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 DM |
21 | |
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 | |
24 | array is indexed by the hash function of the key, so each linked list | |
25 | represents all the hash entries with the same hash value. Each HE contains | |
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 | ||
220 | Stores an SV in a hash. The hash key is specified as C<key> and C<klen> is | |
221 | the length of the key. The C<hash> parameter is the precomputed hash | |
222 | value; if it is zero then Perl will compute it. The return value will be | |
223 | NULL if the operation failed or if the value did not need to be actually | |
224 | stored within the hash (as in the case of tied hashes). Otherwise it can | |
225 | be dereferenced to get the original C<SV*>. Note that the caller is | |
226 | responsible for suitably incrementing the reference count of C<val> before | |
227 | the call, and decrementing it if the function returned NULL. Effectively | |
228 | a successful hv_store takes ownership of one reference to C<val>. This is | |
229 | usually what you want; a newly created SV has a reference count of one, so | |
230 | if all your code does is create SVs then store them in a hash, hv_store | |
231 | will own the only reference to the new SV, and your code doesn't need to do | |
232 | anything further to tidy up. hv_store is not implemented as a call to | |
233 | hv_store_ent, and does not create a temporary SV for the key, so if your | |
234 | key data is not already in SV form then use hv_store in preference to | |
235 | hv_store_ent. | |
236 | ||
237 | See L<perlguts/"Understanding the Magic of Tied Hashes and Arrays"> for more | |
238 | information on how to use this function on tied hashes. | |
239 | ||
34a6f7b4 NC |
240 | =for apidoc hv_store_ent |
241 | ||
242 | Stores C<val> in a hash. The hash key is specified as C<key>. The C<hash> | |
243 | parameter is the precomputed hash value; if it is zero then Perl will | |
244 | compute it. The return value is the new hash entry so created. It will be | |
245 | NULL if the operation failed or if the value did not need to be actually | |
246 | stored within the hash (as in the case of tied hashes). Otherwise the | |
247 | contents of the return value can be accessed using the C<He?> macros | |
248 | described here. Note that the caller is responsible for suitably | |
249 | incrementing the reference count of C<val> before the call, and | |
250 | decrementing it if the function returned NULL. Effectively a successful | |
251 | hv_store_ent takes ownership of one reference to C<val>. This is | |
252 | usually what you want; a newly created SV has a reference count of one, so | |
253 | if all your code does is create SVs then store them in a hash, hv_store | |
254 | will own the only reference to the new SV, and your code doesn't need to do | |
255 | anything further to tidy up. Note that hv_store_ent only reads the C<key>; | |
256 | unlike C<val> it does not take ownership of it, so maintaining the correct | |
257 | reference count on C<key> is entirely the caller's responsibility. hv_store | |
258 | is not implemented as a call to hv_store_ent, and does not create a temporary | |
259 | SV for the key, so if your key data is not already in SV form then use | |
260 | hv_store in preference to hv_store_ent. | |
261 | ||
262 | See L<perlguts/"Understanding the Magic of Tied Hashes and Arrays"> for more | |
263 | information on how to use this function on tied hashes. | |
264 | ||
34a6f7b4 NC |
265 | =for apidoc hv_exists |
266 | ||
267 | Returns a boolean indicating whether the specified hash key exists. The | |
268 | C<klen> is the length of the key. | |
269 | ||
954c1994 GS |
270 | =for apidoc hv_fetch |
271 | ||
272 | Returns the SV which corresponds to the specified key in the hash. The | |
273 | C<klen> is the length of the key. If C<lval> is set then the fetch will be | |
274 | part of a store. Check that the return value is non-null before | |
d1be9408 | 275 | dereferencing it to an C<SV*>. |
954c1994 | 276 | |
96f1132b | 277 | See L<perlguts/"Understanding the Magic of Tied Hashes and Arrays"> for more |
954c1994 GS |
278 | information on how to use this function on tied hashes. |
279 | ||
34a6f7b4 NC |
280 | =for apidoc hv_exists_ent |
281 | ||
282 | Returns a boolean indicating whether the specified hash key exists. C<hash> | |
283 | can be a valid precomputed hash value, or 0 to ask for it to be | |
284 | computed. | |
285 | ||
286 | =cut | |
287 | */ | |
288 | ||
d1be9408 | 289 | /* returns an HE * structure with the all fields set */ |
fde52b5c | 290 | /* note that hent_val will be a mortal sv for MAGICAL hashes */ |
954c1994 GS |
291 | /* |
292 | =for apidoc hv_fetch_ent | |
293 | ||
294 | Returns the hash entry which corresponds to the specified key in the hash. | |
295 | C<hash> must be a valid precomputed hash number for the given C<key>, or 0 | |
296 | if you want the function to compute it. IF C<lval> is set then the fetch | |
297 | will be part of a store. Make sure the return value is non-null before | |
b24b84ef | 298 | accessing it. The return value when C<hv> is a tied hash is a pointer to a |
954c1994 | 299 | static location, so be sure to make a copy of the structure if you need to |
1c846c1f | 300 | store it somewhere. |
954c1994 | 301 | |
96f1132b | 302 | See L<perlguts/"Understanding the Magic of Tied Hashes and Arrays"> for more |
954c1994 GS |
303 | information on how to use this function on tied hashes. |
304 | ||
305 | =cut | |
306 | */ | |
307 | ||
a038e571 NC |
308 | /* Common code for hv_delete()/hv_exists()/hv_fetch()/hv_store() */ |
309 | void * | |
310 | Perl_hv_common_key_len(pTHX_ HV *hv, const char *key, I32 klen_i32, | |
311 | const int action, SV *val, const U32 hash) | |
312 | { | |
313 | STRLEN klen; | |
314 | int flags; | |
315 | ||
7918f24d NC |
316 | PERL_ARGS_ASSERT_HV_COMMON_KEY_LEN; |
317 | ||
a038e571 NC |
318 | if (klen_i32 < 0) { |
319 | klen = -klen_i32; | |
320 | flags = HVhek_UTF8; | |
321 | } else { | |
322 | klen = klen_i32; | |
323 | flags = 0; | |
324 | } | |
325 | return hv_common(hv, NULL, key, klen, flags, action, val, hash); | |
326 | } | |
327 | ||
63c89345 | 328 | void * |
d3ba3f5c NC |
329 | Perl_hv_common(pTHX_ HV *hv, SV *keysv, const char *key, STRLEN klen, |
330 | int flags, int action, SV *val, register U32 hash) | |
113738bb | 331 | { |
27da23d5 | 332 | dVAR; |
b2c64049 | 333 | XPVHV* xhv; |
b2c64049 NC |
334 | HE *entry; |
335 | HE **oentry; | |
fde52b5c | 336 | SV *sv; |
da58a35d | 337 | bool is_utf8; |
113738bb | 338 | int masked_flags; |
3c84c864 | 339 | const int return_svp = action & HV_FETCH_JUST_SV; |
fde52b5c | 340 | |
341 | if (!hv) | |
a4fc7abc | 342 | return NULL; |
8265e3d1 NC |
343 | if (SvTYPE(hv) == SVTYPEMASK) |
344 | return NULL; | |
345 | ||
346 | assert(SvTYPE(hv) == SVt_PVHV); | |
fde52b5c | 347 | |
bdee33e4 | 348 | if (SvSMAGICAL(hv) && SvGMAGICAL(hv) && !(action & HV_DISABLE_UVAR_XKEY)) { |
fda2d18a | 349 | MAGIC* mg; |
ad64d0ec | 350 | if ((mg = mg_find((const SV *)hv, PERL_MAGIC_uvar))) { |
fda2d18a NC |
351 | struct ufuncs * const uf = (struct ufuncs *)mg->mg_ptr; |
352 | if (uf->uf_set == NULL) { | |
353 | SV* obj = mg->mg_obj; | |
354 | ||
355 | if (!keysv) { | |
59cd0e26 NC |
356 | keysv = newSVpvn_flags(key, klen, SVs_TEMP | |
357 | ((flags & HVhek_UTF8) | |
358 | ? SVf_UTF8 : 0)); | |
fda2d18a NC |
359 | } |
360 | ||
361 | mg->mg_obj = keysv; /* pass key */ | |
362 | uf->uf_index = action; /* pass action */ | |
ad64d0ec | 363 | magic_getuvar(MUTABLE_SV(hv), mg); |
fda2d18a NC |
364 | keysv = mg->mg_obj; /* may have changed */ |
365 | mg->mg_obj = obj; | |
366 | ||
367 | /* If the key may have changed, then we need to invalidate | |
368 | any passed-in computed hash value. */ | |
369 | hash = 0; | |
370 | } | |
371 | } | |
bdee33e4 | 372 | } |
113738bb | 373 | if (keysv) { |
e593d2fe AE |
374 | if (flags & HVhek_FREEKEY) |
375 | Safefree(key); | |
5c144d81 | 376 | key = SvPV_const(keysv, klen); |
113738bb | 377 | is_utf8 = (SvUTF8(keysv) != 0); |
44b87b50 NC |
378 | if (SvIsCOW_shared_hash(keysv)) { |
379 | flags = HVhek_KEYCANONICAL | (is_utf8 ? HVhek_UTF8 : 0); | |
380 | } else { | |
381 | flags = 0; | |
382 | } | |
113738bb | 383 | } else { |
c1fe5510 | 384 | is_utf8 = ((flags & HVhek_UTF8) ? TRUE : FALSE); |
113738bb | 385 | } |
113738bb | 386 | |
9dbc5603 | 387 | if (action & HV_DELETE) { |
3c84c864 NC |
388 | return (void *) hv_delete_common(hv, keysv, key, klen, |
389 | flags | (is_utf8 ? HVhek_UTF8 : 0), | |
390 | action, hash); | |
9dbc5603 NC |
391 | } |
392 | ||
b2c64049 | 393 | xhv = (XPVHV*)SvANY(hv); |
7f66fda2 | 394 | if (SvMAGICAL(hv)) { |
6136c704 | 395 | if (SvRMAGICAL(hv) && !(action & (HV_FETCH_ISSTORE|HV_FETCH_ISEXISTS))) { |
ad64d0ec NC |
396 | if (mg_find((const SV *)hv, PERL_MAGIC_tied) |
397 | || SvGMAGICAL((const SV *)hv)) | |
e62cc96a | 398 | { |
3c84c864 | 399 | /* FIXME should be able to skimp on the HE/HEK here when |
7f66fda2 | 400 | HV_FETCH_JUST_SV is true. */ |
7f66fda2 | 401 | if (!keysv) { |
740cce10 NC |
402 | keysv = newSVpvn_utf8(key, klen, is_utf8); |
403 | } else { | |
7f66fda2 | 404 | keysv = newSVsv(keysv); |
113738bb | 405 | } |
44a2ac75 | 406 | sv = sv_newmortal(); |
ad64d0ec | 407 | mg_copy(MUTABLE_SV(hv), sv, (char *)keysv, HEf_SVKEY); |
7f66fda2 NC |
408 | |
409 | /* grab a fake HE/HEK pair from the pool or make a new one */ | |
410 | entry = PL_hv_fetch_ent_mh; | |
411 | if (entry) | |
412 | PL_hv_fetch_ent_mh = HeNEXT(entry); | |
413 | else { | |
414 | char *k; | |
415 | entry = new_HE(); | |
ad64d0ec | 416 | Newx(k, HEK_BASESIZE + sizeof(const SV *), char); |
7f66fda2 NC |
417 | HeKEY_hek(entry) = (HEK*)k; |
418 | } | |
4608196e | 419 | HeNEXT(entry) = NULL; |
7f66fda2 NC |
420 | HeSVKEY_set(entry, keysv); |
421 | HeVAL(entry) = sv; | |
422 | sv_upgrade(sv, SVt_PVLV); | |
423 | LvTYPE(sv) = 'T'; | |
424 | /* so we can free entry when freeing sv */ | |
ad64d0ec | 425 | LvTARG(sv) = MUTABLE_SV(entry); |
7f66fda2 NC |
426 | |
427 | /* XXX remove at some point? */ | |
428 | if (flags & HVhek_FREEKEY) | |
429 | Safefree(key); | |
430 | ||
3c84c864 NC |
431 | if (return_svp) { |
432 | return entry ? (void *) &HeVAL(entry) : NULL; | |
433 | } | |
434 | return (void *) entry; | |
113738bb | 435 | } |
7f66fda2 | 436 | #ifdef ENV_IS_CASELESS |
ad64d0ec | 437 | else if (mg_find((const SV *)hv, PERL_MAGIC_env)) { |
7f66fda2 NC |
438 | U32 i; |
439 | for (i = 0; i < klen; ++i) | |
440 | if (isLOWER(key[i])) { | |
086cb327 NC |
441 | /* Would be nice if we had a routine to do the |
442 | copy and upercase in a single pass through. */ | |
0bd48802 | 443 | const char * const nkey = strupr(savepvn(key,klen)); |
086cb327 NC |
444 | /* Note that this fetch is for nkey (the uppercased |
445 | key) whereas the store is for key (the original) */ | |
63c89345 NC |
446 | void *result = hv_common(hv, NULL, nkey, klen, |
447 | HVhek_FREEKEY, /* free nkey */ | |
448 | 0 /* non-LVAL fetch */ | |
3c84c864 NC |
449 | | HV_DISABLE_UVAR_XKEY |
450 | | return_svp, | |
63c89345 NC |
451 | NULL /* no value */, |
452 | 0 /* compute hash */); | |
26488bcf | 453 | if (!result && (action & HV_FETCH_LVALUE)) { |
086cb327 NC |
454 | /* This call will free key if necessary. |
455 | Do it this way to encourage compiler to tail | |
456 | call optimise. */ | |
63c89345 NC |
457 | result = hv_common(hv, keysv, key, klen, flags, |
458 | HV_FETCH_ISSTORE | |
3c84c864 NC |
459 | | HV_DISABLE_UVAR_XKEY |
460 | | return_svp, | |
63c89345 | 461 | newSV(0), hash); |
086cb327 NC |
462 | } else { |
463 | if (flags & HVhek_FREEKEY) | |
464 | Safefree(key); | |
465 | } | |
63c89345 | 466 | return result; |
7f66fda2 | 467 | } |
902173a3 | 468 | } |
7f66fda2 NC |
469 | #endif |
470 | } /* ISFETCH */ | |
471 | else if (SvRMAGICAL(hv) && (action & HV_FETCH_ISEXISTS)) { | |
ad64d0ec NC |
472 | if (mg_find((const SV *)hv, PERL_MAGIC_tied) |
473 | || SvGMAGICAL((const SV *)hv)) { | |
b2c64049 NC |
474 | /* I don't understand why hv_exists_ent has svret and sv, |
475 | whereas hv_exists only had one. */ | |
9d4ba2ae | 476 | SV * const svret = sv_newmortal(); |
b2c64049 | 477 | sv = sv_newmortal(); |
7f66fda2 NC |
478 | |
479 | if (keysv || is_utf8) { | |
480 | if (!keysv) { | |
740cce10 | 481 | keysv = newSVpvn_utf8(key, klen, TRUE); |
7f66fda2 NC |
482 | } else { |
483 | keysv = newSVsv(keysv); | |
484 | } | |
ad64d0ec | 485 | mg_copy(MUTABLE_SV(hv), sv, (char *)sv_2mortal(keysv), HEf_SVKEY); |
b2c64049 | 486 | } else { |
ad64d0ec | 487 | mg_copy(MUTABLE_SV(hv), sv, key, klen); |
7f66fda2 | 488 | } |
b2c64049 NC |
489 | if (flags & HVhek_FREEKEY) |
490 | Safefree(key); | |
7f66fda2 NC |
491 | magic_existspack(svret, mg_find(sv, PERL_MAGIC_tiedelem)); |
492 | /* This cast somewhat evil, but I'm merely using NULL/ | |
493 | not NULL to return the boolean exists. | |
494 | And I know hv is not NULL. */ | |
3c84c864 | 495 | return SvTRUE(svret) ? (void *)hv : NULL; |
e7152ba2 | 496 | } |
7f66fda2 | 497 | #ifdef ENV_IS_CASELESS |
ad64d0ec | 498 | else if (mg_find((const SV *)hv, PERL_MAGIC_env)) { |
7f66fda2 | 499 | /* XXX This code isn't UTF8 clean. */ |
a15d23f8 | 500 | char * const keysave = (char * const)key; |
b2c64049 NC |
501 | /* Will need to free this, so set FREEKEY flag. */ |
502 | key = savepvn(key,klen); | |
503 | key = (const char*)strupr((char*)key); | |
6136c704 | 504 | is_utf8 = FALSE; |
7f66fda2 | 505 | hash = 0; |
8b4f7dd5 | 506 | keysv = 0; |
b2c64049 NC |
507 | |
508 | if (flags & HVhek_FREEKEY) { | |
509 | Safefree(keysave); | |
510 | } | |
511 | flags |= HVhek_FREEKEY; | |
7f66fda2 | 512 | } |
902173a3 | 513 | #endif |
7f66fda2 | 514 | } /* ISEXISTS */ |
b2c64049 NC |
515 | else if (action & HV_FETCH_ISSTORE) { |
516 | bool needs_copy; | |
517 | bool needs_store; | |
518 | hv_magic_check (hv, &needs_copy, &needs_store); | |
519 | if (needs_copy) { | |
a3b680e6 | 520 | const bool save_taint = PL_tainted; |
b2c64049 NC |
521 | if (keysv || is_utf8) { |
522 | if (!keysv) { | |
740cce10 | 523 | keysv = newSVpvn_utf8(key, klen, TRUE); |
b2c64049 NC |
524 | } |
525 | if (PL_tainting) | |
526 | PL_tainted = SvTAINTED(keysv); | |
527 | keysv = sv_2mortal(newSVsv(keysv)); | |
ad64d0ec | 528 | mg_copy(MUTABLE_SV(hv), val, (char*)keysv, HEf_SVKEY); |
b2c64049 | 529 | } else { |
ad64d0ec | 530 | mg_copy(MUTABLE_SV(hv), val, key, klen); |
b2c64049 NC |
531 | } |
532 | ||
533 | TAINT_IF(save_taint); | |
1baaf5d7 | 534 | if (!needs_store) { |
b2c64049 NC |
535 | if (flags & HVhek_FREEKEY) |
536 | Safefree(key); | |
4608196e | 537 | return NULL; |
b2c64049 NC |
538 | } |
539 | #ifdef ENV_IS_CASELESS | |
ad64d0ec | 540 | else if (mg_find((const SV *)hv, PERL_MAGIC_env)) { |
b2c64049 NC |
541 | /* XXX This code isn't UTF8 clean. */ |
542 | const char *keysave = key; | |
543 | /* Will need to free this, so set FREEKEY flag. */ | |
544 | key = savepvn(key,klen); | |
545 | key = (const char*)strupr((char*)key); | |
6136c704 | 546 | is_utf8 = FALSE; |
b2c64049 | 547 | hash = 0; |
8b4f7dd5 | 548 | keysv = 0; |
b2c64049 NC |
549 | |
550 | if (flags & HVhek_FREEKEY) { | |
551 | Safefree(keysave); | |
552 | } | |
553 | flags |= HVhek_FREEKEY; | |
554 | } | |
555 | #endif | |
556 | } | |
557 | } /* ISSTORE */ | |
7f66fda2 | 558 | } /* SvMAGICAL */ |
fde52b5c | 559 | |
7b2c381c | 560 | if (!HvARRAY(hv)) { |
b2c64049 | 561 | if ((action & (HV_FETCH_LVALUE | HV_FETCH_ISSTORE)) |
fde52b5c | 562 | #ifdef DYNAMIC_ENV_FETCH /* if it's an %ENV lookup, we may get it on the fly */ |
ad64d0ec NC |
563 | || (SvRMAGICAL((const SV *)hv) |
564 | && mg_find((const SV *)hv, PERL_MAGIC_env)) | |
fde52b5c | 565 | #endif |
d58e6666 NC |
566 | ) { |
567 | char *array; | |
a02a5408 | 568 | Newxz(array, |
cbec9347 | 569 | PERL_HV_ARRAY_ALLOC_BYTES(xhv->xhv_max+1 /* HvMAX(hv)+1 */), |
d58e6666 NC |
570 | char); |
571 | HvARRAY(hv) = (HE**)array; | |
572 | } | |
7f66fda2 NC |
573 | #ifdef DYNAMIC_ENV_FETCH |
574 | else if (action & HV_FETCH_ISEXISTS) { | |
575 | /* for an %ENV exists, if we do an insert it's by a recursive | |
576 | store call, so avoid creating HvARRAY(hv) right now. */ | |
577 | } | |
578 | #endif | |
113738bb NC |
579 | else { |
580 | /* XXX remove at some point? */ | |
581 | if (flags & HVhek_FREEKEY) | |
582 | Safefree(key); | |
583 | ||
3c84c864 | 584 | return NULL; |
113738bb | 585 | } |
fde52b5c | 586 | } |
587 | ||
44b87b50 | 588 | if (is_utf8 & !(flags & HVhek_KEYCANONICAL)) { |
41d88b63 | 589 | char * const keysave = (char *)key; |
f9a63242 | 590 | key = (char*)bytes_from_utf8((U8*)key, &klen, &is_utf8); |
19692e8d | 591 | if (is_utf8) |
c1fe5510 NC |
592 | flags |= HVhek_UTF8; |
593 | else | |
594 | flags &= ~HVhek_UTF8; | |
7f66fda2 NC |
595 | if (key != keysave) { |
596 | if (flags & HVhek_FREEKEY) | |
597 | Safefree(keysave); | |
19692e8d | 598 | flags |= HVhek_WASUTF8 | HVhek_FREEKEY; |
527df579 NC |
599 | /* If the caller calculated a hash, it was on the sequence of |
600 | octets that are the UTF-8 form. We've now changed the sequence | |
601 | of octets stored to that of the equivalent byte representation, | |
602 | so the hash we need is different. */ | |
603 | hash = 0; | |
7f66fda2 | 604 | } |
19692e8d | 605 | } |
f9a63242 | 606 | |
4b5190b5 NC |
607 | if (HvREHASH(hv)) { |
608 | PERL_HASH_INTERNAL(hash, key, klen); | |
b2c64049 NC |
609 | /* We don't have a pointer to the hv, so we have to replicate the |
610 | flag into every HEK, so that hv_iterkeysv can see it. */ | |
611 | /* And yes, you do need this even though you are not "storing" because | |
fdcd69b6 NC |
612 | you can flip the flags below if doing an lval lookup. (And that |
613 | was put in to give the semantics Andreas was expecting.) */ | |
614 | flags |= HVhek_REHASH; | |
4b5190b5 | 615 | } else if (!hash) { |
113738bb | 616 | if (keysv && (SvIsCOW_shared_hash(keysv))) { |
c158a4fd | 617 | hash = SvSHARED_HASH(keysv); |
46187eeb NC |
618 | } else { |
619 | PERL_HASH(hash, key, klen); | |
620 | } | |
621 | } | |
effa1e2d | 622 | |
113738bb NC |
623 | masked_flags = (flags & HVhek_MASK); |
624 | ||
7f66fda2 | 625 | #ifdef DYNAMIC_ENV_FETCH |
4608196e | 626 | if (!HvARRAY(hv)) entry = NULL; |
7f66fda2 NC |
627 | else |
628 | #endif | |
b2c64049 | 629 | { |
7b2c381c | 630 | entry = (HvARRAY(hv))[hash & (I32) HvMAX(hv)]; |
b2c64049 | 631 | } |
0298d7b9 | 632 | for (; entry; entry = HeNEXT(entry)) { |
fde52b5c | 633 | if (HeHASH(entry) != hash) /* strings can't be equal */ |
634 | continue; | |
eb160463 | 635 | if (HeKLEN(entry) != (I32)klen) |
fde52b5c | 636 | continue; |
1c846c1f | 637 | if (HeKEY(entry) != key && memNE(HeKEY(entry),key,klen)) /* is this it? */ |
fde52b5c | 638 | continue; |
113738bb | 639 | if ((HeKFLAGS(entry) ^ masked_flags) & HVhek_UTF8) |
c3654f1a | 640 | continue; |
b2c64049 NC |
641 | |
642 | if (action & (HV_FETCH_LVALUE|HV_FETCH_ISSTORE)) { | |
643 | if (HeKFLAGS(entry) != masked_flags) { | |
644 | /* We match if HVhek_UTF8 bit in our flags and hash key's | |
645 | match. But if entry was set previously with HVhek_WASUTF8 | |
646 | and key now doesn't (or vice versa) then we should change | |
647 | the key's flag, as this is assignment. */ | |
648 | if (HvSHAREKEYS(hv)) { | |
649 | /* Need to swap the key we have for a key with the flags we | |
650 | need. As keys are shared we can't just write to the | |
651 | flag, so we share the new one, unshare the old one. */ | |
6136c704 | 652 | HEK * const new_hek = share_hek_flags(key, klen, hash, |
6e838c70 | 653 | masked_flags); |
b2c64049 NC |
654 | unshare_hek (HeKEY_hek(entry)); |
655 | HeKEY_hek(entry) = new_hek; | |
656 | } | |
5d2b1485 NC |
657 | else if (hv == PL_strtab) { |
658 | /* PL_strtab is usually the only hash without HvSHAREKEYS, | |
659 | so putting this test here is cheap */ | |
660 | if (flags & HVhek_FREEKEY) | |
661 | Safefree(key); | |
662 | Perl_croak(aTHX_ S_strtab_error, | |
663 | action & HV_FETCH_LVALUE ? "fetch" : "store"); | |
664 | } | |
b2c64049 NC |
665 | else |
666 | HeKFLAGS(entry) = masked_flags; | |
667 | if (masked_flags & HVhek_ENABLEHVKFLAGS) | |
668 | HvHASKFLAGS_on(hv); | |
669 | } | |
670 | if (HeVAL(entry) == &PL_sv_placeholder) { | |
671 | /* yes, can store into placeholder slot */ | |
672 | if (action & HV_FETCH_LVALUE) { | |
673 | if (SvMAGICAL(hv)) { | |
674 | /* This preserves behaviour with the old hv_fetch | |
675 | implementation which at this point would bail out | |
676 | with a break; (at "if we find a placeholder, we | |
677 | pretend we haven't found anything") | |
678 | ||
679 | That break mean that if a placeholder were found, it | |
680 | caused a call into hv_store, which in turn would | |
681 | check magic, and if there is no magic end up pretty | |
682 | much back at this point (in hv_store's code). */ | |
683 | break; | |
684 | } | |
685 | /* LVAL fetch which actaully needs a store. */ | |
561b68a9 | 686 | val = newSV(0); |
ca732855 | 687 | HvPLACEHOLDERS(hv)--; |
b2c64049 NC |
688 | } else { |
689 | /* store */ | |
690 | if (val != &PL_sv_placeholder) | |
ca732855 | 691 | HvPLACEHOLDERS(hv)--; |
b2c64049 NC |
692 | } |
693 | HeVAL(entry) = val; | |
694 | } else if (action & HV_FETCH_ISSTORE) { | |
cefd5c7c | 695 | SvREFCNT_dec(HeVAL(entry)); |
b2c64049 NC |
696 | HeVAL(entry) = val; |
697 | } | |
27bcc0a7 | 698 | } else if (HeVAL(entry) == &PL_sv_placeholder) { |
b2c64049 NC |
699 | /* if we find a placeholder, we pretend we haven't found |
700 | anything */ | |
8aacddc1 | 701 | break; |
b2c64049 | 702 | } |
113738bb NC |
703 | if (flags & HVhek_FREEKEY) |
704 | Safefree(key); | |
3c84c864 NC |
705 | if (return_svp) { |
706 | return entry ? (void *) &HeVAL(entry) : NULL; | |
707 | } | |
fde52b5c | 708 | return entry; |
709 | } | |
710 | #ifdef DYNAMIC_ENV_FETCH /* %ENV lookup? If so, try to fetch the value now */ | |
0ed29950 | 711 | if (!(action & HV_FETCH_ISSTORE) |
ad64d0ec NC |
712 | && SvRMAGICAL((const SV *)hv) |
713 | && mg_find((const SV *)hv, PERL_MAGIC_env)) { | |
a6c40364 | 714 | unsigned long len; |
9d4ba2ae | 715 | const char * const env = PerlEnv_ENVgetenv_len(key,&len); |
a6c40364 GS |
716 | if (env) { |
717 | sv = newSVpvn(env,len); | |
718 | SvTAINTED_on(sv); | |
d3ba3f5c | 719 | return hv_common(hv, keysv, key, klen, flags, |
3c84c864 NC |
720 | HV_FETCH_ISSTORE|HV_DISABLE_UVAR_XKEY|return_svp, |
721 | sv, hash); | |
a6c40364 | 722 | } |
fde52b5c | 723 | } |
724 | #endif | |
7f66fda2 NC |
725 | |
726 | if (!entry && SvREADONLY(hv) && !(action & HV_FETCH_ISEXISTS)) { | |
c445ea15 | 727 | hv_notallowed(flags, key, klen, |
c8cd6465 NC |
728 | "Attempt to access disallowed key '%"SVf"' in" |
729 | " a restricted hash"); | |
1b1f1335 | 730 | } |
b2c64049 NC |
731 | if (!(action & (HV_FETCH_LVALUE|HV_FETCH_ISSTORE))) { |
732 | /* Not doing some form of store, so return failure. */ | |
733 | if (flags & HVhek_FREEKEY) | |
734 | Safefree(key); | |
3c84c864 | 735 | return NULL; |
b2c64049 | 736 | } |
113738bb | 737 | if (action & HV_FETCH_LVALUE) { |
df5f182b | 738 | val = action & HV_FETCH_EMPTY_HE ? NULL : newSV(0); |
b2c64049 NC |
739 | if (SvMAGICAL(hv)) { |
740 | /* At this point the old hv_fetch code would call to hv_store, | |
741 | which in turn might do some tied magic. So we need to make that | |
742 | magic check happen. */ | |
743 | /* gonna assign to this, so it better be there */ | |
fda2d18a NC |
744 | /* If a fetch-as-store fails on the fetch, then the action is to |
745 | recurse once into "hv_store". If we didn't do this, then that | |
746 | recursive call would call the key conversion routine again. | |
747 | However, as we replace the original key with the converted | |
748 | key, this would result in a double conversion, which would show | |
749 | up as a bug if the conversion routine is not idempotent. */ | |
d3ba3f5c | 750 | return hv_common(hv, keysv, key, klen, flags, |
3c84c864 NC |
751 | HV_FETCH_ISSTORE|HV_DISABLE_UVAR_XKEY|return_svp, |
752 | val, hash); | |
b2c64049 NC |
753 | /* XXX Surely that could leak if the fetch-was-store fails? |
754 | Just like the hv_fetch. */ | |
113738bb NC |
755 | } |
756 | } | |
757 | ||
b2c64049 NC |
758 | /* Welcome to hv_store... */ |
759 | ||
7b2c381c | 760 | if (!HvARRAY(hv)) { |
b2c64049 NC |
761 | /* Not sure if we can get here. I think the only case of oentry being |
762 | NULL is for %ENV with dynamic env fetch. But that should disappear | |
763 | with magic in the previous code. */ | |
d58e6666 | 764 | char *array; |
a02a5408 | 765 | Newxz(array, |
b2c64049 | 766 | PERL_HV_ARRAY_ALLOC_BYTES(xhv->xhv_max+1 /* HvMAX(hv)+1 */), |
d58e6666 NC |
767 | char); |
768 | HvARRAY(hv) = (HE**)array; | |
b2c64049 NC |
769 | } |
770 | ||
7b2c381c | 771 | oentry = &(HvARRAY(hv))[hash & (I32) xhv->xhv_max]; |
ab4af705 | 772 | |
b2c64049 NC |
773 | entry = new_HE(); |
774 | /* share_hek_flags will do the free for us. This might be considered | |
775 | bad API design. */ | |
776 | if (HvSHAREKEYS(hv)) | |
6e838c70 | 777 | HeKEY_hek(entry) = share_hek_flags(key, klen, hash, flags); |
5d2b1485 NC |
778 | else if (hv == PL_strtab) { |
779 | /* PL_strtab is usually the only hash without HvSHAREKEYS, so putting | |
780 | this test here is cheap */ | |
781 | if (flags & HVhek_FREEKEY) | |
782 | Safefree(key); | |
783 | Perl_croak(aTHX_ S_strtab_error, | |
784 | action & HV_FETCH_LVALUE ? "fetch" : "store"); | |
785 | } | |
b2c64049 NC |
786 | else /* gotta do the real thing */ |
787 | HeKEY_hek(entry) = save_hek_flags(key, klen, hash, flags); | |
788 | HeVAL(entry) = val; | |
789 | HeNEXT(entry) = *oentry; | |
790 | *oentry = entry; | |
791 | ||
792 | if (val == &PL_sv_placeholder) | |
ca732855 | 793 | HvPLACEHOLDERS(hv)++; |
b2c64049 NC |
794 | if (masked_flags & HVhek_ENABLEHVKFLAGS) |
795 | HvHASKFLAGS_on(hv); | |
796 | ||
0298d7b9 NC |
797 | { |
798 | const HE *counter = HeNEXT(entry); | |
799 | ||
4c7185a0 | 800 | xhv->xhv_keys++; /* HvTOTALKEYS(hv)++ */ |
0298d7b9 | 801 | if (!counter) { /* initial entry? */ |
5ac36297 | 802 | } else if (xhv->xhv_keys > xhv->xhv_max) { |
0298d7b9 NC |
803 | hsplit(hv); |
804 | } else if(!HvREHASH(hv)) { | |
805 | U32 n_links = 1; | |
806 | ||
807 | while ((counter = HeNEXT(counter))) | |
808 | n_links++; | |
809 | ||
810 | if (n_links > HV_MAX_LENGTH_BEFORE_SPLIT) { | |
811 | /* Use only the old HvKEYS(hv) > HvMAX(hv) condition to limit | |
812 | bucket splits on a rehashed hash, as we're not going to | |
813 | split it again, and if someone is lucky (evil) enough to | |
814 | get all the keys in one list they could exhaust our memory | |
815 | as we repeatedly double the number of buckets on every | |
816 | entry. Linear search feels a less worse thing to do. */ | |
817 | hsplit(hv); | |
818 | } | |
819 | } | |
fde52b5c | 820 | } |
b2c64049 | 821 | |
3c84c864 NC |
822 | if (return_svp) { |
823 | return entry ? (void *) &HeVAL(entry) : NULL; | |
824 | } | |
825 | return (void *) entry; | |
fde52b5c | 826 | } |
827 | ||
864dbfa3 | 828 | STATIC void |
b0e6ae5b | 829 | S_hv_magic_check(HV *hv, bool *needs_copy, bool *needs_store) |
d0066dc7 | 830 | { |
a3b680e6 | 831 | const MAGIC *mg = SvMAGIC(hv); |
7918f24d NC |
832 | |
833 | PERL_ARGS_ASSERT_HV_MAGIC_CHECK; | |
834 | ||
d0066dc7 OT |
835 | *needs_copy = FALSE; |
836 | *needs_store = TRUE; | |
837 | while (mg) { | |
838 | if (isUPPER(mg->mg_type)) { | |
839 | *needs_copy = TRUE; | |
d60c5a05 | 840 | if (mg->mg_type == PERL_MAGIC_tied) { |
d0066dc7 | 841 | *needs_store = FALSE; |
4ab2a30b | 842 | return; /* We've set all there is to set. */ |
d0066dc7 OT |
843 | } |
844 | } | |
845 | mg = mg->mg_moremagic; | |
846 | } | |
847 | } | |
848 | ||
954c1994 | 849 | /* |
a3bcc51e TP |
850 | =for apidoc hv_scalar |
851 | ||
852 | Evaluates the hash in scalar context and returns the result. Handles magic when the hash is tied. | |
853 | ||
854 | =cut | |
855 | */ | |
856 | ||
857 | SV * | |
858 | Perl_hv_scalar(pTHX_ HV *hv) | |
859 | { | |
a3bcc51e | 860 | SV *sv; |
823a54a3 | 861 | |
7918f24d NC |
862 | PERL_ARGS_ASSERT_HV_SCALAR; |
863 | ||
823a54a3 | 864 | if (SvRMAGICAL(hv)) { |
ad64d0ec | 865 | MAGIC * const mg = mg_find((const SV *)hv, PERL_MAGIC_tied); |
823a54a3 AL |
866 | if (mg) |
867 | return magic_scalarpack(hv, mg); | |
868 | } | |
a3bcc51e TP |
869 | |
870 | sv = sv_newmortal(); | |
f4431c56 | 871 | if (HvTOTALKEYS((const HV *)hv)) |
a3bcc51e TP |
872 | Perl_sv_setpvf(aTHX_ sv, "%ld/%ld", |
873 | (long)HvFILL(hv), (long)HvMAX(hv) + 1); | |
874 | else | |
875 | sv_setiv(sv, 0); | |
876 | ||
877 | return sv; | |
878 | } | |
879 | ||
880 | /* | |
954c1994 GS |
881 | =for apidoc hv_delete |
882 | ||
3025a2e4 CS |
883 | Deletes a key/value pair in the hash. The value's SV is removed from the |
884 | hash, made mortal, and returned to the caller. The C<klen> is the length of | |
885 | the key. The C<flags> value will normally be zero; if set to G_DISCARD then | |
886 | NULL will be returned. NULL will also be returned if the key is not found. | |
954c1994 | 887 | |
954c1994 GS |
888 | =for apidoc hv_delete_ent |
889 | ||
3025a2e4 CS |
890 | Deletes a key/value pair in the hash. The value SV is removed from the hash, |
891 | made mortal, and returned to the caller. The C<flags> value will normally be | |
892 | zero; if set to G_DISCARD then NULL will be returned. NULL will also be | |
893 | returned if the key is not found. C<hash> can be a valid precomputed hash | |
894 | value, or 0 to ask for it to be computed. | |
954c1994 GS |
895 | |
896 | =cut | |
897 | */ | |
898 | ||
8f8d40ab | 899 | STATIC SV * |
cd6d36ac NC |
900 | S_hv_delete_common(pTHX_ HV *hv, SV *keysv, const char *key, STRLEN klen, |
901 | int k_flags, I32 d_flags, U32 hash) | |
f1317c8d | 902 | { |
27da23d5 | 903 | dVAR; |
cbec9347 | 904 | register XPVHV* xhv; |
fde52b5c | 905 | register HE *entry; |
906 | register HE **oentry; | |
9e720f71 | 907 | HE *const *first_entry; |
9dbc5603 | 908 | bool is_utf8 = (k_flags & HVhek_UTF8) ? TRUE : FALSE; |
7a9669ca | 909 | int masked_flags; |
1c846c1f | 910 | |
fde52b5c | 911 | if (SvRMAGICAL(hv)) { |
0a0bb7c7 OT |
912 | bool needs_copy; |
913 | bool needs_store; | |
914 | hv_magic_check (hv, &needs_copy, &needs_store); | |
915 | ||
f1317c8d | 916 | if (needs_copy) { |
6136c704 | 917 | SV *sv; |
63c89345 NC |
918 | entry = (HE *) hv_common(hv, keysv, key, klen, |
919 | k_flags & ~HVhek_FREEKEY, | |
920 | HV_FETCH_LVALUE|HV_DISABLE_UVAR_XKEY, | |
921 | NULL, hash); | |
7a9669ca | 922 | sv = entry ? HeVAL(entry) : NULL; |
f1317c8d NC |
923 | if (sv) { |
924 | if (SvMAGICAL(sv)) { | |
925 | mg_clear(sv); | |
926 | } | |
927 | if (!needs_store) { | |
928 | if (mg_find(sv, PERL_MAGIC_tiedelem)) { | |
929 | /* No longer an element */ | |
930 | sv_unmagic(sv, PERL_MAGIC_tiedelem); | |
931 | return sv; | |
932 | } | |
a0714e2c | 933 | return NULL; /* element cannot be deleted */ |
f1317c8d | 934 | } |
902173a3 | 935 | #ifdef ENV_IS_CASELESS |
ad64d0ec | 936 | else if (mg_find((const SV *)hv, PERL_MAGIC_env)) { |
8167a60a | 937 | /* XXX This code isn't UTF8 clean. */ |
59cd0e26 | 938 | keysv = newSVpvn_flags(key, klen, SVs_TEMP); |
8167a60a NC |
939 | if (k_flags & HVhek_FREEKEY) { |
940 | Safefree(key); | |
941 | } | |
942 | key = strupr(SvPVX(keysv)); | |
943 | is_utf8 = 0; | |
944 | k_flags = 0; | |
945 | hash = 0; | |
7f66fda2 | 946 | } |
510ac311 | 947 | #endif |
2fd1c6b8 | 948 | } |
2fd1c6b8 | 949 | } |
fde52b5c | 950 | } |
cbec9347 | 951 | xhv = (XPVHV*)SvANY(hv); |
7b2c381c | 952 | if (!HvARRAY(hv)) |
a0714e2c | 953 | return NULL; |
fde52b5c | 954 | |
19692e8d | 955 | if (is_utf8) { |
c445ea15 | 956 | const char * const keysave = key; |
b464bac0 | 957 | key = (char*)bytes_from_utf8((U8*)key, &klen, &is_utf8); |
cd6d36ac | 958 | |
19692e8d | 959 | if (is_utf8) |
cd6d36ac NC |
960 | k_flags |= HVhek_UTF8; |
961 | else | |
962 | k_flags &= ~HVhek_UTF8; | |
7f66fda2 NC |
963 | if (key != keysave) { |
964 | if (k_flags & HVhek_FREEKEY) { | |
965 | /* This shouldn't happen if our caller does what we expect, | |
966 | but strictly the API allows it. */ | |
967 | Safefree(keysave); | |
968 | } | |
969 | k_flags |= HVhek_WASUTF8 | HVhek_FREEKEY; | |
970 | } | |
ad64d0ec | 971 | HvHASKFLAGS_on(MUTABLE_SV(hv)); |
19692e8d | 972 | } |
f9a63242 | 973 | |
4b5190b5 NC |
974 | if (HvREHASH(hv)) { |
975 | PERL_HASH_INTERNAL(hash, key, klen); | |
976 | } else if (!hash) { | |
7a9669ca | 977 | if (keysv && (SvIsCOW_shared_hash(keysv))) { |
c158a4fd | 978 | hash = SvSHARED_HASH(keysv); |
7a9669ca NC |
979 | } else { |
980 | PERL_HASH(hash, key, klen); | |
981 | } | |
4b5190b5 | 982 | } |
fde52b5c | 983 | |
7a9669ca NC |
984 | masked_flags = (k_flags & HVhek_MASK); |
985 | ||
9e720f71 | 986 | first_entry = 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 | } |
6136c704 | 1015 | if (SvREADONLY(hv) && HeVAL(entry) && SvREADONLY(HeVAL(entry))) { |
d4c19fe8 | 1016 | hv_notallowed(k_flags, key, klen, |
c8cd6465 NC |
1017 | "Attempt to delete readonly key '%"SVf"' from" |
1018 | " a restricted hash"); | |
8aacddc1 | 1019 | } |
b84d0860 NC |
1020 | if (k_flags & HVhek_FREEKEY) |
1021 | Safefree(key); | |
8aacddc1 | 1022 | |
35759254 | 1023 | /* If this is a stash and the key ends with ::, then someone is |
0c3bb3c2 | 1024 | * deleting a package. |
0c3bb3c2 | 1025 | */ |
78b79c77 | 1026 | if (HeVAL(entry) && HvENAME_get(hv)) { |
0290c710 | 1027 | gv = (GV *)HeVAL(entry); |
35759254 FC |
1028 | if (keysv) key = SvPV(keysv, klen); |
1029 | if (klen > 1 && key[klen-2] == ':' && key[klen-1] == ':' | |
e0a52395 | 1030 | && (klen != 6 || hv!=PL_defstash || memNE(key,"main::",6)) |
0290c710 | 1031 | && SvTYPE(gv) == SVt_PVGV && (stash = GvHV((GV *)gv)) |
0c3bb3c2 | 1032 | && HvENAME_get(stash)) { |
0290c710 FC |
1033 | /* A previous version of this code checked that the |
1034 | * GV was still in the symbol table by fetching the | |
1035 | * GV with its name. That is not necessary (and | |
1036 | * sometimes incorrect), as HvENAME cannot be set | |
1037 | * on hv if it is not in the symtab. */ | |
f3d2f32d | 1038 | mro_changes = 2; |
0c3bb3c2 FC |
1039 | /* Hang on to it for a bit. */ |
1040 | SvREFCNT_inc_simple_void_NN( | |
0290c710 | 1041 | sv_2mortal((SV *)gv) |
35759254 FC |
1042 | ); |
1043 | } | |
f3d2f32d FC |
1044 | else if (klen == 3 && strnEQ(key, "ISA", 3)) |
1045 | mro_changes = 1; | |
35759254 FC |
1046 | } |
1047 | ||
cd6d36ac | 1048 | if (d_flags & G_DISCARD) |
a0714e2c | 1049 | sv = NULL; |
94f7643d | 1050 | else { |
79d01fbf | 1051 | sv = sv_2mortal(HeVAL(entry)); |
7996736c | 1052 | HeVAL(entry) = &PL_sv_placeholder; |
94f7643d | 1053 | } |
8aacddc1 NIS |
1054 | |
1055 | /* | |
1056 | * If a restricted hash, rather than really deleting the entry, put | |
1057 | * a placeholder there. This marks the key as being "approved", so | |
1058 | * we can still access via not-really-existing key without raising | |
1059 | * an error. | |
1060 | */ | |
1061 | if (SvREADONLY(hv)) { | |
754604c4 | 1062 | SvREFCNT_dec(HeVAL(entry)); |
7996736c | 1063 | HeVAL(entry) = &PL_sv_placeholder; |
8aacddc1 NIS |
1064 | /* We'll be saving this slot, so the number of allocated keys |
1065 | * doesn't go down, but the number placeholders goes up */ | |
ca732855 | 1066 | HvPLACEHOLDERS(hv)++; |
8aacddc1 | 1067 | } else { |
a26e96df | 1068 | *oentry = HeNEXT(entry); |
b79f7545 | 1069 | if (SvOOK(hv) && entry == HvAUX(hv)->xhv_eiter /* HvEITER(hv) */) |
8aacddc1 NIS |
1070 | HvLAZYDEL_on(hv); |
1071 | else | |
1072 | hv_free_ent(hv, entry); | |
4c7185a0 | 1073 | xhv->xhv_keys--; /* HvTOTALKEYS(hv)-- */ |
574c8022 | 1074 | if (xhv->xhv_keys == 0) |
19692e8d | 1075 | HvHASKFLAGS_off(hv); |
8aacddc1 | 1076 | } |
0c3bb3c2 | 1077 | |
f3d2f32d FC |
1078 | if (mro_changes == 1) mro_isa_changed_in(hv); |
1079 | else if (mro_changes == 2) | |
0290c710 | 1080 | mro_package_moved(NULL, stash, gv, NULL, 1); |
0c3bb3c2 | 1081 | |
79072805 LW |
1082 | return sv; |
1083 | } | |
8aacddc1 | 1084 | if (SvREADONLY(hv)) { |
d4c19fe8 | 1085 | hv_notallowed(k_flags, key, klen, |
c8cd6465 NC |
1086 | "Attempt to delete disallowed key '%"SVf"' from" |
1087 | " a restricted hash"); | |
8aacddc1 NIS |
1088 | } |
1089 | ||
19692e8d | 1090 | if (k_flags & HVhek_FREEKEY) |
f9a63242 | 1091 | Safefree(key); |
a0714e2c | 1092 | return NULL; |
79072805 LW |
1093 | } |
1094 | ||
76e3520e | 1095 | STATIC void |
cea2e8a9 | 1096 | S_hsplit(pTHX_ HV *hv) |
79072805 | 1097 | { |
97aff369 | 1098 | dVAR; |
1e05feb3 | 1099 | register XPVHV* const xhv = (XPVHV*)SvANY(hv); |
a3b680e6 | 1100 | const I32 oldsize = (I32) xhv->xhv_max+1; /* HvMAX(hv)+1 (sick) */ |
79072805 LW |
1101 | register I32 newsize = oldsize * 2; |
1102 | register I32 i; | |
7b2c381c | 1103 | char *a = (char*) HvARRAY(hv); |
72311751 | 1104 | register HE **aep; |
4b5190b5 NC |
1105 | int longest_chain = 0; |
1106 | int was_shared; | |
79072805 | 1107 | |
7918f24d NC |
1108 | PERL_ARGS_ASSERT_HSPLIT; |
1109 | ||
18026298 | 1110 | /*PerlIO_printf(PerlIO_stderr(), "hsplit called for %p which had %d\n", |
6c9570dc | 1111 | (void*)hv, (int) oldsize);*/ |
18026298 | 1112 | |
5d88ecd7 | 1113 | if (HvPLACEHOLDERS_get(hv) && !SvREADONLY(hv)) { |
18026298 NC |
1114 | /* Can make this clear any placeholders first for non-restricted hashes, |
1115 | even though Storable rebuilds restricted hashes by putting in all the | |
1116 | placeholders (first) before turning on the readonly flag, because | |
1117 | Storable always pre-splits the hash. */ | |
1118 | hv_clear_placeholders(hv); | |
1119 | } | |
1120 | ||
3280af22 | 1121 | PL_nomemok = TRUE; |
8d6dde3e | 1122 | #if defined(STRANGE_MALLOC) || defined(MYMALLOC) |
b79f7545 NC |
1123 | Renew(a, PERL_HV_ARRAY_ALLOC_BYTES(newsize) |
1124 | + (SvOOK(hv) ? sizeof(struct xpvhv_aux) : 0), char); | |
422a93e5 | 1125 | if (!a) { |
4a33f861 | 1126 | PL_nomemok = FALSE; |
422a93e5 GA |
1127 | return; |
1128 | } | |
b79f7545 | 1129 | if (SvOOK(hv)) { |
31f0e52e | 1130 | Move(&a[oldsize * sizeof(HE*)], &a[newsize * sizeof(HE*)], 1, struct xpvhv_aux); |
b79f7545 | 1131 | } |
4633a7c4 | 1132 | #else |
a02a5408 | 1133 | Newx(a, PERL_HV_ARRAY_ALLOC_BYTES(newsize) |
b79f7545 | 1134 | + (SvOOK(hv) ? sizeof(struct xpvhv_aux) : 0), char); |
422a93e5 | 1135 | if (!a) { |
3280af22 | 1136 | PL_nomemok = FALSE; |
422a93e5 GA |
1137 | return; |
1138 | } | |
7b2c381c | 1139 | Copy(HvARRAY(hv), a, oldsize * sizeof(HE*), char); |
b79f7545 NC |
1140 | if (SvOOK(hv)) { |
1141 | Copy(HvAUX(hv), &a[newsize * sizeof(HE*)], 1, struct xpvhv_aux); | |
1142 | } | |
9a87bd09 | 1143 | Safefree(HvARRAY(hv)); |
4633a7c4 LW |
1144 | #endif |
1145 | ||
3280af22 | 1146 | PL_nomemok = FALSE; |
72311751 | 1147 | Zero(&a[oldsize * sizeof(HE*)], (newsize-oldsize) * sizeof(HE*), char); /* zero 2nd half*/ |
cbec9347 | 1148 | xhv->xhv_max = --newsize; /* HvMAX(hv) = --newsize */ |
7b2c381c | 1149 | HvARRAY(hv) = (HE**) a; |
72311751 | 1150 | aep = (HE**)a; |
79072805 | 1151 | |
72311751 | 1152 | for (i=0; i<oldsize; i++,aep++) { |
4b5190b5 NC |
1153 | int left_length = 0; |
1154 | int right_length = 0; | |
a50a3493 NC |
1155 | HE **oentry = aep; |
1156 | HE *entry = *aep; | |
a3b680e6 | 1157 | register HE **bep; |
4b5190b5 | 1158 | |
a50a3493 | 1159 | if (!entry) /* non-existent */ |
79072805 | 1160 | continue; |
72311751 | 1161 | bep = aep+oldsize; |
4c9d89c5 | 1162 | do { |
eb160463 | 1163 | if ((HeHASH(entry) & newsize) != (U32)i) { |
fde52b5c | 1164 | *oentry = HeNEXT(entry); |
72311751 | 1165 | HeNEXT(entry) = *bep; |
72311751 | 1166 | *bep = entry; |
4b5190b5 | 1167 | right_length++; |
79072805 | 1168 | } |
4b5190b5 | 1169 | else { |
fde52b5c | 1170 | oentry = &HeNEXT(entry); |
4b5190b5 NC |
1171 | left_length++; |
1172 | } | |
4c9d89c5 NC |
1173 | entry = *oentry; |
1174 | } while (entry); | |
4b5190b5 NC |
1175 | /* I think we don't actually need to keep track of the longest length, |
1176 | merely flag if anything is too long. But for the moment while | |
1177 | developing this code I'll track it. */ | |
1178 | if (left_length > longest_chain) | |
1179 | longest_chain = left_length; | |
1180 | if (right_length > longest_chain) | |
1181 | longest_chain = right_length; | |
1182 | } | |
1183 | ||
1184 | ||
1185 | /* Pick your policy for "hashing isn't working" here: */ | |
fdcd69b6 | 1186 | if (longest_chain <= HV_MAX_LENGTH_BEFORE_SPLIT /* split worked? */ |
4b5190b5 NC |
1187 | || HvREHASH(hv)) { |
1188 | return; | |
79072805 | 1189 | } |
4b5190b5 NC |
1190 | |
1191 | if (hv == PL_strtab) { | |
1192 | /* Urg. Someone is doing something nasty to the string table. | |
1193 | Can't win. */ | |
1194 | return; | |
1195 | } | |
1196 | ||
1197 | /* Awooga. Awooga. Pathological data. */ | |
6c9570dc | 1198 | /*PerlIO_printf(PerlIO_stderr(), "%p %d of %d with %d/%d buckets\n", (void*)hv, |
4b5190b5 NC |
1199 | longest_chain, HvTOTALKEYS(hv), HvFILL(hv), 1+HvMAX(hv));*/ |
1200 | ||
1201 | ++newsize; | |
a02a5408 | 1202 | Newxz(a, PERL_HV_ARRAY_ALLOC_BYTES(newsize) |
b79f7545 NC |
1203 | + (SvOOK(hv) ? sizeof(struct xpvhv_aux) : 0), char); |
1204 | if (SvOOK(hv)) { | |
1205 | Copy(HvAUX(hv), &a[newsize * sizeof(HE*)], 1, struct xpvhv_aux); | |
1206 | } | |
1207 | ||
4b5190b5 NC |
1208 | was_shared = HvSHAREKEYS(hv); |
1209 | ||
4b5190b5 NC |
1210 | HvSHAREKEYS_off(hv); |
1211 | HvREHASH_on(hv); | |
1212 | ||
7b2c381c | 1213 | aep = HvARRAY(hv); |
4b5190b5 NC |
1214 | |
1215 | for (i=0; i<newsize; i++,aep++) { | |
a3b680e6 | 1216 | register HE *entry = *aep; |
4b5190b5 NC |
1217 | while (entry) { |
1218 | /* We're going to trash this HE's next pointer when we chain it | |
1219 | into the new hash below, so store where we go next. */ | |
9d4ba2ae | 1220 | HE * const next = HeNEXT(entry); |
4b5190b5 | 1221 | UV hash; |
a3b680e6 | 1222 | HE **bep; |
4b5190b5 NC |
1223 | |
1224 | /* Rehash it */ | |
1225 | PERL_HASH_INTERNAL(hash, HeKEY(entry), HeKLEN(entry)); | |
1226 | ||
1227 | if (was_shared) { | |
1228 | /* Unshare it. */ | |
aec46f14 | 1229 | HEK * const new_hek |
4b5190b5 NC |
1230 | = save_hek_flags(HeKEY(entry), HeKLEN(entry), |
1231 | hash, HeKFLAGS(entry)); | |
1232 | unshare_hek (HeKEY_hek(entry)); | |
1233 | HeKEY_hek(entry) = new_hek; | |
1234 | } else { | |
1235 | /* Not shared, so simply write the new hash in. */ | |
1236 | HeHASH(entry) = hash; | |
1237 | } | |
1238 | /*PerlIO_printf(PerlIO_stderr(), "%d ", HeKFLAGS(entry));*/ | |
1239 | HEK_REHASH_on(HeKEY_hek(entry)); | |
1240 | /*PerlIO_printf(PerlIO_stderr(), "%d\n", HeKFLAGS(entry));*/ | |
1241 | ||
1242 | /* Copy oentry to the correct new chain. */ | |
1243 | bep = ((HE**)a) + (hash & (I32) xhv->xhv_max); | |
4b5190b5 NC |
1244 | HeNEXT(entry) = *bep; |
1245 | *bep = entry; | |
1246 | ||
1247 | entry = next; | |
1248 | } | |
1249 | } | |
7b2c381c NC |
1250 | Safefree (HvARRAY(hv)); |
1251 | HvARRAY(hv) = (HE **)a; | |
79072805 LW |
1252 | } |
1253 | ||
72940dca | 1254 | void |
864dbfa3 | 1255 | Perl_hv_ksplit(pTHX_ HV *hv, IV newmax) |
72940dca | 1256 | { |
97aff369 | 1257 | dVAR; |
cbec9347 | 1258 | register XPVHV* xhv = (XPVHV*)SvANY(hv); |
a3b680e6 | 1259 | const I32 oldsize = (I32) xhv->xhv_max+1; /* HvMAX(hv)+1 (sick) */ |
72940dca | 1260 | register I32 newsize; |
1261 | register I32 i; | |
72311751 GS |
1262 | register char *a; |
1263 | register HE **aep; | |
72940dca | 1264 | |
7918f24d NC |
1265 | PERL_ARGS_ASSERT_HV_KSPLIT; |
1266 | ||
72940dca | 1267 | newsize = (I32) newmax; /* possible truncation here */ |
1268 | if (newsize != newmax || newmax <= oldsize) | |
1269 | return; | |
1270 | while ((newsize & (1 + ~newsize)) != newsize) { | |
1271 | newsize &= ~(newsize & (1 + ~newsize)); /* get proper power of 2 */ | |
1272 | } | |
1273 | if (newsize < newmax) | |
1274 | newsize *= 2; | |
1275 | if (newsize < newmax) | |
1276 | return; /* overflow detection */ | |
1277 | ||
7b2c381c | 1278 | a = (char *) HvARRAY(hv); |
72940dca | 1279 | if (a) { |
3280af22 | 1280 | PL_nomemok = TRUE; |
8d6dde3e | 1281 | #if defined(STRANGE_MALLOC) || defined(MYMALLOC) |
b79f7545 NC |
1282 | Renew(a, PERL_HV_ARRAY_ALLOC_BYTES(newsize) |
1283 | + (SvOOK(hv) ? sizeof(struct xpvhv_aux) : 0), char); | |
8aacddc1 | 1284 | if (!a) { |
4a33f861 | 1285 | PL_nomemok = FALSE; |
422a93e5 GA |
1286 | return; |
1287 | } | |
b79f7545 | 1288 | if (SvOOK(hv)) { |
7a9b70e9 | 1289 | Copy(&a[oldsize * sizeof(HE*)], &a[newsize * sizeof(HE*)], 1, struct xpvhv_aux); |
b79f7545 | 1290 | } |
72940dca | 1291 | #else |
a02a5408 | 1292 | Newx(a, PERL_HV_ARRAY_ALLOC_BYTES(newsize) |
b79f7545 | 1293 | + (SvOOK(hv) ? sizeof(struct xpvhv_aux) : 0), char); |
8aacddc1 | 1294 | if (!a) { |
3280af22 | 1295 | PL_nomemok = FALSE; |
422a93e5 GA |
1296 | return; |
1297 | } | |
7b2c381c | 1298 | Copy(HvARRAY(hv), a, oldsize * sizeof(HE*), char); |
b79f7545 NC |
1299 | if (SvOOK(hv)) { |
1300 | Copy(HvAUX(hv), &a[newsize * sizeof(HE*)], 1, struct xpvhv_aux); | |
1301 | } | |
9a87bd09 | 1302 | Safefree(HvARRAY(hv)); |
72940dca | 1303 | #endif |
3280af22 | 1304 | PL_nomemok = FALSE; |
72311751 | 1305 | Zero(&a[oldsize * sizeof(HE*)], (newsize-oldsize) * sizeof(HE*), char); /* zero 2nd half*/ |
72940dca | 1306 | } |
1307 | else { | |
a02a5408 | 1308 | Newxz(a, PERL_HV_ARRAY_ALLOC_BYTES(newsize), char); |
72940dca | 1309 | } |
cbec9347 | 1310 | xhv->xhv_max = --newsize; /* HvMAX(hv) = --newsize */ |
7b2c381c | 1311 | HvARRAY(hv) = (HE **) a; |
f4431c56 | 1312 | if (!xhv->xhv_keys /* !HvTOTALKEYS(hv) */) /* skip rest if no entries */ |
72940dca | 1313 | return; |
1314 | ||
72311751 GS |
1315 | aep = (HE**)a; |
1316 | for (i=0; i<oldsize; i++,aep++) { | |
a50a3493 NC |
1317 | HE **oentry = aep; |
1318 | HE *entry = *aep; | |
1319 | ||
1320 | if (!entry) /* non-existent */ | |
72940dca | 1321 | continue; |
4c9d89c5 | 1322 | do { |
6136c704 AL |
1323 | register I32 j = (HeHASH(entry) & newsize); |
1324 | ||
1325 | if (j != i) { | |
72940dca | 1326 | j -= i; |
1327 | *oentry = HeNEXT(entry); | |
4d0fbddd | 1328 | HeNEXT(entry) = aep[j]; |
72311751 | 1329 | aep[j] = entry; |
72940dca | 1330 | } |
1331 | else | |
1332 | oentry = &HeNEXT(entry); | |
4c9d89c5 NC |
1333 | entry = *oentry; |
1334 | } while (entry); | |
72940dca | 1335 | } |
1336 | } | |
1337 | ||
b3ac6de7 | 1338 | HV * |
864dbfa3 | 1339 | Perl_newHVhv(pTHX_ HV *ohv) |
b3ac6de7 | 1340 | { |
749123ff | 1341 | dVAR; |
9d4ba2ae | 1342 | HV * const hv = newHV(); |
f4431c56 | 1343 | STRLEN hv_max; |
4beac62f | 1344 | |
f4431c56 | 1345 | if (!ohv || !HvTOTALKEYS(ohv)) |
4beac62f | 1346 | return hv; |
4beac62f | 1347 | hv_max = HvMAX(ohv); |
b3ac6de7 | 1348 | |
ad64d0ec | 1349 | if (!SvMAGICAL((const SV *)ohv)) { |
b56ba0bf | 1350 | /* It's an ordinary hash, so copy it fast. AMS 20010804 */ |
eb160463 | 1351 | STRLEN i; |
a3b680e6 | 1352 | const bool shared = !!HvSHAREKEYS(ohv); |
aec46f14 | 1353 | HE **ents, ** const oents = (HE **)HvARRAY(ohv); |
ff875642 | 1354 | char *a; |
a02a5408 | 1355 | Newx(a, PERL_HV_ARRAY_ALLOC_BYTES(hv_max+1), char); |
ff875642 | 1356 | ents = (HE**)a; |
b56ba0bf AMS |
1357 | |
1358 | /* In each bucket... */ | |
1359 | for (i = 0; i <= hv_max; i++) { | |
6136c704 | 1360 | HE *prev = NULL; |
aec46f14 | 1361 | HE *oent = oents[i]; |
b56ba0bf AMS |
1362 | |
1363 | if (!oent) { | |
1364 | ents[i] = NULL; | |
1365 | continue; | |
1366 | } | |
1367 | ||
1368 | /* Copy the linked list of entries. */ | |
aec46f14 | 1369 | for (; oent; oent = HeNEXT(oent)) { |
a3b680e6 AL |
1370 | const U32 hash = HeHASH(oent); |
1371 | const char * const key = HeKEY(oent); | |
1372 | const STRLEN len = HeKLEN(oent); | |
1373 | const int flags = HeKFLAGS(oent); | |
6136c704 | 1374 | HE * const ent = new_HE(); |
c3acb9e0 | 1375 | SV *const val = HeVAL(oent); |
b56ba0bf | 1376 | |
c3acb9e0 | 1377 | HeVAL(ent) = SvIMMORTAL(val) ? val : newSVsv(val); |
19692e8d | 1378 | HeKEY_hek(ent) |
6e838c70 | 1379 | = shared ? share_hek_flags(key, len, hash, flags) |
19692e8d | 1380 | : save_hek_flags(key, len, hash, flags); |
b56ba0bf AMS |
1381 | if (prev) |
1382 | HeNEXT(prev) = ent; | |
1383 | else | |
1384 | ents[i] = ent; | |
1385 | prev = ent; | |
1386 | HeNEXT(ent) = NULL; | |
1387 | } | |
1388 | } | |
1389 | ||
1390 | HvMAX(hv) = hv_max; | |
8aacddc1 | 1391 | HvTOTALKEYS(hv) = HvTOTALKEYS(ohv); |
b56ba0bf | 1392 | HvARRAY(hv) = ents; |
aec46f14 | 1393 | } /* not magical */ |
b56ba0bf AMS |
1394 | else { |
1395 | /* Iterate over ohv, copying keys and values one at a time. */ | |
b3ac6de7 | 1396 | HE *entry; |
bfcb3514 NC |
1397 | const I32 riter = HvRITER_get(ohv); |
1398 | HE * const eiter = HvEITER_get(ohv); | |
f4431c56 | 1399 | STRLEN hv_fill = HvFILL(ohv); |
b56ba0bf AMS |
1400 | |
1401 | /* Can we use fewer buckets? (hv_max is always 2^n-1) */ | |
1402 | while (hv_max && hv_max + 1 >= hv_fill * 2) | |
1403 | hv_max = hv_max / 2; | |
1404 | HvMAX(hv) = hv_max; | |
1405 | ||
4a76a316 | 1406 | hv_iterinit(ohv); |
e16e2ff8 | 1407 | while ((entry = hv_iternext_flags(ohv, 0))) { |
c3acb9e0 | 1408 | SV *const val = HeVAL(entry); |
04fe65b0 | 1409 | (void)hv_store_flags(hv, HeKEY(entry), HeKLEN(entry), |
c3acb9e0 NC |
1410 | SvIMMORTAL(val) ? val : newSVsv(val), |
1411 | HeHASH(entry), HeKFLAGS(entry)); | |
b3ac6de7 | 1412 | } |
bfcb3514 NC |
1413 | HvRITER_set(ohv, riter); |
1414 | HvEITER_set(ohv, eiter); | |
b3ac6de7 | 1415 | } |
1c846c1f | 1416 | |
b3ac6de7 IZ |
1417 | return hv; |
1418 | } | |
1419 | ||
defdfed5 Z |
1420 | /* |
1421 | =for apidoc Am|HV *|hv_copy_hints_hv|HV *ohv | |
1422 | ||
1423 | A specialised version of L</newHVhv> for copying C<%^H>. I<ohv> must be | |
1424 | a pointer to a hash (which may have C<%^H> magic, but should be generally | |
1425 | non-magical), or C<NULL> (interpreted as an empty hash). The content | |
1426 | of I<ohv> is copied to a new hash, which has the C<%^H>-specific magic | |
1427 | added to it. A pointer to the new hash is returned. | |
1428 | ||
1429 | =cut | |
1430 | */ | |
1431 | ||
5b9c0671 NC |
1432 | HV * |
1433 | Perl_hv_copy_hints_hv(pTHX_ HV *const ohv) | |
1434 | { | |
1435 | HV * const hv = newHV(); | |
5b9c0671 | 1436 | |
f4431c56 | 1437 | if (ohv && HvTOTALKEYS(ohv)) { |
5b9c0671 | 1438 | STRLEN hv_max = HvMAX(ohv); |
f4431c56 | 1439 | STRLEN hv_fill = HvFILL(ohv); |
5b9c0671 NC |
1440 | HE *entry; |
1441 | const I32 riter = HvRITER_get(ohv); | |
1442 | HE * const eiter = HvEITER_get(ohv); | |
1443 | ||
1444 | while (hv_max && hv_max + 1 >= hv_fill * 2) | |
1445 | hv_max = hv_max / 2; | |
1446 | HvMAX(hv) = hv_max; | |
1447 | ||
1448 | hv_iterinit(ohv); | |
1449 | while ((entry = hv_iternext_flags(ohv, 0))) { | |
1450 | SV *const sv = newSVsv(HeVAL(entry)); | |
e3b1b6b1 | 1451 | SV *heksv = newSVhek(HeKEY_hek(entry)); |
5b9c0671 | 1452 | sv_magic(sv, NULL, PERL_MAGIC_hintselem, |
e3b1b6b1 RGS |
1453 | (char *)heksv, HEf_SVKEY); |
1454 | SvREFCNT_dec(heksv); | |
04fe65b0 RGS |
1455 | (void)hv_store_flags(hv, HeKEY(entry), HeKLEN(entry), |
1456 | sv, HeHASH(entry), HeKFLAGS(entry)); | |
5b9c0671 NC |
1457 | } |
1458 | HvRITER_set(ohv, riter); | |
1459 | HvEITER_set(ohv, eiter); | |
1460 | } | |
1461 | hv_magic(hv, NULL, PERL_MAGIC_hints); | |
1462 | return hv; | |
1463 | } | |
1464 | ||
79072805 | 1465 | void |
864dbfa3 | 1466 | Perl_hv_free_ent(pTHX_ HV *hv, register HE *entry) |
79072805 | 1467 | { |
97aff369 | 1468 | dVAR; |
16bdeea2 GS |
1469 | SV *val; |
1470 | ||
7918f24d NC |
1471 | PERL_ARGS_ASSERT_HV_FREE_ENT; |
1472 | ||
68dc0745 | 1473 | if (!entry) |
79072805 | 1474 | return; |
16bdeea2 | 1475 | val = HeVAL(entry); |
00169e2c | 1476 | if (val && isGV(val) && isGV_with_GP(val) && GvCVu(val) && HvENAME(hv)) |
803f2748 | 1477 | mro_method_changed_in(hv); /* deletion of method from stash */ |
16bdeea2 | 1478 | SvREFCNT_dec(val); |
68dc0745 | 1479 | if (HeKLEN(entry) == HEf_SVKEY) { |
1480 | SvREFCNT_dec(HeKEY_sv(entry)); | |
8aacddc1 | 1481 | Safefree(HeKEY_hek(entry)); |
44a8e56a | 1482 | } |
1483 | else if (HvSHAREKEYS(hv)) | |
68dc0745 | 1484 | unshare_hek(HeKEY_hek(entry)); |
fde52b5c | 1485 | else |
68dc0745 | 1486 | Safefree(HeKEY_hek(entry)); |
d33b2eba | 1487 | del_HE(entry); |
79072805 LW |
1488 | } |
1489 | ||
f1c32fec | 1490 | |
79072805 | 1491 | void |
864dbfa3 | 1492 | Perl_hv_delayfree_ent(pTHX_ HV *hv, register HE *entry) |
79072805 | 1493 | { |
97aff369 | 1494 | dVAR; |
7918f24d NC |
1495 | |
1496 | PERL_ARGS_ASSERT_HV_DELAYFREE_ENT; | |
1497 | ||
68dc0745 | 1498 | if (!entry) |
79072805 | 1499 | return; |
bc4947fc NC |
1500 | /* SvREFCNT_inc to counter the SvREFCNT_dec in hv_free_ent */ |
1501 | sv_2mortal(SvREFCNT_inc(HeVAL(entry))); /* free between statements */ | |
68dc0745 | 1502 | if (HeKLEN(entry) == HEf_SVKEY) { |
bc4947fc | 1503 | sv_2mortal(SvREFCNT_inc(HeKEY_sv(entry))); |
44a8e56a | 1504 | } |
bc4947fc | 1505 | hv_free_ent(hv, entry); |
79072805 LW |
1506 | } |
1507 | ||
954c1994 GS |
1508 | /* |
1509 | =for apidoc hv_clear | |
1510 | ||
1511 | Clears a hash, making it empty. | |
1512 | ||
1513 | =cut | |
1514 | */ | |
1515 | ||
79072805 | 1516 | void |
864dbfa3 | 1517 | Perl_hv_clear(pTHX_ HV *hv) |
79072805 | 1518 | { |
27da23d5 | 1519 | dVAR; |
cbec9347 | 1520 | register XPVHV* xhv; |
79072805 LW |
1521 | if (!hv) |
1522 | return; | |
49293501 | 1523 | |
ecae49c0 NC |
1524 | DEBUG_A(Perl_hv_assert(aTHX_ hv)); |
1525 | ||
34c3c4e3 DM |
1526 | xhv = (XPVHV*)SvANY(hv); |
1527 | ||
7b2c381c | 1528 | if (SvREADONLY(hv) && HvARRAY(hv) != NULL) { |
34c3c4e3 | 1529 | /* restricted hash: convert all keys to placeholders */ |
b464bac0 AL |
1530 | STRLEN i; |
1531 | for (i = 0; i <= xhv->xhv_max; i++) { | |
7b2c381c | 1532 | HE *entry = (HvARRAY(hv))[i]; |
3a676441 JH |
1533 | for (; entry; entry = HeNEXT(entry)) { |
1534 | /* not already placeholder */ | |
7996736c | 1535 | if (HeVAL(entry) != &PL_sv_placeholder) { |
3a676441 | 1536 | if (HeVAL(entry) && SvREADONLY(HeVAL(entry))) { |
6136c704 | 1537 | SV* const keysv = hv_iterkeysv(entry); |
3a676441 | 1538 | Perl_croak(aTHX_ |
95b63a38 JH |
1539 | "Attempt to delete readonly key '%"SVf"' from a restricted hash", |
1540 | (void*)keysv); | |
3a676441 JH |
1541 | } |
1542 | SvREFCNT_dec(HeVAL(entry)); | |
7996736c | 1543 | HeVAL(entry) = &PL_sv_placeholder; |
ca732855 | 1544 | HvPLACEHOLDERS(hv)++; |
3a676441 | 1545 | } |
34c3c4e3 DM |
1546 | } |
1547 | } | |
df8c6964 | 1548 | goto reset; |
49293501 MS |
1549 | } |
1550 | ||
463ee0b2 | 1551 | hfreeentries(hv); |
ca732855 | 1552 | HvPLACEHOLDERS_set(hv, 0); |
7b2c381c | 1553 | if (HvARRAY(hv)) |
41f62432 | 1554 | Zero(HvARRAY(hv), xhv->xhv_max+1 /* HvMAX(hv)+1 */, HE*); |
a0d0e21e LW |
1555 | |
1556 | if (SvRMAGICAL(hv)) | |
ad64d0ec | 1557 | mg_clear(MUTABLE_SV(hv)); |
574c8022 | 1558 | |
19692e8d | 1559 | HvHASKFLAGS_off(hv); |
bb443f97 | 1560 | HvREHASH_off(hv); |
df8c6964 | 1561 | reset: |
b79f7545 | 1562 | if (SvOOK(hv)) { |
00169e2c | 1563 | if(HvENAME_get(hv)) |
dd69841b | 1564 | mro_isa_changed_in(hv); |
bfcb3514 NC |
1565 | HvEITER_set(hv, NULL); |
1566 | } | |
79072805 LW |
1567 | } |
1568 | ||
3540d4ce AB |
1569 | /* |
1570 | =for apidoc hv_clear_placeholders | |
1571 | ||
1572 | Clears any placeholders from a hash. If a restricted hash has any of its keys | |
1573 | marked as readonly and the key is subsequently deleted, the key is not actually | |
1574 | deleted but is marked by assigning it a value of &PL_sv_placeholder. This tags | |
1575 | it so it will be ignored by future operations such as iterating over the hash, | |
4cdaeff7 | 1576 | but will still allow the hash to have a value reassigned to the key at some |
3540d4ce AB |
1577 | future point. This function clears any such placeholder keys from the hash. |
1578 | See Hash::Util::lock_keys() for an example of its use. | |
1579 | ||
1580 | =cut | |
1581 | */ | |
1582 | ||
1583 | void | |
1584 | Perl_hv_clear_placeholders(pTHX_ HV *hv) | |
1585 | { | |
27da23d5 | 1586 | dVAR; |
b3ca2e83 NC |
1587 | const U32 items = (U32)HvPLACEHOLDERS_get(hv); |
1588 | ||
7918f24d NC |
1589 | PERL_ARGS_ASSERT_HV_CLEAR_PLACEHOLDERS; |
1590 | ||
b3ca2e83 NC |
1591 | if (items) |
1592 | clear_placeholders(hv, items); | |
1593 | } | |
1594 | ||
1595 | static void | |
1596 | S_clear_placeholders(pTHX_ HV *hv, U32 items) | |
1597 | { | |
1598 | dVAR; | |
b464bac0 | 1599 | I32 i; |
d3677389 | 1600 | |
7918f24d NC |
1601 | PERL_ARGS_ASSERT_CLEAR_PLACEHOLDERS; |
1602 | ||
d3677389 NC |
1603 | if (items == 0) |
1604 | return; | |
1605 | ||
b464bac0 | 1606 | i = HvMAX(hv); |
d3677389 NC |
1607 | do { |
1608 | /* Loop down the linked list heads */ | |
6136c704 | 1609 | bool first = TRUE; |
d3677389 | 1610 | HE **oentry = &(HvARRAY(hv))[i]; |
cf6db12b | 1611 | HE *entry; |
d3677389 | 1612 | |
cf6db12b | 1613 | while ((entry = *oentry)) { |
d3677389 NC |
1614 | if (HeVAL(entry) == &PL_sv_placeholder) { |
1615 | *oentry = HeNEXT(entry); | |
2e58978b | 1616 | if (entry == HvEITER_get(hv)) |
d3677389 NC |
1617 | HvLAZYDEL_on(hv); |
1618 | else | |
1619 | hv_free_ent(hv, entry); | |
1620 | ||
1621 | if (--items == 0) { | |
1622 | /* Finished. */ | |
5d88ecd7 | 1623 | HvTOTALKEYS(hv) -= (IV)HvPLACEHOLDERS_get(hv); |
d3677389 NC |
1624 | if (HvKEYS(hv) == 0) |
1625 | HvHASKFLAGS_off(hv); | |
5d88ecd7 | 1626 | HvPLACEHOLDERS_set(hv, 0); |
d3677389 NC |
1627 | return; |
1628 | } | |
213ce8b3 NC |
1629 | } else { |
1630 | oentry = &HeNEXT(entry); | |
6136c704 | 1631 | first = FALSE; |
d3677389 NC |
1632 | } |
1633 | } | |
1634 | } while (--i >= 0); | |
1635 | /* You can't get here, hence assertion should always fail. */ | |
1636 | assert (items == 0); | |
1637 | assert (0); | |
3540d4ce AB |
1638 | } |
1639 | ||
76e3520e | 1640 | STATIC void |
cea2e8a9 | 1641 | S_hfreeentries(pTHX_ HV *hv) |
79072805 | 1642 | { |
23976bdd | 1643 | /* This is the array that we're going to restore */ |
fd7de8a8 | 1644 | HE **const orig_array = HvARRAY(hv); |
2d0d1ecc FC |
1645 | HE **tmp_array = NULL; |
1646 | const bool has_aux = SvOOK(hv); | |
1647 | struct xpvhv_aux * current_aux = NULL; | |
23976bdd | 1648 | int attempts = 100; |
3abe233e | 1649 | |
7918f24d NC |
1650 | PERL_ARGS_ASSERT_HFREEENTRIES; |
1651 | ||
fd7de8a8 | 1652 | if (!orig_array) |
79072805 | 1653 | return; |
a0d0e21e | 1654 | |
23976bdd NC |
1655 | /* orig_array remains unchanged throughout the loop. If after freeing all |
1656 | the entries it turns out that one of the little blighters has triggered | |
1657 | an action that has caused HvARRAY to be re-allocated, then we set | |
1658 | array to the new HvARRAY, and try again. */ | |
1659 | ||
1660 | while (1) { | |
1661 | /* This is the one we're going to try to empty. First time round | |
1662 | it's the original array. (Hopefully there will only be 1 time | |
1663 | round) */ | |
6136c704 | 1664 | HE ** const array = HvARRAY(hv); |
7440661e | 1665 | I32 i = HvMAX(hv); |
23976bdd | 1666 | |
2d0d1ecc FC |
1667 | struct xpvhv_aux *iter = SvOOK(hv) ? HvAUX(hv) : NULL; |
1668 | ||
1669 | /* make everyone else think the array is empty, so that the destructors | |
1670 | * called for freed entries can't recursively mess with us */ | |
1671 | HvARRAY(hv) = NULL; | |
23976bdd NC |
1672 | |
1673 | if (SvOOK(hv)) { | |
7440661e | 1674 | HE *entry; |
2d0d1ecc FC |
1675 | |
1676 | SvFLAGS(hv) &= ~SVf_OOK; /* Goodbye, aux structure. */ | |
1677 | /* What aux structure? */ | |
1678 | /* (But we still have a pointer to it in iter.) */ | |
1679 | ||
47f1cf77 FC |
1680 | /* Copy the name and MRO stuff to a new aux structure |
1681 | if present. */ | |
1682 | if (iter->xhv_name || iter->xhv_mro_meta) { | |
2d0d1ecc FC |
1683 | struct xpvhv_aux * const newaux = hv_auxinit(hv); |
1684 | newaux->xhv_name = iter->xhv_name; | |
1685 | newaux->xhv_name_count = iter->xhv_name_count; | |
1686 | iter->xhv_name = NULL; | |
47f1cf77 FC |
1687 | newaux->xhv_mro_meta = iter->xhv_mro_meta; |
1688 | iter->xhv_mro_meta = NULL; | |
2d0d1ecc FC |
1689 | } |
1690 | ||
47f1cf77 FC |
1691 | /* Because we have taken xhv_name and |
1692 | xhv_mro_meta out, the only allocated | |
2d0d1ecc | 1693 | pointers in the aux structure that might exist are the back- |
47f1cf77 | 1694 | reference array and xhv_eiter. |
2d0d1ecc FC |
1695 | */ |
1696 | ||
cdb996f4 DM |
1697 | /* weak references: if called from sv_clear(), the backrefs |
1698 | * should already have been killed; if there are any left, its | |
1699 | * because we're doing hv_clear() or hv_undef(), and the HV | |
1700 | * will continue to live. | |
1701 | * Because while freeing the entries we fake up a NULL HvARRAY | |
1702 | * (and hence HvAUX), we need to store the backref array | |
1703 | * somewhere else; but it still needs to be visible in case | |
1704 | * any the things we free happen to call sv_del_backref(). | |
1705 | * We do this by storing it in magic instead. | |
1706 | * If, during the entry freeing, a destructor happens to add | |
1707 | * a new weak backref, then sv_add_backref will look in both | |
1708 | * places (magic in HvAUX) for the AV, but will create a new | |
8ac9a371 DM |
1709 | * AV in HvAUX if it can't find one (if it finds it in magic, |
1710 | * it moves it back into HvAUX. So at the end of the iteration | |
1711 | * we have to allow for this. */ | |
1712 | ||
09aad8f0 DM |
1713 | |
1714 | if (iter->xhv_backreferences) { | |
5648c0ae DM |
1715 | if (SvTYPE(iter->xhv_backreferences) == SVt_PVAV) { |
1716 | /* The sv_magic will increase the reference count of the AV, | |
1717 | so we need to drop it first. */ | |
09aad8f0 | 1718 | SvREFCNT_dec(iter->xhv_backreferences); |
5648c0ae DM |
1719 | if (AvFILLp(iter->xhv_backreferences) == -1) { |
1720 | /* Turns out that the array is empty. Just free it. */ | |
1721 | SvREFCNT_dec(iter->xhv_backreferences); | |
1b8791d1 | 1722 | |
5648c0ae DM |
1723 | } else { |
1724 | sv_magic(MUTABLE_SV(hv), | |
1725 | MUTABLE_SV(iter->xhv_backreferences), | |
1726 | PERL_MAGIC_backref, NULL, 0); | |
1727 | } | |
1728 | } | |
1729 | else { | |
1730 | MAGIC *mg; | |
1731 | sv_magic(MUTABLE_SV(hv), NULL, PERL_MAGIC_backref, NULL, 0); | |
1732 | mg = mg_find(MUTABLE_SV(hv), PERL_MAGIC_backref); | |
1733 | mg->mg_obj = (SV*)iter->xhv_backreferences; | |
09aad8f0 DM |
1734 | } |
1735 | iter->xhv_backreferences = NULL; | |
5b285ea4 | 1736 | } |
86f55936 | 1737 | |
23976bdd NC |
1738 | entry = iter->xhv_eiter; /* HvEITER(hv) */ |
1739 | if (entry && HvLAZYDEL(hv)) { /* was deleted earlier? */ | |
1740 | HvLAZYDEL_off(hv); | |
1741 | hv_free_ent(hv, entry); | |
1742 | } | |
1743 | iter->xhv_riter = -1; /* HvRITER(hv) = -1 */ | |
4608196e | 1744 | iter->xhv_eiter = NULL; /* HvEITER(hv) = NULL */ |
b79f7545 | 1745 | |
23976bdd | 1746 | /* There are now no allocated pointers in the aux structure. */ |
a0d0e21e | 1747 | } |
bfcb3514 | 1748 | |
2d0d1ecc FC |
1749 | /* If there are no keys, there is nothing left to free. */ |
1750 | if (!((XPVHV*) SvANY(hv))->xhv_keys) break; | |
23976bdd | 1751 | |
2d0d1ecc FC |
1752 | /* Since we have removed the HvARRAY (and possibly replaced it by |
1753 | calling hv_auxinit), set the number of keys accordingly. */ | |
1754 | ((XPVHV*) SvANY(hv))->xhv_keys = 0; | |
7440661e NC |
1755 | |
1756 | do { | |
1757 | /* Loop down the linked list heads */ | |
1758 | HE *entry = array[i]; | |
1759 | ||
1760 | while (entry) { | |
23976bdd NC |
1761 | register HE * const oentry = entry; |
1762 | entry = HeNEXT(entry); | |
1763 | hv_free_ent(hv, oentry); | |
1764 | } | |
7440661e | 1765 | } while (--i >= 0); |
b79f7545 | 1766 | |
23976bdd NC |
1767 | /* As there are no allocated pointers in the aux structure, it's now |
1768 | safe to free the array we just cleaned up, if it's not the one we're | |
1769 | going to put back. */ | |
1770 | if (array != orig_array) { | |
1771 | Safefree(array); | |
1772 | } | |
b79f7545 | 1773 | |
23976bdd NC |
1774 | if (!HvARRAY(hv)) { |
1775 | /* Good. No-one added anything this time round. */ | |
1776 | break; | |
bfcb3514 | 1777 | } |
b79f7545 | 1778 | |
23976bdd NC |
1779 | if (--attempts == 0) { |
1780 | Perl_die(aTHX_ "panic: hfreeentries failed to free hash - something is repeatedly re-creating entries"); | |
1781 | } | |
6136c704 | 1782 | } |
23976bdd | 1783 | |
2d0d1ecc FC |
1784 | /* Set aside the current array for now, in case we still need it. */ |
1785 | if (SvOOK(hv)) current_aux = HvAUX(hv); | |
1786 | if (HvARRAY(hv) && HvARRAY(hv) != orig_array) | |
1787 | tmp_array = HvARRAY(hv); | |
1788 | ||
23976bdd NC |
1789 | HvARRAY(hv) = orig_array; |
1790 | ||
2d0d1ecc | 1791 | if (has_aux) |
23976bdd | 1792 | SvFLAGS(hv) |= SVf_OOK; |
2d0d1ecc FC |
1793 | else |
1794 | SvFLAGS(hv) &=~SVf_OOK; | |
1795 | ||
47f1cf77 FC |
1796 | /* If the hash was actually a symbol table, put the name and MRO |
1797 | caches back. */ | |
2d0d1ecc FC |
1798 | if (current_aux) { |
1799 | struct xpvhv_aux * const aux | |
1800 | = SvOOK(hv) ? HvAUX(hv) : hv_auxinit(hv); | |
1801 | aux->xhv_name = current_aux->xhv_name; | |
1802 | aux->xhv_name_count = current_aux->xhv_name_count; | |
47f1cf77 | 1803 | aux->xhv_mro_meta = current_aux->xhv_mro_meta; |
1b8791d1 | 1804 | } |
2d0d1ecc FC |
1805 | |
1806 | if (tmp_array) Safefree(tmp_array); | |
79072805 LW |
1807 | } |
1808 | ||
954c1994 GS |
1809 | /* |
1810 | =for apidoc hv_undef | |
1811 | ||
1812 | Undefines the hash. | |
1813 | ||
1814 | =cut | |
1815 | */ | |
1816 | ||
79072805 | 1817 | void |
864dbfa3 | 1818 | Perl_hv_undef(pTHX_ HV *hv) |
79072805 | 1819 | { |
97aff369 | 1820 | dVAR; |
cbec9347 | 1821 | register XPVHV* xhv; |
bfcb3514 | 1822 | const char *name; |
86f55936 | 1823 | |
79072805 LW |
1824 | if (!hv) |
1825 | return; | |
ecae49c0 | 1826 | DEBUG_A(Perl_hv_assert(aTHX_ hv)); |
cbec9347 | 1827 | xhv = (XPVHV*)SvANY(hv); |
dd69841b | 1828 | |
627364f1 | 1829 | if ((name = HvENAME_get(hv)) && PL_phase != PERL_PHASE_DESTRUCT) |
b9e30492 FC |
1830 | { |
1831 | /* Delete the @ISA element before calling mro_package_moved, so it | |
1832 | does not see it. */ | |
1833 | (void)hv_delete(hv, "ISA", 3, G_DISCARD); | |
1834 | mro_package_moved(NULL, hv, NULL, name, HvENAMELEN_get(hv)); | |
1835 | } | |
dd69841b | 1836 | |
00169e2c | 1837 | if (name || (name = HvNAME(hv))) { |
04fe65b0 RGS |
1838 | if (PL_stashcache) |
1839 | (void)hv_delete(PL_stashcache, name, HvNAMELEN_get(hv), G_DISCARD); | |
bd61b366 | 1840 | hv_name_set(hv, NULL, 0, 0); |
85e6fe83 | 1841 | } |
2d0d1ecc | 1842 | hfreeentries(hv); |
47f1cf77 FC |
1843 | if (SvOOK(hv)) { |
1844 | struct xpvhv_aux * const aux = HvAUX(hv); | |
1845 | struct mro_meta *meta; | |
ecbbcad6 FC |
1846 | if (aux->xhv_name) { |
1847 | if (PL_stashcache && (name = HvNAME(hv))) | |
2d0d1ecc FC |
1848 | (void)hv_delete(PL_stashcache, name, HvNAMELEN_get(hv), G_DISCARD); |
1849 | hv_name_set(hv, NULL, 0, 0); | |
47f1cf77 FC |
1850 | } |
1851 | if((meta = aux->xhv_mro_meta)) { | |
1852 | if (meta->mro_linear_all) { | |
1853 | SvREFCNT_dec(MUTABLE_SV(meta->mro_linear_all)); | |
1854 | meta->mro_linear_all = NULL; | |
1855 | /* This is just acting as a shortcut pointer. */ | |
1856 | meta->mro_linear_current = NULL; | |
1857 | } else if (meta->mro_linear_current) { | |
1858 | /* Only the current MRO is stored, so this owns the data. | |
1859 | */ | |
1860 | SvREFCNT_dec(meta->mro_linear_current); | |
1861 | meta->mro_linear_current = NULL; | |
1862 | } | |
1863 | if(meta->mro_nextmethod) SvREFCNT_dec(meta->mro_nextmethod); | |
1864 | SvREFCNT_dec(meta->isa); | |
1865 | Safefree(meta); | |
1866 | aux->xhv_mro_meta = NULL; | |
1867 | } | |
2d0d1ecc | 1868 | } |
b79f7545 NC |
1869 | SvFLAGS(hv) &= ~SVf_OOK; |
1870 | Safefree(HvARRAY(hv)); | |
cbec9347 | 1871 | xhv->xhv_max = 7; /* HvMAX(hv) = 7 (it's a normal hash) */ |
7b2c381c | 1872 | HvARRAY(hv) = 0; |
ca732855 | 1873 | HvPLACEHOLDERS_set(hv, 0); |
a0d0e21e LW |
1874 | |
1875 | if (SvRMAGICAL(hv)) | |
ad64d0ec | 1876 | mg_clear(MUTABLE_SV(hv)); |
79072805 LW |
1877 | } |
1878 | ||
4d0fbddd NC |
1879 | /* |
1880 | =for apidoc hv_fill | |
1881 | ||
1882 | Returns the number of hash buckets that happen to be in use. This function is | |
1883 | wrapped by the macro C<HvFILL>. | |
1884 | ||
1885 | Previously this value was stored in the HV structure, rather than being | |
1886 | calculated on demand. | |
1887 | ||
1888 | =cut | |
1889 | */ | |
1890 | ||
1891 | STRLEN | |
1892 | Perl_hv_fill(pTHX_ HV const *const hv) | |
1893 | { | |
1894 | STRLEN count = 0; | |
1895 | HE **ents = HvARRAY(hv); | |
1896 | ||
1897 | PERL_ARGS_ASSERT_HV_FILL; | |
1898 | ||
1899 | if (ents) { | |
fcd24582 NC |
1900 | HE *const *const last = ents + HvMAX(hv); |
1901 | count = last + 1 - ents; | |
4d0fbddd NC |
1902 | |
1903 | do { | |
fcd24582 NC |
1904 | if (!*ents) |
1905 | --count; | |
1906 | } while (++ents <= last); | |
4d0fbddd NC |
1907 | } |
1908 | return count; | |
1909 | } | |
1910 | ||
b464bac0 | 1911 | static struct xpvhv_aux* |
5f66b61c | 1912 | S_hv_auxinit(HV *hv) { |
bfcb3514 | 1913 | struct xpvhv_aux *iter; |
b79f7545 | 1914 | char *array; |
bfcb3514 | 1915 | |
7918f24d NC |
1916 | PERL_ARGS_ASSERT_HV_AUXINIT; |
1917 | ||
b79f7545 | 1918 | if (!HvARRAY(hv)) { |
a02a5408 | 1919 | Newxz(array, PERL_HV_ARRAY_ALLOC_BYTES(HvMAX(hv) + 1) |
b79f7545 NC |
1920 | + sizeof(struct xpvhv_aux), char); |
1921 | } else { | |
1922 | array = (char *) HvARRAY(hv); | |
1923 | Renew(array, PERL_HV_ARRAY_ALLOC_BYTES(HvMAX(hv) + 1) | |
1924 | + sizeof(struct xpvhv_aux), char); | |
1925 | } | |
1926 | HvARRAY(hv) = (HE**) array; | |
1927 | /* SvOOK_on(hv) attacks the IV flags. */ | |
1928 | SvFLAGS(hv) |= SVf_OOK; | |
1929 | iter = HvAUX(hv); | |
bfcb3514 NC |
1930 | |
1931 | iter->xhv_riter = -1; /* HvRITER(hv) = -1 */ | |
4608196e | 1932 | iter->xhv_eiter = NULL; /* HvEITER(hv) = NULL */ |
bfcb3514 | 1933 | iter->xhv_name = 0; |
b7247a80 | 1934 | iter->xhv_name_count = 0; |
86f55936 | 1935 | iter->xhv_backreferences = 0; |
e1a479c5 | 1936 | iter->xhv_mro_meta = NULL; |
bfcb3514 NC |
1937 | return iter; |
1938 | } | |
1939 | ||
954c1994 GS |
1940 | /* |
1941 | =for apidoc hv_iterinit | |
1942 | ||
1943 | Prepares a starting point to traverse a hash table. Returns the number of | |
b24b84ef | 1944 | keys in the hash (i.e. the same as C<HvKEYS(hv)>). The return value is |
1c846c1f | 1945 | currently only meaningful for hashes without tie magic. |
954c1994 GS |
1946 | |
1947 | NOTE: Before version 5.004_65, C<hv_iterinit> used to return the number of | |
1948 | hash buckets that happen to be in use. If you still need that esoteric | |
b24b84ef | 1949 | value, you can get it through the macro C<HvFILL(hv)>. |
954c1994 | 1950 | |
e16e2ff8 | 1951 | |
954c1994 GS |
1952 | =cut |
1953 | */ | |
1954 | ||
79072805 | 1955 | I32 |
864dbfa3 | 1956 | Perl_hv_iterinit(pTHX_ HV *hv) |
79072805 | 1957 | { |
7918f24d NC |
1958 | PERL_ARGS_ASSERT_HV_ITERINIT; |
1959 | ||
1960 | /* FIXME: Are we not NULL, or do we croak? Place bets now! */ | |
1961 | ||
aa689395 | 1962 | if (!hv) |
cea2e8a9 | 1963 | Perl_croak(aTHX_ "Bad hash"); |
bfcb3514 | 1964 | |
b79f7545 | 1965 | if (SvOOK(hv)) { |
6136c704 | 1966 | struct xpvhv_aux * const iter = HvAUX(hv); |
0bd48802 | 1967 | HE * const entry = iter->xhv_eiter; /* HvEITER(hv) */ |
bfcb3514 NC |
1968 | if (entry && HvLAZYDEL(hv)) { /* was deleted earlier? */ |
1969 | HvLAZYDEL_off(hv); | |
1970 | hv_free_ent(hv, entry); | |
1971 | } | |
1972 | iter->xhv_riter = -1; /* HvRITER(hv) = -1 */ | |
4608196e | 1973 | iter->xhv_eiter = NULL; /* HvEITER(hv) = NULL */ |
bfcb3514 | 1974 | } else { |
6136c704 | 1975 | hv_auxinit(hv); |
72940dca | 1976 | } |
44a2ac75 | 1977 | |
cbec9347 | 1978 | /* used to be xhv->xhv_fill before 5.004_65 */ |
5d88ecd7 | 1979 | return HvTOTALKEYS(hv); |
79072805 | 1980 | } |
bfcb3514 NC |
1981 | |
1982 | I32 * | |
1983 | Perl_hv_riter_p(pTHX_ HV *hv) { | |
1984 | struct xpvhv_aux *iter; | |
1985 | ||
7918f24d NC |
1986 | PERL_ARGS_ASSERT_HV_RITER_P; |
1987 | ||
bfcb3514 NC |
1988 | if (!hv) |
1989 | Perl_croak(aTHX_ "Bad hash"); | |
1990 | ||
6136c704 | 1991 | iter = SvOOK(hv) ? HvAUX(hv) : hv_auxinit(hv); |
bfcb3514 NC |
1992 | return &(iter->xhv_riter); |
1993 | } | |
1994 | ||
1995 | HE ** | |
1996 | Perl_hv_eiter_p(pTHX_ HV *hv) { | |
1997 | struct xpvhv_aux *iter; | |
1998 | ||
7918f24d NC |
1999 | PERL_ARGS_ASSERT_HV_EITER_P; |
2000 | ||
bfcb3514 NC |
2001 | if (!hv) |
2002 | Perl_croak(aTHX_ "Bad hash"); | |
2003 | ||
6136c704 | 2004 | iter = SvOOK(hv) ? HvAUX(hv) : hv_auxinit(hv); |
bfcb3514 NC |
2005 | return &(iter->xhv_eiter); |
2006 | } | |
2007 | ||
2008 | void | |
2009 | Perl_hv_riter_set(pTHX_ HV *hv, I32 riter) { | |
2010 | struct xpvhv_aux *iter; | |
2011 | ||
7918f24d NC |
2012 | PERL_ARGS_ASSERT_HV_RITER_SET; |
2013 | ||
bfcb3514 NC |
2014 | if (!hv) |
2015 | Perl_croak(aTHX_ "Bad hash"); | |
2016 | ||
b79f7545 NC |
2017 | if (SvOOK(hv)) { |
2018 | iter = HvAUX(hv); | |
2019 | } else { | |
bfcb3514 NC |
2020 | if (riter == -1) |
2021 | return; | |
2022 | ||
6136c704 | 2023 | iter = hv_auxinit(hv); |
bfcb3514 NC |
2024 | } |
2025 | iter->xhv_riter = riter; | |
2026 | } | |
2027 | ||
2028 | void | |
2029 | Perl_hv_eiter_set(pTHX_ HV *hv, HE *eiter) { | |
2030 | struct xpvhv_aux *iter; | |
2031 | ||
7918f24d NC |
2032 | PERL_ARGS_ASSERT_HV_EITER_SET; |
2033 | ||
bfcb3514 NC |
2034 | if (!hv) |
2035 | Perl_croak(aTHX_ "Bad hash"); | |
2036 | ||
b79f7545 NC |
2037 | if (SvOOK(hv)) { |
2038 | iter = HvAUX(hv); | |
2039 | } else { | |
bfcb3514 NC |
2040 | /* 0 is the default so don't go malloc()ing a new structure just to |
2041 | hold 0. */ | |
2042 | if (!eiter) | |
2043 | return; | |
2044 | ||
6136c704 | 2045 | iter = hv_auxinit(hv); |
bfcb3514 NC |
2046 | } |
2047 | iter->xhv_eiter = eiter; | |
2048 | } | |
2049 | ||
bfcb3514 | 2050 | void |
4164be69 | 2051 | Perl_hv_name_set(pTHX_ HV *hv, const char *name, U32 len, U32 flags) |
bfcb3514 | 2052 | { |
97aff369 | 2053 | dVAR; |
b79f7545 | 2054 | struct xpvhv_aux *iter; |
7423f6db | 2055 | U32 hash; |
78b79c77 | 2056 | HEK **spot; |
46c461b5 | 2057 | |
7918f24d | 2058 | PERL_ARGS_ASSERT_HV_NAME_SET; |
46c461b5 | 2059 | PERL_UNUSED_ARG(flags); |
bfcb3514 | 2060 | |
4164be69 NC |
2061 | if (len > I32_MAX) |
2062 | Perl_croak(aTHX_ "panic: hv name too long (%"UVuf")", (UV) len); | |
2063 | ||
b79f7545 NC |
2064 | if (SvOOK(hv)) { |
2065 | iter = HvAUX(hv); | |
7423f6db | 2066 | if (iter->xhv_name) { |
b7247a80 | 2067 | if(iter->xhv_name_count) { |
78b79c77 | 2068 | if(!name) { |
b7247a80 | 2069 | HEK ** const name = (HEK **)HvAUX(hv)->xhv_name; |
78b79c77 FC |
2070 | HEK **hekp = name + ( |
2071 | iter->xhv_name_count < 0 | |
2072 | ? -iter->xhv_name_count | |
2073 | : iter->xhv_name_count | |
2074 | ); | |
2075 | while(hekp-- > name+1) | |
b7247a80 | 2076 | unshare_hek_or_pvn(*hekp, 0, 0, 0); |
78b79c77 FC |
2077 | /* The first elem may be null. */ |
2078 | if(*name) unshare_hek_or_pvn(*name, 0, 0, 0); | |
b7247a80 | 2079 | Safefree(name); |
78b79c77 FC |
2080 | spot = &iter->xhv_name; |
2081 | iter->xhv_name_count = 0; | |
2082 | } | |
2083 | else { | |
2084 | spot = (HEK **)iter->xhv_name; | |
2085 | if(iter->xhv_name_count > 0) { | |
2086 | /* shift some things over */ | |
2087 | Renew(spot, iter->xhv_name_count, HEK *); | |
2088 | spot[iter->xhv_name_count++] = spot[1]; | |
2089 | spot[1] = spot[0]; | |
2090 | } | |
2091 | else if(*spot) { | |
2092 | unshare_hek_or_pvn(*spot, 0, 0, 0); | |
2093 | } | |
2094 | } | |
2095 | } | |
2096 | else { | |
2097 | unshare_hek_or_pvn(iter->xhv_name, 0, 0, 0); | |
2098 | spot = &iter->xhv_name; | |
b7247a80 | 2099 | } |
7423f6db | 2100 | } |
78b79c77 | 2101 | else spot = &iter->xhv_name; |
16580ff5 | 2102 | } else { |
bfcb3514 NC |
2103 | if (name == 0) |
2104 | return; | |
2105 | ||
6136c704 | 2106 | iter = hv_auxinit(hv); |
78b79c77 | 2107 | spot = &iter->xhv_name; |
bfcb3514 | 2108 | } |
7423f6db | 2109 | PERL_HASH(hash, name, len); |
78b79c77 | 2110 | *spot = name ? share_hek(name, len, hash) : NULL; |
b7247a80 | 2111 | iter->xhv_name_count = 0; |
bfcb3514 NC |
2112 | } |
2113 | ||
99206677 FC |
2114 | /* |
2115 | =for apidoc hv_ename_add | |
2116 | ||
2117 | Adds a name to a stash's internal list of effective names. See | |
2118 | C<hv_ename_delete>. | |
2119 | ||
2120 | This is called when a stash is assigned to a new location in the symbol | |
2121 | table. | |
2122 | ||
2123 | =cut | |
2124 | */ | |
2125 | ||
ee72b38d | 2126 | void |
78b79c77 | 2127 | Perl_hv_ename_add(pTHX_ HV *hv, const char *name, U32 len) |
ee72b38d FC |
2128 | { |
2129 | dVAR; | |
2130 | struct xpvhv_aux *aux = SvOOK(hv) ? HvAUX(hv) : hv_auxinit(hv); | |
2131 | U32 hash; | |
2132 | ||
78b79c77 | 2133 | PERL_ARGS_ASSERT_HV_ENAME_ADD; |
ee72b38d FC |
2134 | |
2135 | if (len > I32_MAX) | |
2136 | Perl_croak(aTHX_ "panic: hv name too long (%"UVuf")", (UV) len); | |
2137 | ||
2138 | PERL_HASH(hash, name, len); | |
2139 | ||
ee72b38d FC |
2140 | if (aux->xhv_name_count) { |
2141 | HEK ** const xhv_name = (HEK **)aux->xhv_name; | |
78b79c77 FC |
2142 | I32 count = aux->xhv_name_count; |
2143 | HEK **hekp = xhv_name + (count < 0 ? -count : count); | |
ee72b38d FC |
2144 | while (hekp-- > xhv_name) |
2145 | if ( | |
2146 | HEK_LEN(*hekp) == (I32)len && memEQ(HEK_KEY(*hekp), name, len) | |
78b79c77 FC |
2147 | ) { |
2148 | if (hekp == xhv_name && count < 0) | |
2149 | aux->xhv_name_count = -count; | |
2150 | return; | |
2151 | } | |
2152 | if (count < 0) aux->xhv_name_count--, count = -count; | |
2153 | else aux->xhv_name_count++; | |
2154 | Renewc(aux->xhv_name, count + 1, HEK *, HEK); | |
ee72b38d FC |
2155 | ((HEK **)aux->xhv_name)[count] = share_hek(name, len, hash); |
2156 | } | |
2157 | else { | |
2158 | HEK *existing_name = aux->xhv_name; | |
2159 | if ( | |
78b79c77 | 2160 | existing_name && HEK_LEN(existing_name) == (I32)len |
ee72b38d FC |
2161 | && memEQ(HEK_KEY(existing_name), name, len) |
2162 | ) return; | |
96517a90 | 2163 | Newxc(aux->xhv_name, 2, HEK *, HEK); |
78b79c77 | 2164 | aux->xhv_name_count = existing_name ? 2 : -2; |
ee72b38d FC |
2165 | *(HEK **)aux->xhv_name = existing_name; |
2166 | ((HEK **)aux->xhv_name)[1] = share_hek(name, len, hash); | |
2167 | } | |
2168 | } | |
2169 | ||
99206677 FC |
2170 | /* |
2171 | =for apidoc hv_ename_delete | |
2172 | ||
2173 | Removes a name from a stash's internal list of effective names. If this is | |
2174 | the name returned by C<HvENAME>, then another name in the list will take | |
2175 | its place (C<HvENAME> will use it). | |
2176 | ||
2177 | This is called when a stash is deleted from the symbol table. | |
2178 | ||
2179 | =cut | |
2180 | */ | |
2181 | ||
ee72b38d | 2182 | void |
78b79c77 | 2183 | Perl_hv_ename_delete(pTHX_ HV *hv, const char *name, U32 len) |
ee72b38d FC |
2184 | { |
2185 | dVAR; | |
2186 | struct xpvhv_aux *aux; | |
2187 | ||
78b79c77 | 2188 | PERL_ARGS_ASSERT_HV_ENAME_DELETE; |
ee72b38d FC |
2189 | |
2190 | if (len > I32_MAX) | |
2191 | Perl_croak(aTHX_ "panic: hv name too long (%"UVuf")", (UV) len); | |
2192 | ||
2193 | if (!SvOOK(hv)) return; | |
2194 | ||
2195 | aux = HvAUX(hv); | |
2196 | if (!aux->xhv_name) return; | |
2197 | ||
2198 | if (aux->xhv_name_count) { | |
2199 | HEK ** const namep = (HEK **)aux->xhv_name; | |
78b79c77 FC |
2200 | I32 const count = aux->xhv_name_count; |
2201 | HEK **victim = namep + (count < 0 ? -count : count); | |
2202 | while (victim-- > namep + 1) | |
ee72b38d FC |
2203 | if ( |
2204 | HEK_LEN(*victim) == (I32)len | |
2205 | && memEQ(HEK_KEY(*victim), name, len) | |
2206 | ) { | |
2207 | unshare_hek_or_pvn(*victim, 0, 0, 0); | |
78b79c77 FC |
2208 | if (count < 0) ++aux->xhv_name_count; |
2209 | else --aux->xhv_name_count; | |
2210 | if ( | |
2211 | (aux->xhv_name_count == 1 || aux->xhv_name_count == -1) | |
2212 | && !*namep | |
2213 | ) { /* if there are none left */ | |
ee72b38d FC |
2214 | Safefree(namep); |
2215 | aux->xhv_name = NULL; | |
78b79c77 | 2216 | aux->xhv_name_count = 0; |
ee72b38d FC |
2217 | } |
2218 | else { | |
2219 | /* Move the last one back to fill the empty slot. It | |
2220 | does not matter what order they are in. */ | |
78b79c77 | 2221 | *victim = *(namep + (count < 0 ? -count : count) - 1); |
ee72b38d FC |
2222 | } |
2223 | return; | |
2224 | } | |
78b79c77 FC |
2225 | if ( |
2226 | count > 0 && HEK_LEN(*namep) == (I32)len | |
2227 | && memEQ(HEK_KEY(*namep),name,len) | |
2228 | ) { | |
2229 | aux->xhv_name_count = -count; | |
2230 | } | |
ee72b38d FC |
2231 | } |
2232 | else if( | |
2233 | HEK_LEN(aux->xhv_name) == (I32)len | |
2234 | && memEQ(HEK_KEY(aux->xhv_name), name, len) | |
2235 | ) { | |
3f783763 FC |
2236 | const HEK * const namehek = aux->xhv_name; |
2237 | Newxc(aux->xhv_name, 1, HEK *, HEK); | |
2238 | *(const HEK **)aux->xhv_name = namehek; | |
2239 | aux->xhv_name_count = -1; | |
ee72b38d FC |
2240 | } |
2241 | } | |
2242 | ||
86f55936 NC |
2243 | AV ** |
2244 | Perl_hv_backreferences_p(pTHX_ HV *hv) { | |
6136c704 | 2245 | struct xpvhv_aux * const iter = SvOOK(hv) ? HvAUX(hv) : hv_auxinit(hv); |
7918f24d NC |
2246 | |
2247 | PERL_ARGS_ASSERT_HV_BACKREFERENCES_P; | |
96a5add6 | 2248 | PERL_UNUSED_CONTEXT; |
7918f24d | 2249 | |
86f55936 NC |
2250 | return &(iter->xhv_backreferences); |
2251 | } | |
2252 | ||
09aad8f0 DM |
2253 | void |
2254 | Perl_hv_kill_backrefs(pTHX_ HV *hv) { | |
2255 | AV *av; | |
2256 | ||
2257 | PERL_ARGS_ASSERT_HV_KILL_BACKREFS; | |
2258 | ||
2259 | if (!SvOOK(hv)) | |
2260 | return; | |
2261 | ||
2262 | av = HvAUX(hv)->xhv_backreferences; | |
2263 | ||
2264 | if (av) { | |
2265 | HvAUX(hv)->xhv_backreferences = 0; | |
2266 | Perl_sv_kill_backrefs(aTHX_ MUTABLE_SV(hv), av); | |
5648c0ae DM |
2267 | if (SvTYPE(av) == SVt_PVAV) |
2268 | SvREFCNT_dec(av); | |
09aad8f0 DM |
2269 | } |
2270 | } | |
2271 | ||
954c1994 | 2272 | /* |
7a7b9979 NC |
2273 | hv_iternext is implemented as a macro in hv.h |
2274 | ||
954c1994 GS |
2275 | =for apidoc hv_iternext |
2276 | ||
2277 | Returns entries from a hash iterator. See C<hv_iterinit>. | |
2278 | ||
fe7bca90 NC |
2279 | You may call C<hv_delete> or C<hv_delete_ent> on the hash entry that the |
2280 | iterator currently points to, without losing your place or invalidating your | |
2281 | iterator. Note that in this case the current entry is deleted from the hash | |
2282 | with your iterator holding the last reference to it. Your iterator is flagged | |
2283 | to free the entry on the next call to C<hv_iternext>, so you must not discard | |
2284 | your iterator immediately else the entry will leak - call C<hv_iternext> to | |
2285 | trigger the resource deallocation. | |
2286 | ||
fe7bca90 NC |
2287 | =for apidoc hv_iternext_flags |
2288 | ||
2289 | Returns entries from a hash iterator. See C<hv_iterinit> and C<hv_iternext>. | |
2290 | The C<flags> value will normally be zero; if HV_ITERNEXT_WANTPLACEHOLDERS is | |
2291 | set the placeholders keys (for restricted hashes) will be returned in addition | |
2292 | to normal keys. By default placeholders are automatically skipped over. | |
7996736c MHM |
2293 | Currently a placeholder is implemented with a value that is |
2294 | C<&Perl_sv_placeholder>. Note that the implementation of placeholders and | |
fe7bca90 NC |
2295 | restricted hashes may change, and the implementation currently is |
2296 | insufficiently abstracted for any change to be tidy. | |
e16e2ff8 | 2297 | |
fe7bca90 | 2298 | =cut |
e16e2ff8 NC |
2299 | */ |
2300 | ||
2301 | HE * | |
2302 | Perl_hv_iternext_flags(pTHX_ HV *hv, I32 flags) | |
2303 | { | |
27da23d5 | 2304 | dVAR; |
cbec9347 | 2305 | register XPVHV* xhv; |
79072805 | 2306 | register HE *entry; |
a0d0e21e | 2307 | HE *oldentry; |
463ee0b2 | 2308 | MAGIC* mg; |
bfcb3514 | 2309 | struct xpvhv_aux *iter; |
79072805 | 2310 | |
7918f24d NC |
2311 | PERL_ARGS_ASSERT_HV_ITERNEXT_FLAGS; |
2312 | ||
79072805 | 2313 | if (!hv) |
cea2e8a9 | 2314 | Perl_croak(aTHX_ "Bad hash"); |
81714fb9 | 2315 | |
cbec9347 | 2316 | xhv = (XPVHV*)SvANY(hv); |
bfcb3514 | 2317 | |
b79f7545 | 2318 | if (!SvOOK(hv)) { |
bfcb3514 NC |
2319 | /* Too many things (well, pp_each at least) merrily assume that you can |
2320 | call iv_iternext without calling hv_iterinit, so we'll have to deal | |
2321 | with it. */ | |
2322 | hv_iterinit(hv); | |
bfcb3514 | 2323 | } |
b79f7545 | 2324 | iter = HvAUX(hv); |
bfcb3514 NC |
2325 | |
2326 | oldentry = entry = iter->xhv_eiter; /* HvEITER(hv) */ | |
e62cc96a | 2327 | if (SvMAGICAL(hv) && SvRMAGICAL(hv)) { |
ad64d0ec | 2328 | if ( ( mg = mg_find((const SV *)hv, PERL_MAGIC_tied) ) ) { |
e62cc96a YO |
2329 | SV * const key = sv_newmortal(); |
2330 | if (entry) { | |
2331 | sv_setsv(key, HeSVKEY_force(entry)); | |
2332 | SvREFCNT_dec(HeSVKEY(entry)); /* get rid of previous key */ | |
2333 | } | |
2334 | else { | |
2335 | char *k; | |
2336 | HEK *hek; | |
2337 | ||
2338 | /* one HE per MAGICAL hash */ | |
2339 | iter->xhv_eiter = entry = new_HE(); /* HvEITER(hv) = new_HE() */ | |
2340 | Zero(entry, 1, HE); | |
ad64d0ec | 2341 | Newxz(k, HEK_BASESIZE + sizeof(const SV *), char); |
e62cc96a YO |
2342 | hek = (HEK*)k; |
2343 | HeKEY_hek(entry) = hek; | |
2344 | HeKLEN(entry) = HEf_SVKEY; | |
2345 | } | |
ad64d0ec | 2346 | magic_nextpack(MUTABLE_SV(hv),mg,key); |
e62cc96a YO |
2347 | if (SvOK(key)) { |
2348 | /* force key to stay around until next time */ | |
2349 | HeSVKEY_set(entry, SvREFCNT_inc_simple_NN(key)); | |
2350 | return entry; /* beware, hent_val is not set */ | |
2351 | } | |
ef8d46e8 | 2352 | SvREFCNT_dec(HeVAL(entry)); |
e62cc96a YO |
2353 | Safefree(HeKEY_hek(entry)); |
2354 | del_HE(entry); | |
2355 | iter->xhv_eiter = NULL; /* HvEITER(hv) = NULL */ | |
2356 | return NULL; | |
81714fb9 | 2357 | } |
79072805 | 2358 | } |
7ee146b1 | 2359 | #if defined(DYNAMIC_ENV_FETCH) && !defined(__riscos__) /* set up %ENV for iteration */ |
ad64d0ec NC |
2360 | if (!entry && SvRMAGICAL((const SV *)hv) |
2361 | && mg_find((const SV *)hv, PERL_MAGIC_env)) { | |
f675dbe5 | 2362 | prime_env_iter(); |
03026e68 JM |
2363 | #ifdef VMS |
2364 | /* The prime_env_iter() on VMS just loaded up new hash values | |
2365 | * so the iteration count needs to be reset back to the beginning | |
2366 | */ | |
2367 | hv_iterinit(hv); | |
2368 | iter = HvAUX(hv); | |
2369 | oldentry = entry = iter->xhv_eiter; /* HvEITER(hv) */ | |
2370 | #endif | |
2371 | } | |
f675dbe5 | 2372 | #endif |
463ee0b2 | 2373 | |
b79f7545 NC |
2374 | /* hv_iterint now ensures this. */ |
2375 | assert (HvARRAY(hv)); | |
2376 | ||
015a5f36 | 2377 | /* At start of hash, entry is NULL. */ |
fde52b5c | 2378 | if (entry) |
8aacddc1 | 2379 | { |
fde52b5c | 2380 | entry = HeNEXT(entry); |
e16e2ff8 NC |
2381 | if (!(flags & HV_ITERNEXT_WANTPLACEHOLDERS)) { |
2382 | /* | |
2383 | * Skip past any placeholders -- don't want to include them in | |
2384 | * any iteration. | |
2385 | */ | |
7996736c | 2386 | while (entry && HeVAL(entry) == &PL_sv_placeholder) { |
e16e2ff8 NC |
2387 | entry = HeNEXT(entry); |
2388 | } | |
8aacddc1 NIS |
2389 | } |
2390 | } | |
015a5f36 | 2391 | |
9eb4ebd1 NC |
2392 | /* Skip the entire loop if the hash is empty. */ |
2393 | if ((flags & HV_ITERNEXT_WANTPLACEHOLDERS) | |
2394 | ? HvTOTALKEYS(hv) : HvUSEDKEYS(hv)) { | |
900ac051 MM |
2395 | while (!entry) { |
2396 | /* OK. Come to the end of the current list. Grab the next one. */ | |
2397 | ||
2398 | iter->xhv_riter++; /* HvRITER(hv)++ */ | |
2399 | if (iter->xhv_riter > (I32)xhv->xhv_max /* HvRITER(hv) > HvMAX(hv) */) { | |
2400 | /* There is no next one. End of the hash. */ | |
2401 | iter->xhv_riter = -1; /* HvRITER(hv) = -1 */ | |
2402 | break; | |
2403 | } | |
2404 | entry = (HvARRAY(hv))[iter->xhv_riter]; | |
8aacddc1 | 2405 | |
900ac051 MM |
2406 | if (!(flags & HV_ITERNEXT_WANTPLACEHOLDERS)) { |
2407 | /* If we have an entry, but it's a placeholder, don't count it. | |
2408 | Try the next. */ | |
2409 | while (entry && HeVAL(entry) == &PL_sv_placeholder) | |
2410 | entry = HeNEXT(entry); | |
2411 | } | |
2412 | /* Will loop again if this linked list starts NULL | |
2413 | (for HV_ITERNEXT_WANTPLACEHOLDERS) | |
2414 | or if we run through it and find only placeholders. */ | |
015a5f36 | 2415 | } |
fde52b5c | 2416 | } |
79072805 | 2417 | |
72940dca | 2418 | if (oldentry && HvLAZYDEL(hv)) { /* was deleted earlier? */ |
2419 | HvLAZYDEL_off(hv); | |
68dc0745 | 2420 | hv_free_ent(hv, oldentry); |
72940dca | 2421 | } |
a0d0e21e | 2422 | |
fdcd69b6 | 2423 | /*if (HvREHASH(hv) && entry && !HeKREHASH(entry)) |
6c9570dc | 2424 | PerlIO_printf(PerlIO_stderr(), "Awooga %p %p\n", (void*)hv, (void*)entry);*/ |
fdcd69b6 | 2425 | |
bfcb3514 | 2426 | iter->xhv_eiter = entry; /* HvEITER(hv) = entry */ |
79072805 LW |
2427 | return entry; |
2428 | } | |
2429 | ||
954c1994 GS |
2430 | /* |
2431 | =for apidoc hv_iterkey | |
2432 | ||
2433 | Returns the key from the current position of the hash iterator. See | |
2434 | C<hv_iterinit>. | |
2435 | ||
2436 | =cut | |
2437 | */ | |
2438 | ||
79072805 | 2439 | char * |
864dbfa3 | 2440 | Perl_hv_iterkey(pTHX_ register HE *entry, I32 *retlen) |
79072805 | 2441 | { |
7918f24d NC |
2442 | PERL_ARGS_ASSERT_HV_ITERKEY; |
2443 | ||
fde52b5c | 2444 | if (HeKLEN(entry) == HEf_SVKEY) { |
fb73857a | 2445 | STRLEN len; |
0bd48802 | 2446 | char * const p = SvPV(HeKEY_sv(entry), len); |
fb73857a | 2447 | *retlen = len; |
2448 | return p; | |
fde52b5c | 2449 | } |
2450 | else { | |
2451 | *retlen = HeKLEN(entry); | |
2452 | return HeKEY(entry); | |
2453 | } | |
2454 | } | |
2455 | ||
2456 | /* unlike hv_iterval(), this always returns a mortal copy of the key */ | |
954c1994 GS |
2457 | /* |
2458 | =for apidoc hv_iterkeysv | |
2459 | ||
2460 | Returns the key as an C<SV*> from the current position of the hash | |
2461 | iterator. The return value will always be a mortal copy of the key. Also | |
2462 | see C<hv_iterinit>. | |
2463 | ||
2464 | =cut | |
2465 | */ | |
2466 | ||
fde52b5c | 2467 | SV * |
864dbfa3 | 2468 | Perl_hv_iterkeysv(pTHX_ register HE *entry) |
fde52b5c | 2469 | { |
7918f24d NC |
2470 | PERL_ARGS_ASSERT_HV_ITERKEYSV; |
2471 | ||
c1b02ed8 | 2472 | return sv_2mortal(newSVhek(HeKEY_hek(entry))); |
79072805 LW |
2473 | } |
2474 | ||
954c1994 GS |
2475 | /* |
2476 | =for apidoc hv_iterval | |
2477 | ||
2478 | Returns the value from the current position of the hash iterator. See | |
2479 | C<hv_iterkey>. | |
2480 | ||
2481 | =cut | |
2482 | */ | |
2483 | ||
79072805 | 2484 | SV * |
864dbfa3 | 2485 | Perl_hv_iterval(pTHX_ HV *hv, register HE *entry) |
79072805 | 2486 | { |
7918f24d NC |
2487 | PERL_ARGS_ASSERT_HV_ITERVAL; |
2488 | ||
8990e307 | 2489 | if (SvRMAGICAL(hv)) { |
ad64d0ec | 2490 | if (mg_find((const SV *)hv, PERL_MAGIC_tied)) { |
c4420975 | 2491 | SV* const sv = sv_newmortal(); |
bbce6d69 | 2492 | if (HeKLEN(entry) == HEf_SVKEY) |
ad64d0ec | 2493 | mg_copy(MUTABLE_SV(hv), sv, (char*)HeKEY_sv(entry), HEf_SVKEY); |
a3b680e6 | 2494 | else |
ad64d0ec | 2495 | mg_copy(MUTABLE_SV(hv), sv, HeKEY(entry), HeKLEN(entry)); |
463ee0b2 LW |
2496 | return sv; |
2497 | } | |
79072805 | 2498 | } |
fde52b5c | 2499 | return HeVAL(entry); |
79072805 LW |
2500 | } |
2501 | ||
954c1994 GS |
2502 | /* |
2503 | =for apidoc hv_iternextsv | |
2504 | ||
2505 | Performs an C<hv_iternext>, C<hv_iterkey>, and C<hv_iterval> in one | |
2506 | operation. | |
2507 | ||
2508 | =cut | |
2509 | */ | |
2510 | ||
a0d0e21e | 2511 | SV * |
864dbfa3 | 2512 | Perl_hv_iternextsv(pTHX_ HV *hv, char **key, I32 *retlen) |
a0d0e21e | 2513 | { |
0bd48802 AL |
2514 | HE * const he = hv_iternext_flags(hv, 0); |
2515 | ||
7918f24d NC |
2516 | PERL_ARGS_ASSERT_HV_ITERNEXTSV; |
2517 | ||
0bd48802 | 2518 | if (!he) |
a0d0e21e LW |
2519 | return NULL; |
2520 | *key = hv_iterkey(he, retlen); | |
2521 | return hv_iterval(hv, he); | |
2522 | } | |
2523 | ||
954c1994 | 2524 | /* |
bc5cdc23 NC |
2525 | |
2526 | Now a macro in hv.h | |
2527 | ||
954c1994 GS |
2528 | =for apidoc hv_magic |
2529 | ||
2530 | Adds magic to a hash. See C<sv_magic>. | |
2531 | ||
2532 | =cut | |
2533 | */ | |
2534 | ||
bbce6d69 | 2535 | /* possibly free a shared string if no one has access to it |
fde52b5c | 2536 | * len and hash must both be valid for str. |
2537 | */ | |
bbce6d69 | 2538 | void |
864dbfa3 | 2539 | Perl_unsharepvn(pTHX_ const char *str, I32 len, U32 hash) |
fde52b5c | 2540 | { |
19692e8d NC |
2541 | unshare_hek_or_pvn (NULL, str, len, hash); |
2542 | } | |
2543 | ||
2544 | ||
2545 | void | |
2546 | Perl_unshare_hek(pTHX_ HEK *hek) | |
2547 | { | |
bf11fd37 | 2548 | assert(hek); |
19692e8d NC |
2549 | unshare_hek_or_pvn(hek, NULL, 0, 0); |
2550 | } | |
2551 | ||
2552 | /* possibly free a shared string if no one has access to it | |
2553 | hek if non-NULL takes priority over the other 3, else str, len and hash | |
2554 | are used. If so, len and hash must both be valid for str. | |
2555 | */ | |
df132699 | 2556 | STATIC void |
97ddebaf | 2557 | S_unshare_hek_or_pvn(pTHX_ const HEK *hek, const char *str, I32 len, U32 hash) |
19692e8d | 2558 | { |
97aff369 | 2559 | dVAR; |
cbec9347 | 2560 | register XPVHV* xhv; |
20454177 | 2561 | HE *entry; |
fde52b5c | 2562 | register HE **oentry; |
45d1cc86 | 2563 | HE **first; |
c3654f1a | 2564 | bool is_utf8 = FALSE; |
19692e8d | 2565 | int k_flags = 0; |
aec46f14 | 2566 | const char * const save = str; |
cbbf8932 | 2567 | struct shared_he *he = NULL; |
c3654f1a | 2568 | |
19692e8d | 2569 | if (hek) { |
cbae3960 NC |
2570 | /* Find the shared he which is just before us in memory. */ |
2571 | he = (struct shared_he *)(((char *)hek) | |
2572 | - STRUCT_OFFSET(struct shared_he, | |
2573 | shared_he_hek)); | |
2574 | ||
2575 | /* Assert that the caller passed us a genuine (or at least consistent) | |
2576 | shared hek */ | |
2577 | assert (he->shared_he_he.hent_hek == hek); | |
29404ae0 | 2578 | |
de616631 NC |
2579 | if (he->shared_he_he.he_valu.hent_refcount - 1) { |
2580 | --he->shared_he_he.he_valu.hent_refcount; | |
29404ae0 NC |
2581 | return; |
2582 | } | |
29404ae0 | 2583 | |
19692e8d NC |
2584 | hash = HEK_HASH(hek); |
2585 | } else if (len < 0) { | |
2586 | STRLEN tmplen = -len; | |
2587 | is_utf8 = TRUE; | |
2588 | /* See the note in hv_fetch(). --jhi */ | |
2589 | str = (char*)bytes_from_utf8((U8*)str, &tmplen, &is_utf8); | |
2590 | len = tmplen; | |
2591 | if (is_utf8) | |
2592 | k_flags = HVhek_UTF8; | |
2593 | if (str != save) | |
2594 | k_flags |= HVhek_WASUTF8 | HVhek_FREEKEY; | |
c3654f1a | 2595 | } |
1c846c1f | 2596 | |
de616631 | 2597 | /* what follows was the moral equivalent of: |
6b88bc9c | 2598 | if ((Svp = hv_fetch(PL_strtab, tmpsv, FALSE, hash))) { |
a0714e2c | 2599 | if (--*Svp == NULL) |
6b88bc9c | 2600 | hv_delete(PL_strtab, str, len, G_DISCARD, hash); |
bbce6d69 | 2601 | } */ |
cbec9347 | 2602 | xhv = (XPVHV*)SvANY(PL_strtab); |
fde52b5c | 2603 | /* assert(xhv_array != 0) */ |
45d1cc86 | 2604 | first = oentry = &(HvARRAY(PL_strtab))[hash & (I32) HvMAX(PL_strtab)]; |
6c1b96a1 NC |
2605 | if (he) { |
2606 | const HE *const he_he = &(he->shared_he_he); | |
45d1cc86 | 2607 | for (entry = *oentry; entry; oentry = &HeNEXT(entry), entry = *oentry) { |
35ab5632 NC |
2608 | if (entry == he_he) |
2609 | break; | |
19692e8d NC |
2610 | } |
2611 | } else { | |
35a4481c | 2612 | const int flags_masked = k_flags & HVhek_MASK; |
45d1cc86 | 2613 | for (entry = *oentry; entry; oentry = &HeNEXT(entry), entry = *oentry) { |
19692e8d NC |
2614 | if (HeHASH(entry) != hash) /* strings can't be equal */ |
2615 | continue; | |
2616 | if (HeKLEN(entry) != len) | |
2617 | continue; | |
2618 | if (HeKEY(entry) != str && memNE(HeKEY(entry),str,len)) /* is this it? */ | |
2619 | continue; | |
2620 | if (HeKFLAGS(entry) != flags_masked) | |
2621 | continue; | |
19692e8d NC |
2622 | break; |
2623 | } | |
2624 | } | |
2625 | ||
35ab5632 NC |
2626 | if (entry) { |
2627 | if (--entry->he_valu.hent_refcount == 0) { | |
19692e8d | 2628 | *oentry = HeNEXT(entry); |
cbae3960 | 2629 | Safefree(entry); |
4c7185a0 | 2630 | xhv->xhv_keys--; /* HvTOTALKEYS(hv)-- */ |
19692e8d | 2631 | } |
fde52b5c | 2632 | } |
19692e8d | 2633 | |
9b387841 NC |
2634 | if (!entry) |
2635 | Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL), | |
2636 | "Attempt to free non-existent shared string '%s'%s" | |
2637 | pTHX__FORMAT, | |
2638 | hek ? HEK_KEY(hek) : str, | |
2639 | ((k_flags & HVhek_UTF8) ? " (utf8)" : "") pTHX__VALUE); | |
19692e8d NC |
2640 | if (k_flags & HVhek_FREEKEY) |
2641 | Safefree(str); | |
fde52b5c | 2642 | } |
2643 | ||
bbce6d69 | 2644 | /* get a (constant) string ptr from the global string table |
2645 | * string will get added if it is not already there. | |
fde52b5c | 2646 | * len and hash must both be valid for str. |
2647 | */ | |
bbce6d69 | 2648 | HEK * |
864dbfa3 | 2649 | Perl_share_hek(pTHX_ const char *str, I32 len, register U32 hash) |
fde52b5c | 2650 | { |
da58a35d | 2651 | bool is_utf8 = FALSE; |
19692e8d | 2652 | int flags = 0; |
aec46f14 | 2653 | const char * const save = str; |
da58a35d | 2654 | |
7918f24d NC |
2655 | PERL_ARGS_ASSERT_SHARE_HEK; |
2656 | ||
da58a35d | 2657 | if (len < 0) { |
77caf834 | 2658 | STRLEN tmplen = -len; |
da58a35d | 2659 | is_utf8 = TRUE; |
77caf834 JH |
2660 | /* See the note in hv_fetch(). --jhi */ |
2661 | str = (char*)bytes_from_utf8((U8*)str, &tmplen, &is_utf8); | |
2662 | len = tmplen; | |
19692e8d NC |
2663 | /* If we were able to downgrade here, then than means that we were passed |
2664 | in a key which only had chars 0-255, but was utf8 encoded. */ | |
2665 | if (is_utf8) | |
2666 | flags = HVhek_UTF8; | |
2667 | /* If we found we were able to downgrade the string to bytes, then | |
2668 | we should flag that it needs upgrading on keys or each. Also flag | |
2669 | that we need share_hek_flags to free the string. */ | |
2670 | if (str != save) | |
2671 | flags |= HVhek_WASUTF8 | HVhek_FREEKEY; | |
2672 | } | |
2673 | ||
6e838c70 | 2674 | return share_hek_flags (str, len, hash, flags); |
19692e8d NC |
2675 | } |
2676 | ||
6e838c70 | 2677 | STATIC HEK * |
19692e8d NC |
2678 | S_share_hek_flags(pTHX_ const char *str, I32 len, register U32 hash, int flags) |
2679 | { | |
97aff369 | 2680 | dVAR; |
19692e8d | 2681 | register HE *entry; |
35a4481c | 2682 | const int flags_masked = flags & HVhek_MASK; |
263cb4a6 | 2683 | const U32 hindex = hash & (I32) HvMAX(PL_strtab); |
7918f24d NC |
2684 | register XPVHV * const xhv = (XPVHV*)SvANY(PL_strtab); |
2685 | ||
2686 | PERL_ARGS_ASSERT_SHARE_HEK_FLAGS; | |
bbce6d69 | 2687 | |
fde52b5c | 2688 | /* what follows is the moral equivalent of: |
1c846c1f | 2689 | |
6b88bc9c | 2690 | if (!(Svp = hv_fetch(PL_strtab, str, len, FALSE))) |
a0714e2c | 2691 | hv_store(PL_strtab, str, len, NULL, hash); |
fdcd69b6 NC |
2692 | |
2693 | Can't rehash the shared string table, so not sure if it's worth | |
2694 | counting the number of entries in the linked list | |
bbce6d69 | 2695 | */ |
7918f24d | 2696 | |
fde52b5c | 2697 | /* assert(xhv_array != 0) */ |
263cb4a6 NC |
2698 | entry = (HvARRAY(PL_strtab))[hindex]; |
2699 | for (;entry; entry = HeNEXT(entry)) { | |
fde52b5c | 2700 | if (HeHASH(entry) != hash) /* strings can't be equal */ |
2701 | continue; | |
2702 | if (HeKLEN(entry) != len) | |
2703 | continue; | |
1c846c1f | 2704 | if (HeKEY(entry) != str && memNE(HeKEY(entry),str,len)) /* is this it? */ |
fde52b5c | 2705 | continue; |
19692e8d | 2706 | if (HeKFLAGS(entry) != flags_masked) |
c3654f1a | 2707 | continue; |
fde52b5c | 2708 | break; |
2709 | } | |
263cb4a6 NC |
2710 | |
2711 | if (!entry) { | |
45d1cc86 NC |
2712 | /* What used to be head of the list. |
2713 | If this is NULL, then we're the first entry for this slot, which | |
2714 | means we need to increate fill. */ | |
cbae3960 NC |
2715 | struct shared_he *new_entry; |
2716 | HEK *hek; | |
2717 | char *k; | |
263cb4a6 NC |
2718 | HE **const head = &HvARRAY(PL_strtab)[hindex]; |
2719 | HE *const next = *head; | |
cbae3960 NC |
2720 | |
2721 | /* We don't actually store a HE from the arena and a regular HEK. | |
2722 | Instead we allocate one chunk of memory big enough for both, | |
2723 | and put the HEK straight after the HE. This way we can find the | |
2724 | HEK directly from the HE. | |
2725 | */ | |
2726 | ||
a02a5408 | 2727 | Newx(k, STRUCT_OFFSET(struct shared_he, |
cbae3960 NC |
2728 | shared_he_hek.hek_key[0]) + len + 2, char); |
2729 | new_entry = (struct shared_he *)k; | |
2730 | entry = &(new_entry->shared_he_he); | |
2731 | hek = &(new_entry->shared_he_hek); | |
2732 | ||
2733 | Copy(str, HEK_KEY(hek), len, char); | |
2734 | HEK_KEY(hek)[len] = 0; | |
2735 | HEK_LEN(hek) = len; | |
2736 | HEK_HASH(hek) = hash; | |
2737 | HEK_FLAGS(hek) = (unsigned char)flags_masked; | |
2738 | ||
2739 | /* Still "point" to the HEK, so that other code need not know what | |
2740 | we're up to. */ | |
2741 | HeKEY_hek(entry) = hek; | |
de616631 | 2742 | entry->he_valu.hent_refcount = 0; |
263cb4a6 NC |
2743 | HeNEXT(entry) = next; |
2744 | *head = entry; | |
cbae3960 | 2745 | |
4c7185a0 | 2746 | xhv->xhv_keys++; /* HvTOTALKEYS(hv)++ */ |
263cb4a6 | 2747 | if (!next) { /* initial entry? */ |
5ac36297 | 2748 | } else if (xhv->xhv_keys > xhv->xhv_max /* HvKEYS(hv) > HvMAX(hv) */) { |
cbec9347 | 2749 | hsplit(PL_strtab); |
bbce6d69 | 2750 | } |
2751 | } | |
2752 | ||
de616631 | 2753 | ++entry->he_valu.hent_refcount; |
19692e8d NC |
2754 | |
2755 | if (flags & HVhek_FREEKEY) | |
f9a63242 | 2756 | Safefree(str); |
19692e8d | 2757 | |
6e838c70 | 2758 | return HeKEY_hek(entry); |
fde52b5c | 2759 | } |
ecae49c0 | 2760 | |
ca732855 NC |
2761 | I32 * |
2762 | Perl_hv_placeholders_p(pTHX_ HV *hv) | |
2763 | { | |
2764 | dVAR; | |
ad64d0ec | 2765 | MAGIC *mg = mg_find((const SV *)hv, PERL_MAGIC_rhash); |
ca732855 | 2766 | |
7918f24d NC |
2767 | PERL_ARGS_ASSERT_HV_PLACEHOLDERS_P; |
2768 | ||
ca732855 | 2769 | if (!mg) { |
ad64d0ec | 2770 | mg = sv_magicext(MUTABLE_SV(hv), 0, PERL_MAGIC_rhash, 0, 0, 0); |
ca732855 NC |
2771 | |
2772 | if (!mg) { | |
2773 | Perl_die(aTHX_ "panic: hv_placeholders_p"); | |
2774 | } | |
2775 | } | |
2776 | return &(mg->mg_len); | |
2777 | } | |
2778 | ||
2779 | ||
2780 | I32 | |
0c289d13 | 2781 | Perl_hv_placeholders_get(pTHX_ const HV *hv) |
ca732855 NC |
2782 | { |
2783 | dVAR; | |
0c289d13 | 2784 | MAGIC * const mg = mg_find((const SV *)hv, PERL_MAGIC_rhash); |
ca732855 | 2785 | |
7918f24d NC |
2786 | PERL_ARGS_ASSERT_HV_PLACEHOLDERS_GET; |
2787 | ||
ca732855 NC |
2788 | return mg ? mg->mg_len : 0; |
2789 | } | |
2790 | ||
2791 | void | |
ac1e784a | 2792 | Perl_hv_placeholders_set(pTHX_ HV *hv, I32 ph) |
ca732855 NC |
2793 | { |
2794 | dVAR; | |
ad64d0ec | 2795 | MAGIC * const mg = mg_find((const SV *)hv, PERL_MAGIC_rhash); |
ca732855 | 2796 | |
7918f24d NC |
2797 | PERL_ARGS_ASSERT_HV_PLACEHOLDERS_SET; |
2798 | ||
ca732855 NC |
2799 | if (mg) { |
2800 | mg->mg_len = ph; | |
2801 | } else if (ph) { | |
ad64d0ec | 2802 | if (!sv_magicext(MUTABLE_SV(hv), 0, PERL_MAGIC_rhash, 0, 0, ph)) |
ca732855 NC |
2803 | Perl_die(aTHX_ "panic: hv_placeholders_set"); |
2804 | } | |
2805 | /* else we don't need to add magic to record 0 placeholders. */ | |
2806 | } | |
ecae49c0 | 2807 | |
2a49f0f5 | 2808 | STATIC SV * |
7b0bddfa NC |
2809 | S_refcounted_he_value(pTHX_ const struct refcounted_he *he) |
2810 | { | |
0b2d3faa | 2811 | dVAR; |
7b0bddfa | 2812 | SV *value; |
7918f24d NC |
2813 | |
2814 | PERL_ARGS_ASSERT_REFCOUNTED_HE_VALUE; | |
2815 | ||
7b0bddfa NC |
2816 | switch(he->refcounted_he_data[0] & HVrhek_typemask) { |
2817 | case HVrhek_undef: | |
2818 | value = newSV(0); | |
2819 | break; | |
2820 | case HVrhek_delete: | |
2821 | value = &PL_sv_placeholder; | |
2822 | break; | |
2823 | case HVrhek_IV: | |
44ebaf21 NC |
2824 | value = newSViv(he->refcounted_he_val.refcounted_he_u_iv); |
2825 | break; | |
2826 | case HVrhek_UV: | |
2827 | value = newSVuv(he->refcounted_he_val.refcounted_he_u_uv); | |
7b0bddfa NC |
2828 | break; |
2829 | case HVrhek_PV: | |
44ebaf21 | 2830 | case HVrhek_PV_UTF8: |
7b0bddfa NC |
2831 | /* Create a string SV that directly points to the bytes in our |
2832 | structure. */ | |
b9f83d2f | 2833 | value = newSV_type(SVt_PV); |
7b0bddfa NC |
2834 | SvPV_set(value, (char *) he->refcounted_he_data + 1); |
2835 | SvCUR_set(value, he->refcounted_he_val.refcounted_he_u_len); | |
2836 | /* This stops anything trying to free it */ | |
2837 | SvLEN_set(value, 0); | |
2838 | SvPOK_on(value); | |
2839 | SvREADONLY_on(value); | |
44ebaf21 | 2840 | if ((he->refcounted_he_data[0] & HVrhek_typemask) == HVrhek_PV_UTF8) |
7b0bddfa NC |
2841 | SvUTF8_on(value); |
2842 | break; | |
2843 | default: | |
20439bc7 Z |
2844 | Perl_croak(aTHX_ "panic: refcounted_he_value bad flags %"UVxf, |
2845 | (UV)he->refcounted_he_data[0]); | |
7b0bddfa NC |
2846 | } |
2847 | return value; | |
2848 | } | |
2849 | ||
ecae49c0 | 2850 | /* |
20439bc7 | 2851 | =for apidoc m|HV *|refcounted_he_chain_2hv|const struct refcounted_he *c|U32 flags |
8dff4fc5 | 2852 | |
20439bc7 Z |
2853 | Generates and returns a C<HV *> representing the content of a |
2854 | C<refcounted_he> chain. | |
2855 | I<flags> is currently unused and must be zero. | |
8dff4fc5 BM |
2856 | |
2857 | =cut | |
2858 | */ | |
2859 | HV * | |
20439bc7 | 2860 | Perl_refcounted_he_chain_2hv(pTHX_ const struct refcounted_he *chain, U32 flags) |
8dff4fc5 | 2861 | { |
20439bc7 Z |
2862 | dVAR; |
2863 | HV *hv; | |
2864 | U32 placeholders, max; | |
b3ca2e83 | 2865 | |
20439bc7 Z |
2866 | if (flags) |
2867 | Perl_croak(aTHX_ "panic: refcounted_he_chain_2hv bad flags %"UVxf, | |
2868 | (UV)flags); | |
b3ca2e83 | 2869 | |
b3ca2e83 NC |
2870 | /* We could chase the chain once to get an idea of the number of keys, |
2871 | and call ksplit. But for now we'll make a potentially inefficient | |
2872 | hash with only 8 entries in its array. */ | |
20439bc7 Z |
2873 | hv = newHV(); |
2874 | max = HvMAX(hv); | |
b3ca2e83 NC |
2875 | if (!HvARRAY(hv)) { |
2876 | char *array; | |
2877 | Newxz(array, PERL_HV_ARRAY_ALLOC_BYTES(max + 1), char); | |
2878 | HvARRAY(hv) = (HE**)array; | |
2879 | } | |
2880 | ||
20439bc7 | 2881 | placeholders = 0; |
b3ca2e83 | 2882 | while (chain) { |
cbb1fbea | 2883 | #ifdef USE_ITHREADS |
b6bbf3fa | 2884 | U32 hash = chain->refcounted_he_hash; |
cbb1fbea NC |
2885 | #else |
2886 | U32 hash = HEK_HASH(chain->refcounted_he_hek); | |
2887 | #endif | |
b3ca2e83 NC |
2888 | HE **oentry = &((HvARRAY(hv))[hash & max]); |
2889 | HE *entry = *oentry; | |
b6bbf3fa | 2890 | SV *value; |
cbb1fbea | 2891 | |
b3ca2e83 NC |
2892 | for (; entry; entry = HeNEXT(entry)) { |
2893 | if (HeHASH(entry) == hash) { | |
9f769845 NC |
2894 | /* We might have a duplicate key here. If so, entry is older |
2895 | than the key we've already put in the hash, so if they are | |
2896 | the same, skip adding entry. */ | |
2897 | #ifdef USE_ITHREADS | |
2898 | const STRLEN klen = HeKLEN(entry); | |
2899 | const char *const key = HeKEY(entry); | |
2900 | if (klen == chain->refcounted_he_keylen | |
2901 | && (!!HeKUTF8(entry) | |
2902 | == !!(chain->refcounted_he_data[0] & HVhek_UTF8)) | |
2903 | && memEQ(key, REF_HE_KEY(chain), klen)) | |
2904 | goto next_please; | |
2905 | #else | |
2906 | if (HeKEY_hek(entry) == chain->refcounted_he_hek) | |
2907 | goto next_please; | |
2908 | if (HeKLEN(entry) == HEK_LEN(chain->refcounted_he_hek) | |
2909 | && HeKUTF8(entry) == HEK_UTF8(chain->refcounted_he_hek) | |
2910 | && memEQ(HeKEY(entry), HEK_KEY(chain->refcounted_he_hek), | |
2911 | HeKLEN(entry))) | |
2912 | goto next_please; | |
2913 | #endif | |
b3ca2e83 NC |
2914 | } |
2915 | } | |
2916 | assert (!entry); | |
2917 | entry = new_HE(); | |
2918 | ||
cbb1fbea NC |
2919 | #ifdef USE_ITHREADS |
2920 | HeKEY_hek(entry) | |
7b0bddfa | 2921 | = share_hek_flags(REF_HE_KEY(chain), |
b6bbf3fa NC |
2922 | chain->refcounted_he_keylen, |
2923 | chain->refcounted_he_hash, | |
2924 | (chain->refcounted_he_data[0] | |
2925 | & (HVhek_UTF8|HVhek_WASUTF8))); | |
cbb1fbea | 2926 | #else |
71ad1b0c | 2927 | HeKEY_hek(entry) = share_hek_hek(chain->refcounted_he_hek); |
cbb1fbea | 2928 | #endif |
7b0bddfa NC |
2929 | value = refcounted_he_value(chain); |
2930 | if (value == &PL_sv_placeholder) | |
b3ca2e83 | 2931 | placeholders++; |
b6bbf3fa | 2932 | HeVAL(entry) = value; |
b3ca2e83 NC |
2933 | |
2934 | /* Link it into the chain. */ | |
2935 | HeNEXT(entry) = *oentry; | |
b3ca2e83 NC |
2936 | *oentry = entry; |
2937 | ||
2938 | HvTOTALKEYS(hv)++; | |
2939 | ||
2940 | next_please: | |
71ad1b0c | 2941 | chain = chain->refcounted_he_next; |
b3ca2e83 NC |
2942 | } |
2943 | ||
2944 | if (placeholders) { | |
2945 | clear_placeholders(hv, placeholders); | |
2946 | HvTOTALKEYS(hv) -= placeholders; | |
2947 | } | |
2948 | ||
2949 | /* We could check in the loop to see if we encounter any keys with key | |
2950 | flags, but it's probably not worth it, as this per-hash flag is only | |
2951 | really meant as an optimisation for things like Storable. */ | |
2952 | HvHASKFLAGS_on(hv); | |
def9038f | 2953 | DEBUG_A(Perl_hv_assert(aTHX_ hv)); |
b3ca2e83 NC |
2954 | |
2955 | return hv; | |
2956 | } | |
2957 | ||
20439bc7 Z |
2958 | /* |
2959 | =for apidoc m|SV *|refcounted_he_fetch_pvn|const struct refcounted_he *chain|const char *keypv|STRLEN keylen|U32 hash|U32 flags | |
2960 | ||
2961 | Search along a C<refcounted_he> chain for an entry with the key specified | |
2962 | by I<keypv> and I<keylen>. If I<flags> has the C<REFCOUNTED_HE_KEY_UTF8> | |
2963 | bit set, the key octets are interpreted as UTF-8, otherwise they | |
2964 | are interpreted as Latin-1. I<hash> is a precomputed hash of the key | |
2965 | string, or zero if it has not been precomputed. Returns a mortal scalar | |
2966 | representing the value associated with the key, or C<&PL_sv_placeholder> | |
2967 | if there is no value associated with the key. | |
2968 | ||
2969 | =cut | |
2970 | */ | |
2971 | ||
7b0bddfa | 2972 | SV * |
20439bc7 Z |
2973 | Perl_refcounted_he_fetch_pvn(pTHX_ const struct refcounted_he *chain, |
2974 | const char *keypv, STRLEN keylen, U32 hash, U32 flags) | |
7b0bddfa | 2975 | { |
0b2d3faa | 2976 | dVAR; |
20439bc7 Z |
2977 | U8 utf8_flag; |
2978 | PERL_ARGS_ASSERT_REFCOUNTED_HE_FETCH_PVN; | |
7b0bddfa | 2979 | |
20439bc7 Z |
2980 | if (flags & ~REFCOUNTED_HE_KEY_UTF8) |
2981 | Perl_croak(aTHX_ "panic: refcounted_he_fetch_pvn bad flags %"UVxf, | |
2982 | (UV)flags); | |
2983 | if (!chain) | |
2984 | return &PL_sv_placeholder; | |
2985 | if (flags & REFCOUNTED_HE_KEY_UTF8) { | |
2986 | /* For searching purposes, canonicalise to Latin-1 where possible. */ | |
2987 | const char *keyend = keypv + keylen, *p; | |
2988 | STRLEN nonascii_count = 0; | |
2989 | for (p = keypv; p != keyend; p++) { | |
2990 | U8 c = (U8)*p; | |
2991 | if (c & 0x80) { | |
2992 | if (!((c & 0xfe) == 0xc2 && ++p != keyend && | |
2993 | (((U8)*p) & 0xc0) == 0x80)) | |
2994 | goto canonicalised_key; | |
2995 | nonascii_count++; | |
2996 | } | |
cd1d2f8a | 2997 | } |
20439bc7 Z |
2998 | if (nonascii_count) { |
2999 | char *q; | |
3000 | const char *p = keypv, *keyend = keypv + keylen; | |
3001 | keylen -= nonascii_count; | |
3002 | Newx(q, keylen, char); | |
3003 | SAVEFREEPV(q); | |
3004 | keypv = q; | |
3005 | for (; p != keyend; p++, q++) { | |
3006 | U8 c = (U8)*p; | |
3007 | *q = (char) | |
3008 | ((c & 0x80) ? ((c & 0x03) << 6) | (((U8)*++p) & 0x3f) : c); | |
cd1d2f8a NC |
3009 | } |
3010 | } | |
20439bc7 Z |
3011 | flags &= ~REFCOUNTED_HE_KEY_UTF8; |
3012 | canonicalised_key: ; | |
3013 | } | |
3014 | utf8_flag = (flags & REFCOUNTED_HE_KEY_UTF8) ? HVhek_UTF8 : 0; | |
3015 | if (!hash) | |
3016 | PERL_HASH(hash, keypv, keylen); | |
7b0bddfa | 3017 | |
20439bc7 Z |
3018 | for (; chain; chain = chain->refcounted_he_next) { |
3019 | if ( | |
7b0bddfa | 3020 | #ifdef USE_ITHREADS |
20439bc7 Z |
3021 | hash == chain->refcounted_he_hash && |
3022 | keylen == chain->refcounted_he_keylen && | |
3023 | memEQ(REF_HE_KEY(chain), keypv, keylen) && | |
3024 | utf8_flag == (chain->refcounted_he_data[0] & HVhek_UTF8) | |
7b0bddfa | 3025 | #else |
20439bc7 Z |
3026 | hash == HEK_HASH(chain->refcounted_he_hek) && |
3027 | keylen == (STRLEN)HEK_LEN(chain->refcounted_he_hek) && | |
3028 | memEQ(HEK_KEY(chain->refcounted_he_hek), keypv, keylen) && | |
3029 | utf8_flag == (HEK_FLAGS(chain->refcounted_he_hek) & HVhek_UTF8) | |
7b0bddfa | 3030 | #endif |
20439bc7 Z |
3031 | ) |
3032 | return sv_2mortal(refcounted_he_value(chain)); | |
7b0bddfa | 3033 | } |
20439bc7 Z |
3034 | return &PL_sv_placeholder; |
3035 | } | |
7b0bddfa | 3036 | |
20439bc7 Z |
3037 | /* |
3038 | =for apidoc m|SV *|refcounted_he_fetch_pv|const struct refcounted_he *chain|const char *key|U32 hash|U32 flags | |
7b0bddfa | 3039 | |
20439bc7 Z |
3040 | Like L</refcounted_he_fetch_pvn>, but takes a nul-terminated string |
3041 | instead of a string/length pair. | |
3042 | ||
3043 | =cut | |
3044 | */ | |
3045 | ||
3046 | SV * | |
3047 | Perl_refcounted_he_fetch_pv(pTHX_ const struct refcounted_he *chain, | |
3048 | const char *key, U32 hash, U32 flags) | |
3049 | { | |
3050 | PERL_ARGS_ASSERT_REFCOUNTED_HE_FETCH_PV; | |
3051 | return refcounted_he_fetch_pvn(chain, key, strlen(key), hash, flags); | |
7b0bddfa NC |
3052 | } |
3053 | ||
b3ca2e83 | 3054 | /* |
20439bc7 Z |
3055 | =for apidoc m|SV *|refcounted_he_fetch_sv|const struct refcounted_he *chain|SV *key|U32 hash|U32 flags |
3056 | ||
3057 | Like L</refcounted_he_fetch_pvn>, but takes a Perl scalar instead of a | |
3058 | string/length pair. | |
3059 | ||
3060 | =cut | |
3061 | */ | |
b3ca2e83 | 3062 | |
20439bc7 Z |
3063 | SV * |
3064 | Perl_refcounted_he_fetch_sv(pTHX_ const struct refcounted_he *chain, | |
3065 | SV *key, U32 hash, U32 flags) | |
3066 | { | |
3067 | const char *keypv; | |
3068 | STRLEN keylen; | |
3069 | PERL_ARGS_ASSERT_REFCOUNTED_HE_FETCH_SV; | |
3070 | if (flags & REFCOUNTED_HE_KEY_UTF8) | |
3071 | Perl_croak(aTHX_ "panic: refcounted_he_fetch_sv bad flags %"UVxf, | |
3072 | (UV)flags); | |
3073 | keypv = SvPV_const(key, keylen); | |
3074 | if (SvUTF8(key)) | |
3075 | flags |= REFCOUNTED_HE_KEY_UTF8; | |
3076 | if (!hash && SvIsCOW_shared_hash(key)) | |
3077 | hash = SvSHARED_HASH(key); | |
3078 | return refcounted_he_fetch_pvn(chain, keypv, keylen, hash, flags); | |
3079 | } | |
3080 | ||
3081 | /* | |
3082 | =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 | |
3083 | ||
3084 | Creates a new C<refcounted_he>. This consists of a single key/value | |
3085 | pair and a reference to an existing C<refcounted_he> chain (which may | |
3086 | be empty), and thus forms a longer chain. When using the longer chain, | |
3087 | the new key/value pair takes precedence over any entry for the same key | |
3088 | further along the chain. | |
3089 | ||
3090 | The new key is specified by I<keypv> and I<keylen>. If I<flags> has | |
3091 | the C<REFCOUNTED_HE_KEY_UTF8> bit set, the key octets are interpreted | |
3092 | as UTF-8, otherwise they are interpreted as Latin-1. I<hash> is | |
3093 | a precomputed hash of the key string, or zero if it has not been | |
3094 | precomputed. | |
3095 | ||
3096 | I<value> is the scalar value to store for this key. I<value> is copied | |
3097 | by this function, which thus does not take ownership of any reference | |
3098 | to it, and later changes to the scalar will not be reflected in the | |
3099 | value visible in the C<refcounted_he>. Complex types of scalar will not | |
3100 | be stored with referential integrity, but will be coerced to strings. | |
3101 | I<value> may be either null or C<&PL_sv_placeholder> to indicate that no | |
3102 | value is to be associated with the key; this, as with any non-null value, | |
3103 | takes precedence over the existence of a value for the key further along | |
3104 | the chain. | |
3105 | ||
3106 | I<parent> points to the rest of the C<refcounted_he> chain to be | |
3107 | attached to the new C<refcounted_he>. This function takes ownership | |
3108 | of one reference to I<parent>, and returns one reference to the new | |
3109 | C<refcounted_he>. | |
b3ca2e83 NC |
3110 | |
3111 | =cut | |
3112 | */ | |
3113 | ||
3114 | struct refcounted_he * | |
20439bc7 Z |
3115 | Perl_refcounted_he_new_pvn(pTHX_ struct refcounted_he *parent, |
3116 | const char *keypv, STRLEN keylen, U32 hash, SV *value, U32 flags) | |
3117 | { | |
7a89be66 | 3118 | dVAR; |
b6bbf3fa | 3119 | STRLEN value_len = 0; |
95b63a38 | 3120 | const char *value_p = NULL; |
20439bc7 | 3121 | bool is_pv; |
b6bbf3fa | 3122 | char value_type; |
20439bc7 Z |
3123 | char hekflags; |
3124 | STRLEN key_offset = 1; | |
3125 | struct refcounted_he *he; | |
3126 | PERL_ARGS_ASSERT_REFCOUNTED_HE_NEW_PVN; | |
b6bbf3fa | 3127 | |
20439bc7 Z |
3128 | if (!value || value == &PL_sv_placeholder) { |
3129 | value_type = HVrhek_delete; | |
3130 | } else if (SvPOK(value)) { | |
b6bbf3fa NC |
3131 | value_type = HVrhek_PV; |
3132 | } else if (SvIOK(value)) { | |
ad64d0ec | 3133 | value_type = SvUOK((const SV *)value) ? HVrhek_UV : HVrhek_IV; |
b6bbf3fa NC |
3134 | } else if (!SvOK(value)) { |
3135 | value_type = HVrhek_undef; | |
3136 | } else { | |
3137 | value_type = HVrhek_PV; | |
3138 | } | |
20439bc7 Z |
3139 | is_pv = value_type == HVrhek_PV; |
3140 | if (is_pv) { | |
012da8e5 NC |
3141 | /* Do it this way so that the SvUTF8() test is after the SvPV, in case |
3142 | the value is overloaded, and doesn't yet have the UTF-8flag set. */ | |
b6bbf3fa | 3143 | value_p = SvPV_const(value, value_len); |
012da8e5 NC |
3144 | if (SvUTF8(value)) |
3145 | value_type = HVrhek_PV_UTF8; | |
20439bc7 Z |
3146 | key_offset = value_len + 2; |
3147 | } | |
3148 | hekflags = value_type; | |
3149 | ||
3150 | if (flags & REFCOUNTED_HE_KEY_UTF8) { | |
3151 | /* Canonicalise to Latin-1 where possible. */ | |
3152 | const char *keyend = keypv + keylen, *p; | |
3153 | STRLEN nonascii_count = 0; | |
3154 | for (p = keypv; p != keyend; p++) { | |
3155 | U8 c = (U8)*p; | |
3156 | if (c & 0x80) { | |
3157 | if (!((c & 0xfe) == 0xc2 && ++p != keyend && | |
3158 | (((U8)*p) & 0xc0) == 0x80)) | |
3159 | goto canonicalised_key; | |
3160 | nonascii_count++; | |
3161 | } | |
3162 | } | |
3163 | if (nonascii_count) { | |
3164 | char *q; | |
3165 | const char *p = keypv, *keyend = keypv + keylen; | |
3166 | keylen -= nonascii_count; | |
3167 | Newx(q, keylen, char); | |
3168 | SAVEFREEPV(q); | |
3169 | keypv = q; | |
3170 | for (; p != keyend; p++, q++) { | |
3171 | U8 c = (U8)*p; | |
3172 | *q = (char) | |
3173 | ((c & 0x80) ? ((c & 0x03) << 6) | (((U8)*++p) & 0x3f) : c); | |
3174 | } | |
3175 | } | |
3176 | flags &= ~REFCOUNTED_HE_KEY_UTF8; | |
3177 | canonicalised_key: ; | |
b6bbf3fa | 3178 | } |
20439bc7 Z |
3179 | if (flags & REFCOUNTED_HE_KEY_UTF8) |
3180 | hekflags |= HVhek_UTF8; | |
3181 | if (!hash) | |
3182 | PERL_HASH(hash, keypv, keylen); | |
012da8e5 | 3183 | |
0de694c5 | 3184 | #ifdef USE_ITHREADS |
10edeb5d JH |
3185 | he = (struct refcounted_he*) |
3186 | PerlMemShared_malloc(sizeof(struct refcounted_he) - 1 | |
20439bc7 | 3187 | + keylen |
20439bc7 | 3188 | + key_offset); |
0de694c5 NC |
3189 | #else |
3190 | he = (struct refcounted_he*) | |
3191 | PerlMemShared_malloc(sizeof(struct refcounted_he) - 1 | |
3192 | + key_offset); | |
3193 | #endif | |
b3ca2e83 | 3194 | |
71ad1b0c | 3195 | he->refcounted_he_next = parent; |
b6bbf3fa | 3196 | |
012da8e5 | 3197 | if (is_pv) { |
20439bc7 | 3198 | Copy(value_p, he->refcounted_he_data + 1, value_len + 1, char); |
b6bbf3fa | 3199 | he->refcounted_he_val.refcounted_he_u_len = value_len; |
b6bbf3fa | 3200 | } else if (value_type == HVrhek_IV) { |
20439bc7 | 3201 | he->refcounted_he_val.refcounted_he_u_iv = SvIVX(value); |
012da8e5 | 3202 | } else if (value_type == HVrhek_UV) { |
20439bc7 | 3203 | he->refcounted_he_val.refcounted_he_u_uv = SvUVX(value); |
b6bbf3fa NC |
3204 | } |
3205 | ||
cbb1fbea | 3206 | #ifdef USE_ITHREADS |
b6bbf3fa | 3207 | he->refcounted_he_hash = hash; |
20439bc7 Z |
3208 | he->refcounted_he_keylen = keylen; |
3209 | Copy(keypv, he->refcounted_he_data + key_offset, keylen, char); | |
cbb1fbea | 3210 | #else |
20439bc7 | 3211 | he->refcounted_he_hek = share_hek_flags(keypv, keylen, hash, hekflags); |
cbb1fbea | 3212 | #endif |
b6bbf3fa | 3213 | |
20439bc7 | 3214 | he->refcounted_he_data[0] = hekflags; |
b3ca2e83 NC |
3215 | he->refcounted_he_refcnt = 1; |
3216 | ||
3217 | return he; | |
3218 | } | |
3219 | ||
3220 | /* | |
20439bc7 | 3221 | =for apidoc m|struct refcounted_he *|refcounted_he_new_pv|struct refcounted_he *parent|const char *key|U32 hash|SV *value|U32 flags |
b3ca2e83 | 3222 | |
20439bc7 Z |
3223 | Like L</refcounted_he_new_pvn>, but takes a nul-terminated string instead |
3224 | of a string/length pair. | |
3225 | ||
3226 | =cut | |
3227 | */ | |
3228 | ||
3229 | struct refcounted_he * | |
3230 | Perl_refcounted_he_new_pv(pTHX_ struct refcounted_he *parent, | |
3231 | const char *key, U32 hash, SV *value, U32 flags) | |
3232 | { | |
3233 | PERL_ARGS_ASSERT_REFCOUNTED_HE_NEW_PV; | |
3234 | return refcounted_he_new_pvn(parent, key, strlen(key), hash, value, flags); | |
3235 | } | |
3236 | ||
3237 | /* | |
3238 | =for apidoc m|struct refcounted_he *|refcounted_he_new_sv|struct refcounted_he *parent|SV *key|U32 hash|SV *value|U32 flags | |
3239 | ||
3240 | Like L</refcounted_he_new_pvn>, but takes a Perl scalar instead of a | |
3241 | string/length pair. | |
3242 | ||
3243 | =cut | |
3244 | */ | |
3245 | ||
3246 | struct refcounted_he * | |
3247 | Perl_refcounted_he_new_sv(pTHX_ struct refcounted_he *parent, | |
3248 | SV *key, U32 hash, SV *value, U32 flags) | |
3249 | { | |
3250 | const char *keypv; | |
3251 | STRLEN keylen; | |
3252 | PERL_ARGS_ASSERT_REFCOUNTED_HE_NEW_SV; | |
3253 | if (flags & REFCOUNTED_HE_KEY_UTF8) | |
3254 | Perl_croak(aTHX_ "panic: refcounted_he_new_sv bad flags %"UVxf, | |
3255 | (UV)flags); | |
3256 | keypv = SvPV_const(key, keylen); | |
3257 | if (SvUTF8(key)) | |
3258 | flags |= REFCOUNTED_HE_KEY_UTF8; | |
3259 | if (!hash && SvIsCOW_shared_hash(key)) | |
3260 | hash = SvSHARED_HASH(key); | |
3261 | return refcounted_he_new_pvn(parent, keypv, keylen, hash, value, flags); | |
3262 | } | |
3263 | ||
3264 | /* | |
3265 | =for apidoc m|void|refcounted_he_free|struct refcounted_he *he | |
3266 | ||
3267 | Decrements the reference count of a C<refcounted_he> by one. If the | |
3268 | reference count reaches zero the structure's memory is freed, which | |
3269 | (recursively) causes a reduction of its parent C<refcounted_he>'s | |
3270 | reference count. It is safe to pass a null pointer to this function: | |
3271 | no action occurs in this case. | |
b3ca2e83 NC |
3272 | |
3273 | =cut | |
3274 | */ | |
3275 | ||
3276 | void | |
3277 | Perl_refcounted_he_free(pTHX_ struct refcounted_he *he) { | |
53d44271 | 3278 | dVAR; |
57ca3b03 AL |
3279 | PERL_UNUSED_CONTEXT; |
3280 | ||
b3ca2e83 NC |
3281 | while (he) { |
3282 | struct refcounted_he *copy; | |
cbb1fbea | 3283 | U32 new_count; |
b3ca2e83 | 3284 | |
cbb1fbea NC |
3285 | HINTS_REFCNT_LOCK; |
3286 | new_count = --he->refcounted_he_refcnt; | |
3287 | HINTS_REFCNT_UNLOCK; | |
3288 | ||
3289 | if (new_count) { | |
b3ca2e83 | 3290 | return; |
cbb1fbea | 3291 | } |
b3ca2e83 | 3292 | |
b6bbf3fa | 3293 | #ifndef USE_ITHREADS |
71ad1b0c | 3294 | unshare_hek_or_pvn (he->refcounted_he_hek, 0, 0, 0); |
cbb1fbea | 3295 | #endif |
b3ca2e83 | 3296 | copy = he; |
71ad1b0c | 3297 | he = he->refcounted_he_next; |
b6bbf3fa | 3298 | PerlMemShared_free(copy); |
b3ca2e83 NC |
3299 | } |
3300 | } | |
3301 | ||
20439bc7 Z |
3302 | /* |
3303 | =for apidoc m|struct refcounted_he *|refcounted_he_inc|struct refcounted_he *he | |
3304 | ||
3305 | Increment the reference count of a C<refcounted_he>. The pointer to the | |
3306 | C<refcounted_he> is also returned. It is safe to pass a null pointer | |
3307 | to this function: no action occurs and a null pointer is returned. | |
3308 | ||
3309 | =cut | |
3310 | */ | |
3311 | ||
3312 | struct refcounted_he * | |
3313 | Perl_refcounted_he_inc(pTHX_ struct refcounted_he *he) | |
3314 | { | |
3315 | if (he) { | |
3316 | HINTS_REFCNT_LOCK; | |
3317 | he->refcounted_he_refcnt++; | |
3318 | HINTS_REFCNT_UNLOCK; | |
3319 | } | |
3320 | return he; | |
3321 | } | |
3322 | ||
47550813 NC |
3323 | /* pp_entereval is aware that labels are stored with a key ':' at the top of |
3324 | the linked list. */ | |
dca6062a | 3325 | const char * |
ce42d03d | 3326 | Perl_fetch_cop_label(pTHX_ COP *const cop, STRLEN *len, U32 *flags) { |
d6747b7a NC |
3327 | struct refcounted_he *const chain = cop->cop_hints_hash; |
3328 | ||
3329 | PERL_ARGS_ASSERT_FETCH_COP_LABEL; | |
3330 | ||
dca6062a NC |
3331 | if (!chain) |
3332 | return NULL; | |
3333 | #ifdef USE_ITHREADS | |
3334 | if (chain->refcounted_he_keylen != 1) | |
3335 | return NULL; | |
3336 | if (*REF_HE_KEY(chain) != ':') | |
3337 | return NULL; | |
3338 | #else | |
3339 | if ((STRLEN)HEK_LEN(chain->refcounted_he_hek) != 1) | |
3340 | return NULL; | |
3341 | if (*HEK_KEY(chain->refcounted_he_hek) != ':') | |
3342 | return NULL; | |
3343 | #endif | |
012da8e5 NC |
3344 | /* Stop anyone trying to really mess us up by adding their own value for |
3345 | ':' into %^H */ | |
3346 | if ((chain->refcounted_he_data[0] & HVrhek_typemask) != HVrhek_PV | |
3347 | && (chain->refcounted_he_data[0] & HVrhek_typemask) != HVrhek_PV_UTF8) | |
3348 | return NULL; | |
3349 | ||
dca6062a NC |
3350 | if (len) |
3351 | *len = chain->refcounted_he_val.refcounted_he_u_len; | |
3352 | if (flags) { | |
3353 | *flags = ((chain->refcounted_he_data[0] & HVrhek_typemask) | |
3354 | == HVrhek_PV_UTF8) ? SVf_UTF8 : 0; | |
3355 | } | |
3356 | return chain->refcounted_he_data + 1; | |
3357 | } | |
3358 | ||
a77ac40c NC |
3359 | void |
3360 | Perl_store_cop_label(pTHX_ COP *const cop, const char *label, STRLEN len, | |
3361 | U32 flags) | |
012da8e5 | 3362 | { |
20439bc7 | 3363 | SV *labelsv; |
547bb267 NC |
3364 | PERL_ARGS_ASSERT_STORE_COP_LABEL; |
3365 | ||
a77ac40c NC |
3366 | if (flags & ~(SVf_UTF8)) |
3367 | Perl_croak(aTHX_ "panic: store_cop_label illegal flag bits 0x%" UVxf, | |
3368 | (UV)flags); | |
20439bc7 Z |
3369 | labelsv = sv_2mortal(newSVpvn(label, len)); |
3370 | if (flags & SVf_UTF8) | |
3371 | SvUTF8_on(labelsv); | |
a77ac40c | 3372 | cop->cop_hints_hash |
20439bc7 | 3373 | = refcounted_he_new_pvs(cop->cop_hints_hash, ":", labelsv, 0); |
012da8e5 NC |
3374 | } |
3375 | ||
b3ca2e83 | 3376 | /* |
ecae49c0 NC |
3377 | =for apidoc hv_assert |
3378 | ||
3379 | Check that a hash is in an internally consistent state. | |
3380 | ||
3381 | =cut | |
3382 | */ | |
3383 | ||
943795c2 NC |
3384 | #ifdef DEBUGGING |
3385 | ||
ecae49c0 NC |
3386 | void |
3387 | Perl_hv_assert(pTHX_ HV *hv) | |
3388 | { | |
57ca3b03 AL |
3389 | dVAR; |
3390 | HE* entry; | |
3391 | int withflags = 0; | |
3392 | int placeholders = 0; | |
3393 | int real = 0; | |
3394 | int bad = 0; | |
3395 | const I32 riter = HvRITER_get(hv); | |
3396 | HE *eiter = HvEITER_get(hv); | |
3397 | ||
7918f24d NC |
3398 | PERL_ARGS_ASSERT_HV_ASSERT; |
3399 | ||
57ca3b03 AL |
3400 | (void)hv_iterinit(hv); |
3401 | ||
3402 | while ((entry = hv_iternext_flags(hv, HV_ITERNEXT_WANTPLACEHOLDERS))) { | |
3403 | /* sanity check the values */ | |
3404 | if (HeVAL(entry) == &PL_sv_placeholder) | |
3405 | placeholders++; | |
3406 | else | |
3407 | real++; | |
3408 | /* sanity check the keys */ | |
3409 | if (HeSVKEY(entry)) { | |
6f207bd3 | 3410 | NOOP; /* Don't know what to check on SV keys. */ |
57ca3b03 AL |
3411 | } else if (HeKUTF8(entry)) { |
3412 | withflags++; | |
3413 | if (HeKWASUTF8(entry)) { | |
3414 | PerlIO_printf(Perl_debug_log, | |
d2a455e7 | 3415 | "hash key has both WASUTF8 and UTF8: '%.*s'\n", |
57ca3b03 AL |
3416 | (int) HeKLEN(entry), HeKEY(entry)); |
3417 | bad = 1; | |
3418 | } | |
3419 | } else if (HeKWASUTF8(entry)) | |
3420 | withflags++; | |
3421 | } | |
ad64d0ec | 3422 | if (!SvTIED_mg((const SV *)hv, PERL_MAGIC_tied)) { |
57ca3b03 AL |
3423 | static const char bad_count[] = "Count %d %s(s), but hash reports %d\n"; |
3424 | const int nhashkeys = HvUSEDKEYS(hv); | |
3425 | const int nhashplaceholders = HvPLACEHOLDERS_get(hv); | |
3426 | ||
3427 | if (nhashkeys != real) { | |
3428 | PerlIO_printf(Perl_debug_log, bad_count, real, "keys", nhashkeys ); | |
3429 | bad = 1; | |
3430 | } | |
3431 | if (nhashplaceholders != placeholders) { | |
3432 | PerlIO_printf(Perl_debug_log, bad_count, placeholders, "placeholder", nhashplaceholders ); | |
3433 | bad = 1; | |
3434 | } | |
3435 | } | |
3436 | if (withflags && ! HvHASKFLAGS(hv)) { | |
3437 | PerlIO_printf(Perl_debug_log, | |
3438 | "Hash has HASKFLAGS off but I count %d key(s) with flags\n", | |
3439 | withflags); | |
3440 | bad = 1; | |
3441 | } | |
3442 | if (bad) { | |
ad64d0ec | 3443 | sv_dump(MUTABLE_SV(hv)); |
57ca3b03 AL |
3444 | } |
3445 | HvRITER_set(hv, riter); /* Restore hash iterator state */ | |
3446 | HvEITER_set(hv, eiter); | |
ecae49c0 | 3447 | } |
af3babe4 | 3448 | |
943795c2 NC |
3449 | #endif |
3450 | ||
af3babe4 NC |
3451 | /* |
3452 | * Local variables: | |
3453 | * c-indentation-style: bsd | |
3454 | * c-basic-offset: 4 | |
3455 | * indent-tabs-mode: t | |
3456 | * End: | |
3457 | * | |
37442d52 RGS |
3458 | * ex: set ts=8 sts=4 sw=4 noet: |
3459 | */ |