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 | ||
cac9b346 NC |
43 | STATIC void |
44 | S_more_he(pTHX) | |
45 | { | |
97aff369 | 46 | dVAR; |
d8fca402 NC |
47 | /* We could generate this at compile time via (another) auxiliary C |
48 | program? */ | |
49 | const size_t arena_size = Perl_malloc_good_size(PERL_ARENA_SIZE); | |
50 | HE* he = (HE*) Perl_get_arena(aTHX_ arena_size, HE_SVSLOT); | |
51 | HE * const heend = &he[arena_size / sizeof(HE) - 1]; | |
cac9b346 | 52 | |
d2a0f284 | 53 | PL_body_roots[HE_SVSLOT] = he; |
cac9b346 NC |
54 | while (he < heend) { |
55 | HeNEXT(he) = (HE*)(he + 1); | |
56 | he++; | |
57 | } | |
58 | HeNEXT(he) = 0; | |
59 | } | |
60 | ||
c941fb51 NC |
61 | #ifdef PURIFY |
62 | ||
63 | #define new_HE() (HE*)safemalloc(sizeof(HE)) | |
64 | #define del_HE(p) safefree((char*)p) | |
65 | ||
66 | #else | |
67 | ||
76e3520e | 68 | STATIC HE* |
cea2e8a9 | 69 | S_new_he(pTHX) |
4633a7c4 | 70 | { |
97aff369 | 71 | dVAR; |
4633a7c4 | 72 | HE* he; |
0bd48802 | 73 | void ** const root = &PL_body_roots[HE_SVSLOT]; |
6a93a7e5 | 74 | |
6a93a7e5 | 75 | if (!*root) |
cac9b346 | 76 | S_more_he(aTHX); |
10edeb5d | 77 | he = (HE*) *root; |
ce3e5c45 | 78 | assert(he); |
6a93a7e5 | 79 | *root = HeNEXT(he); |
333f433b | 80 | return he; |
4633a7c4 LW |
81 | } |
82 | ||
c941fb51 NC |
83 | #define new_HE() new_he() |
84 | #define del_HE(p) \ | |
85 | STMT_START { \ | |
6a93a7e5 NC |
86 | HeNEXT(p) = (HE*)(PL_body_roots[HE_SVSLOT]); \ |
87 | PL_body_roots[HE_SVSLOT] = p; \ | |
c941fb51 | 88 | } STMT_END |
d33b2eba | 89 | |
d33b2eba | 90 | |
d33b2eba GS |
91 | |
92 | #endif | |
93 | ||
76e3520e | 94 | STATIC HEK * |
5f66b61c | 95 | S_save_hek_flags(const char *str, I32 len, U32 hash, int flags) |
bbce6d69 | 96 | { |
35a4481c | 97 | const int flags_masked = flags & HVhek_MASK; |
bbce6d69 | 98 | char *k; |
99 | register HEK *hek; | |
1c846c1f | 100 | |
7918f24d NC |
101 | PERL_ARGS_ASSERT_SAVE_HEK_FLAGS; |
102 | ||
a02a5408 | 103 | Newx(k, HEK_BASESIZE + len + 2, char); |
bbce6d69 | 104 | hek = (HEK*)k; |
ff68c719 | 105 | Copy(str, HEK_KEY(hek), len, char); |
e05949c7 | 106 | HEK_KEY(hek)[len] = 0; |
ff68c719 | 107 | HEK_LEN(hek) = len; |
108 | HEK_HASH(hek) = hash; | |
45e34800 | 109 | HEK_FLAGS(hek) = (unsigned char)flags_masked | HVhek_UNSHARED; |
dcf933a4 NC |
110 | |
111 | if (flags & HVhek_FREEKEY) | |
112 | Safefree(str); | |
bbce6d69 | 113 | return hek; |
114 | } | |
115 | ||
4a31713e | 116 | /* free the pool of temporary HE/HEK pairs returned by hv_fetch_ent |
dd28f7bb DM |
117 | * for tied hashes */ |
118 | ||
119 | void | |
120 | Perl_free_tied_hv_pool(pTHX) | |
121 | { | |
97aff369 | 122 | dVAR; |
dd28f7bb DM |
123 | HE *he = PL_hv_fetch_ent_mh; |
124 | while (he) { | |
9d4ba2ae | 125 | HE * const ohe = he; |
dd28f7bb | 126 | Safefree(HeKEY_hek(he)); |
dd28f7bb DM |
127 | he = HeNEXT(he); |
128 | del_HE(ohe); | |
129 | } | |
4608196e | 130 | PL_hv_fetch_ent_mh = NULL; |
dd28f7bb DM |
131 | } |
132 | ||
d18c6117 | 133 | #if defined(USE_ITHREADS) |
0bff533c NC |
134 | HEK * |
135 | Perl_hek_dup(pTHX_ HEK *source, CLONE_PARAMS* param) | |
136 | { | |
566771cc | 137 | HEK *shared; |
9d4ba2ae | 138 | |
7918f24d | 139 | PERL_ARGS_ASSERT_HEK_DUP; |
9d4ba2ae | 140 | PERL_UNUSED_ARG(param); |
0bff533c | 141 | |
566771cc NC |
142 | if (!source) |
143 | return NULL; | |
144 | ||
145 | shared = (HEK*)ptr_table_fetch(PL_ptr_table, source); | |
0bff533c NC |
146 | if (shared) { |
147 | /* We already shared this hash key. */ | |
454f1e26 | 148 | (void)share_hek_hek(shared); |
0bff533c NC |
149 | } |
150 | else { | |
658b4a4a | 151 | shared |
6e838c70 NC |
152 | = share_hek_flags(HEK_KEY(source), HEK_LEN(source), |
153 | HEK_HASH(source), HEK_FLAGS(source)); | |
658b4a4a | 154 | ptr_table_store(PL_ptr_table, source, shared); |
0bff533c | 155 | } |
658b4a4a | 156 | return shared; |
0bff533c NC |
157 | } |
158 | ||
d18c6117 | 159 | HE * |
5c4138a0 | 160 | Perl_he_dup(pTHX_ const HE *e, bool shared, CLONE_PARAMS* param) |
d18c6117 GS |
161 | { |
162 | HE *ret; | |
163 | ||
7918f24d NC |
164 | PERL_ARGS_ASSERT_HE_DUP; |
165 | ||
d18c6117 | 166 | if (!e) |
4608196e | 167 | return NULL; |
7766f137 GS |
168 | /* look for it in the table first */ |
169 | ret = (HE*)ptr_table_fetch(PL_ptr_table, e); | |
170 | if (ret) | |
171 | return ret; | |
172 | ||
173 | /* create anew and remember what it is */ | |
d33b2eba | 174 | ret = new_HE(); |
7766f137 GS |
175 | ptr_table_store(PL_ptr_table, e, ret); |
176 | ||
d2d73c3e | 177 | HeNEXT(ret) = he_dup(HeNEXT(e),shared, param); |
dd28f7bb DM |
178 | if (HeKLEN(e) == HEf_SVKEY) { |
179 | char *k; | |
ad64d0ec | 180 | Newx(k, HEK_BASESIZE + sizeof(const SV *), char); |
dd28f7bb | 181 | HeKEY_hek(ret) = (HEK*)k; |
a09252eb | 182 | HeKEY_sv(ret) = sv_dup_inc(HeKEY_sv(e), param); |
dd28f7bb | 183 | } |
c21d1a0f | 184 | else if (shared) { |
0bff533c NC |
185 | /* This is hek_dup inlined, which seems to be important for speed |
186 | reasons. */ | |
1b6737cc | 187 | HEK * const source = HeKEY_hek(e); |
658b4a4a | 188 | HEK *shared = (HEK*)ptr_table_fetch(PL_ptr_table, source); |
c21d1a0f NC |
189 | |
190 | if (shared) { | |
191 | /* We already shared this hash key. */ | |
454f1e26 | 192 | (void)share_hek_hek(shared); |
c21d1a0f NC |
193 | } |
194 | else { | |
658b4a4a | 195 | shared |
6e838c70 NC |
196 | = share_hek_flags(HEK_KEY(source), HEK_LEN(source), |
197 | HEK_HASH(source), HEK_FLAGS(source)); | |
658b4a4a | 198 | ptr_table_store(PL_ptr_table, source, shared); |
c21d1a0f | 199 | } |
658b4a4a | 200 | HeKEY_hek(ret) = shared; |
c21d1a0f | 201 | } |
d18c6117 | 202 | else |
19692e8d NC |
203 | HeKEY_hek(ret) = save_hek_flags(HeKEY(e), HeKLEN(e), HeHASH(e), |
204 | HeKFLAGS(e)); | |
a09252eb | 205 | HeVAL(ret) = sv_dup_inc(HeVAL(e), param); |
d18c6117 GS |
206 | return ret; |
207 | } | |
208 | #endif /* USE_ITHREADS */ | |
209 | ||
1b1f1335 | 210 | static void |
2393f1b9 JH |
211 | S_hv_notallowed(pTHX_ int flags, const char *key, I32 klen, |
212 | const char *msg) | |
1b1f1335 | 213 | { |
1b6737cc | 214 | SV * const sv = sv_newmortal(); |
7918f24d NC |
215 | |
216 | PERL_ARGS_ASSERT_HV_NOTALLOWED; | |
217 | ||
19692e8d | 218 | if (!(flags & HVhek_FREEKEY)) { |
1b1f1335 NIS |
219 | sv_setpvn(sv, key, klen); |
220 | } | |
221 | else { | |
222 | /* Need to free saved eventually assign to mortal SV */ | |
34c3c4e3 | 223 | /* XXX is this line an error ???: SV *sv = sv_newmortal(); */ |
1b1f1335 NIS |
224 | sv_usepvn(sv, (char *) key, klen); |
225 | } | |
19692e8d | 226 | if (flags & HVhek_UTF8) { |
1b1f1335 NIS |
227 | SvUTF8_on(sv); |
228 | } | |
be2597df | 229 | Perl_croak(aTHX_ msg, SVfARG(sv)); |
1b1f1335 NIS |
230 | } |
231 | ||
fde52b5c | 232 | /* (klen == HEf_SVKEY) is special for MAGICAL hv entries, meaning key slot |
233 | * contains an SV* */ | |
234 | ||
34a6f7b4 NC |
235 | /* |
236 | =for apidoc hv_store | |
237 | ||
238 | Stores an SV in a hash. The hash key is specified as C<key> and C<klen> is | |
239 | the length of the key. The C<hash> parameter is the precomputed hash | |
240 | value; if it is zero then Perl will compute it. The return value will be | |
241 | NULL if the operation failed or if the value did not need to be actually | |
242 | stored within the hash (as in the case of tied hashes). Otherwise it can | |
243 | be dereferenced to get the original C<SV*>. Note that the caller is | |
244 | responsible for suitably incrementing the reference count of C<val> before | |
245 | the call, and decrementing it if the function returned NULL. Effectively | |
246 | a successful hv_store takes ownership of one reference to C<val>. This is | |
247 | usually what you want; a newly created SV has a reference count of one, so | |
248 | if all your code does is create SVs then store them in a hash, hv_store | |
249 | will own the only reference to the new SV, and your code doesn't need to do | |
250 | anything further to tidy up. hv_store is not implemented as a call to | |
251 | hv_store_ent, and does not create a temporary SV for the key, so if your | |
252 | key data is not already in SV form then use hv_store in preference to | |
253 | hv_store_ent. | |
254 | ||
255 | See L<perlguts/"Understanding the Magic of Tied Hashes and Arrays"> for more | |
256 | information on how to use this function on tied hashes. | |
257 | ||
34a6f7b4 NC |
258 | =for apidoc hv_store_ent |
259 | ||
260 | Stores C<val> in a hash. The hash key is specified as C<key>. The C<hash> | |
261 | parameter is the precomputed hash value; if it is zero then Perl will | |
262 | compute it. The return value is the new hash entry so created. It will be | |
263 | NULL if the operation failed or if the value did not need to be actually | |
264 | stored within the hash (as in the case of tied hashes). Otherwise the | |
265 | contents of the return value can be accessed using the C<He?> macros | |
266 | described here. Note that the caller is responsible for suitably | |
267 | incrementing the reference count of C<val> before the call, and | |
268 | decrementing it if the function returned NULL. Effectively a successful | |
269 | hv_store_ent takes ownership of one reference to C<val>. This is | |
270 | usually what you want; a newly created SV has a reference count of one, so | |
271 | if all your code does is create SVs then store them in a hash, hv_store | |
272 | will own the only reference to the new SV, and your code doesn't need to do | |
273 | anything further to tidy up. Note that hv_store_ent only reads the C<key>; | |
274 | unlike C<val> it does not take ownership of it, so maintaining the correct | |
275 | reference count on C<key> is entirely the caller's responsibility. hv_store | |
276 | is not implemented as a call to hv_store_ent, and does not create a temporary | |
277 | SV for the key, so if your key data is not already in SV form then use | |
278 | hv_store in preference to hv_store_ent. | |
279 | ||
280 | See L<perlguts/"Understanding the Magic of Tied Hashes and Arrays"> for more | |
281 | information on how to use this function on tied hashes. | |
282 | ||
34a6f7b4 NC |
283 | =for apidoc hv_exists |
284 | ||
285 | Returns a boolean indicating whether the specified hash key exists. The | |
286 | C<klen> is the length of the key. | |
287 | ||
954c1994 GS |
288 | =for apidoc hv_fetch |
289 | ||
290 | Returns the SV which corresponds to the specified key in the hash. The | |
291 | C<klen> is the length of the key. If C<lval> is set then the fetch will be | |
292 | part of a store. Check that the return value is non-null before | |
d1be9408 | 293 | dereferencing it to an C<SV*>. |
954c1994 | 294 | |
96f1132b | 295 | See L<perlguts/"Understanding the Magic of Tied Hashes and Arrays"> for more |
954c1994 GS |
296 | information on how to use this function on tied hashes. |
297 | ||
34a6f7b4 NC |
298 | =for apidoc hv_exists_ent |
299 | ||
300 | Returns a boolean indicating whether the specified hash key exists. C<hash> | |
301 | can be a valid precomputed hash value, or 0 to ask for it to be | |
302 | computed. | |
303 | ||
304 | =cut | |
305 | */ | |
306 | ||
d1be9408 | 307 | /* returns an HE * structure with the all fields set */ |
fde52b5c | 308 | /* note that hent_val will be a mortal sv for MAGICAL hashes */ |
954c1994 GS |
309 | /* |
310 | =for apidoc hv_fetch_ent | |
311 | ||
312 | Returns the hash entry which corresponds to the specified key in the hash. | |
313 | C<hash> must be a valid precomputed hash number for the given C<key>, or 0 | |
314 | if you want the function to compute it. IF C<lval> is set then the fetch | |
315 | will be part of a store. Make sure the return value is non-null before | |
316 | accessing it. The return value when C<tb> is a tied hash is a pointer to a | |
317 | static location, so be sure to make a copy of the structure if you need to | |
1c846c1f | 318 | store it somewhere. |
954c1994 | 319 | |
96f1132b | 320 | See L<perlguts/"Understanding the Magic of Tied Hashes and Arrays"> for more |
954c1994 GS |
321 | information on how to use this function on tied hashes. |
322 | ||
323 | =cut | |
324 | */ | |
325 | ||
a038e571 NC |
326 | /* Common code for hv_delete()/hv_exists()/hv_fetch()/hv_store() */ |
327 | void * | |
328 | Perl_hv_common_key_len(pTHX_ HV *hv, const char *key, I32 klen_i32, | |
329 | const int action, SV *val, const U32 hash) | |
330 | { | |
331 | STRLEN klen; | |
332 | int flags; | |
333 | ||
7918f24d NC |
334 | PERL_ARGS_ASSERT_HV_COMMON_KEY_LEN; |
335 | ||
a038e571 NC |
336 | if (klen_i32 < 0) { |
337 | klen = -klen_i32; | |
338 | flags = HVhek_UTF8; | |
339 | } else { | |
340 | klen = klen_i32; | |
341 | flags = 0; | |
342 | } | |
343 | return hv_common(hv, NULL, key, klen, flags, action, val, hash); | |
344 | } | |
345 | ||
63c89345 | 346 | void * |
d3ba3f5c NC |
347 | Perl_hv_common(pTHX_ HV *hv, SV *keysv, const char *key, STRLEN klen, |
348 | int flags, int action, SV *val, register U32 hash) | |
113738bb | 349 | { |
27da23d5 | 350 | dVAR; |
b2c64049 | 351 | XPVHV* xhv; |
b2c64049 NC |
352 | HE *entry; |
353 | HE **oentry; | |
fde52b5c | 354 | SV *sv; |
da58a35d | 355 | bool is_utf8; |
113738bb | 356 | int masked_flags; |
3c84c864 | 357 | const int return_svp = action & HV_FETCH_JUST_SV; |
fde52b5c | 358 | |
359 | if (!hv) | |
a4fc7abc | 360 | return NULL; |
8265e3d1 NC |
361 | if (SvTYPE(hv) == SVTYPEMASK) |
362 | return NULL; | |
363 | ||
364 | assert(SvTYPE(hv) == SVt_PVHV); | |
fde52b5c | 365 | |
bdee33e4 | 366 | if (SvSMAGICAL(hv) && SvGMAGICAL(hv) && !(action & HV_DISABLE_UVAR_XKEY)) { |
fda2d18a | 367 | MAGIC* mg; |
ad64d0ec | 368 | if ((mg = mg_find((const SV *)hv, PERL_MAGIC_uvar))) { |
fda2d18a NC |
369 | struct ufuncs * const uf = (struct ufuncs *)mg->mg_ptr; |
370 | if (uf->uf_set == NULL) { | |
371 | SV* obj = mg->mg_obj; | |
372 | ||
373 | if (!keysv) { | |
59cd0e26 NC |
374 | keysv = newSVpvn_flags(key, klen, SVs_TEMP | |
375 | ((flags & HVhek_UTF8) | |
376 | ? SVf_UTF8 : 0)); | |
fda2d18a NC |
377 | } |
378 | ||
379 | mg->mg_obj = keysv; /* pass key */ | |
380 | uf->uf_index = action; /* pass action */ | |
ad64d0ec | 381 | magic_getuvar(MUTABLE_SV(hv), mg); |
fda2d18a NC |
382 | keysv = mg->mg_obj; /* may have changed */ |
383 | mg->mg_obj = obj; | |
384 | ||
385 | /* If the key may have changed, then we need to invalidate | |
386 | any passed-in computed hash value. */ | |
387 | hash = 0; | |
388 | } | |
389 | } | |
bdee33e4 | 390 | } |
113738bb | 391 | if (keysv) { |
e593d2fe AE |
392 | if (flags & HVhek_FREEKEY) |
393 | Safefree(key); | |
5c144d81 | 394 | key = SvPV_const(keysv, klen); |
113738bb | 395 | is_utf8 = (SvUTF8(keysv) != 0); |
44b87b50 NC |
396 | if (SvIsCOW_shared_hash(keysv)) { |
397 | flags = HVhek_KEYCANONICAL | (is_utf8 ? HVhek_UTF8 : 0); | |
398 | } else { | |
399 | flags = 0; | |
400 | } | |
113738bb | 401 | } else { |
c1fe5510 | 402 | is_utf8 = ((flags & HVhek_UTF8) ? TRUE : FALSE); |
113738bb | 403 | } |
113738bb | 404 | |
9dbc5603 | 405 | if (action & HV_DELETE) { |
3c84c864 NC |
406 | return (void *) hv_delete_common(hv, keysv, key, klen, |
407 | flags | (is_utf8 ? HVhek_UTF8 : 0), | |
408 | action, hash); | |
9dbc5603 NC |
409 | } |
410 | ||
b2c64049 | 411 | xhv = (XPVHV*)SvANY(hv); |
7f66fda2 | 412 | if (SvMAGICAL(hv)) { |
6136c704 | 413 | if (SvRMAGICAL(hv) && !(action & (HV_FETCH_ISSTORE|HV_FETCH_ISEXISTS))) { |
ad64d0ec NC |
414 | if (mg_find((const SV *)hv, PERL_MAGIC_tied) |
415 | || SvGMAGICAL((const SV *)hv)) | |
e62cc96a | 416 | { |
3c84c864 | 417 | /* FIXME should be able to skimp on the HE/HEK here when |
7f66fda2 | 418 | HV_FETCH_JUST_SV is true. */ |
7f66fda2 | 419 | if (!keysv) { |
740cce10 NC |
420 | keysv = newSVpvn_utf8(key, klen, is_utf8); |
421 | } else { | |
7f66fda2 | 422 | keysv = newSVsv(keysv); |
113738bb | 423 | } |
44a2ac75 | 424 | sv = sv_newmortal(); |
ad64d0ec | 425 | mg_copy(MUTABLE_SV(hv), sv, (char *)keysv, HEf_SVKEY); |
7f66fda2 NC |
426 | |
427 | /* grab a fake HE/HEK pair from the pool or make a new one */ | |
428 | entry = PL_hv_fetch_ent_mh; | |
429 | if (entry) | |
430 | PL_hv_fetch_ent_mh = HeNEXT(entry); | |
431 | else { | |
432 | char *k; | |
433 | entry = new_HE(); | |
ad64d0ec | 434 | Newx(k, HEK_BASESIZE + sizeof(const SV *), char); |
7f66fda2 NC |
435 | HeKEY_hek(entry) = (HEK*)k; |
436 | } | |
4608196e | 437 | HeNEXT(entry) = NULL; |
7f66fda2 NC |
438 | HeSVKEY_set(entry, keysv); |
439 | HeVAL(entry) = sv; | |
440 | sv_upgrade(sv, SVt_PVLV); | |
441 | LvTYPE(sv) = 'T'; | |
442 | /* so we can free entry when freeing sv */ | |
ad64d0ec | 443 | LvTARG(sv) = MUTABLE_SV(entry); |
7f66fda2 NC |
444 | |
445 | /* XXX remove at some point? */ | |
446 | if (flags & HVhek_FREEKEY) | |
447 | Safefree(key); | |
448 | ||
3c84c864 NC |
449 | if (return_svp) { |
450 | return entry ? (void *) &HeVAL(entry) : NULL; | |
451 | } | |
452 | return (void *) entry; | |
113738bb | 453 | } |
7f66fda2 | 454 | #ifdef ENV_IS_CASELESS |
ad64d0ec | 455 | else if (mg_find((const SV *)hv, PERL_MAGIC_env)) { |
7f66fda2 NC |
456 | U32 i; |
457 | for (i = 0; i < klen; ++i) | |
458 | if (isLOWER(key[i])) { | |
086cb327 NC |
459 | /* Would be nice if we had a routine to do the |
460 | copy and upercase in a single pass through. */ | |
0bd48802 | 461 | const char * const nkey = strupr(savepvn(key,klen)); |
086cb327 NC |
462 | /* Note that this fetch is for nkey (the uppercased |
463 | key) whereas the store is for key (the original) */ | |
63c89345 NC |
464 | void *result = hv_common(hv, NULL, nkey, klen, |
465 | HVhek_FREEKEY, /* free nkey */ | |
466 | 0 /* non-LVAL fetch */ | |
3c84c864 NC |
467 | | HV_DISABLE_UVAR_XKEY |
468 | | return_svp, | |
63c89345 NC |
469 | NULL /* no value */, |
470 | 0 /* compute hash */); | |
26488bcf | 471 | if (!result && (action & HV_FETCH_LVALUE)) { |
086cb327 NC |
472 | /* This call will free key if necessary. |
473 | Do it this way to encourage compiler to tail | |
474 | call optimise. */ | |
63c89345 NC |
475 | result = hv_common(hv, keysv, key, klen, flags, |
476 | HV_FETCH_ISSTORE | |
3c84c864 NC |
477 | | HV_DISABLE_UVAR_XKEY |
478 | | return_svp, | |
63c89345 | 479 | newSV(0), hash); |
086cb327 NC |
480 | } else { |
481 | if (flags & HVhek_FREEKEY) | |
482 | Safefree(key); | |
483 | } | |
63c89345 | 484 | return result; |
7f66fda2 | 485 | } |
902173a3 | 486 | } |
7f66fda2 NC |
487 | #endif |
488 | } /* ISFETCH */ | |
489 | else if (SvRMAGICAL(hv) && (action & HV_FETCH_ISEXISTS)) { | |
ad64d0ec NC |
490 | if (mg_find((const SV *)hv, PERL_MAGIC_tied) |
491 | || SvGMAGICAL((const SV *)hv)) { | |
b2c64049 NC |
492 | /* I don't understand why hv_exists_ent has svret and sv, |
493 | whereas hv_exists only had one. */ | |
9d4ba2ae | 494 | SV * const svret = sv_newmortal(); |
b2c64049 | 495 | sv = sv_newmortal(); |
7f66fda2 NC |
496 | |
497 | if (keysv || is_utf8) { | |
498 | if (!keysv) { | |
740cce10 | 499 | keysv = newSVpvn_utf8(key, klen, TRUE); |
7f66fda2 NC |
500 | } else { |
501 | keysv = newSVsv(keysv); | |
502 | } | |
ad64d0ec | 503 | mg_copy(MUTABLE_SV(hv), sv, (char *)sv_2mortal(keysv), HEf_SVKEY); |
b2c64049 | 504 | } else { |
ad64d0ec | 505 | mg_copy(MUTABLE_SV(hv), sv, key, klen); |
7f66fda2 | 506 | } |
b2c64049 NC |
507 | if (flags & HVhek_FREEKEY) |
508 | Safefree(key); | |
7f66fda2 NC |
509 | magic_existspack(svret, mg_find(sv, PERL_MAGIC_tiedelem)); |
510 | /* This cast somewhat evil, but I'm merely using NULL/ | |
511 | not NULL to return the boolean exists. | |
512 | And I know hv is not NULL. */ | |
3c84c864 | 513 | return SvTRUE(svret) ? (void *)hv : NULL; |
e7152ba2 | 514 | } |
7f66fda2 | 515 | #ifdef ENV_IS_CASELESS |
ad64d0ec | 516 | else if (mg_find((const SV *)hv, PERL_MAGIC_env)) { |
7f66fda2 | 517 | /* XXX This code isn't UTF8 clean. */ |
a15d23f8 | 518 | char * const keysave = (char * const)key; |
b2c64049 NC |
519 | /* Will need to free this, so set FREEKEY flag. */ |
520 | key = savepvn(key,klen); | |
521 | key = (const char*)strupr((char*)key); | |
6136c704 | 522 | is_utf8 = FALSE; |
7f66fda2 | 523 | hash = 0; |
8b4f7dd5 | 524 | keysv = 0; |
b2c64049 NC |
525 | |
526 | if (flags & HVhek_FREEKEY) { | |
527 | Safefree(keysave); | |
528 | } | |
529 | flags |= HVhek_FREEKEY; | |
7f66fda2 | 530 | } |
902173a3 | 531 | #endif |
7f66fda2 | 532 | } /* ISEXISTS */ |
b2c64049 NC |
533 | else if (action & HV_FETCH_ISSTORE) { |
534 | bool needs_copy; | |
535 | bool needs_store; | |
536 | hv_magic_check (hv, &needs_copy, &needs_store); | |
537 | if (needs_copy) { | |
a3b680e6 | 538 | const bool save_taint = PL_tainted; |
b2c64049 NC |
539 | if (keysv || is_utf8) { |
540 | if (!keysv) { | |
740cce10 | 541 | keysv = newSVpvn_utf8(key, klen, TRUE); |
b2c64049 NC |
542 | } |
543 | if (PL_tainting) | |
544 | PL_tainted = SvTAINTED(keysv); | |
545 | keysv = sv_2mortal(newSVsv(keysv)); | |
ad64d0ec | 546 | mg_copy(MUTABLE_SV(hv), val, (char*)keysv, HEf_SVKEY); |
b2c64049 | 547 | } else { |
ad64d0ec | 548 | mg_copy(MUTABLE_SV(hv), val, key, klen); |
b2c64049 NC |
549 | } |
550 | ||
551 | TAINT_IF(save_taint); | |
1baaf5d7 | 552 | if (!needs_store) { |
b2c64049 NC |
553 | if (flags & HVhek_FREEKEY) |
554 | Safefree(key); | |
4608196e | 555 | return NULL; |
b2c64049 NC |
556 | } |
557 | #ifdef ENV_IS_CASELESS | |
ad64d0ec | 558 | else if (mg_find((const SV *)hv, PERL_MAGIC_env)) { |
b2c64049 NC |
559 | /* XXX This code isn't UTF8 clean. */ |
560 | const char *keysave = key; | |
561 | /* Will need to free this, so set FREEKEY flag. */ | |
562 | key = savepvn(key,klen); | |
563 | key = (const char*)strupr((char*)key); | |
6136c704 | 564 | is_utf8 = FALSE; |
b2c64049 | 565 | hash = 0; |
8b4f7dd5 | 566 | keysv = 0; |
b2c64049 NC |
567 | |
568 | if (flags & HVhek_FREEKEY) { | |
569 | Safefree(keysave); | |
570 | } | |
571 | flags |= HVhek_FREEKEY; | |
572 | } | |
573 | #endif | |
574 | } | |
575 | } /* ISSTORE */ | |
7f66fda2 | 576 | } /* SvMAGICAL */ |
fde52b5c | 577 | |
7b2c381c | 578 | if (!HvARRAY(hv)) { |
b2c64049 | 579 | if ((action & (HV_FETCH_LVALUE | HV_FETCH_ISSTORE)) |
fde52b5c | 580 | #ifdef DYNAMIC_ENV_FETCH /* if it's an %ENV lookup, we may get it on the fly */ |
ad64d0ec NC |
581 | || (SvRMAGICAL((const SV *)hv) |
582 | && mg_find((const SV *)hv, PERL_MAGIC_env)) | |
fde52b5c | 583 | #endif |
d58e6666 NC |
584 | ) { |
585 | char *array; | |
a02a5408 | 586 | Newxz(array, |
cbec9347 | 587 | PERL_HV_ARRAY_ALLOC_BYTES(xhv->xhv_max+1 /* HvMAX(hv)+1 */), |
d58e6666 NC |
588 | char); |
589 | HvARRAY(hv) = (HE**)array; | |
590 | } | |
7f66fda2 NC |
591 | #ifdef DYNAMIC_ENV_FETCH |
592 | else if (action & HV_FETCH_ISEXISTS) { | |
593 | /* for an %ENV exists, if we do an insert it's by a recursive | |
594 | store call, so avoid creating HvARRAY(hv) right now. */ | |
595 | } | |
596 | #endif | |
113738bb NC |
597 | else { |
598 | /* XXX remove at some point? */ | |
599 | if (flags & HVhek_FREEKEY) | |
600 | Safefree(key); | |
601 | ||
3c84c864 | 602 | return NULL; |
113738bb | 603 | } |
fde52b5c | 604 | } |
605 | ||
44b87b50 | 606 | if (is_utf8 & !(flags & HVhek_KEYCANONICAL)) { |
41d88b63 | 607 | char * const keysave = (char *)key; |
f9a63242 | 608 | key = (char*)bytes_from_utf8((U8*)key, &klen, &is_utf8); |
19692e8d | 609 | if (is_utf8) |
c1fe5510 NC |
610 | flags |= HVhek_UTF8; |
611 | else | |
612 | flags &= ~HVhek_UTF8; | |
7f66fda2 NC |
613 | if (key != keysave) { |
614 | if (flags & HVhek_FREEKEY) | |
615 | Safefree(keysave); | |
19692e8d | 616 | flags |= HVhek_WASUTF8 | HVhek_FREEKEY; |
527df579 NC |
617 | /* If the caller calculated a hash, it was on the sequence of |
618 | octets that are the UTF-8 form. We've now changed the sequence | |
619 | of octets stored to that of the equivalent byte representation, | |
620 | so the hash we need is different. */ | |
621 | hash = 0; | |
7f66fda2 | 622 | } |
19692e8d | 623 | } |
f9a63242 | 624 | |
4b5190b5 NC |
625 | if (HvREHASH(hv)) { |
626 | PERL_HASH_INTERNAL(hash, key, klen); | |
b2c64049 NC |
627 | /* We don't have a pointer to the hv, so we have to replicate the |
628 | flag into every HEK, so that hv_iterkeysv can see it. */ | |
629 | /* And yes, you do need this even though you are not "storing" because | |
fdcd69b6 NC |
630 | you can flip the flags below if doing an lval lookup. (And that |
631 | was put in to give the semantics Andreas was expecting.) */ | |
632 | flags |= HVhek_REHASH; | |
4b5190b5 | 633 | } else if (!hash) { |
113738bb | 634 | if (keysv && (SvIsCOW_shared_hash(keysv))) { |
c158a4fd | 635 | hash = SvSHARED_HASH(keysv); |
46187eeb NC |
636 | } else { |
637 | PERL_HASH(hash, key, klen); | |
638 | } | |
639 | } | |
effa1e2d | 640 | |
113738bb NC |
641 | masked_flags = (flags & HVhek_MASK); |
642 | ||
7f66fda2 | 643 | #ifdef DYNAMIC_ENV_FETCH |
4608196e | 644 | if (!HvARRAY(hv)) entry = NULL; |
7f66fda2 NC |
645 | else |
646 | #endif | |
b2c64049 | 647 | { |
7b2c381c | 648 | entry = (HvARRAY(hv))[hash & (I32) HvMAX(hv)]; |
b2c64049 | 649 | } |
0298d7b9 | 650 | for (; entry; entry = HeNEXT(entry)) { |
fde52b5c | 651 | if (HeHASH(entry) != hash) /* strings can't be equal */ |
652 | continue; | |
eb160463 | 653 | if (HeKLEN(entry) != (I32)klen) |
fde52b5c | 654 | continue; |
1c846c1f | 655 | if (HeKEY(entry) != key && memNE(HeKEY(entry),key,klen)) /* is this it? */ |
fde52b5c | 656 | continue; |
113738bb | 657 | if ((HeKFLAGS(entry) ^ masked_flags) & HVhek_UTF8) |
c3654f1a | 658 | continue; |
b2c64049 NC |
659 | |
660 | if (action & (HV_FETCH_LVALUE|HV_FETCH_ISSTORE)) { | |
661 | if (HeKFLAGS(entry) != masked_flags) { | |
662 | /* We match if HVhek_UTF8 bit in our flags and hash key's | |
663 | match. But if entry was set previously with HVhek_WASUTF8 | |
664 | and key now doesn't (or vice versa) then we should change | |
665 | the key's flag, as this is assignment. */ | |
666 | if (HvSHAREKEYS(hv)) { | |
667 | /* Need to swap the key we have for a key with the flags we | |
668 | need. As keys are shared we can't just write to the | |
669 | flag, so we share the new one, unshare the old one. */ | |
6136c704 | 670 | HEK * const new_hek = share_hek_flags(key, klen, hash, |
6e838c70 | 671 | masked_flags); |
b2c64049 NC |
672 | unshare_hek (HeKEY_hek(entry)); |
673 | HeKEY_hek(entry) = new_hek; | |
674 | } | |
5d2b1485 NC |
675 | else if (hv == PL_strtab) { |
676 | /* PL_strtab is usually the only hash without HvSHAREKEYS, | |
677 | so putting this test here is cheap */ | |
678 | if (flags & HVhek_FREEKEY) | |
679 | Safefree(key); | |
680 | Perl_croak(aTHX_ S_strtab_error, | |
681 | action & HV_FETCH_LVALUE ? "fetch" : "store"); | |
682 | } | |
b2c64049 NC |
683 | else |
684 | HeKFLAGS(entry) = masked_flags; | |
685 | if (masked_flags & HVhek_ENABLEHVKFLAGS) | |
686 | HvHASKFLAGS_on(hv); | |
687 | } | |
688 | if (HeVAL(entry) == &PL_sv_placeholder) { | |
689 | /* yes, can store into placeholder slot */ | |
690 | if (action & HV_FETCH_LVALUE) { | |
691 | if (SvMAGICAL(hv)) { | |
692 | /* This preserves behaviour with the old hv_fetch | |
693 | implementation which at this point would bail out | |
694 | with a break; (at "if we find a placeholder, we | |
695 | pretend we haven't found anything") | |
696 | ||
697 | That break mean that if a placeholder were found, it | |
698 | caused a call into hv_store, which in turn would | |
699 | check magic, and if there is no magic end up pretty | |
700 | much back at this point (in hv_store's code). */ | |
701 | break; | |
702 | } | |
703 | /* LVAL fetch which actaully needs a store. */ | |
561b68a9 | 704 | val = newSV(0); |
ca732855 | 705 | HvPLACEHOLDERS(hv)--; |
b2c64049 NC |
706 | } else { |
707 | /* store */ | |
708 | if (val != &PL_sv_placeholder) | |
ca732855 | 709 | HvPLACEHOLDERS(hv)--; |
b2c64049 NC |
710 | } |
711 | HeVAL(entry) = val; | |
712 | } else if (action & HV_FETCH_ISSTORE) { | |
713 | SvREFCNT_dec(HeVAL(entry)); | |
714 | HeVAL(entry) = val; | |
715 | } | |
27bcc0a7 | 716 | } else if (HeVAL(entry) == &PL_sv_placeholder) { |
b2c64049 NC |
717 | /* if we find a placeholder, we pretend we haven't found |
718 | anything */ | |
8aacddc1 | 719 | break; |
b2c64049 | 720 | } |
113738bb NC |
721 | if (flags & HVhek_FREEKEY) |
722 | Safefree(key); | |
3c84c864 NC |
723 | if (return_svp) { |
724 | return entry ? (void *) &HeVAL(entry) : NULL; | |
725 | } | |
fde52b5c | 726 | return entry; |
727 | } | |
728 | #ifdef DYNAMIC_ENV_FETCH /* %ENV lookup? If so, try to fetch the value now */ | |
0ed29950 | 729 | if (!(action & HV_FETCH_ISSTORE) |
ad64d0ec NC |
730 | && SvRMAGICAL((const SV *)hv) |
731 | && mg_find((const SV *)hv, PERL_MAGIC_env)) { | |
a6c40364 | 732 | unsigned long len; |
9d4ba2ae | 733 | const char * const env = PerlEnv_ENVgetenv_len(key,&len); |
a6c40364 GS |
734 | if (env) { |
735 | sv = newSVpvn(env,len); | |
736 | SvTAINTED_on(sv); | |
d3ba3f5c | 737 | return hv_common(hv, keysv, key, klen, flags, |
3c84c864 NC |
738 | HV_FETCH_ISSTORE|HV_DISABLE_UVAR_XKEY|return_svp, |
739 | sv, hash); | |
a6c40364 | 740 | } |
fde52b5c | 741 | } |
742 | #endif | |
7f66fda2 NC |
743 | |
744 | if (!entry && SvREADONLY(hv) && !(action & HV_FETCH_ISEXISTS)) { | |
c445ea15 | 745 | hv_notallowed(flags, key, klen, |
c8cd6465 NC |
746 | "Attempt to access disallowed key '%"SVf"' in" |
747 | " a restricted hash"); | |
1b1f1335 | 748 | } |
b2c64049 NC |
749 | if (!(action & (HV_FETCH_LVALUE|HV_FETCH_ISSTORE))) { |
750 | /* Not doing some form of store, so return failure. */ | |
751 | if (flags & HVhek_FREEKEY) | |
752 | Safefree(key); | |
3c84c864 | 753 | return NULL; |
b2c64049 | 754 | } |
113738bb | 755 | if (action & HV_FETCH_LVALUE) { |
561b68a9 | 756 | val = newSV(0); |
b2c64049 NC |
757 | if (SvMAGICAL(hv)) { |
758 | /* At this point the old hv_fetch code would call to hv_store, | |
759 | which in turn might do some tied magic. So we need to make that | |
760 | magic check happen. */ | |
761 | /* gonna assign to this, so it better be there */ | |
fda2d18a NC |
762 | /* If a fetch-as-store fails on the fetch, then the action is to |
763 | recurse once into "hv_store". If we didn't do this, then that | |
764 | recursive call would call the key conversion routine again. | |
765 | However, as we replace the original key with the converted | |
766 | key, this would result in a double conversion, which would show | |
767 | up as a bug if the conversion routine is not idempotent. */ | |
d3ba3f5c | 768 | return hv_common(hv, keysv, key, klen, flags, |
3c84c864 NC |
769 | HV_FETCH_ISSTORE|HV_DISABLE_UVAR_XKEY|return_svp, |
770 | val, hash); | |
b2c64049 NC |
771 | /* XXX Surely that could leak if the fetch-was-store fails? |
772 | Just like the hv_fetch. */ | |
113738bb NC |
773 | } |
774 | } | |
775 | ||
b2c64049 NC |
776 | /* Welcome to hv_store... */ |
777 | ||
7b2c381c | 778 | if (!HvARRAY(hv)) { |
b2c64049 NC |
779 | /* Not sure if we can get here. I think the only case of oentry being |
780 | NULL is for %ENV with dynamic env fetch. But that should disappear | |
781 | with magic in the previous code. */ | |
d58e6666 | 782 | char *array; |
a02a5408 | 783 | Newxz(array, |
b2c64049 | 784 | PERL_HV_ARRAY_ALLOC_BYTES(xhv->xhv_max+1 /* HvMAX(hv)+1 */), |
d58e6666 NC |
785 | char); |
786 | HvARRAY(hv) = (HE**)array; | |
b2c64049 NC |
787 | } |
788 | ||
7b2c381c | 789 | oentry = &(HvARRAY(hv))[hash & (I32) xhv->xhv_max]; |
ab4af705 | 790 | |
b2c64049 NC |
791 | entry = new_HE(); |
792 | /* share_hek_flags will do the free for us. This might be considered | |
793 | bad API design. */ | |
794 | if (HvSHAREKEYS(hv)) | |
6e838c70 | 795 | HeKEY_hek(entry) = share_hek_flags(key, klen, hash, flags); |
5d2b1485 NC |
796 | else if (hv == PL_strtab) { |
797 | /* PL_strtab is usually the only hash without HvSHAREKEYS, so putting | |
798 | this test here is cheap */ | |
799 | if (flags & HVhek_FREEKEY) | |
800 | Safefree(key); | |
801 | Perl_croak(aTHX_ S_strtab_error, | |
802 | action & HV_FETCH_LVALUE ? "fetch" : "store"); | |
803 | } | |
b2c64049 NC |
804 | else /* gotta do the real thing */ |
805 | HeKEY_hek(entry) = save_hek_flags(key, klen, hash, flags); | |
806 | HeVAL(entry) = val; | |
807 | HeNEXT(entry) = *oentry; | |
808 | *oentry = entry; | |
809 | ||
810 | if (val == &PL_sv_placeholder) | |
ca732855 | 811 | HvPLACEHOLDERS(hv)++; |
b2c64049 NC |
812 | if (masked_flags & HVhek_ENABLEHVKFLAGS) |
813 | HvHASKFLAGS_on(hv); | |
814 | ||
0298d7b9 NC |
815 | { |
816 | const HE *counter = HeNEXT(entry); | |
817 | ||
4c7185a0 | 818 | xhv->xhv_keys++; /* HvTOTALKEYS(hv)++ */ |
0298d7b9 | 819 | if (!counter) { /* initial entry? */ |
5ac36297 | 820 | } else if (xhv->xhv_keys > xhv->xhv_max) { |
0298d7b9 NC |
821 | hsplit(hv); |
822 | } else if(!HvREHASH(hv)) { | |
823 | U32 n_links = 1; | |
824 | ||
825 | while ((counter = HeNEXT(counter))) | |
826 | n_links++; | |
827 | ||
828 | if (n_links > HV_MAX_LENGTH_BEFORE_SPLIT) { | |
829 | /* Use only the old HvKEYS(hv) > HvMAX(hv) condition to limit | |
830 | bucket splits on a rehashed hash, as we're not going to | |
831 | split it again, and if someone is lucky (evil) enough to | |
832 | get all the keys in one list they could exhaust our memory | |
833 | as we repeatedly double the number of buckets on every | |
834 | entry. Linear search feels a less worse thing to do. */ | |
835 | hsplit(hv); | |
836 | } | |
837 | } | |
fde52b5c | 838 | } |
b2c64049 | 839 | |
3c84c864 NC |
840 | if (return_svp) { |
841 | return entry ? (void *) &HeVAL(entry) : NULL; | |
842 | } | |
843 | return (void *) entry; | |
fde52b5c | 844 | } |
845 | ||
864dbfa3 | 846 | STATIC void |
b0e6ae5b | 847 | S_hv_magic_check(HV *hv, bool *needs_copy, bool *needs_store) |
d0066dc7 | 848 | { |
a3b680e6 | 849 | const MAGIC *mg = SvMAGIC(hv); |
7918f24d NC |
850 | |
851 | PERL_ARGS_ASSERT_HV_MAGIC_CHECK; | |
852 | ||
d0066dc7 OT |
853 | *needs_copy = FALSE; |
854 | *needs_store = TRUE; | |
855 | while (mg) { | |
856 | if (isUPPER(mg->mg_type)) { | |
857 | *needs_copy = TRUE; | |
d60c5a05 | 858 | if (mg->mg_type == PERL_MAGIC_tied) { |
d0066dc7 | 859 | *needs_store = FALSE; |
4ab2a30b | 860 | return; /* We've set all there is to set. */ |
d0066dc7 OT |
861 | } |
862 | } | |
863 | mg = mg->mg_moremagic; | |
864 | } | |
865 | } | |
866 | ||
954c1994 | 867 | /* |
a3bcc51e TP |
868 | =for apidoc hv_scalar |
869 | ||
870 | Evaluates the hash in scalar context and returns the result. Handles magic when the hash is tied. | |
871 | ||
872 | =cut | |
873 | */ | |
874 | ||
875 | SV * | |
876 | Perl_hv_scalar(pTHX_ HV *hv) | |
877 | { | |
a3bcc51e | 878 | SV *sv; |
823a54a3 | 879 | |
7918f24d NC |
880 | PERL_ARGS_ASSERT_HV_SCALAR; |
881 | ||
823a54a3 | 882 | if (SvRMAGICAL(hv)) { |
ad64d0ec | 883 | MAGIC * const mg = mg_find((const SV *)hv, PERL_MAGIC_tied); |
823a54a3 AL |
884 | if (mg) |
885 | return magic_scalarpack(hv, mg); | |
886 | } | |
a3bcc51e TP |
887 | |
888 | sv = sv_newmortal(); | |
f4431c56 | 889 | if (HvTOTALKEYS((const HV *)hv)) |
a3bcc51e TP |
890 | Perl_sv_setpvf(aTHX_ sv, "%ld/%ld", |
891 | (long)HvFILL(hv), (long)HvMAX(hv) + 1); | |
892 | else | |
893 | sv_setiv(sv, 0); | |
894 | ||
895 | return sv; | |
896 | } | |
897 | ||
898 | /* | |
954c1994 GS |
899 | =for apidoc hv_delete |
900 | ||
901 | Deletes a key/value pair in the hash. The value SV is removed from the | |
1c846c1f | 902 | hash and returned to the caller. The C<klen> is the length of the key. |
954c1994 GS |
903 | The C<flags> value will normally be zero; if set to G_DISCARD then NULL |
904 | will be returned. | |
905 | ||
954c1994 GS |
906 | =for apidoc hv_delete_ent |
907 | ||
908 | Deletes a key/value pair in the hash. The value SV is removed from the | |
909 | hash and returned to the caller. The C<flags> value will normally be zero; | |
910 | if set to G_DISCARD then NULL will be returned. C<hash> can be a valid | |
911 | precomputed hash value, or 0 to ask for it to be computed. | |
912 | ||
913 | =cut | |
914 | */ | |
915 | ||
8f8d40ab | 916 | STATIC SV * |
cd6d36ac NC |
917 | S_hv_delete_common(pTHX_ HV *hv, SV *keysv, const char *key, STRLEN klen, |
918 | int k_flags, I32 d_flags, U32 hash) | |
f1317c8d | 919 | { |
27da23d5 | 920 | dVAR; |
cbec9347 | 921 | register XPVHV* xhv; |
fde52b5c | 922 | register HE *entry; |
923 | register HE **oentry; | |
9e720f71 | 924 | HE *const *first_entry; |
9dbc5603 | 925 | bool is_utf8 = (k_flags & HVhek_UTF8) ? TRUE : FALSE; |
7a9669ca | 926 | int masked_flags; |
1c846c1f | 927 | |
fde52b5c | 928 | if (SvRMAGICAL(hv)) { |
0a0bb7c7 OT |
929 | bool needs_copy; |
930 | bool needs_store; | |
931 | hv_magic_check (hv, &needs_copy, &needs_store); | |
932 | ||
f1317c8d | 933 | if (needs_copy) { |
6136c704 | 934 | SV *sv; |
63c89345 NC |
935 | entry = (HE *) hv_common(hv, keysv, key, klen, |
936 | k_flags & ~HVhek_FREEKEY, | |
937 | HV_FETCH_LVALUE|HV_DISABLE_UVAR_XKEY, | |
938 | NULL, hash); | |
7a9669ca | 939 | sv = entry ? HeVAL(entry) : NULL; |
f1317c8d NC |
940 | if (sv) { |
941 | if (SvMAGICAL(sv)) { | |
942 | mg_clear(sv); | |
943 | } | |
944 | if (!needs_store) { | |
945 | if (mg_find(sv, PERL_MAGIC_tiedelem)) { | |
946 | /* No longer an element */ | |
947 | sv_unmagic(sv, PERL_MAGIC_tiedelem); | |
948 | return sv; | |
949 | } | |
a0714e2c | 950 | return NULL; /* element cannot be deleted */ |
f1317c8d | 951 | } |
902173a3 | 952 | #ifdef ENV_IS_CASELESS |
ad64d0ec | 953 | else if (mg_find((const SV *)hv, PERL_MAGIC_env)) { |
8167a60a | 954 | /* XXX This code isn't UTF8 clean. */ |
59cd0e26 | 955 | keysv = newSVpvn_flags(key, klen, SVs_TEMP); |
8167a60a NC |
956 | if (k_flags & HVhek_FREEKEY) { |
957 | Safefree(key); | |
958 | } | |
959 | key = strupr(SvPVX(keysv)); | |
960 | is_utf8 = 0; | |
961 | k_flags = 0; | |
962 | hash = 0; | |
7f66fda2 | 963 | } |
510ac311 | 964 | #endif |
2fd1c6b8 | 965 | } |
2fd1c6b8 | 966 | } |
fde52b5c | 967 | } |
cbec9347 | 968 | xhv = (XPVHV*)SvANY(hv); |
7b2c381c | 969 | if (!HvARRAY(hv)) |
a0714e2c | 970 | return NULL; |
fde52b5c | 971 | |
19692e8d | 972 | if (is_utf8) { |
c445ea15 | 973 | const char * const keysave = key; |
b464bac0 | 974 | key = (char*)bytes_from_utf8((U8*)key, &klen, &is_utf8); |
cd6d36ac | 975 | |
19692e8d | 976 | if (is_utf8) |
cd6d36ac NC |
977 | k_flags |= HVhek_UTF8; |
978 | else | |
979 | k_flags &= ~HVhek_UTF8; | |
7f66fda2 NC |
980 | if (key != keysave) { |
981 | if (k_flags & HVhek_FREEKEY) { | |
982 | /* This shouldn't happen if our caller does what we expect, | |
983 | but strictly the API allows it. */ | |
984 | Safefree(keysave); | |
985 | } | |
986 | k_flags |= HVhek_WASUTF8 | HVhek_FREEKEY; | |
987 | } | |
ad64d0ec | 988 | HvHASKFLAGS_on(MUTABLE_SV(hv)); |
19692e8d | 989 | } |
f9a63242 | 990 | |
4b5190b5 NC |
991 | if (HvREHASH(hv)) { |
992 | PERL_HASH_INTERNAL(hash, key, klen); | |
993 | } else if (!hash) { | |
7a9669ca | 994 | if (keysv && (SvIsCOW_shared_hash(keysv))) { |
c158a4fd | 995 | hash = SvSHARED_HASH(keysv); |
7a9669ca NC |
996 | } else { |
997 | PERL_HASH(hash, key, klen); | |
998 | } | |
4b5190b5 | 999 | } |
fde52b5c | 1000 | |
7a9669ca NC |
1001 | masked_flags = (k_flags & HVhek_MASK); |
1002 | ||
9e720f71 | 1003 | first_entry = oentry = &(HvARRAY(hv))[hash & (I32) HvMAX(hv)]; |
fde52b5c | 1004 | entry = *oentry; |
9e720f71 | 1005 | for (; entry; oentry = &HeNEXT(entry), entry = *oentry) { |
6136c704 | 1006 | SV *sv; |
fde52b5c | 1007 | if (HeHASH(entry) != hash) /* strings can't be equal */ |
1008 | continue; | |
eb160463 | 1009 | if (HeKLEN(entry) != (I32)klen) |
fde52b5c | 1010 | continue; |
1c846c1f | 1011 | if (HeKEY(entry) != key && memNE(HeKEY(entry),key,klen)) /* is this it? */ |
fde52b5c | 1012 | continue; |
7a9669ca | 1013 | if ((HeKFLAGS(entry) ^ masked_flags) & HVhek_UTF8) |
c3654f1a | 1014 | continue; |
8aacddc1 | 1015 | |
5d2b1485 NC |
1016 | if (hv == PL_strtab) { |
1017 | if (k_flags & HVhek_FREEKEY) | |
1018 | Safefree(key); | |
1019 | Perl_croak(aTHX_ S_strtab_error, "delete"); | |
1020 | } | |
1021 | ||
8aacddc1 | 1022 | /* if placeholder is here, it's already been deleted.... */ |
6136c704 AL |
1023 | if (HeVAL(entry) == &PL_sv_placeholder) { |
1024 | if (k_flags & HVhek_FREEKEY) | |
1025 | Safefree(key); | |
1026 | return NULL; | |
8aacddc1 | 1027 | } |
6136c704 | 1028 | if (SvREADONLY(hv) && HeVAL(entry) && SvREADONLY(HeVAL(entry))) { |
d4c19fe8 | 1029 | hv_notallowed(k_flags, key, klen, |
c8cd6465 NC |
1030 | "Attempt to delete readonly key '%"SVf"' from" |
1031 | " a restricted hash"); | |
8aacddc1 | 1032 | } |
b84d0860 NC |
1033 | if (k_flags & HVhek_FREEKEY) |
1034 | Safefree(key); | |
8aacddc1 | 1035 | |
cd6d36ac | 1036 | if (d_flags & G_DISCARD) |
a0714e2c | 1037 | sv = NULL; |
94f7643d | 1038 | else { |
79d01fbf | 1039 | sv = sv_2mortal(HeVAL(entry)); |
7996736c | 1040 | HeVAL(entry) = &PL_sv_placeholder; |
94f7643d | 1041 | } |
8aacddc1 NIS |
1042 | |
1043 | /* | |
1044 | * If a restricted hash, rather than really deleting the entry, put | |
1045 | * a placeholder there. This marks the key as being "approved", so | |
1046 | * we can still access via not-really-existing key without raising | |
1047 | * an error. | |
1048 | */ | |
1049 | if (SvREADONLY(hv)) { | |
754604c4 | 1050 | SvREFCNT_dec(HeVAL(entry)); |
7996736c | 1051 | HeVAL(entry) = &PL_sv_placeholder; |
8aacddc1 NIS |
1052 | /* We'll be saving this slot, so the number of allocated keys |
1053 | * doesn't go down, but the number placeholders goes up */ | |
ca732855 | 1054 | HvPLACEHOLDERS(hv)++; |
8aacddc1 | 1055 | } else { |
a26e96df | 1056 | *oentry = HeNEXT(entry); |
b79f7545 | 1057 | if (SvOOK(hv) && entry == HvAUX(hv)->xhv_eiter /* HvEITER(hv) */) |
8aacddc1 NIS |
1058 | HvLAZYDEL_on(hv); |
1059 | else | |
1060 | hv_free_ent(hv, entry); | |
4c7185a0 | 1061 | xhv->xhv_keys--; /* HvTOTALKEYS(hv)-- */ |
574c8022 | 1062 | if (xhv->xhv_keys == 0) |
19692e8d | 1063 | HvHASKFLAGS_off(hv); |
8aacddc1 | 1064 | } |
79072805 LW |
1065 | return sv; |
1066 | } | |
8aacddc1 | 1067 | if (SvREADONLY(hv)) { |
d4c19fe8 | 1068 | hv_notallowed(k_flags, key, klen, |
c8cd6465 NC |
1069 | "Attempt to delete disallowed key '%"SVf"' from" |
1070 | " a restricted hash"); | |
8aacddc1 NIS |
1071 | } |
1072 | ||
19692e8d | 1073 | if (k_flags & HVhek_FREEKEY) |
f9a63242 | 1074 | Safefree(key); |
a0714e2c | 1075 | return NULL; |
79072805 LW |
1076 | } |
1077 | ||
76e3520e | 1078 | STATIC void |
cea2e8a9 | 1079 | S_hsplit(pTHX_ HV *hv) |
79072805 | 1080 | { |
97aff369 | 1081 | dVAR; |
1e05feb3 | 1082 | register XPVHV* const xhv = (XPVHV*)SvANY(hv); |
a3b680e6 | 1083 | const I32 oldsize = (I32) xhv->xhv_max+1; /* HvMAX(hv)+1 (sick) */ |
79072805 LW |
1084 | register I32 newsize = oldsize * 2; |
1085 | register I32 i; | |
7b2c381c | 1086 | char *a = (char*) HvARRAY(hv); |
72311751 | 1087 | register HE **aep; |
79072805 | 1088 | register HE **oentry; |
4b5190b5 NC |
1089 | int longest_chain = 0; |
1090 | int was_shared; | |
79072805 | 1091 | |
7918f24d NC |
1092 | PERL_ARGS_ASSERT_HSPLIT; |
1093 | ||
18026298 | 1094 | /*PerlIO_printf(PerlIO_stderr(), "hsplit called for %p which had %d\n", |
6c9570dc | 1095 | (void*)hv, (int) oldsize);*/ |
18026298 | 1096 | |
5d88ecd7 | 1097 | if (HvPLACEHOLDERS_get(hv) && !SvREADONLY(hv)) { |
18026298 NC |
1098 | /* Can make this clear any placeholders first for non-restricted hashes, |
1099 | even though Storable rebuilds restricted hashes by putting in all the | |
1100 | placeholders (first) before turning on the readonly flag, because | |
1101 | Storable always pre-splits the hash. */ | |
1102 | hv_clear_placeholders(hv); | |
1103 | } | |
1104 | ||
3280af22 | 1105 | PL_nomemok = TRUE; |
8d6dde3e | 1106 | #if defined(STRANGE_MALLOC) || defined(MYMALLOC) |
b79f7545 NC |
1107 | Renew(a, PERL_HV_ARRAY_ALLOC_BYTES(newsize) |
1108 | + (SvOOK(hv) ? sizeof(struct xpvhv_aux) : 0), char); | |
422a93e5 | 1109 | if (!a) { |
4a33f861 | 1110 | PL_nomemok = FALSE; |
422a93e5 GA |
1111 | return; |
1112 | } | |
b79f7545 | 1113 | if (SvOOK(hv)) { |
31f0e52e | 1114 | Move(&a[oldsize * sizeof(HE*)], &a[newsize * sizeof(HE*)], 1, struct xpvhv_aux); |
b79f7545 | 1115 | } |
4633a7c4 | 1116 | #else |
a02a5408 | 1117 | Newx(a, PERL_HV_ARRAY_ALLOC_BYTES(newsize) |
b79f7545 | 1118 | + (SvOOK(hv) ? sizeof(struct xpvhv_aux) : 0), char); |
422a93e5 | 1119 | if (!a) { |
3280af22 | 1120 | PL_nomemok = FALSE; |
422a93e5 GA |
1121 | return; |
1122 | } | |
7b2c381c | 1123 | Copy(HvARRAY(hv), a, oldsize * sizeof(HE*), char); |
b79f7545 NC |
1124 | if (SvOOK(hv)) { |
1125 | Copy(HvAUX(hv), &a[newsize * sizeof(HE*)], 1, struct xpvhv_aux); | |
1126 | } | |
fba3b22e | 1127 | if (oldsize >= 64) { |
7b2c381c | 1128 | offer_nice_chunk(HvARRAY(hv), |
b79f7545 NC |
1129 | PERL_HV_ARRAY_ALLOC_BYTES(oldsize) |
1130 | + (SvOOK(hv) ? sizeof(struct xpvhv_aux) : 0)); | |
4633a7c4 LW |
1131 | } |
1132 | else | |
7b2c381c | 1133 | Safefree(HvARRAY(hv)); |
4633a7c4 LW |
1134 | #endif |
1135 | ||
3280af22 | 1136 | PL_nomemok = FALSE; |
72311751 | 1137 | Zero(&a[oldsize * sizeof(HE*)], (newsize-oldsize) * sizeof(HE*), char); /* zero 2nd half*/ |
cbec9347 | 1138 | xhv->xhv_max = --newsize; /* HvMAX(hv) = --newsize */ |
7b2c381c | 1139 | HvARRAY(hv) = (HE**) a; |
72311751 | 1140 | aep = (HE**)a; |
79072805 | 1141 | |
72311751 | 1142 | for (i=0; i<oldsize; i++,aep++) { |
4b5190b5 NC |
1143 | int left_length = 0; |
1144 | int right_length = 0; | |
a3b680e6 AL |
1145 | register HE *entry; |
1146 | register HE **bep; | |
4b5190b5 | 1147 | |
72311751 | 1148 | if (!*aep) /* non-existent */ |
79072805 | 1149 | continue; |
72311751 GS |
1150 | bep = aep+oldsize; |
1151 | for (oentry = aep, entry = *aep; entry; entry = *oentry) { | |
eb160463 | 1152 | if ((HeHASH(entry) & newsize) != (U32)i) { |
fde52b5c | 1153 | *oentry = HeNEXT(entry); |
72311751 | 1154 | HeNEXT(entry) = *bep; |
72311751 | 1155 | *bep = entry; |
4b5190b5 | 1156 | right_length++; |
79072805 LW |
1157 | continue; |
1158 | } | |
4b5190b5 | 1159 | else { |
fde52b5c | 1160 | oentry = &HeNEXT(entry); |
4b5190b5 NC |
1161 | left_length++; |
1162 | } | |
79072805 | 1163 | } |
4b5190b5 NC |
1164 | /* I think we don't actually need to keep track of the longest length, |
1165 | merely flag if anything is too long. But for the moment while | |
1166 | developing this code I'll track it. */ | |
1167 | if (left_length > longest_chain) | |
1168 | longest_chain = left_length; | |
1169 | if (right_length > longest_chain) | |
1170 | longest_chain = right_length; | |
1171 | } | |
1172 | ||
1173 | ||
1174 | /* Pick your policy for "hashing isn't working" here: */ | |
fdcd69b6 | 1175 | if (longest_chain <= HV_MAX_LENGTH_BEFORE_SPLIT /* split worked? */ |
4b5190b5 NC |
1176 | || HvREHASH(hv)) { |
1177 | return; | |
79072805 | 1178 | } |
4b5190b5 NC |
1179 | |
1180 | if (hv == PL_strtab) { | |
1181 | /* Urg. Someone is doing something nasty to the string table. | |
1182 | Can't win. */ | |
1183 | return; | |
1184 | } | |
1185 | ||
1186 | /* Awooga. Awooga. Pathological data. */ | |
6c9570dc | 1187 | /*PerlIO_printf(PerlIO_stderr(), "%p %d of %d with %d/%d buckets\n", (void*)hv, |
4b5190b5 NC |
1188 | longest_chain, HvTOTALKEYS(hv), HvFILL(hv), 1+HvMAX(hv));*/ |
1189 | ||
1190 | ++newsize; | |
a02a5408 | 1191 | Newxz(a, PERL_HV_ARRAY_ALLOC_BYTES(newsize) |
b79f7545 NC |
1192 | + (SvOOK(hv) ? sizeof(struct xpvhv_aux) : 0), char); |
1193 | if (SvOOK(hv)) { | |
1194 | Copy(HvAUX(hv), &a[newsize * sizeof(HE*)], 1, struct xpvhv_aux); | |
1195 | } | |
1196 | ||
4b5190b5 NC |
1197 | was_shared = HvSHAREKEYS(hv); |
1198 | ||
4b5190b5 NC |
1199 | HvSHAREKEYS_off(hv); |
1200 | HvREHASH_on(hv); | |
1201 | ||
7b2c381c | 1202 | aep = HvARRAY(hv); |
4b5190b5 NC |
1203 | |
1204 | for (i=0; i<newsize; i++,aep++) { | |
a3b680e6 | 1205 | register HE *entry = *aep; |
4b5190b5 NC |
1206 | while (entry) { |
1207 | /* We're going to trash this HE's next pointer when we chain it | |
1208 | into the new hash below, so store where we go next. */ | |
9d4ba2ae | 1209 | HE * const next = HeNEXT(entry); |
4b5190b5 | 1210 | UV hash; |
a3b680e6 | 1211 | HE **bep; |
4b5190b5 NC |
1212 | |
1213 | /* Rehash it */ | |
1214 | PERL_HASH_INTERNAL(hash, HeKEY(entry), HeKLEN(entry)); | |
1215 | ||
1216 | if (was_shared) { | |
1217 | /* Unshare it. */ | |
aec46f14 | 1218 | HEK * const new_hek |
4b5190b5 NC |
1219 | = save_hek_flags(HeKEY(entry), HeKLEN(entry), |
1220 | hash, HeKFLAGS(entry)); | |
1221 | unshare_hek (HeKEY_hek(entry)); | |
1222 | HeKEY_hek(entry) = new_hek; | |
1223 | } else { | |
1224 | /* Not shared, so simply write the new hash in. */ | |
1225 | HeHASH(entry) = hash; | |
1226 | } | |
1227 | /*PerlIO_printf(PerlIO_stderr(), "%d ", HeKFLAGS(entry));*/ | |
1228 | HEK_REHASH_on(HeKEY_hek(entry)); | |
1229 | /*PerlIO_printf(PerlIO_stderr(), "%d\n", HeKFLAGS(entry));*/ | |
1230 | ||
1231 | /* Copy oentry to the correct new chain. */ | |
1232 | bep = ((HE**)a) + (hash & (I32) xhv->xhv_max); | |
4b5190b5 NC |
1233 | HeNEXT(entry) = *bep; |
1234 | *bep = entry; | |
1235 | ||
1236 | entry = next; | |
1237 | } | |
1238 | } | |
7b2c381c NC |
1239 | Safefree (HvARRAY(hv)); |
1240 | HvARRAY(hv) = (HE **)a; | |
79072805 LW |
1241 | } |
1242 | ||
72940dca | 1243 | void |
864dbfa3 | 1244 | Perl_hv_ksplit(pTHX_ HV *hv, IV newmax) |
72940dca | 1245 | { |
97aff369 | 1246 | dVAR; |
cbec9347 | 1247 | register XPVHV* xhv = (XPVHV*)SvANY(hv); |
a3b680e6 | 1248 | const I32 oldsize = (I32) xhv->xhv_max+1; /* HvMAX(hv)+1 (sick) */ |
72940dca | 1249 | register I32 newsize; |
1250 | register I32 i; | |
72311751 GS |
1251 | register char *a; |
1252 | register HE **aep; | |
72940dca | 1253 | register HE *entry; |
1254 | register HE **oentry; | |
1255 | ||
7918f24d NC |
1256 | PERL_ARGS_ASSERT_HV_KSPLIT; |
1257 | ||
72940dca | 1258 | newsize = (I32) newmax; /* possible truncation here */ |
1259 | if (newsize != newmax || newmax <= oldsize) | |
1260 | return; | |
1261 | while ((newsize & (1 + ~newsize)) != newsize) { | |
1262 | newsize &= ~(newsize & (1 + ~newsize)); /* get proper power of 2 */ | |
1263 | } | |
1264 | if (newsize < newmax) | |
1265 | newsize *= 2; | |
1266 | if (newsize < newmax) | |
1267 | return; /* overflow detection */ | |
1268 | ||
7b2c381c | 1269 | a = (char *) HvARRAY(hv); |
72940dca | 1270 | if (a) { |
3280af22 | 1271 | PL_nomemok = TRUE; |
8d6dde3e | 1272 | #if defined(STRANGE_MALLOC) || defined(MYMALLOC) |
b79f7545 NC |
1273 | Renew(a, PERL_HV_ARRAY_ALLOC_BYTES(newsize) |
1274 | + (SvOOK(hv) ? sizeof(struct xpvhv_aux) : 0), char); | |
8aacddc1 | 1275 | if (!a) { |
4a33f861 | 1276 | PL_nomemok = FALSE; |
422a93e5 GA |
1277 | return; |
1278 | } | |
b79f7545 | 1279 | if (SvOOK(hv)) { |
7a9b70e9 | 1280 | Copy(&a[oldsize * sizeof(HE*)], &a[newsize * sizeof(HE*)], 1, struct xpvhv_aux); |
b79f7545 | 1281 | } |
72940dca | 1282 | #else |
a02a5408 | 1283 | Newx(a, PERL_HV_ARRAY_ALLOC_BYTES(newsize) |
b79f7545 | 1284 | + (SvOOK(hv) ? sizeof(struct xpvhv_aux) : 0), char); |
8aacddc1 | 1285 | if (!a) { |
3280af22 | 1286 | PL_nomemok = FALSE; |
422a93e5 GA |
1287 | return; |
1288 | } | |
7b2c381c | 1289 | Copy(HvARRAY(hv), a, oldsize * sizeof(HE*), char); |
b79f7545 NC |
1290 | if (SvOOK(hv)) { |
1291 | Copy(HvAUX(hv), &a[newsize * sizeof(HE*)], 1, struct xpvhv_aux); | |
1292 | } | |
fba3b22e | 1293 | if (oldsize >= 64) { |
7b2c381c | 1294 | offer_nice_chunk(HvARRAY(hv), |
b79f7545 NC |
1295 | PERL_HV_ARRAY_ALLOC_BYTES(oldsize) |
1296 | + (SvOOK(hv) ? sizeof(struct xpvhv_aux) : 0)); | |
72940dca | 1297 | } |
1298 | else | |
7b2c381c | 1299 | Safefree(HvARRAY(hv)); |
72940dca | 1300 | #endif |
3280af22 | 1301 | PL_nomemok = FALSE; |
72311751 | 1302 | Zero(&a[oldsize * sizeof(HE*)], (newsize-oldsize) * sizeof(HE*), char); /* zero 2nd half*/ |
72940dca | 1303 | } |
1304 | else { | |
a02a5408 | 1305 | Newxz(a, PERL_HV_ARRAY_ALLOC_BYTES(newsize), char); |
72940dca | 1306 | } |
cbec9347 | 1307 | xhv->xhv_max = --newsize; /* HvMAX(hv) = --newsize */ |
7b2c381c | 1308 | HvARRAY(hv) = (HE **) a; |
f4431c56 | 1309 | if (!xhv->xhv_keys /* !HvTOTALKEYS(hv) */) /* skip rest if no entries */ |
72940dca | 1310 | return; |
1311 | ||
72311751 GS |
1312 | aep = (HE**)a; |
1313 | for (i=0; i<oldsize; i++,aep++) { | |
1314 | if (!*aep) /* non-existent */ | |
72940dca | 1315 | continue; |
72311751 | 1316 | for (oentry = aep, entry = *aep; entry; entry = *oentry) { |
6136c704 AL |
1317 | register I32 j = (HeHASH(entry) & newsize); |
1318 | ||
1319 | if (j != i) { | |
72940dca | 1320 | j -= i; |
1321 | *oentry = HeNEXT(entry); | |
4d0fbddd | 1322 | HeNEXT(entry) = aep[j]; |
72311751 | 1323 | aep[j] = entry; |
72940dca | 1324 | continue; |
1325 | } | |
1326 | else | |
1327 | oentry = &HeNEXT(entry); | |
1328 | } | |
72940dca | 1329 | } |
1330 | } | |
1331 | ||
b3ac6de7 | 1332 | HV * |
864dbfa3 | 1333 | Perl_newHVhv(pTHX_ HV *ohv) |
b3ac6de7 | 1334 | { |
749123ff | 1335 | dVAR; |
9d4ba2ae | 1336 | HV * const hv = newHV(); |
f4431c56 | 1337 | STRLEN hv_max; |
4beac62f | 1338 | |
f4431c56 | 1339 | if (!ohv || !HvTOTALKEYS(ohv)) |
4beac62f | 1340 | return hv; |
4beac62f | 1341 | hv_max = HvMAX(ohv); |
b3ac6de7 | 1342 | |
ad64d0ec | 1343 | if (!SvMAGICAL((const SV *)ohv)) { |
b56ba0bf | 1344 | /* It's an ordinary hash, so copy it fast. AMS 20010804 */ |
eb160463 | 1345 | STRLEN i; |
a3b680e6 | 1346 | const bool shared = !!HvSHAREKEYS(ohv); |
aec46f14 | 1347 | HE **ents, ** const oents = (HE **)HvARRAY(ohv); |
ff875642 | 1348 | char *a; |
a02a5408 | 1349 | Newx(a, PERL_HV_ARRAY_ALLOC_BYTES(hv_max+1), char); |
ff875642 | 1350 | ents = (HE**)a; |
b56ba0bf AMS |
1351 | |
1352 | /* In each bucket... */ | |
1353 | for (i = 0; i <= hv_max; i++) { | |
6136c704 | 1354 | HE *prev = NULL; |
aec46f14 | 1355 | HE *oent = oents[i]; |
b56ba0bf AMS |
1356 | |
1357 | if (!oent) { | |
1358 | ents[i] = NULL; | |
1359 | continue; | |
1360 | } | |
1361 | ||
1362 | /* Copy the linked list of entries. */ | |
aec46f14 | 1363 | for (; oent; oent = HeNEXT(oent)) { |
a3b680e6 AL |
1364 | const U32 hash = HeHASH(oent); |
1365 | const char * const key = HeKEY(oent); | |
1366 | const STRLEN len = HeKLEN(oent); | |
1367 | const int flags = HeKFLAGS(oent); | |
6136c704 | 1368 | HE * const ent = new_HE(); |
c3acb9e0 | 1369 | SV *const val = HeVAL(oent); |
b56ba0bf | 1370 | |
c3acb9e0 | 1371 | HeVAL(ent) = SvIMMORTAL(val) ? val : newSVsv(val); |
19692e8d | 1372 | HeKEY_hek(ent) |
6e838c70 | 1373 | = shared ? share_hek_flags(key, len, hash, flags) |
19692e8d | 1374 | : save_hek_flags(key, len, hash, flags); |
b56ba0bf AMS |
1375 | if (prev) |
1376 | HeNEXT(prev) = ent; | |
1377 | else | |
1378 | ents[i] = ent; | |
1379 | prev = ent; | |
1380 | HeNEXT(ent) = NULL; | |
1381 | } | |
1382 | } | |
1383 | ||
1384 | HvMAX(hv) = hv_max; | |
8aacddc1 | 1385 | HvTOTALKEYS(hv) = HvTOTALKEYS(ohv); |
b56ba0bf | 1386 | HvARRAY(hv) = ents; |
aec46f14 | 1387 | } /* not magical */ |
b56ba0bf AMS |
1388 | else { |
1389 | /* Iterate over ohv, copying keys and values one at a time. */ | |
b3ac6de7 | 1390 | HE *entry; |
bfcb3514 NC |
1391 | const I32 riter = HvRITER_get(ohv); |
1392 | HE * const eiter = HvEITER_get(ohv); | |
f4431c56 | 1393 | STRLEN hv_fill = HvFILL(ohv); |
b56ba0bf AMS |
1394 | |
1395 | /* Can we use fewer buckets? (hv_max is always 2^n-1) */ | |
1396 | while (hv_max && hv_max + 1 >= hv_fill * 2) | |
1397 | hv_max = hv_max / 2; | |
1398 | HvMAX(hv) = hv_max; | |
1399 | ||
4a76a316 | 1400 | hv_iterinit(ohv); |
e16e2ff8 | 1401 | while ((entry = hv_iternext_flags(ohv, 0))) { |
c3acb9e0 | 1402 | SV *const val = HeVAL(entry); |
04fe65b0 | 1403 | (void)hv_store_flags(hv, HeKEY(entry), HeKLEN(entry), |
c3acb9e0 NC |
1404 | SvIMMORTAL(val) ? val : newSVsv(val), |
1405 | HeHASH(entry), HeKFLAGS(entry)); | |
b3ac6de7 | 1406 | } |
bfcb3514 NC |
1407 | HvRITER_set(ohv, riter); |
1408 | HvEITER_set(ohv, eiter); | |
b3ac6de7 | 1409 | } |
1c846c1f | 1410 | |
b3ac6de7 IZ |
1411 | return hv; |
1412 | } | |
1413 | ||
5b9c0671 NC |
1414 | /* A rather specialised version of newHVhv for copying %^H, ensuring all the |
1415 | magic stays on it. */ | |
1416 | HV * | |
1417 | Perl_hv_copy_hints_hv(pTHX_ HV *const ohv) | |
1418 | { | |
1419 | HV * const hv = newHV(); | |
5b9c0671 | 1420 | |
f4431c56 | 1421 | if (ohv && HvTOTALKEYS(ohv)) { |
5b9c0671 | 1422 | STRLEN hv_max = HvMAX(ohv); |
f4431c56 | 1423 | STRLEN hv_fill = HvFILL(ohv); |
5b9c0671 NC |
1424 | HE *entry; |
1425 | const I32 riter = HvRITER_get(ohv); | |
1426 | HE * const eiter = HvEITER_get(ohv); | |
1427 | ||
1428 | while (hv_max && hv_max + 1 >= hv_fill * 2) | |
1429 | hv_max = hv_max / 2; | |
1430 | HvMAX(hv) = hv_max; | |
1431 | ||
1432 | hv_iterinit(ohv); | |
1433 | while ((entry = hv_iternext_flags(ohv, 0))) { | |
1434 | SV *const sv = newSVsv(HeVAL(entry)); | |
e3b1b6b1 | 1435 | SV *heksv = newSVhek(HeKEY_hek(entry)); |
5b9c0671 | 1436 | sv_magic(sv, NULL, PERL_MAGIC_hintselem, |
e3b1b6b1 RGS |
1437 | (char *)heksv, HEf_SVKEY); |
1438 | SvREFCNT_dec(heksv); | |
04fe65b0 RGS |
1439 | (void)hv_store_flags(hv, HeKEY(entry), HeKLEN(entry), |
1440 | sv, HeHASH(entry), HeKFLAGS(entry)); | |
5b9c0671 NC |
1441 | } |
1442 | HvRITER_set(ohv, riter); | |
1443 | HvEITER_set(ohv, eiter); | |
1444 | } | |
1445 | hv_magic(hv, NULL, PERL_MAGIC_hints); | |
1446 | return hv; | |
1447 | } | |
1448 | ||
79072805 | 1449 | void |
864dbfa3 | 1450 | Perl_hv_free_ent(pTHX_ HV *hv, register HE *entry) |
79072805 | 1451 | { |
97aff369 | 1452 | dVAR; |
16bdeea2 GS |
1453 | SV *val; |
1454 | ||
7918f24d NC |
1455 | PERL_ARGS_ASSERT_HV_FREE_ENT; |
1456 | ||
68dc0745 | 1457 | if (!entry) |
79072805 | 1458 | return; |
16bdeea2 | 1459 | val = HeVAL(entry); |
6eb667c6 | 1460 | if (HvNAME(hv) && anonymise_cv(HvNAME_HEK(hv), val) && GvCVu(val)) |
f1c32fec | 1461 | mro_method_changed_in(hv); |
16bdeea2 | 1462 | SvREFCNT_dec(val); |
68dc0745 | 1463 | if (HeKLEN(entry) == HEf_SVKEY) { |
1464 | SvREFCNT_dec(HeKEY_sv(entry)); | |
8aacddc1 | 1465 | Safefree(HeKEY_hek(entry)); |
44a8e56a | 1466 | } |
1467 | else if (HvSHAREKEYS(hv)) | |
68dc0745 | 1468 | unshare_hek(HeKEY_hek(entry)); |
fde52b5c | 1469 | else |
68dc0745 | 1470 | Safefree(HeKEY_hek(entry)); |
d33b2eba | 1471 | del_HE(entry); |
79072805 LW |
1472 | } |
1473 | ||
f1c32fec | 1474 | static I32 |
6eb667c6 | 1475 | S_anonymise_cv(pTHX_ HEK *stash, SV *val) |
f1c32fec BM |
1476 | { |
1477 | CV *cv; | |
1478 | ||
1479 | PERL_ARGS_ASSERT_ANONYMISE_CV; | |
1480 | ||
1481 | if (val && isGV(val) && isGV_with_GP(val) && (cv = GvCV(val))) { | |
1482 | if ((SV *)CvGV(cv) == val) { | |
f1c32fec BM |
1483 | GV *anongv; |
1484 | ||
6eb667c6 NC |
1485 | if (stash) { |
1486 | SV *gvname = newSVhek(stash); | |
1487 | sv_catpvs(gvname, "::__ANON__"); | |
1488 | anongv = gv_fetchsv(gvname, GV_ADDMULTI, SVt_PVCV); | |
1489 | SvREFCNT_dec(gvname); | |
1490 | } else { | |
1491 | anongv = gv_fetchpvs("__ANON__::__ANON__", GV_ADDMULTI, | |
1492 | SVt_PVCV); | |
1493 | } | |
f1c32fec BM |
1494 | CvGV(cv) = anongv; |
1495 | CvANON_on(cv); | |
1496 | return 1; | |
1497 | } | |
1498 | } | |
1499 | return 0; | |
1500 | } | |
1501 | ||
79072805 | 1502 | void |
864dbfa3 | 1503 | Perl_hv_delayfree_ent(pTHX_ HV *hv, register HE *entry) |
79072805 | 1504 | { |
97aff369 | 1505 | dVAR; |
7918f24d NC |
1506 | |
1507 | PERL_ARGS_ASSERT_HV_DELAYFREE_ENT; | |
1508 | ||
68dc0745 | 1509 | if (!entry) |
79072805 | 1510 | return; |
bc4947fc NC |
1511 | /* SvREFCNT_inc to counter the SvREFCNT_dec in hv_free_ent */ |
1512 | sv_2mortal(SvREFCNT_inc(HeVAL(entry))); /* free between statements */ | |
68dc0745 | 1513 | if (HeKLEN(entry) == HEf_SVKEY) { |
bc4947fc | 1514 | sv_2mortal(SvREFCNT_inc(HeKEY_sv(entry))); |
44a8e56a | 1515 | } |
bc4947fc | 1516 | hv_free_ent(hv, entry); |
79072805 LW |
1517 | } |
1518 | ||
954c1994 GS |
1519 | /* |
1520 | =for apidoc hv_clear | |
1521 | ||
1522 | Clears a hash, making it empty. | |
1523 | ||
1524 | =cut | |
1525 | */ | |
1526 | ||
79072805 | 1527 | void |
864dbfa3 | 1528 | Perl_hv_clear(pTHX_ HV *hv) |
79072805 | 1529 | { |
27da23d5 | 1530 | dVAR; |
cbec9347 | 1531 | register XPVHV* xhv; |
79072805 LW |
1532 | if (!hv) |
1533 | return; | |
49293501 | 1534 | |
ecae49c0 NC |
1535 | DEBUG_A(Perl_hv_assert(aTHX_ hv)); |
1536 | ||
34c3c4e3 DM |
1537 | xhv = (XPVHV*)SvANY(hv); |
1538 | ||
7b2c381c | 1539 | if (SvREADONLY(hv) && HvARRAY(hv) != NULL) { |
34c3c4e3 | 1540 | /* restricted hash: convert all keys to placeholders */ |
b464bac0 AL |
1541 | STRLEN i; |
1542 | for (i = 0; i <= xhv->xhv_max; i++) { | |
7b2c381c | 1543 | HE *entry = (HvARRAY(hv))[i]; |
3a676441 JH |
1544 | for (; entry; entry = HeNEXT(entry)) { |
1545 | /* not already placeholder */ | |
7996736c | 1546 | if (HeVAL(entry) != &PL_sv_placeholder) { |
3a676441 | 1547 | if (HeVAL(entry) && SvREADONLY(HeVAL(entry))) { |
6136c704 | 1548 | SV* const keysv = hv_iterkeysv(entry); |
3a676441 | 1549 | Perl_croak(aTHX_ |
95b63a38 JH |
1550 | "Attempt to delete readonly key '%"SVf"' from a restricted hash", |
1551 | (void*)keysv); | |
3a676441 JH |
1552 | } |
1553 | SvREFCNT_dec(HeVAL(entry)); | |
7996736c | 1554 | HeVAL(entry) = &PL_sv_placeholder; |
ca732855 | 1555 | HvPLACEHOLDERS(hv)++; |
3a676441 | 1556 | } |
34c3c4e3 DM |
1557 | } |
1558 | } | |
df8c6964 | 1559 | goto reset; |
49293501 MS |
1560 | } |
1561 | ||
463ee0b2 | 1562 | hfreeentries(hv); |
ca732855 | 1563 | HvPLACEHOLDERS_set(hv, 0); |
7b2c381c | 1564 | if (HvARRAY(hv)) |
41f62432 | 1565 | Zero(HvARRAY(hv), xhv->xhv_max+1 /* HvMAX(hv)+1 */, HE*); |
a0d0e21e LW |
1566 | |
1567 | if (SvRMAGICAL(hv)) | |
ad64d0ec | 1568 | mg_clear(MUTABLE_SV(hv)); |
574c8022 | 1569 | |
19692e8d | 1570 | HvHASKFLAGS_off(hv); |
bb443f97 | 1571 | HvREHASH_off(hv); |
df8c6964 | 1572 | reset: |
b79f7545 | 1573 | if (SvOOK(hv)) { |
dd69841b BB |
1574 | if(HvNAME_get(hv)) |
1575 | mro_isa_changed_in(hv); | |
bfcb3514 NC |
1576 | HvEITER_set(hv, NULL); |
1577 | } | |
79072805 LW |
1578 | } |
1579 | ||
3540d4ce AB |
1580 | /* |
1581 | =for apidoc hv_clear_placeholders | |
1582 | ||
1583 | Clears any placeholders from a hash. If a restricted hash has any of its keys | |
1584 | marked as readonly and the key is subsequently deleted, the key is not actually | |
1585 | deleted but is marked by assigning it a value of &PL_sv_placeholder. This tags | |
1586 | it so it will be ignored by future operations such as iterating over the hash, | |
4cdaeff7 | 1587 | but will still allow the hash to have a value reassigned to the key at some |
3540d4ce AB |
1588 | future point. This function clears any such placeholder keys from the hash. |
1589 | See Hash::Util::lock_keys() for an example of its use. | |
1590 | ||
1591 | =cut | |
1592 | */ | |
1593 | ||
1594 | void | |
1595 | Perl_hv_clear_placeholders(pTHX_ HV *hv) | |
1596 | { | |
27da23d5 | 1597 | dVAR; |
b3ca2e83 NC |
1598 | const U32 items = (U32)HvPLACEHOLDERS_get(hv); |
1599 | ||
7918f24d NC |
1600 | PERL_ARGS_ASSERT_HV_CLEAR_PLACEHOLDERS; |
1601 | ||
b3ca2e83 NC |
1602 | if (items) |
1603 | clear_placeholders(hv, items); | |
1604 | } | |
1605 | ||
1606 | static void | |
1607 | S_clear_placeholders(pTHX_ HV *hv, U32 items) | |
1608 | { | |
1609 | dVAR; | |
b464bac0 | 1610 | I32 i; |
d3677389 | 1611 | |
7918f24d NC |
1612 | PERL_ARGS_ASSERT_CLEAR_PLACEHOLDERS; |
1613 | ||
d3677389 NC |
1614 | if (items == 0) |
1615 | return; | |
1616 | ||
b464bac0 | 1617 | i = HvMAX(hv); |
d3677389 NC |
1618 | do { |
1619 | /* Loop down the linked list heads */ | |
6136c704 | 1620 | bool first = TRUE; |
d3677389 | 1621 | HE **oentry = &(HvARRAY(hv))[i]; |
cf6db12b | 1622 | HE *entry; |
d3677389 | 1623 | |
cf6db12b | 1624 | while ((entry = *oentry)) { |
d3677389 NC |
1625 | if (HeVAL(entry) == &PL_sv_placeholder) { |
1626 | *oentry = HeNEXT(entry); | |
2e58978b | 1627 | if (entry == HvEITER_get(hv)) |
d3677389 NC |
1628 | HvLAZYDEL_on(hv); |
1629 | else | |
1630 | hv_free_ent(hv, entry); | |
1631 | ||
1632 | if (--items == 0) { | |
1633 | /* Finished. */ | |
5d88ecd7 | 1634 | HvTOTALKEYS(hv) -= (IV)HvPLACEHOLDERS_get(hv); |
d3677389 NC |
1635 | if (HvKEYS(hv) == 0) |
1636 | HvHASKFLAGS_off(hv); | |
5d88ecd7 | 1637 | HvPLACEHOLDERS_set(hv, 0); |
d3677389 NC |
1638 | return; |
1639 | } | |
213ce8b3 NC |
1640 | } else { |
1641 | oentry = &HeNEXT(entry); | |
6136c704 | 1642 | first = FALSE; |
d3677389 NC |
1643 | } |
1644 | } | |
1645 | } while (--i >= 0); | |
1646 | /* You can't get here, hence assertion should always fail. */ | |
1647 | assert (items == 0); | |
1648 | assert (0); | |
3540d4ce AB |
1649 | } |
1650 | ||
76e3520e | 1651 | STATIC void |
cea2e8a9 | 1652 | S_hfreeentries(pTHX_ HV *hv) |
79072805 | 1653 | { |
23976bdd | 1654 | /* This is the array that we're going to restore */ |
fd7de8a8 | 1655 | HE **const orig_array = HvARRAY(hv); |
23976bdd NC |
1656 | HEK *name; |
1657 | int attempts = 100; | |
3abe233e | 1658 | |
7918f24d NC |
1659 | PERL_ARGS_ASSERT_HFREEENTRIES; |
1660 | ||
fd7de8a8 | 1661 | if (!orig_array) |
79072805 | 1662 | return; |
a0d0e21e | 1663 | |
f1c32fec BM |
1664 | if (HvNAME(hv) && orig_array != NULL) { |
1665 | /* symbol table: make all the contained subs ANON */ | |
1666 | STRLEN i; | |
1667 | XPVHV *xhv = (XPVHV*)SvANY(hv); | |
1668 | ||
1669 | for (i = 0; i <= xhv->xhv_max; i++) { | |
1670 | HE *entry = (HvARRAY(hv))[i]; | |
1671 | for (; entry; entry = HeNEXT(entry)) { | |
1672 | SV *val = HeVAL(entry); | |
1673 | /* we need to put the subs in the __ANON__ symtable, as | |
1674 | * this one is being cleared. */ | |
1675 | anonymise_cv(NULL, val); | |
1676 | } | |
1677 | } | |
1678 | } | |
1679 | ||
23976bdd NC |
1680 | if (SvOOK(hv)) { |
1681 | /* If the hash is actually a symbol table with a name, look after the | |
1682 | name. */ | |
1683 | struct xpvhv_aux *iter = HvAUX(hv); | |
1684 | ||
1685 | name = iter->xhv_name; | |
1686 | iter->xhv_name = NULL; | |
1687 | } else { | |
1688 | name = NULL; | |
1689 | } | |
1690 | ||
23976bdd NC |
1691 | /* orig_array remains unchanged throughout the loop. If after freeing all |
1692 | the entries it turns out that one of the little blighters has triggered | |
1693 | an action that has caused HvARRAY to be re-allocated, then we set | |
1694 | array to the new HvARRAY, and try again. */ | |
1695 | ||
1696 | while (1) { | |
1697 | /* This is the one we're going to try to empty. First time round | |
1698 | it's the original array. (Hopefully there will only be 1 time | |
1699 | round) */ | |
6136c704 | 1700 | HE ** const array = HvARRAY(hv); |
7440661e | 1701 | I32 i = HvMAX(hv); |
23976bdd NC |
1702 | |
1703 | /* Because we have taken xhv_name out, the only allocated pointer | |
1704 | in the aux structure that might exist is the backreference array. | |
1705 | */ | |
1706 | ||
1707 | if (SvOOK(hv)) { | |
7440661e | 1708 | HE *entry; |
e1a479c5 | 1709 | struct mro_meta *meta; |
23976bdd NC |
1710 | struct xpvhv_aux *iter = HvAUX(hv); |
1711 | /* If there are weak references to this HV, we need to avoid | |
1712 | freeing them up here. In particular we need to keep the AV | |
1713 | visible as what we're deleting might well have weak references | |
1714 | back to this HV, so the for loop below may well trigger | |
1715 | the removal of backreferences from this array. */ | |
1716 | ||
1717 | if (iter->xhv_backreferences) { | |
1718 | /* So donate them to regular backref magic to keep them safe. | |
1719 | The sv_magic will increase the reference count of the AV, | |
1720 | so we need to drop it first. */ | |
5b285ea4 | 1721 | SvREFCNT_dec(iter->xhv_backreferences); |
23976bdd NC |
1722 | if (AvFILLp(iter->xhv_backreferences) == -1) { |
1723 | /* Turns out that the array is empty. Just free it. */ | |
1724 | SvREFCNT_dec(iter->xhv_backreferences); | |
1b8791d1 | 1725 | |
23976bdd | 1726 | } else { |
ad64d0ec NC |
1727 | sv_magic(MUTABLE_SV(hv), |
1728 | MUTABLE_SV(iter->xhv_backreferences), | |
23976bdd NC |
1729 | PERL_MAGIC_backref, NULL, 0); |
1730 | } | |
1731 | iter->xhv_backreferences = NULL; | |
5b285ea4 | 1732 | } |
86f55936 | 1733 | |
23976bdd NC |
1734 | entry = iter->xhv_eiter; /* HvEITER(hv) */ |
1735 | if (entry && HvLAZYDEL(hv)) { /* was deleted earlier? */ | |
1736 | HvLAZYDEL_off(hv); | |
1737 | hv_free_ent(hv, entry); | |
1738 | } | |
1739 | iter->xhv_riter = -1; /* HvRITER(hv) = -1 */ | |
4608196e | 1740 | iter->xhv_eiter = NULL; /* HvEITER(hv) = NULL */ |
b79f7545 | 1741 | |
e1a479c5 | 1742 | if((meta = iter->xhv_mro_meta)) { |
9953ff72 NC |
1743 | if (meta->mro_linear_all) { |
1744 | SvREFCNT_dec(MUTABLE_SV(meta->mro_linear_all)); | |
1745 | meta->mro_linear_all = NULL; | |
553e831a | 1746 | /* This is just acting as a shortcut pointer. */ |
3a6fa573 NC |
1747 | meta->mro_linear_current = NULL; |
1748 | } else if (meta->mro_linear_current) { | |
553e831a NC |
1749 | /* Only the current MRO is stored, so this owns the data. |
1750 | */ | |
3a6fa573 NC |
1751 | SvREFCNT_dec(meta->mro_linear_current); |
1752 | meta->mro_linear_current = NULL; | |
553e831a | 1753 | } |
e1a479c5 | 1754 | if(meta->mro_nextmethod) SvREFCNT_dec(meta->mro_nextmethod); |
a49ba3fc | 1755 | SvREFCNT_dec(meta->isa); |
e1a479c5 BB |
1756 | Safefree(meta); |
1757 | iter->xhv_mro_meta = NULL; | |
1758 | } | |
1759 | ||
23976bdd | 1760 | /* There are now no allocated pointers in the aux structure. */ |
2f86008e | 1761 | |
23976bdd NC |
1762 | SvFLAGS(hv) &= ~SVf_OOK; /* Goodbye, aux structure. */ |
1763 | /* What aux structure? */ | |
a0d0e21e | 1764 | } |
bfcb3514 | 1765 | |
23976bdd NC |
1766 | /* make everyone else think the array is empty, so that the destructors |
1767 | * called for freed entries can't recusively mess with us */ | |
1768 | HvARRAY(hv) = NULL; | |
23976bdd NC |
1769 | ((XPVHV*) SvANY(hv))->xhv_keys = 0; |
1770 | ||
7440661e NC |
1771 | |
1772 | do { | |
1773 | /* Loop down the linked list heads */ | |
1774 | HE *entry = array[i]; | |
1775 | ||
1776 | while (entry) { | |
23976bdd NC |
1777 | register HE * const oentry = entry; |
1778 | entry = HeNEXT(entry); | |
1779 | hv_free_ent(hv, oentry); | |
1780 | } | |
7440661e | 1781 | } while (--i >= 0); |
b79f7545 | 1782 | |
23976bdd NC |
1783 | /* As there are no allocated pointers in the aux structure, it's now |
1784 | safe to free the array we just cleaned up, if it's not the one we're | |
1785 | going to put back. */ | |
1786 | if (array != orig_array) { | |
1787 | Safefree(array); | |
1788 | } | |
b79f7545 | 1789 | |
23976bdd NC |
1790 | if (!HvARRAY(hv)) { |
1791 | /* Good. No-one added anything this time round. */ | |
1792 | break; | |
bfcb3514 | 1793 | } |
b79f7545 | 1794 | |
23976bdd NC |
1795 | if (SvOOK(hv)) { |
1796 | /* Someone attempted to iterate or set the hash name while we had | |
1797 | the array set to 0. We'll catch backferences on the next time | |
1798 | round the while loop. */ | |
1799 | assert(HvARRAY(hv)); | |
1b8791d1 | 1800 | |
23976bdd NC |
1801 | if (HvAUX(hv)->xhv_name) { |
1802 | unshare_hek_or_pvn(HvAUX(hv)->xhv_name, 0, 0, 0); | |
1803 | } | |
1804 | } | |
1805 | ||
1806 | if (--attempts == 0) { | |
1807 | Perl_die(aTHX_ "panic: hfreeentries failed to free hash - something is repeatedly re-creating entries"); | |
1808 | } | |
6136c704 | 1809 | } |
23976bdd NC |
1810 | |
1811 | HvARRAY(hv) = orig_array; | |
1812 | ||
1813 | /* If the hash was actually a symbol table, put the name back. */ | |
1814 | if (name) { | |
1815 | /* We have restored the original array. If name is non-NULL, then | |
1816 | the original array had an aux structure at the end. So this is | |
1817 | valid: */ | |
1818 | SvFLAGS(hv) |= SVf_OOK; | |
1819 | HvAUX(hv)->xhv_name = name; | |
1b8791d1 | 1820 | } |
79072805 LW |
1821 | } |
1822 | ||
954c1994 GS |
1823 | /* |
1824 | =for apidoc hv_undef | |
1825 | ||
1826 | Undefines the hash. | |
1827 | ||
1828 | =cut | |
1829 | */ | |
1830 | ||
79072805 | 1831 | void |
864dbfa3 | 1832 | Perl_hv_undef(pTHX_ HV *hv) |
79072805 | 1833 | { |
97aff369 | 1834 | dVAR; |
cbec9347 | 1835 | register XPVHV* xhv; |
bfcb3514 | 1836 | const char *name; |
86f55936 | 1837 | |
79072805 LW |
1838 | if (!hv) |
1839 | return; | |
ecae49c0 | 1840 | DEBUG_A(Perl_hv_assert(aTHX_ hv)); |
cbec9347 | 1841 | xhv = (XPVHV*)SvANY(hv); |
dd69841b | 1842 | |
0fa56319 | 1843 | if ((name = HvNAME_get(hv)) && !PL_dirty) |
dd69841b BB |
1844 | mro_isa_changed_in(hv); |
1845 | ||
463ee0b2 | 1846 | hfreeentries(hv); |
dd69841b | 1847 | if (name) { |
04fe65b0 RGS |
1848 | if (PL_stashcache) |
1849 | (void)hv_delete(PL_stashcache, name, HvNAMELEN_get(hv), G_DISCARD); | |
bd61b366 | 1850 | hv_name_set(hv, NULL, 0, 0); |
85e6fe83 | 1851 | } |
b79f7545 NC |
1852 | SvFLAGS(hv) &= ~SVf_OOK; |
1853 | Safefree(HvARRAY(hv)); | |
cbec9347 | 1854 | xhv->xhv_max = 7; /* HvMAX(hv) = 7 (it's a normal hash) */ |
7b2c381c | 1855 | HvARRAY(hv) = 0; |
ca732855 | 1856 | HvPLACEHOLDERS_set(hv, 0); |
a0d0e21e LW |
1857 | |
1858 | if (SvRMAGICAL(hv)) | |
ad64d0ec | 1859 | mg_clear(MUTABLE_SV(hv)); |
79072805 LW |
1860 | } |
1861 | ||
4d0fbddd NC |
1862 | /* |
1863 | =for apidoc hv_fill | |
1864 | ||
1865 | Returns the number of hash buckets that happen to be in use. This function is | |
1866 | wrapped by the macro C<HvFILL>. | |
1867 | ||
1868 | Previously this value was stored in the HV structure, rather than being | |
1869 | calculated on demand. | |
1870 | ||
1871 | =cut | |
1872 | */ | |
1873 | ||
1874 | STRLEN | |
1875 | Perl_hv_fill(pTHX_ HV const *const hv) | |
1876 | { | |
1877 | STRLEN count = 0; | |
1878 | HE **ents = HvARRAY(hv); | |
1879 | ||
1880 | PERL_ARGS_ASSERT_HV_FILL; | |
1881 | ||
1882 | if (ents) { | |
fcd24582 NC |
1883 | HE *const *const last = ents + HvMAX(hv); |
1884 | count = last + 1 - ents; | |
4d0fbddd NC |
1885 | |
1886 | do { | |
fcd24582 NC |
1887 | if (!*ents) |
1888 | --count; | |
1889 | } while (++ents <= last); | |
4d0fbddd NC |
1890 | } |
1891 | return count; | |
1892 | } | |
1893 | ||
b464bac0 | 1894 | static struct xpvhv_aux* |
5f66b61c | 1895 | S_hv_auxinit(HV *hv) { |
bfcb3514 | 1896 | struct xpvhv_aux *iter; |
b79f7545 | 1897 | char *array; |
bfcb3514 | 1898 | |
7918f24d NC |
1899 | PERL_ARGS_ASSERT_HV_AUXINIT; |
1900 | ||
b79f7545 | 1901 | if (!HvARRAY(hv)) { |
a02a5408 | 1902 | Newxz(array, PERL_HV_ARRAY_ALLOC_BYTES(HvMAX(hv) + 1) |
b79f7545 NC |
1903 | + sizeof(struct xpvhv_aux), char); |
1904 | } else { | |
1905 | array = (char *) HvARRAY(hv); | |
1906 | Renew(array, PERL_HV_ARRAY_ALLOC_BYTES(HvMAX(hv) + 1) | |
1907 | + sizeof(struct xpvhv_aux), char); | |
1908 | } | |
1909 | HvARRAY(hv) = (HE**) array; | |
1910 | /* SvOOK_on(hv) attacks the IV flags. */ | |
1911 | SvFLAGS(hv) |= SVf_OOK; | |
1912 | iter = HvAUX(hv); | |
bfcb3514 NC |
1913 | |
1914 | iter->xhv_riter = -1; /* HvRITER(hv) = -1 */ | |
4608196e | 1915 | iter->xhv_eiter = NULL; /* HvEITER(hv) = NULL */ |
bfcb3514 | 1916 | iter->xhv_name = 0; |
86f55936 | 1917 | iter->xhv_backreferences = 0; |
e1a479c5 | 1918 | iter->xhv_mro_meta = NULL; |
bfcb3514 NC |
1919 | return iter; |
1920 | } | |
1921 | ||
954c1994 GS |
1922 | /* |
1923 | =for apidoc hv_iterinit | |
1924 | ||
1925 | Prepares a starting point to traverse a hash table. Returns the number of | |
1926 | keys in the hash (i.e. the same as C<HvKEYS(tb)>). The return value is | |
1c846c1f | 1927 | currently only meaningful for hashes without tie magic. |
954c1994 GS |
1928 | |
1929 | NOTE: Before version 5.004_65, C<hv_iterinit> used to return the number of | |
1930 | hash buckets that happen to be in use. If you still need that esoteric | |
1931 | value, you can get it through the macro C<HvFILL(tb)>. | |
1932 | ||
e16e2ff8 | 1933 | |
954c1994 GS |
1934 | =cut |
1935 | */ | |
1936 | ||
79072805 | 1937 | I32 |
864dbfa3 | 1938 | Perl_hv_iterinit(pTHX_ HV *hv) |
79072805 | 1939 | { |
7918f24d NC |
1940 | PERL_ARGS_ASSERT_HV_ITERINIT; |
1941 | ||
1942 | /* FIXME: Are we not NULL, or do we croak? Place bets now! */ | |
1943 | ||
aa689395 | 1944 | if (!hv) |
cea2e8a9 | 1945 | Perl_croak(aTHX_ "Bad hash"); |
bfcb3514 | 1946 | |
b79f7545 | 1947 | if (SvOOK(hv)) { |
6136c704 | 1948 | struct xpvhv_aux * const iter = HvAUX(hv); |
0bd48802 | 1949 | HE * const entry = iter->xhv_eiter; /* HvEITER(hv) */ |
bfcb3514 NC |
1950 | if (entry && HvLAZYDEL(hv)) { /* was deleted earlier? */ |
1951 | HvLAZYDEL_off(hv); | |
1952 | hv_free_ent(hv, entry); | |
1953 | } | |
1954 | iter->xhv_riter = -1; /* HvRITER(hv) = -1 */ | |
4608196e | 1955 | iter->xhv_eiter = NULL; /* HvEITER(hv) = NULL */ |
bfcb3514 | 1956 | } else { |
6136c704 | 1957 | hv_auxinit(hv); |
72940dca | 1958 | } |
44a2ac75 | 1959 | |
cbec9347 | 1960 | /* used to be xhv->xhv_fill before 5.004_65 */ |
5d88ecd7 | 1961 | return HvTOTALKEYS(hv); |
79072805 | 1962 | } |
bfcb3514 NC |
1963 | |
1964 | I32 * | |
1965 | Perl_hv_riter_p(pTHX_ HV *hv) { | |
1966 | struct xpvhv_aux *iter; | |
1967 | ||
7918f24d NC |
1968 | PERL_ARGS_ASSERT_HV_RITER_P; |
1969 | ||
bfcb3514 NC |
1970 | if (!hv) |
1971 | Perl_croak(aTHX_ "Bad hash"); | |
1972 | ||
6136c704 | 1973 | iter = SvOOK(hv) ? HvAUX(hv) : hv_auxinit(hv); |
bfcb3514 NC |
1974 | return &(iter->xhv_riter); |
1975 | } | |
1976 | ||
1977 | HE ** | |
1978 | Perl_hv_eiter_p(pTHX_ HV *hv) { | |
1979 | struct xpvhv_aux *iter; | |
1980 | ||
7918f24d NC |
1981 | PERL_ARGS_ASSERT_HV_EITER_P; |
1982 | ||
bfcb3514 NC |
1983 | if (!hv) |
1984 | Perl_croak(aTHX_ "Bad hash"); | |
1985 | ||
6136c704 | 1986 | iter = SvOOK(hv) ? HvAUX(hv) : hv_auxinit(hv); |
bfcb3514 NC |
1987 | return &(iter->xhv_eiter); |
1988 | } | |
1989 | ||
1990 | void | |
1991 | Perl_hv_riter_set(pTHX_ HV *hv, I32 riter) { | |
1992 | struct xpvhv_aux *iter; | |
1993 | ||
7918f24d NC |
1994 | PERL_ARGS_ASSERT_HV_RITER_SET; |
1995 | ||
bfcb3514 NC |
1996 | if (!hv) |
1997 | Perl_croak(aTHX_ "Bad hash"); | |
1998 | ||
b79f7545 NC |
1999 | if (SvOOK(hv)) { |
2000 | iter = HvAUX(hv); | |
2001 | } else { | |
bfcb3514 NC |
2002 | if (riter == -1) |
2003 | return; | |
2004 | ||
6136c704 | 2005 | iter = hv_auxinit(hv); |
bfcb3514 NC |
2006 | } |
2007 | iter->xhv_riter = riter; | |
2008 | } | |
2009 | ||
2010 | void | |
2011 | Perl_hv_eiter_set(pTHX_ HV *hv, HE *eiter) { | |
2012 | struct xpvhv_aux *iter; | |
2013 | ||
7918f24d NC |
2014 | PERL_ARGS_ASSERT_HV_EITER_SET; |
2015 | ||
bfcb3514 NC |
2016 | if (!hv) |
2017 | Perl_croak(aTHX_ "Bad hash"); | |
2018 | ||
b79f7545 NC |
2019 | if (SvOOK(hv)) { |
2020 | iter = HvAUX(hv); | |
2021 | } else { | |
bfcb3514 NC |
2022 | /* 0 is the default so don't go malloc()ing a new structure just to |
2023 | hold 0. */ | |
2024 | if (!eiter) | |
2025 | return; | |
2026 | ||
6136c704 | 2027 | iter = hv_auxinit(hv); |
bfcb3514 NC |
2028 | } |
2029 | iter->xhv_eiter = eiter; | |
2030 | } | |
2031 | ||
bfcb3514 | 2032 | void |
4164be69 | 2033 | Perl_hv_name_set(pTHX_ HV *hv, const char *name, U32 len, U32 flags) |
bfcb3514 | 2034 | { |
97aff369 | 2035 | dVAR; |
b79f7545 | 2036 | struct xpvhv_aux *iter; |
7423f6db | 2037 | U32 hash; |
46c461b5 | 2038 | |
7918f24d | 2039 | PERL_ARGS_ASSERT_HV_NAME_SET; |
46c461b5 | 2040 | PERL_UNUSED_ARG(flags); |
bfcb3514 | 2041 | |
4164be69 NC |
2042 | if (len > I32_MAX) |
2043 | Perl_croak(aTHX_ "panic: hv name too long (%"UVuf")", (UV) len); | |
2044 | ||
b79f7545 NC |
2045 | if (SvOOK(hv)) { |
2046 | iter = HvAUX(hv); | |
7423f6db NC |
2047 | if (iter->xhv_name) { |
2048 | unshare_hek_or_pvn(iter->xhv_name, 0, 0, 0); | |
2049 | } | |
16580ff5 | 2050 | } else { |
bfcb3514 NC |
2051 | if (name == 0) |
2052 | return; | |
2053 | ||
6136c704 | 2054 | iter = hv_auxinit(hv); |
bfcb3514 | 2055 | } |
7423f6db | 2056 | PERL_HASH(hash, name, len); |
adf4e37a | 2057 | iter->xhv_name = name ? share_hek(name, len, hash) : NULL; |
bfcb3514 NC |
2058 | } |
2059 | ||
86f55936 NC |
2060 | AV ** |
2061 | Perl_hv_backreferences_p(pTHX_ HV *hv) { | |
6136c704 | 2062 | struct xpvhv_aux * const iter = SvOOK(hv) ? HvAUX(hv) : hv_auxinit(hv); |
7918f24d NC |
2063 | |
2064 | PERL_ARGS_ASSERT_HV_BACKREFERENCES_P; | |
96a5add6 | 2065 | PERL_UNUSED_CONTEXT; |
7918f24d | 2066 | |
86f55936 NC |
2067 | return &(iter->xhv_backreferences); |
2068 | } | |
2069 | ||
2070 | void | |
2071 | Perl_hv_kill_backrefs(pTHX_ HV *hv) { | |
2072 | AV *av; | |
2073 | ||
7918f24d NC |
2074 | PERL_ARGS_ASSERT_HV_KILL_BACKREFS; |
2075 | ||
86f55936 NC |
2076 | if (!SvOOK(hv)) |
2077 | return; | |
2078 | ||
2079 | av = HvAUX(hv)->xhv_backreferences; | |
2080 | ||
2081 | if (av) { | |
2082 | HvAUX(hv)->xhv_backreferences = 0; | |
ad64d0ec | 2083 | Perl_sv_kill_backrefs(aTHX_ MUTABLE_SV(hv), av); |
b17f5ab7 | 2084 | SvREFCNT_dec(av); |
86f55936 NC |
2085 | } |
2086 | } | |
2087 | ||
954c1994 | 2088 | /* |
7a7b9979 NC |
2089 | hv_iternext is implemented as a macro in hv.h |
2090 | ||
954c1994 GS |
2091 | =for apidoc hv_iternext |
2092 | ||
2093 | Returns entries from a hash iterator. See C<hv_iterinit>. | |
2094 | ||
fe7bca90 NC |
2095 | You may call C<hv_delete> or C<hv_delete_ent> on the hash entry that the |
2096 | iterator currently points to, without losing your place or invalidating your | |
2097 | iterator. Note that in this case the current entry is deleted from the hash | |
2098 | with your iterator holding the last reference to it. Your iterator is flagged | |
2099 | to free the entry on the next call to C<hv_iternext>, so you must not discard | |
2100 | your iterator immediately else the entry will leak - call C<hv_iternext> to | |
2101 | trigger the resource deallocation. | |
2102 | ||
fe7bca90 NC |
2103 | =for apidoc hv_iternext_flags |
2104 | ||
2105 | Returns entries from a hash iterator. See C<hv_iterinit> and C<hv_iternext>. | |
2106 | The C<flags> value will normally be zero; if HV_ITERNEXT_WANTPLACEHOLDERS is | |
2107 | set the placeholders keys (for restricted hashes) will be returned in addition | |
2108 | to normal keys. By default placeholders are automatically skipped over. | |
7996736c MHM |
2109 | Currently a placeholder is implemented with a value that is |
2110 | C<&Perl_sv_placeholder>. Note that the implementation of placeholders and | |
fe7bca90 NC |
2111 | restricted hashes may change, and the implementation currently is |
2112 | insufficiently abstracted for any change to be tidy. | |
e16e2ff8 | 2113 | |
fe7bca90 | 2114 | =cut |
e16e2ff8 NC |
2115 | */ |
2116 | ||
2117 | HE * | |
2118 | Perl_hv_iternext_flags(pTHX_ HV *hv, I32 flags) | |
2119 | { | |
27da23d5 | 2120 | dVAR; |
cbec9347 | 2121 | register XPVHV* xhv; |
79072805 | 2122 | register HE *entry; |
a0d0e21e | 2123 | HE *oldentry; |
463ee0b2 | 2124 | MAGIC* mg; |
bfcb3514 | 2125 | struct xpvhv_aux *iter; |
79072805 | 2126 | |
7918f24d NC |
2127 | PERL_ARGS_ASSERT_HV_ITERNEXT_FLAGS; |
2128 | ||
79072805 | 2129 | if (!hv) |
cea2e8a9 | 2130 | Perl_croak(aTHX_ "Bad hash"); |
81714fb9 | 2131 | |
cbec9347 | 2132 | xhv = (XPVHV*)SvANY(hv); |
bfcb3514 | 2133 | |
b79f7545 | 2134 | if (!SvOOK(hv)) { |
bfcb3514 NC |
2135 | /* Too many things (well, pp_each at least) merrily assume that you can |
2136 | call iv_iternext without calling hv_iterinit, so we'll have to deal | |
2137 | with it. */ | |
2138 | hv_iterinit(hv); | |
bfcb3514 | 2139 | } |
b79f7545 | 2140 | iter = HvAUX(hv); |
bfcb3514 NC |
2141 | |
2142 | oldentry = entry = iter->xhv_eiter; /* HvEITER(hv) */ | |
e62cc96a | 2143 | if (SvMAGICAL(hv) && SvRMAGICAL(hv)) { |
ad64d0ec | 2144 | if ( ( mg = mg_find((const SV *)hv, PERL_MAGIC_tied) ) ) { |
e62cc96a YO |
2145 | SV * const key = sv_newmortal(); |
2146 | if (entry) { | |
2147 | sv_setsv(key, HeSVKEY_force(entry)); | |
2148 | SvREFCNT_dec(HeSVKEY(entry)); /* get rid of previous key */ | |
2149 | } | |
2150 | else { | |
2151 | char *k; | |
2152 | HEK *hek; | |
2153 | ||
2154 | /* one HE per MAGICAL hash */ | |
2155 | iter->xhv_eiter = entry = new_HE(); /* HvEITER(hv) = new_HE() */ | |
2156 | Zero(entry, 1, HE); | |
ad64d0ec | 2157 | Newxz(k, HEK_BASESIZE + sizeof(const SV *), char); |
e62cc96a YO |
2158 | hek = (HEK*)k; |
2159 | HeKEY_hek(entry) = hek; | |
2160 | HeKLEN(entry) = HEf_SVKEY; | |
2161 | } | |
ad64d0ec | 2162 | magic_nextpack(MUTABLE_SV(hv),mg,key); |
e62cc96a YO |
2163 | if (SvOK(key)) { |
2164 | /* force key to stay around until next time */ | |
2165 | HeSVKEY_set(entry, SvREFCNT_inc_simple_NN(key)); | |
2166 | return entry; /* beware, hent_val is not set */ | |
2167 | } | |
ef8d46e8 | 2168 | SvREFCNT_dec(HeVAL(entry)); |
e62cc96a YO |
2169 | Safefree(HeKEY_hek(entry)); |
2170 | del_HE(entry); | |
2171 | iter->xhv_eiter = NULL; /* HvEITER(hv) = NULL */ | |
2172 | return NULL; | |
81714fb9 | 2173 | } |
79072805 | 2174 | } |
7ee146b1 | 2175 | #if defined(DYNAMIC_ENV_FETCH) && !defined(__riscos__) /* set up %ENV for iteration */ |
ad64d0ec NC |
2176 | if (!entry && SvRMAGICAL((const SV *)hv) |
2177 | && mg_find((const SV *)hv, PERL_MAGIC_env)) { | |
f675dbe5 | 2178 | prime_env_iter(); |
03026e68 JM |
2179 | #ifdef VMS |
2180 | /* The prime_env_iter() on VMS just loaded up new hash values | |
2181 | * so the iteration count needs to be reset back to the beginning | |
2182 | */ | |
2183 | hv_iterinit(hv); | |
2184 | iter = HvAUX(hv); | |
2185 | oldentry = entry = iter->xhv_eiter; /* HvEITER(hv) */ | |
2186 | #endif | |
2187 | } | |
f675dbe5 | 2188 | #endif |
463ee0b2 | 2189 | |
b79f7545 NC |
2190 | /* hv_iterint now ensures this. */ |
2191 | assert (HvARRAY(hv)); | |
2192 | ||
015a5f36 | 2193 | /* At start of hash, entry is NULL. */ |
fde52b5c | 2194 | if (entry) |
8aacddc1 | 2195 | { |
fde52b5c | 2196 | entry = HeNEXT(entry); |
e16e2ff8 NC |
2197 | if (!(flags & HV_ITERNEXT_WANTPLACEHOLDERS)) { |
2198 | /* | |
2199 | * Skip past any placeholders -- don't want to include them in | |
2200 | * any iteration. | |
2201 | */ | |
7996736c | 2202 | while (entry && HeVAL(entry) == &PL_sv_placeholder) { |
e16e2ff8 NC |
2203 | entry = HeNEXT(entry); |
2204 | } | |
8aacddc1 NIS |
2205 | } |
2206 | } | |
015a5f36 | 2207 | |
9eb4ebd1 NC |
2208 | /* Skip the entire loop if the hash is empty. */ |
2209 | if ((flags & HV_ITERNEXT_WANTPLACEHOLDERS) | |
2210 | ? HvTOTALKEYS(hv) : HvUSEDKEYS(hv)) { | |
900ac051 MM |
2211 | while (!entry) { |
2212 | /* OK. Come to the end of the current list. Grab the next one. */ | |
2213 | ||
2214 | iter->xhv_riter++; /* HvRITER(hv)++ */ | |
2215 | if (iter->xhv_riter > (I32)xhv->xhv_max /* HvRITER(hv) > HvMAX(hv) */) { | |
2216 | /* There is no next one. End of the hash. */ | |
2217 | iter->xhv_riter = -1; /* HvRITER(hv) = -1 */ | |
2218 | break; | |
2219 | } | |
2220 | entry = (HvARRAY(hv))[iter->xhv_riter]; | |
8aacddc1 | 2221 | |
900ac051 MM |
2222 | if (!(flags & HV_ITERNEXT_WANTPLACEHOLDERS)) { |
2223 | /* If we have an entry, but it's a placeholder, don't count it. | |
2224 | Try the next. */ | |
2225 | while (entry && HeVAL(entry) == &PL_sv_placeholder) | |
2226 | entry = HeNEXT(entry); | |
2227 | } | |
2228 | /* Will loop again if this linked list starts NULL | |
2229 | (for HV_ITERNEXT_WANTPLACEHOLDERS) | |
2230 | or if we run through it and find only placeholders. */ | |
015a5f36 | 2231 | } |
fde52b5c | 2232 | } |
79072805 | 2233 | |
72940dca | 2234 | if (oldentry && HvLAZYDEL(hv)) { /* was deleted earlier? */ |
2235 | HvLAZYDEL_off(hv); | |
68dc0745 | 2236 | hv_free_ent(hv, oldentry); |
72940dca | 2237 | } |
a0d0e21e | 2238 | |
fdcd69b6 | 2239 | /*if (HvREHASH(hv) && entry && !HeKREHASH(entry)) |
6c9570dc | 2240 | PerlIO_printf(PerlIO_stderr(), "Awooga %p %p\n", (void*)hv, (void*)entry);*/ |
fdcd69b6 | 2241 | |
bfcb3514 | 2242 | iter->xhv_eiter = entry; /* HvEITER(hv) = entry */ |
79072805 LW |
2243 | return entry; |
2244 | } | |
2245 | ||
954c1994 GS |
2246 | /* |
2247 | =for apidoc hv_iterkey | |
2248 | ||
2249 | Returns the key from the current position of the hash iterator. See | |
2250 | C<hv_iterinit>. | |
2251 | ||
2252 | =cut | |
2253 | */ | |
2254 | ||
79072805 | 2255 | char * |
864dbfa3 | 2256 | Perl_hv_iterkey(pTHX_ register HE *entry, I32 *retlen) |
79072805 | 2257 | { |
7918f24d NC |
2258 | PERL_ARGS_ASSERT_HV_ITERKEY; |
2259 | ||
fde52b5c | 2260 | if (HeKLEN(entry) == HEf_SVKEY) { |
fb73857a | 2261 | STRLEN len; |
0bd48802 | 2262 | char * const p = SvPV(HeKEY_sv(entry), len); |
fb73857a | 2263 | *retlen = len; |
2264 | return p; | |
fde52b5c | 2265 | } |
2266 | else { | |
2267 | *retlen = HeKLEN(entry); | |
2268 | return HeKEY(entry); | |
2269 | } | |
2270 | } | |
2271 | ||
2272 | /* unlike hv_iterval(), this always returns a mortal copy of the key */ | |
954c1994 GS |
2273 | /* |
2274 | =for apidoc hv_iterkeysv | |
2275 | ||
2276 | Returns the key as an C<SV*> from the current position of the hash | |
2277 | iterator. The return value will always be a mortal copy of the key. Also | |
2278 | see C<hv_iterinit>. | |
2279 | ||
2280 | =cut | |
2281 | */ | |
2282 | ||
fde52b5c | 2283 | SV * |
864dbfa3 | 2284 | Perl_hv_iterkeysv(pTHX_ register HE *entry) |
fde52b5c | 2285 | { |
7918f24d NC |
2286 | PERL_ARGS_ASSERT_HV_ITERKEYSV; |
2287 | ||
c1b02ed8 | 2288 | return sv_2mortal(newSVhek(HeKEY_hek(entry))); |
79072805 LW |
2289 | } |
2290 | ||
954c1994 GS |
2291 | /* |
2292 | =for apidoc hv_iterval | |
2293 | ||
2294 | Returns the value from the current position of the hash iterator. See | |
2295 | C<hv_iterkey>. | |
2296 | ||
2297 | =cut | |
2298 | */ | |
2299 | ||
79072805 | 2300 | SV * |
864dbfa3 | 2301 | Perl_hv_iterval(pTHX_ HV *hv, register HE *entry) |
79072805 | 2302 | { |
7918f24d NC |
2303 | PERL_ARGS_ASSERT_HV_ITERVAL; |
2304 | ||
8990e307 | 2305 | if (SvRMAGICAL(hv)) { |
ad64d0ec | 2306 | if (mg_find((const SV *)hv, PERL_MAGIC_tied)) { |
c4420975 | 2307 | SV* const sv = sv_newmortal(); |
bbce6d69 | 2308 | if (HeKLEN(entry) == HEf_SVKEY) |
ad64d0ec | 2309 | mg_copy(MUTABLE_SV(hv), sv, (char*)HeKEY_sv(entry), HEf_SVKEY); |
a3b680e6 | 2310 | else |
ad64d0ec | 2311 | mg_copy(MUTABLE_SV(hv), sv, HeKEY(entry), HeKLEN(entry)); |
463ee0b2 LW |
2312 | return sv; |
2313 | } | |
79072805 | 2314 | } |
fde52b5c | 2315 | return HeVAL(entry); |
79072805 LW |
2316 | } |
2317 | ||
954c1994 GS |
2318 | /* |
2319 | =for apidoc hv_iternextsv | |
2320 | ||
2321 | Performs an C<hv_iternext>, C<hv_iterkey>, and C<hv_iterval> in one | |
2322 | operation. | |
2323 | ||
2324 | =cut | |
2325 | */ | |
2326 | ||
a0d0e21e | 2327 | SV * |
864dbfa3 | 2328 | Perl_hv_iternextsv(pTHX_ HV *hv, char **key, I32 *retlen) |
a0d0e21e | 2329 | { |
0bd48802 AL |
2330 | HE * const he = hv_iternext_flags(hv, 0); |
2331 | ||
7918f24d NC |
2332 | PERL_ARGS_ASSERT_HV_ITERNEXTSV; |
2333 | ||
0bd48802 | 2334 | if (!he) |
a0d0e21e LW |
2335 | return NULL; |
2336 | *key = hv_iterkey(he, retlen); | |
2337 | return hv_iterval(hv, he); | |
2338 | } | |
2339 | ||
954c1994 | 2340 | /* |
bc5cdc23 NC |
2341 | |
2342 | Now a macro in hv.h | |
2343 | ||
954c1994 GS |
2344 | =for apidoc hv_magic |
2345 | ||
2346 | Adds magic to a hash. See C<sv_magic>. | |
2347 | ||
2348 | =cut | |
2349 | */ | |
2350 | ||
bbce6d69 | 2351 | /* possibly free a shared string if no one has access to it |
fde52b5c | 2352 | * len and hash must both be valid for str. |
2353 | */ | |
bbce6d69 | 2354 | void |
864dbfa3 | 2355 | Perl_unsharepvn(pTHX_ const char *str, I32 len, U32 hash) |
fde52b5c | 2356 | { |
19692e8d NC |
2357 | unshare_hek_or_pvn (NULL, str, len, hash); |
2358 | } | |
2359 | ||
2360 | ||
2361 | void | |
2362 | Perl_unshare_hek(pTHX_ HEK *hek) | |
2363 | { | |
bf11fd37 | 2364 | assert(hek); |
19692e8d NC |
2365 | unshare_hek_or_pvn(hek, NULL, 0, 0); |
2366 | } | |
2367 | ||
2368 | /* possibly free a shared string if no one has access to it | |
2369 | hek if non-NULL takes priority over the other 3, else str, len and hash | |
2370 | are used. If so, len and hash must both be valid for str. | |
2371 | */ | |
df132699 | 2372 | STATIC void |
97ddebaf | 2373 | S_unshare_hek_or_pvn(pTHX_ const HEK *hek, const char *str, I32 len, U32 hash) |
19692e8d | 2374 | { |
97aff369 | 2375 | dVAR; |
cbec9347 | 2376 | register XPVHV* xhv; |
20454177 | 2377 | HE *entry; |
fde52b5c | 2378 | register HE **oentry; |
45d1cc86 | 2379 | HE **first; |
c3654f1a | 2380 | bool is_utf8 = FALSE; |
19692e8d | 2381 | int k_flags = 0; |
aec46f14 | 2382 | const char * const save = str; |
cbbf8932 | 2383 | struct shared_he *he = NULL; |
c3654f1a | 2384 | |
19692e8d | 2385 | if (hek) { |
cbae3960 NC |
2386 | /* Find the shared he which is just before us in memory. */ |
2387 | he = (struct shared_he *)(((char *)hek) | |
2388 | - STRUCT_OFFSET(struct shared_he, | |
2389 | shared_he_hek)); | |
2390 | ||
2391 | /* Assert that the caller passed us a genuine (or at least consistent) | |
2392 | shared hek */ | |
2393 | assert (he->shared_he_he.hent_hek == hek); | |
29404ae0 | 2394 | |
de616631 NC |
2395 | if (he->shared_he_he.he_valu.hent_refcount - 1) { |
2396 | --he->shared_he_he.he_valu.hent_refcount; | |
29404ae0 NC |
2397 | return; |
2398 | } | |
29404ae0 | 2399 | |
19692e8d NC |
2400 | hash = HEK_HASH(hek); |
2401 | } else if (len < 0) { | |
2402 | STRLEN tmplen = -len; | |
2403 | is_utf8 = TRUE; | |
2404 | /* See the note in hv_fetch(). --jhi */ | |
2405 | str = (char*)bytes_from_utf8((U8*)str, &tmplen, &is_utf8); | |
2406 | len = tmplen; | |
2407 | if (is_utf8) | |
2408 | k_flags = HVhek_UTF8; | |
2409 | if (str != save) | |
2410 | k_flags |= HVhek_WASUTF8 | HVhek_FREEKEY; | |
c3654f1a | 2411 | } |
1c846c1f | 2412 | |
de616631 | 2413 | /* what follows was the moral equivalent of: |
6b88bc9c | 2414 | if ((Svp = hv_fetch(PL_strtab, tmpsv, FALSE, hash))) { |
a0714e2c | 2415 | if (--*Svp == NULL) |
6b88bc9c | 2416 | hv_delete(PL_strtab, str, len, G_DISCARD, hash); |
bbce6d69 | 2417 | } */ |
cbec9347 | 2418 | xhv = (XPVHV*)SvANY(PL_strtab); |
fde52b5c | 2419 | /* assert(xhv_array != 0) */ |
45d1cc86 | 2420 | first = oentry = &(HvARRAY(PL_strtab))[hash & (I32) HvMAX(PL_strtab)]; |
6c1b96a1 NC |
2421 | if (he) { |
2422 | const HE *const he_he = &(he->shared_he_he); | |
45d1cc86 | 2423 | for (entry = *oentry; entry; oentry = &HeNEXT(entry), entry = *oentry) { |
35ab5632 NC |
2424 | if (entry == he_he) |
2425 | break; | |
19692e8d NC |
2426 | } |
2427 | } else { | |
35a4481c | 2428 | const int flags_masked = k_flags & HVhek_MASK; |
45d1cc86 | 2429 | for (entry = *oentry; entry; oentry = &HeNEXT(entry), entry = *oentry) { |
19692e8d NC |
2430 | if (HeHASH(entry) != hash) /* strings can't be equal */ |
2431 | continue; | |
2432 | if (HeKLEN(entry) != len) | |
2433 | continue; | |
2434 | if (HeKEY(entry) != str && memNE(HeKEY(entry),str,len)) /* is this it? */ | |
2435 | continue; | |
2436 | if (HeKFLAGS(entry) != flags_masked) | |
2437 | continue; | |
19692e8d NC |
2438 | break; |
2439 | } | |
2440 | } | |
2441 | ||
35ab5632 NC |
2442 | if (entry) { |
2443 | if (--entry->he_valu.hent_refcount == 0) { | |
19692e8d | 2444 | *oentry = HeNEXT(entry); |
cbae3960 | 2445 | Safefree(entry); |
4c7185a0 | 2446 | xhv->xhv_keys--; /* HvTOTALKEYS(hv)-- */ |
19692e8d | 2447 | } |
fde52b5c | 2448 | } |
19692e8d | 2449 | |
9b387841 NC |
2450 | if (!entry) |
2451 | Perl_ck_warner_d(aTHX_ packWARN(WARN_INTERNAL), | |
2452 | "Attempt to free non-existent shared string '%s'%s" | |
2453 | pTHX__FORMAT, | |
2454 | hek ? HEK_KEY(hek) : str, | |
2455 | ((k_flags & HVhek_UTF8) ? " (utf8)" : "") pTHX__VALUE); | |
19692e8d NC |
2456 | if (k_flags & HVhek_FREEKEY) |
2457 | Safefree(str); | |
fde52b5c | 2458 | } |
2459 | ||
bbce6d69 | 2460 | /* get a (constant) string ptr from the global string table |
2461 | * string will get added if it is not already there. | |
fde52b5c | 2462 | * len and hash must both be valid for str. |
2463 | */ | |
bbce6d69 | 2464 | HEK * |
864dbfa3 | 2465 | Perl_share_hek(pTHX_ const char *str, I32 len, register U32 hash) |
fde52b5c | 2466 | { |
da58a35d | 2467 | bool is_utf8 = FALSE; |
19692e8d | 2468 | int flags = 0; |
aec46f14 | 2469 | const char * const save = str; |
da58a35d | 2470 | |
7918f24d NC |
2471 | PERL_ARGS_ASSERT_SHARE_HEK; |
2472 | ||
da58a35d | 2473 | if (len < 0) { |
77caf834 | 2474 | STRLEN tmplen = -len; |
da58a35d | 2475 | is_utf8 = TRUE; |
77caf834 JH |
2476 | /* See the note in hv_fetch(). --jhi */ |
2477 | str = (char*)bytes_from_utf8((U8*)str, &tmplen, &is_utf8); | |
2478 | len = tmplen; | |
19692e8d NC |
2479 | /* If we were able to downgrade here, then than means that we were passed |
2480 | in a key which only had chars 0-255, but was utf8 encoded. */ | |
2481 | if (is_utf8) | |
2482 | flags = HVhek_UTF8; | |
2483 | /* If we found we were able to downgrade the string to bytes, then | |
2484 | we should flag that it needs upgrading on keys or each. Also flag | |
2485 | that we need share_hek_flags to free the string. */ | |
2486 | if (str != save) | |
2487 | flags |= HVhek_WASUTF8 | HVhek_FREEKEY; | |
2488 | } | |
2489 | ||
6e838c70 | 2490 | return share_hek_flags (str, len, hash, flags); |
19692e8d NC |
2491 | } |
2492 | ||
6e838c70 | 2493 | STATIC HEK * |
19692e8d NC |
2494 | S_share_hek_flags(pTHX_ const char *str, I32 len, register U32 hash, int flags) |
2495 | { | |
97aff369 | 2496 | dVAR; |
19692e8d | 2497 | register HE *entry; |
35a4481c | 2498 | const int flags_masked = flags & HVhek_MASK; |
263cb4a6 | 2499 | const U32 hindex = hash & (I32) HvMAX(PL_strtab); |
7918f24d NC |
2500 | register XPVHV * const xhv = (XPVHV*)SvANY(PL_strtab); |
2501 | ||
2502 | PERL_ARGS_ASSERT_SHARE_HEK_FLAGS; | |
bbce6d69 | 2503 | |
fde52b5c | 2504 | /* what follows is the moral equivalent of: |
1c846c1f | 2505 | |
6b88bc9c | 2506 | if (!(Svp = hv_fetch(PL_strtab, str, len, FALSE))) |
a0714e2c | 2507 | hv_store(PL_strtab, str, len, NULL, hash); |
fdcd69b6 NC |
2508 | |
2509 | Can't rehash the shared string table, so not sure if it's worth | |
2510 | counting the number of entries in the linked list | |
bbce6d69 | 2511 | */ |
7918f24d | 2512 | |
fde52b5c | 2513 | /* assert(xhv_array != 0) */ |
263cb4a6 NC |
2514 | entry = (HvARRAY(PL_strtab))[hindex]; |
2515 | for (;entry; entry = HeNEXT(entry)) { | |
fde52b5c | 2516 | if (HeHASH(entry) != hash) /* strings can't be equal */ |
2517 | continue; | |
2518 | if (HeKLEN(entry) != len) | |
2519 | continue; | |
1c846c1f | 2520 | if (HeKEY(entry) != str && memNE(HeKEY(entry),str,len)) /* is this it? */ |
fde52b5c | 2521 | continue; |
19692e8d | 2522 | if (HeKFLAGS(entry) != flags_masked) |
c3654f1a | 2523 | continue; |
fde52b5c | 2524 | break; |
2525 | } | |
263cb4a6 NC |
2526 | |
2527 | if (!entry) { | |
45d1cc86 NC |
2528 | /* What used to be head of the list. |
2529 | If this is NULL, then we're the first entry for this slot, which | |
2530 | means we need to increate fill. */ | |
cbae3960 NC |
2531 | struct shared_he *new_entry; |
2532 | HEK *hek; | |
2533 | char *k; | |
263cb4a6 NC |
2534 | HE **const head = &HvARRAY(PL_strtab)[hindex]; |
2535 | HE *const next = *head; | |
cbae3960 NC |
2536 | |
2537 | /* We don't actually store a HE from the arena and a regular HEK. | |
2538 | Instead we allocate one chunk of memory big enough for both, | |
2539 | and put the HEK straight after the HE. This way we can find the | |
2540 | HEK directly from the HE. | |
2541 | */ | |
2542 | ||
a02a5408 | 2543 | Newx(k, STRUCT_OFFSET(struct shared_he, |
cbae3960 NC |
2544 | shared_he_hek.hek_key[0]) + len + 2, char); |
2545 | new_entry = (struct shared_he *)k; | |
2546 | entry = &(new_entry->shared_he_he); | |
2547 | hek = &(new_entry->shared_he_hek); | |
2548 | ||
2549 | Copy(str, HEK_KEY(hek), len, char); | |
2550 | HEK_KEY(hek)[len] = 0; | |
2551 | HEK_LEN(hek) = len; | |
2552 | HEK_HASH(hek) = hash; | |
2553 | HEK_FLAGS(hek) = (unsigned char)flags_masked; | |
2554 | ||
2555 | /* Still "point" to the HEK, so that other code need not know what | |
2556 | we're up to. */ | |
2557 | HeKEY_hek(entry) = hek; | |
de616631 | 2558 | entry->he_valu.hent_refcount = 0; |
263cb4a6 NC |
2559 | HeNEXT(entry) = next; |
2560 | *head = entry; | |
cbae3960 | 2561 | |
4c7185a0 | 2562 | xhv->xhv_keys++; /* HvTOTALKEYS(hv)++ */ |
263cb4a6 | 2563 | if (!next) { /* initial entry? */ |
5ac36297 | 2564 | } else if (xhv->xhv_keys > xhv->xhv_max /* HvKEYS(hv) > HvMAX(hv) */) { |
cbec9347 | 2565 | hsplit(PL_strtab); |
bbce6d69 | 2566 | } |
2567 | } | |
2568 | ||
de616631 | 2569 | ++entry->he_valu.hent_refcount; |
19692e8d NC |
2570 | |
2571 | if (flags & HVhek_FREEKEY) | |
f9a63242 | 2572 | Safefree(str); |
19692e8d | 2573 | |
6e838c70 | 2574 | return HeKEY_hek(entry); |
fde52b5c | 2575 | } |
ecae49c0 | 2576 | |
ca732855 NC |
2577 | I32 * |
2578 | Perl_hv_placeholders_p(pTHX_ HV *hv) | |
2579 | { | |
2580 | dVAR; | |
ad64d0ec | 2581 | MAGIC *mg = mg_find((const SV *)hv, PERL_MAGIC_rhash); |
ca732855 | 2582 | |
7918f24d NC |
2583 | PERL_ARGS_ASSERT_HV_PLACEHOLDERS_P; |
2584 | ||
ca732855 | 2585 | if (!mg) { |
ad64d0ec | 2586 | mg = sv_magicext(MUTABLE_SV(hv), 0, PERL_MAGIC_rhash, 0, 0, 0); |
ca732855 NC |
2587 | |
2588 | if (!mg) { | |
2589 | Perl_die(aTHX_ "panic: hv_placeholders_p"); | |
2590 | } | |
2591 | } | |
2592 | return &(mg->mg_len); | |
2593 | } | |
2594 | ||
2595 | ||
2596 | I32 | |
0c289d13 | 2597 | Perl_hv_placeholders_get(pTHX_ const HV *hv) |
ca732855 NC |
2598 | { |
2599 | dVAR; | |
0c289d13 | 2600 | MAGIC * const mg = mg_find((const SV *)hv, PERL_MAGIC_rhash); |
ca732855 | 2601 | |
7918f24d NC |
2602 | PERL_ARGS_ASSERT_HV_PLACEHOLDERS_GET; |
2603 | ||
ca732855 NC |
2604 | return mg ? mg->mg_len : 0; |
2605 | } | |
2606 | ||
2607 | void | |
ac1e784a | 2608 | Perl_hv_placeholders_set(pTHX_ HV *hv, I32 ph) |
ca732855 NC |
2609 | { |
2610 | dVAR; | |
ad64d0ec | 2611 | MAGIC * const mg = mg_find((const SV *)hv, PERL_MAGIC_rhash); |
ca732855 | 2612 | |
7918f24d NC |
2613 | PERL_ARGS_ASSERT_HV_PLACEHOLDERS_SET; |
2614 | ||
ca732855 NC |
2615 | if (mg) { |
2616 | mg->mg_len = ph; | |
2617 | } else if (ph) { | |
ad64d0ec | 2618 | if (!sv_magicext(MUTABLE_SV(hv), 0, PERL_MAGIC_rhash, 0, 0, ph)) |
ca732855 NC |
2619 | Perl_die(aTHX_ "panic: hv_placeholders_set"); |
2620 | } | |
2621 | /* else we don't need to add magic to record 0 placeholders. */ | |
2622 | } | |
ecae49c0 | 2623 | |
2a49f0f5 | 2624 | STATIC SV * |
7b0bddfa NC |
2625 | S_refcounted_he_value(pTHX_ const struct refcounted_he *he) |
2626 | { | |
0b2d3faa | 2627 | dVAR; |
7b0bddfa | 2628 | SV *value; |
7918f24d NC |
2629 | |
2630 | PERL_ARGS_ASSERT_REFCOUNTED_HE_VALUE; | |
2631 | ||
7b0bddfa NC |
2632 | switch(he->refcounted_he_data[0] & HVrhek_typemask) { |
2633 | case HVrhek_undef: | |
2634 | value = newSV(0); | |
2635 | break; | |
2636 | case HVrhek_delete: | |
2637 | value = &PL_sv_placeholder; | |
2638 | break; | |
2639 | case HVrhek_IV: | |
44ebaf21 NC |
2640 | value = newSViv(he->refcounted_he_val.refcounted_he_u_iv); |
2641 | break; | |
2642 | case HVrhek_UV: | |
2643 | value = newSVuv(he->refcounted_he_val.refcounted_he_u_uv); | |
7b0bddfa NC |
2644 | break; |
2645 | case HVrhek_PV: | |
44ebaf21 | 2646 | case HVrhek_PV_UTF8: |
7b0bddfa NC |
2647 | /* Create a string SV that directly points to the bytes in our |
2648 | structure. */ | |
b9f83d2f | 2649 | value = newSV_type(SVt_PV); |
7b0bddfa NC |
2650 | SvPV_set(value, (char *) he->refcounted_he_data + 1); |
2651 | SvCUR_set(value, he->refcounted_he_val.refcounted_he_u_len); | |
2652 | /* This stops anything trying to free it */ | |
2653 | SvLEN_set(value, 0); | |
2654 | SvPOK_on(value); | |
2655 | SvREADONLY_on(value); | |
44ebaf21 | 2656 | if ((he->refcounted_he_data[0] & HVrhek_typemask) == HVrhek_PV_UTF8) |
7b0bddfa NC |
2657 | SvUTF8_on(value); |
2658 | break; | |
2659 | default: | |
2660 | Perl_croak(aTHX_ "panic: refcounted_he_value bad flags %x", | |
2661 | he->refcounted_he_data[0]); | |
2662 | } | |
2663 | return value; | |
2664 | } | |
2665 | ||
ecae49c0 | 2666 | /* |
b3ca2e83 NC |
2667 | =for apidoc refcounted_he_chain_2hv |
2668 | ||
abc25d8c | 2669 | Generates and returns a C<HV *> by walking up the tree starting at the passed |
b3ca2e83 NC |
2670 | in C<struct refcounted_he *>. |
2671 | ||
2672 | =cut | |
2673 | */ | |
2674 | HV * | |
2675 | Perl_refcounted_he_chain_2hv(pTHX_ const struct refcounted_he *chain) | |
2676 | { | |
7a89be66 | 2677 | dVAR; |
b3ca2e83 NC |
2678 | HV *hv = newHV(); |
2679 | U32 placeholders = 0; | |
2680 | /* We could chase the chain once to get an idea of the number of keys, | |
2681 | and call ksplit. But for now we'll make a potentially inefficient | |
2682 | hash with only 8 entries in its array. */ | |
2683 | const U32 max = HvMAX(hv); | |
2684 | ||
2685 | if (!HvARRAY(hv)) { | |
2686 | char *array; | |
2687 | Newxz(array, PERL_HV_ARRAY_ALLOC_BYTES(max + 1), char); | |
2688 | HvARRAY(hv) = (HE**)array; | |
2689 | } | |
2690 | ||
2691 | while (chain) { | |
cbb1fbea | 2692 | #ifdef USE_ITHREADS |
b6bbf3fa | 2693 | U32 hash = chain->refcounted_he_hash; |
cbb1fbea NC |
2694 | #else |
2695 | U32 hash = HEK_HASH(chain->refcounted_he_hek); | |
2696 | #endif | |
b3ca2e83 NC |
2697 | HE **oentry = &((HvARRAY(hv))[hash & max]); |
2698 | HE *entry = *oentry; | |
b6bbf3fa | 2699 | SV *value; |
cbb1fbea | 2700 | |
b3ca2e83 NC |
2701 | for (; entry; entry = HeNEXT(entry)) { |
2702 | if (HeHASH(entry) == hash) { | |
9f769845 NC |
2703 | /* We might have a duplicate key here. If so, entry is older |
2704 | than the key we've already put in the hash, so if they are | |
2705 | the same, skip adding entry. */ | |
2706 | #ifdef USE_ITHREADS | |
2707 | const STRLEN klen = HeKLEN(entry); | |
2708 | const char *const key = HeKEY(entry); | |
2709 | if (klen == chain->refcounted_he_keylen | |
2710 | && (!!HeKUTF8(entry) | |
2711 | == !!(chain->refcounted_he_data[0] & HVhek_UTF8)) | |
2712 | && memEQ(key, REF_HE_KEY(chain), klen)) | |
2713 | goto next_please; | |
2714 | #else | |
2715 | if (HeKEY_hek(entry) == chain->refcounted_he_hek) | |
2716 | goto next_please; | |
2717 | if (HeKLEN(entry) == HEK_LEN(chain->refcounted_he_hek) | |
2718 | && HeKUTF8(entry) == HEK_UTF8(chain->refcounted_he_hek) | |
2719 | && memEQ(HeKEY(entry), HEK_KEY(chain->refcounted_he_hek), | |
2720 | HeKLEN(entry))) | |
2721 | goto next_please; | |
2722 | #endif | |
b3ca2e83 NC |
2723 | } |
2724 | } | |
2725 | assert (!entry); | |
2726 | entry = new_HE(); | |
2727 | ||
cbb1fbea NC |
2728 | #ifdef USE_ITHREADS |
2729 | HeKEY_hek(entry) | |
7b0bddfa | 2730 | = share_hek_flags(REF_HE_KEY(chain), |
b6bbf3fa NC |
2731 | chain->refcounted_he_keylen, |
2732 | chain->refcounted_he_hash, | |
2733 | (chain->refcounted_he_data[0] | |
2734 | & (HVhek_UTF8|HVhek_WASUTF8))); | |
cbb1fbea | 2735 | #else |
71ad1b0c | 2736 | HeKEY_hek(entry) = share_hek_hek(chain->refcounted_he_hek); |
cbb1fbea | 2737 | #endif |
7b0bddfa NC |
2738 | value = refcounted_he_value(chain); |
2739 | if (value == &PL_sv_placeholder) | |
b3ca2e83 | 2740 | placeholders++; |
b6bbf3fa | 2741 | HeVAL(entry) = value; |
b3ca2e83 NC |
2742 | |
2743 | /* Link it into the chain. */ | |
2744 | HeNEXT(entry) = *oentry; | |
b3ca2e83 NC |
2745 | *oentry = entry; |
2746 | ||
2747 | HvTOTALKEYS(hv)++; | |
2748 | ||
2749 | next_please: | |
71ad1b0c | 2750 | chain = chain->refcounted_he_next; |
b3ca2e83 NC |
2751 | } |
2752 | ||
2753 | if (placeholders) { | |
2754 | clear_placeholders(hv, placeholders); | |
2755 | HvTOTALKEYS(hv) -= placeholders; | |
2756 | } | |
2757 | ||
2758 | /* We could check in the loop to see if we encounter any keys with key | |
2759 | flags, but it's probably not worth it, as this per-hash flag is only | |
2760 | really meant as an optimisation for things like Storable. */ | |
2761 | HvHASKFLAGS_on(hv); | |
def9038f | 2762 | DEBUG_A(Perl_hv_assert(aTHX_ hv)); |
b3ca2e83 NC |
2763 | |
2764 | return hv; | |
2765 | } | |
2766 | ||
7b0bddfa NC |
2767 | SV * |
2768 | Perl_refcounted_he_fetch(pTHX_ const struct refcounted_he *chain, SV *keysv, | |
2769 | const char *key, STRLEN klen, int flags, U32 hash) | |
2770 | { | |
0b2d3faa | 2771 | dVAR; |
7b0bddfa NC |
2772 | /* Just to be awkward, if you're using this interface the UTF-8-or-not-ness |
2773 | of your key has to exactly match that which is stored. */ | |
2774 | SV *value = &PL_sv_placeholder; | |
7b0bddfa | 2775 | |
cd1d2f8a NC |
2776 | if (chain) { |
2777 | /* No point in doing any of this if there's nothing to find. */ | |
2778 | bool is_utf8; | |
7b0bddfa | 2779 | |
cd1d2f8a NC |
2780 | if (keysv) { |
2781 | if (flags & HVhek_FREEKEY) | |
2782 | Safefree(key); | |
2783 | key = SvPV_const(keysv, klen); | |
2784 | flags = 0; | |
2785 | is_utf8 = (SvUTF8(keysv) != 0); | |
2786 | } else { | |
2787 | is_utf8 = ((flags & HVhek_UTF8) ? TRUE : FALSE); | |
2788 | } | |
2789 | ||
2790 | if (!hash) { | |
2791 | if (keysv && (SvIsCOW_shared_hash(keysv))) { | |
2792 | hash = SvSHARED_HASH(keysv); | |
2793 | } else { | |
2794 | PERL_HASH(hash, key, klen); | |
2795 | } | |
2796 | } | |
7b0bddfa | 2797 | |
cd1d2f8a | 2798 | for (; chain; chain = chain->refcounted_he_next) { |
7b0bddfa | 2799 | #ifdef USE_ITHREADS |
cd1d2f8a NC |
2800 | if (hash != chain->refcounted_he_hash) |
2801 | continue; | |
2802 | if (klen != chain->refcounted_he_keylen) | |
2803 | continue; | |
2804 | if (memNE(REF_HE_KEY(chain),key,klen)) | |
2805 | continue; | |
2806 | if (!!is_utf8 != !!(chain->refcounted_he_data[0] & HVhek_UTF8)) | |
2807 | continue; | |
7b0bddfa | 2808 | #else |
cd1d2f8a NC |
2809 | if (hash != HEK_HASH(chain->refcounted_he_hek)) |
2810 | continue; | |
2811 | if (klen != (STRLEN)HEK_LEN(chain->refcounted_he_hek)) | |
2812 | continue; | |
2813 | if (memNE(HEK_KEY(chain->refcounted_he_hek),key,klen)) | |
2814 | continue; | |
2815 | if (!!is_utf8 != !!HEK_UTF8(chain->refcounted_he_hek)) | |
2816 | continue; | |
7b0bddfa NC |
2817 | #endif |
2818 | ||
cd1d2f8a NC |
2819 | value = sv_2mortal(refcounted_he_value(chain)); |
2820 | break; | |
2821 | } | |
7b0bddfa NC |
2822 | } |
2823 | ||
2824 | if (flags & HVhek_FREEKEY) | |
2825 | Safefree(key); | |
2826 | ||
2827 | return value; | |
2828 | } | |
2829 | ||
b3ca2e83 NC |
2830 | /* |
2831 | =for apidoc refcounted_he_new | |
2832 | ||
ec2a1de7 NC |
2833 | Creates a new C<struct refcounted_he>. As S<key> is copied, and value is |
2834 | stored in a compact form, all references remain the property of the caller. | |
2835 | The C<struct refcounted_he> is returned with a reference count of 1. | |
b3ca2e83 NC |
2836 | |
2837 | =cut | |
2838 | */ | |
2839 | ||
2840 | struct refcounted_he * | |
2841 | Perl_refcounted_he_new(pTHX_ struct refcounted_he *const parent, | |
2842 | SV *const key, SV *const value) { | |
7a89be66 | 2843 | dVAR; |
b6bbf3fa NC |
2844 | STRLEN key_len; |
2845 | const char *key_p = SvPV_const(key, key_len); | |
2846 | STRLEN value_len = 0; | |
95b63a38 | 2847 | const char *value_p = NULL; |
b6bbf3fa NC |
2848 | char value_type; |
2849 | char flags; | |
d8c5b3c5 | 2850 | bool is_utf8 = SvUTF8(key) ? TRUE : FALSE; |
b6bbf3fa NC |
2851 | |
2852 | if (SvPOK(value)) { | |
2853 | value_type = HVrhek_PV; | |
2854 | } else if (SvIOK(value)) { | |
ad64d0ec | 2855 | value_type = SvUOK((const SV *)value) ? HVrhek_UV : HVrhek_IV; |
b6bbf3fa NC |
2856 | } else if (value == &PL_sv_placeholder) { |
2857 | value_type = HVrhek_delete; | |
2858 | } else if (!SvOK(value)) { | |
2859 | value_type = HVrhek_undef; | |
2860 | } else { | |
2861 | value_type = HVrhek_PV; | |
2862 | } | |
b3ca2e83 | 2863 | |
b6bbf3fa | 2864 | if (value_type == HVrhek_PV) { |
012da8e5 NC |
2865 | /* Do it this way so that the SvUTF8() test is after the SvPV, in case |
2866 | the value is overloaded, and doesn't yet have the UTF-8flag set. */ | |
b6bbf3fa | 2867 | value_p = SvPV_const(value, value_len); |
012da8e5 NC |
2868 | if (SvUTF8(value)) |
2869 | value_type = HVrhek_PV_UTF8; | |
b6bbf3fa | 2870 | } |
012da8e5 NC |
2871 | flags = value_type; |
2872 | ||
2873 | if (is_utf8) { | |
2874 | /* Hash keys are always stored normalised to (yes) ISO-8859-1. | |
2875 | As we're going to be building hash keys from this value in future, | |
2876 | normalise it now. */ | |
2877 | key_p = (char*)bytes_from_utf8((const U8*)key_p, &key_len, &is_utf8); | |
2878 | flags |= is_utf8 ? HVhek_UTF8 : HVhek_WASUTF8; | |
2879 | } | |
2880 | ||
2881 | return refcounted_he_new_common(parent, key_p, key_len, flags, value_type, | |
2882 | ((value_type == HVrhek_PV | |
2883 | || value_type == HVrhek_PV_UTF8) ? | |
2884 | (void *)value_p : (void *)value), | |
2885 | value_len); | |
2886 | } | |
2887 | ||
1da83c39 | 2888 | static struct refcounted_he * |
012da8e5 NC |
2889 | S_refcounted_he_new_common(pTHX_ struct refcounted_he *const parent, |
2890 | const char *const key_p, const STRLEN key_len, | |
2891 | const char flags, char value_type, | |
2892 | const void *value, const STRLEN value_len) { | |
2893 | dVAR; | |
2894 | struct refcounted_he *he; | |
2895 | U32 hash; | |
2896 | const bool is_pv = value_type == HVrhek_PV || value_type == HVrhek_PV_UTF8; | |
2897 | STRLEN key_offset = is_pv ? value_len + 2 : 1; | |
2898 | ||
2899 | PERL_ARGS_ASSERT_REFCOUNTED_HE_NEW_COMMON; | |
b6bbf3fa | 2900 | |
b6bbf3fa | 2901 | #ifdef USE_ITHREADS |
10edeb5d JH |
2902 | he = (struct refcounted_he*) |
2903 | PerlMemShared_malloc(sizeof(struct refcounted_he) - 1 | |
2904 | + key_len | |
2905 | + key_offset); | |
6cef672b | 2906 | #else |
10edeb5d JH |
2907 | he = (struct refcounted_he*) |
2908 | PerlMemShared_malloc(sizeof(struct refcounted_he) - 1 | |
2909 | + key_offset); | |
6cef672b | 2910 | #endif |
b3ca2e83 | 2911 | |
71ad1b0c | 2912 | he->refcounted_he_next = parent; |
b6bbf3fa | 2913 | |
012da8e5 NC |
2914 | if (is_pv) { |
2915 | Copy((char *)value, he->refcounted_he_data + 1, value_len + 1, char); | |
b6bbf3fa | 2916 | he->refcounted_he_val.refcounted_he_u_len = value_len; |
b6bbf3fa | 2917 | } else if (value_type == HVrhek_IV) { |
ad64d0ec | 2918 | he->refcounted_he_val.refcounted_he_u_iv = SvIVX((const SV *)value); |
012da8e5 | 2919 | } else if (value_type == HVrhek_UV) { |
ad64d0ec | 2920 | he->refcounted_he_val.refcounted_he_u_uv = SvUVX((const SV *)value); |
b6bbf3fa NC |
2921 | } |
2922 | ||
b6bbf3fa NC |
2923 | PERL_HASH(hash, key_p, key_len); |
2924 | ||
cbb1fbea | 2925 | #ifdef USE_ITHREADS |
b6bbf3fa NC |
2926 | he->refcounted_he_hash = hash; |
2927 | he->refcounted_he_keylen = key_len; | |
2928 | Copy(key_p, he->refcounted_he_data + key_offset, key_len, char); | |
cbb1fbea | 2929 | #else |
b6bbf3fa | 2930 | he->refcounted_he_hek = share_hek_flags(key_p, key_len, hash, flags); |
cbb1fbea | 2931 | #endif |
b6bbf3fa NC |
2932 | |
2933 | if (flags & HVhek_WASUTF8) { | |
2934 | /* If it was downgraded from UTF-8, then the pointer returned from | |
2935 | bytes_from_utf8 is an allocated pointer that we must free. */ | |
2936 | Safefree(key_p); | |
2937 | } | |
2938 | ||
2939 | he->refcounted_he_data[0] = flags; | |
b3ca2e83 NC |
2940 | he->refcounted_he_refcnt = 1; |
2941 | ||
2942 | return he; | |
2943 | } | |
2944 | ||
2945 | /* | |
2946 | =for apidoc refcounted_he_free | |
2947 | ||
2948 | Decrements the reference count of the passed in C<struct refcounted_he *> | |
2949 | by one. If the reference count reaches zero the structure's memory is freed, | |
2950 | and C<refcounted_he_free> iterates onto the parent node. | |
2951 | ||
2952 | =cut | |
2953 | */ | |
2954 | ||
2955 | void | |
2956 | Perl_refcounted_he_free(pTHX_ struct refcounted_he *he) { | |
53d44271 | 2957 | dVAR; |
57ca3b03 AL |
2958 | PERL_UNUSED_CONTEXT; |
2959 | ||
b3ca2e83 NC |
2960 | while (he) { |
2961 | struct refcounted_he *copy; | |
cbb1fbea | 2962 | U32 new_count; |
b3ca2e83 | 2963 | |
cbb1fbea NC |
2964 | HINTS_REFCNT_LOCK; |
2965 | new_count = --he->refcounted_he_refcnt; | |
2966 | HINTS_REFCNT_UNLOCK; | |
2967 | ||
2968 | if (new_count) { | |
b3ca2e83 | 2969 | return; |
cbb1fbea | 2970 | } |
b3ca2e83 | 2971 | |
b6bbf3fa | 2972 | #ifndef USE_ITHREADS |
71ad1b0c | 2973 | unshare_hek_or_pvn (he->refcounted_he_hek, 0, 0, 0); |
cbb1fbea | 2974 | #endif |
b3ca2e83 | 2975 | copy = he; |
71ad1b0c | 2976 | he = he->refcounted_he_next; |
b6bbf3fa | 2977 | PerlMemShared_free(copy); |
b3ca2e83 NC |
2978 | } |
2979 | } | |
2980 | ||
47550813 NC |
2981 | /* pp_entereval is aware that labels are stored with a key ':' at the top of |
2982 | the linked list. */ | |
dca6062a NC |
2983 | const char * |
2984 | Perl_fetch_cop_label(pTHX_ struct refcounted_he *const chain, STRLEN *len, | |
2985 | U32 *flags) { | |
2986 | if (!chain) | |
2987 | return NULL; | |
2988 | #ifdef USE_ITHREADS | |
2989 | if (chain->refcounted_he_keylen != 1) | |
2990 | return NULL; | |
2991 | if (*REF_HE_KEY(chain) != ':') | |
2992 | return NULL; | |
2993 | #else | |
2994 | if ((STRLEN)HEK_LEN(chain->refcounted_he_hek) != 1) | |
2995 | return NULL; | |
2996 | if (*HEK_KEY(chain->refcounted_he_hek) != ':') | |
2997 | return NULL; | |
2998 | #endif | |
012da8e5 NC |
2999 | /* Stop anyone trying to really mess us up by adding their own value for |
3000 | ':' into %^H */ | |
3001 | if ((chain->refcounted_he_data[0] & HVrhek_typemask) != HVrhek_PV | |
3002 | && (chain->refcounted_he_data[0] & HVrhek_typemask) != HVrhek_PV_UTF8) | |
3003 | return NULL; | |
3004 | ||
dca6062a NC |
3005 | if (len) |
3006 | *len = chain->refcounted_he_val.refcounted_he_u_len; | |
3007 | if (flags) { | |
3008 | *flags = ((chain->refcounted_he_data[0] & HVrhek_typemask) | |
3009 | == HVrhek_PV_UTF8) ? SVf_UTF8 : 0; | |
3010 | } | |
3011 | return chain->refcounted_he_data + 1; | |
3012 | } | |
3013 | ||
012da8e5 NC |
3014 | /* As newSTATEOP currently gets passed plain char* labels, we will only provide |
3015 | that interface. Once it works out how to pass in length and UTF-8 ness, this | |
3016 | function will need superseding. */ | |
3017 | struct refcounted_he * | |
3018 | Perl_store_cop_label(pTHX_ struct refcounted_he *const chain, const char *label) | |
3019 | { | |
547bb267 NC |
3020 | PERL_ARGS_ASSERT_STORE_COP_LABEL; |
3021 | ||
012da8e5 NC |
3022 | return refcounted_he_new_common(chain, ":", 1, HVrhek_PV, HVrhek_PV, |
3023 | label, strlen(label)); | |
3024 | } | |
3025 | ||
b3ca2e83 | 3026 | /* |
ecae49c0 NC |
3027 | =for apidoc hv_assert |
3028 | ||
3029 | Check that a hash is in an internally consistent state. | |
3030 | ||
3031 | =cut | |
3032 | */ | |
3033 | ||
943795c2 NC |
3034 | #ifdef DEBUGGING |
3035 | ||
ecae49c0 NC |
3036 | void |
3037 | Perl_hv_assert(pTHX_ HV *hv) | |
3038 | { | |
57ca3b03 AL |
3039 | dVAR; |
3040 | HE* entry; | |
3041 | int withflags = 0; | |
3042 | int placeholders = 0; | |
3043 | int real = 0; | |
3044 | int bad = 0; | |
3045 | const I32 riter = HvRITER_get(hv); | |
3046 | HE *eiter = HvEITER_get(hv); | |
3047 | ||
7918f24d NC |
3048 | PERL_ARGS_ASSERT_HV_ASSERT; |
3049 | ||
57ca3b03 AL |
3050 | (void)hv_iterinit(hv); |
3051 | ||
3052 | while ((entry = hv_iternext_flags(hv, HV_ITERNEXT_WANTPLACEHOLDERS))) { | |
3053 | /* sanity check the values */ | |
3054 | if (HeVAL(entry) == &PL_sv_placeholder) | |
3055 | placeholders++; | |
3056 | else | |
3057 | real++; | |
3058 | /* sanity check the keys */ | |
3059 | if (HeSVKEY(entry)) { | |
6f207bd3 | 3060 | NOOP; /* Don't know what to check on SV keys. */ |
57ca3b03 AL |
3061 | } else if (HeKUTF8(entry)) { |
3062 | withflags++; | |
3063 | if (HeKWASUTF8(entry)) { | |
3064 | PerlIO_printf(Perl_debug_log, | |
d2a455e7 | 3065 | "hash key has both WASUTF8 and UTF8: '%.*s'\n", |
57ca3b03 AL |
3066 | (int) HeKLEN(entry), HeKEY(entry)); |
3067 | bad = 1; | |
3068 | } | |
3069 | } else if (HeKWASUTF8(entry)) | |
3070 | withflags++; | |
3071 | } | |
ad64d0ec | 3072 | if (!SvTIED_mg((const SV *)hv, PERL_MAGIC_tied)) { |
57ca3b03 AL |
3073 | static const char bad_count[] = "Count %d %s(s), but hash reports %d\n"; |
3074 | const int nhashkeys = HvUSEDKEYS(hv); | |
3075 | const int nhashplaceholders = HvPLACEHOLDERS_get(hv); | |
3076 | ||
3077 | if (nhashkeys != real) { | |
3078 | PerlIO_printf(Perl_debug_log, bad_count, real, "keys", nhashkeys ); | |
3079 | bad = 1; | |
3080 | } | |
3081 | if (nhashplaceholders != placeholders) { | |
3082 | PerlIO_printf(Perl_debug_log, bad_count, placeholders, "placeholder", nhashplaceholders ); | |
3083 | bad = 1; | |
3084 | } | |
3085 | } | |
3086 | if (withflags && ! HvHASKFLAGS(hv)) { | |
3087 | PerlIO_printf(Perl_debug_log, | |
3088 | "Hash has HASKFLAGS off but I count %d key(s) with flags\n", | |
3089 | withflags); | |
3090 | bad = 1; | |
3091 | } | |
3092 | if (bad) { | |
ad64d0ec | 3093 | sv_dump(MUTABLE_SV(hv)); |
57ca3b03 AL |
3094 | } |
3095 | HvRITER_set(hv, riter); /* Restore hash iterator state */ | |
3096 | HvEITER_set(hv, eiter); | |
ecae49c0 | 3097 | } |
af3babe4 | 3098 | |
943795c2 NC |
3099 | #endif |
3100 | ||
af3babe4 NC |
3101 | /* |
3102 | * Local variables: | |
3103 | * c-indentation-style: bsd | |
3104 | * c-basic-offset: 4 | |
3105 | * indent-tabs-mode: t | |
3106 | * End: | |
3107 | * | |
37442d52 RGS |
3108 | * ex: set ts=8 sts=4 sw=4 noet: |
3109 | */ |