tied hashes. | |
268 | ||
34a6f7b4 NC |
269 | =for apidoc hv_exists |
270 | ||
271 | Returns a boolean indicating whether the specified hash key exists. The | |
a05d6c5d TC |
272 | absolute value of C<klen> is the length of the key. If C<klen> is |
273 | negative the key is assumed to be in UTF-8-encoded Unicode. | |
34a6f7b4 | 274 | |
954c1994 GS |
275 | =for apidoc hv_fetch |
276 | ||
a05d6c5d TC |
277 | Returns the SV which corresponds to the specified key in the hash. |
278 | The absolute value of C<klen> is the length of the key. If C<klen> is | |
279 | negative the key is assumed to be in UTF-8-encoded Unicode. If | |
280 | C<lval> is set then the fetch will be part of a store. Check that the | |
281 | return value is non-null before dereferencing it to an C<SV*>. | |
954c1994 | 282 | |
96f1132b | 283 | See L<perlguts/"Understanding the Magic of Tied Hashes and Arrays"> for more |
954c1994 GS |
284 | information on how to use this function on tied hashes. |
285 | ||
34a6f7b4 NC |
286 | =for apidoc hv_exists_ent |
287 | ||
db4fbf16 FC |
288 | Returns a boolean indicating whether |
289 | the specified hash key exists. C<hash> | |
34a6f7b4 NC |
290 | can be a valid precomputed hash value, or 0 to ask for it to be |
291 | computed. | |
292 | ||
293 | =cut | |
294 | */ | |
295 | ||
d1be9408 | 296 | /* returns an HE * structure with the all fields set */ |
fde52b5c | 297 | /* note that hent_val will be a mortal sv for MAGICAL hashes */ |
954c1994 GS |
298 | /* |
299 | =for apidoc hv_fetch_ent | |
300 | ||
301 | Returns the hash entry which corresponds to the specified key in the hash. | |
302 | C<hash> must be a valid precomputed hash number for the given C<key>, or 0 | |
303 | if you want the function to compute it. IF C<lval> is set then the fetch | |
304 | will be part of a store. Make sure the return value is non-null before | |
b24b84ef | 305 | accessing it. The return value when C<hv> is a tied hash is a pointer to a |
954c1994 | 306 | static location, so be sure to make a copy of the structure if you need to |
1c846c1f | 307 | store it somewhere. |
954c1994 | 308 | |
96f1132b | 309 | See L<perlguts/"Understanding the Magic of Tied Hashes and Arrays"> for more |
954c1994 GS |
310 | information on how to use this function on tied hashes. |
311 | ||
312 | =cut | |
313 | */ | |
314 | ||
a038e571 NC |
315 | /* Common code for hv_delete()/hv_exists()/hv_fetch()/hv_store() */ |
316 | void * | |
317 | Perl_hv_common_key_len(pTHX_ HV *hv, const char *key, I32 klen_i32, | |
318 | const int action, SV *val, const U32 hash) | |
319 | { | |
320 | STRLEN klen; | |
321 | int flags; | |
322 | ||
7918f24d NC |
323 | PERL_ARGS_ASSERT_HV_COMMON_KEY_LEN; |
324 | ||
a038e571 NC |
325 | if (klen_i32 < 0) { |
326 | klen = -klen_i32; | |
327 | flags = HVhek_UTF8; | |
328 | } else { | |
329 | klen = klen_i32; | |
330 | flags = 0; | |
331 | } | |
332 | return hv_common(hv, NULL, key, klen, flags, action, val, hash); | |
333 | } | |
334 | ||
63c89345 | 335 | void * |
d3ba3f5c NC |
336 | Perl_hv_common(pTHX_ HV *hv, SV *keysv, const char *key, STRLEN klen, |
337 | int flags, int action, SV *val, register U32 hash) | |
113738bb | 338 | { |
27da23d5 | 339 | dVAR; |
b2c64049 | 340 | XPVHV* xhv; |
b2c64049 NC |
341 | HE *entry; |
342 | HE **oentry; | |
fde52b5c | 343 | SV *sv; |
da58a35d | 344 | bool is_utf8; |
113738bb | 345 | int masked_flags; |
3c84c864 | 346 | const int return_svp = action & HV_FETCH_JUST_SV; |
fde52b5c | 347 | |
348 | if (!hv) | |
a4fc7abc | 349 | return NULL; |
e4787c0c | 350 | if (SvTYPE(hv) == (svtype)SVTYPEMASK) |
8265e3d1 NC |
351 | return NULL; |
352 | ||
353 | assert(SvTYPE(hv) == SVt_PVHV); | |
fde52b5c | 354 | |
bdee33e4 | 355 | if (SvSMAGICAL(hv) && SvGMAGICAL(hv) && !(action & HV_DISABLE_UVAR_XKEY)) { |
fda2d18a | 356 | MAGIC* mg; |
ad64d0ec | 357 | if ((mg = mg_find((const SV *)hv, PERL_MAGIC_uvar))) { |
fda2d18a NC |
358 | struct ufuncs * const uf = (struct ufuncs *)mg->mg_ptr; |
359 | if (uf->uf_set == NULL) { | |
360 | SV* obj = mg->mg_obj; | |
361 | ||
362 | if (!keysv) { | |
59cd0e26 NC |
363 | keysv = newSVpvn_flags(key, klen, SVs_TEMP | |
364 | ((flags & HVhek_UTF8) | |
365 | ? SVf_UTF8 : 0)); | |
fda2d18a NC |
366 | } |
367 | ||
368 | mg->mg_obj = keysv; /* pass key */ | |
369 | uf->uf_index = action; /* pass action */ | |
ad64d0ec | 370 | magic_getuvar(MUTABLE_SV(hv), mg); |
fda2d18a NC |
371 | keysv = mg->mg_obj; /* may have changed */ |
372 | mg->mg_obj = obj; | |
373 | ||
374 | /* If the key may have changed, then we need to invalidate | |
375 | any passed-in computed hash value. */ | |
376 | hash = 0; | |
377 | } | |
378 | } | |
bdee33e4 | 379 | } |
113738bb | 380 | if (keysv) { |
e593d2fe AE |
381 | if (flags & HVhek_FREEKEY) |
382 | Safefree(key); | |
5c144d81 | 383 | key = SvPV_const(keysv, klen); |
113738bb | 384 | is_utf8 = (SvUTF8(keysv) != 0); |
44b87b50 NC |
385 | if (SvIsCOW_shared_hash(keysv)) { |
386 | flags = HVhek_KEYCANONICAL | (is_utf8 ? HVhek_UTF8 : 0); | |
387 | } else { | |
388 | flags = 0; | |
389 | } | |
113738bb | 390 | } else { |
c1fe5510 | 391 | is_utf8 = ((flags & HVhek_UTF8) ? TRUE : FALSE); |
113738bb | 392 | } |
113738bb | 393 | |
9dbc5603 | 394 | if (action & HV_DELETE) { |
3c84c864 NC |
395 | return (void *) hv_delete_common(hv, keysv, key, klen, |
396 | flags | (is_utf8 ? HVhek_UTF8 : 0), | |
397 | action, hash); | |
9dbc5603 NC |
398 | } |
399 | ||
b2c64049 | 400 | xhv = (XPVHV*)SvANY(hv); |
7f66fda2 | 401 | if (SvMAGICAL(hv)) { |
6136c704 | 402 | if (SvRMAGICAL(hv) && !(action & (HV_FETCH_ISSTORE|HV_FETCH_ISEXISTS))) { |
ad64d0ec NC |
403 | if (mg_find((const SV *)hv, PERL_MAGIC_tied) |
404 | || SvGMAGICAL((const SV *)hv)) | |
e62cc96a | 405 | { |
3c84c864 | 406 | /* FIXME should be able to skimp on the HE/HEK here when |
7f66fda2 | 407 | HV_FETCH_JUST_SV is true. */ |
7f66fda2 | 408 | if (!keysv) { |
740cce10 NC |
409 | keysv = newSVpvn_utf8(key, klen, is_utf8); |
410 | } else { | |
7f66fda2 | 411 | keysv = newSVsv(keysv); |
113738bb | 412 | } |
44a2ac75 | 413 | sv = sv_newmortal(); |
ad64d0ec | 414 | mg_copy(MUTABLE_SV(hv), sv, (char *)keysv, HEf_SVKEY); |
7f66fda2 NC |
415 | |
416 | /* grab a fake HE/HEK pair from the pool or make a new one */ | |
417 | entry = PL_hv_fetch_ent_mh; | |
418 | if (entry) | |
419 | PL_hv_fetch_ent_mh = HeNEXT(entry); | |
420 | else { | |
421 | char *k; | |
422 | entry = new_HE(); | |
ad64d0ec | 423 | Newx(k, HEK_BASESIZE + sizeof(const SV *), char); |
7f66fda2 NC |
424 | HeKEY_hek(entry) = (HEK*)k; |
425 | } | |
4608196e | 426 | HeNEXT(entry) = NULL; |
7f66fda2 NC |
427 | HeSVKEY_set(entry, keysv); |
428 | HeVAL(entry) = sv; | |
429 | sv_upgrade(sv, SVt_PVLV); | |
430 | LvTYPE(sv) = 'T'; | |
431 | /* so we can free entry when freeing sv */ | |
ad64d0ec | 432 | LvTARG(sv) = MUTABLE_SV(entry); |
7f66fda2 NC |
433 | |
434 | /* XXX remove at some point? */ | |
435 | if (flags & HVhek_FREEKEY) | |
436 | Safefree(key); | |
437 | ||
3c84c864 NC |
438 | if (return_svp) { |
439 | return entry ? (void *) &HeVAL(entry) : NULL; | |
440 | } | |
441 | return (void *) entry; | |
113738bb | 442 | } |
7f66fda2 | 443 | #ifdef ENV_IS_CASELESS |
ad64d0ec | 444 | else if (mg_find((const SV *)hv, PERL_MAGIC_env)) { |
7f66fda2 NC |
445 | U32 i; |
446 | for (i = 0; i < klen; ++i) | |
447 | if (isLOWER(key[i])) { | |
086cb327 NC |
448 | /* Would be nice if we had a routine to do the |
449 | copy and upercase in a single pass through. */ | |
0bd48802 | 450 | const char * const nkey = strupr(savepvn(key,klen)); |
086cb327 NC |
451 | /* Note that this fetch is for nkey (the uppercased |
452 | key) whereas the store is for key (the original) */ | |
63c89345 NC |
453 | void *result = hv_common(hv, NULL, nkey, klen, |
454 | HVhek_FREEKEY, /* free nkey */ | |
455 | 0 /* non-LVAL fetch */ | |
3c84c864 NC |
456 | | HV_DISABLE_UVAR_XKEY |
457 | | return_svp, | |
63c89345 NC |
458 | NULL /* no value */, |
459 | 0 /* compute hash */); | |
26488bcf | 460 | if (!result && (action & HV_FETCH_LVALUE)) { |
086cb327 NC |
461 | /* This call will free key if necessary. |
462 | Do it this way to encourage compiler to tail | |
463 | call optimise. */ | |
63c89345 NC |
464 | result = hv_common(hv, keysv, key, klen, flags, |
465 | HV_FETCH_ISSTORE | |
3c84c864 NC |
466 | | HV_DISABLE_UVAR_XKEY |
467 | | return_svp, | |
63c89345 | 468 | newSV(0), hash); |
086cb327 NC |
469 | } else { |
470 | if (flags & HVhek_FREEKEY) | |
471 | Safefree(key); | |
472 | } | |
63c89345 | 473 | return result; |
7f66fda2 | 474 | } |
902173a3 | 475 | } |
7f66fda2 NC |
476 | #endif |
477 | } /* ISFETCH */ | |
478 | else if (SvRMAGICAL(hv) && (action & HV_FETCH_ISEXISTS)) { | |
ad64d0ec NC |
479 | if (mg_find((const SV *)hv, PERL_MAGIC_tied) |
480 | || SvGMAGICAL((const SV *)hv)) { | |
b2c64049 NC |
481 | /* I don't understand why hv_exists_ent has svret and sv, |
482 | whereas hv_exists only had one. */ | |
9d4ba2ae | 483 | SV * const svret = sv_newmortal(); |
b2c64049 | 484 | sv = sv_newmortal(); |
7f66fda2 NC |
485 | |
486 | if (keysv || is_utf8) { | |
487 | if (!keysv) { | |
740cce10 | 488 | keysv = newSVpvn_utf8(key, klen, TRUE); |
7f66fda2 NC |
489 | } else { |
490 | keysv = newSVsv(keysv); | |
491 | } | |
ad64d0ec | 492 | mg_copy(MUTABLE_SV(hv), sv, (char *)sv_2mortal(keysv), HEf_SVKEY); |
b2c64049 | 493 | } else { |
ad64d0ec | 494 | mg_copy(MUTABLE_SV(hv), sv, key, klen); |
7f66fda2 | 495 | } |
b2c64049 NC |
496 | if (flags & HVhek_FREEKEY) |
497 | Safefree(key); | |
7f66fda2 NC |
498 | magic_existspack(svret, mg_find(sv, PERL_MAGIC_tiedelem)); |
499 | /* This cast somewhat evil, but I'm merely using NULL/ | |
500 | not NULL to return the boolean exists. | |
501 | And I know hv is not NULL. */ | |
3c84c864 | 502 | return SvTRUE(svret) ? (void *)hv : NULL; |
e7152ba2 | 503 | } |
7f66fda2 | 504 | #ifdef ENV_IS_CASELESS |
ad64d0ec | 505 | else if (mg_find((const SV *)hv, PERL_MAGIC_env)) { |
7f66fda2 | 506 | /* XXX This code isn't UTF8 clean. */ |
a15d23f8 | 507 | char * const keysave = (char * const)key; |
b2c64049 NC |
508 | /* Will need to free this, so set FREEKEY flag. */ |
509 | key = savepvn(key,klen); | |
510 | key = (const char*)strupr((char*)key); | |
6136c704 | 511 | is_utf8 = FALSE; |
7f66fda2 | 512 | hash = 0; |
8b4f7dd5 | 513 | keysv = 0; |
b2c64049 NC |
514 | |
515 | if (flags & HVhek_FREEKEY) { | |
516 | Safefree(keysave); | |
517 | } | |
518 | flags |= HVhek_FREEKEY; | |
7f66fda2 | 519 | } |
902173a3 | 520 | #endif |
7f66fda2 | 521 | } /* ISEXISTS */ |
b2c64049 NC |
522 | else if (action & HV_FETCH_ISSTORE) { |
523 | bool needs_copy; | |
524 | bool needs_store; | |
525 | hv_magic_check (hv, &needs_copy, &needs_store); | |
526 | if (needs_copy) { | |
a3b680e6 | 527 | const bool save_taint = PL_tainted; |
b2c64049 NC |
528 | if (keysv || is_utf8) { |
529 | if (!keysv) { | |
740cce10 | 530 | keysv = newSVpvn_utf8(key, klen, TRUE); |
b2c64049 NC |
531 | } |
532 | if (PL_tainting) | |
533 | PL_tainted = SvTAINTED(keysv); | |
534 | keysv = sv_2mortal(newSVsv(keysv)); | |
ad64d0ec | 535 | mg_copy(MUTABLE_SV(hv), val, (char*)keysv, HEf_SVKEY); |
b2c64049 | 536 | } else { |
ad64d0ec | 537 | mg_copy(MUTABLE_SV(hv), val, key, klen); |
b2c64049 NC |
538 | } |
539 | ||
540 | TAINT_IF(save_taint); | |
1baaf5d7 | 541 | if (!needs_store) { |
b2c64049 NC |
542 | if (flags & HVhek_FREEKEY) |
543 | Safefree(key); | |
4608196e | 544 | return NULL; |
b2c64049 NC |
545 | } |
546 | #ifdef ENV_IS_CASELESS | |
ad64d0ec | 547 | else if (mg_find((const SV *)hv, PERL_MAGIC_env)) { |
b2c64049 NC |
548 | /* XXX This code isn't UTF8 clean. */ |
549 | const char *keysave = key; | |
550 | /* Will need to free this, so set FREEKEY flag. */ | |
551 | key = savepvn(key,klen); | |
552 | key = (const char*)strupr((char*)key); | |
6136c704 | 553 | is_utf8 = FALSE; |
b2c64049 | 554 | hash = 0; |
8b4f7dd5 | 555 | keysv = 0; |
b2c64049 NC |
556 | |
557 | if (flags & HVhek_FREEKEY) { | |
558 | Safefree(keysave); | |
559 | } | |
560 | flags |= HVhek_FREEKEY; | |
561 | } | |
562 | #endif | |
563 | } | |
564 | } /* ISSTORE */ | |
7f66fda2 | 565 | } /* SvMAGICAL */ |
fde52b5c | 566 | |
7b2c381c | 567 | if (!HvARRAY(hv)) { |
b2c64049 | 568 | if ((action & (HV_FETCH_LVALUE | HV_FETCH_ISSTORE)) |
fde52b5c | 569 | #ifdef DYNAMIC_ENV_FETCH /* if it's an %ENV lookup, we may get it on the fly */ |
ad64d0ec NC |
570 | || (SvRMAGICAL((const SV *)hv) |
571 | && mg_find((const SV *)hv, PERL_MAGIC_env)) | |
fde52b5c | 572 | #endif |
d58e6666 NC |
573 | ) { |
574 | char *array; | |
a02a5408 | 575 | Newxz(array, |
cbec9347 | 576 | PERL_HV_ARRAY_ALLOC_BYTES(xhv->xhv_max+1 /* HvMAX(hv)+1 */), |
d58e6666 NC |
577 | char); |
578 | HvARRAY(hv) = (HE**)array; | |
579 | } | |
7f66fda2 NC |
580 | #ifdef DYNAMIC_ENV_FETCH |
581 | else if (action & HV_FETCH_ISEXISTS) { | |
582 | /* for an %ENV exists, if we do an insert it's by a recursive | |
583 | store call, so avoid creating HvARRAY(hv) right now. */ | |
584 | } | |
585 | #endif | |
113738bb NC |
586 | else { |
587 | /* XXX remove at some point? */ | |
588 | if (flags & HVhek_FREEKEY) | |
589 | Safefree(key); | |
590 | ||
3c84c864 | 591 | return NULL; |
113738bb | 592 | } |
fde52b5c | 593 | } |
594 | ||
44b87b50 | 595 | if (is_utf8 & !(flags & HVhek_KEYCANONICAL)) { |
41d88b63 | 596 | char * const keysave = (char *)key; |
f9a63242 | 597 | key = (char*)bytes_from_utf8((U8*)key, &klen, &is_utf8); |
19692e8d | 598 | if (is_utf8) |
c1fe5510 NC |
599 | flags |= HVhek_UTF8; |
600 | else | |
601 | flags &= ~HVhek_UTF8; | |
7f66fda2 NC |
602 | if (key != keysave) { |
603 | if (flags & HVhek_FREEKEY) | |
604 | Safefree(keysave); | |
19692e8d | 605 | flags |= HVhek_WASUTF8 | HVhek_FREEKEY; |
527df579 NC |
606 | /* If the caller calculated a hash, it was on the sequence of |
607 | octets that are the UTF-8 form. We've now changed the sequence | |
608 | of octets stored to that of the equivalent byte representation, | |
609 | so the hash we need is different. */ | |
610 | hash = 0; | |
7f66fda2 | 611 | } |
19692e8d | 612 | } |
f9a63242 | 613 | |
f8d50d94 DM |
614 | if (HvREHASH(hv) || (!hash && !(keysv && (SvIsCOW_shared_hash(keysv))))) |
615 | PERL_HASH_INTERNAL_(hash, key, klen, HvREHASH(hv)); | |
616 | else if (!hash) | |
617 | hash = SvSHARED_HASH(keysv); | |
618 | ||
619 | /* We don't have a pointer to the hv, so we have to replicate the | |
620 | flag into every HEK, so that hv_iterkeysv can see it. | |
621 | And yes, you do need this even though you are not "storing" because | |
622 | you can flip the flags below if doing an lval lookup. (And that | |
623 | was put in to give the semantics Andreas was expecting.) */ | |
624 | if (HvREHASH(hv)) | |
fdcd69b6 | 625 | flags |= HVhek_REHASH; |
effa1e2d | 626 | |
113738bb NC |
627 | masked_flags = (flags & HVhek_MASK); |
628 | ||
7f66fda2 | 629 | #ifdef DYNAMIC_ENV_FETCH |
4608196e | 630 | if (!HvARRAY(hv)) entry = NULL; |
7f66fda2 NC |
631 | else |
632 | #endif | |
b2c64049 | 633 | { |
7b2c381c | 634 | entry = (HvARRAY(hv))[hash & (I32) HvMAX(hv)]; |
b2c64049 | 635 | } |
0298d7b9 | 636 | for (; entry; entry = HeNEXT(entry)) { |
fde52b5c | 637 | if (HeHASH(entry) != hash) /* strings can't be equal */ |
638 | continue; | |
eb160463 | 639 | if (HeKLEN(entry) != (I32)klen) |
fde52b5c | 640 | continue; |
1c846c1f | 641 | if (HeKEY(entry) != key && memNE(HeKEY(entry),key,klen)) /* is this it? */ |
fde52b5c | 642 | continue; |
113738bb | 643 | if ((HeKFLAGS(entry) ^ masked_flags) & HVhek_UTF8) |
c3654f1a | 644 | continue; |
b2c64049 NC |
645 | |
646 | if (action & (HV_FETCH_LVALUE|HV_FETCH_ISSTORE)) { | |
647 | if (HeKFLAGS(entry) != masked_flags) { | |
648 | /* We match if HVhek_UTF8 bit in our flags and hash key's | |
649 | match. But if entry was set previously with HVhek_WASUTF8 | |
650 | and key now doesn't (or vice versa) then we should change | |
651 | the key's flag, as this is assignment. */ | |
652 | if (HvSHAREKEYS(hv)) { | |
653 | /* Need to swap the key we have for a key with the flags we | |
654 | need. As keys are shared we can't just write to the | |
655 | flag, so we share the new one, unshare the old one. */ | |
6136c704 | 656 | HEK * const new_hek = share_hek_flags(key, klen, hash, |
6e838c70 | 657 | masked_flags); |
b2c64049 NC |
658 | unshare_hek (HeKEY_hek(entry)); |
659 | HeKEY_hek(entry) = new_hek; | |
660 | } | |
5d2b1485 NC |
661 | else if (hv == PL_strtab) { |
662 | /* PL_strtab is usually the only hash without HvSHAREKEYS, | |
663 | so putting this test here is cheap */ | |
664 | if (flags & HVhek_FREEKEY) | |
665 | Safefree(key); | |
666 | Perl_croak(aTHX_ S_strtab_error, | |
667 | action & HV_FETCH_LVALUE ? "fetch" : "store"); | |
668 | } | |
b2c64049 NC |
669 | else |
670 | HeKFLAGS(entry) = masked_flags; | |
671 | if (masked_flags & HVhek_ENABLEHVKFLAGS) | |
672 | HvHASKFLAGS_on(hv); | |
673 | } | |
674 | if (HeVAL(entry) == &PL_sv_placeholder) { | |
675 | /* yes, can store into placeholder slot */ | |
676 | if (action & HV_FETCH_LVALUE) { | |
677 | if (SvMAGICAL(hv)) { | |
678 | /* This preserves behaviour with the old hv_fetch | |
679 | implementation which at this point would bail out | |
680 | with a break; (at "if we find a placeholder, we | |
681 | pretend we haven't found anything") | |
682 | ||
683 | That break mean that if a placeholder were found, it | |
684 | caused a call into hv_store, which in turn would | |
685 | check magic, and if there is no magic end up pretty | |
686 | much back at this point (in hv_store's code). */ | |
687 | break; | |
688 | } | |
486ec47a | 689 | /* LVAL fetch which actually needs a store. */ |
561b68a9 | 690 | val = newSV(0); |
ca732855 | 691 | HvPLACEHOLDERS(hv)--; |
b2c64049 NC |
692 | } else { |
693 | /* store */ | |
694 | if (val != &PL_sv_placeholder) | |
ca732855 | 695 | HvPLACEHOLDERS(hv)--; |
b2c64049 NC |
696 | } |
697 | HeVAL(entry) = val; | |
698 | } else if (action & HV_FETCH_ISSTORE) { | |
cefd5c7c | 699 | SvREFCNT_dec(HeVAL(entry)); |
b2c64049 NC |
700 | HeVAL(entry) = val; |
701 | } | |
27bcc0a7 | 702 | } else if (HeVAL(entry) == &PL_sv_placeholder) { |
b2c64049 NC |
703 | /* if we find a placeholder, we pretend we haven't found |
704 | anything */ | |
8aacddc1 | 705 | break; |
b2c64049 | 706 | } |
113738bb NC |
707 | if (flags & HVhek_FREEKEY) |
708 | Safefree(key); | |
3c84c864 NC |
709 | if (return_svp) { |
710 | return entry ? (void *) &HeVAL(entry) : NULL; | |
711 | } | |
fde52b5c | 712 | return entry; |
713 | } | |
714 | #ifdef DYNAMIC_ENV_FETCH /* %ENV lookup? If so, try to fetch the value now */ | |
0ed29950 | 715 | if (!(action & HV_FETCH_ISSTORE) |
ad64d0ec NC |
716 | && SvRMAGICAL((const SV *)hv) |
717 | && mg_find((const SV *)hv, PERL_MAGIC_env)) { | |
a6c40364 | 718 | unsigned long len; |
9d4ba2ae | 719 | const char * const env = PerlEnv_ENVgetenv_len(key,&len); |
a6c40364 GS |
720 | if (env) { |
721 | sv = newSVpvn(env,len); | |
722 | SvTAINTED_on(sv); | |
d3ba3f5c | 723 | return hv_common(hv, keysv, key, klen, flags, |
3c84c864 NC |
724 | HV_FETCH_ISSTORE|HV_DISABLE_UVAR_XKEY|return_svp, |
725 | sv, hash); | |
a6c40364 | 726 | } |
fde52b5c | 727 | } |
728 | #endif | |
7f66fda2 NC |
729 | |
730 | if (!entry && SvREADONLY(hv) && !(action & HV_FETCH_ISEXISTS)) { | |
c445ea15 | 731 | hv_notallowed(flags, key, klen, |
c8cd6465 NC |
732 | "Attempt to access disallowed key '%"SVf"' in" |
733 | " a restricted hash"); | |
1b1f1335 | 734 | } |
b2c64049 NC |
735 | if (!(action & (HV_FETCH_LVALUE|HV_FETCH_ISSTORE))) { |
736 | /* Not doing some form of store, so return failure. */ | |
737 | if (flags & HVhek_FREEKEY) | |
738 | Safefree(key); | |
3c84c864 | 739 | return NULL; |
b2c64049 | 740 | } |
113738bb | 741 | if (action & HV_FETCH_LVALUE) { |
df5f182b | 742 | val = action & HV_FETCH_EMPTY_HE ? NULL : newSV(0); |
b2c64049 NC |
743 | if (SvMAGICAL(hv)) { |
744 | /* At this point the old hv_fetch code would call to hv_store, | |
745 | which in turn might do some tied magic. So we need to make that | |
746 | magic check happen. */ | |
747 | /* gonna assign to this, so it better be there */ | |
fda2d18a NC |
748 | /* If a fetch-as-store fails on the fetch, then the action is to |
749 | recurse once into "hv_store". If we didn't do this, then that | |
750 | recursive call would call the key conversion routine again. | |
751 | However, as we replace the original key with the converted | |
752 | key, this would result in a double conversion, which would show | |
753 | up as a bug if the conversion routine is not idempotent. */ | |
d3ba3f5c | 754 | return hv_common(hv, keysv, key, klen, flags, |
3c84c864 NC |
755 | HV_FETCH_ISSTORE|HV_DISABLE_UVAR_XKEY|return_svp, |
756 | val, hash); | |
b2c64049 NC |
757 | /* XXX Surely that could leak if the fetch-was-store fails? |
758 | Just like the hv_fetch. */ | |
113738bb NC |
759 | } |
760 | } | |
761 | ||
b2c64049 NC |
762 | /* Welcome to hv_store... */ |
763 | ||
7b2c381c | 764 | if (!HvARRAY(hv)) { |
b2c64049 NC |
765 | /* Not sure if we can get here. I think the only case of oentry being |
766 | NULL is for %ENV with dynamic env fetch. But that should disappear | |
767 | with magic in the previous code. */ | |
d58e6666 | 768 | char *array; |
a02a5408 | 769 | Newxz(array, |
b2c64049 | 770 | PERL_HV_ARRAY_ALLOC_BYTES(xhv->xhv_max+1 /* HvMAX(hv)+1 */), |
d58e6666 NC |
771 | char); |
772 | HvARRAY(hv) = (HE**)array; | |
b2c64049 NC |
773 | } |
774 | ||
7b2c381c | 775 | oentry = &(HvARRAY(hv))[hash & (I32) xhv->xhv_max]; |
ab4af705 | 776 | |
b2c64049 NC |
777 | entry = new_HE(); |
778 | /* share_hek_flags will do the free for us. This might be considered | |
779 | bad API design. */ | |
780 | if (HvSHAREKEYS(hv)) | |
6e838c70 | 781 | HeKEY_hek(entry) = share_hek_flags(key, klen, hash, flags); |
5d2b1485 NC |
782 | else if (hv == PL_strtab) { |
783 | /* PL_strtab is usually the only hash without HvSHAREKEYS, so putting | |
784 | this test here is cheap */ | |
785 | if (flags & HVhek_FREEKEY) | |
786 | Safefree(key); | |
787 | Perl_croak(aTHX_ S_strtab_error, | |
788 | action & HV_FETCH_LVALUE ? "fetch" : "store"); | |
789 | } | |
b2c64049 NC |
790 | else /* gotta do the real thing */ |
791 | HeKEY_hek(entry) = save_hek_flags(key, klen, hash, flags); | |
792 | HeVAL(entry) = val; | |
793 | HeNEXT(entry) = *oentry; | |
794 | *oentry = entry; | |
795 | ||
796 | if (val == &PL_sv_placeholder) | |
ca732855 | 797 | HvPLACEHOLDERS(hv)++; |
b2c64049 NC |
798 | if (masked_flags & HVhek_ENABLEHVKFLAGS) |
799 | HvHASKFLAGS_on(hv); | |
800 | ||
0298d7b9 NC |
801 | { |
802 | const HE *counter = HeNEXT(entry); | |
803 | ||
4c7185a0 | 804 | xhv->xhv_keys++; /* HvTOTALKEYS(hv)++ */ |
0298d7b9 | 805 | if (!counter) { /* initial entry? */ |
5ac36297 | 806 | } else if (xhv->xhv_keys > xhv->xhv_max) { |
1b95d04f | 807 | /* Use only the old HvUSEDKEYS(hv) > HvMAX(hv) condition to limit |
5b430b36 MW |
808 | bucket splits on a rehashed hash, as we're not going to |
809 | split it again, and if someone is lucky (evil) enough to | |
810 | get all the keys in one list they could exhaust our memory | |
811 | as we repeatedly double the number of buckets on every | |
812 | entry. Linear search feels a less worse thing to do. */ | |
0298d7b9 NC |
813 | hsplit(hv); |
814 | } else if(!HvREHASH(hv)) { | |
815 | U32 n_links = 1; | |
816 | ||
817 | while ((counter = HeNEXT(counter))) | |
818 | n_links++; | |
819 | ||
820 | if (n_links > HV_MAX_LENGTH_BEFORE_SPLIT) { | |
0298d7b9 NC |
821 | hsplit(hv); |
822 | } | |
823 | } | |
fde52b5c | 824 | } |
b2c64049 | 825 | |
3c84c864 NC |
826 | if (return_svp) { |
827 | return entry ? (void *) &HeVAL(entry) : NULL; | |
828 | } | |
829 | return (void *) entry; | |
fde52b5c | 830 | } |
831 | ||
864dbfa3 | 832 | STATIC void |
b0e6ae5b | 833 | S_hv_magic_check(HV *hv, bool *needs_copy, bool *needs_store) |
d0066dc7 | 834 | { |
a3b680e6 | 835 | const MAGIC *mg = SvMAGIC(hv); |
7918f24d NC |
836 | |
837 | PERL_ARGS_ASSERT_HV_MAGIC_CHECK; | |
838 | ||
d0066dc7 OT |
839 | *needs_copy = FALSE; |
840 | *needs_store = TRUE; | |
841 | while (mg) { | |
842 | if (isUPPER(mg->mg_type)) { | |
843 | *needs_copy = TRUE; | |
d60c5a05 | 844 | if (mg->mg_type == PERL_MAGIC_tied) { |
d0066dc7 | 845 | *needs_store = FALSE; |
4ab2a30b | 846 | return; /* We've set all there is to set. */ |
d0066dc7 OT |
847 | } |
848 | } | |
849 | mg = mg->mg_moremagic; | |
850 | } | |
851 | } | |
852 | ||
954c1994 | 853 | /* |
a3bcc51e TP |
854 | =for apidoc hv_scalar |
855 | ||
856 | Evaluates the hash in scalar context and returns the result. Handles magic when the hash is tied. | |
857 | ||
858 | =cut | |
859 | */ | |
860 | ||
861 | SV * | |
862 | Perl_hv_scalar(pTHX_ HV *hv) | |
863 | { | |
a3bcc51e | 864 | SV *sv; |
823a54a3 | 865 | |
7918f24d NC |
866 | PERL_ARGS_ASSERT_HV_SCALAR; |
867 | ||
823a54a3 | 868 | if (SvRMAGICAL(hv)) { |
ad64d0ec | 869 | MAGIC * const mg = mg_find((const SV *)hv, PERL_MAGIC_tied); |
823a54a3 AL |
870 | if (mg) |
871 | return magic_scalarpack(hv, mg); | |
872 | } | |
a3bcc51e TP |
873 | |
874 | sv = sv_newmortal(); | |
f4431c56 | 875 | if (HvTOTALKEYS((const HV *)hv)) |
a3bcc51e TP |
876 | Perl_sv_setpvf(aTHX_ sv, "%ld/%ld", |
877 | (long)HvFILL(hv), (long)HvMAX(hv) + 1); | |
878 | else | |
879 | sv_setiv(sv, 0); | |
880 | ||
881 | return sv; | |
882 | } | |
883 | ||
884 | /* | |
954c1994 GS |
885 | =for apidoc hv_delete |
886 | ||
a05d6c5d TC |
887 | Deletes a key/value pair in the hash. The value's SV is removed from |
888 | the hash, made mortal, and returned to the caller. The absolute | |
889 | value of C<klen> is the length of the key. If C<klen> is negative the | |
890 | key is assumed to be in UTF-8-encoded Unicode. The C<flags> value | |
891 | will normally be zero; if set to G_DISCARD then NULL will be returned. | |
892 | NULL will also be returned if the key is not found. | |
954c1994 | 893 | |
954c1994 GS |
894 | =for apidoc hv_delete_ent |
895 | ||
3025a2e4 CS |
896 | Deletes a key/value pair in the hash. The value SV is removed from the hash, |
897 | made mortal, and returned to the caller. The C<flags> value will normally be | |
898 | zero; if set to G_DISCARD then NULL will be returned. NULL will also be | |
899 | returned if the key is not found. C<hash> can be a valid precomputed hash | |
900 | value, or 0 to ask for it to be computed. | |
954c1994 GS |
901 | |
902 | =cut | |
903 | */ | |
904 | ||
8f8d40ab | 905 | STATIC SV * |
cd6d36ac NC |
906 | S_hv_delete_common(pTHX_ HV *hv, SV *keysv, const char *key, STRLEN klen, |
907 | int k_flags, I32 d_flags, U32 hash) | |
f1317c8d | 908 | { |
27da23d5 | 909 | dVAR; |
cbec9347 | 910 | register XPVHV* xhv; |
fde52b5c | 911 | register HE *entry; |
912 | register HE **oentry; | |
9dbc5603 | 913 | bool is_utf8 = (k_flags & HVhek_UTF8) ? TRUE : FALSE; |
7a9669ca | 914 | int masked_flags; |
1c846c1f | 915 | |
fde52b5c | 916 | if (SvRMAGICAL(hv)) { |
0a0bb7c7 OT |
917 | bool needs_copy; |
918 | bool needs_store; | |
919 | hv_magic_check (hv, &needs_copy, &needs_store); | |
920 | ||
f1317c8d | 921 | if (needs_copy) { |
6136c704 | 922 | SV *sv; |
63c89345 NC |
923 | entry = (HE *) hv_common(hv, keysv, key, klen, |
924 | k_flags & ~HVhek_FREEKEY, | |
925 | HV_FETCH_LVALUE|HV_DISABLE_UVAR_XKEY, | |
926 | NULL, hash); | |
7a9669ca | 927 | sv = entry ? HeVAL(entry) : NULL; |
f1317c8d NC |
928 | if (sv) { |
929 | if (SvMAGICAL(sv)) { | |
930 | mg_clear(sv); | |
931 | } | |
932 | if (!needs_store) { | |
933 | if (mg_find(sv, PERL_MAGIC_tiedelem)) { | |
934 | /* No longer an element */ | |
935 | sv_unmagic(sv, PERL_MAGIC_tiedelem); | |
936 | return sv; | |
937 | } | |
a0714e2c | 938 | return NULL; /* element cannot be deleted */ |
f1317c8d | 939 | } |
902173a3 | 940 | #ifdef ENV_IS_CASELESS |
ad64d0ec | 941 | else if (mg_find((const SV *)hv, PERL_MAGIC_env)) { |
8167a60a | 942 | /* XXX This code isn't UTF8 clean. */ |
59cd0e26 | 943 | keysv = newSVpvn_flags(key, klen, SVs_TEMP); |
8167a60a NC |
944 | if (k_flags & HVhek_FREEKEY) { |
945 | Safefree(key); | |
946 | } | |
947 | key = strupr(SvPVX(keysv)); | |
948 | is_utf8 = 0; | |
949 | k_flags = 0; | |
950 | hash = 0; | |
7f66fda2 | 951 | } |
510ac311 | 952 | #endif |
2fd1c6b8 | 953 | } |
2fd1c6b8 | 954 | } |
fde52b5c | 955 | } |
cbec9347 | 956 | xhv = (XPVHV*)SvANY(hv); |
7b2c381c | 957 | if (!HvARRAY(hv)) |
a0714e2c | 958 | return NULL; |
fde52b5c | 959 | |
19692e8d | 960 | if (is_utf8) { |
c445ea15 | 961 | const char * const keysave = key; |
b464bac0 | 962 | key = (char*)bytes_from_utf8((U8*)key, &klen, &is_utf8); |
cd6d36ac | 963 | |
19692e8d | 964 | if (is_utf8) |
cd6d36ac NC |
965 | k_flags |= HVhek_UTF8; |
966 | else | |
967 | k_flags &= ~HVhek_UTF8; | |
7f66fda2 NC |
968 | if (key != keysave) { |
969 | if (k_flags & HVhek_FREEKEY) { | |
970 | /* This shouldn't happen if our caller does what we expect, | |
971 | but strictly the API allows it. */ | |
972 | Safefree(keysave); | |
973 | } | |
974 | k_flags |= HVhek_WASUTF8 | HVhek_FREEKEY; | |
975 | } | |
ad64d0ec | 976 | HvHASKFLAGS_on(MUTABLE_SV(hv)); |
19692e8d | 977 | } |
f9a63242 | 978 | |
f8d50d94 DM |
979 | if (HvREHASH(hv) || (!hash && !(keysv && (SvIsCOW_shared_hash(keysv))))) |
980 | PERL_HASH_INTERNAL_(hash, key, klen, HvREHASH(hv)); | |
981 | else if (!hash) | |
982 | hash = SvSHARED_HASH(keysv); | |
fde52b5c | 983 | |
7a9669ca NC |
984 | masked_flags = (k_flags & HVhek_MASK); |
985 | ||
9de10d5c | 986 | oentry = &(HvARRAY(hv))[hash & (I32) HvMAX(hv)]; |
fde52b5c | 987 | entry = *oentry; |
9e720f71 | 988 | for (; entry; oentry = &HeNEXT(entry), entry = *oentry) { |
6136c704 | 989 | SV *sv; |
f3d2f32d | 990 | U8 mro_changes = 0; /* 1 = isa; 2 = package moved */ |
0290c710 | 991 | GV *gv = NULL; |
0c3bb3c2 FC |
992 | HV *stash = NULL; |
993 | ||
fde52b5c | 994 | if (HeHASH(entry) != hash) /* strings can't be equal */ |
995 | continue; | |
eb160463 | 996 | if (HeKLEN(entry) != (I32)klen) |
fde52b5c | 997 | continue; |
1c846c1f | 998 | if (HeKEY(entry) != key && memNE(HeKEY(entry),key,klen)) /* is this it? */ |
fde52b5c | 999 | continue; |
7a9669ca | 1000 | if ((HeKFLAGS(entry) ^ masked_flags) & HVhek_UTF8) |
c3654f1a | 1001 | continue; |
8aacddc1 | 1002 | |
5d2b1485 NC |
1003 | if (hv == PL_strtab) { |
1004 | if (k_flags & HVhek_FREEKEY) | |
1005 | Safefree(key); | |
1006 | Perl_croak(aTHX_ S_strtab_error, "delete"); | |
1007 | } | |
1008 | ||
8aacddc1 | 1009 | /* if placeholder is here, it's already been deleted.... */ |
6136c704 AL |
1010 | if (HeVAL(entry) == &PL_sv_placeholder) { |
1011 | if (k_flags & HVhek_FREEKEY) | |
1012 | Safefree(key); | |
1013 | return NULL; | |
8aacddc1 | 1014 | } |
e5accad2 FC |
1015 | if (SvREADONLY(hv) && HeVAL(entry) && SvREADONLY(HeVAL(entry)) |
1016 | && !SvIsCOW(HeVAL(entry))) { | |
d4c19fe8 | 1017 | hv_notallowed(k_flags, key, klen, |
c8cd6465 NC |
1018 | "Attempt to delete readonly key '%"SVf"' from" |
1019 | " a restricted hash"); | |
8aacddc1 | 1020 | } |
b84d0860 NC |
1021 | if (k_flags & HVhek_FREEKEY) |
1022 | Safefree(key); | |
8aacddc1 | 1023 | |
35759254 | 1024 | /* If this is a stash and the key ends with ::, then someone is |
0c3bb3c2 | 1025 | * deleting a package. |
0c3bb3c2 | 1026 | */ |
78b79c77 | 1027 | if (HeVAL(entry) && HvENAME_get(hv)) { |
0290c710 | 1028 | gv = (GV *)HeVAL(entry); |
35759254 | 1029 | if (keysv) key = SvPV(keysv, klen); |
1f656fcf FC |
1030 | if (( |
1031 | (klen > 1 && key[klen-2] == ':' && key[klen-1] == ':') | |
1032 | || | |
1033 | (klen == 1 && key[0] == ':') | |
1034 | ) | |
e0a52395 | 1035 | && (klen != 6 || hv!=PL_defstash || memNE(key,"main::",6)) |
0290c710 | 1036 | && SvTYPE(gv) == SVt_PVGV && (stash = GvHV((GV *)gv)) |
0c3bb3c2 | 1037 | && HvENAME_get(stash)) { |
0290c710 FC |
1038 | /* A previous version of this code checked that the |
1039 | * GV was still in the symbol table by fetching the | |
1040 | * GV with its name. That is not necessary (and | |
1041 | * sometimes incorrect), as HvENAME cannot be set | |
1042 | * on hv if it is not in the symtab. */ | |
f3d2f32d | 1043 | mro_changes = 2; |
0c3bb3c2 FC |
1044 | /* Hang on to it for a bit. */ |
1045 | SvREFCNT_inc_simple_void_NN( | |
0290c710 | 1046 | sv_2mortal((SV *)gv) |
35759254 FC |
1047 | ); |
1048 | } | |
f3d2f32d FC |
1049 | else if (klen == 3 && strnEQ(key, "ISA", 3)) |
1050 | mro_changes = 1; | |
35759254 FC |
1051 | } |
1052 | ||
8571a3cc FC |
1053 | sv = d_flags & G_DISCARD ? HeVAL(entry) : sv_2mortal(HeVAL(entry)); |
1054 | HeVAL(entry) = &PL_sv_placeholder; | |
5743f2a3 FC |
1055 | if (sv) { |
1056 | /* deletion of method from stash */ | |
1057 | if (isGV(sv) && isGV_with_GP(sv) && GvCVu(sv) | |
1058 | && HvENAME_get(hv)) | |
1059 | mro_method_changed_in(hv); | |
5743f2a3 | 1060 | } |
8aacddc1 NIS |
1061 | |
1062 | /* | |
1063 | * If a restricted hash, rather than really deleting the entry, put | |
1064 | * a placeholder there. This marks the key as being "approved", so | |
1065 | * we can still access via not-really-existing key without raising | |
1066 | * an error. | |
1067 | */ | |
f50383f5 | 1068 | if (SvREADONLY(hv)) |
8aacddc1 NIS |
1069 | /* We'll be saving this slot, so the number of allocated keys |
1070 | * doesn't go down, but the number placeholders goes up */ | |
ca732855 | 1071 | HvPLACEHOLDERS(hv)++; |
f50383f5 | 1072 | else { |
a26e96df | 1073 | *oentry = HeNEXT(entry); |
b79f7545 | 1074 | if (SvOOK(hv) && entry == HvAUX(hv)->xhv_eiter /* HvEITER(hv) */) |
8aacddc1 | 1075 | HvLAZYDEL_on(hv); |
ae199939 TH |
1076 | else { |
1077 | if (SvOOK(hv) && HvLAZYDEL(hv) && | |
1078 | entry == HeNEXT(HvAUX(hv)->xhv_eiter)) | |
1079 | HeNEXT(HvAUX(hv)->xhv_eiter) = HeNEXT(entry); | |
8aacddc1 | 1080 | hv_free_ent(hv, entry); |
ae199939 | 1081 | } |
4c7185a0 | 1082 | xhv->xhv_keys--; /* HvTOTALKEYS(hv)-- */ |
574c8022 | 1083 | if (xhv->xhv_keys == 0) |
19692e8d | 1084 | HvHASKFLAGS_off(hv); |
8aacddc1 | 1085 | } |
0c3bb3c2 | 1086 | |
3b2cd809 FC |
1087 | if (d_flags & G_DISCARD) { |
1088 | SvREFCNT_dec(sv); | |
1089 | sv = NULL; | |
1090 | } | |
1091 | ||
f3d2f32d FC |
1092 | if (mro_changes == 1) mro_isa_changed_in(hv); |
1093 | else if (mro_changes == 2) | |
afdbe55d | 1094 | mro_package_moved(NULL, stash, gv, 1); |
0c3bb3c2 | 1095 | |
79072805 LW |
1096 | return sv; |
1097 | } | |
8aacddc1 | 1098 | if (SvREADONLY(hv)) { |
d4c19fe8 | 1099 | hv_notallowed(k_flags, key, klen, |
c8cd6465 NC |
1100 | "Attempt to delete disallowed key '%"SVf"' from" |
1101 | " a restricted hash"); | |
8aacddc1 NIS |
1102 | } |
1103 | ||
19692e8d | 1104 | if (k_flags & HVhek_FREEKEY) |
f9a63242 | 1105 | Safefree(key); |
a0714e2c | 1106 | return NULL; |
79072805 LW |
1107 | } |
1108 | ||
76e3520e | 1109 | STATIC void |
cea2e8a9 | 1110 | S_hsplit(pTHX_ HV *hv) |
79072805 | 1111 | { |
97aff369 | 1112 | dVAR; |
1e05feb3 | 1113 | register XPVHV* const xhv = (XPVHV*)SvANY(hv); |
a3b680e6 | 1114 | const I32 oldsize = (I32) xhv->xhv_max+1; /* HvMAX(hv)+1 (sick) */ |
79072805 LW |
1115 | register I32 newsize = oldsize * 2; |
1116 | register I32 i; | |
7b2c381c | 1117 | char *a = (char*) HvARRAY(hv); |
72311751 | 1118 | register HE **aep; |
4b5190b5 NC |
1119 | int longest_chain = 0; |
1120 | int was_shared; | |
79072805 | 1121 | |
7918f24d NC |
1122 | PERL_ARGS_ASSERT_HSPLIT; |
1123 | ||
18026298 | 1124 | /*PerlIO_printf(PerlIO_stderr(), "hsplit called for %p which had %d\n", |
6c9570dc | 1125 | (void*)hv, (int) oldsize);*/ |
18026298 | 1126 | |
5d88ecd7 | 1127 | if (HvPLACEHOLDERS_get(hv) && !SvREADONLY(hv)) { |
18026298 NC |
1128 | /* Can make this clear any placeholders first for non-restricted hashes, |
1129 | even though Storable rebuilds restricted hashes by putting in all the | |
1130 | placeholders (first) before turning on the readonly flag, because | |
1131 | Storable always pre-splits the hash. */ | |
1132 | hv_clear_placeholders(hv); | |
1133 | } | |
1134 | ||
3280af22 | 1135 | PL_nomemok = TRUE; |
8d6dde3e | 1136 | #if defined(STRANGE_MALLOC) || defined(MYMALLOC) |
b79f7545 NC |
1137 | Renew(a, PERL_HV_ARRAY_ALLOC_BYTES(newsize) |
1138 | + (SvOOK(hv) ? sizeof(struct xpvhv_aux) : 0), char); | |
422a93e5 | 1139 | if (!a) { |
4a33f861 | 1140 | PL_nomemok = FALSE; |
422a93e5 GA |
1141 | return; |
1142 | } | |
b79f7545 | 1143 | if (SvOOK(hv)) { |
31f0e52e | 1144 | Move(&a[oldsize * sizeof(HE*)], &a[newsize * sizeof(HE*)], 1, struct xpvhv_aux); |
b79f7545 | 1145 | } |
4633a7c4 | 1146 | #else |
a02a5408 | 1147 | Newx(a, PERL_HV_ARRAY_ALLOC_BYTES(newsize) |
b79f7545 | 1148 | + (SvOOK(hv) ? sizeof(struct xpvhv_aux) : 0), char); |
422a93e5 | 1149 | if (!a) { |
3280af22 | 1150 | PL_nomemok = FALSE; |
422a93e5 GA |
1151 | return; |
1152 | } | |
7b2c381c | 1153 | Copy(HvARRAY(hv), a, oldsize * sizeof(HE*), char); |
b79f7545 NC |
1154 | if (SvOOK(hv)) { |
1155 | Copy(HvAUX(hv), &a[newsize * sizeof(HE*)], 1, struct xpvhv_aux); | |
1156 | } | |
9a87bd09 | 1157 | Safefree(HvARRAY(hv)); |
4633a7c4 LW |
1158 | #endif |
1159 | ||
3280af22 | 1160 | PL_nomemok = FALSE; |
72311751 | 1161 | Zero(&a[oldsize * sizeof(HE*)], (newsize-oldsize) * sizeof(HE*), char); /* zero 2nd half*/ |
cbec9347 | 1162 | xhv->xhv_max = --newsize; /* HvMAX(hv) = --newsize */ |
7b2c381c | 1163 | HvARRAY(hv) = (HE**) a; |
72311751 | 1164 | aep = (HE**)a; |
79072805 | 1165 | |
72311751 | 1166 | for (i=0; i<oldsize; i++,aep++) { |
4b5190b5 NC |
1167 | int left_length = 0; |
1168 | int right_length = 0; | |
a50a3493 NC |
1169 | HE **oentry = aep; |
1170 | HE *entry = *aep; | |
a3b680e6 | 1171 | register HE **bep; |
4b5190b5 | 1172 | |
a50a3493 | 1173 | if (!entry) /* non-existent */ |
79072805 | 1174 | continue; |
72311751 | 1175 | bep = aep+oldsize; |
4c9d89c5 | 1176 | do { |
eb160463 | 1177 | if ((HeHASH(entry) & newsize) != (U32)i) { |
fde52b5c | 1178 | *oentry = HeNEXT(entry); |
72311751 | 1179 | HeNEXT(entry) = *bep; |
72311751 | 1180 | *bep = entry; |
4b5190b5 | 1181 | right_length++; |
79072805 | 1182 | } |
4b5190b5 | 1183 | else { |
fde52b5c | 1184 | oentry = &HeNEXT(entry); |
4b5190b5 NC |
1185 | left_length++; |
1186 | } | |
4c9d89c5 NC |
1187 | entry = *oentry; |
1188 | } while (entry); | |
4b5190b5 NC |
1189 | /* I think we don't actually need to keep track of the longest length, |
1190 | merely flag if anything is too long. But for the moment while | |
1191 | developing this code I'll track it. */ | |
1192 | if (left_length > longest_chain) | |
1193 | longest_chain = left_length; | |
1194 | if (right_length > longest_chain) | |
1195 | longest_chain = right_length; | |
1196 | } | |
1197 | ||
1198 | ||
1199 | /* Pick your policy for "hashing isn't working" here: */ | |
fdcd69b6 | 1200 | if (longest_chain <= HV_MAX_LENGTH_BEFORE_SPLIT /* split worked? */ |
4b5190b5 NC |
1201 | || HvREHASH(hv)) { |
1202 | return; | |
79072805 | 1203 | } |
4b5190b5 NC |
1204 | |
1205 | if (hv == PL_strtab) { | |
1206 | /* Urg. Someone is doing something nasty to the string table. | |
1207 | Can't win. */ | |
1208 | return; | |
1209 | } | |
1210 | ||
1211 | /* Awooga. Awooga. Pathological data. */ | |
6c9570dc | 1212 | /*PerlIO_printf(PerlIO_stderr(), "%p %d of %d with %d/%d buckets\n", (void*)hv, |
4b5190b5 NC |
1213 | longest_chain, HvTOTALKEYS(hv), HvFILL(hv), 1+HvMAX(hv));*/ |
1214 | ||
1215 | ++newsize; | |
a02a5408 | 1216 | Newxz(a, PERL_HV_ARRAY_ALLOC_BYTES(newsize) |
b79f7545 NC |
1217 | + (SvOOK(hv) ? sizeof(struct xpvhv_aux) : 0), char); |
1218 | if (SvOOK(hv)) { | |
1219 | Copy(HvAUX(hv), &a[newsize * sizeof(HE*)], 1, struct xpvhv_aux); | |
1220 | } | |
1221 | ||
4b5190b5 NC |
1222 | was_shared = HvSHAREKEYS(hv); |
1223 | ||
4b5190b5 NC |
1224 | HvSHAREKEYS_off(hv); |
1225 | HvREHASH_on(hv); | |
1226 | ||
7b2c381c | 1227 | aep = HvARRAY(hv); |
4b5190b5 NC |
1228 | |
1229 | for (i=0; i<newsize; i++,aep++) { | |
a3b680e6 | 1230 | register HE *entry = *aep; |
4b5190b5 NC |
1231 | while (entry) { |
1232 | /* We're going to trash this HE's next pointer when we chain it | |
1233 | into the new hash below, so store where we go next. */ | |
9d4ba2ae | 1234 | HE * const next = HeNEXT(entry); |
4b5190b5 | 1235 | UV hash; |
a3b680e6 | 1236 | HE **bep; |
4b5190b5 NC |
1237 | |
1238 | /* Rehash it */ | |
1239 | PERL_HASH_INTERNAL(hash, HeKEY(entry), HeKLEN(entry)); | |
1240 | ||
1241 | if (was_shared) { | |
1242 | /* Unshare it. */ | |
aec46f14 | 1243 | HEK * const new_hek |
4b5190b5 NC |
1244 | = save_hek_flags(HeKEY(entry), HeKLEN(entry), |
1245 | hash, HeKFLAGS(entry)); | |
1246 | unshare_hek (HeKEY_hek(entry)); | |
1247 | HeKEY_hek(entry) = new_hek; | |
1248 | } else { | |
1249 | /* Not shared, so simply write the new hash in. */ | |
1250 | HeHASH(entry) = hash; | |
1251 | } | |
1252 | /*PerlIO_printf(PerlIO_stderr(), "%d ", HeKFLAGS(entry));*/ | |
1253 | HEK_REHASH_on(HeKEY_hek(entry)); | |
1254 | /*PerlIO_printf(PerlIO_stderr(), "%d\n", HeKFLAGS(entry));*/ | |
1255 | ||
1256 | /* Copy oentry to the correct new chain. */ | |
1257 | bep = ((HE**)a) + (hash & (I32) xhv->xhv_max); | |
4b5190b5 NC |
1258 | HeNEXT(entry) = *bep; |
1259 | *bep = entry; | |
1260 | ||
1261 | entry = next; | |
1262 | } | |
1263 | } | |
7b2c381c NC |
1264 | Safefree (HvARRAY(hv)); |
1265 | HvARRAY(hv) = (HE **)a; | |
79072805 LW |
1266 | } |
1267 | ||
72940dca | 1268 | void |
864dbfa3 | 1269 | Perl_hv_ksplit(pTHX_ HV *hv, IV newmax) |
72940dca | 1270 | { |
97aff369 | 1271 | dVAR; |
cbec9347 | 1272 | register XPVHV* xhv = (XPVHV*)SvANY(hv); |
a3b680e6 | 1273 | const I32 oldsize = (I32) xhv->xhv_max+1; /* HvMAX(hv)+1 (sick) */ |
72940dca | 1274 | register I32 newsize; |
1275 | register I32 i; | |
72311751 GS |
1276 | register char *a; |
1277 | register HE **aep; | |
72940dca | 1278 | |
7918f24d NC |
1279 | PERL_ARGS_ASSERT_HV_KSPLIT; |
1280 | ||
72940dca | 1281 | newsize = (I32) newmax; /* possible truncation here */ |
1282 | if (newsize != newmax || newmax <= oldsize) | |
1283 | return; | |
1284 | while ((newsize & (1 + ~newsize)) != newsize) { | |
1285 | newsize &= ~(newsize & (1 + ~newsize)); /* get proper power of 2 */ | |
1286 | } | |
1287 | if (newsize < newmax) | |
1288 | newsize *= 2; | |
1289 | if (newsize < newmax) | |
1290 | return; /* overflow detection */ | |
1291 | ||
7b2c381c | 1292 | a = (char *) HvARRAY(hv); |
72940dca | 1293 | if (a) { |
3280af22 | 1294 | PL_nomemok = TRUE; |
8d6dde3e | 1295 | #if defined(STRANGE_MALLOC) || defined(MYMALLOC) |
b79f7545 NC |
1296 | Renew(a, PERL_HV_ARRAY_ALLOC_BYTES(newsize) |
1297 | + (SvOOK(hv) ? sizeof(struct xpvhv_aux) : 0), char); | |
8aacddc1 | 1298 | if (!a) { |
4a33f861 | 1299 | PL_nomemok = FALSE; |
422a93e5 GA |
1300 | return; |
1301 | } | |
b79f7545 | 1302 | if (SvOOK(hv)) { |
7a9b70e9 | 1303 | Copy(&a[oldsize * sizeof(HE*)], &a[newsize * sizeof(HE*)], 1, struct xpvhv_aux); |
b79f7545 | 1304 | } |
72940dca | 1305 | #else |
a02a5408 | 1306 | Newx(a, PERL_HV_ARRAY_ALLOC_BYTES(newsize) |
b79f7545 | 1307 | + (SvOOK(hv) ? sizeof(struct xpvhv_aux) : 0), char); |
8aacddc1 | 1308 | if (!a) { |
3280af22 | 1309 | PL_nomemok = FALSE; |
422a93e5 GA |
1310 | return; |
1311 | } | |
7b2c381c | 1312 | Copy(HvARRAY(hv), a, oldsize * sizeof(HE*), char); |
b79f7545 NC |
1313 | if (SvOOK(hv)) { |
1314 | Copy(HvAUX(hv), &a[newsize * sizeof(HE*)], 1, struct xpvhv_aux); | |
1315 | } | |
9a87bd09 | 1316 | Safefree(HvARRAY(hv)); |
72940dca | 1317 | #endif |
3280af22 | 1318 | PL_nomemok = FALSE; |
72311751 | 1319 | Zero(&a[oldsize * sizeof(HE*)], (newsize-oldsize) * sizeof(HE*), char); /* zero 2nd half*/ |
72940dca | 1320 | } |
1321 | else { | |
a02a5408 | 1322 | Newxz(a, PERL_HV_ARRAY_ALLOC_BYTES(newsize), char); |
72940dca | 1323 | } |
cbec9347 | 1324 | xhv->xhv_max = --newsize; /* HvMAX(hv) = --newsize */ |
7b2c381c | 1325 | HvARRAY(hv) = (HE **) a; |
f4431c56 | 1326 | if (!xhv->xhv_keys /* !HvTOTALKEYS(hv) */) /* skip rest if no entries */ |
72940dca | 1327 | return; |
1328 | ||
72311751 GS |
1329 | aep = (HE**)a; |
1330 | for (i=0; i<oldsize; i++,aep++) { | |
a50a3493 NC |
1331 | HE **oentry = aep; |
1332 | HE *entry = *aep; | |
1333 | ||
1334 | if (!entry) /* non-existent */ | |
72940dca | 1335 | continue; |
4c9d89c5 | 1336 | do { |
6136c704 AL |
1337 | register I32 j = (HeHASH(entry) & newsize); |
1338 | ||
1339 | if (j != i) { | |
72940dca | 1340 | j -= i; |
1341 | *oentry = HeNEXT(entry); | |
4d0fbddd | 1342 | HeNEXT(entry) = aep[j]; |
72311751 | 1343 | aep[j] = entry; |
72940dca | 1344 | } |
1345 | else | |
1346 | oentry = &HeNEXT(entry); | |
4c9d89c5 NC |
1347 | entry = *oentry; |
1348 | } while (entry); | |
72940dca | 1349 | } |
1350 | } | |
1351 | ||
b3ac6de7 | 1352 | HV * |
864dbfa3 | 1353 | Perl_newHVhv(pTHX_ HV *ohv) |
b3ac6de7 | 1354 | { |
749123ff | 1355 | dVAR; |
9d4ba2ae | 1356 | HV * const hv = newHV(); |
f4431c56 | 1357 | STRLEN hv_max; |
4beac62f | 1358 | |
3f4d1d78 | 1359 | if (!ohv || (!HvTOTALKEYS(ohv) && !SvMAGICAL((const SV *)ohv))) |
4beac62f | 1360 | return hv; |
4beac62f | 1361 | hv_max = HvMAX(ohv); |
b3ac6de7 | 1362 | |
ad64d0ec | 1363 | if (!SvMAGICAL((const SV *)ohv)) { |
b56ba0bf | 1364 | /* It's an ordinary hash, so copy it fast. AMS 20010804 */ |
eb160463 | 1365 | STRLEN i; |
a3b680e6 | 1366 | const bool shared = !!HvSHAREKEYS(ohv); |
aec46f14 | 1367 | HE **ents, ** const oents = (HE **)HvARRAY(ohv); |
ff875642 | 1368 | char *a; |
a02a5408 | 1369 | Newx(a, PERL_HV_ARRAY_ALLOC_BYTES(hv_max+1), char); |
ff875642 | 1370 | ents = (HE**)a; |
b56ba0bf AMS |
1371 | |
1372 | /* In each bucket... */ | |
1373 | for (i = 0; i <= hv_max; i++) { | |
6136c704 | 1374 | HE *prev = NULL; |
aec46f14 | 1375 | HE *oent = oents[i]; |
b56ba0bf AMS |
1376 | |
1377 | if (!oent) { | |
1378 | ents[i] = NULL; | |
1379 | continue; | |
1380 | } | |
1381 | ||
1382 | /* Copy the linked list of entries. */ | |
aec46f14 | 1383 | for (; oent; oent = HeNEXT(oent)) { |
a3b680e6 AL |
1384 | const U32 hash = HeHASH(oent); |
1385 | const char * const key = HeKEY(oent); | |
1386 | const STRLEN len = HeKLEN(oent); | |
1387 | const int flags = HeKFLAGS(oent); | |
6136c704 | 1388 | HE * const ent = new_HE(); |
c3acb9e0 | 1389 | SV *const val = HeVAL(oent); |
b56ba0bf | 1390 | |
c3acb9e0 | 1391 | HeVAL(ent) = SvIMMORTAL(val) ? val : newSVsv(val); |
19692e8d | 1392 | HeKEY_hek(ent) |
6e838c70 | 1393 | = shared ? share_hek_flags(key, len, hash, flags) |
19692e8d | 1394 | : save_hek_flags(key, len, hash, flags); |
b56ba0bf AMS |
1395 | if (prev) |
1396 | HeNEXT(prev) = ent; | |
1397 | else | |
1398 | ents[i] = ent; | |
1399 | prev = ent; | |
1400 | HeNEXT(ent) = NULL; | |
1401 | } | |
1402 | } | |
1403 | ||
1404 | HvMAX(hv) = hv_max; | |
8aacddc1 | 1405 | HvTOTALKEYS(hv) = HvTOTALKEYS(ohv); |
b56ba0bf | 1406 | HvARRAY(hv) = ents; |
aec46f14 | 1407 | } /* not magical */ |
b56ba0bf AMS |
1408 | else { |
1409 | /* Iterate over ohv, copying keys and values one at a time. */ | |
b3ac6de7 | 1410 | HE *entry; |
bfcb3514 NC |
1411 | const I32 riter = HvRITER_get(ohv); |
1412 | HE * const eiter = HvEITER_get(ohv); | |
f4431c56 | 1413 | STRLEN hv_fill = HvFILL(ohv); |
b56ba0bf AMS |
1414 | |
1415 | /* Can we use fewer buckets? (hv_max is always 2^n-1) */ | |
1416 | while (hv_max && hv_max + 1 >= hv_fill * 2) | |
1417 | hv_max = hv_max / 2; | |
1418 | HvMAX(hv) = hv_max; | |
1419 | ||
4a76a316 | 1420 | hv_iterinit(ohv); |
e16e2ff8 | 1421 | while ((entry = hv_iternext_flags(ohv, 0))) { |
3f4d1d78 FC |
1422 | SV *val = hv_iterval(ohv,entry); |
1423 | SV * const keysv = HeSVKEY(entry); | |
1424 | val = SvIMMORTAL(val) ? val : newSVsv(val); | |
1425 | if (keysv) | |
1426 | (void)hv_store_ent(hv, keysv, val, 0); | |
1427 | else | |
1428 | (void)hv_store_flags(hv, HeKEY(entry), HeKLEN(entry), val, | |
c3acb9e0 | 1429 | HeHASH(entry), HeKFLAGS(entry)); |
b3ac6de7 | 1430 | } |
bfcb3514 NC |
1431 | HvRITER_set(ohv, riter); |
1432 | HvEITER_set(ohv, eiter); | |
b3ac6de7 | 1433 | } |
1c846c1f | 1434 | |
b3ac6de7 IZ |
1435 | return hv; |
1436 | } | |
1437 | ||
defdfed5 Z |
1438 | /* |
1439 | =for apidoc Am|HV *|hv_copy_hints_hv|HV *ohv | |
1440 | ||
1441 | A specialised version of L</newHVhv> for copying C<%^H>. I<ohv> must be | |
1442 | a pointer to a hash (which may have C<%^H> magic, but should be generally | |
1443 | non-magical), or C<NULL> (interpreted as an empty hash). The content | |
1444 | of I<ohv> is copied to a new hash, which has the C<%^H>-specific magic | |
1445 | added to it. A pointer to the new hash is returned. | |
1446 | ||
1447 | =cut | |
1448 | */ | |
1449 | ||
5b9c0671 NC |
1450 | HV * |
1451 | Perl_hv_copy_hints_hv(pTHX_ HV *const ohv) | |
1452 | { | |
1453 | HV * const hv = newHV(); | |
5b9c0671 | 1454 | |
cb1f05e8 | 1455 | if (ohv) { |
5b9c0671 | 1456 | STRLEN hv_max = HvMAX(ohv); |
f4431c56 | 1457 | STRLEN hv_fill = HvFILL(ohv); |
5b9c0671 NC |
1458 | HE *entry; |
1459 | const I32 riter = HvRITER_get(ohv); | |
1460 | HE * const eiter = HvEITER_get(ohv); | |
1461 | ||
1462 | while (hv_max && hv_max + 1 >= hv_fill * 2) | |
1463 | hv_max = hv_max / 2; | |
1464 | HvMAX(hv) = hv_max; | |
1465 | ||
1466 | hv_iterinit(ohv); | |
1467 | while ((entry = hv_iternext_flags(ohv, 0))) { | |
cb1f05e8 | 1468 | SV *const sv = newSVsv(hv_iterval(ohv,entry)); |
7ef9d42c FC |
1469 | SV *heksv = HeSVKEY(entry); |
1470 | if (!heksv && sv) heksv = newSVhek(HeKEY_hek(entry)); | |
95cf2368 | 1471 | if (sv) sv_magic(sv, NULL, PERL_MAGIC_hintselem, |
e3b1b6b1 | 1472 | (char *)heksv, HEf_SVKEY); |
7ef9d42c FC |
1473 | if (heksv == HeSVKEY(entry)) |
1474 | (void)hv_store_ent(hv, heksv, sv, 0); | |
1475 | else { | |
1476 | (void)hv_common(hv, heksv, HeKEY(entry), HeKLEN(entry), | |
1477 | HeKFLAGS(entry), HV_FETCH_ISSTORE|HV_FETCH_JUST_SV, sv, HeHASH(entry)); | |
1478 | SvREFCNT_dec(heksv); | |
1479 | } | |
5b9c0671 NC |
1480 | } |
1481 | HvRITER_set(ohv, riter); | |
1482 | HvEITER_set(ohv, eiter); | |
1483 | } | |
1484 | hv_magic(hv, NULL, PERL_MAGIC_hints); | |
1485 | return hv; | |
1486 | } | |
1487 | ||
e0171a1a DM |
1488 | /* like hv_free_ent, but returns the SV rather than freeing it */ |
1489 | STATIC SV* | |
1490 | S_hv_free_ent_ret(pTHX_ HV *hv, register HE *entry) | |
79072805 | 1491 | { |
97aff369 | 1492 | dVAR; |
16bdeea2 GS |
1493 | SV *val; |
1494 | ||
e0171a1a | 1495 | PERL_ARGS_ASSERT_HV_FREE_ENT_RET; |
7918f24d | 1496 | |
68dc0745 | 1497 | if (!entry) |
e0171a1a | 1498 | return NULL; |
16bdeea2 | 1499 | val = HeVAL(entry); |
68dc0745 | 1500 | if (HeKLEN(entry) == HEf_SVKEY) { |
1501 | SvREFCNT_dec(HeKEY_sv(entry)); | |
8aacddc1 | 1502 | Safefree(HeKEY_hek(entry)); |
44a8e56a | 1503 | } |
1504 | else if (HvSHAREKEYS(hv)) | |
68dc0745 | 1505 | unshare_hek(HeKEY_hek(entry)); |
fde52b5c | 1506 | else |
68dc0745 | 1507 | Safefree(HeKEY_hek(entry)); |
d33b2eba | 1508 | del_HE(entry); |
e0171a1a DM |
1509 | return val; |
1510 | } | |
1511 | ||
1512 | ||
1513 | void | |
1514 | Perl_hv_free_ent(pTHX_ HV *hv, register HE *entry) | |
1515 | { | |
1516 | dVAR; | |
1517 | SV *val; | |
1518 | ||
1519 | PERL_ARGS_ASSERT_HV_FREE_ENT; | |
1520 | ||
1521 | if (!entry) | |
1522 | return; | |
1523 | val = hv_free_ent_ret(hv, entry); | |
272e8453 | 1524 | SvREFCNT_dec(val); |
79072805 LW |
1525 | } |
1526 | ||
f1c32fec | 1527 | |
79072805 | 1528 | void |
864dbfa3 | 1529 | Perl_hv_delayfree_ent(pTHX_ HV *hv, register HE *entry) |
79072805 | 1530 | { |
97aff369 | 1531 | dVAR; |
7918f24d NC |
1532 | |
1533 | PERL_ARGS_ASSERT_HV_DELAYFREE_ENT; | |
1534 | ||
68dc0745 | 1535 | if (!entry) |
79072805 | 1536 | return; |
bc4947fc NC |
1537 | /* SvREFCNT_inc to counter the SvREFCNT_dec in hv_free_ent */ |
1538 | sv_2mortal(SvREFCNT_inc(HeVAL(entry))); /* free between statements */ | |
68dc0745 | 1539 | if (HeKLEN(entry) == HEf_SVKEY) { |
bc4947fc | 1540 | sv_2mortal(SvREFCNT_inc(HeKEY_sv(entry))); |
44a8e56a | 1541 | } |
bc4947fc | 1542 | hv_free_ent(hv, entry); |
79072805 LW |
1543 | } |
1544 | ||
954c1994 GS |
1545 | /* |
1546 | =for apidoc hv_clear | |
1547 | ||
c2217cd3 | 1548 | Frees the all the elements of a hash, leaving it empty. |
8b9a1153 FC |
1549 | The XS equivalent of C<%hash = ()>. See also L</hv_undef>. |
1550 | ||
1551 | If any destructors are triggered as a result, the hv itself may | |
1552 | be freed. | |
954c1994 GS |
1553 | |
1554 | =cut | |
1555 | */ | |
1556 | ||
79072805 | 1557 | void |
864dbfa3 | 1558 | Perl_hv_clear(pTHX_ HV *hv) |
79072805 | 1559 | { |
27da23d5 | 1560 | dVAR; |
cbec9347 | 1561 | register XPVHV* xhv; |
79072805 LW |
1562 | if (!hv) |
1563 | return; | |
49293501 | 1564 | |
ecae49c0 NC |
1565 | DEBUG_A(Perl_hv_assert(aTHX_ hv)); |
1566 | ||
34c3c4e3 DM |
1567 | xhv = (XPVHV*)SvANY(hv); |
1568 | ||
8505eec0 FC |
1569 | ENTER; |
1570 | SAVEFREESV(SvREFCNT_inc_simple_NN(hv)); | |
7b2c381c | 1571 | if (SvREADONLY(hv) && HvARRAY(hv) != NULL) { |
34c3c4e3 | 1572 | /* restricted hash: convert all keys to placeholders */ |
b464bac0 AL |
1573 | STRLEN i; |
1574 | for (i = 0; i <= xhv->xhv_max; i++) { | |
7b2c381c | 1575 | HE *entry = (HvARRAY(hv))[i]; |
3a676441 JH |
1576 | for (; entry; entry = HeNEXT(entry)) { |
1577 | /* not already placeholder */ | |
7996736c | 1578 | if (HeVAL(entry) != &PL_sv_placeholder) { |
fb2352eb FC |
1579 | if (HeVAL(entry) && SvREADONLY(HeVAL(entry)) |
1580 | && !SvIsCOW(HeVAL(entry))) { | |
6136c704 | 1581 | SV* const keysv = hv_iterkeysv(entry); |
3a676441 | 1582 | Perl_croak(aTHX_ |
95b63a38 JH |
1583 | "Attempt to delete readonly key '%"SVf"' from a restricted hash", |
1584 | (void*)keysv); | |
3a676441 JH |
1585 | } |
1586 | SvREFCNT_dec(HeVAL(entry)); | |
7996736c | 1587 | HeVAL(entry) = &PL_sv_placeholder; |
ca732855 | 1588 | HvPLACEHOLDERS(hv)++; |
3a676441 | 1589 | } |
34c3c4e3 DM |
1590 | } |
1591 | } | |
49293501 | 1592 | } |
afbbf215 DM |
1593 | else { |
1594 | hfreeentries(hv); | |
1595 | HvPLACEHOLDERS_set(hv, 0); | |
49293501 | 1596 | |
afbbf215 DM |
1597 | if (SvRMAGICAL(hv)) |
1598 | mg_clear(MUTABLE_SV(hv)); | |
574c8022 | 1599 | |
afbbf215 DM |
1600 | HvHASKFLAGS_off(hv); |
1601 | HvREHASH_off(hv); | |
1602 | } | |
b79f7545 | 1603 | if (SvOOK(hv)) { |
00169e2c | 1604 | if(HvENAME_get(hv)) |
dd69841b | 1605 | mro_isa_changed_in(hv); |
bfcb3514 NC |
1606 | HvEITER_set(hv, NULL); |
1607 | } | |
8505eec0 | 1608 | LEAVE; |
79072805 LW |
1609 | } |
1610 | ||
3540d4ce AB |
1611 | /* |
1612 | =for apidoc hv_clear_placeholders | |
1613 | ||
1614 | Clears any placeholders from a hash. If a restricted hash has any of its keys | |
1615 | marked as readonly and the key is subsequently deleted, the key is not actually | |
1616 | deleted but is marked by assigning it a value of &PL_sv_placeholder. This tags | |
1617 | it so it will be ignored by future operations such as iterating over the hash, | |
4cdaeff7 | 1618 | but will still allow the hash to have a value reassigned to the key at some |
3540d4ce AB |
1619 | future point. This function clears any such placeholder keys from the hash. |
1620 | See Hash::Util::lock_keys() for an example of its use. | |
1621 | ||
1622 | =cut | |
1623 | */ | |
1624 | ||
1625 | void | |
1626 | Perl_hv_clear_placeholders(pTHX_ HV *hv) | |
1627 | { | |
27da23d5 | 1628 | dVAR; |
b3ca2e83 NC |
1629 | const U32 items = (U32)HvPLACEHOLDERS_get(hv); |
1630 | ||
7918f24d NC |
1631 | PERL_ARGS_ASSERT_HV_CLEAR_PLACEHOLDERS; |
1632 | ||
b3ca2e83 NC |
1633 | if (items) |
1634 | clear_placeholders(hv, items); | |
1635 | } | |
1636 | ||
1637 | static void | |
1638 | S_clear_placeholders(pTHX_ HV *hv, U32 items) | |
1639 | { | |
1640 | dVAR; | |
b464bac0 | 1641 | I32 i; |
d3677389 | 1642 | |
7918f24d NC |
1643 | PERL_ARGS_ASSERT_CLEAR_PLACEHOLDERS; |
1644 | ||
d3677389 NC |
1645 | if (items == 0) |
1646 | return; | |
1647 | ||
b464bac0 | 1648 | i = HvMAX(hv); |
d3677389 NC |
1649 | do { |
1650 | /* Loop down the linked list heads */ | |
d3677389 | 1651 | HE **oentry = &(HvARRAY(hv))[i]; |
cf6db12b | 1652 | HE *entry; |
d3677389 | 1653 | |
cf6db12b | 1654 | while ((entry = *oentry)) { |
d3677389 NC |
1655 | if (HeVAL(entry) == &PL_sv_placeholder) { |
1656 | *oentry = HeNEXT(entry); | |
2e58978b | 1657 | if (entry == HvEITER_get(hv)) |
d3677389 | 1658 | HvLAZYDEL_on(hv); |
ae199939 TH |
1659 | else { |
1660 | if (SvOOK(hv) && HvLAZYDEL(hv) && | |
1661 | entry == HeNEXT(HvAUX(hv)->xhv_eiter)) | |
1662 | HeNEXT(HvAUX(hv)->xhv_eiter) = HeNEXT(entry); | |
d3677389 | 1663 | hv_free_ent(hv, entry); |
ae199939 | 1664 | } |
d3677389 NC |
1665 | |
1666 | if (--items == 0) { | |
1667 | /* Finished. */ | |
5d88ecd7 | 1668 | HvTOTALKEYS(hv) -= (IV)HvPLACEHOLDERS_get(hv); |
1b95d04f | 1669 | if (HvUSEDKEYS(hv) == 0) |
d3677389 | 1670 | HvHASKFLAGS_off(hv); |
5d88ecd7 | 1671 | HvPLACEHOLDERS_set(hv, 0); |
d3677389 NC |
1672 | return; |
1673 | } | |
213ce8b3 NC |
1674 | } else { |
1675 | oentry = &HeNEXT(entry); | |
d3677389 NC |
1676 | } |
1677 | } | |
1678 | } while (--i >= 0); | |
1679 | /* You can't get here, hence assertion should always fail. */ | |
1680 | assert (items == 0); | |
1681 | assert (0); | |
3540d4ce AB |
1682 | } |
1683 | ||
76e3520e | 1684 | STATIC void |
cea2e8a9 | 1685 | S_hfreeentries(pTHX_ HV *hv) |
79072805 | 1686 | { |
e0171a1a | 1687 | STRLEN index = 0; |
7d6175ef | 1688 | XPVHV * const xhv = (XPVHV*)SvANY(hv); |
6d1c68e6 | 1689 | SV *sv; |
3abe233e | 1690 | |
7918f24d NC |
1691 | PERL_ARGS_ASSERT_HFREEENTRIES; |
1692 | ||
6d1c68e6 FC |
1693 | while ((sv = Perl_hfree_next_entry(aTHX_ hv, &index))||xhv->xhv_keys) { |
1694 | SvREFCNT_dec(sv); | |
e0171a1a DM |
1695 | } |
1696 | } | |
23976bdd | 1697 | |
b79f7545 | 1698 | |
e0171a1a DM |
1699 | /* hfree_next_entry() |
1700 | * For use only by S_hfreeentries() and sv_clear(). | |
1701 | * Delete the next available HE from hv and return the associated SV. | |
7d6175ef FC |
1702 | * Returns null on empty hash. Nevertheless null is not a reliable |
1703 | * indicator that the hash is empty, as the deleted entry may have a | |
1704 | * null value. | |
e0171a1a DM |
1705 | * indexp is a pointer to the current index into HvARRAY. The index should |
1706 | * initially be set to 0. hfree_next_entry() may update it. */ | |
1707 | ||
1708 | SV* | |
1709 | Perl_hfree_next_entry(pTHX_ HV *hv, STRLEN *indexp) | |
1710 | { | |
1711 | struct xpvhv_aux *iter; | |
1712 | HE *entry; | |
1713 | HE ** array; | |
1714 | #ifdef DEBUGGING | |
1715 | STRLEN orig_index = *indexp; | |
1716 | #endif | |
1717 | ||
1718 | PERL_ARGS_ASSERT_HFREE_NEXT_ENTRY; | |
1719 | ||
e0171a1a DM |
1720 | if (SvOOK(hv) && ((iter = HvAUX(hv))) |
1721 | && ((entry = iter->xhv_eiter)) ) | |
1722 | { | |
1723 | /* the iterator may get resurrected after each | |
1724 | * destructor call, so check each time */ | |
1725 | if (entry && HvLAZYDEL(hv)) { /* was deleted earlier? */ | |
1726 | HvLAZYDEL_off(hv); | |
922090f4 DM |
1727 | hv_free_ent(hv, entry); |
1728 | /* warning: at this point HvARRAY may have been | |
1729 | * re-allocated, HvMAX changed etc */ | |
23976bdd | 1730 | } |
e0171a1a DM |
1731 | iter->xhv_riter = -1; /* HvRITER(hv) = -1 */ |
1732 | iter->xhv_eiter = NULL; /* HvEITER(hv) = NULL */ | |
1733 | } | |
1734 | ||
00a1a643 DM |
1735 | if (!((XPVHV*)SvANY(hv))->xhv_keys) |
1736 | return NULL; | |
1737 | ||
e0171a1a DM |
1738 | array = HvARRAY(hv); |
1739 | assert(array); | |
1740 | while ( ! ((entry = array[*indexp])) ) { | |
1741 | if ((*indexp)++ >= HvMAX(hv)) | |
1742 | *indexp = 0; | |
1743 | assert(*indexp != orig_index); | |
1744 | } | |
1745 | array[*indexp] = HeNEXT(entry); | |
1746 | ((XPVHV*) SvANY(hv))->xhv_keys--; | |
1747 | ||
1748 | if ( PL_phase != PERL_PHASE_DESTRUCT && HvENAME(hv) | |
1749 | && HeVAL(entry) && isGV(HeVAL(entry)) | |
1750 | && GvHV(HeVAL(entry)) && HvENAME(GvHV(HeVAL(entry))) | |
1751 | ) { | |
1752 | STRLEN klen; | |
1753 | const char * const key = HePV(entry,klen); | |
1754 | if ((klen > 1 && key[klen-1]==':' && key[klen-2]==':') | |
1755 | || (klen == 1 && key[0] == ':')) { | |
1756 | mro_package_moved( | |
1757 | NULL, GvHV(HeVAL(entry)), | |
1758 | (GV *)HeVAL(entry), 0 | |
1759 | ); | |
1760 | } | |
1761 | } | |
1762 | return hv_free_ent_ret(hv, entry); | |
79072805 LW |
1763 | } |
1764 | ||
e0171a1a | 1765 | |
954c1994 GS |
1766 | /* |
1767 | =for apidoc hv_undef | |
1768 | ||
8b9a1153 | 1769 | Undefines the hash. The XS equivalent of C<undef(%hash)>. |
c2217cd3 DM |
1770 | |
1771 | As well as freeing all the elements of the hash (like hv_clear()), this | |
1772 | also frees any auxiliary data and storage associated with the hash. | |
8b9a1153 FC |
1773 | |
1774 | If any destructors are triggered as a result, the hv itself may | |
1775 | be freed. | |
1776 | ||
c2217cd3 | 1777 | See also L</hv_clear>. |
954c1994 GS |
1778 | |
1779 | =cut | |
1780 | */ | |
1781 | ||
79072805 | 1782 | void |
8581adba | 1783 | Perl_hv_undef_flags(pTHX_ HV *hv, U32 flags) |
79072805 | 1784 | { |
97aff369 | 1785 | dVAR; |
cbec9347 | 1786 | register XPVHV* xhv; |
bfcb3514 | 1787 | const char *name; |
8505eec0 | 1788 | const bool save = !!SvREFCNT(hv); |
86f55936 | 1789 | |
79072805 LW |
1790 | if (!hv) |
1791 | return; | |
ecae49c0 | 1792 | DEBUG_A(Perl_hv_assert(aTHX_ hv)); |
cbec9347 | 1793 | xhv = (XPVHV*)SvANY(hv); |
dd69841b | 1794 | |
745edda6 FC |
1795 | /* The name must be deleted before the call to hfreeeeentries so that |
1796 | CVs are anonymised properly. But the effective name must be pre- | |
1797 | served until after that call (and only deleted afterwards if the | |
1798 | call originated from sv_clear). For stashes with one name that is | |
1799 | both the canonical name and the effective name, hv_name_set has to | |
1800 | allocate an array for storing the effective name. We can skip that | |
1801 | during global destruction, as it does not matter where the CVs point | |
1802 | if they will be freed anyway. */ | |
104d7b69 DM |
1803 | /* note that the code following prior to hfreeentries is duplicated |
1804 | * in sv_clear(), and changes here should be done there too */ | |
745edda6 | 1805 | if (PL_phase != PERL_PHASE_DESTRUCT && (name = HvNAME(hv))) { |
04fe65b0 | 1806 | if (PL_stashcache) |
4643eb69 BF |
1807 | (void)hv_delete(PL_stashcache, name, |
1808 | HEK_UTF8(HvNAME_HEK(hv)) ? -HvNAMELEN_get(hv) : HvNAMELEN_get(hv), | |
1809 | G_DISCARD | |
1810 | ); | |
bd61b366 | 1811 | hv_name_set(hv, NULL, 0, 0); |
85e6fe83 | 1812 | } |
8505eec0 FC |
1813 | if (save) { |
1814 | ENTER; | |
1815 | SAVEFREESV(SvREFCNT_inc_simple_NN(hv)); | |
1816 | } | |
2d0d1ecc | 1817 | hfreeentries(hv); |
47f1cf77 FC |
1818 | if (SvOOK(hv)) { |
1819 | struct xpvhv_aux * const aux = HvAUX(hv); | |
1820 | struct mro_meta *meta; | |
745edda6 FC |
1821 | |
1822 | if ((name = HvENAME_get(hv))) { | |
5f243b5f | 1823 | if (PL_phase != PERL_PHASE_DESTRUCT) |
745edda6 | 1824 | mro_isa_changed_in(hv); |
745edda6 FC |
1825 | if (PL_stashcache) |
1826 | (void)hv_delete( | |
4643eb69 BF |
1827 | PL_stashcache, name, |
1828 | HEK_UTF8(HvENAME_HEK(hv)) ? -HvENAMELEN_get(hv) : HvENAMELEN_get(hv), | |
1829 | G_DISCARD | |
745edda6 FC |
1830 | ); |
1831 | } | |
1832 | ||
1833 | /* If this call originated from sv_clear, then we must check for | |
1834 | * effective names that need freeing, as well as the usual name. */ | |
1835 | name = HvNAME(hv); | |
15d9236d | 1836 | if (flags & HV_NAME_SETALL ? !!aux->xhv_name_u.xhvnameu_name : !!name) { |
745edda6 | 1837 | if (name && PL_stashcache) |
4643eb69 | 1838 | (void)hv_delete(PL_stashcache, name, (HEK_UTF8(HvNAME_HEK(hv)) ? -HvNAMELEN_get(hv) : HvNAMELEN_get(hv)), G_DISCARD); |
745edda6 | 1839 | hv_name_set(hv, NULL, 0, flags); |
47f1cf77 FC |
1840 | } |
1841 | if((meta = aux->xhv_mro_meta)) { | |
1842 | if (meta->mro_linear_all) { | |
1843 | SvREFCNT_dec(MUTABLE_SV(meta->mro_linear_all)); | |
1844 | meta->mro_linear_all = NULL; | |
1845 | /* This is just acting as a shortcut pointer. */ | |
1846 | meta->mro_linear_current = NULL; | |
1847 | } else if (meta->mro_linear_current) { | |
1848 | /* Only the current MRO is stored, so this owns the data. | |
1849 | */ | |
1850 | SvREFCNT_dec(meta->mro_linear_current); | |
1851 | meta->mro_linear_current = NULL; | |
1852 | } | |
9bfbb681 | 1853 | SvREFCNT_dec(meta->mro_nextmethod); |
47f1cf77 FC |
1854 | SvREFCNT_dec(meta->isa); |
1855 | Safefree(meta); | |
1856 | aux->xhv_mro_meta = NULL; | |
1857 | } | |
3b37eb24 | 1858 | if (!aux->xhv_name_u.xhvnameu_name && ! aux->xhv_backreferences) |
745edda6 | 1859 | SvFLAGS(hv) &= ~SVf_OOK; |
745edda6 FC |
1860 | } |
1861 | if (!SvOOK(hv)) { | |
1862 | Safefree(HvARRAY(hv)); | |
1863 | xhv->xhv_max = 7; /* HvMAX(hv) = 7 (it's a normal hash) */ | |
1864 | HvARRAY(hv) = 0; | |
2d0d1ecc | 1865 | } |
5bec93be DM |
1866 | /* if we're freeing the HV, the SvMAGIC field has been reused for |
1867 | * other purposes, and so there can't be any placeholder magic */ | |
1868 | if (SvREFCNT(hv)) | |
1869 | HvPLACEHOLDERS_set(hv, 0); | |
a0d0e21e LW |
1870 | |
1871 | if (SvRMAGICAL(hv)) | |
ad64d0ec | 1872 | mg_clear(MUTABLE_SV(hv)); |
8505eec0 | 1873 | if (save) LEAVE; |
79072805 LW |
1874 | } |
1875 | ||
4d0fbddd NC |
1876 | /* |
1877 | =for apidoc hv_fill | |
1878 | ||
1879 | Returns the number of hash buckets that happen to be in use. This function is | |
1880 | wrapped by the macro C<HvFILL>. | |
1881 | ||
1882 | Previously this value was stored in the HV structure, rather than being | |
1883 | calculated on demand. | |
1884 | ||
1885 | =cut | |
1886 | */ | |
1887 | ||
1888 | STRLEN | |
1889 | Perl_hv_fill(pTHX_ HV const *const hv) | |
1890 | { | |
1891 | STRLEN count = 0; | |
1892 | HE **ents = HvARRAY(hv); | |
1893 | ||
1894 | PERL_ARGS_ASSERT_HV_FILL; | |
1895 | ||
1896 | if (ents) { | |
fcd24582 NC |
1897 | HE *const *const last = ents + HvMAX(hv); |
1898 | count = last + 1 - ents; | |
4d0fbddd NC |
1899 | |
1900 | do { | |
fcd24582 NC |
1901 | if (!*ents) |
1902 | --count; | |
1903 | } while (++ents <= last); | |
4d0fbddd NC |
1904 | } |
1905 | return count; | |
1906 | } | |
1907 | ||
b464bac0 | 1908 | static struct xpvhv_aux* |
5f66b61c | 1909 | S_hv_auxinit(HV *hv) { |
bfcb3514 | 1910 | struct xpvhv_aux *iter; |
b79f7545 | 1911 | char *array; |
bfcb3514 | 1912 | |
7918f24d NC |
1913 | PERL_ARGS_ASSERT_HV_AUXINIT; |
1914 | ||
b79f7545 | 1915 | if (!HvARRAY(hv)) { |
a02a5408 | 1916 | Newxz(array, PERL_HV_ARRAY_ALLOC_BYTES(HvMAX(hv) + 1) |
b79f7545 NC |
1917 | + sizeof(struct xpvhv_aux), char); |
1918 | } else { | |
1919 | array = (char *) HvARRAY(hv); | |
1920 | Renew(array, PERL_HV_ARRAY_ALLOC_BYTES(HvMAX(hv) + 1) | |
1921 | + sizeof(struct xpvhv_aux), char); | |
1922 | } | |
1923 | HvARRAY(hv) = (HE**) array; | |
a82b195b | 1924 | SvOOK_on(hv); |
b79f7545 | 1925 | iter = HvAUX(hv); |
bfcb3514 NC |
1926 | |
1927 | iter->xhv_riter = -1; /* HvRITER(hv) = -1 */ | |
4608196e | 1928 | iter->xhv_eiter = NULL; /* HvEITER(hv) = NULL */ |
15d9236d | 1929 | iter->xhv_name_u.xhvnameu_name = 0; |
b7247a80 | 1930 | iter->xhv_name_count = 0; |
86f55936 | 1931 | iter->xhv_backreferences = 0; |
e1a479c5 | 1932 | iter->xhv_mro_meta = NULL; |
bfcb3514 NC |
1933 | return iter; |
1934 | } | |
1935 | ||
954c1994 GS |
1936 | /* |
1937 | =for apidoc hv_iterinit | |
1938 | ||
1939 | Prepares a starting point to traverse a hash table. Returns the number of | |
1b95d04f | 1940 | keys in the hash (i.e. the same as C<HvUSEDKEYS(hv)>). The return value is |
1c846c1f | 1941 | currently only meaningful for hashes without tie magic. |
954c1994 GS |
1942 | |
1943 | NOTE: Before version 5.004_65, C<hv_iterinit> used to return the number of | |
1944 | hash buckets that happen to be in use. If you still need that esoteric | |
b24b84ef | 1945 | value, you can get it through the macro C<HvFILL(hv)>. |
954c1994 | 1946 | |
e16e2ff8 | 1947 | |
954c1994 GS |
1948 | =cut |
1949 | */ | |
1950 | ||
79072805 | 1951 | I32 |
864dbfa3 | 1952 | Perl_hv_iterinit(pTHX_ HV *hv) |
79072805 | 1953 | { |
7918f24d NC |
1954 | PERL_ARGS_ASSERT_HV_ITERINIT; |
1955 | ||
1956 | /* FIXME: Are we not NULL, or do we croak? Place bets now! */ | |
1957 | ||
aa689395 | 1958 | if (!hv) |
cea2e8a9 | 1959 | Perl_croak(aTHX_ "Bad hash"); |
bfcb3514 | 1960 | |
b79f7545 | 1961 | if (SvOOK(hv)) { |
6136c704 | 1962 | struct xpvhv_aux * const iter = HvAUX(hv); |
0bd48802 | 1963 | HE * const entry = iter->xhv_eiter; /* HvEITER(hv) */ |
bfcb3514 NC |
1964 | if (entry && HvLAZYDEL(hv)) { /* was deleted earlier? */ |
1965 | HvLAZYDEL_off(hv); | |
1966 | hv_free_ent(hv, entry); | |
1967 | } | |
1968 | iter->xhv_riter = -1; /* HvRITER(hv) = -1 */ | |
4608196e | 1969 | iter->xhv_eiter = NULL; /* HvEITER(hv) = NULL */ |
bfcb3514 | 1970 | } else { |
6136c704 | 1971 | hv_auxinit(hv); |
72940dca | 1972 | } |
44a2ac75 | 1973 | |
cbec9347 | 1974 | /* used to be xhv->xhv_fill before 5.004_65 */ |
5d88ecd7 | 1975 | return HvTOTALKEYS(hv); |
79072805 | 1976 | } |
bfcb3514 NC |
1977 | |
1978 | I32 * | |
1979 | Perl_hv_riter_p(pTHX_ HV *hv) { | |
1980 | struct xpvhv_aux *iter; | |
1981 | ||
7918f24d NC |
1982 | PERL_ARGS_ASSERT_HV_RITER_P; |
1983 | ||
bfcb3514 NC |
1984 | if (!hv) |
1985 | Perl_croak(aTHX_ "Bad hash"); | |
1986 | ||
6136c704 | 1987 | iter = SvOOK(hv) ? HvAUX(hv) : hv_auxinit(hv); |
bfcb3514 NC |
1988 | return &(iter->xhv_riter); |
1989 | } | |
1990 | ||
1991 | HE ** | |
1992 | Perl_hv_eiter_p(pTHX_ HV *hv) { | |
1993 | struct xpvhv_aux *iter; | |
1994 | ||
7918f24d NC |
1995 | PERL_ARGS_ASSERT_HV_EITER_P; |
1996 | ||
bfcb3514 NC |
1997 | if (!hv) |
1998 | Perl_croak(aTHX_ "Bad hash"); | |
1999 | ||
6136c704 | 2000 | iter = SvOOK(hv) ? HvAUX(hv) : hv_auxinit(hv); |
bfcb3514 NC |
2001 | return &(iter->xhv_eiter); |
2002 | } | |
2003 | ||
2004 | void | |
2005 | Perl_hv_riter_set(pTHX_ HV *hv, I32 riter) { | |
2006 | struct xpvhv_aux *iter; | |
2007 | ||
7918f24d NC |
2008 | PERL_ARGS_ASSERT_HV_RITER_SET; |
2009 | ||
bfcb3514 NC |
2010 | if (!hv) |
2011 | Perl_croak(aTHX_ "Bad hash"); | |
2012 | ||
b79f7545 NC |
2013 | if (SvOOK(hv)) { |
2014 | iter = HvAUX(hv); | |
2015 | } else { | |
bfcb3514 NC |
2016 | if (riter == -1) |
2017 | return; | |
2018 | ||
6136c704 | 2019 | iter = hv_auxinit(hv); |
bfcb3514 NC |
2020 | } |
2021 | iter->xhv_riter = riter; | |
2022 | } | |
2023 | ||
2024 | void | |
2025 | Perl_hv_eiter_set(pTHX_ HV *hv, HE *eiter) { | |
2026 | struct xpvhv_aux *iter; | |
2027 | ||
7918f24d NC |
2028 | PERL_ARGS_ASSERT_HV_EITER_SET; |
2029 | ||
bfcb3514 NC |
2030 | if (!hv) |
2031 | Perl_croak(aTHX_ "Bad hash"); | |
2032 | ||
b79f7545 NC |
2033 | if (SvOOK(hv)) { |
2034 | iter = HvAUX(hv); | |
2035 | } else { | |
bfcb3514 NC |
2036 | /* 0 is the default so don't go malloc()ing a new structure just to |
2037 | hold 0. */ | |
2038 | if (!eiter) | |
2039 | return; | |
2040 | ||
6136c704 | 2041 | iter = hv_auxinit(hv); |
bfcb3514 NC |
2042 | } |
2043 | iter->xhv_eiter = eiter; | |
2044 | } | |
2045 | ||
bfcb3514 | 2046 | void |
4164be69 | 2047 | Perl_hv_name_set(pTHX_ HV *hv, const char *name, U32 len, U32 flags) |
bfcb3514 | 2048 | { |
97aff369 | 2049 | dVAR; |
b79f7545 | 2050 | struct xpvhv_aux *iter; |
7423f6db | 2051 | U32 hash; |
78b79c77 | 2052 | HEK **spot; |
46c461b5 | 2053 | |
7918f24d | 2054 | PERL_ARGS_ASSERT_HV_NAME_SET; |
bfcb3514 | 2055 | |
4164be69 NC |
2056 | if (len > I32_MAX) |
2057 | Perl_croak(aTHX_ "panic: hv name too long (%"UVuf")", (UV) len); | |
2058 | ||
b79f7545 NC |
2059 | if (SvOOK(hv)) { |
2060 | iter = HvAUX(hv); | |
15d9236d | 2061 | if (iter->xhv_name_u.xhvnameu_name) { |
b7247a80 | 2062 | if(iter->xhv_name_count) { |
745edda6 | 2063 | if(flags & HV_NAME_SETALL) { |
15d9236d | 2064 | HEK ** const name = HvAUX(hv)->xhv_name_u.xhvnameu_names; |
78b79c77 FC |
2065 | HEK **hekp = name + ( |
2066 | iter->xhv_name_count < 0 | |
2067 | ? -iter->xhv_name_count | |
2068 | : iter->xhv_name_count | |
2069 | ); | |
2070 | while(hekp-- > name+1) | |
b7247a80 | 2071 | unshare_hek_or_pvn(*hekp, 0, 0, 0); |
78b79c77 FC |
2072 | /* The first elem may be null. */ |
2073 | if(*name) unshare_hek_or_pvn(*name, 0, 0, 0); | |
b7247a80 | 2074 | Safefree(name); |
15d9236d | 2075 | spot = &iter->xhv_name_u.xhvnameu_name; |
78b79c77 FC |
2076 | iter->xhv_name_count = 0; |
2077 | } | |
2078 | else { | |
78b79c77 FC |
2079 | if(iter->xhv_name_count > 0) { |
2080 | /* shift some things over */ | |
15d9236d NC |
2081 | Renew( |
2082 | iter->xhv_name_u.xhvnameu_names, iter->xhv_name_count + 1, HEK * | |
4c2bfb4f | 2083 | ); |
15d9236d | 2084 | spot = iter->xhv_name_u.xhvnameu_names; |
4c2bfb4f | 2085 | spot[iter->xhv_name_count] = spot[1]; |
78b79c77 | 2086 | spot[1] = spot[0]; |
4c2bfb4f | 2087 | iter->xhv_name_count = -(iter->xhv_name_count + 1); |
78b79c77 | 2088 | } |
15d9236d | 2089 | else if(*(spot = iter->xhv_name_u.xhvnameu_names)) { |
78b79c77 FC |
2090 | unshare_hek_or_pvn(*spot, 0, 0, 0); |
2091 | } | |
2092 | } | |
2093 | } | |
745edda6 | 2094 | else if (flags & HV_NAME_SETALL) { |
15d9236d NC |
2095 | unshare_hek_or_pvn(iter->xhv_name_u.xhvnameu_name, 0, 0, 0); |
2096 | spot = &iter->xhv_name_u.xhvnameu_name; | |
b7247a80 | 2097 | } |
745edda6 | 2098 | else { |
15d9236d NC |
2099 | HEK * const existing_name = iter->xhv_name_u.xhvnameu_name; |
2100 | Newx(iter->xhv_name_u.xhvnameu_names, 2, HEK *); | |
745edda6 | 2101 | iter->xhv_name_count = -2; |
15d9236d | 2102 | spot = iter->xhv_name_u.xhvnameu_names; |
745edda6 FC |
2103 | spot[1] = existing_name; |
2104 | } | |
7423f6db | 2105 | } |
15d9236d | 2106 | else { spot = &iter->xhv_name_u.xhvnameu_name; iter->xhv_name_count = 0; } |
16580ff5 | 2107 | } else { |
bfcb3514 NC |
2108 | if (name == 0) |
2109 | return; | |
2110 | ||
6136c704 | 2111 | iter = hv_auxinit(hv); |
15d9236d | 2112 | spot = &iter->xhv_name_u.xhvnameu_name; |
bfcb3514 | 2113 | } |
7423f6db | 2114 | PERL_HASH(hash, name, len); |
c60dbbc3 | 2115 | *spot = name ? share_hek(name, flags & SVf_UTF8 ? -(I32)len : (I32)len, hash) : NULL; |
4643eb69 BF |
2116 | } |
2117 | ||
2118 | /* | |
2119 | This is basically sv_eq_flags() in sv.c, but we avoid the magic | |
2120 | and bytes checking. | |
2121 | */ | |
2122 | ||
2123 | STATIC I32 | |
2124 | hek_eq_pvn_flags(pTHX_ const HEK *hek, const char* pv, const I32 pvlen, const U32 flags) { | |
2125 | if ( (HEK_UTF8(hek) ? 1 : 0) != (flags & SVf_UTF8 ? 1 : 0) ) { | |
2126 | if (flags & SVf_UTF8) | |
2127 | return (bytes_cmp_utf8( | |
2128 | (const U8*)HEK_KEY(hek), HEK_LEN(hek), | |
2129 | (const U8*)pv, pvlen) == 0); | |
2130 | else | |
2131 | return (bytes_cmp_utf8( | |
2132 | (const U8*)pv, pvlen, | |
2133 | (const U8*)HEK_KEY(hek), HEK_LEN(hek)) == 0); | |
2134 | } | |
2135 | else | |
d35fec6c | 2136 | return HEK_LEN(hek) == pvlen && ((HEK_KEY(hek) == pv) |
4643eb69 | 2137 | || memEQ(HEK_KEY(hek), pv, pvlen)); |
bfcb3514 NC |
2138 | } |
2139 | ||
99206677 FC |
2140 | /* |
2141 | =for apidoc hv_ename_add | |
2142 | ||
db4fbf16 | 2143 | Adds a name to a stash's internal list of effective names. See |
99206677 FC |
2144 | C<hv_ename_delete>. |
2145 | ||
2146 | This is called when a stash is assigned to a new location in the symbol | |
2147 | table. | |
2148 | ||
2149 | =cut | |
2150 | */ | |
2151 | ||
ee72b38d | 2152 | void |
27a1175b | 2153 | Perl_hv_ename_add(pTHX_ HV *hv, const char *name, U32 len, U32 flags) |
ee72b38d FC |
2154 | { |
2155 | dVAR; | |
2156 | struct xpvhv_aux *aux = SvOOK(hv) ? HvAUX(hv) : hv_auxinit(hv); | |
2157 | U32 hash; | |
2158 | ||
78b79c77 | 2159 | PERL_ARGS_ASSERT_HV_ENAME_ADD; |
ee72b38d FC |
2160 | |
2161 | if (len > I32_MAX) | |
2162 | Perl_croak(aTHX_ "panic: hv name too long (%"UVuf")", (UV) len); | |
2163 | ||
2164 | PERL_HASH(hash, name, len); | |
2165 | ||
ee72b38d | 2166 | if (aux->xhv_name_count) { |
15d9236d | 2167 | HEK ** const xhv_name = aux->xhv_name_u.xhvnameu_names; |
78b79c77 FC |
2168 | I32 count = aux->xhv_name_count; |
2169 | HEK **hekp = xhv_name + (count < 0 ? -count : count); | |
ee72b38d FC |
2170 | while (hekp-- > xhv_name) |
2171 | if ( | |
4643eb69 BF |
2172 | (HEK_UTF8(*hekp) || (flags & SVf_UTF8)) |
2173 | ? hek_eq_pvn_flags(aTHX_ *hekp, name, (I32)len, flags) | |
2174 | : (HEK_LEN(*hekp) == (I32)len && memEQ(HEK_KEY(*hekp), name, len)) | |
2175 | ) { | |
78b79c77 FC |
2176 | if (hekp == xhv_name && count < 0) |
2177 | aux->xhv_name_count = -count; | |
2178 | return; | |
2179 | } | |
2180 | if (count < 0) aux->xhv_name_count--, count = -count; | |
2181 | else aux->xhv_name_count++; | |
15d9236d | 2182 | Renew(aux->xhv_name_u.xhvnameu_names, count + 1, HEK *); |
c60dbbc3 | 2183 | (aux->xhv_name_u.xhvnameu_names)[count] = share_hek(name, (flags & SVf_UTF8 ? -(I32)len : (I32)len), hash); |
ee72b38d FC |
2184 | } |
2185 | else { | |
15d9236d | 2186 | HEK *existing_name = aux->xhv_name_u.xhvnameu_name; |
ee72b38d | 2187 | if ( |
4643eb69 BF |
2188 | existing_name && ( |
2189 | (HEK_UTF8(existing_name) || (flags & SVf_UTF8)) | |
2190 | ? hek_eq_pvn_flags(aTHX_ existing_name, name, (I32)len, flags) | |
2191 | : (HEK_LEN(existing_name) == (I32)len && memEQ(HEK_KEY(existing_name), name, len)) | |
2192 | ) | |
ee72b38d | 2193 | ) return; |
15d9236d | 2194 | Newx(aux->xhv_name_u.xhvnameu_names, 2, HEK *); |
78b79c77 | 2195 | aux->xhv_name_count = existing_name ? 2 : -2; |
15d9236d | 2196 | *aux->xhv_name_u.xhvnameu_names = existing_name; |
c60dbbc3 | 2197 | (aux->xhv_name_u.xhvnameu_names)[1] = share_hek(name, (flags & SVf_UTF8 ? -(I32)len : (I32)len), hash); |
ee72b38d FC |
2198 | } |
2199 | } | |
2200 | ||
99206677 FC |
2201 | /* |
2202 | =for apidoc hv_ename_delete | |
2203 | ||
db4fbf16 | 2204 | Removes a name from a stash's internal list of effective names. If this is |
99206677 FC |
2205 | the name returned by C<HvENAME>, then another name in the list will take |
2206 | its place (C<HvENAME> will use it). | |
2207 | ||
2208 | This is called when a stash is deleted from the symbol table. | |
2209 | ||
2210 | =cut | |
2211 | */ | |
2212 | ||
ee72b38d | 2213 | void |
27a1175b | 2214 | Perl_hv_ename_delete(pTHX_ HV *hv, const char *name, U32 len, U32 flags) |
ee72b38d FC |
2215 | { |
2216 | dVAR; | |
2217 | struct xpvhv_aux *aux; | |
2218 | ||
78b79c77 | 2219 | PERL_ARGS_ASSERT_HV_ENAME_DELETE; |
ee72b38d FC |
2220 | |
2221 | if (len > I32_MAX) | |
2222 | Perl_croak(aTHX_ "panic: hv name too long (%"UVuf")", (UV) len); | |
2223 | ||
2224 | if (!SvOOK(hv)) return; | |
2225 | ||
2226 | aux = HvAUX(hv); | |
15d9236d | 2227 | if (!aux->xhv_name_u.xhvnameu_name) return; |
ee72b38d FC |
2228 | |
2229 | if (aux->xhv_name_count) { | |
15d9236d | 2230 | HEK ** const namep = aux->xhv_name_u.xhvnameu_names; |
78b79c77 FC |
2231 | I32 const count = aux->xhv_name_count; |
2232 | HEK **victim = namep + (count < 0 ? -count : count); | |
2233 | while (victim-- > namep + 1) | |
ee72b38d | 2234 | if ( |
4643eb69 BF |
2235 | (HEK_UTF8(*victim) || (flags & SVf_UTF8)) |
2236 | ? hek_eq_pvn_flags(aTHX_ *victim, name, (I32)len, flags) | |
2237 | : (HEK_LEN(*victim) == (I32)len && memEQ(HEK_KEY(*victim), name, len)) | |
ee72b38d FC |
2238 | ) { |
2239 | unshare_hek_or_pvn(*victim, 0, 0, 0); | |
78b79c77 FC |
2240 | if (count < 0) ++aux->xhv_name_count; |
2241 | else --aux->xhv_name_count; | |
2242 | if ( | |
2243 | (aux->xhv_name_count == 1 || aux->xhv_name_count == -1) | |
2244 | && !*namep | |
2245 | ) { /* if there are none left */ | |
ee72b38d | 2246 | Safefree(namep); |
15d9236d | 2247 | aux->xhv_name_u.xhvnameu_names = NULL; |
78b79c77 | 2248 | aux->xhv_name_count = 0; |
ee72b38d FC |
2249 | } |
2250 | else { | |
2251 | /* Move the last one back to fill the empty slot. It | |
2252 | does not matter what order they are in. */ | |
78b79c77 | 2253 | *victim = *(namep + (count < 0 ? -count : count) - 1); |
ee72b38d FC |
2254 | } |
2255 | return; | |
2256 | } | |
78b79c77 | 2257 | if ( |
4643eb69 BF |
2258 | count > 0 && (HEK_UTF8(*namep) || (flags & SVf_UTF8)) |
2259 | ? hek_eq_pvn_flags(aTHX_ *namep, name, (I32)len, flags) | |
2260 | : (HEK_LEN(*namep) == (I32)len && memEQ(HEK_KEY(*namep), name, len)) | |
78b79c77 FC |
2261 | ) { |
2262 | aux->xhv_name_count = -count; | |
2263 | } | |
ee72b38d FC |
2264 | } |
2265 | else if( | |
4643eb69 BF |
2266 | (HEK_UTF8(aux->xhv_name_u.xhvnameu_name) || (flags & SVf_UTF8)) |
2267 | ? hek_eq_pvn_flags(aTHX_ aux->xhv_name_u.xhvnameu_name, name, (I32)len, flags) | |
2268 | : (HEK_LEN(aux->xhv_name_u.xhvnameu_name) == (I32)len && | |
2269 | memEQ(HEK_KEY(aux->xhv_name_u.xhvnameu_name), name, len)) | |
ee72b38d | 2270 | ) { |
15d9236d NC |
2271 | HEK * const namehek = aux->xhv_name_u.xhvnameu_name; |
2272 | Newx(aux->xhv_name_u.xhvnameu_names, 1, HEK *); | |
2273 | *aux->xhv_name_u.xhvnameu_names = namehek; | |
3f783763 | 2274 | aux->xhv_name_count = -1; |
ee72b38d FC |
2275 | } |
2276 | } | |
2277 | ||
86f55936 NC |
2278 | AV ** |
2279 | Perl_hv_backreferences_p(pTHX_ HV *hv) { | |
6136c704 | 2280 | struct xpvhv_aux * const iter = SvOOK(hv) ? HvAUX(hv) : hv_auxinit(hv); |
7918f24d NC |
2281 | |
2282 | PERL_ARGS_ASSERT_HV_BACKREFERENCES_P; | |
96a5add6 | 2283 | PERL_UNUSED_CONTEXT; |
7918f24d | 2284 | |
86f55936 NC |
2285 | return &(iter->xhv_backreferences); |
2286 | } | |
2287 | ||
09aad8f0 DM |
2288 | void |
2289 | Perl_hv_kill_backrefs(pTHX_ HV *hv) { | |
2290 | AV *av; | |
2291 | ||
2292 | PERL_ARGS_ASSERT_HV_KILL_BACKREFS; | |
2293 | ||
2294 | if (!SvOOK(hv)) | |
2295 | return; | |
2296 | ||
2297 | av = HvAUX(hv)->xhv_backreferences; | |
2298 | ||
2299 | if (av) { | |
2300 | HvAUX(hv)->xhv_backreferences = 0; | |
2301 | Perl_sv_kill_backrefs(aTHX_ MUTABLE_SV(hv), av); | |
5648c0ae DM |
2302 | if (SvTYPE(av) == SVt_PVAV) |
2303 | SvREFCNT_dec(av); | |
09aad8f0 DM |
2304 | } |
2305 | } | |
2306 | ||
954c1994 | 2307 | /* |
7a7b9979 NC |
2308 | hv_iternext is implemented as a macro in hv.h |
2309 | ||
954c1994 GS |
2310 | =for apidoc hv_iternext |
2311 | ||
2312 | Returns entries from a hash iterator. See C<hv_iterinit>. | |
2313 | ||
fe7bca90 NC |
2314 | You may call C<hv_delete> or C<hv_delete_ent> on the hash entry that the |
2315 | iterator currently points to, without losing your place or invalidating your | |
2316 | iterator. Note that in this case the current entry is deleted from the hash | |
2317 | with your iterator holding the last reference to it. Your iterator is flagged | |
2318 | to free the entry on the next call to C<hv_iternext>, so you must not discard | |
2319 | your iterator immediately else the entry will leak - call C<hv_iternext> to | |
2320 | trigger the resource deallocation. | |
2321 | ||
fe7bca90 NC |
2322 | =for apidoc hv_iternext_flags |
2323 | ||
2324 | Returns entries from a hash iterator. See C<hv_iterinit> and C<hv_iternext>. | |
2325 | The C<flags> value will normally be zero; if HV_ITERNEXT_WANTPLACEHOLDERS is | |
2326 | set the placeholders keys (for restricted hashes) will be returned in addition | |
2327 | to normal keys. By default placeholders are automatically skipped over. | |
7996736c | 2328 | Currently a placeholder is implemented with a value that is |
990c89d7 | 2329 | C<&PL_sv_placeholder>. Note that the implementation of placeholders and |
fe7bca90 NC |
2330 | restricted hashes may change, and the implementation currently is |
2331 | insufficiently abstracted for any change to be tidy. | |
e16e2ff8 | 2332 | |
fe7bca90 | 2333 | =cut |
e16e2ff8 NC |
2334 | */ |
2335 | ||
2336 | HE * | |
2337 | Perl_hv_iternext_flags(pTHX_ HV *hv, I32 flags) | |
2338 | { | |
27da23d5 | 2339 | dVAR; |
cbec9347 | 2340 | register XPVHV* xhv; |
79072805 | 2341 | register HE *entry; |
a0d0e21e | 2342 | HE *oldentry; |
463ee0b2 | 2343 | MAGIC* mg; |
bfcb3514 | 2344 | struct xpvhv_aux *iter; |
79072805 | 2345 | |
7918f24d NC |
2346 | PERL_ARGS_ASSERT_HV_ITERNEXT_FLAGS; |
2347 | ||
79072805 | 2348 | if (!hv) |
cea2e8a9 | 2349 | Perl_croak(aTHX_ "Bad hash"); |
81714fb9 | 2350 | |
cbec9347 | 2351 | xhv = (XPVHV*)SvANY(hv); |
bfcb3514 | 2352 | |
b79f7545 | 2353 | if (!SvOOK(hv)) { |
bfcb3514 NC |
2354 | /* Too many things (well, pp_each at least) merrily assume that you can |
2355 | call iv_iternext without calling hv_iterinit, so we'll have to deal | |
2356 | with it. */ | |
2357 | hv_iterinit(hv); | |
bfcb3514 | 2358 | } |
b79f7545 | 2359 | iter = HvAUX(hv); |
bfcb3514 NC |
2360 | |
2361 | oldentry = entry = iter->xhv_eiter; /* HvEITER(hv) */ | |
e62cc96a | 2362 | if (SvMAGICAL(hv) && SvRMAGICAL(hv)) { |
ad64d0ec | 2363 | if ( ( mg = mg_find((const SV *)hv, PERL_MAGIC_tied) ) ) { |
e62cc96a YO |
2364 | SV * const key = sv_newmortal(); |
2365 | if (entry) { | |
2366 | sv_setsv(key, HeSVKEY_force(entry)); | |
2367 | SvREFCNT_dec(HeSVKEY(entry)); /* get rid of previous key */ | |
2368 | } | |
2369 | else { | |
2370 | char *k; | |
2371 | HEK *hek; | |
2372 | ||
2373 | /* one HE per MAGICAL hash */ | |
2374 | iter->xhv_eiter = entry = new_HE(); /* HvEITER(hv) = new_HE() */ | |
2375 | Zero(entry, 1, HE); | |
ad64d0ec | 2376 | Newxz(k, HEK_BASESIZE + sizeof(const SV *), char); |
e62cc96a YO |
2377 | hek = (HEK*)k; |
2378 | HeKEY_hek(entry) = hek; | |
2379 | HeKLEN(entry) = HEf_SVKEY; | |
2380 | } | |
ad64d0ec | 2381 | magic_nextpack(MUTABLE_SV(hv),mg,key); |
e62cc96a YO |
2382 | if (SvOK(key)) { |
2383 | /* force key to stay around until next time */ | |
2384 | HeSVKEY_set(entry, SvREFCNT_inc_simple_NN(key)); | |
2385 | return entry; /* beware, hent_val is not set */ | |
2386 | } | |
ef8d46e8 | 2387 | SvREFCNT_dec(HeVAL(entry)); |
e62cc96a YO |
2388 | Safefree(HeKEY_hek(entry)); |
2389 | del_HE(entry); | |
2390 | iter->xhv_eiter = NULL; /* HvEITER(hv) = NULL */ | |
2391 | return NULL; | |
81714fb9 | 2392 | } |
79072805 | 2393 | } |
7ee146b1 | 2394 | #if defined(DYNAMIC_ENV_FETCH) && !defined(__riscos__) /* set up %ENV for iteration */ |
ad64d0ec NC |
2395 | if (!entry && SvRMAGICAL((const SV *)hv) |
2396 | && mg_find((const SV *)hv, PERL_MAGIC_env)) { | |
f675dbe5 | 2397 | prime_env_iter(); |
03026e68 JM |
2398 | #ifdef VMS |
2399 | /* The prime_env_iter() on VMS just loaded up new hash values | |
2400 | * so the iteration count needs to be reset back to the beginning | |
2401 | */ | |
2402 | hv_iterinit(hv); | |
2403 | iter = HvAUX(hv); | |
2404 | oldentry = entry = iter->xhv_eiter; /* HvEITER(hv) */ | |
2405 | #endif | |
2406 | } | |
f675dbe5 | 2407 | #endif |
463ee0b2 | 2408 | |
bfaf5b52 | 2409 | /* hv_iterinit now ensures this. */ |
b79f7545 NC |
2410 | assert (HvARRAY(hv)); |
2411 | ||
015a5f36 | 2412 | /* At start of hash, entry is NULL. */ |
fde52b5c | 2413 | if (entry) |
8aacddc1 | 2414 | { |
fde52b5c | 2415 | entry = HeNEXT(entry); |
e16e2ff8 NC |
2416 | if (!(flags & HV_ITERNEXT_WANTPLACEHOLDERS)) { |
2417 | /* | |
2418 | * Skip past any placeholders -- don't want to include them in | |
2419 | * any iteration. | |
2420 | */ | |
7996736c | 2421 | while (entry && HeVAL(entry) == &PL_sv_placeholder) { |
e16e2ff8 NC |
2422 | entry = HeNEXT(entry); |
2423 | } | |
8aacddc1 NIS |
2424 | } |
2425 | } | |
015a5f36 | 2426 | |
9eb4ebd1 NC |
2427 | /* Skip the entire loop if the hash is empty. */ |
2428 | if ((flags & HV_ITERNEXT_WANTPLACEHOLDERS) | |
2429 | ? HvTOTALKEYS(hv) : HvUSEDKEYS(hv)) { | |
900ac051 MM |
2430 | while (!entry) { |
2431 | /* OK. Come to the end of the current list. Grab the next one. */ | |
2432 | ||
2433 | iter->xhv_riter++; /* HvRITER(hv)++ */ | |
2434 | if (iter->xhv_riter > (I32)xhv->xhv_max /* HvRITER(hv) > HvMAX(hv) */) { | |
2435 | /* There is no next one. End of the hash. */ | |
2436 | iter->xhv_riter = -1; /* HvRITER(hv) = -1 */ | |
2437 | break; | |
2438 | } | |
2439 | entry = (HvARRAY(hv))[iter->xhv_riter]; | |
8aacddc1 | 2440 | |
900ac051 MM |
2441 | if (!(flags & HV_ITERNEXT_WANTPLACEHOLDERS)) { |
2442 | /* If we have an entry, but it's a placeholder, don't count it. | |
2443 | Try the next. */ | |
2444 | while (entry && HeVAL(entry) == &PL_sv_placeholder) | |
2445 | entry = HeNEXT(entry); | |
2446 | } | |
2447 | /* Will loop again if this linked list starts NULL | |
2448 | (for HV_ITERNEXT_WANTPLACEHOLDERS) | |
2449 | or if we run through it and find only placeholders. */ | |
015a5f36 | 2450 | } |
fde52b5c | 2451 | } |
41aa816f | 2452 | else iter->xhv_riter = -1; |
79072805 | 2453 | |
72940dca | 2454 | if (oldentry && HvLAZYDEL(hv)) { /* was deleted earlier? */ |
2455 | HvLAZYDEL_off(hv); | |
68dc0745 | 2456 | hv_free_ent(hv, oldentry); |
72940dca | 2457 | } |
a0d0e21e | 2458 | |
fdcd69b6 | 2459 | /*if (HvREHASH(hv) && entry && !HeKREHASH(entry)) |
6c9570dc | 2460 | PerlIO_printf(PerlIO_stderr(), "Awooga %p %p\n", (void*)hv, (void*)entry);*/ |
fdcd69b6 | 2461 | |
bfcb3514 | 2462 | iter->xhv_eiter = entry; /* HvEITER(hv) = entry */ |
79072805 LW |
2463 | return entry; |
2464 | } | |
2465 | ||
954c1994 GS |
2466 | /* |
2467 | =for apidoc hv_iterkey | |
2468 | ||
2469 | Returns the key from the current position of the hash iterator. See | |
2470 | C<hv_iterinit>. | |
2471 | ||
2472 | =cut | |
2473 | */ | |
2474 | ||
79072805 | 2475 | char * |
864dbfa3 | 2476 | Perl_hv_iterkey(pTHX_ register HE *entry, I32 *retlen) |
79072805 | 2477 | { |
7918f24d NC |
2478 | PERL_ARGS_ASSERT_HV_ITERKEY; |
2479 | ||
fde52b5c | 2480 | if (HeKLEN(entry) == HEf_SVKEY) { |
fb73857a | 2481 | STRLEN len; |
0bd48802 | 2482 | char * const p = SvPV(HeKEY_sv(entry), len); |
fb73857a | 2483 | *retlen = len; |
2484 | return p; | |
fde52b5c | 2485 | } |
2486 | else { | |
2487 | *retlen = HeKLEN(entry); | |
2488 | return HeKEY(entry); | |
2489 | } | |
2490 | } | |
2491 | ||
2492 | /* unlike hv_iterval(), this always returns a mortal copy of the key */ | |
954c1994 GS |
2493 | /* |
2494 | =for apidoc hv_iterkeysv | |
2495 | ||
2496 | Returns the key as an C<SV*> from the current position of the hash | |
2497 | iterator. The return value will always be a mortal copy of the key. Also | |
2498 | see C<hv_iterinit>. | |
2499 | ||
2500 | =cut | |
2501 | */ | |
2502 | ||
fde52b5c | 2503 | SV * |
864dbfa3 | 2504 | Perl_hv_iterkeysv(pTHX_ register HE *entry) |
fde52b5c | 2505 | { |
7918f24d NC |
2506 | PERL_ARGS_ASSERT_HV_ITERKEYSV; |
2507 | ||
c1b02ed8 | 2508 | return sv_2mortal(newSVhek(HeKEY_hek(entry))); |
79072805 LW |
2509 | } |
2510 | ||
954c1994 GS |
2511 | /* |
2512 | =for apidoc hv_iterval | |
2513 | ||
2514 | Returns the value from the current position of the hash iterator. See | |
2515 | C<hv_iterkey>. | |
2516 | ||
2517 | =cut | |
2518 | */ | |
2519 | ||
79072805 | 2520 | SV * |
864dbfa3 | 2521 | Perl_hv_iterval(pTHX_ HV *hv, register HE *entry) |
79072805 | 2522 | { |
7918f24d NC |
2523 | PERL_ARGS_ASSERT_HV_ITERVAL; |
2524 | ||
8990e307 | 2525 | if (SvRMAGICAL(hv)) { |
ad64d0ec | 2526 | if (mg_find((const SV *)hv, PERL_MAGIC_tied)) { |
c4420975 | 2527 | SV* const sv = sv_newmortal(); |
bbce6d69 | 2528 | if (HeKLEN(entry) == HEf_SVKEY) |
ad64d0ec | 2529 | mg_copy(MUTABLE_SV(hv), sv, (char*)HeKEY_sv(entry), HEf_SVKEY); |
a3b680e6 | 2530 | else |
ad64d0ec | 2531 | mg_copy(MUTABLE_SV(hv), sv, HeKEY(entry), HeKLEN(entry)); |
463ee0b2 LW |
2532 | return sv; |
2533 | } | |
79072805 | 2534 | } |
fde52b5c | 2535 | return HeVAL(entry); |
79072805 LW |
2536 | } |
2537 | ||
954c1994 GS |
2538 | /* |
2539 | =for apidoc hv_iternextsv | |
2540 | ||
2541 | Performs an C<hv_iternext>, C<hv_iterkey>, and C<hv_iterval> in one | |
2542 | operation. | |
2543 | ||
2544 | =cut | |
2545 | */ | |
2546 | ||
a0d0e21e | 2547 | SV * |
864dbfa3 | 2548 | Perl_hv_iternextsv(pTHX_ HV *hv, char **key, I32 *retlen) |
a0d0e21e | 2549 | { |
0bd48802 AL |
2550 | HE * const he = hv_iternext_flags(hv, 0); |
2551 | ||
7918f24d NC |
2552 | PERL_ARGS_ASSERT_HV_ITERNEXTSV; |
2553 | ||
0bd48802 | 2554 | if (!he) |
a0d0e21e LW |
2555 | return NULL; |
2556 | *key = hv_iterkey(he, retlen); | |
2557 | return hv_iterval(hv, he); | |
2558 | } | |
2559 | ||
954c1994 | 2560 | /* |
bc5cdc23 NC |
2561 | |
2562 | Now a macro in hv.h | |
2563 | ||
954c1994 GS |
2564 | =for apidoc hv_magic |
2565 | ||
2566 | Adds magic to a hash. See C<sv_magic>. | |
2567 | ||
2568 | =cut | |
2569 | */ | |
2570 | ||
bbce6d69 | 2571 | /* possibly free a shared string if no one has access to it |
fde52b5c | 2572 | * len and hash must both be valid for str. |
2573 | */ | |
bbce6d69 | 2574 | void |
864dbfa3 | 2575 | Perl_unsharepvn(pTHX_ const char *str, I32 len, U32 hash) |
fde52b5c | 2576 | { |
19692e8d NC |
2577 | unshare_hek_or_pvn (NULL, str, len, hash); |
2578 | } | |
2579 | ||
2580 | ||
2581 | void | |
2582 | Perl_unshare_hek(pTHX_ HEK *hek) | |
2583 | { | |
bf11fd37 | 2584 | assert(hek); |
19692e8d NC |
2585 | unshare_hek_or_pvn(hek, NULL, 0, 0); |
2586 | } | |
2587 | ||
2588 | /* possibly free a shared string if no one has access to it | |
2589 | hek if non-NULL takes priority over the other 3, else str, len and hash | |
2590 | are used. If so, len and hash must both be valid for str. | |
2591 | */ | |
df132699 | 2592 | STATIC void |
97ddebaf | 2593 | S_unshare_hek_or_pvn(pTHX_ const HEK *hek, const char *str, I32 len, U32 hash) |
19692e8d | 2594 | { |
97aff369 | 2595 | dVAR; |
cbec9347 | 2596 | register XPVHV* xhv; |
20454177 | 2597 | HE *entry; |
fde52b5c | 2598 | register HE **oentry; |
c3654f1a | 2599 | bool is_utf8 = FALSE; |
19692e8d | 2600 | int k_flags = 0; |
aec46f14 | 2601 | const char * const save = str; |
cbbf8932 | 2602 | struct shared_he *he = NULL; |
c3654f1a | 2603 | |
19692e8d | 2604 | if (hek) { |
cbae3960 NC |
2605 | /* Find the shared he which is just before us in memory. */ |
2606 | he = (struct shared_he *)(((char *)hek) | |
2607 | - STRUCT_OFFSET(struct shared_he, | |
2608 | shared_he_hek)); | |
2609 | ||
2610 | /* Assert that the caller passed us a genuine (or at least consistent) | |
2611 | shared hek */ | |
2612 | assert (he->shared_he_he.hent_hek == hek); | |
29404ae0 | 2613 | |
de616631 NC |
2614 | if (he->shared_he_he.he_valu.hent_refcount - 1) { |
2615 | --he->shared_he_he.he_valu.hent_refcount; | |
29404ae0 NC |
2616 | return; |
2617 | } | |
29404ae0 | 2618 | |
19692e8d NC |
2619 | hash = HEK_HASH(hek); |
2620 | } else if (len < 0) { | |
2621 | STRLEN tmplen = -len; | |
2622 | is_utf8 = TRUE; | |
2623 | /* See the note in hv_fetch(). --jhi */ | |
2624 | str = (char*)bytes_from_utf8((U8*)str, &tmplen, &is_utf8); | |
2625 | len = tmplen; | |
2626 | if (is_utf8) | |
2627 | k_flags = HVhek_UTF8; | |
2628 | if (str != save) | |
2629 | k_flags |= HVhek_WASUTF8 | HVhek_FREEKEY; | |
c3654f1a | 2630 | } |
1c846c1f | 2631 | |
de616631 | 2632 | /* what follows was the moral equivalent of: |
6b88bc9c | 2633 | if ((Svp = hv_fetch(PL_strtab, tmpsv, FALSE, hash))) { |
a0714e2c | 2634 | if (--*Svp == NULL) |
6b88bc9c | 2635 | hv_delete(PL_strtab, str, len, G_DISCARD, hash); |
bbce6d69 | 2636 | } */ |
cbec9347 | 2637 | xhv = (XPVHV*)SvANY(PL_strtab); |
fde52b5c | 2638 | /* assert(xhv_array != 0) */ |
9de10d5c | 2639 | oentry = &(HvARRAY(PL_strtab))[hash & (I32) HvMAX(PL_strtab)]; |
6c1b96a1 NC |
2640 | if (he) { |
2641 | const HE *const he_he = &(he->shared_he_he); | |
45d1cc86 | 2642 | for (entry = *oentry; entry; oentry = &HeNEXT(entry), entry = *oentry) { |
35ab5632 NC |
2643 | if (entry == he_he) |
2644 | break; | |
19692e8d NC |
2645 | } |
2646 | } else { | |
35a4481c | 2647 | const int flags_masked = k_flags & HVhek_MASK; |
45d1cc86 | 2648 | for (entry = *oentry; entry; oentry = &HeNEXT(entry), entry = *oentry) { |
19692e8d NC |
2649 | if (HeHASH(entry) != hash) /* strings can't be equal */ |
2650 | continue; | |
2651 | if (HeKLEN(entry) != len) | |
2652 | continue; | |
2653 | if (HeKEY(entry) != str && memNE(HeKEY(entry),str,len)) /* is this it? */ | |
2654 | continue; | |
2655 | if (HeKFLAGS(entry) != flags_masked) | |
2656 | continue; | |
19692e8d NC |
2657 | break; |
2658 | } | |
2659 | } | |
2660 | ||
35ab5632 NC |
2661 | if (entry) { |
2662 | if (--entry->he_valu.hent_refcount == 0) { | |
19692e8d | 2663 | *oentry = HeNEXT(entry); |
cbae3960 | 2664 | Safefree(entry); |
4c7185a0 | 2665 | xhv->xhv_keys--; /* HvTOTALKEYS(hv)-- */ |
19692e8d | 2666 | } |
fde52b5c | 2667 | } |
19692e8d | 2668 | |
9b387841 NC |
2669 | if (!entry) |
2670 | Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL), | |
12578ffb | 2671 | "Attempt to free nonexistent shared string '%s'%s" |
9b387841 NC |
2672 | pTHX__FORMAT, |
2673 | hek ? HEK_KEY(hek) : str, | |
2674 | ((k_flags & HVhek_UTF8) ? " (utf8)" : "") pTHX__VALUE); | |
19692e8d NC |
2675 | if (k_flags & HVhek_FREEKEY) |
2676 | Safefree(str); | |
fde52b5c | 2677 | } |
2678 | ||
bbce6d69 | 2679 | /* get a (constant) string ptr from the global string table |
2680 | * string will get added if it is not already there. | |
fde52b5c | 2681 | * len and hash must both be valid for str. |
2682 | */ | |
bbce6d69 | 2683 | HEK * |
864dbfa3 | 2684 | Perl_share_hek(pTHX_ const char *str, I32 len, register U32 hash) |
fde52b5c | 2685 | { |
da58a35d | 2686 | bool is_utf8 = FALSE; |
19692e8d | 2687 | int flags = 0; |
aec46f14 | 2688 | const char * const save = str; |
da58a35d | 2689 | |
7918f24d NC |
2690 | PERL_ARGS_ASSERT_SHARE_HEK; |
2691 | ||
da58a35d | 2692 | if (len < 0) { |
77caf834 | 2693 | STRLEN tmplen = -len; |
da58a35d | 2694 | is_utf8 = TRUE; |
77caf834 JH |
2695 | /* See the note in hv_fetch(). --jhi */ |
2696 | str = (char*)bytes_from_utf8((U8*)str, &tmplen, &is_utf8); | |
2697 | len = tmplen; | |
19692e8d NC |
2698 | /* If we were able to downgrade here, then than means that we were passed |
2699 | in a key which only had chars 0-255, but was utf8 encoded. */ | |
2700 | if (is_utf8) | |
2701 | flags = HVhek_UTF8; | |
2702 | /* If we found we were able to downgrade the string to bytes, then | |
2703 | we should flag that it needs upgrading on keys or each. Also flag | |
2704 | that we need share_hek_flags to free the string. */ | |
4643eb69 BF |
2705 | if (str != save) { |
2706 | PERL_HASH(hash, str, len); | |
19692e8d | 2707 | flags |= HVhek_WASUTF8 | HVhek_FREEKEY; |
4643eb69 | 2708 | } |
19692e8d NC |
2709 | } |
2710 | ||
6e838c70 | 2711 | return share_hek_flags (str, len, hash, flags); |
19692e8d NC |
2712 | } |
2713 | ||
6e838c70 | 2714 | STATIC HEK * |
19692e8d NC |
2715 | S_share_hek_flags(pTHX_ const char *str, I32 len, register U32 hash, int flags) |
2716 | { | |
97aff369 | 2717 | dVAR; |
19692e8d | 2718 | register HE *entry; |
35a4481c | 2719 | const int flags_masked = flags & HVhek_MASK; |
263cb4a6 | 2720 | const U32 hindex = hash & (I32) HvMAX(PL_strtab); |
7918f24d NC |
2721 | register XPVHV * const xhv = (XPVHV*)SvANY(PL_strtab); |
2722 | ||
2723 | PERL_ARGS_ASSERT_SHARE_HEK_FLAGS; | |
bbce6d69 | 2724 | |
fde52b5c | 2725 | /* what follows is the moral equivalent of: |
1c846c1f | 2726 | |
6b88bc9c | 2727 | if (!(Svp = hv_fetch(PL_strtab, str, len, FALSE))) |
a0714e2c | 2728 | hv_store(PL_strtab, str, len, NULL, hash); |
fdcd69b6 NC |
2729 | |
2730 | Can't rehash the shared string table, so not sure if it's worth | |
2731 | counting the number of entries in the linked list | |
bbce6d69 | 2732 | */ |
7918f24d | 2733 | |
fde52b5c | 2734 | /* assert(xhv_array != 0) */ |
263cb4a6 NC |
2735 | entry = (HvARRAY(PL_strtab))[hindex]; |
2736 | for (;entry; entry = HeNEXT(entry)) { | |
fde52b5c | 2737 | if (HeHASH(entry) != hash) /* strings can't be equal */ |
2738 | continue; | |
2739 | if (HeKLEN(entry) != len) | |
2740 | continue; | |
1c846c1f | 2741 | if (HeKEY(entry) != str && memNE(HeKEY(entry),str,len)) /* is this it? */ |
fde52b5c | 2742 | continue; |
19692e8d | 2743 | if (HeKFLAGS(entry) != flags_masked) |
c3654f1a | 2744 | continue; |
fde52b5c | 2745 | break; |
2746 | } | |
263cb4a6 NC |
2747 | |
2748 | if (!entry) { | |
45d1cc86 NC |
2749 | /* What used to be head of the list. |
2750 | If this is NULL, then we're the first entry for this slot, which | |
2751 | means we need to increate fill. */ | |
cbae3960 NC |
2752 | struct shared_he *new_entry; |
2753 | HEK *hek; | |
2754 | char *k; | |
263cb4a6 NC |
2755 | HE **const head = &HvARRAY(PL_strtab)[hindex]; |
2756 | HE *const next = *head; | |
cbae3960 NC |
2757 | |
2758 | /* We don't actually store a HE from the arena and a regular HEK. | |
2759 | Instead we allocate one chunk of memory big enough for both, | |
2760 | and put the HEK straight after the HE. This way we can find the | |
f52337cf | 2761 | HE directly from the HEK. |
cbae3960 NC |
2762 | */ |
2763 | ||
a02a5408 | 2764 | Newx(k, STRUCT_OFFSET(struct shared_he, |
cbae3960 NC |
2765 | shared_he_hek.hek_key[0]) + len + 2, char); |
2766 | new_entry = (struct shared_he *)k; | |
2767 | entry = &(new_entry->shared_he_he); | |
2768 | hek = &(new_entry->shared_he_hek); | |
2769 | ||
2770 | Copy(str, HEK_KEY(hek), len, char); | |
2771 | HEK_KEY(hek)[len] = 0; | |
2772 | HEK_LEN(hek) = len; | |
2773 | HEK_HASH(hek) = hash; | |
2774 | HEK_FLAGS(hek) = (unsigned char)flags_masked; | |
2775 | ||
2776 | /* Still "point" to the HEK, so that other code need not know what | |
2777 | we're up to. */ | |
2778 | HeKEY_hek(entry) = hek; | |
de616631 | 2779 | entry->he_valu.hent_refcount = 0; |
263cb4a6 NC |
2780 | HeNEXT(entry) = next; |
2781 | *head = entry; | |
cbae3960 | 2782 | |
4c7185a0 | 2783 | xhv->xhv_keys++; /* HvTOTALKEYS(hv)++ */ |
263cb4a6 | 2784 | if (!next) { /* initial entry? */ |
1b95d04f | 2785 | } else if (xhv->xhv_keys > xhv->xhv_max /* HvUSEDKEYS(hv) > HvMAX(hv) */) { |
cbec9347 | 2786 | hsplit(PL_strtab); |
bbce6d69 | 2787 | } |
2788 | } | |
2789 | ||
de616631 | 2790 | ++entry->he_valu.hent_refcount; |
19692e8d NC |
2791 | |
2792 | if (flags & HVhek_FREEKEY) | |
f9a63242 | 2793 | Safefree(str); |
19692e8d | 2794 | |
6e838c70 | 2795 | return HeKEY_hek(entry); |
fde52b5c | 2796 | } |
ecae49c0 | 2797 | |
ca732855 NC |
2798 | I32 * |
2799 | Perl_hv_placeholders_p(pTHX_ HV *hv) | |
2800 | { | |
2801 | dVAR; | |
ad64d0ec | 2802 | MAGIC *mg = mg_find((const SV *)hv, PERL_MAGIC_rhash); |
ca732855 | 2803 | |
7918f24d NC |
2804 | PERL_ARGS_ASSERT_HV_PLACEHOLDERS_P; |
2805 | ||
ca732855 | 2806 | if (!mg) { |
ad64d0ec | 2807 | mg = sv_magicext(MUTABLE_SV(hv), 0, PERL_MAGIC_rhash, 0, 0, 0); |
ca732855 NC |
2808 | |
2809 | if (!mg) { | |
2810 | Perl_die(aTHX_ "panic: hv_placeholders_p"); | |
2811 | } | |
2812 | } | |
2813 | return &(mg->mg_len); | |
2814 | } | |
2815 | ||
2816 | ||
2817 | I32 | |
0c289d13 | 2818 | Perl_hv_placeholders_get(pTHX_ const HV *hv) |
ca732855 NC |
2819 | { |
2820 | dVAR; | |
0c289d13 | 2821 | MAGIC * const mg = mg_find((const SV *)hv, PERL_MAGIC_rhash); |
ca732855 | 2822 | |
7918f24d NC |
2823 | PERL_ARGS_ASSERT_HV_PLACEHOLDERS_GET; |
2824 | ||
ca732855 NC |
2825 | return mg ? mg->mg_len : 0; |
2826 | } | |
2827 | ||
2828 | void | |
ac1e784a | 2829 | Perl_hv_placeholders_set(pTHX_ HV *hv, I32 ph) |
ca732855 NC |
2830 | { |
2831 | dVAR; | |
ad64d0ec | 2832 | MAGIC * const mg = mg_find((const SV *)hv, PERL_MAGIC_rhash); |
ca732855 | 2833 | |
7918f24d NC |
2834 | PERL_ARGS_ASSERT_HV_PLACEHOLDERS_SET; |
2835 | ||
ca732855 NC |
2836 | if (mg) { |
2837 | mg->mg_len = ph; | |
2838 | } else if (ph) { | |
ad64d0ec | 2839 | if (!sv_magicext(MUTABLE_SV(hv), 0, PERL_MAGIC_rhash, 0, 0, ph)) |
ca732855 NC |
2840 | Perl_die(aTHX_ "panic: hv_placeholders_set"); |
2841 | } | |
2842 | /* else we don't need to add magic to record 0 placeholders. */ | |
2843 | } | |
ecae49c0 | 2844 | |
2a49f0f5 | 2845 | STATIC SV * |
7b0bddfa NC |
2846 | S_refcounted_he_value(pTHX_ const struct refcounted_he *he) |
2847 | { | |
0b2d3faa | 2848 | dVAR; |
7b0bddfa | 2849 | SV *value; |
7918f24d NC |
2850 | |
2851 | PERL_ARGS_ASSERT_REFCOUNTED_HE_VALUE; | |
2852 | ||
7b0bddfa NC |
2853 | switch(he->refcounted_he_data[0] & HVrhek_typemask) { |
2854 | case HVrhek_undef: | |
2855 | value = newSV(0); | |
2856 | break; | |
2857 | case HVrhek_delete: | |
2858 | value = &PL_sv_placeholder; | |
2859 | break; | |
2860 | case HVrhek_IV: | |
44ebaf21 NC |
2861 | value = newSViv(he->refcounted_he_val.refcounted_he_u_iv); |
2862 | break; | |
2863 | case HVrhek_UV: | |
2864 | value = newSVuv(he->refcounted_he_val.refcounted_he_u_uv); | |
7b0bddfa NC |
2865 | break; |
2866 | case HVrhek_PV: | |
44ebaf21 | 2867 | case HVrhek_PV_UTF8: |
7b0bddfa NC |
2868 | /* Create a string SV that directly points to the bytes in our |
2869 | structure. */ | |
b9f83d2f | 2870 | value = newSV_type(SVt_PV); |
7b0bddfa NC |
2871 | SvPV_set(value, (char *) he->refcounted_he_data + 1); |
2872 | SvCUR_set(value, he->refcounted_he_val.refcounted_he_u_len); | |
2873 | /* This stops anything trying to free it */ | |
2874 | SvLEN_set(value, 0); | |
2875 | SvPOK_on(value); | |
2876 | SvREADONLY_on(value); | |
44ebaf21 | 2877 | if ((he->refcounted_he_data[0] & HVrhek_typemask) == HVrhek_PV_UTF8) |
7b0bddfa NC |
2878 | SvUTF8_on(value); |
2879 | break; | |
2880 | default: | |
20439bc7 Z |
2881 | Perl_croak(aTHX_ "panic: refcounted_he_value bad flags %"UVxf, |
2882 | (UV)he->refcounted_he_data[0]); | |
7b0bddfa NC |
2883 | } |
2884 | return value; | |
2885 | } | |
2886 | ||
ecae49c0 | 2887 | /* |
20439bc7 | 2888 | =for apidoc m|HV *|refcounted_he_chain_2hv|const struct refcounted_he *c|U32 flags |
8dff4fc5 | 2889 | |
20439bc7 Z |
2890 | Generates and returns a C<HV *> representing the content of a |
2891 | C<refcounted_he> chain. | |
2892 | I<flags> is currently unused and must be zero. | |
8dff4fc5 BM |
2893 | |
2894 | =cut | |
2895 | */ | |
2896 | HV * | |
20439bc7 | 2897 | Perl_refcounted_he_chain_2hv(pTHX_ const struct refcounted_he *chain, U32 flags) |
8dff4fc5 | 2898 | { |
20439bc7 Z |
2899 | dVAR; |
2900 | HV *hv; | |
2901 | U32 placeholders, max; | |
b3ca2e83 | 2902 | |
20439bc7 Z |
2903 | if (flags) |
2904 | Perl_croak(aTHX_ "panic: refcounted_he_chain_2hv bad flags %"UVxf, | |
2905 | (UV)flags); | |
b3ca2e83 | 2906 | |
b3ca2e83 NC |
2907 | /* We could chase the chain once to get an idea of the number of keys, |
2908 | and call ksplit. But for now we'll make a potentially inefficient | |
2909 | hash with only 8 entries in its array. */ | |
20439bc7 Z |
2910 | hv = newHV(); |
2911 | max = HvMAX(hv); | |
b3ca2e83 NC |
2912 | if (!HvARRAY(hv)) { |
2913 | char *array; | |
2914 | Newxz(array, PERL_HV_ARRAY_ALLOC_BYTES(max + 1), char); | |
2915 | HvARRAY(hv) = (HE**)array; | |
2916 | } | |
2917 | ||
20439bc7 | 2918 | placeholders = 0; |
b3ca2e83 | 2919 | while (chain) { |
cbb1fbea | 2920 | #ifdef USE_ITHREADS |
b6bbf3fa | 2921 | U32 hash = chain->refcounted_he_hash; |
cbb1fbea NC |
2922 | #else |
2923 | U32 hash = HEK_HASH(chain->refcounted_he_hek); | |
2924 | #endif | |
b3ca2e83 NC |
2925 | HE **oentry = &((HvARRAY(hv))[hash & max]); |
2926 | HE *entry = *oentry; | |
b6bbf3fa | 2927 | SV *value; |
cbb1fbea | 2928 | |
b3ca2e83 NC |
2929 | for (; entry; entry = HeNEXT(entry)) { |
2930 | if (HeHASH(entry) == hash) { | |
9f769845 NC |
2931 | /* We might have a duplicate key here. If so, entry is older |
2932 | than the key we've already put in the hash, so if they are | |
2933 | the same, skip adding entry. */ | |
2934 | #ifdef USE_ITHREADS | |
2935 | const STRLEN klen = HeKLEN(entry); | |
2936 | const char *const key = HeKEY(entry); | |
2937 | if (klen == chain->refcounted_he_keylen | |
2938 | && (!!HeKUTF8(entry) | |
2939 | == !!(chain->refcounted_he_data[0] & HVhek_UTF8)) | |
2940 | && memEQ(key, REF_HE_KEY(chain), klen)) | |
2941 | goto next_please; | |
2942 | #else | |
2943 | if (HeKEY_hek(entry) == chain->refcounted_he_hek) | |
2944 | goto next_please; | |
2945 | if (HeKLEN(entry) == HEK_LEN(chain->refcounted_he_hek) | |
2946 | && HeKUTF8(entry) == HEK_UTF8(chain->refcounted_he_hek) | |
2947 | && memEQ(HeKEY(entry), HEK_KEY(chain->refcounted_he_hek), | |
2948 | HeKLEN(entry))) | |
2949 | goto next_please; | |
2950 | #endif | |
b3ca2e83 NC |
2951 | } |
2952 | } | |
2953 | assert (!entry); | |
2954 | entry = new_HE(); | |
2955 | ||
cbb1fbea NC |
2956 | #ifdef USE_ITHREADS |
2957 | HeKEY_hek(entry) | |
7b0bddfa | 2958 | = share_hek_flags(REF_HE_KEY(chain), |
b6bbf3fa NC |
2959 | chain->refcounted_he_keylen, |
2960 | chain->refcounted_he_hash, | |
2961 | (chain->refcounted_he_data[0] | |
2962 | & (HVhek_UTF8|HVhek_WASUTF8))); | |
cbb1fbea | 2963 | #else |
71ad1b0c | 2964 | HeKEY_hek(entry) = share_hek_hek(chain->refcounted_he_hek); |
cbb1fbea | 2965 | #endif |
7b0bddfa NC |
2966 | value = refcounted_he_value(chain); |
2967 | if (value == &PL_sv_placeholder) | |
b3ca2e83 | 2968 | placeholders++; |
b6bbf3fa | 2969 | HeVAL(entry) = value; |
b3ca2e83 NC |
2970 | |
2971 | /* Link it into the chain. */ | |
2972 | HeNEXT(entry) = *oentry; | |
b3ca2e83 NC |
2973 | *oentry = entry; |
2974 | ||
2975 | HvTOTALKEYS(hv)++; | |
2976 | ||
2977 | next_please: | |
71ad1b0c | 2978 | chain = chain->refcounted_he_next; |
b3ca2e83 NC |
2979 | } |
2980 | ||
2981 | if (placeholders) { | |
2982 | clear_placeholders(hv, placeholders); | |
2983 | HvTOTALKEYS(hv) -= placeholders; | |
2984 | } | |
2985 | ||
2986 | /* We could check in the loop to see if we encounter any keys with key | |
2987 | flags, but it's probably not worth it, as this per-hash flag is only | |
2988 | really meant as an optimisation for things like Storable. */ | |
2989 | HvHASKFLAGS_on(hv); | |
def9038f | 2990 | DEBUG_A(Perl_hv_assert(aTHX_ hv)); |
b3ca2e83 NC |
2991 | |
2992 | return hv; | |
2993 | } | |
2994 | ||
20439bc7 Z |
2995 | /* |
2996 | =for apidoc m|SV *|refcounted_he_fetch_pvn|const struct refcounted_he *chain|const char *keypv|STRLEN keylen|U32 hash|U32 flags | |
2997 | ||
2998 | Search along a C<refcounted_he> chain for an entry with the key specified | |
2999 | by I<keypv> and I<keylen>. If I<flags> has the C<REFCOUNTED_HE_KEY_UTF8> | |
3000 | bit set, the key octets are interpreted as UTF-8, otherwise they | |
3001 | are interpreted as Latin-1. I<hash> is a precomputed hash of the key | |
3002 | string, or zero if it has not been precomputed. Returns a mortal scalar | |
3003 | representing the value associated with the key, or C<&PL_sv_placeholder> | |
3004 | if there is no value associated with the key. | |
3005 | ||
3006 | =cut | |
3007 | */ | |
3008 | ||
7b0bddfa | 3009 | SV * |
20439bc7 Z |
3010 | Perl_refcounted_he_fetch_pvn(pTHX_ const struct refcounted_he *chain, |
3011 | const char *keypv, STRLEN keylen, U32 hash, U32 flags) | |
7b0bddfa | 3012 | { |
0b2d3faa | 3013 | dVAR; |
20439bc7 Z |
3014 | U8 utf8_flag; |
3015 | PERL_ARGS_ASSERT_REFCOUNTED_HE_FETCH_PVN; | |
7b0bddfa | 3016 | |
94250aee | 3017 | if (flags & ~(REFCOUNTED_HE_KEY_UTF8|REFCOUNTED_HE_EXISTS)) |
20439bc7 Z |
3018 | Perl_croak(aTHX_ "panic: refcounted_he_fetch_pvn bad flags %"UVxf, |
3019 | (UV)flags); | |
3020 | if (!chain) | |
3021 | return &PL_sv_placeholder; | |
3022 | if (flags & REFCOUNTED_HE_KEY_UTF8) { | |
3023 | /* For searching purposes, canonicalise to Latin-1 where possible. */ | |
3024 | const char *keyend = keypv + keylen, *p; | |
3025 | STRLEN nonascii_count = 0; | |
3026 | for (p = keypv; p != keyend; p++) { | |
3027 | U8 c = (U8)*p; | |
3028 | if (c & 0x80) { | |
3029 | if (!((c & 0xfe) == 0xc2 && ++p != keyend && | |
3030 | (((U8)*p) & 0xc0) == 0x80)) | |
3031 | goto canonicalised_key; | |
3032 | nonascii_count++; | |
3033 | } | |
cd1d2f8a | 3034 | } |
20439bc7 Z |
3035 | if (nonascii_count) { |
3036 | char *q; | |
3037 | const char *p = keypv, *keyend = keypv + keylen; | |
3038 | keylen -= nonascii_count; | |
3039 | Newx(q, keylen, char); | |
3040 | SAVEFREEPV(q); | |
3041 | keypv = q; | |
3042 | for (; p != keyend; p++, q++) { | |
3043 | U8 c = (U8)*p; | |
3044 | *q = (char) | |
3045 | ((c & 0x80) ? ((c & 0x03) << 6) | (((U8)*++p) & 0x3f) : c); | |
cd1d2f8a NC |
3046 | } |
3047 | } | |
20439bc7 Z |
3048 | flags &= ~REFCOUNTED_HE_KEY_UTF8; |
3049 | canonicalised_key: ; | |
3050 | } | |
3051 | utf8_flag = (flags & REFCOUNTED_HE_KEY_UTF8) ? HVhek_UTF8 : 0; | |
3052 | if (!hash) | |
3053 | PERL_HASH(hash, keypv, keylen); | |
7b0bddfa | 3054 | |
20439bc7 Z |
3055 | for (; chain; chain = chain->refcounted_he_next) { |
3056 | if ( | |
7b0bddfa | 3057 | #ifdef USE_ITHREADS |
20439bc7 Z |
3058 | hash == chain->refcounted_he_hash && |
3059 | keylen == chain->refcounted_he_keylen && | |
3060 | memEQ(REF_HE_KEY(chain), keypv, keylen) && | |
3061 | utf8_flag == (chain->refcounted_he_data[0] & HVhek_UTF8) | |
7b0bddfa | 3062 | #else |
20439bc7 Z |
3063 | hash == HEK_HASH(chain->refcounted_he_hek) && |
3064 | keylen == (STRLEN)HEK_LEN(chain->refcounted_he_hek) && | |
3065 | memEQ(HEK_KEY(chain->refcounted_he_hek), keypv, keylen) && | |
3066 | utf8_flag == (HEK_FLAGS(chain->refcounted_he_hek) & HVhek_UTF8) | |
7b0bddfa | 3067 | #endif |
ef8156f5 NC |
3068 | ) { |
3069 | if (flags & REFCOUNTED_HE_EXISTS) | |
3070 | return (chain->refcounted_he_data[0] & HVrhek_typemask) | |
3071 | == HVrhek_delete | |
3072 | ? NULL : &PL_sv_yes; | |
3073 | return sv_2mortal(refcounted_he_value(chain)); | |
3074 | } | |
94250aee FC |
3075 | } |
3076 | return flags & REFCOUNTED_HE_EXISTS ? NULL : &PL_sv_placeholder; | |
20439bc7 | 3077 | } |
7b0bddfa | 3078 | |
20439bc7 Z |
3079 | /* |
3080 | =for apidoc m|SV *|refcounted_he_fetch_pv|const struct refcounted_he *chain|const char *key|U32 hash|U32 flags | |
7b0bddfa | 3081 | |
20439bc7 Z |
3082 | Like L</refcounted_he_fetch_pvn>, but takes a nul-terminated string |
3083 | instead of a string/length pair. | |
3084 | ||
3085 | =cut | |
3086 | */ | |
3087 | ||
3088 | SV * | |
3089 | Perl_refcounted_he_fetch_pv(pTHX_ const struct refcounted_he *chain, | |
3090 | const char *key, U32 hash, U32 flags) | |
3091 | { | |
3092 | PERL_ARGS_ASSERT_REFCOUNTED_HE_FETCH_PV; | |
3093 | return refcounted_he_fetch_pvn(chain, key, strlen(key), hash, flags); | |
7b0bddfa NC |
3094 | } |
3095 | ||
b3ca2e83 | 3096 | /* |
20439bc7 Z |
3097 | =for apidoc m|SV *|refcounted_he_fetch_sv|const struct refcounted_he *chain|SV *key|U32 hash|U32 flags |
3098 | ||
3099 | Like L</refcounted_he_fetch_pvn>, but takes a Perl scalar instead of a | |
3100 | string/length pair. | |
3101 | ||
3102 | =cut | |
3103 | */ | |
b3ca2e83 | 3104 | |
20439bc7 Z |
3105 | SV * |
3106 | Perl_refcounted_he_fetch_sv(pTHX_ const struct refcounted_he *chain, | |
3107 | SV *key, U32 hash, U32 flags) | |
3108 | { | |
3109 | const char *keypv; | |
3110 | STRLEN keylen; | |
3111 | PERL_ARGS_ASSERT_REFCOUNTED_HE_FETCH_SV; | |
3112 | if (flags & REFCOUNTED_HE_KEY_UTF8) | |
3113 | Perl_croak(aTHX_ "panic: refcounted_he_fetch_sv bad flags %"UVxf, | |
3114 | (UV)flags); | |
3115 | keypv = SvPV_const(key, keylen); | |
3116 | if (SvUTF8(key)) | |
3117 | flags |= REFCOUNTED_HE_KEY_UTF8; | |
3118 | if (!hash && SvIsCOW_shared_hash(key)) | |
3119 | hash = SvSHARED_HASH(key); | |
3120 | return refcounted_he_fetch_pvn(chain, keypv, keylen, hash, flags); | |
3121 | } | |
3122 | ||
3123 | /* | |
3124 | =for apidoc m|struct refcounted_he *|refcounted_he_new_pvn|struct refcounted_he *parent|const char *keypv|STRLEN keylen|U32 hash|SV *value|U32 flags | |
3125 | ||
3126 | Creates a new C<refcounted_he>. This consists of a single key/value | |
3127 | pair and a reference to an existing C<refcounted_he> chain (which may | |
3128 | be empty), and thus forms a longer chain. When using the longer chain, | |
3129 | the new key/value pair takes precedence over any entry for the same key | |
3130 | further along the chain. | |
3131 | ||
3132 | The new key is specified by I<keypv> and I<keylen>. If I<flags> has | |
3133 | the C<REFCOUNTED_HE_KEY_UTF8> bit set, the key octets are interpreted | |
3134 | as UTF-8, otherwise they are interpreted as Latin-1. I<hash> is | |
3135 | a precomputed hash of the key string, or zero if it has not been | |
3136 | precomputed. | |
3137 | ||
3138 | I<value> is the scalar value to store for this key. I<value> is copied | |
3139 | by this function, which thus does not take ownership of any reference | |
3140 | to it, and later changes to the scalar will not be reflected in the | |
3141 | value visible in the C<refcounted_he>. Complex types of scalar will not | |
3142 | be stored with referential integrity, but will be coerced to strings. | |
3143 | I<value> may be either null or C<&PL_sv_placeholder> to indicate that no | |
3144 | value is to be associated with the key; this, as with any non-null value, | |
3145 | takes precedence over the existence of a value for the key further along | |
3146 | the chain. | |
3147 | ||
3148 | I<parent> points to the rest of the C<refcounted_he> chain to be | |
3149 | attached to the new C<refcounted_he>. This function takes ownership | |
3150 | of one reference to I<parent>, and returns one reference to the new | |
3151 | C<refcounted_he>. | |
b3ca2e83 NC |
3152 | |
3153 | =cut | |
3154 | */ | |
3155 | ||
3156 | struct refcounted_he * | |
20439bc7 Z |
3157 | Perl_refcounted_he_new_pvn(pTHX_ struct refcounted_he *parent, |
3158 | const char *keypv, STRLEN keylen, U32 hash, SV *value, U32 flags) | |
3159 | { | |
7a89be66 | 3160 | dVAR; |
b6bbf3fa | 3161 | STRLEN value_len = 0; |
95b63a38 | 3162 | const char *value_p = NULL; |
20439bc7 | 3163 | bool is_pv; |
b6bbf3fa | 3164 | char value_type; |
20439bc7 Z |
3165 | char hekflags; |
3166 | STRLEN key_offset = 1; | |
3167 | struct refcounted_he *he; | |
3168 | PERL_ARGS_ASSERT_REFCOUNTED_HE_NEW_PVN; | |
b6bbf3fa | 3169 | |
20439bc7 Z |
3170 | if (!value || value == &PL_sv_placeholder) { |
3171 | value_type = HVrhek_delete; | |
3172 | } else if (SvPOK(value)) { | |
b6bbf3fa NC |
3173 | value_type = HVrhek_PV; |
3174 | } else if (SvIOK(value)) { | |
ad64d0ec | 3175 | value_type = SvUOK((const SV *)value) ? HVrhek_UV : HVrhek_IV; |
b6bbf3fa NC |
3176 | } else if (!SvOK(value)) { |
3177 | value_type = HVrhek_undef; | |
3178 | } else { | |
3179 | value_type = HVrhek_PV; | |
3180 | } | |
20439bc7 Z |
3181 | is_pv = value_type == HVrhek_PV; |
3182 | if (is_pv) { | |
012da8e5 NC |
3183 | /* Do it this way so that the SvUTF8() test is after the SvPV, in case |
3184 | the value is overloaded, and doesn't yet have the UTF-8flag set. */ | |
b6bbf3fa | 3185 | value_p = SvPV_const(value, value_len); |
012da8e5 NC |
3186 | if (SvUTF8(value)) |
3187 | value_type = HVrhek_PV_UTF8; | |
20439bc7 Z |
3188 | key_offset = value_len + 2; |
3189 | } | |
3190 | hekflags = value_type; | |
3191 | ||
3192 | if (flags & REFCOUNTED_HE_KEY_UTF8) { | |
3193 | /* Canonicalise to Latin-1 where possible. */ | |
3194 | const char *keyend = keypv + keylen, *p; | |
3195 | STRLEN nonascii_count = 0; | |
3196 | for (p = keypv; p != keyend; p++) { | |
3197 | U8 c = (U8)*p; | |
3198 | if (c & 0x80) { | |
3199 | if (!((c & 0xfe) == 0xc2 && ++p != keyend && | |
3200 | (((U8)*p) & 0xc0) == 0x80)) | |
3201 | goto canonicalised_key; | |
3202 | nonascii_count++; | |
3203 | } | |
3204 | } | |
3205 | if (nonascii_count) { | |
3206 | char *q; | |
3207 | const char *p = keypv, *keyend = keypv + keylen; | |
3208 | keylen -= nonascii_count; | |
3209 | Newx(q, keylen, char); | |
3210 | SAVEFREEPV(q); | |
3211 | keypv = q; | |
3212 | for (; p != keyend; p++, q++) { | |
3213 | U8 c = (U8)*p; | |
3214 | *q = (char) | |
3215 | ((c & 0x80) ? ((c & 0x03) << 6) | (((U8)*++p) & 0x3f) : c); | |
3216 | } | |
3217 | } | |
3218 | flags &= ~REFCOUNTED_HE_KEY_UTF8; | |
3219 | canonicalised_key: ; | |
b6bbf3fa | 3220 | } |
20439bc7 Z |
3221 | if (flags & REFCOUNTED_HE_KEY_UTF8) |
3222 | hekflags |= HVhek_UTF8; | |
3223 | if (!hash) | |
3224 | PERL_HASH(hash, keypv, keylen); | |
012da8e5 | 3225 | |
0de694c5 | 3226 | #ifdef USE_ITHREADS |
10edeb5d JH |
3227 | he = (struct refcounted_he*) |
3228 | PerlMemShared_malloc(sizeof(struct refcounted_he) - 1 | |
20439bc7 | 3229 | + keylen |
20439bc7 | 3230 | + key_offset); |
0de694c5 NC |
3231 | #else |
3232 | he = (struct refcounted_he*) | |
3233 | PerlMemShared_malloc(sizeof(struct refcounted_he) - 1 | |
3234 | + key_offset); | |
3235 | #endif | |
b3ca2e83 | 3236 | |
71ad1b0c | 3237 | he->refcounted_he_next = parent; |
b6bbf3fa | 3238 | |
012da8e5 | 3239 | if (is_pv) { |
20439bc7 | 3240 | Copy(value_p, he->refcounted_he_data + 1, value_len + 1, char); |
b6bbf3fa | 3241 | he->refcounted_he_val.refcounted_he_u_len = value_len; |
b6bbf3fa | 3242 | } else if (value_type == HVrhek_IV) { |
20439bc7 | 3243 | he->refcounted_he_val.refcounted_he_u_iv = SvIVX(value); |
012da8e5 | 3244 | } else if (value_type == HVrhek_UV) { |
20439bc7 | 3245 | he->refcounted_he_val.refcounted_he_u_uv = SvUVX(value); |
b6bbf3fa NC |
3246 | } |
3247 | ||
cbb1fbea | 3248 | #ifdef USE_ITHREADS |
b6bbf3fa | 3249 | he->refcounted_he_hash = hash; |
20439bc7 Z |
3250 | he->refcounted_he_keylen = keylen; |
3251 | Copy(keypv, he->refcounted_he_data + key_offset, keylen, char); | |
cbb1fbea | 3252 | #else |
20439bc7 | 3253 | he->refcounted_he_hek = share_hek_flags(keypv, keylen, hash, hekflags); |
cbb1fbea | 3254 | #endif |
b6bbf3fa | 3255 | |
20439bc7 | 3256 | he->refcounted_he_data[0] = hekflags; |
b3ca2e83 NC |
3257 | he->refcounted_he_refcnt = 1; |
3258 | ||
3259 | return he; | |
3260 | } | |
3261 | ||
3262 | /* | |
20439bc7 | 3263 | =for apidoc m|struct refcounted_he *|refcounted_he_new_pv|struct refcounted_he *parent|const char *key|U32 hash|SV *value|U32 flags |
b3ca2e83 | 3264 | |
20439bc7 Z |
3265 | Like L</refcounted_he_new_pvn>, but takes a nul-terminated string instead |
3266 | of a string/length pair. | |
3267 | ||
3268 | =cut | |
3269 | */ | |
3270 | ||
3271 | struct refcounted_he * | |
3272 | Perl_refcounted_he_new_pv(pTHX_ struct refcounted_he *parent, | |
3273 | const char *key, U32 hash, SV *value, U32 flags) | |
3274 | { | |
3275 | PERL_ARGS_ASSERT_REFCOUNTED_HE_NEW_PV; | |
3276 | return refcounted_he_new_pvn(parent, key, strlen(key), hash, value, flags); | |
3277 | } | |
3278 | ||
3279 | /* | |
3280 | =for apidoc m|struct refcounted_he *|refcounted_he_new_sv|struct refcounted_he *parent|SV *key|U32 hash|SV *value|U32 flags | |
3281 | ||
3282 | Like L</refcounted_he_new_pvn>, but takes a Perl scalar instead of a | |
3283 | string/length pair. | |
3284 | ||
3285 | =cut | |
3286 | */ | |
3287 | ||
3288 | struct refcounted_he * | |
3289 | Perl_refcounted_he_new_sv(pTHX_ struct refcounted_he *parent, | |
3290 | SV *key, U32 hash, SV *value, U32 flags) | |
3291 | { | |
3292 | const char *keypv; | |
3293 | STRLEN keylen; | |
3294 | PERL_ARGS_ASSERT_REFCOUNTED_HE_NEW_SV; | |
3295 | if (flags & REFCOUNTED_HE_KEY_UTF8) | |
3296 | Perl_croak(aTHX_ "panic: refcounted_he_new_sv bad flags %"UVxf, | |
3297 | (UV)flags); | |
3298 | keypv = SvPV_const(key, keylen); | |
3299 | if (SvUTF8(key)) | |
3300 | flags |= REFCOUNTED_HE_KEY_UTF8; | |
3301 | if (!hash && SvIsCOW_shared_hash(key)) | |
3302 | hash = SvSHARED_HASH(key); | |
3303 | return refcounted_he_new_pvn(parent, keypv, keylen, hash, value, flags); | |
3304 | } | |
3305 | ||
3306 | /* | |
3307 | =for apidoc m|void|refcounted_he_free|struct refcounted_he *he | |
3308 | ||
3309 | Decrements the reference count of a C<refcounted_he> by one. If the | |
3310 | reference count reaches zero the structure's memory is freed, which | |
3311 | (recursively) causes a reduction of its parent C<refcounted_he>'s | |
3312 | reference count. It is safe to pass a null pointer to this function: | |
3313 | no action occurs in this case. | |
b3ca2e83 NC |
3314 | |
3315 | =cut | |
3316 | */ | |
3317 | ||
3318 | void | |
3319 | Perl_refcounted_he_free(pTHX_ struct refcounted_he *he) { | |
53d44271 | 3320 | dVAR; |
57ca3b03 AL |
3321 | PERL_UNUSED_CONTEXT; |
3322 | ||
b3ca2e83 NC |
3323 | while (he) { |
3324 | struct refcounted_he *copy; | |
cbb1fbea | 3325 | U32 new_count; |
b3ca2e83 | 3326 | |
cbb1fbea NC |
3327 | HINTS_REFCNT_LOCK; |
3328 | new_count = --he->refcounted_he_refcnt; | |
3329 | HINTS_REFCNT_UNLOCK; | |
3330 | ||
3331 | if (new_count) { | |
b3ca2e83 | 3332 | return; |
cbb1fbea | 3333 | } |
b3ca2e83 | 3334 | |
b6bbf3fa | 3335 | #ifndef USE_ITHREADS |
71ad1b0c | 3336 | unshare_hek_or_pvn (he->refcounted_he_hek, 0, 0, 0); |
cbb1fbea | 3337 | #endif |
b3ca2e83 | 3338 | copy = he; |
71ad1b0c | 3339 | he = he->refcounted_he_next; |
b6bbf3fa | 3340 | PerlMemShared_free(copy); |
b3ca2e83 NC |
3341 | } |
3342 | } | |
3343 | ||
20439bc7 Z |
3344 | /* |
3345 | =for apidoc m|struct refcounted_he *|refcounted_he_inc|struct refcounted_he *he | |
3346 | ||
3347 | Increment the reference count of a C<refcounted_he>. The pointer to the | |
3348 | C<refcounted_he> is also returned. It is safe to pass a null pointer | |
3349 | to this function: no action occurs and a null pointer is returned. | |
3350 | ||
3351 | =cut | |
3352 | */ | |
3353 | ||
3354 | struct refcounted_he * | |
3355 | Perl_refcounted_he_inc(pTHX_ struct refcounted_he *he) | |
3356 | { | |
09ddd873 | 3357 | dVAR; |
20439bc7 Z |
3358 | if (he) { |
3359 | HINTS_REFCNT_LOCK; | |
3360 | he->refcounted_he_refcnt++; | |
3361 | HINTS_REFCNT_UNLOCK; | |
3362 | } | |
3363 | return he; | |
3364 | } | |
3365 | ||
8375c93e | 3366 | /* |
aebc0cbe | 3367 | =for apidoc cop_fetch_label |
8375c93e RU |
3368 | |
3369 | Returns the label attached to a cop. | |
3370 | The flags pointer may be set to C<SVf_UTF8> or 0. | |
3371 | ||
3372 | =cut | |
3373 | */ | |
3374 | ||
47550813 NC |
3375 | /* pp_entereval is aware that labels are stored with a key ':' at the top of |
3376 | the linked list. */ | |
dca6062a | 3377 | const char * |
aebc0cbe | 3378 | Perl_cop_fetch_label(pTHX_ COP *const cop, STRLEN *len, U32 *flags) { |
d6747b7a NC |
3379 | struct refcounted_he *const chain = cop->cop_hints_hash; |
3380 | ||
aebc0cbe | 3381 | PERL_ARGS_ASSERT_COP_FETCH_LABEL; |
d6747b7a | 3382 | |
dca6062a NC |
3383 | if (!chain) |
3384 | return NULL; | |
3385 | #ifdef USE_ITHREADS | |
3386 | if (chain->refcounted_he_keylen != 1) | |
3387 | return NULL; | |
3388 | if (*REF_HE_KEY(chain) != ':') | |
3389 | return NULL; | |
3390 | #else | |
3391 | if ((STRLEN)HEK_LEN(chain->refcounted_he_hek) != 1) | |
3392 | return NULL; | |
3393 | if (*HEK_KEY(chain->refcounted_he_hek) != ':') | |
3394 | return NULL; | |
3395 | #endif | |
012da8e5 NC |
3396 | /* Stop anyone trying to really mess us up by adding their own value for |
3397 | ':' into %^H */ | |
3398 | if ((chain->refcounted_he_data[0] & HVrhek_typemask) != HVrhek_PV | |
3399 | && (chain->refcounted_he_data[0] & HVrhek_typemask) != HVrhek_PV_UTF8) | |
3400 | return NULL; | |
3401 | ||
dca6062a NC |
3402 | if (len) |
3403 | *len = chain->refcounted_he_val.refcounted_he_u_len; | |
3404 | if (flags) { | |
3405 | *flags = ((chain->refcounted_he_data[0] & HVrhek_typemask) | |
3406 | == HVrhek_PV_UTF8) ? SVf_UTF8 : 0; | |
3407 | } | |
3408 | return chain->refcounted_he_data + 1; | |
3409 | } | |
3410 | ||
8375c93e | 3411 | /* |
aebc0cbe | 3412 | =for apidoc cop_store_label |
8375c93e RU |
3413 | |
3414 | Save a label into a C<cop_hints_hash>. You need to set flags to C<SVf_UTF8> | |
3415 | for a utf-8 label. | |
3416 | ||
3417 | =cut | |
3418 | */ | |
3419 | ||
a77ac40c | 3420 | void |
aebc0cbe | 3421 | Perl_cop_store_label(pTHX_ COP *const cop, const char *label, STRLEN len, |
a77ac40c | 3422 | U32 flags) |
012da8e5 | 3423 | { |
20439bc7 | 3424 | SV *labelsv; |
aebc0cbe | 3425 | PERL_ARGS_ASSERT_COP_STORE_LABEL; |
547bb267 | 3426 | |
a77ac40c | 3427 | if (flags & ~(SVf_UTF8)) |
aebc0cbe | 3428 | Perl_croak(aTHX_ "panic: cop_store_label illegal flag bits 0x%" UVxf, |
a77ac40c | 3429 | (UV)flags); |
a3179684 | 3430 | labelsv = newSVpvn_flags(label, len, SVs_TEMP); |
20439bc7 Z |
3431 | if (flags & SVf_UTF8) |
3432 | SvUTF8_on(labelsv); | |
a77ac40c | 3433 | cop->cop_hints_hash |
20439bc7 | 3434 | = refcounted_he_new_pvs(cop->cop_hints_hash, ":", labelsv, 0); |
012da8e5 NC |
3435 | } |
3436 | ||
b3ca2e83 | 3437 | /* |
ecae49c0 NC |
3438 | =for apidoc hv_assert |
3439 | ||
3440 | Check that a hash is in an internally consistent state. | |
3441 | ||
3442 | =cut | |
3443 | */ | |
3444 | ||
943795c2 NC |
3445 | #ifdef DEBUGGING |
3446 | ||
ecae49c0 NC |
3447 | void |
3448 | Perl_hv_assert(pTHX_ HV *hv) | |
3449 | { | |
57ca3b03 AL |
3450 | dVAR; |
3451 | HE* entry; | |
3452 | int withflags = 0; | |
3453 | int placeholders = 0; | |
3454 | int real = 0; | |
3455 | int bad = 0; | |
3456 | const I32 riter = HvRITER_get(hv); | |
3457 | HE *eiter = HvEITER_get(hv); | |
3458 | ||
7918f24d NC |
3459 | PERL_ARGS_ASSERT_HV_ASSERT; |
3460 | ||
57ca3b03 AL |
3461 | (void)hv_iterinit(hv); |
3462 | ||
3463 | while ((entry = hv_iternext_flags(hv, HV_ITERNEXT_WANTPLACEHOLDERS))) { | |
3464 | /* sanity check the values */ | |
3465 | if (HeVAL(entry) == &PL_sv_placeholder) | |
3466 | placeholders++; | |
3467 | else | |
3468 | real++; | |
3469 | /* sanity check the keys */ | |
3470 | if (HeSVKEY(entry)) { | |
6f207bd3 | 3471 | NOOP; /* Don't know what to check on SV keys. */ |
57ca3b03 AL |
3472 | } else if (HeKUTF8(entry)) { |
3473 | withflags++; | |
3474 | if (HeKWASUTF8(entry)) { | |
3475 | PerlIO_printf(Perl_debug_log, | |
d2a455e7 | 3476 | "hash key has both WASUTF8 and UTF8: '%.*s'\n", |
57ca3b03 AL |
3477 | (int) HeKLEN(entry), HeKEY(entry)); |
3478 | bad = 1; | |
3479 | } | |
3480 | } else if (HeKWASUTF8(entry)) | |
3481 | withflags++; | |
3482 | } | |
ad64d0ec | 3483 | if (!SvTIED_mg((const SV *)hv, PERL_MAGIC_tied)) { |
57ca3b03 AL |
3484 | static const char bad_count[] = "Count %d %s(s), but hash reports %d\n"; |
3485 | const int nhashkeys = HvUSEDKEYS(hv); | |
3486 | const int nhashplaceholders = HvPLACEHOLDERS_get(hv); | |
3487 | ||
3488 | if (nhashkeys != real) { | |
3489 | PerlIO_printf(Perl_debug_log, bad_count, real, "keys", nhashkeys ); | |
3490 | bad = 1; | |
3491 | } | |
3492 | if (nhashplaceholders != placeholders) { | |
3493 | PerlIO_printf(Perl_debug_log, bad_count, placeholders, "placeholder", nhashplaceholders ); | |
3494 | bad = 1; | |
3495 | } | |
3496 | } | |
3497 | if (withflags && ! HvHASKFLAGS(hv)) { | |
3498 | PerlIO_printf(Perl_debug_log, | |
3499 | "Hash has HASKFLAGS off but I count %d key(s) with flags\n", | |
3500 | withflags); | |
3501 | bad = 1; | |
3502 | } | |
3503 | if (bad) { | |
ad64d0ec | 3504 | sv_dump(MUTABLE_SV(hv)); |
57ca3b03 AL |
3505 | } |
3506 | HvRITER_set(hv, riter); /* Restore hash iterator state */ | |
3507 | HvEITER_set(hv, eiter); | |
ecae49c0 | 3508 | } |
af3babe4 | 3509 | |
943795c2 NC |
3510 | #endif |
3511 | ||
af3babe4 NC |
3512 | /* |
3513 | * Local variables: | |
3514 | * c-indentation-style: bsd | |
3515 | * c-basic-offset: 4 | |
14d04a33 | 3516 | * indent-tabs-mode: nil |
af3babe4 NC |
3517 | * End: |
3518 | * | |
14d04a33 | 3519 | * ex: set ts=8 sts=4 sw=4 et: |
37442d52 | 3520 | */ |