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 | ||
a05d6c5d TC |
220 | Stores an SV in a hash. The hash key is specified as C<key> and the |
221 | absolute value of C<klen> is the length of the key. If C<klen> is | |
222 | negative the key is assumed to be in UTF-8-encoded Unicode. The | |
223 | C<hash> parameter is the precomputed hash value; if it is zero then | |
224 | Perl will compute it. | |
225 | ||
226 | The return value will be | |
34a6f7b4 NC |
227 | NULL if the operation failed or if the value did not need to be actually |
228 | stored within the hash (as in the case of tied hashes). Otherwise it can | |
229 | be dereferenced to get the original C<SV*>. Note that the caller is | |
230 | responsible for suitably incrementing the reference count of C<val> before | |
231 | the call, and decrementing it if the function returned NULL. Effectively | |
232 | a successful hv_store takes ownership of one reference to C<val>. This is | |
233 | usually what you want; a newly created SV has a reference count of one, so | |
234 | if all your code does is create SVs then store them in a hash, hv_store | |
235 | will own the only reference to the new SV, and your code doesn't need to do | |
236 | anything further to tidy up. hv_store is not implemented as a call to | |
237 | hv_store_ent, and does not create a temporary SV for the key, so if your | |
238 | key data is not already in SV form then use hv_store in preference to | |
239 | hv_store_ent. | |
240 | ||
241 | See L<perlguts/"Understanding the Magic of Tied Hashes and Arrays"> for more | |
242 | information on how to use this function on tied hashes. | |
243 | ||
34a6f7b4 NC |
244 | =for apidoc hv_store_ent |
245 | ||
246 | Stores C<val> in a hash. The hash key is specified as C<key>. The C<hash> | |
247 | parameter is the precomputed hash value; if it is zero then Perl will | |
248 | compute it. The return value is the new hash entry so created. It will be | |
249 | NULL if the operation failed or if the value did not need to be actually | |
250 | stored within the hash (as in the case of tied hashes). Otherwise the | |
251 | contents of the return value can be accessed using the C<He?> macros | |
252 | described here. Note that the caller is responsible for suitably | |
253 | incrementing the reference count of C<val> before the call, and | |
254 | decrementing it if the function returned NULL. Effectively a successful | |
255 | hv_store_ent takes ownership of one reference to C<val>. This is | |
256 | usually what you want; a newly created SV has a reference count of one, so | |
257 | if all your code does is create SVs then store them in a hash, hv_store | |
258 | will own the only reference to the new SV, and your code doesn't need to do | |
259 | anything further to tidy up. Note that hv_store_ent only reads the C<key>; | |
260 | unlike C<val> it does not take ownership of it, so maintaining the correct | |
261 | reference count on C<key> is entirely the caller's responsibility. hv_store | |
262 | is not implemented as a call to hv_store_ent, and does not create a temporary | |
263 | SV for the key, so if your key data is not already in SV form then use | |
264 | hv_store in preference to hv_store_ent. | |
265 | ||
266 | See L<perlguts/"Understanding the Magic of Tied Hashes and Arrays"> for more | |
267 | information on how to use this function on tied hashes. | |
268 | ||
34a6f7b4 NC |
269 | =for apidoc hv_exists |
270 | ||
271 | Returns a boolean indicating whether the specified hash key exists. The | |
a05d6c5d TC |
272 | absolute value of C<klen> is the length of the key. If C<klen> is |
273 | negative the key is assumed to be in UTF-8-encoded Unicode. | |
34a6f7b4 | 274 | |
954c1994 GS |
275 | =for apidoc hv_fetch |
276 | ||
a05d6c5d TC |
277 | Returns the SV which corresponds to the specified key in the hash. |
278 | The absolute value of C<klen> is the length of the key. If C<klen> is | |
279 | negative the key is assumed to be in UTF-8-encoded Unicode. If | |
280 | C<lval> is set then the fetch will be part of a store. Check that the | |
281 | return value is non-null before dereferencing it to an C<SV*>. | |
954c1994 | 282 | |
96f1132b | 283 | See L<perlguts/"Understanding the Magic of Tied Hashes and Arrays"> for more |
954c1994 GS |
284 | information on how to use this function on tied hashes. |
285 | ||
34a6f7b4 NC |
286 | =for apidoc hv_exists_ent |
287 | ||
288 | Returns a boolean indicating whether the specified hash key exists. C<hash> | |
289 | can be a valid precomputed hash value, or 0 to ask for it to be | |
290 | computed. | |
291 | ||
292 | =cut | |
293 | */ | |
294 | ||
d1be9408 | 295 | /* returns an HE * structure with the all fields set */ |
fde52b5c | 296 | /* note that hent_val will be a mortal sv for MAGICAL hashes */ |
954c1994 GS |
297 | /* |
298 | =for apidoc hv_fetch_ent | |
299 | ||
300 | Returns the hash entry which corresponds to the specified key in the hash. | |
301 | C<hash> must be a valid precomputed hash number for the given C<key>, or 0 | |
302 | if you want the function to compute it. IF C<lval> is set then the fetch | |
303 | will be part of a store. Make sure the return value is non-null before | |
b24b84ef | 304 | accessing it. The return value when C<hv> is a tied hash is a pointer to a |
954c1994 | 305 | static location, so be sure to make a copy of the structure if you need to |
1c846c1f | 306 | store it somewhere. |
954c1994 | 307 | |
96f1132b | 308 | See L<perlguts/"Understanding the Magic of Tied Hashes and Arrays"> for more |
954c1994 GS |
309 | information on how to use this function on tied hashes. |
310 | ||
311 | =cut | |
312 | */ | |
313 | ||
a038e571 NC |
314 | /* Common code for hv_delete()/hv_exists()/hv_fetch()/hv_store() */ |
315 | void * | |
316 | Perl_hv_common_key_len(pTHX_ HV *hv, const char *key, I32 klen_i32, | |
317 | const int action, SV *val, const U32 hash) | |
318 | { | |
319 | STRLEN klen; | |
320 | int flags; | |
321 | ||
7918f24d NC |
322 | PERL_ARGS_ASSERT_HV_COMMON_KEY_LEN; |
323 | ||
a038e571 NC |
324 | if (klen_i32 < 0) { |
325 | klen = -klen_i32; | |
326 | flags = HVhek_UTF8; | |
327 | } else { | |
328 | klen = klen_i32; | |
329 | flags = 0; | |
330 | } | |
331 | return hv_common(hv, NULL, key, klen, flags, action, val, hash); | |
332 | } | |
333 | ||
63c89345 | 334 | void * |
d3ba3f5c NC |
335 | Perl_hv_common(pTHX_ HV *hv, SV *keysv, const char *key, STRLEN klen, |
336 | int flags, int action, SV *val, register U32 hash) | |
113738bb | 337 | { |
27da23d5 | 338 | dVAR; |
b2c64049 | 339 | XPVHV* xhv; |
b2c64049 NC |
340 | HE *entry; |
341 | HE **oentry; | |
fde52b5c | 342 | SV *sv; |
da58a35d | 343 | bool is_utf8; |
113738bb | 344 | int masked_flags; |
3c84c864 | 345 | const int return_svp = action & HV_FETCH_JUST_SV; |
fde52b5c | 346 | |
347 | if (!hv) | |
a4fc7abc | 348 | return NULL; |
e4787c0c | 349 | if (SvTYPE(hv) == (svtype)SVTYPEMASK) |
8265e3d1 NC |
350 | return NULL; |
351 | ||
352 | assert(SvTYPE(hv) == SVt_PVHV); | |
fde52b5c | 353 | |
bdee33e4 | 354 | if (SvSMAGICAL(hv) && SvGMAGICAL(hv) && !(action & HV_DISABLE_UVAR_XKEY)) { |
fda2d18a | 355 | MAGIC* mg; |
ad64d0ec | 356 | if ((mg = mg_find((const SV *)hv, PERL_MAGIC_uvar))) { |
fda2d18a NC |
357 | struct ufuncs * const uf = (struct ufuncs *)mg->mg_ptr; |
358 | if (uf->uf_set == NULL) { | |
359 | SV* obj = mg->mg_obj; | |
360 | ||
361 | if (!keysv) { | |
59cd0e26 NC |
362 | keysv = newSVpvn_flags(key, klen, SVs_TEMP | |
363 | ((flags & HVhek_UTF8) | |
364 | ? SVf_UTF8 : 0)); | |
fda2d18a NC |
365 | } |
366 | ||
367 | mg->mg_obj = keysv; /* pass key */ | |
368 | uf->uf_index = action; /* pass action */ | |
ad64d0ec | 369 | magic_getuvar(MUTABLE_SV(hv), mg); |
fda2d18a NC |
370 | keysv = mg->mg_obj; /* may have changed */ |
371 | mg->mg_obj = obj; | |
372 | ||
373 | /* If the key may have changed, then we need to invalidate | |
374 | any passed-in computed hash value. */ | |
375 | hash = 0; | |
376 | } | |
377 | } | |
bdee33e4 | 378 | } |
113738bb | 379 | if (keysv) { |
e593d2fe AE |
380 | if (flags & HVhek_FREEKEY) |
381 | Safefree(key); | |
5c144d81 | 382 | key = SvPV_const(keysv, klen); |
113738bb | 383 | is_utf8 = (SvUTF8(keysv) != 0); |
44b87b50 NC |
384 | if (SvIsCOW_shared_hash(keysv)) { |
385 | flags = HVhek_KEYCANONICAL | (is_utf8 ? HVhek_UTF8 : 0); | |
386 | } else { | |
387 | flags = 0; | |
388 | } | |
113738bb | 389 | } else { |
c1fe5510 | 390 | is_utf8 = ((flags & HVhek_UTF8) ? TRUE : FALSE); |
113738bb | 391 | } |
113738bb | 392 | |
9dbc5603 | 393 | if (action & HV_DELETE) { |
3c84c864 NC |
394 | return (void *) hv_delete_common(hv, keysv, key, klen, |
395 | flags | (is_utf8 ? HVhek_UTF8 : 0), | |
396 | action, hash); | |
9dbc5603 NC |
397 | } |
398 | ||
b2c64049 | 399 | xhv = (XPVHV*)SvANY(hv); |
7f66fda2 | 400 | if (SvMAGICAL(hv)) { |
6136c704 | 401 | if (SvRMAGICAL(hv) && !(action & (HV_FETCH_ISSTORE|HV_FETCH_ISEXISTS))) { |
ad64d0ec NC |
402 | if (mg_find((const SV *)hv, PERL_MAGIC_tied) |
403 | || SvGMAGICAL((const SV *)hv)) | |
e62cc96a | 404 | { |
3c84c864 | 405 | /* FIXME should be able to skimp on the HE/HEK here when |
7f66fda2 | 406 | HV_FETCH_JUST_SV is true. */ |
7f66fda2 | 407 | if (!keysv) { |
740cce10 NC |
408 | keysv = newSVpvn_utf8(key, klen, is_utf8); |
409 | } else { | |
7f66fda2 | 410 | keysv = newSVsv(keysv); |
113738bb | 411 | } |
44a2ac75 | 412 | sv = sv_newmortal(); |
ad64d0ec | 413 | mg_copy(MUTABLE_SV(hv), sv, (char *)keysv, HEf_SVKEY); |
7f66fda2 NC |
414 | |
415 | /* grab a fake HE/HEK pair from the pool or make a new one */ | |
416 | entry = PL_hv_fetch_ent_mh; | |
417 | if (entry) | |
418 | PL_hv_fetch_ent_mh = HeNEXT(entry); | |
419 | else { | |
420 | char *k; | |
421 | entry = new_HE(); | |
ad64d0ec | 422 | Newx(k, HEK_BASESIZE + sizeof(const SV *), char); |
7f66fda2 NC |
423 | HeKEY_hek(entry) = (HEK*)k; |
424 | } | |
4608196e | 425 | HeNEXT(entry) = NULL; |
7f66fda2 NC |
426 | HeSVKEY_set(entry, keysv); |
427 | HeVAL(entry) = sv; | |
428 | sv_upgrade(sv, SVt_PVLV); | |
429 | LvTYPE(sv) = 'T'; | |
430 | /* so we can free entry when freeing sv */ | |
ad64d0ec | 431 | LvTARG(sv) = MUTABLE_SV(entry); |
7f66fda2 NC |
432 | |
433 | /* XXX remove at some point? */ | |
434 | if (flags & HVhek_FREEKEY) | |
435 | Safefree(key); | |
436 | ||
3c84c864 NC |
437 | if (return_svp) { |
438 | return entry ? (void *) &HeVAL(entry) : NULL; | |
439 | } | |
440 | return (void *) entry; | |
113738bb | 441 | } |
7f66fda2 | 442 | #ifdef ENV_IS_CASELESS |
ad64d0ec | 443 | else if (mg_find((const SV *)hv, PERL_MAGIC_env)) { |
7f66fda2 NC |
444 | U32 i; |
445 | for (i = 0; i < klen; ++i) | |
446 | if (isLOWER(key[i])) { | |
086cb327 NC |
447 | /* Would be nice if we had a routine to do the |
448 | copy and upercase in a single pass through. */ | |
0bd48802 | 449 | const char * const nkey = strupr(savepvn(key,klen)); |
086cb327 NC |
450 | /* Note that this fetch is for nkey (the uppercased |
451 | key) whereas the store is for key (the original) */ | |
63c89345 NC |
452 | void *result = hv_common(hv, NULL, nkey, klen, |
453 | HVhek_FREEKEY, /* free nkey */ | |
454 | 0 /* non-LVAL fetch */ | |
3c84c864 NC |
455 | | HV_DISABLE_UVAR_XKEY |
456 | | return_svp, | |
63c89345 NC |
457 | NULL /* no value */, |
458 | 0 /* compute hash */); | |
26488bcf | 459 | if (!result && (action & HV_FETCH_LVALUE)) { |
086cb327 NC |
460 | /* This call will free key if necessary. |
461 | Do it this way to encourage compiler to tail | |
462 | call optimise. */ | |
63c89345 NC |
463 | result = hv_common(hv, keysv, key, klen, flags, |
464 | HV_FETCH_ISSTORE | |
3c84c864 NC |
465 | | HV_DISABLE_UVAR_XKEY |
466 | | return_svp, | |
63c89345 | 467 | newSV(0), hash); |
086cb327 NC |
468 | } else { |
469 | if (flags & HVhek_FREEKEY) | |
470 | Safefree(key); | |
471 | } | |
63c89345 | 472 | return result; |
7f66fda2 | 473 | } |
902173a3 | 474 | } |
7f66fda2 NC |
475 | #endif |
476 | } /* ISFETCH */ | |
477 | else if (SvRMAGICAL(hv) && (action & HV_FETCH_ISEXISTS)) { | |
ad64d0ec NC |
478 | if (mg_find((const SV *)hv, PERL_MAGIC_tied) |
479 | || SvGMAGICAL((const SV *)hv)) { | |
b2c64049 NC |
480 | /* I don't understand why hv_exists_ent has svret and sv, |
481 | whereas hv_exists only had one. */ | |
9d4ba2ae | 482 | SV * const svret = sv_newmortal(); |
b2c64049 | 483 | sv = sv_newmortal(); |
7f66fda2 NC |
484 | |
485 | if (keysv || is_utf8) { | |
486 | if (!keysv) { | |
740cce10 | 487 | keysv = newSVpvn_utf8(key, klen, TRUE); |
7f66fda2 NC |
488 | } else { |
489 | keysv = newSVsv(keysv); | |
490 | } | |
ad64d0ec | 491 | mg_copy(MUTABLE_SV(hv), sv, (char *)sv_2mortal(keysv), HEf_SVKEY); |
b2c64049 | 492 | } else { |
ad64d0ec | 493 | mg_copy(MUTABLE_SV(hv), sv, key, klen); |
7f66fda2 | 494 | } |
b2c64049 NC |
495 | if (flags & HVhek_FREEKEY) |
496 | Safefree(key); | |
7f66fda2 NC |
497 | magic_existspack(svret, mg_find(sv, PERL_MAGIC_tiedelem)); |
498 | /* This cast somewhat evil, but I'm merely using NULL/ | |
499 | not NULL to return the boolean exists. | |
500 | And I know hv is not NULL. */ | |
3c84c864 | 501 | return SvTRUE(svret) ? (void *)hv : NULL; |
e7152ba2 | 502 | } |
7f66fda2 | 503 | #ifdef ENV_IS_CASELESS |
ad64d0ec | 504 | else if (mg_find((const SV *)hv, PERL_MAGIC_env)) { |
7f66fda2 | 505 | /* XXX This code isn't UTF8 clean. */ |
a15d23f8 | 506 | char * const keysave = (char * const)key; |
b2c64049 NC |
507 | /* Will need to free this, so set FREEKEY flag. */ |
508 | key = savepvn(key,klen); | |
509 | key = (const char*)strupr((char*)key); | |
6136c704 | 510 | is_utf8 = FALSE; |
7f66fda2 | 511 | hash = 0; |
8b4f7dd5 | 512 | keysv = 0; |
b2c64049 NC |
513 | |
514 | if (flags & HVhek_FREEKEY) { | |
515 | Safefree(keysave); | |
516 | } | |
517 | flags |= HVhek_FREEKEY; | |
7f66fda2 | 518 | } |
902173a3 | 519 | #endif |
7f66fda2 | 520 | } /* ISEXISTS */ |
b2c64049 NC |
521 | else if (action & HV_FETCH_ISSTORE) { |
522 | bool needs_copy; | |
523 | bool needs_store; | |
524 | hv_magic_check (hv, &needs_copy, &needs_store); | |
525 | if (needs_copy) { | |
a3b680e6 | 526 | const bool save_taint = PL_tainted; |
b2c64049 NC |
527 | if (keysv || is_utf8) { |
528 | if (!keysv) { | |
740cce10 | 529 | keysv = newSVpvn_utf8(key, klen, TRUE); |
b2c64049 NC |
530 | } |
531 | if (PL_tainting) | |
532 | PL_tainted = SvTAINTED(keysv); | |
533 | keysv = sv_2mortal(newSVsv(keysv)); | |
ad64d0ec | 534 | mg_copy(MUTABLE_SV(hv), val, (char*)keysv, HEf_SVKEY); |
b2c64049 | 535 | } else { |
ad64d0ec | 536 | mg_copy(MUTABLE_SV(hv), val, key, klen); |
b2c64049 NC |
537 | } |
538 | ||
539 | TAINT_IF(save_taint); | |
1baaf5d7 | 540 | if (!needs_store) { |
b2c64049 NC |
541 | if (flags & HVhek_FREEKEY) |
542 | Safefree(key); | |
4608196e | 543 | return NULL; |
b2c64049 NC |
544 | } |
545 | #ifdef ENV_IS_CASELESS | |
ad64d0ec | 546 | else if (mg_find((const SV *)hv, PERL_MAGIC_env)) { |
b2c64049 NC |
547 | /* XXX This code isn't UTF8 clean. */ |
548 | const char *keysave = key; | |
549 | /* Will need to free this, so set FREEKEY flag. */ | |
550 | key = savepvn(key,klen); | |
551 | key = (const char*)strupr((char*)key); | |
6136c704 | 552 | is_utf8 = FALSE; |
b2c64049 | 553 | hash = 0; |
8b4f7dd5 | 554 | keysv = 0; |
b2c64049 NC |
555 | |
556 | if (flags & HVhek_FREEKEY) { | |
557 | Safefree(keysave); | |
558 | } | |
559 | flags |= HVhek_FREEKEY; | |
560 | } | |
561 | #endif | |
562 | } | |
563 | } /* ISSTORE */ | |
7f66fda2 | 564 | } /* SvMAGICAL */ |
fde52b5c | 565 | |
7b2c381c | 566 | if (!HvARRAY(hv)) { |
b2c64049 | 567 | if ((action & (HV_FETCH_LVALUE | HV_FETCH_ISSTORE)) |
fde52b5c | 568 | #ifdef DYNAMIC_ENV_FETCH /* if it's an %ENV lookup, we may get it on the fly */ |
ad64d0ec NC |
569 | || (SvRMAGICAL((const SV *)hv) |
570 | && mg_find((const SV *)hv, PERL_MAGIC_env)) | |
fde52b5c | 571 | #endif |
d58e6666 NC |
572 | ) { |
573 | char *array; | |
a02a5408 | 574 | Newxz(array, |
cbec9347 | 575 | PERL_HV_ARRAY_ALLOC_BYTES(xhv->xhv_max+1 /* HvMAX(hv)+1 */), |
d58e6666 NC |
576 | char); |
577 | HvARRAY(hv) = (HE**)array; | |
578 | } | |
7f66fda2 NC |
579 | #ifdef DYNAMIC_ENV_FETCH |
580 | else if (action & HV_FETCH_ISEXISTS) { | |
581 | /* for an %ENV exists, if we do an insert it's by a recursive | |
582 | store call, so avoid creating HvARRAY(hv) right now. */ | |
583 | } | |
584 | #endif | |
113738bb NC |
585 | else { |
586 | /* XXX remove at some point? */ | |
587 | if (flags & HVhek_FREEKEY) | |
588 | Safefree(key); | |
589 | ||
3c84c864 | 590 | return NULL; |
113738bb | 591 | } |
fde52b5c | 592 | } |
593 | ||
44b87b50 | 594 | if (is_utf8 & !(flags & HVhek_KEYCANONICAL)) { |
41d88b63 | 595 | char * const keysave = (char *)key; |
f9a63242 | 596 | key = (char*)bytes_from_utf8((U8*)key, &klen, &is_utf8); |
19692e8d | 597 | if (is_utf8) |
c1fe5510 NC |
598 | flags |= HVhek_UTF8; |
599 | else | |
600 | flags &= ~HVhek_UTF8; | |
7f66fda2 NC |
601 | if (key != keysave) { |
602 | if (flags & HVhek_FREEKEY) | |
603 | Safefree(keysave); | |
19692e8d | 604 | flags |= HVhek_WASUTF8 | HVhek_FREEKEY; |
527df579 NC |
605 | /* If the caller calculated a hash, it was on the sequence of |
606 | octets that are the UTF-8 form. We've now changed the sequence | |
607 | of octets stored to that of the equivalent byte representation, | |
608 | so the hash we need is different. */ | |
609 | hash = 0; | |
7f66fda2 | 610 | } |
19692e8d | 611 | } |
f9a63242 | 612 | |
f8d50d94 DM |
613 | if (HvREHASH(hv) || (!hash && !(keysv && (SvIsCOW_shared_hash(keysv))))) |
614 | PERL_HASH_INTERNAL_(hash, key, klen, HvREHASH(hv)); | |
615 | else if (!hash) | |
616 | hash = SvSHARED_HASH(keysv); | |
617 | ||
618 | /* We don't have a pointer to the hv, so we have to replicate the | |
619 | flag into every HEK, so that hv_iterkeysv can see it. | |
620 | And yes, you do need this even though you are not "storing" because | |
621 | you can flip the flags below if doing an lval lookup. (And that | |
622 | was put in to give the semantics Andreas was expecting.) */ | |
623 | if (HvREHASH(hv)) | |
fdcd69b6 | 624 | flags |= HVhek_REHASH; |
effa1e2d | 625 | |
113738bb NC |
626 | masked_flags = (flags & HVhek_MASK); |
627 | ||
7f66fda2 | 628 | #ifdef DYNAMIC_ENV_FETCH |
4608196e | 629 | if (!HvARRAY(hv)) entry = NULL; |
7f66fda2 NC |
630 | else |
631 | #endif | |
b2c64049 | 632 | { |
7b2c381c | 633 | entry = (HvARRAY(hv))[hash & (I32) HvMAX(hv)]; |
b2c64049 | 634 | } |
0298d7b9 | 635 | for (; entry; entry = HeNEXT(entry)) { |
fde52b5c | 636 | if (HeHASH(entry) != hash) /* strings can't be equal */ |
637 | continue; | |
eb160463 | 638 | if (HeKLEN(entry) != (I32)klen) |
fde52b5c | 639 | continue; |
1c846c1f | 640 | if (HeKEY(entry) != key && memNE(HeKEY(entry),key,klen)) /* is this it? */ |
fde52b5c | 641 | continue; |
113738bb | 642 | if ((HeKFLAGS(entry) ^ masked_flags) & HVhek_UTF8) |
c3654f1a | 643 | continue; |
b2c64049 NC |
644 | |
645 | if (action & (HV_FETCH_LVALUE|HV_FETCH_ISSTORE)) { | |
646 | if (HeKFLAGS(entry) != masked_flags) { | |
647 | /* We match if HVhek_UTF8 bit in our flags and hash key's | |
648 | match. But if entry was set previously with HVhek_WASUTF8 | |
649 | and key now doesn't (or vice versa) then we should change | |
650 | the key's flag, as this is assignment. */ | |
651 | if (HvSHAREKEYS(hv)) { | |
652 | /* Need to swap the key we have for a key with the flags we | |
653 | need. As keys are shared we can't just write to the | |
654 | flag, so we share the new one, unshare the old one. */ | |
6136c704 | 655 | HEK * const new_hek = share_hek_flags(key, klen, hash, |
6e838c70 | 656 | masked_flags); |
b2c64049 NC |
657 | unshare_hek (HeKEY_hek(entry)); |
658 | HeKEY_hek(entry) = new_hek; | |
659 | } | |
5d2b1485 NC |
660 | else if (hv == PL_strtab) { |
661 | /* PL_strtab is usually the only hash without HvSHAREKEYS, | |
662 | so putting this test here is cheap */ | |
663 | if (flags & HVhek_FREEKEY) | |
664 | Safefree(key); | |
665 | Perl_croak(aTHX_ S_strtab_error, | |
666 | action & HV_FETCH_LVALUE ? "fetch" : "store"); | |
667 | } | |
b2c64049 NC |
668 | else |
669 | HeKFLAGS(entry) = masked_flags; | |
670 | if (masked_flags & HVhek_ENABLEHVKFLAGS) | |
671 | HvHASKFLAGS_on(hv); | |
672 | } | |
673 | if (HeVAL(entry) == &PL_sv_placeholder) { | |
674 | /* yes, can store into placeholder slot */ | |
675 | if (action & HV_FETCH_LVALUE) { | |
676 | if (SvMAGICAL(hv)) { | |
677 | /* This preserves behaviour with the old hv_fetch | |
678 | implementation which at this point would bail out | |
679 | with a break; (at "if we find a placeholder, we | |
680 | pretend we haven't found anything") | |
681 | ||
682 | That break mean that if a placeholder were found, it | |
683 | caused a call into hv_store, which in turn would | |
684 | check magic, and if there is no magic end up pretty | |
685 | much back at this point (in hv_store's code). */ | |
686 | break; | |
687 | } | |
486ec47a | 688 | /* LVAL fetch which actually needs a store. */ |
561b68a9 | 689 | val = newSV(0); |
ca732855 | 690 | HvPLACEHOLDERS(hv)--; |
b2c64049 NC |
691 | } else { |
692 | /* store */ | |
693 | if (val != &PL_sv_placeholder) | |
ca732855 | 694 | HvPLACEHOLDERS(hv)--; |
b2c64049 NC |
695 | } |
696 | HeVAL(entry) = val; | |
697 | } else if (action & HV_FETCH_ISSTORE) { | |
cefd5c7c | 698 | SvREFCNT_dec(HeVAL(entry)); |
b2c64049 NC |
699 | HeVAL(entry) = val; |
700 | } | |
27bcc0a7 | 701 | } else if (HeVAL(entry) == &PL_sv_placeholder) { |
b2c64049 NC |
702 | /* if we find a placeholder, we pretend we haven't found |
703 | anything */ | |
8aacddc1 | 704 | break; |
b2c64049 | 705 | } |
113738bb NC |
706 | if (flags & HVhek_FREEKEY) |
707 | Safefree(key); | |
3c84c864 NC |
708 | if (return_svp) { |
709 | return entry ? (void *) &HeVAL(entry) : NULL; | |
710 | } | |
fde52b5c | 711 | return entry; |
712 | } | |
713 | #ifdef DYNAMIC_ENV_FETCH /* %ENV lookup? If so, try to fetch the value now */ | |
0ed29950 | 714 | if (!(action & HV_FETCH_ISSTORE) |
ad64d0ec NC |
715 | && SvRMAGICAL((const SV *)hv) |
716 | && mg_find((const SV *)hv, PERL_MAGIC_env)) { | |
a6c40364 | 717 | unsigned long len; |
9d4ba2ae | 718 | const char * const env = PerlEnv_ENVgetenv_len(key,&len); |
a6c40364 GS |
719 | if (env) { |
720 | sv = newSVpvn(env,len); | |
721 | SvTAINTED_on(sv); | |
d3ba3f5c | 722 | return hv_common(hv, keysv, key, klen, flags, |
3c84c864 NC |
723 | HV_FETCH_ISSTORE|HV_DISABLE_UVAR_XKEY|return_svp, |
724 | sv, hash); | |
a6c40364 | 725 | } |
fde52b5c | 726 | } |
727 | #endif | |
7f66fda2 NC |
728 | |
729 | if (!entry && SvREADONLY(hv) && !(action & HV_FETCH_ISEXISTS)) { | |
c445ea15 | 730 | hv_notallowed(flags, key, klen, |
c8cd6465 NC |
731 | "Attempt to access disallowed key '%"SVf"' in" |
732 | " a restricted hash"); | |
1b1f1335 | 733 | } |
b2c64049 NC |
734 | if (!(action & (HV_FETCH_LVALUE|HV_FETCH_ISSTORE))) { |
735 | /* Not doing some form of store, so return failure. */ | |
736 | if (flags & HVhek_FREEKEY) | |
737 | Safefree(key); | |
3c84c864 | 738 | return NULL; |
b2c64049 | 739 | } |
113738bb | 740 | if (action & HV_FETCH_LVALUE) { |
df5f182b | 741 | val = action & HV_FETCH_EMPTY_HE ? NULL : newSV(0); |
b2c64049 NC |
742 | if (SvMAGICAL(hv)) { |
743 | /* At this point the old hv_fetch code would call to hv_store, | |
744 | which in turn might do some tied magic. So we need to make that | |
745 | magic check happen. */ | |
746 | /* gonna assign to this, so it better be there */ | |
fda2d18a NC |
747 | /* If a fetch-as-store fails on the fetch, then the action is to |
748 | recurse once into "hv_store". If we didn't do this, then that | |
749 | recursive call would call the key conversion routine again. | |
750 | However, as we replace the original key with the converted | |
751 | key, this would result in a double conversion, which would show | |
752 | up as a bug if the conversion routine is not idempotent. */ | |
d3ba3f5c | 753 | return hv_common(hv, keysv, key, klen, flags, |
3c84c864 NC |
754 | HV_FETCH_ISSTORE|HV_DISABLE_UVAR_XKEY|return_svp, |
755 | val, hash); | |
b2c64049 NC |
756 | /* XXX Surely that could leak if the fetch-was-store fails? |
757 | Just like the hv_fetch. */ | |
113738bb NC |
758 | } |
759 | } | |
760 | ||
b2c64049 NC |
761 | /* Welcome to hv_store... */ |
762 | ||
7b2c381c | 763 | if (!HvARRAY(hv)) { |
b2c64049 NC |
764 | /* Not sure if we can get here. I think the only case of oentry being |
765 | NULL is for %ENV with dynamic env fetch. But that should disappear | |
766 | with magic in the previous code. */ | |
d58e6666 | 767 | char *array; |
a02a5408 | 768 | Newxz(array, |
b2c64049 | 769 | PERL_HV_ARRAY_ALLOC_BYTES(xhv->xhv_max+1 /* HvMAX(hv)+1 */), |
d58e6666 NC |
770 | char); |
771 | HvARRAY(hv) = (HE**)array; | |
b2c64049 NC |
772 | } |
773 | ||
7b2c381c | 774 | oentry = &(HvARRAY(hv))[hash & (I32) xhv->xhv_max]; |
ab4af705 | 775 | |
b2c64049 NC |
776 | entry = new_HE(); |
777 | /* share_hek_flags will do the free for us. This might be considered | |
778 | bad API design. */ | |
779 | if (HvSHAREKEYS(hv)) | |
6e838c70 | 780 | HeKEY_hek(entry) = share_hek_flags(key, klen, hash, flags); |
5d2b1485 NC |
781 | else if (hv == PL_strtab) { |
782 | /* PL_strtab is usually the only hash without HvSHAREKEYS, so putting | |
783 | this test here is cheap */ | |
784 | if (flags & HVhek_FREEKEY) | |
785 | Safefree(key); | |
786 | Perl_croak(aTHX_ S_strtab_error, | |
787 | action & HV_FETCH_LVALUE ? "fetch" : "store"); | |
788 | } | |
b2c64049 NC |
789 | else /* gotta do the real thing */ |
790 | HeKEY_hek(entry) = save_hek_flags(key, klen, hash, flags); | |
791 | HeVAL(entry) = val; | |
792 | HeNEXT(entry) = *oentry; | |
793 | *oentry = entry; | |
794 | ||
795 | if (val == &PL_sv_placeholder) | |
ca732855 | 796 | HvPLACEHOLDERS(hv)++; |
b2c64049 NC |
797 | if (masked_flags & HVhek_ENABLEHVKFLAGS) |
798 | HvHASKFLAGS_on(hv); | |
799 | ||
0298d7b9 NC |
800 | { |
801 | const HE *counter = HeNEXT(entry); | |
802 | ||
4c7185a0 | 803 | xhv->xhv_keys++; /* HvTOTALKEYS(hv)++ */ |
0298d7b9 | 804 | if (!counter) { /* initial entry? */ |
5ac36297 | 805 | } else if (xhv->xhv_keys > xhv->xhv_max) { |
1b95d04f | 806 | /* Use only the old HvUSEDKEYS(hv) > HvMAX(hv) condition to limit |
5b430b36 MW |
807 | bucket splits on a rehashed hash, as we're not going to |
808 | split it again, and if someone is lucky (evil) enough to | |
809 | get all the keys in one list they could exhaust our memory | |
810 | as we repeatedly double the number of buckets on every | |
811 | entry. Linear search feels a less worse thing to do. */ | |
0298d7b9 NC |
812 | hsplit(hv); |
813 | } else if(!HvREHASH(hv)) { | |
814 | U32 n_links = 1; | |
815 | ||
816 | while ((counter = HeNEXT(counter))) | |
817 | n_links++; | |
818 | ||
819 | if (n_links > HV_MAX_LENGTH_BEFORE_SPLIT) { | |
0298d7b9 NC |
820 | hsplit(hv); |
821 | } | |
822 | } | |
fde52b5c | 823 | } |
b2c64049 | 824 | |
3c84c864 NC |
825 | if (return_svp) { |
826 | return entry ? (void *) &HeVAL(entry) : NULL; | |
827 | } | |
828 | return (void *) entry; | |
fde52b5c | 829 | } |
830 | ||
864dbfa3 | 831 | STATIC void |
b0e6ae5b | 832 | S_hv_magic_check(HV *hv, bool *needs_copy, bool *needs_store) |
d0066dc7 | 833 | { |
a3b680e6 | 834 | const MAGIC *mg = SvMAGIC(hv); |
7918f24d NC |
835 | |
836 | PERL_ARGS_ASSERT_HV_MAGIC_CHECK; | |
837 | ||
d0066dc7 OT |
838 | *needs_copy = FALSE; |
839 | *needs_store = TRUE; | |
840 | while (mg) { | |
841 | if (isUPPER(mg->mg_type)) { | |
842 | *needs_copy = TRUE; | |
d60c5a05 | 843 | if (mg->mg_type == PERL_MAGIC_tied) { |
d0066dc7 | 844 | *needs_store = FALSE; |
4ab2a30b | 845 | return; /* We've set all there is to set. */ |
d0066dc7 OT |
846 | } |
847 | } | |
848 | mg = mg->mg_moremagic; | |
849 | } | |
850 | } | |
851 | ||
954c1994 | 852 | /* |
a3bcc51e TP |
853 | =for apidoc hv_scalar |
854 | ||
855 | Evaluates the hash in scalar context and returns the result. Handles magic when the hash is tied. | |
856 | ||
857 | =cut | |
858 | */ | |
859 | ||
860 | SV * | |
861 | Perl_hv_scalar(pTHX_ HV *hv) | |
862 | { | |
a3bcc51e | 863 | SV *sv; |
823a54a3 | 864 | |
7918f24d NC |
865 | PERL_ARGS_ASSERT_HV_SCALAR; |
866 | ||
823a54a3 | 867 | if (SvRMAGICAL(hv)) { |
ad64d0ec | 868 | MAGIC * const mg = mg_find((const SV *)hv, PERL_MAGIC_tied); |
823a54a3 AL |
869 | if (mg) |
870 | return magic_scalarpack(hv, mg); | |
871 | } | |
a3bcc51e TP |
872 | |
873 | sv = sv_newmortal(); | |
f4431c56 | 874 | if (HvTOTALKEYS((const HV *)hv)) |
a3bcc51e TP |
875 | Perl_sv_setpvf(aTHX_ sv, "%ld/%ld", |
876 | (long)HvFILL(hv), (long)HvMAX(hv) + 1); | |
877 | else | |
878 | sv_setiv(sv, 0); | |
879 | ||
880 | return sv; | |
881 | } | |
882 | ||
883 | /* | |
954c1994 GS |
884 | =for apidoc hv_delete |
885 | ||
a05d6c5d TC |
886 | Deletes a key/value pair in the hash. The value's SV is removed from |
887 | the hash, made mortal, and returned to the caller. The absolute | |
888 | value of C<klen> is the length of the key. If C<klen> is negative the | |
889 | key is assumed to be in UTF-8-encoded Unicode. The C<flags> value | |
890 | will normally be zero; if set to G_DISCARD then NULL will be returned. | |
891 | NULL will also be returned if the key is not found. | |
954c1994 | 892 | |
954c1994 GS |
893 | =for apidoc hv_delete_ent |
894 | ||
3025a2e4 CS |
895 | Deletes a key/value pair in the hash. The value SV is removed from the hash, |
896 | made mortal, and returned to the caller. The C<flags> value will normally be | |
897 | zero; if set to G_DISCARD then NULL will be returned. NULL will also be | |
898 | returned if the key is not found. C<hash> can be a valid precomputed hash | |
899 | value, or 0 to ask for it to be computed. | |
954c1994 GS |
900 | |
901 | =cut | |
902 | */ | |
903 | ||
8f8d40ab | 904 | STATIC SV * |
cd6d36ac NC |
905 | S_hv_delete_common(pTHX_ HV *hv, SV *keysv, const char *key, STRLEN klen, |
906 | int k_flags, I32 d_flags, U32 hash) | |
f1317c8d | 907 | { |
27da23d5 | 908 | dVAR; |
cbec9347 | 909 | register XPVHV* xhv; |
fde52b5c | 910 | register HE *entry; |
911 | register HE **oentry; | |
9dbc5603 | 912 | bool is_utf8 = (k_flags & HVhek_UTF8) ? TRUE : FALSE; |
7a9669ca | 913 | int masked_flags; |
1c846c1f | 914 | |
fde52b5c | 915 | if (SvRMAGICAL(hv)) { |
0a0bb7c7 OT |
916 | bool needs_copy; |
917 | bool needs_store; | |
918 | hv_magic_check (hv, &needs_copy, &needs_store); | |
919 | ||
f1317c8d | 920 | if (needs_copy) { |
6136c704 | 921 | SV *sv; |
63c89345 NC |
922 | entry = (HE *) hv_common(hv, keysv, key, klen, |
923 | k_flags & ~HVhek_FREEKEY, | |
924 | HV_FETCH_LVALUE|HV_DISABLE_UVAR_XKEY, | |
925 | NULL, hash); | |
7a9669ca | 926 | sv = entry ? HeVAL(entry) : NULL; |
f1317c8d NC |
927 | if (sv) { |
928 | if (SvMAGICAL(sv)) { | |
929 | mg_clear(sv); | |
930 | } | |
931 | if (!needs_store) { | |
932 | if (mg_find(sv, PERL_MAGIC_tiedelem)) { | |
933 | /* No longer an element */ | |
934 | sv_unmagic(sv, PERL_MAGIC_tiedelem); | |
935 | return sv; | |
936 | } | |
a0714e2c | 937 | return NULL; /* element cannot be deleted */ |
f1317c8d | 938 | } |
902173a3 | 939 | #ifdef ENV_IS_CASELESS |
ad64d0ec | 940 | else if (mg_find((const SV *)hv, PERL_MAGIC_env)) { |
8167a60a | 941 | /* XXX This code isn't UTF8 clean. */ |
59cd0e26 | 942 | keysv = newSVpvn_flags(key, klen, SVs_TEMP); |
8167a60a NC |
943 | if (k_flags & HVhek_FREEKEY) { |
944 | Safefree(key); | |
945 | } | |
946 | key = strupr(SvPVX(keysv)); | |
947 | is_utf8 = 0; | |
948 | k_flags = 0; | |
949 | hash = 0; | |
7f66fda2 | 950 | } |
510ac311 | 951 | #endif |
2fd1c6b8 | 952 | } |
2fd1c6b8 | 953 | } |
fde52b5c | 954 | } |
cbec9347 | 955 | xhv = (XPVHV*)SvANY(hv); |
7b2c381c | 956 | if (!HvARRAY(hv)) |
a0714e2c | 957 | return NULL; |
fde52b5c | 958 | |
19692e8d | 959 | if (is_utf8) { |
c445ea15 | 960 | const char * const keysave = key; |
b464bac0 | 961 | key = (char*)bytes_from_utf8((U8*)key, &klen, &is_utf8); |
cd6d36ac | 962 | |
19692e8d | 963 | if (is_utf8) |
cd6d36ac NC |
964 | k_flags |= HVhek_UTF8; |
965 | else | |
966 | k_flags &= ~HVhek_UTF8; | |
7f66fda2 NC |
967 | if (key != keysave) { |
968 | if (k_flags & HVhek_FREEKEY) { | |
969 | /* This shouldn't happen if our caller does what we expect, | |
970 | but strictly the API allows it. */ | |
971 | Safefree(keysave); | |
972 | } | |
973 | k_flags |= HVhek_WASUTF8 | HVhek_FREEKEY; | |
974 | } | |
ad64d0ec | 975 | HvHASKFLAGS_on(MUTABLE_SV(hv)); |
19692e8d | 976 | } |
f9a63242 | 977 | |
f8d50d94 DM |
978 | if (HvREHASH(hv) || (!hash && !(keysv && (SvIsCOW_shared_hash(keysv))))) |
979 | PERL_HASH_INTERNAL_(hash, key, klen, HvREHASH(hv)); | |
980 | else if (!hash) | |
981 | hash = SvSHARED_HASH(keysv); | |
fde52b5c | 982 | |
7a9669ca NC |
983 | masked_flags = (k_flags & HVhek_MASK); |
984 | ||
9de10d5c | 985 | oentry = &(HvARRAY(hv))[hash & (I32) HvMAX(hv)]; |
fde52b5c | 986 | entry = *oentry; |
9e720f71 | 987 | for (; entry; oentry = &HeNEXT(entry), entry = *oentry) { |
6136c704 | 988 | SV *sv; |
f3d2f32d | 989 | U8 mro_changes = 0; /* 1 = isa; 2 = package moved */ |
0290c710 | 990 | GV *gv = NULL; |
0c3bb3c2 FC |
991 | HV *stash = NULL; |
992 | ||
fde52b5c | 993 | if (HeHASH(entry) != hash) /* strings can't be equal */ |
994 | continue; | |
eb160463 | 995 | if (HeKLEN(entry) != (I32)klen) |
fde52b5c | 996 | continue; |
1c846c1f | 997 | if (HeKEY(entry) != key && memNE(HeKEY(entry),key,klen)) /* is this it? */ |
fde52b5c | 998 | continue; |
7a9669ca | 999 | if ((HeKFLAGS(entry) ^ masked_flags) & HVhek_UTF8) |
c3654f1a | 1000 | continue; |
8aacddc1 | 1001 | |
5d2b1485 NC |
1002 | if (hv == PL_strtab) { |
1003 | if (k_flags & HVhek_FREEKEY) | |
1004 | Safefree(key); | |
1005 | Perl_croak(aTHX_ S_strtab_error, "delete"); | |
1006 | } | |
1007 | ||
8aacddc1 | 1008 | /* if placeholder is here, it's already been deleted.... */ |
6136c704 AL |
1009 | if (HeVAL(entry) == &PL_sv_placeholder) { |
1010 | if (k_flags & HVhek_FREEKEY) | |
1011 | Safefree(key); | |
1012 | return NULL; | |
8aacddc1 | 1013 | } |
e5accad2 FC |
1014 | if (SvREADONLY(hv) && HeVAL(entry) && SvREADONLY(HeVAL(entry)) |
1015 | && !SvIsCOW(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 | 1028 | if (keysv) key = SvPV(keysv, klen); |
1f656fcf FC |
1029 | if (( |
1030 | (klen > 1 && key[klen-2] == ':' && key[klen-1] == ':') | |
1031 | || | |
1032 | (klen == 1 && key[0] == ':') | |
1033 | ) | |
e0a52395 | 1034 | && (klen != 6 || hv!=PL_defstash || memNE(key,"main::",6)) |
0290c710 | 1035 | && SvTYPE(gv) == SVt_PVGV && (stash = GvHV((GV *)gv)) |
0c3bb3c2 | 1036 | && HvENAME_get(stash)) { |
0290c710 FC |
1037 | /* A previous version of this code checked that the |
1038 | * GV was still in the symbol table by fetching the | |
1039 | * GV with its name. That is not necessary (and | |
1040 | * sometimes incorrect), as HvENAME cannot be set | |
1041 | * on hv if it is not in the symtab. */ | |
f3d2f32d | 1042 | mro_changes = 2; |
0c3bb3c2 FC |
1043 | /* Hang on to it for a bit. */ |
1044 | SvREFCNT_inc_simple_void_NN( | |
0290c710 | 1045 | sv_2mortal((SV *)gv) |
35759254 FC |
1046 | ); |
1047 | } | |
f3d2f32d FC |
1048 | else if (klen == 3 && strnEQ(key, "ISA", 3)) |
1049 | mro_changes = 1; | |
35759254 FC |
1050 | } |
1051 | ||
f50383f5 TH |
1052 | if (d_flags & G_DISCARD) { |
1053 | sv = HeVAL(entry); | |
70582212 | 1054 | HeVAL(entry) = &PL_sv_placeholder; |
f50383f5 TH |
1055 | if (sv) { |
1056 | /* deletion of method from stash */ | |
1057 | if (isGV(sv) && isGV_with_GP(sv) && GvCVu(sv) | |
1058 | && HvENAME_get(hv)) | |
1059 | mro_method_changed_in(hv); | |
1060 | SvREFCNT_dec(sv); | |
1061 | sv = NULL; | |
1062 | } | |
70582212 FC |
1063 | } |
1064 | else { | |
1065 | sv = sv_2mortal(HeVAL(entry)); | |
1066 | HeVAL(entry) = &PL_sv_placeholder; | |
1067 | } | |
8aacddc1 NIS |
1068 | |
1069 | /* | |
1070 | * If a restricted hash, rather than really deleting the entry, put | |
1071 | * a placeholder there. This marks the key as being "approved", so | |
1072 | * we can still access via not-really-existing key without raising | |
1073 | * an error. | |
1074 | */ | |
f50383f5 | 1075 | if (SvREADONLY(hv)) |
8aacddc1 NIS |
1076 | /* We'll be saving this slot, so the number of allocated keys |
1077 | * doesn't go down, but the number placeholders goes up */ | |
ca732855 | 1078 | HvPLACEHOLDERS(hv)++; |
f50383f5 | 1079 | else { |
a26e96df | 1080 | *oentry = HeNEXT(entry); |
b79f7545 | 1081 | if (SvOOK(hv) && entry == HvAUX(hv)->xhv_eiter /* HvEITER(hv) */) |
8aacddc1 | 1082 | HvLAZYDEL_on(hv); |
ae199939 TH |
1083 | else { |
1084 | if (SvOOK(hv) && HvLAZYDEL(hv) && | |
1085 | entry == HeNEXT(HvAUX(hv)->xhv_eiter)) | |
1086 | HeNEXT(HvAUX(hv)->xhv_eiter) = HeNEXT(entry); | |
8aacddc1 | 1087 | hv_free_ent(hv, entry); |
ae199939 | 1088 | } |
4c7185a0 | 1089 | xhv->xhv_keys--; /* HvTOTALKEYS(hv)-- */ |
574c8022 | 1090 | if (xhv->xhv_keys == 0) |
19692e8d | 1091 | HvHASKFLAGS_off(hv); |
8aacddc1 | 1092 | } |
0c3bb3c2 | 1093 | |
f3d2f32d FC |
1094 | if (mro_changes == 1) mro_isa_changed_in(hv); |
1095 | else if (mro_changes == 2) | |
afdbe55d | 1096 | mro_package_moved(NULL, stash, gv, 1); |
0c3bb3c2 | 1097 | |
79072805 LW |
1098 | return sv; |
1099 | } | |
8aacddc1 | 1100 | if (SvREADONLY(hv)) { |
d4c19fe8 | 1101 | hv_notallowed(k_flags, key, klen, |
c8cd6465 NC |
1102 | "Attempt to delete disallowed key '%"SVf"' from" |
1103 | " a restricted hash"); | |
8aacddc1 NIS |
1104 | } |
1105 | ||
19692e8d | 1106 | if (k_flags & HVhek_FREEKEY) |
f9a63242 | 1107 | Safefree(key); |
a0714e2c | 1108 | return NULL; |
79072805 LW |
1109 | } |
1110 | ||
76e3520e | 1111 | STATIC void |
cea2e8a9 | 1112 | S_hsplit(pTHX_ HV *hv) |
79072805 | 1113 | { |
97aff369 | 1114 | dVAR; |
1e05feb3 | 1115 | register XPVHV* const xhv = (XPVHV*)SvANY(hv); |
a3b680e6 | 1116 | const I32 oldsize = (I32) xhv->xhv_max+1; /* HvMAX(hv)+1 (sick) */ |
79072805 LW |
1117 | register I32 newsize = oldsize * 2; |
1118 | register I32 i; | |
7b2c381c | 1119 | char *a = (char*) HvARRAY(hv); |
72311751 | 1120 | register HE **aep; |
4b5190b5 NC |
1121 | int longest_chain = 0; |
1122 | int was_shared; | |
79072805 | 1123 | |
7918f24d NC |
1124 | PERL_ARGS_ASSERT_HSPLIT; |
1125 | ||
18026298 | 1126 | /*PerlIO_printf(PerlIO_stderr(), "hsplit called for %p which had %d\n", |
6c9570dc | 1127 | (void*)hv, (int) oldsize);*/ |
18026298 | 1128 | |
5d88ecd7 | 1129 | if (HvPLACEHOLDERS_get(hv) && !SvREADONLY(hv)) { |
18026298 NC |
1130 | /* Can make this clear any placeholders first for non-restricted hashes, |
1131 | even though Storable rebuilds restricted hashes by putting in all the | |
1132 | placeholders (first) before turning on the readonly flag, because | |
1133 | Storable always pre-splits the hash. */ | |
1134 | hv_clear_placeholders(hv); | |
1135 | } | |
1136 | ||
3280af22 | 1137 | PL_nomemok = TRUE; |
8d6dde3e | 1138 | #if defined(STRANGE_MALLOC) || defined(MYMALLOC) |
b79f7545 NC |
1139 | Renew(a, PERL_HV_ARRAY_ALLOC_BYTES(newsize) |
1140 | + (SvOOK(hv) ? sizeof(struct xpvhv_aux) : 0), char); | |
422a93e5 | 1141 | if (!a) { |
4a33f861 | 1142 | PL_nomemok = FALSE; |
422a93e5 GA |
1143 | return; |
1144 | } | |
b79f7545 | 1145 | if (SvOOK(hv)) { |
31f0e52e | 1146 | Move(&a[oldsize * sizeof(HE*)], &a[newsize * sizeof(HE*)], 1, struct xpvhv_aux); |
b79f7545 | 1147 | } |
4633a7c4 | 1148 | #else |
a02a5408 | 1149 | Newx(a, PERL_HV_ARRAY_ALLOC_BYTES(newsize) |
b79f7545 | 1150 | + (SvOOK(hv) ? sizeof(struct xpvhv_aux) : 0), char); |
422a93e5 | 1151 | if (!a) { |
3280af22 | 1152 | PL_nomemok = FALSE; |
422a93e5 GA |
1153 | return; |
1154 | } | |
7b2c381c | 1155 | Copy(HvARRAY(hv), a, oldsize * sizeof(HE*), char); |
b79f7545 NC |
1156 | if (SvOOK(hv)) { |
1157 | Copy(HvAUX(hv), &a[newsize * sizeof(HE*)], 1, struct xpvhv_aux); | |
1158 | } | |
9a87bd09 | 1159 | Safefree(HvARRAY(hv)); |
4633a7c4 LW |
1160 | #endif |
1161 | ||
3280af22 | 1162 | PL_nomemok = FALSE; |
72311751 | 1163 | Zero(&a[oldsize * sizeof(HE*)], (newsize-oldsize) * sizeof(HE*), char); /* zero 2nd half*/ |
cbec9347 | 1164 | xhv->xhv_max = --newsize; /* HvMAX(hv) = --newsize */ |
7b2c381c | 1165 | HvARRAY(hv) = (HE**) a; |
72311751 | 1166 | aep = (HE**)a; |
79072805 | 1167 | |
72311751 | 1168 | for (i=0; i<oldsize; i++,aep++) { |
4b5190b5 NC |
1169 | int left_length = 0; |
1170 | int right_length = 0; | |
a50a3493 NC |
1171 | HE **oentry = aep; |
1172 | HE *entry = *aep; | |
a3b680e6 | 1173 | register HE **bep; |
4b5190b5 | 1174 | |
a50a3493 | 1175 | if (!entry) /* non-existent */ |
79072805 | 1176 | continue; |
72311751 | 1177 | bep = aep+oldsize; |
4c9d89c5 | 1178 | do { |
eb160463 | 1179 | if ((HeHASH(entry) & newsize) != (U32)i) { |
fde52b5c | 1180 | *oentry = HeNEXT(entry); |
72311751 | 1181 | HeNEXT(entry) = *bep; |
72311751 | 1182 | *bep = entry; |
4b5190b5 | 1183 | right_length++; |
79072805 | 1184 | } |
4b5190b5 | 1185 | else { |
fde52b5c | 1186 | oentry = &HeNEXT(entry); |
4b5190b5 NC |
1187 | left_length++; |
1188 | } | |
4c9d89c5 NC |
1189 | entry = *oentry; |
1190 | } while (entry); | |
4b5190b5 NC |
1191 | /* I think we don't actually need to keep track of the longest length, |
1192 | merely flag if anything is too long. But for the moment while | |
1193 | developing this code I'll track it. */ | |
1194 | if (left_length > longest_chain) | |
1195 | longest_chain = left_length; | |
1196 | if (right_length > longest_chain) | |
1197 | longest_chain = right_length; | |
1198 | } | |
1199 | ||
1200 | ||
1201 | /* Pick your policy for "hashing isn't working" here: */ | |
fdcd69b6 | 1202 | if (longest_chain <= HV_MAX_LENGTH_BEFORE_SPLIT /* split worked? */ |
4b5190b5 NC |
1203 | || HvREHASH(hv)) { |
1204 | return; | |
79072805 | 1205 | } |
4b5190b5 NC |
1206 | |
1207 | if (hv == PL_strtab) { | |
1208 | /* Urg. Someone is doing something nasty to the string table. | |
1209 | Can't win. */ | |
1210 | return; | |
1211 | } | |
1212 | ||
1213 | /* Awooga. Awooga. Pathological data. */ | |
6c9570dc | 1214 | /*PerlIO_printf(PerlIO_stderr(), "%p %d of %d with %d/%d buckets\n", (void*)hv, |
4b5190b5 NC |
1215 | longest_chain, HvTOTALKEYS(hv), HvFILL(hv), 1+HvMAX(hv));*/ |
1216 | ||
1217 | ++newsize; | |
a02a5408 | 1218 | Newxz(a, PERL_HV_ARRAY_ALLOC_BYTES(newsize) |
b79f7545 NC |
1219 | + (SvOOK(hv) ? sizeof(struct xpvhv_aux) : 0), char); |
1220 | if (SvOOK(hv)) { | |
1221 | Copy(HvAUX(hv), &a[newsize * sizeof(HE*)], 1, struct xpvhv_aux); | |
1222 | } | |
1223 | ||
4b5190b5 NC |
1224 | was_shared = HvSHAREKEYS(hv); |
1225 | ||
4b5190b5 NC |
1226 | HvSHAREKEYS_off(hv); |
1227 | HvREHASH_on(hv); | |
1228 | ||
7b2c381c | 1229 | aep = HvARRAY(hv); |
4b5190b5 NC |
1230 | |
1231 | for (i=0; i<newsize; i++,aep++) { | |
a3b680e6 | 1232 | register HE *entry = *aep; |
4b5190b5 NC |
1233 | while (entry) { |
1234 | /* We're going to trash this HE's next pointer when we chain it | |
1235 | into the new hash below, so store where we go next. */ | |
9d4ba2ae | 1236 | HE * const next = HeNEXT(entry); |
4b5190b5 | 1237 | UV hash; |
a3b680e6 | 1238 | HE **bep; |
4b5190b5 NC |
1239 | |
1240 | /* Rehash it */ | |
1241 | PERL_HASH_INTERNAL(hash, HeKEY(entry), HeKLEN(entry)); | |
1242 | ||
1243 | if (was_shared) { | |
1244 | /* Unshare it. */ | |
aec46f14 | 1245 | HEK * const new_hek |
4b5190b5 NC |
1246 | = save_hek_flags(HeKEY(entry), HeKLEN(entry), |
1247 | hash, HeKFLAGS(entry)); | |
1248 | unshare_hek (HeKEY_hek(entry)); | |
1249 | HeKEY_hek(entry) = new_hek; | |
1250 | } else { | |
1251 | /* Not shared, so simply write the new hash in. */ | |
1252 | HeHASH(entry) = hash; | |
1253 | } | |
1254 | /*PerlIO_printf(PerlIO_stderr(), "%d ", HeKFLAGS(entry));*/ | |
1255 | HEK_REHASH_on(HeKEY_hek(entry)); | |
1256 | /*PerlIO_printf(PerlIO_stderr(), "%d\n", HeKFLAGS(entry));*/ | |
1257 | ||
1258 | /* Copy oentry to the correct new chain. */ | |
1259 | bep = ((HE**)a) + (hash & (I32) xhv->xhv_max); | |
4b5190b5 NC |
1260 | HeNEXT(entry) = *bep; |
1261 | *bep = entry; | |
1262 | ||
1263 | entry = next; | |
1264 | } | |
1265 | } | |
7b2c381c NC |
1266 | Safefree (HvARRAY(hv)); |
1267 | HvARRAY(hv) = (HE **)a; | |
79072805 LW |
1268 | } |
1269 | ||
72940dca | 1270 | void |
864dbfa3 | 1271 | Perl_hv_ksplit(pTHX_ HV *hv, IV newmax) |
72940dca | 1272 | { |
97aff369 | 1273 | dVAR; |
cbec9347 | 1274 | register XPVHV* xhv = (XPVHV*)SvANY(hv); |
a3b680e6 | 1275 | const I32 oldsize = (I32) xhv->xhv_max+1; /* HvMAX(hv)+1 (sick) */ |
72940dca | 1276 | register I32 newsize; |
1277 | register I32 i; | |
72311751 GS |
1278 | register char *a; |
1279 | register HE **aep; | |
72940dca | 1280 | |
7918f24d NC |
1281 | PERL_ARGS_ASSERT_HV_KSPLIT; |
1282 | ||
72940dca | 1283 | newsize = (I32) newmax; /* possible truncation here */ |
1284 | if (newsize != newmax || newmax <= oldsize) | |
1285 | return; | |
1286 | while ((newsize & (1 + ~newsize)) != newsize) { | |
1287 | newsize &= ~(newsize & (1 + ~newsize)); /* get proper power of 2 */ | |
1288 | } | |
1289 | if (newsize < newmax) | |
1290 | newsize *= 2; | |
1291 | if (newsize < newmax) | |
1292 | return; /* overflow detection */ | |
1293 | ||
7b2c381c | 1294 | a = (char *) HvARRAY(hv); |
72940dca | 1295 | if (a) { |
3280af22 | 1296 | PL_nomemok = TRUE; |
8d6dde3e | 1297 | #if defined(STRANGE_MALLOC) || defined(MYMALLOC) |
b79f7545 NC |
1298 | Renew(a, PERL_HV_ARRAY_ALLOC_BYTES(newsize) |
1299 | + (SvOOK(hv) ? sizeof(struct xpvhv_aux) : 0), char); | |
8aacddc1 | 1300 | if (!a) { |
4a33f861 | 1301 | PL_nomemok = FALSE; |
422a93e5 GA |
1302 | return; |
1303 | } | |
b79f7545 | 1304 | if (SvOOK(hv)) { |
7a9b70e9 | 1305 | Copy(&a[oldsize * sizeof(HE*)], &a[newsize * sizeof(HE*)], 1, struct xpvhv_aux); |
b79f7545 | 1306 | } |
72940dca | 1307 | #else |
a02a5408 | 1308 | Newx(a, PERL_HV_ARRAY_ALLOC_BYTES(newsize) |
b79f7545 | 1309 | + (SvOOK(hv) ? sizeof(struct xpvhv_aux) : 0), char); |
8aacddc1 | 1310 | if (!a) { |
3280af22 | 1311 | PL_nomemok = FALSE; |
422a93e5 GA |
1312 | return; |
1313 | } | |
7b2c381c | 1314 | Copy(HvARRAY(hv), a, oldsize * sizeof(HE*), char); |
b79f7545 NC |
1315 | if (SvOOK(hv)) { |
1316 | Copy(HvAUX(hv), &a[newsize * sizeof(HE*)], 1, struct xpvhv_aux); | |
1317 | } | |
9a87bd09 | 1318 | Safefree(HvARRAY(hv)); |
72940dca | 1319 | #endif |
3280af22 | 1320 | PL_nomemok = FALSE; |
72311751 | 1321 | Zero(&a[oldsize * sizeof(HE*)], (newsize-oldsize) * sizeof(HE*), char); /* zero 2nd half*/ |
72940dca | 1322 | } |
1323 | else { | |
a02a5408 | 1324 | Newxz(a, PERL_HV_ARRAY_ALLOC_BYTES(newsize), char); |
72940dca | 1325 | } |
cbec9347 | 1326 | xhv->xhv_max = --newsize; /* HvMAX(hv) = --newsize */ |
7b2c381c | 1327 | HvARRAY(hv) = (HE **) a; |
f4431c56 | 1328 | if (!xhv->xhv_keys /* !HvTOTALKEYS(hv) */) /* skip rest if no entries */ |
72940dca | 1329 | return; |
1330 | ||
72311751 GS |
1331 | aep = (HE**)a; |
1332 | for (i=0; i<oldsize; i++,aep++) { | |
a50a3493 NC |
1333 | HE **oentry = aep; |
1334 | HE *entry = *aep; | |
1335 | ||
1336 | if (!entry) /* non-existent */ | |
72940dca | 1337 | continue; |
4c9d89c5 | 1338 | do { |
6136c704 AL |
1339 | register I32 j = (HeHASH(entry) & newsize); |
1340 | ||
1341 | if (j != i) { | |
72940dca | 1342 | j -= i; |
1343 | *oentry = HeNEXT(entry); | |
4d0fbddd | 1344 | HeNEXT(entry) = aep[j]; |
72311751 | 1345 | aep[j] = entry; |
72940dca | 1346 | } |
1347 | else | |
1348 | oentry = &HeNEXT(entry); | |
4c9d89c5 NC |
1349 | entry = *oentry; |
1350 | } while (entry); | |
72940dca | 1351 | } |
1352 | } | |
1353 | ||
b3ac6de7 | 1354 | HV * |
864dbfa3 | 1355 | Perl_newHVhv(pTHX_ HV *ohv) |
b3ac6de7 | 1356 | { |
749123ff | 1357 | dVAR; |
9d4ba2ae | 1358 | HV * const hv = newHV(); |
f4431c56 | 1359 | STRLEN hv_max; |
4beac62f | 1360 | |
f4431c56 | 1361 | if (!ohv || !HvTOTALKEYS(ohv)) |
4beac62f | 1362 | return hv; |
4beac62f | 1363 | hv_max = HvMAX(ohv); |
b3ac6de7 | 1364 | |
ad64d0ec | 1365 | if (!SvMAGICAL((const SV *)ohv)) { |
b56ba0bf | 1366 | /* It's an ordinary hash, so copy it fast. AMS 20010804 */ |
eb160463 | 1367 | STRLEN i; |
a3b680e6 | 1368 | const bool shared = !!HvSHAREKEYS(ohv); |
aec46f14 | 1369 | HE **ents, ** const oents = (HE **)HvARRAY(ohv); |
ff875642 | 1370 | char *a; |
a02a5408 | 1371 | Newx(a, PERL_HV_ARRAY_ALLOC_BYTES(hv_max+1), char); |
ff875642 | 1372 | ents = (HE**)a; |
b56ba0bf AMS |
1373 | |
1374 | /* In each bucket... */ | |
1375 | for (i = 0; i <= hv_max; i++) { | |
6136c704 | 1376 | HE *prev = NULL; |
aec46f14 | 1377 | HE *oent = oents[i]; |
b56ba0bf AMS |
1378 | |
1379 | if (!oent) { | |
1380 | ents[i] = NULL; | |
1381 | continue; | |
1382 | } | |
1383 | ||
1384 | /* Copy the linked list of entries. */ | |
aec46f14 | 1385 | for (; oent; oent = HeNEXT(oent)) { |
a3b680e6 AL |
1386 | const U32 hash = HeHASH(oent); |
1387 | const char * const key = HeKEY(oent); | |
1388 | const STRLEN len = HeKLEN(oent); | |
1389 | const int flags = HeKFLAGS(oent); | |
6136c704 | 1390 | HE * const ent = new_HE(); |
c3acb9e0 | 1391 | SV *const val = HeVAL(oent); |
b56ba0bf | 1392 | |
c3acb9e0 | 1393 | HeVAL(ent) = SvIMMORTAL(val) ? val : newSVsv(val); |
19692e8d | 1394 | HeKEY_hek(ent) |
6e838c70 | 1395 | = shared ? share_hek_flags(key, len, hash, flags) |
19692e8d | 1396 | : save_hek_flags(key, len, hash, flags); |
b56ba0bf AMS |
1397 | if (prev) |
1398 | HeNEXT(prev) = ent; | |
1399 | else | |
1400 | ents[i] = ent; | |
1401 | prev = ent; | |
1402 | HeNEXT(ent) = NULL; | |
1403 | } | |
1404 | } | |
1405 | ||
1406 | HvMAX(hv) = hv_max; | |
8aacddc1 | 1407 | HvTOTALKEYS(hv) = HvTOTALKEYS(ohv); |
b56ba0bf | 1408 | HvARRAY(hv) = ents; |
aec46f14 | 1409 | } /* not magical */ |
b56ba0bf AMS |
1410 | else { |
1411 | /* Iterate over ohv, copying keys and values one at a time. */ | |
b3ac6de7 | 1412 | HE *entry; |
bfcb3514 NC |
1413 | const I32 riter = HvRITER_get(ohv); |
1414 | HE * const eiter = HvEITER_get(ohv); | |
f4431c56 | 1415 | STRLEN hv_fill = HvFILL(ohv); |
b56ba0bf AMS |
1416 | |
1417 | /* Can we use fewer buckets? (hv_max is always 2^n-1) */ | |
1418 | while (hv_max && hv_max + 1 >= hv_fill * 2) | |
1419 | hv_max = hv_max / 2; | |
1420 | HvMAX(hv) = hv_max; | |
1421 | ||
4a76a316 | 1422 | hv_iterinit(ohv); |
e16e2ff8 | 1423 | while ((entry = hv_iternext_flags(ohv, 0))) { |
c3acb9e0 | 1424 | SV *const val = HeVAL(entry); |
04fe65b0 | 1425 | (void)hv_store_flags(hv, HeKEY(entry), HeKLEN(entry), |
c3acb9e0 NC |
1426 | SvIMMORTAL(val) ? val : newSVsv(val), |
1427 | HeHASH(entry), HeKFLAGS(entry)); | |
b3ac6de7 | 1428 | } |
bfcb3514 NC |
1429 | HvRITER_set(ohv, riter); |
1430 | HvEITER_set(ohv, eiter); | |
b3ac6de7 | 1431 | } |
1c846c1f | 1432 | |
b3ac6de7 IZ |
1433 | return hv; |
1434 | } | |
1435 | ||
defdfed5 Z |
1436 | /* |
1437 | =for apidoc Am|HV *|hv_copy_hints_hv|HV *ohv | |
1438 | ||
1439 | A specialised version of L</newHVhv> for copying C<%^H>. I<ohv> must be | |
1440 | a pointer to a hash (which may have C<%^H> magic, but should be generally | |
1441 | non-magical), or C<NULL> (interpreted as an empty hash). The content | |
1442 | of I<ohv> is copied to a new hash, which has the C<%^H>-specific magic | |
1443 | added to it. A pointer to the new hash is returned. | |
1444 | ||
1445 | =cut | |
1446 | */ | |
1447 | ||
5b9c0671 NC |
1448 | HV * |
1449 | Perl_hv_copy_hints_hv(pTHX_ HV *const ohv) | |
1450 | { | |
1451 | HV * const hv = newHV(); | |
5b9c0671 | 1452 | |
f4431c56 | 1453 | if (ohv && HvTOTALKEYS(ohv)) { |
5b9c0671 | 1454 | STRLEN hv_max = HvMAX(ohv); |
f4431c56 | 1455 | STRLEN hv_fill = HvFILL(ohv); |
5b9c0671 NC |
1456 | HE *entry; |
1457 | const I32 riter = HvRITER_get(ohv); | |
1458 | HE * const eiter = HvEITER_get(ohv); | |
1459 | ||
1460 | while (hv_max && hv_max + 1 >= hv_fill * 2) | |
1461 | hv_max = hv_max / 2; | |
1462 | HvMAX(hv) = hv_max; | |
1463 | ||
1464 | hv_iterinit(ohv); | |
1465 | while ((entry = hv_iternext_flags(ohv, 0))) { | |
1466 | SV *const sv = newSVsv(HeVAL(entry)); | |
7ef9d42c FC |
1467 | SV *heksv = HeSVKEY(entry); |
1468 | if (!heksv && sv) heksv = newSVhek(HeKEY_hek(entry)); | |
95cf2368 | 1469 | if (sv) sv_magic(sv, NULL, PERL_MAGIC_hintselem, |
e3b1b6b1 | 1470 | (char *)heksv, HEf_SVKEY); |
7ef9d42c FC |
1471 | if (heksv == HeSVKEY(entry)) |
1472 | (void)hv_store_ent(hv, heksv, sv, 0); | |
1473 | else { | |
1474 | (void)hv_common(hv, heksv, HeKEY(entry), HeKLEN(entry), | |
1475 | HeKFLAGS(entry), HV_FETCH_ISSTORE|HV_FETCH_JUST_SV, sv, HeHASH(entry)); | |
1476 | SvREFCNT_dec(heksv); | |
1477 | } | |
5b9c0671 NC |
1478 | } |
1479 | HvRITER_set(ohv, riter); | |
1480 | HvEITER_set(ohv, eiter); | |
1481 | } | |
1482 | hv_magic(hv, NULL, PERL_MAGIC_hints); | |
1483 | return hv; | |
1484 | } | |
1485 | ||
e0171a1a DM |
1486 | /* like hv_free_ent, but returns the SV rather than freeing it */ |
1487 | STATIC SV* | |
1488 | S_hv_free_ent_ret(pTHX_ HV *hv, register HE *entry) | |
79072805 | 1489 | { |
97aff369 | 1490 | dVAR; |
16bdeea2 GS |
1491 | SV *val; |
1492 | ||
e0171a1a | 1493 | PERL_ARGS_ASSERT_HV_FREE_ENT_RET; |
7918f24d | 1494 | |
68dc0745 | 1495 | if (!entry) |
e0171a1a | 1496 | return NULL; |
16bdeea2 | 1497 | val = HeVAL(entry); |
00169e2c | 1498 | if (val && isGV(val) && isGV_with_GP(val) && GvCVu(val) && HvENAME(hv)) |
803f2748 | 1499 | mro_method_changed_in(hv); /* deletion of method from stash */ |
68dc0745 | 1500 | if (HeKLEN(entry) == HEf_SVKEY) { |
1501 | SvREFCNT_dec(HeKEY_sv(entry)); | |
8aacddc1 | 1502 | Safefree(HeKEY_hek(entry)); |
44a8e56a | 1503 | } |
1504 | else if (HvSHAREKEYS(hv)) | |
68dc0745 | 1505 | unshare_hek(HeKEY_hek(entry)); |
fde52b5c | 1506 | else |
68dc0745 | 1507 | Safefree(HeKEY_hek(entry)); |
d33b2eba | 1508 | del_HE(entry); |
e0171a1a DM |
1509 | return val; |
1510 | } | |
1511 | ||
1512 | ||
1513 | void | |
1514 | Perl_hv_free_ent(pTHX_ HV *hv, register HE *entry) | |
1515 | { | |
1516 | dVAR; | |
1517 | SV *val; | |
1518 | ||
1519 | PERL_ARGS_ASSERT_HV_FREE_ENT; | |
1520 | ||
1521 | if (!entry) | |
1522 | return; | |
1523 | val = hv_free_ent_ret(hv, entry); | |
272e8453 | 1524 | SvREFCNT_dec(val); |
79072805 LW |
1525 | } |
1526 | ||
f1c32fec | 1527 | |
79072805 | 1528 | void |
864dbfa3 | 1529 | Perl_hv_delayfree_ent(pTHX_ HV *hv, register HE *entry) |
79072805 | 1530 | { |
97aff369 | 1531 | dVAR; |
7918f24d NC |
1532 | |
1533 | PERL_ARGS_ASSERT_HV_DELAYFREE_ENT; | |
1534 | ||
68dc0745 | 1535 | if (!entry) |
79072805 | 1536 | return; |
bc4947fc NC |
1537 | /* SvREFCNT_inc to counter the SvREFCNT_dec in hv_free_ent */ |
1538 | sv_2mortal(SvREFCNT_inc(HeVAL(entry))); /* free between statements */ | |
68dc0745 | 1539 | if (HeKLEN(entry) == HEf_SVKEY) { |
bc4947fc | 1540 | sv_2mortal(SvREFCNT_inc(HeKEY_sv(entry))); |
44a8e56a | 1541 | } |
bc4947fc | 1542 | hv_free_ent(hv, entry); |
79072805 LW |
1543 | } |
1544 | ||
954c1994 GS |
1545 | /* |
1546 | =for apidoc hv_clear | |
1547 | ||
c2217cd3 DM |
1548 | Frees the all the elements of a hash, leaving it empty. |
1549 | The XS equivalent of %hash = (). See also L</hv_undef>. | |
954c1994 GS |
1550 | |
1551 | =cut | |
1552 | */ | |
1553 | ||
79072805 | 1554 | void |
864dbfa3 | 1555 | Perl_hv_clear(pTHX_ HV *hv) |
79072805 | 1556 | { |
27da23d5 | 1557 | dVAR; |
cbec9347 | 1558 | register XPVHV* xhv; |
79072805 LW |
1559 | if (!hv) |
1560 | return; | |
49293501 | 1561 | |
ecae49c0 NC |
1562 | DEBUG_A(Perl_hv_assert(aTHX_ hv)); |
1563 | ||
34c3c4e3 DM |
1564 | xhv = (XPVHV*)SvANY(hv); |
1565 | ||
7b2c381c | 1566 | if (SvREADONLY(hv) && HvARRAY(hv) != NULL) { |
34c3c4e3 | 1567 | /* restricted hash: convert all keys to placeholders */ |
b464bac0 AL |
1568 | STRLEN i; |
1569 | for (i = 0; i <= xhv->xhv_max; i++) { | |
7b2c381c | 1570 | HE *entry = (HvARRAY(hv))[i]; |
3a676441 JH |
1571 | for (; entry; entry = HeNEXT(entry)) { |
1572 | /* not already placeholder */ | |
7996736c | 1573 | if (HeVAL(entry) != &PL_sv_placeholder) { |
fb2352eb FC |
1574 | if (HeVAL(entry) && SvREADONLY(HeVAL(entry)) |
1575 | && !SvIsCOW(HeVAL(entry))) { | |
6136c704 | 1576 | SV* const keysv = hv_iterkeysv(entry); |
3a676441 | 1577 | Perl_croak(aTHX_ |
95b63a38 JH |
1578 | "Attempt to delete readonly key '%"SVf"' from a restricted hash", |
1579 | (void*)keysv); | |
3a676441 JH |
1580 | } |
1581 | SvREFCNT_dec(HeVAL(entry)); | |
7996736c | 1582 | HeVAL(entry) = &PL_sv_placeholder; |
ca732855 | 1583 | HvPLACEHOLDERS(hv)++; |
3a676441 | 1584 | } |
34c3c4e3 DM |
1585 | } |
1586 | } | |
49293501 | 1587 | } |
afbbf215 DM |
1588 | else { |
1589 | hfreeentries(hv); | |
1590 | HvPLACEHOLDERS_set(hv, 0); | |
49293501 | 1591 | |
afbbf215 DM |
1592 | if (SvRMAGICAL(hv)) |
1593 | mg_clear(MUTABLE_SV(hv)); | |
574c8022 | 1594 | |
afbbf215 DM |
1595 | HvHASKFLAGS_off(hv); |
1596 | HvREHASH_off(hv); | |
1597 | } | |
b79f7545 | 1598 | if (SvOOK(hv)) { |
00169e2c | 1599 | if(HvENAME_get(hv)) |
dd69841b | 1600 | mro_isa_changed_in(hv); |
bfcb3514 NC |
1601 | HvEITER_set(hv, NULL); |
1602 | } | |
79072805 LW |
1603 | } |
1604 | ||
3540d4ce AB |
1605 | /* |
1606 | =for apidoc hv_clear_placeholders | |
1607 | ||
1608 | Clears any placeholders from a hash. If a restricted hash has any of its keys | |
1609 | marked as readonly and the key is subsequently deleted, the key is not actually | |
1610 | deleted but is marked by assigning it a value of &PL_sv_placeholder. This tags | |
1611 | it so it will be ignored by future operations such as iterating over the hash, | |
4cdaeff7 | 1612 | but will still allow the hash to have a value reassigned to the key at some |
3540d4ce AB |
1613 | future point. This function clears any such placeholder keys from the hash. |
1614 | See Hash::Util::lock_keys() for an example of its use. | |
1615 | ||
1616 | =cut | |
1617 | */ | |
1618 | ||
1619 | void | |
1620 | Perl_hv_clear_placeholders(pTHX_ HV *hv) | |
1621 | { | |
27da23d5 | 1622 | dVAR; |
b3ca2e83 NC |
1623 | const U32 items = (U32)HvPLACEHOLDERS_get(hv); |
1624 | ||
7918f24d NC |
1625 | PERL_ARGS_ASSERT_HV_CLEAR_PLACEHOLDERS; |
1626 | ||
b3ca2e83 NC |
1627 | if (items) |
1628 | clear_placeholders(hv, items); | |
1629 | } | |
1630 | ||
1631 | static void | |
1632 | S_clear_placeholders(pTHX_ HV *hv, U32 items) | |
1633 | { | |
1634 | dVAR; | |
b464bac0 | 1635 | I32 i; |
d3677389 | 1636 | |
7918f24d NC |
1637 | PERL_ARGS_ASSERT_CLEAR_PLACEHOLDERS; |
1638 | ||
d3677389 NC |
1639 | if (items == 0) |
1640 | return; | |
1641 | ||
b464bac0 | 1642 | i = HvMAX(hv); |
d3677389 NC |
1643 | do { |
1644 | /* Loop down the linked list heads */ | |
d3677389 | 1645 | HE **oentry = &(HvARRAY(hv))[i]; |
cf6db12b | 1646 | HE *entry; |
d3677389 | 1647 | |
cf6db12b | 1648 | while ((entry = *oentry)) { |
d3677389 NC |
1649 | if (HeVAL(entry) == &PL_sv_placeholder) { |
1650 | *oentry = HeNEXT(entry); | |
2e58978b | 1651 | if (entry == HvEITER_get(hv)) |
d3677389 | 1652 | HvLAZYDEL_on(hv); |
ae199939 TH |
1653 | else { |
1654 | if (SvOOK(hv) && HvLAZYDEL(hv) && | |
1655 | entry == HeNEXT(HvAUX(hv)->xhv_eiter)) | |
1656 | HeNEXT(HvAUX(hv)->xhv_eiter) = HeNEXT(entry); | |
d3677389 | 1657 | hv_free_ent(hv, entry); |
ae199939 | 1658 | } |
d3677389 NC |
1659 | |
1660 | if (--items == 0) { | |
1661 | /* Finished. */ | |
5d88ecd7 | 1662 | HvTOTALKEYS(hv) -= (IV)HvPLACEHOLDERS_get(hv); |
1b95d04f | 1663 | if (HvUSEDKEYS(hv) == 0) |
d3677389 | 1664 | HvHASKFLAGS_off(hv); |
5d88ecd7 | 1665 | HvPLACEHOLDERS_set(hv, 0); |
d3677389 NC |
1666 | return; |
1667 | } | |
213ce8b3 NC |
1668 | } else { |
1669 | oentry = &HeNEXT(entry); | |
d3677389 NC |
1670 | } |
1671 | } | |
1672 | } while (--i >= 0); | |
1673 | /* You can't get here, hence assertion should always fail. */ | |
1674 | assert (items == 0); | |
1675 | assert (0); | |
3540d4ce AB |
1676 | } |
1677 | ||
76e3520e | 1678 | STATIC void |
cea2e8a9 | 1679 | S_hfreeentries(pTHX_ HV *hv) |
79072805 | 1680 | { |
e0171a1a | 1681 | STRLEN index = 0; |
7d6175ef | 1682 | XPVHV * const xhv = (XPVHV*)SvANY(hv); |
6d1c68e6 | 1683 | SV *sv; |
3abe233e | 1684 | |
7918f24d NC |
1685 | PERL_ARGS_ASSERT_HFREEENTRIES; |
1686 | ||
6d1c68e6 FC |
1687 | while ((sv = Perl_hfree_next_entry(aTHX_ hv, &index))||xhv->xhv_keys) { |
1688 | SvREFCNT_dec(sv); | |
e0171a1a DM |
1689 | } |
1690 | } | |
23976bdd | 1691 | |
b79f7545 | 1692 | |
e0171a1a DM |
1693 | /* hfree_next_entry() |
1694 | * For use only by S_hfreeentries() and sv_clear(). | |
1695 | * Delete the next available HE from hv and return the associated SV. | |
7d6175ef FC |
1696 | * Returns null on empty hash. Nevertheless null is not a reliable |
1697 | * indicator that the hash is empty, as the deleted entry may have a | |
1698 | * null value. | |
e0171a1a DM |
1699 | * indexp is a pointer to the current index into HvARRAY. The index should |
1700 | * initially be set to 0. hfree_next_entry() may update it. */ | |
1701 | ||
1702 | SV* | |
1703 | Perl_hfree_next_entry(pTHX_ HV *hv, STRLEN *indexp) | |
1704 | { | |
1705 | struct xpvhv_aux *iter; | |
1706 | HE *entry; | |
1707 | HE ** array; | |
1708 | #ifdef DEBUGGING | |
1709 | STRLEN orig_index = *indexp; | |
1710 | #endif | |
1711 | ||
1712 | PERL_ARGS_ASSERT_HFREE_NEXT_ENTRY; | |
1713 | ||
e0171a1a DM |
1714 | if (SvOOK(hv) && ((iter = HvAUX(hv))) |
1715 | && ((entry = iter->xhv_eiter)) ) | |
1716 | { | |
1717 | /* the iterator may get resurrected after each | |
1718 | * destructor call, so check each time */ | |
1719 | if (entry && HvLAZYDEL(hv)) { /* was deleted earlier? */ | |
1720 | HvLAZYDEL_off(hv); | |
922090f4 DM |
1721 | hv_free_ent(hv, entry); |
1722 | /* warning: at this point HvARRAY may have been | |
1723 | * re-allocated, HvMAX changed etc */ | |
23976bdd | 1724 | } |
e0171a1a DM |
1725 | iter->xhv_riter = -1; /* HvRITER(hv) = -1 */ |
1726 | iter->xhv_eiter = NULL; /* HvEITER(hv) = NULL */ | |
1727 | } | |
1728 | ||
00a1a643 DM |
1729 | if (!((XPVHV*)SvANY(hv))->xhv_keys) |
1730 | return NULL; | |
1731 | ||
e0171a1a DM |
1732 | array = HvARRAY(hv); |
1733 | assert(array); | |
1734 | while ( ! ((entry = array[*indexp])) ) { | |
1735 | if ((*indexp)++ >= HvMAX(hv)) | |
1736 | *indexp = 0; | |
1737 | assert(*indexp != orig_index); | |
1738 | } | |
1739 | array[*indexp] = HeNEXT(entry); | |
1740 | ((XPVHV*) SvANY(hv))->xhv_keys--; | |
1741 | ||
1742 | if ( PL_phase != PERL_PHASE_DESTRUCT && HvENAME(hv) | |
1743 | && HeVAL(entry) && isGV(HeVAL(entry)) | |
1744 | && GvHV(HeVAL(entry)) && HvENAME(GvHV(HeVAL(entry))) | |
1745 | ) { | |
1746 | STRLEN klen; | |
1747 | const char * const key = HePV(entry,klen); | |
1748 | if ((klen > 1 && key[klen-1]==':' && key[klen-2]==':') | |
1749 | || (klen == 1 && key[0] == ':')) { | |
1750 | mro_package_moved( | |
1751 | NULL, GvHV(HeVAL(entry)), | |
1752 | (GV *)HeVAL(entry), 0 | |
1753 | ); | |
1754 | } | |
1755 | } | |
1756 | return hv_free_ent_ret(hv, entry); | |
79072805 LW |
1757 | } |
1758 | ||
e0171a1a | 1759 | |
954c1994 GS |
1760 | /* |
1761 | =for apidoc hv_undef | |
1762 | ||
c2217cd3 DM |
1763 | Undefines the hash. The XS equivalent of undef(%hash). |
1764 | ||
1765 | As well as freeing all the elements of the hash (like hv_clear()), this | |
1766 | also frees any auxiliary data and storage associated with the hash. | |
1767 | See also L</hv_clear>. | |
954c1994 GS |
1768 | |
1769 | =cut | |
1770 | */ | |
1771 | ||
79072805 | 1772 | void |
8581adba | 1773 | Perl_hv_undef_flags(pTHX_ HV *hv, U32 flags) |
79072805 | 1774 | { |
97aff369 | 1775 | dVAR; |
cbec9347 | 1776 | register XPVHV* xhv; |
bfcb3514 | 1777 | const char *name; |
86f55936 | 1778 | |
79072805 LW |
1779 | if (!hv) |
1780 | return; | |
ecae49c0 | 1781 | DEBUG_A(Perl_hv_assert(aTHX_ hv)); |
cbec9347 | 1782 | xhv = (XPVHV*)SvANY(hv); |
dd69841b | 1783 | |
745edda6 FC |
1784 | /* The name must be deleted before the call to hfreeeeentries so that |
1785 | CVs are anonymised properly. But the effective name must be pre- | |
1786 | served until after that call (and only deleted afterwards if the | |
1787 | call originated from sv_clear). For stashes with one name that is | |
1788 | both the canonical name and the effective name, hv_name_set has to | |
1789 | allocate an array for storing the effective name. We can skip that | |
1790 | during global destruction, as it does not matter where the CVs point | |
1791 | if they will be freed anyway. */ | |
104d7b69 DM |
1792 | /* note that the code following prior to hfreeentries is duplicated |
1793 | * in sv_clear(), and changes here should be done there too */ | |
745edda6 | 1794 | if (PL_phase != PERL_PHASE_DESTRUCT && (name = HvNAME(hv))) { |
04fe65b0 | 1795 | if (PL_stashcache) |
4643eb69 BF |
1796 | (void)hv_delete(PL_stashcache, name, |
1797 | HEK_UTF8(HvNAME_HEK(hv)) ? -HvNAMELEN_get(hv) : HvNAMELEN_get(hv), | |
1798 | G_DISCARD | |
1799 | ); | |
bd61b366 | 1800 | hv_name_set(hv, NULL, 0, 0); |
85e6fe83 | 1801 | } |
2d0d1ecc | 1802 | hfreeentries(hv); |
47f1cf77 FC |
1803 | if (SvOOK(hv)) { |
1804 | struct xpvhv_aux * const aux = HvAUX(hv); | |
1805 | struct mro_meta *meta; | |
745edda6 FC |
1806 | |
1807 | if ((name = HvENAME_get(hv))) { | |
5f243b5f | 1808 | if (PL_phase != PERL_PHASE_DESTRUCT) |
745edda6 | 1809 | mro_isa_changed_in(hv); |
745edda6 FC |
1810 | if (PL_stashcache) |
1811 | (void)hv_delete( | |
4643eb69 BF |
1812 | PL_stashcache, name, |
1813 | HEK_UTF8(HvENAME_HEK(hv)) ? -HvENAMELEN_get(hv) : HvENAMELEN_get(hv), | |
1814 | G_DISCARD | |
745edda6 FC |
1815 | ); |
1816 | } | |
1817 | ||
1818 | /* If this call originated from sv_clear, then we must check for | |
1819 | * effective names that need freeing, as well as the usual name. */ | |
1820 | name = HvNAME(hv); | |
15d9236d | 1821 | if (flags & HV_NAME_SETALL ? !!aux->xhv_name_u.xhvnameu_name : !!name) { |
745edda6 | 1822 | if (name && PL_stashcache) |
4643eb69 | 1823 | (void)hv_delete(PL_stashcache, name, (HEK_UTF8(HvNAME_HEK(hv)) ? -HvNAMELEN_get(hv) : HvNAMELEN_get(hv)), G_DISCARD); |
745edda6 | 1824 | hv_name_set(hv, NULL, 0, flags); |
47f1cf77 FC |
1825 | } |
1826 | if((meta = aux->xhv_mro_meta)) { | |
1827 | if (meta->mro_linear_all) { | |
1828 | SvREFCNT_dec(MUTABLE_SV(meta->mro_linear_all)); | |
1829 | meta->mro_linear_all = NULL; | |
1830 | /* This is just acting as a shortcut pointer. */ | |
1831 | meta->mro_linear_current = NULL; | |
1832 | } else if (meta->mro_linear_current) { | |
1833 | /* Only the current MRO is stored, so this owns the data. | |
1834 | */ | |
1835 | SvREFCNT_dec(meta->mro_linear_current); | |
1836 | meta->mro_linear_current = NULL; | |
1837 | } | |
9bfbb681 | 1838 | SvREFCNT_dec(meta->mro_nextmethod); |
47f1cf77 FC |
1839 | SvREFCNT_dec(meta->isa); |
1840 | Safefree(meta); | |
1841 | aux->xhv_mro_meta = NULL; | |
1842 | } | |
3b37eb24 | 1843 | if (!aux->xhv_name_u.xhvnameu_name && ! aux->xhv_backreferences) |
745edda6 | 1844 | SvFLAGS(hv) &= ~SVf_OOK; |
745edda6 FC |
1845 | } |
1846 | if (!SvOOK(hv)) { | |
1847 | Safefree(HvARRAY(hv)); | |
1848 | xhv->xhv_max = 7; /* HvMAX(hv) = 7 (it's a normal hash) */ | |
1849 | HvARRAY(hv) = 0; | |
2d0d1ecc | 1850 | } |
ca732855 | 1851 | HvPLACEHOLDERS_set(hv, 0); |
a0d0e21e LW |
1852 | |
1853 | if (SvRMAGICAL(hv)) | |
ad64d0ec | 1854 | mg_clear(MUTABLE_SV(hv)); |
79072805 LW |
1855 | } |
1856 | ||
4d0fbddd NC |
1857 | /* |
1858 | =for apidoc hv_fill | |
1859 | ||
1860 | Returns the number of hash buckets that happen to be in use. This function is | |
1861 | wrapped by the macro C<HvFILL>. | |
1862 | ||
1863 | Previously this value was stored in the HV structure, rather than being | |
1864 | calculated on demand. | |
1865 | ||
1866 | =cut | |
1867 | */ | |
1868 | ||
1869 | STRLEN | |
1870 | Perl_hv_fill(pTHX_ HV const *const hv) | |
1871 | { | |
1872 | STRLEN count = 0; | |
1873 | HE **ents = HvARRAY(hv); | |
1874 | ||
1875 | PERL_ARGS_ASSERT_HV_FILL; | |
1876 | ||
1877 | if (ents) { | |
fcd24582 NC |
1878 | HE *const *const last = ents + HvMAX(hv); |
1879 | count = last + 1 - ents; | |
4d0fbddd NC |
1880 | |
1881 | do { | |
fcd24582 NC |
1882 | if (!*ents) |
1883 | --count; | |
1884 | } while (++ents <= last); | |
4d0fbddd NC |
1885 | } |
1886 | return count; | |
1887 | } | |
1888 | ||
b464bac0 | 1889 | static struct xpvhv_aux* |
5f66b61c | 1890 | S_hv_auxinit(HV *hv) { |
bfcb3514 | 1891 | struct xpvhv_aux *iter; |
b79f7545 | 1892 | char *array; |
bfcb3514 | 1893 | |
7918f24d NC |
1894 | PERL_ARGS_ASSERT_HV_AUXINIT; |
1895 | ||
b79f7545 | 1896 | if (!HvARRAY(hv)) { |
a02a5408 | 1897 | Newxz(array, PERL_HV_ARRAY_ALLOC_BYTES(HvMAX(hv) + 1) |
b79f7545 NC |
1898 | + sizeof(struct xpvhv_aux), char); |
1899 | } else { | |
1900 | array = (char *) HvARRAY(hv); | |
1901 | Renew(array, PERL_HV_ARRAY_ALLOC_BYTES(HvMAX(hv) + 1) | |
1902 | + sizeof(struct xpvhv_aux), char); | |
1903 | } | |
1904 | HvARRAY(hv) = (HE**) array; | |
a82b195b | 1905 | SvOOK_on(hv); |
b79f7545 | 1906 | iter = HvAUX(hv); |
bfcb3514 NC |
1907 | |
1908 | iter->xhv_riter = -1; /* HvRITER(hv) = -1 */ | |
4608196e | 1909 | iter->xhv_eiter = NULL; /* HvEITER(hv) = NULL */ |
15d9236d | 1910 | iter->xhv_name_u.xhvnameu_name = 0; |
b7247a80 | 1911 | iter->xhv_name_count = 0; |
86f55936 | 1912 | iter->xhv_backreferences = 0; |
e1a479c5 | 1913 | iter->xhv_mro_meta = NULL; |
bfcb3514 NC |
1914 | return iter; |
1915 | } | |
1916 | ||
954c1994 GS |
1917 | /* |
1918 | =for apidoc hv_iterinit | |
1919 | ||
1920 | Prepares a starting point to traverse a hash table. Returns the number of | |
1b95d04f | 1921 | keys in the hash (i.e. the same as C<HvUSEDKEYS(hv)>). The return value is |
1c846c1f | 1922 | currently only meaningful for hashes without tie magic. |
954c1994 GS |
1923 | |
1924 | NOTE: Before version 5.004_65, C<hv_iterinit> used to return the number of | |
1925 | hash buckets that happen to be in use. If you still need that esoteric | |
b24b84ef | 1926 | value, you can get it through the macro C<HvFILL(hv)>. |
954c1994 | 1927 | |
e16e2ff8 | 1928 | |
954c1994 GS |
1929 | =cut |
1930 | */ | |
1931 | ||
79072805 | 1932 | I32 |
864dbfa3 | 1933 | Perl_hv_iterinit(pTHX_ HV *hv) |
79072805 | 1934 | { |
7918f24d NC |
1935 | PERL_ARGS_ASSERT_HV_ITERINIT; |
1936 | ||
1937 | /* FIXME: Are we not NULL, or do we croak? Place bets now! */ | |
1938 | ||
aa689395 | 1939 | if (!hv) |
cea2e8a9 | 1940 | Perl_croak(aTHX_ "Bad hash"); |
bfcb3514 | 1941 | |
b79f7545 | 1942 | if (SvOOK(hv)) { |
6136c704 | 1943 | struct xpvhv_aux * const iter = HvAUX(hv); |
0bd48802 | 1944 | HE * const entry = iter->xhv_eiter; /* HvEITER(hv) */ |
bfcb3514 NC |
1945 | if (entry && HvLAZYDEL(hv)) { /* was deleted earlier? */ |
1946 | HvLAZYDEL_off(hv); | |
1947 | hv_free_ent(hv, entry); | |
1948 | } | |
1949 | iter->xhv_riter = -1; /* HvRITER(hv) = -1 */ | |
4608196e | 1950 | iter->xhv_eiter = NULL; /* HvEITER(hv) = NULL */ |
bfcb3514 | 1951 | } else { |
6136c704 | 1952 | hv_auxinit(hv); |
72940dca | 1953 | } |
44a2ac75 | 1954 | |
cbec9347 | 1955 | /* used to be xhv->xhv_fill before 5.004_65 */ |
5d88ecd7 | 1956 | return HvTOTALKEYS(hv); |
79072805 | 1957 | } |
bfcb3514 NC |
1958 | |
1959 | I32 * | |
1960 | Perl_hv_riter_p(pTHX_ HV *hv) { | |
1961 | struct xpvhv_aux *iter; | |
1962 | ||
7918f24d NC |
1963 | PERL_ARGS_ASSERT_HV_RITER_P; |
1964 | ||
bfcb3514 NC |
1965 | if (!hv) |
1966 | Perl_croak(aTHX_ "Bad hash"); | |
1967 | ||
6136c704 | 1968 | iter = SvOOK(hv) ? HvAUX(hv) : hv_auxinit(hv); |
bfcb3514 NC |
1969 | return &(iter->xhv_riter); |
1970 | } | |
1971 | ||
1972 | HE ** | |
1973 | Perl_hv_eiter_p(pTHX_ HV *hv) { | |
1974 | struct xpvhv_aux *iter; | |
1975 | ||
7918f24d NC |
1976 | PERL_ARGS_ASSERT_HV_EITER_P; |
1977 | ||
bfcb3514 NC |
1978 | if (!hv) |
1979 | Perl_croak(aTHX_ "Bad hash"); | |
1980 | ||
6136c704 | 1981 | iter = SvOOK(hv) ? HvAUX(hv) : hv_auxinit(hv); |
bfcb3514 NC |
1982 | return &(iter->xhv_eiter); |
1983 | } | |
1984 | ||
1985 | void | |
1986 | Perl_hv_riter_set(pTHX_ HV *hv, I32 riter) { | |
1987 | struct xpvhv_aux *iter; | |
1988 | ||
7918f24d NC |
1989 | PERL_ARGS_ASSERT_HV_RITER_SET; |
1990 | ||
bfcb3514 NC |
1991 | if (!hv) |
1992 | Perl_croak(aTHX_ "Bad hash"); | |
1993 | ||
b79f7545 NC |
1994 | if (SvOOK(hv)) { |
1995 | iter = HvAUX(hv); | |
1996 | } else { | |
bfcb3514 NC |
1997 | if (riter == -1) |
1998 | return; | |
1999 | ||
6136c704 | 2000 | iter = hv_auxinit(hv); |
bfcb3514 NC |
2001 | } |
2002 | iter->xhv_riter = riter; | |
2003 | } | |
2004 | ||
2005 | void | |
2006 | Perl_hv_eiter_set(pTHX_ HV *hv, HE *eiter) { | |
2007 | struct xpvhv_aux *iter; | |
2008 | ||
7918f24d NC |
2009 | PERL_ARGS_ASSERT_HV_EITER_SET; |
2010 | ||
bfcb3514 NC |
2011 | if (!hv) |
2012 | Perl_croak(aTHX_ "Bad hash"); | |
2013 | ||
b79f7545 NC |
2014 | if (SvOOK(hv)) { |
2015 | iter = HvAUX(hv); | |
2016 | } else { | |
bfcb3514 NC |
2017 | /* 0 is the default so don't go malloc()ing a new structure just to |
2018 | hold 0. */ | |
2019 | if (!eiter) | |
2020 | return; | |
2021 | ||
6136c704 | 2022 | iter = hv_auxinit(hv); |
bfcb3514 NC |
2023 | } |
2024 | iter->xhv_eiter = eiter; | |
2025 | } | |
2026 | ||
bfcb3514 | 2027 | void |
4164be69 | 2028 | Perl_hv_name_set(pTHX_ HV *hv, const char *name, U32 len, U32 flags) |
bfcb3514 | 2029 | { |
97aff369 | 2030 | dVAR; |
b79f7545 | 2031 | struct xpvhv_aux *iter; |
7423f6db | 2032 | U32 hash; |
78b79c77 | 2033 | HEK **spot; |
46c461b5 | 2034 | |
7918f24d | 2035 | PERL_ARGS_ASSERT_HV_NAME_SET; |
bfcb3514 | 2036 | |
4164be69 NC |
2037 | if (len > I32_MAX) |
2038 | Perl_croak(aTHX_ "panic: hv name too long (%"UVuf")", (UV) len); | |
2039 | ||
b79f7545 NC |
2040 | if (SvOOK(hv)) { |
2041 | iter = HvAUX(hv); | |
15d9236d | 2042 | if (iter->xhv_name_u.xhvnameu_name) { |
b7247a80 | 2043 | if(iter->xhv_name_count) { |
745edda6 | 2044 | if(flags & HV_NAME_SETALL) { |
15d9236d | 2045 | HEK ** const name = HvAUX(hv)->xhv_name_u.xhvnameu_names; |
78b79c77 FC |
2046 | HEK **hekp = name + ( |
2047 | iter->xhv_name_count < 0 | |
2048 | ? -iter->xhv_name_count | |
2049 | : iter->xhv_name_count | |
2050 | ); | |
2051 | while(hekp-- > name+1) | |
b7247a80 | 2052 | unshare_hek_or_pvn(*hekp, 0, 0, 0); |
78b79c77 FC |
2053 | /* The first elem may be null. */ |
2054 | if(*name) unshare_hek_or_pvn(*name, 0, 0, 0); | |
b7247a80 | 2055 | Safefree(name); |
15d9236d | 2056 | spot = &iter->xhv_name_u.xhvnameu_name; |
78b79c77 FC |
2057 | iter->xhv_name_count = 0; |
2058 | } | |
2059 | else { | |
78b79c77 FC |
2060 | if(iter->xhv_name_count > 0) { |
2061 | /* shift some things over */ | |
15d9236d NC |
2062 | Renew( |
2063 | iter->xhv_name_u.xhvnameu_names, iter->xhv_name_count + 1, HEK * | |
4c2bfb4f | 2064 | ); |
15d9236d | 2065 | spot = iter->xhv_name_u.xhvnameu_names; |
4c2bfb4f | 2066 | spot[iter->xhv_name_count] = spot[1]; |
78b79c77 | 2067 | spot[1] = spot[0]; |
4c2bfb4f | 2068 | iter->xhv_name_count = -(iter->xhv_name_count + 1); |
78b79c77 | 2069 | } |
15d9236d | 2070 | else if(*(spot = iter->xhv_name_u.xhvnameu_names)) { |
78b79c77 FC |
2071 | unshare_hek_or_pvn(*spot, 0, 0, 0); |
2072 | } | |
2073 | } | |
2074 | } | |
745edda6 | 2075 | else if (flags & HV_NAME_SETALL) { |
15d9236d NC |
2076 | unshare_hek_or_pvn(iter->xhv_name_u.xhvnameu_name, 0, 0, 0); |
2077 | spot = &iter->xhv_name_u.xhvnameu_name; | |
b7247a80 | 2078 | } |
745edda6 | 2079 | else { |
15d9236d NC |
2080 | HEK * const existing_name = iter->xhv_name_u.xhvnameu_name; |
2081 | Newx(iter->xhv_name_u.xhvnameu_names, 2, HEK *); | |
745edda6 | 2082 | iter->xhv_name_count = -2; |
15d9236d | 2083 | spot = iter->xhv_name_u.xhvnameu_names; |
745edda6 FC |
2084 | spot[1] = existing_name; |
2085 | } | |
7423f6db | 2086 | } |
15d9236d | 2087 | else { spot = &iter->xhv_name_u.xhvnameu_name; iter->xhv_name_count = 0; } |
16580ff5 | 2088 | } else { |
bfcb3514 NC |
2089 | if (name == 0) |
2090 | return; | |
2091 | ||
6136c704 | 2092 | iter = hv_auxinit(hv); |
15d9236d | 2093 | spot = &iter->xhv_name_u.xhvnameu_name; |
bfcb3514 | 2094 | } |
7423f6db | 2095 | PERL_HASH(hash, name, len); |
c60dbbc3 | 2096 | *spot = name ? share_hek(name, flags & SVf_UTF8 ? -(I32)len : (I32)len, hash) : NULL; |
4643eb69 BF |
2097 | } |
2098 | ||
2099 | /* | |
2100 | This is basically sv_eq_flags() in sv.c, but we avoid the magic | |
2101 | and bytes checking. | |
2102 | */ | |
2103 | ||
2104 | STATIC I32 | |
2105 | hek_eq_pvn_flags(pTHX_ const HEK *hek, const char* pv, const I32 pvlen, const U32 flags) { | |
2106 | if ( (HEK_UTF8(hek) ? 1 : 0) != (flags & SVf_UTF8 ? 1 : 0) ) { | |
2107 | if (flags & SVf_UTF8) | |
2108 | return (bytes_cmp_utf8( | |
2109 | (const U8*)HEK_KEY(hek), HEK_LEN(hek), | |
2110 | (const U8*)pv, pvlen) == 0); | |
2111 | else | |
2112 | return (bytes_cmp_utf8( | |
2113 | (const U8*)pv, pvlen, | |
2114 | (const U8*)HEK_KEY(hek), HEK_LEN(hek)) == 0); | |
2115 | } | |
2116 | else | |
d35fec6c | 2117 | return HEK_LEN(hek) == pvlen && ((HEK_KEY(hek) == pv) |
4643eb69 | 2118 | || memEQ(HEK_KEY(hek), pv, pvlen)); |
bfcb3514 NC |
2119 | } |
2120 | ||
99206677 FC |
2121 | /* |
2122 | =for apidoc hv_ename_add | |
2123 | ||
2124 | Adds a name to a stash's internal list of effective names. See | |
2125 | C<hv_ename_delete>. | |
2126 | ||
2127 | This is called when a stash is assigned to a new location in the symbol | |
2128 | table. | |
2129 | ||
2130 | =cut | |
2131 | */ | |
2132 | ||
ee72b38d | 2133 | void |
27a1175b | 2134 | Perl_hv_ename_add(pTHX_ HV *hv, const char *name, U32 len, U32 flags) |
ee72b38d FC |
2135 | { |
2136 | dVAR; | |
2137 | struct xpvhv_aux *aux = SvOOK(hv) ? HvAUX(hv) : hv_auxinit(hv); | |
2138 | U32 hash; | |
2139 | ||
78b79c77 | 2140 | PERL_ARGS_ASSERT_HV_ENAME_ADD; |
ee72b38d FC |
2141 | |
2142 | if (len > I32_MAX) | |
2143 | Perl_croak(aTHX_ "panic: hv name too long (%"UVuf")", (UV) len); | |
2144 | ||
2145 | PERL_HASH(hash, name, len); | |
2146 | ||
ee72b38d | 2147 | if (aux->xhv_name_count) { |
15d9236d | 2148 | HEK ** const xhv_name = aux->xhv_name_u.xhvnameu_names; |
78b79c77 FC |
2149 | I32 count = aux->xhv_name_count; |
2150 | HEK **hekp = xhv_name + (count < 0 ? -count : count); | |
ee72b38d FC |
2151 | while (hekp-- > xhv_name) |
2152 | if ( | |
4643eb69 BF |
2153 | (HEK_UTF8(*hekp) || (flags & SVf_UTF8)) |
2154 | ? hek_eq_pvn_flags(aTHX_ *hekp, name, (I32)len, flags) | |
2155 | : (HEK_LEN(*hekp) == (I32)len && memEQ(HEK_KEY(*hekp), name, len)) | |
2156 | ) { | |
78b79c77 FC |
2157 | if (hekp == xhv_name && count < 0) |
2158 | aux->xhv_name_count = -count; | |
2159 | return; | |
2160 | } | |
2161 | if (count < 0) aux->xhv_name_count--, count = -count; | |
2162 | else aux->xhv_name_count++; | |
15d9236d | 2163 | Renew(aux->xhv_name_u.xhvnameu_names, count + 1, HEK *); |
c60dbbc3 | 2164 | (aux->xhv_name_u.xhvnameu_names)[count] = share_hek(name, (flags & SVf_UTF8 ? -(I32)len : (I32)len), hash); |
ee72b38d FC |
2165 | } |
2166 | else { | |
15d9236d | 2167 | HEK *existing_name = aux->xhv_name_u.xhvnameu_name; |
ee72b38d | 2168 | if ( |
4643eb69 BF |
2169 | existing_name && ( |
2170 | (HEK_UTF8(existing_name) || (flags & SVf_UTF8)) | |
2171 | ? hek_eq_pvn_flags(aTHX_ existing_name, name, (I32)len, flags) | |
2172 | : (HEK_LEN(existing_name) == (I32)len && memEQ(HEK_KEY(existing_name), name, len)) | |
2173 | ) | |
ee72b38d | 2174 | ) return; |
15d9236d | 2175 | Newx(aux->xhv_name_u.xhvnameu_names, 2, HEK *); |
78b79c77 | 2176 | aux->xhv_name_count = existing_name ? 2 : -2; |
15d9236d | 2177 | *aux->xhv_name_u.xhvnameu_names = existing_name; |
c60dbbc3 | 2178 | (aux->xhv_name_u.xhvnameu_names)[1] = share_hek(name, (flags & SVf_UTF8 ? -(I32)len : (I32)len), hash); |
ee72b38d FC |
2179 | } |
2180 | } | |
2181 | ||
99206677 FC |
2182 | /* |
2183 | =for apidoc hv_ename_delete | |
2184 | ||
2185 | Removes a name from a stash's internal list of effective names. If this is | |
2186 | the name returned by C<HvENAME>, then another name in the list will take | |
2187 | its place (C<HvENAME> will use it). | |
2188 | ||
2189 | This is called when a stash is deleted from the symbol table. | |
2190 | ||
2191 | =cut | |
2192 | */ | |
2193 | ||
ee72b38d | 2194 | void |
27a1175b | 2195 | Perl_hv_ename_delete(pTHX_ HV *hv, const char *name, U32 len, U32 flags) |
ee72b38d FC |
2196 | { |
2197 | dVAR; | |
2198 | struct xpvhv_aux *aux; | |
2199 | ||
78b79c77 | 2200 | PERL_ARGS_ASSERT_HV_ENAME_DELETE; |
ee72b38d FC |
2201 | |
2202 | if (len > I32_MAX) | |
2203 | Perl_croak(aTHX_ "panic: hv name too long (%"UVuf")", (UV) len); | |
2204 | ||
2205 | if (!SvOOK(hv)) return; | |
2206 | ||
2207 | aux = HvAUX(hv); | |
15d9236d | 2208 | if (!aux->xhv_name_u.xhvnameu_name) return; |
ee72b38d FC |
2209 | |
2210 | if (aux->xhv_name_count) { | |
15d9236d | 2211 | HEK ** const namep = aux->xhv_name_u.xhvnameu_names; |
78b79c77 FC |
2212 | I32 const count = aux->xhv_name_count; |
2213 | HEK **victim = namep + (count < 0 ? -count : count); | |
2214 | while (victim-- > namep + 1) | |
ee72b38d | 2215 | if ( |
4643eb69 BF |
2216 | (HEK_UTF8(*victim) || (flags & SVf_UTF8)) |
2217 | ? hek_eq_pvn_flags(aTHX_ *victim, name, (I32)len, flags) | |
2218 | : (HEK_LEN(*victim) == (I32)len && memEQ(HEK_KEY(*victim), name, len)) | |
ee72b38d FC |
2219 | ) { |
2220 | unshare_hek_or_pvn(*victim, 0, 0, 0); | |
78b79c77 FC |
2221 | if (count < 0) ++aux->xhv_name_count; |
2222 | else --aux->xhv_name_count; | |
2223 | if ( | |
2224 | (aux->xhv_name_count == 1 || aux->xhv_name_count == -1) | |
2225 | && !*namep | |
2226 | ) { /* if there are none left */ | |
ee72b38d | 2227 | Safefree(namep); |
15d9236d | 2228 | aux->xhv_name_u.xhvnameu_names = NULL; |
78b79c77 | 2229 | aux->xhv_name_count = 0; |
ee72b38d FC |
2230 | } |
2231 | else { | |
2232 | /* Move the last one back to fill the empty slot. It | |
2233 | does not matter what order they are in. */ | |
78b79c77 | 2234 | *victim = *(namep + (count < 0 ? -count : count) - 1); |
ee72b38d FC |
2235 | } |
2236 | return; | |
2237 | } | |
78b79c77 | 2238 | if ( |
4643eb69 BF |
2239 | count > 0 && (HEK_UTF8(*namep) || (flags & SVf_UTF8)) |
2240 | ? hek_eq_pvn_flags(aTHX_ *namep, name, (I32)len, flags) | |
2241 | : (HEK_LEN(*namep) == (I32)len && memEQ(HEK_KEY(*namep), name, len)) | |
78b79c77 FC |
2242 | ) { |
2243 | aux->xhv_name_count = -count; | |
2244 | } | |
ee72b38d FC |
2245 | } |
2246 | else if( | |
4643eb69 BF |
2247 | (HEK_UTF8(aux->xhv_name_u.xhvnameu_name) || (flags & SVf_UTF8)) |
2248 | ? hek_eq_pvn_flags(aTHX_ aux->xhv_name_u.xhvnameu_name, name, (I32)len, flags) | |
2249 | : (HEK_LEN(aux->xhv_name_u.xhvnameu_name) == (I32)len && | |
2250 | memEQ(HEK_KEY(aux->xhv_name_u.xhvnameu_name), name, len)) | |
ee72b38d | 2251 | ) { |
15d9236d NC |
2252 | HEK * const namehek = aux->xhv_name_u.xhvnameu_name; |
2253 | Newx(aux->xhv_name_u.xhvnameu_names, 1, HEK *); | |
2254 | *aux->xhv_name_u.xhvnameu_names = namehek; | |
3f783763 | 2255 | aux->xhv_name_count = -1; |
ee72b38d FC |
2256 | } |
2257 | } | |
2258 | ||
86f55936 NC |
2259 | AV ** |
2260 | Perl_hv_backreferences_p(pTHX_ HV *hv) { | |
6136c704 | 2261 | struct xpvhv_aux * const iter = SvOOK(hv) ? HvAUX(hv) : hv_auxinit(hv); |
7918f24d NC |
2262 | |
2263 | PERL_ARGS_ASSERT_HV_BACKREFERENCES_P; | |
96a5add6 | 2264 | PERL_UNUSED_CONTEXT; |
7918f24d | 2265 | |
86f55936 NC |
2266 | return &(iter->xhv_backreferences); |
2267 | } | |
2268 | ||
09aad8f0 DM |
2269 | void |
2270 | Perl_hv_kill_backrefs(pTHX_ HV *hv) { | |
2271 | AV *av; | |
2272 | ||
2273 | PERL_ARGS_ASSERT_HV_KILL_BACKREFS; | |
2274 | ||
2275 | if (!SvOOK(hv)) | |
2276 | return; | |
2277 | ||
2278 | av = HvAUX(hv)->xhv_backreferences; | |
2279 | ||
2280 | if (av) { | |
2281 | HvAUX(hv)->xhv_backreferences = 0; | |
2282 | Perl_sv_kill_backrefs(aTHX_ MUTABLE_SV(hv), av); | |
5648c0ae DM |
2283 | if (SvTYPE(av) == SVt_PVAV) |
2284 | SvREFCNT_dec(av); | |
09aad8f0 DM |
2285 | } |
2286 | } | |
2287 | ||
954c1994 | 2288 | /* |
7a7b9979 NC |
2289 | hv_iternext is implemented as a macro in hv.h |
2290 | ||
954c1994 GS |
2291 | =for apidoc hv_iternext |
2292 | ||
2293 | Returns entries from a hash iterator. See C<hv_iterinit>. | |
2294 | ||
fe7bca90 NC |
2295 | You may call C<hv_delete> or C<hv_delete_ent> on the hash entry that the |
2296 | iterator currently points to, without losing your place or invalidating your | |
2297 | iterator. Note that in this case the current entry is deleted from the hash | |
2298 | with your iterator holding the last reference to it. Your iterator is flagged | |
2299 | to free the entry on the next call to C<hv_iternext>, so you must not discard | |
2300 | your iterator immediately else the entry will leak - call C<hv_iternext> to | |
2301 | trigger the resource deallocation. | |
2302 | ||
fe7bca90 NC |
2303 | =for apidoc hv_iternext_flags |
2304 | ||
2305 | Returns entries from a hash iterator. See C<hv_iterinit> and C<hv_iternext>. | |
2306 | The C<flags> value will normally be zero; if HV_ITERNEXT_WANTPLACEHOLDERS is | |
2307 | set the placeholders keys (for restricted hashes) will be returned in addition | |
2308 | to normal keys. By default placeholders are automatically skipped over. | |
7996736c MHM |
2309 | Currently a placeholder is implemented with a value that is |
2310 | C<&Perl_sv_placeholder>. Note that the implementation of placeholders and | |
fe7bca90 NC |
2311 | restricted hashes may change, and the implementation currently is |
2312 | insufficiently abstracted for any change to be tidy. | |
e16e2ff8 | 2313 | |
fe7bca90 | 2314 | =cut |
e16e2ff8 NC |
2315 | */ |
2316 | ||
2317 | HE * | |
2318 | Perl_hv_iternext_flags(pTHX_ HV *hv, I32 flags) | |
2319 | { | |
27da23d5 | 2320 | dVAR; |
cbec9347 | 2321 | register XPVHV* xhv; |
79072805 | 2322 | register HE *entry; |
a0d0e21e | 2323 | HE *oldentry; |
463ee0b2 | 2324 | MAGIC* mg; |
bfcb3514 | 2325 | struct xpvhv_aux *iter; |
79072805 | 2326 | |
7918f24d NC |
2327 | PERL_ARGS_ASSERT_HV_ITERNEXT_FLAGS; |
2328 | ||
79072805 | 2329 | if (!hv) |
cea2e8a9 | 2330 | Perl_croak(aTHX_ "Bad hash"); |
81714fb9 | 2331 | |
cbec9347 | 2332 | xhv = (XPVHV*)SvANY(hv); |
bfcb3514 | 2333 | |
b79f7545 | 2334 | if (!SvOOK(hv)) { |
bfcb3514 NC |
2335 | /* Too many things (well, pp_each at least) merrily assume that you can |
2336 | call iv_iternext without calling hv_iterinit, so we'll have to deal | |
2337 | with it. */ | |
2338 | hv_iterinit(hv); | |
bfcb3514 | 2339 | } |
b79f7545 | 2340 | iter = HvAUX(hv); |
bfcb3514 NC |
2341 | |
2342 | oldentry = entry = iter->xhv_eiter; /* HvEITER(hv) */ | |
e62cc96a | 2343 | if (SvMAGICAL(hv) && SvRMAGICAL(hv)) { |
ad64d0ec | 2344 | if ( ( mg = mg_find((const SV *)hv, PERL_MAGIC_tied) ) ) { |
e62cc96a YO |
2345 | SV * const key = sv_newmortal(); |
2346 | if (entry) { | |
2347 | sv_setsv(key, HeSVKEY_force(entry)); | |
2348 | SvREFCNT_dec(HeSVKEY(entry)); /* get rid of previous key */ | |
2349 | } | |
2350 | else { | |
2351 | char *k; | |
2352 | HEK *hek; | |
2353 | ||
2354 | /* one HE per MAGICAL hash */ | |
2355 | iter->xhv_eiter = entry = new_HE(); /* HvEITER(hv) = new_HE() */ | |
2356 | Zero(entry, 1, HE); | |
ad64d0ec | 2357 | Newxz(k, HEK_BASESIZE + sizeof(const SV *), char); |
e62cc96a YO |
2358 | hek = (HEK*)k; |
2359 | HeKEY_hek(entry) = hek; | |
2360 | HeKLEN(entry) = HEf_SVKEY; | |
2361 | } | |
ad64d0ec | 2362 | magic_nextpack(MUTABLE_SV(hv),mg,key); |
e62cc96a YO |
2363 | if (SvOK(key)) { |
2364 | /* force key to stay around until next time */ | |
2365 | HeSVKEY_set(entry, SvREFCNT_inc_simple_NN(key)); | |
2366 | return entry; /* beware, hent_val is not set */ | |
2367 | } | |
ef8d46e8 | 2368 | SvREFCNT_dec(HeVAL(entry)); |
e62cc96a YO |
2369 | Safefree(HeKEY_hek(entry)); |
2370 | del_HE(entry); | |
2371 | iter->xhv_eiter = NULL; /* HvEITER(hv) = NULL */ | |
2372 | return NULL; | |
81714fb9 | 2373 | } |
79072805 | 2374 | } |
7ee146b1 | 2375 | #if defined(DYNAMIC_ENV_FETCH) && !defined(__riscos__) /* set up %ENV for iteration */ |
ad64d0ec NC |
2376 | if (!entry && SvRMAGICAL((const SV *)hv) |
2377 | && mg_find((const SV *)hv, PERL_MAGIC_env)) { | |
f675dbe5 | 2378 | prime_env_iter(); |
03026e68 JM |
2379 | #ifdef VMS |
2380 | /* The prime_env_iter() on VMS just loaded up new hash values | |
2381 | * so the iteration count needs to be reset back to the beginning | |
2382 | */ | |
2383 | hv_iterinit(hv); | |
2384 | iter = HvAUX(hv); | |
2385 | oldentry = entry = iter->xhv_eiter; /* HvEITER(hv) */ | |
2386 | #endif | |
2387 | } | |
f675dbe5 | 2388 | #endif |
463ee0b2 | 2389 | |
bfaf5b52 | 2390 | /* hv_iterinit now ensures this. */ |
b79f7545 NC |
2391 | assert (HvARRAY(hv)); |
2392 | ||
015a5f36 | 2393 | /* At start of hash, entry is NULL. */ |
fde52b5c | 2394 | if (entry) |
8aacddc1 | 2395 | { |
fde52b5c | 2396 | entry = HeNEXT(entry); |
e16e2ff8 NC |
2397 | if (!(flags & HV_ITERNEXT_WANTPLACEHOLDERS)) { |
2398 | /* | |
2399 | * Skip past any placeholders -- don't want to include them in | |
2400 | * any iteration. | |
2401 | */ | |
7996736c | 2402 | while (entry && HeVAL(entry) == &PL_sv_placeholder) { |
e16e2ff8 NC |
2403 | entry = HeNEXT(entry); |
2404 | } | |
8aacddc1 NIS |
2405 | } |
2406 | } | |
015a5f36 | 2407 | |
9eb4ebd1 NC |
2408 | /* Skip the entire loop if the hash is empty. */ |
2409 | if ((flags & HV_ITERNEXT_WANTPLACEHOLDERS) | |
2410 | ? HvTOTALKEYS(hv) : HvUSEDKEYS(hv)) { | |
900ac051 MM |
2411 | while (!entry) { |
2412 | /* OK. Come to the end of the current list. Grab the next one. */ | |
2413 | ||
2414 | iter->xhv_riter++; /* HvRITER(hv)++ */ | |
2415 | if (iter->xhv_riter > (I32)xhv->xhv_max /* HvRITER(hv) > HvMAX(hv) */) { | |
2416 | /* There is no next one. End of the hash. */ | |
2417 | iter->xhv_riter = -1; /* HvRITER(hv) = -1 */ | |
2418 | break; | |
2419 | } | |
2420 | entry = (HvARRAY(hv))[iter->xhv_riter]; | |
8aacddc1 | 2421 | |
900ac051 MM |
2422 | if (!(flags & HV_ITERNEXT_WANTPLACEHOLDERS)) { |
2423 | /* If we have an entry, but it's a placeholder, don't count it. | |
2424 | Try the next. */ | |
2425 | while (entry && HeVAL(entry) == &PL_sv_placeholder) | |
2426 | entry = HeNEXT(entry); | |
2427 | } | |
2428 | /* Will loop again if this linked list starts NULL | |
2429 | (for HV_ITERNEXT_WANTPLACEHOLDERS) | |
2430 | or if we run through it and find only placeholders. */ | |
015a5f36 | 2431 | } |
fde52b5c | 2432 | } |
41aa816f | 2433 | else iter->xhv_riter = -1; |
79072805 | 2434 | |
72940dca | 2435 | if (oldentry && HvLAZYDEL(hv)) { /* was deleted earlier? */ |
2436 | HvLAZYDEL_off(hv); | |
68dc0745 | 2437 | hv_free_ent(hv, oldentry); |
72940dca | 2438 | } |
a0d0e21e | 2439 | |
fdcd69b6 | 2440 | /*if (HvREHASH(hv) && entry && !HeKREHASH(entry)) |
6c9570dc | 2441 | PerlIO_printf(PerlIO_stderr(), "Awooga %p %p\n", (void*)hv, (void*)entry);*/ |
fdcd69b6 | 2442 | |
bfcb3514 | 2443 | iter->xhv_eiter = entry; /* HvEITER(hv) = entry */ |
79072805 LW |
2444 | return entry; |
2445 | } | |
2446 | ||
954c1994 GS |
2447 | /* |
2448 | =for apidoc hv_iterkey | |
2449 | ||
2450 | Returns the key from the current position of the hash iterator. See | |
2451 | C<hv_iterinit>. | |
2452 | ||
2453 | =cut | |
2454 | */ | |
2455 | ||
79072805 | 2456 | char * |
864dbfa3 | 2457 | Perl_hv_iterkey(pTHX_ register HE *entry, I32 *retlen) |
79072805 | 2458 | { |
7918f24d NC |
2459 | PERL_ARGS_ASSERT_HV_ITERKEY; |
2460 | ||
fde52b5c | 2461 | if (HeKLEN(entry) == HEf_SVKEY) { |
fb73857a | 2462 | STRLEN len; |
0bd48802 | 2463 | char * const p = SvPV(HeKEY_sv(entry), len); |
fb73857a | 2464 | *retlen = len; |
2465 | return p; | |
fde52b5c | 2466 | } |
2467 | else { | |
2468 | *retlen = HeKLEN(entry); | |
2469 | return HeKEY(entry); | |
2470 | } | |
2471 | } | |
2472 | ||
2473 | /* unlike hv_iterval(), this always returns a mortal copy of the key */ | |
954c1994 GS |
2474 | /* |
2475 | =for apidoc hv_iterkeysv | |
2476 | ||
2477 | Returns the key as an C<SV*> from the current position of the hash | |
2478 | iterator. The return value will always be a mortal copy of the key. Also | |
2479 | see C<hv_iterinit>. | |
2480 | ||
2481 | =cut | |
2482 | */ | |
2483 | ||
fde52b5c | 2484 | SV * |
864dbfa3 | 2485 | Perl_hv_iterkeysv(pTHX_ register HE *entry) |
fde52b5c | 2486 | { |
7918f24d NC |
2487 | PERL_ARGS_ASSERT_HV_ITERKEYSV; |
2488 | ||
c1b02ed8 | 2489 | return sv_2mortal(newSVhek(HeKEY_hek(entry))); |
79072805 LW |
2490 | } |
2491 | ||
954c1994 GS |
2492 | /* |
2493 | =for apidoc hv_iterval | |
2494 | ||
2495 | Returns the value from the current position of the hash iterator. See | |
2496 | C<hv_iterkey>. | |
2497 | ||
2498 | =cut | |
2499 | */ | |
2500 | ||
79072805 | 2501 | SV * |
864dbfa3 | 2502 | Perl_hv_iterval(pTHX_ HV *hv, register HE *entry) |
79072805 | 2503 | { |
7918f24d NC |
2504 | PERL_ARGS_ASSERT_HV_ITERVAL; |
2505 | ||
8990e307 | 2506 | if (SvRMAGICAL(hv)) { |
ad64d0ec | 2507 | if (mg_find((const SV *)hv, PERL_MAGIC_tied)) { |
c4420975 | 2508 | SV* const sv = sv_newmortal(); |
bbce6d69 | 2509 | if (HeKLEN(entry) == HEf_SVKEY) |
ad64d0ec | 2510 | mg_copy(MUTABLE_SV(hv), sv, (char*)HeKEY_sv(entry), HEf_SVKEY); |
a3b680e6 | 2511 | else |
ad64d0ec | 2512 | mg_copy(MUTABLE_SV(hv), sv, HeKEY(entry), HeKLEN(entry)); |
463ee0b2 LW |
2513 | return sv; |
2514 | } | |
79072805 | 2515 | } |
fde52b5c | 2516 | return HeVAL(entry); |
79072805 LW |
2517 | } |
2518 | ||
954c1994 GS |
2519 | /* |
2520 | =for apidoc hv_iternextsv | |
2521 | ||
2522 | Performs an C<hv_iternext>, C<hv_iterkey>, and C<hv_iterval> in one | |
2523 | operation. | |
2524 | ||
2525 | =cut | |
2526 | */ | |
2527 | ||
a0d0e21e | 2528 | SV * |
864dbfa3 | 2529 | Perl_hv_iternextsv(pTHX_ HV *hv, char **key, I32 *retlen) |
a0d0e21e | 2530 | { |
0bd48802 AL |
2531 | HE * const he = hv_iternext_flags(hv, 0); |
2532 | ||
7918f24d NC |
2533 | PERL_ARGS_ASSERT_HV_ITERNEXTSV; |
2534 | ||
0bd48802 | 2535 | if (!he) |
a0d0e21e LW |
2536 | return NULL; |
2537 | *key = hv_iterkey(he, retlen); | |
2538 | return hv_iterval(hv, he); | |
2539 | } | |
2540 | ||
954c1994 | 2541 | /* |
bc5cdc23 NC |
2542 | |
2543 | Now a macro in hv.h | |
2544 | ||
954c1994 GS |
2545 | =for apidoc hv_magic |
2546 | ||
2547 | Adds magic to a hash. See C<sv_magic>. | |
2548 | ||
2549 | =cut | |
2550 | */ | |
2551 | ||
bbce6d69 | 2552 | /* possibly free a shared string if no one has access to it |
fde52b5c | 2553 | * len and hash must both be valid for str. |
2554 | */ | |
bbce6d69 | 2555 | void |
864dbfa3 | 2556 | Perl_unsharepvn(pTHX_ const char *str, I32 len, U32 hash) |
fde52b5c | 2557 | { |
19692e8d NC |
2558 | unshare_hek_or_pvn (NULL, str, len, hash); |
2559 | } | |
2560 | ||
2561 | ||
2562 | void | |
2563 | Perl_unshare_hek(pTHX_ HEK *hek) | |
2564 | { | |
bf11fd37 | 2565 | assert(hek); |
19692e8d NC |
2566 | unshare_hek_or_pvn(hek, NULL, 0, 0); |
2567 | } | |
2568 | ||
2569 | /* possibly free a shared string if no one has access to it | |
2570 | hek if non-NULL takes priority over the other 3, else str, len and hash | |
2571 | are used. If so, len and hash must both be valid for str. | |
2572 | */ | |
df132699 | 2573 | STATIC void |
97ddebaf | 2574 | S_unshare_hek_or_pvn(pTHX_ const HEK *hek, const char *str, I32 len, U32 hash) |
19692e8d | 2575 | { |
97aff369 | 2576 | dVAR; |
cbec9347 | 2577 | register XPVHV* xhv; |
20454177 | 2578 | HE *entry; |
fde52b5c | 2579 | register HE **oentry; |
c3654f1a | 2580 | bool is_utf8 = FALSE; |
19692e8d | 2581 | int k_flags = 0; |
aec46f14 | 2582 | const char * const save = str; |
cbbf8932 | 2583 | struct shared_he *he = NULL; |
c3654f1a | 2584 | |
19692e8d | 2585 | if (hek) { |
cbae3960 NC |
2586 | /* Find the shared he which is just before us in memory. */ |
2587 | he = (struct shared_he *)(((char *)hek) | |
2588 | - STRUCT_OFFSET(struct shared_he, | |
2589 | shared_he_hek)); | |
2590 | ||
2591 | /* Assert that the caller passed us a genuine (or at least consistent) | |
2592 | shared hek */ | |
2593 | assert (he->shared_he_he.hent_hek == hek); | |
29404ae0 | 2594 | |
de616631 NC |
2595 | if (he->shared_he_he.he_valu.hent_refcount - 1) { |
2596 | --he->shared_he_he.he_valu.hent_refcount; | |
29404ae0 NC |
2597 | return; |
2598 | } | |
29404ae0 | 2599 | |
19692e8d NC |
2600 | hash = HEK_HASH(hek); |
2601 | } else if (len < 0) { | |
2602 | STRLEN tmplen = -len; | |
2603 | is_utf8 = TRUE; | |
2604 | /* See the note in hv_fetch(). --jhi */ | |
2605 | str = (char*)bytes_from_utf8((U8*)str, &tmplen, &is_utf8); | |
2606 | len = tmplen; | |
2607 | if (is_utf8) | |
2608 | k_flags = HVhek_UTF8; | |
2609 | if (str != save) | |
2610 | k_flags |= HVhek_WASUTF8 | HVhek_FREEKEY; | |
c3654f1a | 2611 | } |
1c846c1f | 2612 | |
de616631 | 2613 | /* what follows was the moral equivalent of: |
6b88bc9c | 2614 | if ((Svp = hv_fetch(PL_strtab, tmpsv, FALSE, hash))) { |
a0714e2c | 2615 | if (--*Svp == NULL) |
6b88bc9c | 2616 | hv_delete(PL_strtab, str, len, G_DISCARD, hash); |
bbce6d69 | 2617 | } */ |
cbec9347 | 2618 | xhv = (XPVHV*)SvANY(PL_strtab); |
fde52b5c | 2619 | /* assert(xhv_array != 0) */ |
9de10d5c | 2620 | oentry = &(HvARRAY(PL_strtab))[hash & (I32) HvMAX(PL_strtab)]; |
6c1b96a1 NC |
2621 | if (he) { |
2622 | const HE *const he_he = &(he->shared_he_he); | |
45d1cc86 | 2623 | for (entry = *oentry; entry; oentry = &HeNEXT(entry), entry = *oentry) { |
35ab5632 NC |
2624 | if (entry == he_he) |
2625 | break; | |
19692e8d NC |
2626 | } |
2627 | } else { | |
35a4481c | 2628 | const int flags_masked = k_flags & HVhek_MASK; |
45d1cc86 | 2629 | for (entry = *oentry; entry; oentry = &HeNEXT(entry), entry = *oentry) { |
19692e8d NC |
2630 | if (HeHASH(entry) != hash) /* strings can't be equal */ |
2631 | continue; | |
2632 | if (HeKLEN(entry) != len) | |
2633 | continue; | |
2634 | if (HeKEY(entry) != str && memNE(HeKEY(entry),str,len)) /* is this it? */ | |
2635 | continue; | |
2636 | if (HeKFLAGS(entry) != flags_masked) | |
2637 | continue; | |
19692e8d NC |
2638 | break; |
2639 | } | |
2640 | } | |
2641 | ||
35ab5632 NC |
2642 | if (entry) { |
2643 | if (--entry->he_valu.hent_refcount == 0) { | |
19692e8d | 2644 | *oentry = HeNEXT(entry); |
cbae3960 | 2645 | Safefree(entry); |
4c7185a0 | 2646 | xhv->xhv_keys--; /* HvTOTALKEYS(hv)-- */ |
19692e8d | 2647 | } |
fde52b5c | 2648 | } |
19692e8d | 2649 | |
9b387841 NC |
2650 | if (!entry) |
2651 | Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL), | |
12578ffb | 2652 | "Attempt to free nonexistent shared string '%s'%s" |
9b387841 NC |
2653 | pTHX__FORMAT, |
2654 | hek ? HEK_KEY(hek) : str, | |
2655 | ((k_flags & HVhek_UTF8) ? " (utf8)" : "") pTHX__VALUE); | |
19692e8d NC |
2656 | if (k_flags & HVhek_FREEKEY) |
2657 | Safefree(str); | |
fde52b5c | 2658 | } |
2659 | ||
bbce6d69 | 2660 | /* get a (constant) string ptr from the global string table |
2661 | * string will get added if it is not already there. | |
fde52b5c | 2662 | * len and hash must both be valid for str. |
2663 | */ | |
bbce6d69 | 2664 | HEK * |
864dbfa3 | 2665 | Perl_share_hek(pTHX_ const char *str, I32 len, register U32 hash) |
fde52b5c | 2666 | { |
da58a35d | 2667 | bool is_utf8 = FALSE; |
19692e8d | 2668 | int flags = 0; |
aec46f14 | 2669 | const char * const save = str; |
da58a35d | 2670 | |
7918f24d NC |
2671 | PERL_ARGS_ASSERT_SHARE_HEK; |
2672 | ||
da58a35d | 2673 | if (len < 0) { |
77caf834 | 2674 | STRLEN tmplen = -len; |
da58a35d | 2675 | is_utf8 = TRUE; |
77caf834 JH |
2676 | /* See the note in hv_fetch(). --jhi */ |
2677 | str = (char*)bytes_from_utf8((U8*)str, &tmplen, &is_utf8); | |
2678 | len = tmplen; | |
19692e8d NC |
2679 | /* If we were able to downgrade here, then than means that we were passed |
2680 | in a key which only had chars 0-255, but was utf8 encoded. */ | |
2681 | if (is_utf8) | |
2682 | flags = HVhek_UTF8; | |
2683 | /* If we found we were able to downgrade the string to bytes, then | |
2684 | we should flag that it needs upgrading on keys or each. Also flag | |
2685 | that we need share_hek_flags to free the string. */ | |
4643eb69 BF |
2686 | if (str != save) { |
2687 | PERL_HASH(hash, str, len); | |
19692e8d | 2688 | flags |= HVhek_WASUTF8 | HVhek_FREEKEY; |
4643eb69 | 2689 | } |
19692e8d NC |
2690 | } |
2691 | ||
6e838c70 | 2692 | return share_hek_flags (str, len, hash, flags); |
19692e8d NC |
2693 | } |
2694 | ||
6e838c70 | 2695 | STATIC HEK * |
19692e8d NC |
2696 | S_share_hek_flags(pTHX_ const char *str, I32 len, register U32 hash, int flags) |
2697 | { | |
97aff369 | 2698 | dVAR; |
19692e8d | 2699 | register HE *entry; |
35a4481c | 2700 | const int flags_masked = flags & HVhek_MASK; |
263cb4a6 | 2701 | const U32 hindex = hash & (I32) HvMAX(PL_strtab); |
7918f24d NC |
2702 | register XPVHV * const xhv = (XPVHV*)SvANY(PL_strtab); |
2703 | ||
2704 | PERL_ARGS_ASSERT_SHARE_HEK_FLAGS; | |
bbce6d69 | 2705 | |
fde52b5c | 2706 | /* what follows is the moral equivalent of: |
1c846c1f | 2707 | |
6b88bc9c | 2708 | if (!(Svp = hv_fetch(PL_strtab, str, len, FALSE))) |
a0714e2c | 2709 | hv_store(PL_strtab, str, len, NULL, hash); |
fdcd69b6 NC |
2710 | |
2711 | Can't rehash the shared string table, so not sure if it's worth | |
2712 | counting the number of entries in the linked list | |
bbce6d69 | 2713 | */ |
7918f24d | 2714 | |
fde52b5c | 2715 | /* assert(xhv_array != 0) */ |
263cb4a6 NC |
2716 | entry = (HvARRAY(PL_strtab))[hindex]; |
2717 | for (;entry; entry = HeNEXT(entry)) { | |
fde52b5c | 2718 | if (HeHASH(entry) != hash) /* strings can't be equal */ |
2719 | continue; | |
2720 | if (HeKLEN(entry) != len) | |
2721 | continue; | |
1c846c1f | 2722 | if (HeKEY(entry) != str && memNE(HeKEY(entry),str,len)) /* is this it? */ |
fde52b5c | 2723 | continue; |
19692e8d | 2724 | if (HeKFLAGS(entry) != flags_masked) |
c3654f1a | 2725 | continue; |
fde52b5c | 2726 | break; |
2727 | } | |
263cb4a6 NC |
2728 | |
2729 | if (!entry) { | |
45d1cc86 NC |
2730 | /* What used to be head of the list. |
2731 | If this is NULL, then we're the first entry for this slot, which | |
2732 | means we need to increate fill. */ | |
cbae3960 NC |
2733 | struct shared_he *new_entry; |
2734 | HEK *hek; | |
2735 | char *k; | |
263cb4a6 NC |
2736 | HE **const head = &HvARRAY(PL_strtab)[hindex]; |
2737 | HE *const next = *head; | |
cbae3960 NC |
2738 | |
2739 | /* We don't actually store a HE from the arena and a regular HEK. | |
2740 | Instead we allocate one chunk of memory big enough for both, | |
2741 | and put the HEK straight after the HE. This way we can find the | |
f52337cf | 2742 | HE directly from the HEK. |
cbae3960 NC |
2743 | */ |
2744 | ||
a02a5408 | 2745 | Newx(k, STRUCT_OFFSET(struct shared_he, |
cbae3960 NC |
2746 | shared_he_hek.hek_key[0]) + len + 2, char); |
2747 | new_entry = (struct shared_he *)k; | |
2748 | entry = &(new_entry->shared_he_he); | |
2749 | hek = &(new_entry->shared_he_hek); | |
2750 | ||
2751 | Copy(str, HEK_KEY(hek), len, char); | |
2752 | HEK_KEY(hek)[len] = 0; | |
2753 | HEK_LEN(hek) = len; | |
2754 | HEK_HASH(hek) = hash; | |
2755 | HEK_FLAGS(hek) = (unsigned char)flags_masked; | |
2756 | ||
2757 | /* Still "point" to the HEK, so that other code need not know what | |
2758 | we're up to. */ | |
2759 | HeKEY_hek(entry) = hek; | |
de616631 | 2760 | entry->he_valu.hent_refcount = 0; |
263cb4a6 NC |
2761 | HeNEXT(entry) = next; |
2762 | *head = entry; | |
cbae3960 | 2763 | |
4c7185a0 | 2764 | xhv->xhv_keys++; /* HvTOTALKEYS(hv)++ */ |
263cb4a6 | 2765 | if (!next) { /* initial entry? */ |
1b95d04f | 2766 | } else if (xhv->xhv_keys > xhv->xhv_max /* HvUSEDKEYS(hv) > HvMAX(hv) */) { |
cbec9347 | 2767 | hsplit(PL_strtab); |
bbce6d69 | 2768 | } |
2769 | } | |
2770 | ||
de616631 | 2771 | ++entry->he_valu.hent_refcount; |
19692e8d NC |
2772 | |
2773 | if (flags & HVhek_FREEKEY) | |
f9a63242 | 2774 | Safefree(str); |
19692e8d | 2775 | |
6e838c70 | 2776 | return HeKEY_hek(entry); |
fde52b5c | 2777 | } |
ecae49c0 | 2778 | |
ca732855 NC |
2779 | I32 * |
2780 | Perl_hv_placeholders_p(pTHX_ HV *hv) | |
2781 | { | |
2782 | dVAR; | |
ad64d0ec | 2783 | MAGIC *mg = mg_find((const SV *)hv, PERL_MAGIC_rhash); |
ca732855 | 2784 | |
7918f24d NC |
2785 | PERL_ARGS_ASSERT_HV_PLACEHOLDERS_P; |
2786 | ||
ca732855 | 2787 | if (!mg) { |
ad64d0ec | 2788 | mg = sv_magicext(MUTABLE_SV(hv), 0, PERL_MAGIC_rhash, 0, 0, 0); |
ca732855 NC |
2789 | |
2790 | if (!mg) { | |
2791 | Perl_die(aTHX_ "panic: hv_placeholders_p"); | |
2792 | } | |
2793 | } | |
2794 | return &(mg->mg_len); | |
2795 | } | |
2796 | ||
2797 | ||
2798 | I32 | |
0c289d13 | 2799 | Perl_hv_placeholders_get(pTHX_ const HV *hv) |
ca732855 NC |
2800 | { |
2801 | dVAR; | |
0c289d13 | 2802 | MAGIC * const mg = mg_find((const SV *)hv, PERL_MAGIC_rhash); |
ca732855 | 2803 | |
7918f24d NC |
2804 | PERL_ARGS_ASSERT_HV_PLACEHOLDERS_GET; |
2805 | ||
ca732855 NC |
2806 | return mg ? mg->mg_len : 0; |
2807 | } | |
2808 | ||
2809 | void | |
ac1e784a | 2810 | Perl_hv_placeholders_set(pTHX_ HV *hv, I32 ph) |
ca732855 NC |
2811 | { |
2812 | dVAR; | |
ad64d0ec | 2813 | MAGIC * const mg = mg_find((const SV *)hv, PERL_MAGIC_rhash); |
ca732855 | 2814 | |
7918f24d NC |
2815 | PERL_ARGS_ASSERT_HV_PLACEHOLDERS_SET; |
2816 | ||
ca732855 NC |
2817 | if (mg) { |
2818 | mg->mg_len = ph; | |
2819 | } else if (ph) { | |
ad64d0ec | 2820 | if (!sv_magicext(MUTABLE_SV(hv), 0, PERL_MAGIC_rhash, 0, 0, ph)) |
ca732855 NC |
2821 | Perl_die(aTHX_ "panic: hv_placeholders_set"); |
2822 | } | |
2823 | /* else we don't need to add magic to record 0 placeholders. */ | |
2824 | } | |
ecae49c0 | 2825 | |
2a49f0f5 | 2826 | STATIC SV * |
7b0bddfa NC |
2827 | S_refcounted_he_value(pTHX_ const struct refcounted_he *he) |
2828 | { | |
0b2d3faa | 2829 | dVAR; |
7b0bddfa | 2830 | SV *value; |
7918f24d NC |
2831 | |
2832 | PERL_ARGS_ASSERT_REFCOUNTED_HE_VALUE; | |
2833 | ||
7b0bddfa NC |
2834 | switch(he->refcounted_he_data[0] & HVrhek_typemask) { |
2835 | case HVrhek_undef: | |
2836 | value = newSV(0); | |
2837 | break; | |
2838 | case HVrhek_delete: | |
2839 | value = &PL_sv_placeholder; | |
2840 | break; | |
2841 | case HVrhek_IV: | |
44ebaf21 NC |
2842 | value = newSViv(he->refcounted_he_val.refcounted_he_u_iv); |
2843 | break; | |
2844 | case HVrhek_UV: | |
2845 | value = newSVuv(he->refcounted_he_val.refcounted_he_u_uv); | |
7b0bddfa NC |
2846 | break; |
2847 | case HVrhek_PV: | |
44ebaf21 | 2848 | case HVrhek_PV_UTF8: |
7b0bddfa NC |
2849 | /* Create a string SV that directly points to the bytes in our |
2850 | structure. */ | |
b9f83d2f | 2851 | value = newSV_type(SVt_PV); |
7b0bddfa NC |
2852 | SvPV_set(value, (char *) he->refcounted_he_data + 1); |
2853 | SvCUR_set(value, he->refcounted_he_val.refcounted_he_u_len); | |
2854 | /* This stops anything trying to free it */ | |
2855 | SvLEN_set(value, 0); | |
2856 | SvPOK_on(value); | |
2857 | SvREADONLY_on(value); | |
44ebaf21 | 2858 | if ((he->refcounted_he_data[0] & HVrhek_typemask) == HVrhek_PV_UTF8) |
7b0bddfa NC |
2859 | SvUTF8_on(value); |
2860 | break; | |
2861 | default: | |
20439bc7 Z |
2862 | Perl_croak(aTHX_ "panic: refcounted_he_value bad flags %"UVxf, |
2863 | (UV)he->refcounted_he_data[0]); | |
7b0bddfa NC |
2864 | } |
2865 | return value; | |
2866 | } | |
2867 | ||
ecae49c0 | 2868 | /* |
20439bc7 | 2869 | =for apidoc m|HV *|refcounted_he_chain_2hv|const struct refcounted_he *c|U32 flags |
8dff4fc5 | 2870 | |
20439bc7 Z |
2871 | Generates and returns a C<HV *> representing the content of a |
2872 | C<refcounted_he> chain. | |
2873 | I<flags> is currently unused and must be zero. | |
8dff4fc5 BM |
2874 | |
2875 | =cut | |
2876 | */ | |
2877 | HV * | |
20439bc7 | 2878 | Perl_refcounted_he_chain_2hv(pTHX_ const struct refcounted_he *chain, U32 flags) |
8dff4fc5 | 2879 | { |
20439bc7 Z |
2880 | dVAR; |
2881 | HV *hv; | |
2882 | U32 placeholders, max; | |
b3ca2e83 | 2883 | |
20439bc7 Z |
2884 | if (flags) |
2885 | Perl_croak(aTHX_ "panic: refcounted_he_chain_2hv bad flags %"UVxf, | |
2886 | (UV)flags); | |
b3ca2e83 | 2887 | |
b3ca2e83 NC |
2888 | /* We could chase the chain once to get an idea of the number of keys, |
2889 | and call ksplit. But for now we'll make a potentially inefficient | |
2890 | hash with only 8 entries in its array. */ | |
20439bc7 Z |
2891 | hv = newHV(); |
2892 | max = HvMAX(hv); | |
b3ca2e83 NC |
2893 | if (!HvARRAY(hv)) { |
2894 | char *array; | |
2895 | Newxz(array, PERL_HV_ARRAY_ALLOC_BYTES(max + 1), char); | |
2896 | HvARRAY(hv) = (HE**)array; | |
2897 | } | |
2898 | ||
20439bc7 | 2899 | placeholders = 0; |
b3ca2e83 | 2900 | while (chain) { |
cbb1fbea | 2901 | #ifdef USE_ITHREADS |
b6bbf3fa | 2902 | U32 hash = chain->refcounted_he_hash; |
cbb1fbea NC |
2903 | #else |
2904 | U32 hash = HEK_HASH(chain->refcounted_he_hek); | |
2905 | #endif | |
b3ca2e83 NC |
2906 | HE **oentry = &((HvARRAY(hv))[hash & max]); |
2907 | HE *entry = *oentry; | |
b6bbf3fa | 2908 | SV *value; |
cbb1fbea | 2909 | |
b3ca2e83 NC |
2910 | for (; entry; entry = HeNEXT(entry)) { |
2911 | if (HeHASH(entry) == hash) { | |
9f769845 NC |
2912 | /* We might have a duplicate key here. If so, entry is older |
2913 | than the key we've already put in the hash, so if they are | |
2914 | the same, skip adding entry. */ | |
2915 | #ifdef USE_ITHREADS | |
2916 | const STRLEN klen = HeKLEN(entry); | |
2917 | const char *const key = HeKEY(entry); | |
2918 | if (klen == chain->refcounted_he_keylen | |
2919 | && (!!HeKUTF8(entry) | |
2920 | == !!(chain->refcounted_he_data[0] & HVhek_UTF8)) | |
2921 | && memEQ(key, REF_HE_KEY(chain), klen)) | |
2922 | goto next_please; | |
2923 | #else | |
2924 | if (HeKEY_hek(entry) == chain->refcounted_he_hek) | |
2925 | goto next_please; | |
2926 | if (HeKLEN(entry) == HEK_LEN(chain->refcounted_he_hek) | |
2927 | && HeKUTF8(entry) == HEK_UTF8(chain->refcounted_he_hek) | |
2928 | && memEQ(HeKEY(entry), HEK_KEY(chain->refcounted_he_hek), | |
2929 | HeKLEN(entry))) | |
2930 | goto next_please; | |
2931 | #endif | |
b3ca2e83 NC |
2932 | } |
2933 | } | |
2934 | assert (!entry); | |
2935 | entry = new_HE(); | |
2936 | ||
cbb1fbea NC |
2937 | #ifdef USE_ITHREADS |
2938 | HeKEY_hek(entry) | |
7b0bddfa | 2939 | = share_hek_flags(REF_HE_KEY(chain), |
b6bbf3fa NC |
2940 | chain->refcounted_he_keylen, |
2941 | chain->refcounted_he_hash, | |
2942 | (chain->refcounted_he_data[0] | |
2943 | & (HVhek_UTF8|HVhek_WASUTF8))); | |
cbb1fbea | 2944 | #else |
71ad1b0c | 2945 | HeKEY_hek(entry) = share_hek_hek(chain->refcounted_he_hek); |
cbb1fbea | 2946 | #endif |
7b0bddfa NC |
2947 | value = refcounted_he_value(chain); |
2948 | if (value == &PL_sv_placeholder) | |
b3ca2e83 | 2949 | placeholders++; |
b6bbf3fa | 2950 | HeVAL(entry) = value; |
b3ca2e83 NC |
2951 | |
2952 | /* Link it into the chain. */ | |
2953 | HeNEXT(entry) = *oentry; | |
b3ca2e83 NC |
2954 | *oentry = entry; |
2955 | ||
2956 | HvTOTALKEYS(hv)++; | |
2957 | ||
2958 | next_please: | |
71ad1b0c | 2959 | chain = chain->refcounted_he_next; |
b3ca2e83 NC |
2960 | } |
2961 | ||
2962 | if (placeholders) { | |
2963 | clear_placeholders(hv, placeholders); | |
2964 | HvTOTALKEYS(hv) -= placeholders; | |
2965 | } | |
2966 | ||
2967 | /* We could check in the loop to see if we encounter any keys with key | |
2968 | flags, but it's probably not worth it, as this per-hash flag is only | |
2969 | really meant as an optimisation for things like Storable. */ | |
2970 | HvHASKFLAGS_on(hv); | |
def9038f | 2971 | DEBUG_A(Perl_hv_assert(aTHX_ hv)); |
b3ca2e83 NC |
2972 | |
2973 | return hv; | |
2974 | } | |
2975 | ||
20439bc7 Z |
2976 | /* |
2977 | =for apidoc m|SV *|refcounted_he_fetch_pvn|const struct refcounted_he *chain|const char *keypv|STRLEN keylen|U32 hash|U32 flags | |
2978 | ||
2979 | Search along a C<refcounted_he> chain for an entry with the key specified | |
2980 | by I<keypv> and I<keylen>. If I<flags> has the C<REFCOUNTED_HE_KEY_UTF8> | |
2981 | bit set, the key octets are interpreted as UTF-8, otherwise they | |
2982 | are interpreted as Latin-1. I<hash> is a precomputed hash of the key | |
2983 | string, or zero if it has not been precomputed. Returns a mortal scalar | |
2984 | representing the value associated with the key, or C<&PL_sv_placeholder> | |
2985 | if there is no value associated with the key. | |
2986 | ||
2987 | =cut | |
2988 | */ | |
2989 | ||
7b0bddfa | 2990 | SV * |
20439bc7 Z |
2991 | Perl_refcounted_he_fetch_pvn(pTHX_ const struct refcounted_he *chain, |
2992 | const char *keypv, STRLEN keylen, U32 hash, U32 flags) | |
7b0bddfa | 2993 | { |
0b2d3faa | 2994 | dVAR; |
20439bc7 Z |
2995 | U8 utf8_flag; |
2996 | PERL_ARGS_ASSERT_REFCOUNTED_HE_FETCH_PVN; | |
7b0bddfa | 2997 | |
20439bc7 Z |
2998 | if (flags & ~REFCOUNTED_HE_KEY_UTF8) |
2999 | Perl_croak(aTHX_ "panic: refcounted_he_fetch_pvn bad flags %"UVxf, | |
3000 | (UV)flags); | |
3001 | if (!chain) | |
3002 | return &PL_sv_placeholder; | |
3003 | if (flags & REFCOUNTED_HE_KEY_UTF8) { | |
3004 | /* For searching purposes, canonicalise to Latin-1 where possible. */ | |
3005 | const char *keyend = keypv + keylen, *p; | |
3006 | STRLEN nonascii_count = 0; | |
3007 | for (p = keypv; p != keyend; p++) { | |
3008 | U8 c = (U8)*p; | |
3009 | if (c & 0x80) { | |
3010 | if (!((c & 0xfe) == 0xc2 && ++p != keyend && | |
3011 | (((U8)*p) & 0xc0) == 0x80)) | |
3012 | goto canonicalised_key; | |
3013 | nonascii_count++; | |
3014 | } | |
cd1d2f8a | 3015 | } |
20439bc7 Z |
3016 | if (nonascii_count) { |
3017 | char *q; | |
3018 | const char *p = keypv, *keyend = keypv + keylen; | |
3019 | keylen -= nonascii_count; | |
3020 | Newx(q, keylen, char); | |
3021 | SAVEFREEPV(q); | |
3022 | keypv = q; | |
3023 | for (; p != keyend; p++, q++) { | |
3024 | U8 c = (U8)*p; | |
3025 | *q = (char) | |
3026 | ((c & 0x80) ? ((c & 0x03) << 6) | (((U8)*++p) & 0x3f) : c); | |
cd1d2f8a NC |
3027 | } |
3028 | } | |
20439bc7 Z |
3029 | flags &= ~REFCOUNTED_HE_KEY_UTF8; |
3030 | canonicalised_key: ; | |
3031 | } | |
3032 | utf8_flag = (flags & REFCOUNTED_HE_KEY_UTF8) ? HVhek_UTF8 : 0; | |
3033 | if (!hash) | |
3034 | PERL_HASH(hash, keypv, keylen); | |
7b0bddfa | 3035 | |
20439bc7 Z |
3036 | for (; chain; chain = chain->refcounted_he_next) { |
3037 | if ( | |
7b0bddfa | 3038 | #ifdef USE_ITHREADS |
20439bc7 Z |
3039 | hash == chain->refcounted_he_hash && |
3040 | keylen == chain->refcounted_he_keylen && | |
3041 | memEQ(REF_HE_KEY(chain), keypv, keylen) && | |
3042 | utf8_flag == (chain->refcounted_he_data[0] & HVhek_UTF8) | |
7b0bddfa | 3043 | #else |
20439bc7 Z |
3044 | hash == HEK_HASH(chain->refcounted_he_hek) && |
3045 | keylen == (STRLEN)HEK_LEN(chain->refcounted_he_hek) && | |
3046 | memEQ(HEK_KEY(chain->refcounted_he_hek), keypv, keylen) && | |
3047 | utf8_flag == (HEK_FLAGS(chain->refcounted_he_hek) & HVhek_UTF8) | |
7b0bddfa | 3048 | #endif |
20439bc7 Z |
3049 | ) |
3050 | return sv_2mortal(refcounted_he_value(chain)); | |
7b0bddfa | 3051 | } |
20439bc7 Z |
3052 | return &PL_sv_placeholder; |
3053 | } | |
7b0bddfa | 3054 | |
20439bc7 Z |
3055 | /* |
3056 | =for apidoc m|SV *|refcounted_he_fetch_pv|const struct refcounted_he *chain|const char *key|U32 hash|U32 flags | |
7b0bddfa | 3057 | |
20439bc7 Z |
3058 | Like L</refcounted_he_fetch_pvn>, but takes a nul-terminated string |
3059 | instead of a string/length pair. | |
3060 | ||
3061 | =cut | |
3062 | */ | |
3063 | ||
3064 | SV * | |
3065 | Perl_refcounted_he_fetch_pv(pTHX_ const struct refcounted_he *chain, | |
3066 | const char *key, U32 hash, U32 flags) | |
3067 | { | |
3068 | PERL_ARGS_ASSERT_REFCOUNTED_HE_FETCH_PV; | |
3069 | return refcounted_he_fetch_pvn(chain, key, strlen(key), hash, flags); | |
7b0bddfa NC |
3070 | } |
3071 | ||
b3ca2e83 | 3072 | /* |
20439bc7 Z |
3073 | =for apidoc m|SV *|refcounted_he_fetch_sv|const struct refcounted_he *chain|SV *key|U32 hash|U32 flags |
3074 | ||
3075 | Like L</refcounted_he_fetch_pvn>, but takes a Perl scalar instead of a | |
3076 | string/length pair. | |
3077 | ||
3078 | =cut | |
3079 | */ | |
b3ca2e83 | 3080 | |
20439bc7 Z |
3081 | SV * |
3082 | Perl_refcounted_he_fetch_sv(pTHX_ const struct refcounted_he *chain, | |
3083 | SV *key, U32 hash, U32 flags) | |
3084 | { | |
3085 | const char *keypv; | |
3086 | STRLEN keylen; | |
3087 | PERL_ARGS_ASSERT_REFCOUNTED_HE_FETCH_SV; | |
3088 | if (flags & REFCOUNTED_HE_KEY_UTF8) | |
3089 | Perl_croak(aTHX_ "panic: refcounted_he_fetch_sv bad flags %"UVxf, | |
3090 | (UV)flags); | |
3091 | keypv = SvPV_const(key, keylen); | |
3092 | if (SvUTF8(key)) | |
3093 | flags |= REFCOUNTED_HE_KEY_UTF8; | |
3094 | if (!hash && SvIsCOW_shared_hash(key)) | |
3095 | hash = SvSHARED_HASH(key); | |
3096 | return refcounted_he_fetch_pvn(chain, keypv, keylen, hash, flags); | |
3097 | } | |
3098 | ||
3099 | /* | |
3100 | =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 | |
3101 | ||
3102 | Creates a new C<refcounted_he>. This consists of a single key/value | |
3103 | pair and a reference to an existing C<refcounted_he> chain (which may | |
3104 | be empty), and thus forms a longer chain. When using the longer chain, | |
3105 | the new key/value pair takes precedence over any entry for the same key | |
3106 | further along the chain. | |
3107 | ||
3108 | The new key is specified by I<keypv> and I<keylen>. If I<flags> has | |
3109 | the C<REFCOUNTED_HE_KEY_UTF8> bit set, the key octets are interpreted | |
3110 | as UTF-8, otherwise they are interpreted as Latin-1. I<hash> is | |
3111 | a precomputed hash of the key string, or zero if it has not been | |
3112 | precomputed. | |
3113 | ||
3114 | I<value> is the scalar value to store for this key. I<value> is copied | |
3115 | by this function, which thus does not take ownership of any reference | |
3116 | to it, and later changes to the scalar will not be reflected in the | |
3117 | value visible in the C<refcounted_he>. Complex types of scalar will not | |
3118 | be stored with referential integrity, but will be coerced to strings. | |
3119 | I<value> may be either null or C<&PL_sv_placeholder> to indicate that no | |
3120 | value is to be associated with the key; this, as with any non-null value, | |
3121 | takes precedence over the existence of a value for the key further along | |
3122 | the chain. | |
3123 | ||
3124 | I<parent> points to the rest of the C<refcounted_he> chain to be | |
3125 | attached to the new C<refcounted_he>. This function takes ownership | |
3126 | of one reference to I<parent>, and returns one reference to the new | |
3127 | C<refcounted_he>. | |
b3ca2e83 NC |
3128 | |
3129 | =cut | |
3130 | */ | |
3131 | ||
3132 | struct refcounted_he * | |
20439bc7 Z |
3133 | Perl_refcounted_he_new_pvn(pTHX_ struct refcounted_he *parent, |
3134 | const char *keypv, STRLEN keylen, U32 hash, SV *value, U32 flags) | |
3135 | { | |
7a89be66 | 3136 | dVAR; |
b6bbf3fa | 3137 | STRLEN value_len = 0; |
95b63a38 | 3138 | const char *value_p = NULL; |
20439bc7 | 3139 | bool is_pv; |
b6bbf3fa | 3140 | char value_type; |
20439bc7 Z |
3141 | char hekflags; |
3142 | STRLEN key_offset = 1; | |
3143 | struct refcounted_he *he; | |
3144 | PERL_ARGS_ASSERT_REFCOUNTED_HE_NEW_PVN; | |
b6bbf3fa | 3145 | |
20439bc7 Z |
3146 | if (!value || value == &PL_sv_placeholder) { |
3147 | value_type = HVrhek_delete; | |
3148 | } else if (SvPOK(value)) { | |
b6bbf3fa NC |
3149 | value_type = HVrhek_PV; |
3150 | } else if (SvIOK(value)) { | |
ad64d0ec | 3151 | value_type = SvUOK((const SV *)value) ? HVrhek_UV : HVrhek_IV; |
b6bbf3fa NC |
3152 | } else if (!SvOK(value)) { |
3153 | value_type = HVrhek_undef; | |
3154 | } else { | |
3155 | value_type = HVrhek_PV; | |
3156 | } | |
20439bc7 Z |
3157 | is_pv = value_type == HVrhek_PV; |
3158 | if (is_pv) { | |
012da8e5 NC |
3159 | /* Do it this way so that the SvUTF8() test is after the SvPV, in case |
3160 | the value is overloaded, and doesn't yet have the UTF-8flag set. */ | |
b6bbf3fa | 3161 | value_p = SvPV_const(value, value_len); |
012da8e5 NC |
3162 | if (SvUTF8(value)) |
3163 | value_type = HVrhek_PV_UTF8; | |
20439bc7 Z |
3164 | key_offset = value_len + 2; |
3165 | } | |
3166 | hekflags = value_type; | |
3167 | ||
3168 | if (flags & REFCOUNTED_HE_KEY_UTF8) { | |
3169 | /* Canonicalise to Latin-1 where possible. */ | |
3170 | const char *keyend = keypv + keylen, *p; | |
3171 | STRLEN nonascii_count = 0; | |
3172 | for (p = keypv; p != keyend; p++) { | |
3173 | U8 c = (U8)*p; | |
3174 | if (c & 0x80) { | |
3175 | if (!((c & 0xfe) == 0xc2 && ++p != keyend && | |
3176 | (((U8)*p) & 0xc0) == 0x80)) | |
3177 | goto canonicalised_key; | |
3178 | nonascii_count++; | |
3179 | } | |
3180 | } | |
3181 | if (nonascii_count) { | |
3182 | char *q; | |
3183 | const char *p = keypv, *keyend = keypv + keylen; | |
3184 | keylen -= nonascii_count; | |
3185 | Newx(q, keylen, char); | |
3186 | SAVEFREEPV(q); | |
3187 | keypv = q; | |
3188 | for (; p != keyend; p++, q++) { | |
3189 | U8 c = (U8)*p; | |
3190 | *q = (char) | |
3191 | ((c & 0x80) ? ((c & 0x03) << 6) | (((U8)*++p) & 0x3f) : c); | |
3192 | } | |
3193 | } | |
3194 | flags &= ~REFCOUNTED_HE_KEY_UTF8; | |
3195 | canonicalised_key: ; | |
b6bbf3fa | 3196 | } |
20439bc7 Z |
3197 | if (flags & REFCOUNTED_HE_KEY_UTF8) |
3198 | hekflags |= HVhek_UTF8; | |
3199 | if (!hash) | |
3200 | PERL_HASH(hash, keypv, keylen); | |
012da8e5 | 3201 | |
0de694c5 | 3202 | #ifdef USE_ITHREADS |
10edeb5d JH |
3203 | he = (struct refcounted_he*) |
3204 | PerlMemShared_malloc(sizeof(struct refcounted_he) - 1 | |
20439bc7 | 3205 | + keylen |
20439bc7 | 3206 | + key_offset); |
0de694c5 NC |
3207 | #else |
3208 | he = (struct refcounted_he*) | |
3209 | PerlMemShared_malloc(sizeof(struct refcounted_he) - 1 | |
3210 | + key_offset); | |
3211 | #endif | |
b3ca2e83 | 3212 | |
71ad1b0c | 3213 | he->refcounted_he_next = parent; |
b6bbf3fa | 3214 | |
012da8e5 | 3215 | if (is_pv) { |
20439bc7 | 3216 | Copy(value_p, he->refcounted_he_data + 1, value_len + 1, char); |
b6bbf3fa | 3217 | he->refcounted_he_val.refcounted_he_u_len = value_len; |
b6bbf3fa | 3218 | } else if (value_type == HVrhek_IV) { |
20439bc7 | 3219 | he->refcounted_he_val.refcounted_he_u_iv = SvIVX(value); |
012da8e5 | 3220 | } else if (value_type == HVrhek_UV) { |
20439bc7 | 3221 | he->refcounted_he_val.refcounted_he_u_uv = SvUVX(value); |
b6bbf3fa NC |
3222 | } |
3223 | ||
cbb1fbea | 3224 | #ifdef USE_ITHREADS |
b6bbf3fa | 3225 | he->refcounted_he_hash = hash; |
20439bc7 Z |
3226 | he->refcounted_he_keylen = keylen; |
3227 | Copy(keypv, he->refcounted_he_data + key_offset, keylen, char); | |
cbb1fbea | 3228 | #else |
20439bc7 | 3229 | he->refcounted_he_hek = share_hek_flags(keypv, keylen, hash, hekflags); |
cbb1fbea | 3230 | #endif |
b6bbf3fa | 3231 | |
20439bc7 | 3232 | he->refcounted_he_data[0] = hekflags; |
b3ca2e83 NC |
3233 | he->refcounted_he_refcnt = 1; |
3234 | ||
3235 | return he; | |
3236 | } | |
3237 | ||
3238 | /* | |
20439bc7 | 3239 | =for apidoc m|struct refcounted_he *|refcounted_he_new_pv|struct refcounted_he *parent|const char *key|U32 hash|SV *value|U32 flags |
b3ca2e83 | 3240 | |
20439bc7 Z |
3241 | Like L</refcounted_he_new_pvn>, but takes a nul-terminated string instead |
3242 | of a string/length pair. | |
3243 | ||
3244 | =cut | |
3245 | */ | |
3246 | ||
3247 | struct refcounted_he * | |
3248 | Perl_refcounted_he_new_pv(pTHX_ struct refcounted_he *parent, | |
3249 | const char *key, U32 hash, SV *value, U32 flags) | |
3250 | { | |
3251 | PERL_ARGS_ASSERT_REFCOUNTED_HE_NEW_PV; | |
3252 | return refcounted_he_new_pvn(parent, key, strlen(key), hash, value, flags); | |
3253 | } | |
3254 | ||
3255 | /* | |
3256 | =for apidoc m|struct refcounted_he *|refcounted_he_new_sv|struct refcounted_he *parent|SV *key|U32 hash|SV *value|U32 flags | |
3257 | ||
3258 | Like L</refcounted_he_new_pvn>, but takes a Perl scalar instead of a | |
3259 | string/length pair. | |
3260 | ||
3261 | =cut | |
3262 | */ | |
3263 | ||
3264 | struct refcounted_he * | |
3265 | Perl_refcounted_he_new_sv(pTHX_ struct refcounted_he *parent, | |
3266 | SV *key, U32 hash, SV *value, U32 flags) | |
3267 | { | |
3268 | const char *keypv; | |
3269 | STRLEN keylen; | |
3270 | PERL_ARGS_ASSERT_REFCOUNTED_HE_NEW_SV; | |
3271 | if (flags & REFCOUNTED_HE_KEY_UTF8) | |
3272 | Perl_croak(aTHX_ "panic: refcounted_he_new_sv bad flags %"UVxf, | |
3273 | (UV)flags); | |
3274 | keypv = SvPV_const(key, keylen); | |
3275 | if (SvUTF8(key)) | |
3276 | flags |= REFCOUNTED_HE_KEY_UTF8; | |
3277 | if (!hash && SvIsCOW_shared_hash(key)) | |
3278 | hash = SvSHARED_HASH(key); | |
3279 | return refcounted_he_new_pvn(parent, keypv, keylen, hash, value, flags); | |
3280 | } | |
3281 | ||
3282 | /* | |
3283 | =for apidoc m|void|refcounted_he_free|struct refcounted_he *he | |
3284 | ||
3285 | Decrements the reference count of a C<refcounted_he> by one. If the | |
3286 | reference count reaches zero the structure's memory is freed, which | |
3287 | (recursively) causes a reduction of its parent C<refcounted_he>'s | |
3288 | reference count. It is safe to pass a null pointer to this function: | |
3289 | no action occurs in this case. | |
b3ca2e83 NC |
3290 | |
3291 | =cut | |
3292 | */ | |
3293 | ||
3294 | void | |
3295 | Perl_refcounted_he_free(pTHX_ struct refcounted_he *he) { | |
53d44271 | 3296 | dVAR; |
57ca3b03 AL |
3297 | PERL_UNUSED_CONTEXT; |
3298 | ||
b3ca2e83 NC |
3299 | while (he) { |
3300 | struct refcounted_he *copy; | |
cbb1fbea | 3301 | U32 new_count; |
b3ca2e83 | 3302 | |
cbb1fbea NC |
3303 | HINTS_REFCNT_LOCK; |
3304 | new_count = --he->refcounted_he_refcnt; | |
3305 | HINTS_REFCNT_UNLOCK; | |
3306 | ||
3307 | if (new_count) { | |
b3ca2e83 | 3308 | return; |
cbb1fbea | 3309 | } |
b3ca2e83 | 3310 | |
b6bbf3fa | 3311 | #ifndef USE_ITHREADS |
71ad1b0c | 3312 | unshare_hek_or_pvn (he->refcounted_he_hek, 0, 0, 0); |
cbb1fbea | 3313 | #endif |
b3ca2e83 | 3314 | copy = he; |
71ad1b0c | 3315 | he = he->refcounted_he_next; |
b6bbf3fa | 3316 | PerlMemShared_free(copy); |
b3ca2e83 NC |
3317 | } |
3318 | } | |
3319 | ||
20439bc7 Z |
3320 | /* |
3321 | =for apidoc m|struct refcounted_he *|refcounted_he_inc|struct refcounted_he *he | |
3322 | ||
3323 | Increment the reference count of a C<refcounted_he>. The pointer to the | |
3324 | C<refcounted_he> is also returned. It is safe to pass a null pointer | |
3325 | to this function: no action occurs and a null pointer is returned. | |
3326 | ||
3327 | =cut | |
3328 | */ | |
3329 | ||
3330 | struct refcounted_he * | |
3331 | Perl_refcounted_he_inc(pTHX_ struct refcounted_he *he) | |
3332 | { | |
09ddd873 | 3333 | dVAR; |
20439bc7 Z |
3334 | if (he) { |
3335 | HINTS_REFCNT_LOCK; | |
3336 | he->refcounted_he_refcnt++; | |
3337 | HINTS_REFCNT_UNLOCK; | |
3338 | } | |
3339 | return he; | |
3340 | } | |
3341 | ||
8375c93e | 3342 | /* |
aebc0cbe | 3343 | =for apidoc cop_fetch_label |
8375c93e RU |
3344 | |
3345 | Returns the label attached to a cop. | |
3346 | The flags pointer may be set to C<SVf_UTF8> or 0. | |
3347 | ||
3348 | =cut | |
3349 | */ | |
3350 | ||
47550813 NC |
3351 | /* pp_entereval is aware that labels are stored with a key ':' at the top of |
3352 | the linked list. */ | |
dca6062a | 3353 | const char * |
aebc0cbe | 3354 | Perl_cop_fetch_label(pTHX_ COP *const cop, STRLEN *len, U32 *flags) { |
d6747b7a NC |
3355 | struct refcounted_he *const chain = cop->cop_hints_hash; |
3356 | ||
aebc0cbe | 3357 | PERL_ARGS_ASSERT_COP_FETCH_LABEL; |
d6747b7a | 3358 | |
dca6062a NC |
3359 | if (!chain) |
3360 | return NULL; | |
3361 | #ifdef USE_ITHREADS | |
3362 | if (chain->refcounted_he_keylen != 1) | |
3363 | return NULL; | |
3364 | if (*REF_HE_KEY(chain) != ':') | |
3365 | return NULL; | |
3366 | #else | |
3367 | if ((STRLEN)HEK_LEN(chain->refcounted_he_hek) != 1) | |
3368 | return NULL; | |
3369 | if (*HEK_KEY(chain->refcounted_he_hek) != ':') | |
3370 | return NULL; | |
3371 | #endif | |
012da8e5 NC |
3372 | /* Stop anyone trying to really mess us up by adding their own value for |
3373 | ':' into %^H */ | |
3374 | if ((chain->refcounted_he_data[0] & HVrhek_typemask) != HVrhek_PV | |
3375 | && (chain->refcounted_he_data[0] & HVrhek_typemask) != HVrhek_PV_UTF8) | |
3376 | return NULL; | |
3377 | ||
dca6062a NC |
3378 | if (len) |
3379 | *len = chain->refcounted_he_val.refcounted_he_u_len; | |
3380 | if (flags) { | |
3381 | *flags = ((chain->refcounted_he_data[0] & HVrhek_typemask) | |
3382 | == HVrhek_PV_UTF8) ? SVf_UTF8 : 0; | |
3383 | } | |
3384 | return chain->refcounted_he_data + 1; | |
3385 | } | |
3386 | ||
8375c93e | 3387 | /* |
aebc0cbe | 3388 | =for apidoc cop_store_label |
8375c93e RU |
3389 | |
3390 | Save a label into a C<cop_hints_hash>. You need to set flags to C<SVf_UTF8> | |
3391 | for a utf-8 label. | |
3392 | ||
3393 | =cut | |
3394 | */ | |
3395 | ||
a77ac40c | 3396 | void |
aebc0cbe | 3397 | Perl_cop_store_label(pTHX_ COP *const cop, const char *label, STRLEN len, |
a77ac40c | 3398 | U32 flags) |
012da8e5 | 3399 | { |
20439bc7 | 3400 | SV *labelsv; |
aebc0cbe | 3401 | PERL_ARGS_ASSERT_COP_STORE_LABEL; |
547bb267 | 3402 | |
a77ac40c | 3403 | if (flags & ~(SVf_UTF8)) |
aebc0cbe | 3404 | Perl_croak(aTHX_ "panic: cop_store_label illegal flag bits 0x%" UVxf, |
a77ac40c | 3405 | (UV)flags); |
a3179684 | 3406 | labelsv = newSVpvn_flags(label, len, SVs_TEMP); |
20439bc7 Z |
3407 | if (flags & SVf_UTF8) |
3408 | SvUTF8_on(labelsv); | |
a77ac40c | 3409 | cop->cop_hints_hash |
20439bc7 | 3410 | = refcounted_he_new_pvs(cop->cop_hints_hash, ":", labelsv, 0); |
012da8e5 NC |
3411 | } |
3412 | ||
b3ca2e83 | 3413 | /* |
ecae49c0 NC |
3414 | =for apidoc hv_assert |
3415 | ||
3416 | Check that a hash is in an internally consistent state. | |
3417 | ||
3418 | =cut | |
3419 | */ | |
3420 | ||
943795c2 NC |
3421 | #ifdef DEBUGGING |
3422 | ||
ecae49c0 NC |
3423 | void |
3424 | Perl_hv_assert(pTHX_ HV *hv) | |
3425 | { | |
57ca3b03 AL |
3426 | dVAR; |
3427 | HE* entry; | |
3428 | int withflags = 0; | |
3429 | int placeholders = 0; | |
3430 | int real = 0; | |
3431 | int bad = 0; | |
3432 | const I32 riter = HvRITER_get(hv); | |
3433 | HE *eiter = HvEITER_get(hv); | |
3434 | ||
7918f24d NC |
3435 | PERL_ARGS_ASSERT_HV_ASSERT; |
3436 | ||
57ca3b03 AL |
3437 | (void)hv_iterinit(hv); |
3438 | ||
3439 | while ((entry = hv_iternext_flags(hv, HV_ITERNEXT_WANTPLACEHOLDERS))) { | |
3440 | /* sanity check the values */ | |
3441 | if (HeVAL(entry) == &PL_sv_placeholder) | |
3442 | placeholders++; | |
3443 | else | |
3444 | real++; | |
3445 | /* sanity check the keys */ | |
3446 | if (HeSVKEY(entry)) { | |
6f207bd3 | 3447 | NOOP; /* Don't know what to check on SV keys. */ |
57ca3b03 AL |
3448 | } else if (HeKUTF8(entry)) { |
3449 | withflags++; | |
3450 | if (HeKWASUTF8(entry)) { | |
3451 | PerlIO_printf(Perl_debug_log, | |
d2a455e7 | 3452 | "hash key has both WASUTF8 and UTF8: '%.*s'\n", |
57ca3b03 AL |
3453 | (int) HeKLEN(entry), HeKEY(entry)); |
3454 | bad = 1; | |
3455 | } | |
3456 | } else if (HeKWASUTF8(entry)) | |
3457 | withflags++; | |
3458 | } | |
ad64d0ec | 3459 | if (!SvTIED_mg((const SV *)hv, PERL_MAGIC_tied)) { |
57ca3b03 AL |
3460 | static const char bad_count[] = "Count %d %s(s), but hash reports %d\n"; |
3461 | const int nhashkeys = HvUSEDKEYS(hv); | |
3462 | const int nhashplaceholders = HvPLACEHOLDERS_get(hv); | |
3463 | ||
3464 | if (nhashkeys != real) { | |
3465 | PerlIO_printf(Perl_debug_log, bad_count, real, "keys", nhashkeys ); | |
3466 | bad = 1; | |
3467 | } | |
3468 | if (nhashplaceholders != placeholders) { | |
3469 | PerlIO_printf(Perl_debug_log, bad_count, placeholders, "placeholder", nhashplaceholders ); | |
3470 | bad = 1; | |
3471 | } | |
3472 | } | |
3473 | if (withflags && ! HvHASKFLAGS(hv)) { | |
3474 | PerlIO_printf(Perl_debug_log, | |
3475 | "Hash has HASKFLAGS off but I count %d key(s) with flags\n", | |
3476 | withflags); | |
3477 | bad = 1; | |
3478 | } | |
3479 | if (bad) { | |
ad64d0ec | 3480 | sv_dump(MUTABLE_SV(hv)); |
57ca3b03 AL |
3481 | } |
3482 | HvRITER_set(hv, riter); /* Restore hash iterator state */ | |
3483 | HvEITER_set(hv, eiter); | |
ecae49c0 | 3484 | } |
af3babe4 | 3485 | |
943795c2 NC |
3486 | #endif |
3487 | ||
af3babe4 NC |
3488 | /* |
3489 | * Local variables: | |
3490 | * c-indentation-style: bsd | |
3491 | * c-basic-offset: 4 | |
3492 | * indent-tabs-mode: t | |
3493 | * End: | |
3494 | * | |
37442d52 RGS |
3495 | * ex: set ts=8 sts=4 sw=4 noet: |
3496 | */ |