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 | { | |
658b4a4a | 137 | HEK *shared = (HEK*)ptr_table_fetch(PL_ptr_table, source); |
9d4ba2ae | 138 | |
7918f24d | 139 | PERL_ARGS_ASSERT_HEK_DUP; |
9d4ba2ae | 140 | PERL_UNUSED_ARG(param); |
0bff533c NC |
141 | |
142 | if (shared) { | |
143 | /* We already shared this hash key. */ | |
454f1e26 | 144 | (void)share_hek_hek(shared); |
0bff533c NC |
145 | } |
146 | else { | |
658b4a4a | 147 | shared |
6e838c70 NC |
148 | = share_hek_flags(HEK_KEY(source), HEK_LEN(source), |
149 | HEK_HASH(source), HEK_FLAGS(source)); | |
658b4a4a | 150 | ptr_table_store(PL_ptr_table, source, shared); |
0bff533c | 151 | } |
658b4a4a | 152 | return shared; |
0bff533c NC |
153 | } |
154 | ||
d18c6117 | 155 | HE * |
5c4138a0 | 156 | Perl_he_dup(pTHX_ const HE *e, bool shared, CLONE_PARAMS* param) |
d18c6117 GS |
157 | { |
158 | HE *ret; | |
159 | ||
7918f24d NC |
160 | PERL_ARGS_ASSERT_HE_DUP; |
161 | ||
d18c6117 | 162 | if (!e) |
4608196e | 163 | return NULL; |
7766f137 GS |
164 | /* look for it in the table first */ |
165 | ret = (HE*)ptr_table_fetch(PL_ptr_table, e); | |
166 | if (ret) | |
167 | return ret; | |
168 | ||
169 | /* create anew and remember what it is */ | |
d33b2eba | 170 | ret = new_HE(); |
7766f137 GS |
171 | ptr_table_store(PL_ptr_table, e, ret); |
172 | ||
d2d73c3e | 173 | HeNEXT(ret) = he_dup(HeNEXT(e),shared, param); |
dd28f7bb DM |
174 | if (HeKLEN(e) == HEf_SVKEY) { |
175 | char *k; | |
ad64d0ec | 176 | Newx(k, HEK_BASESIZE + sizeof(const SV *), char); |
dd28f7bb | 177 | HeKEY_hek(ret) = (HEK*)k; |
d2d73c3e | 178 | HeKEY_sv(ret) = SvREFCNT_inc(sv_dup(HeKEY_sv(e), param)); |
dd28f7bb | 179 | } |
c21d1a0f | 180 | else if (shared) { |
0bff533c NC |
181 | /* This is hek_dup inlined, which seems to be important for speed |
182 | reasons. */ | |
1b6737cc | 183 | HEK * const source = HeKEY_hek(e); |
658b4a4a | 184 | HEK *shared = (HEK*)ptr_table_fetch(PL_ptr_table, source); |
c21d1a0f NC |
185 | |
186 | if (shared) { | |
187 | /* We already shared this hash key. */ | |
454f1e26 | 188 | (void)share_hek_hek(shared); |
c21d1a0f NC |
189 | } |
190 | else { | |
658b4a4a | 191 | shared |
6e838c70 NC |
192 | = share_hek_flags(HEK_KEY(source), HEK_LEN(source), |
193 | HEK_HASH(source), HEK_FLAGS(source)); | |
658b4a4a | 194 | ptr_table_store(PL_ptr_table, source, shared); |
c21d1a0f | 195 | } |
658b4a4a | 196 | HeKEY_hek(ret) = shared; |
c21d1a0f | 197 | } |
d18c6117 | 198 | else |
19692e8d NC |
199 | HeKEY_hek(ret) = save_hek_flags(HeKEY(e), HeKLEN(e), HeHASH(e), |
200 | HeKFLAGS(e)); | |
d2d73c3e | 201 | HeVAL(ret) = SvREFCNT_inc(sv_dup(HeVAL(e), param)); |
d18c6117 GS |
202 | return ret; |
203 | } | |
204 | #endif /* USE_ITHREADS */ | |
205 | ||
1b1f1335 | 206 | static void |
2393f1b9 JH |
207 | S_hv_notallowed(pTHX_ int flags, const char *key, I32 klen, |
208 | const char *msg) | |
1b1f1335 | 209 | { |
1b6737cc | 210 | SV * const sv = sv_newmortal(); |
7918f24d NC |
211 | |
212 | PERL_ARGS_ASSERT_HV_NOTALLOWED; | |
213 | ||
19692e8d | 214 | if (!(flags & HVhek_FREEKEY)) { |
1b1f1335 NIS |
215 | sv_setpvn(sv, key, klen); |
216 | } | |
217 | else { | |
218 | /* Need to free saved eventually assign to mortal SV */ | |
34c3c4e3 | 219 | /* XXX is this line an error ???: SV *sv = sv_newmortal(); */ |
1b1f1335 NIS |
220 | sv_usepvn(sv, (char *) key, klen); |
221 | } | |
19692e8d | 222 | if (flags & HVhek_UTF8) { |
1b1f1335 NIS |
223 | SvUTF8_on(sv); |
224 | } | |
be2597df | 225 | Perl_croak(aTHX_ msg, SVfARG(sv)); |
1b1f1335 NIS |
226 | } |
227 | ||
fde52b5c | 228 | /* (klen == HEf_SVKEY) is special for MAGICAL hv entries, meaning key slot |
229 | * contains an SV* */ | |
230 | ||
34a6f7b4 NC |
231 | /* |
232 | =for apidoc hv_store | |
233 | ||
234 | Stores an SV in a hash. The hash key is specified as C<key> and C<klen> is | |
235 | the length of the key. The C<hash> parameter is the precomputed hash | |
236 | value; if it is zero then Perl will compute it. The return value will be | |
237 | NULL if the operation failed or if the value did not need to be actually | |
238 | stored within the hash (as in the case of tied hashes). Otherwise it can | |
239 | be dereferenced to get the original C<SV*>. Note that the caller is | |
240 | responsible for suitably incrementing the reference count of C<val> before | |
241 | the call, and decrementing it if the function returned NULL. Effectively | |
242 | a successful hv_store takes ownership of one reference to C<val>. This is | |
243 | usually what you want; a newly created SV has a reference count of one, so | |
244 | if all your code does is create SVs then store them in a hash, hv_store | |
245 | will own the only reference to the new SV, and your code doesn't need to do | |
246 | anything further to tidy up. hv_store is not implemented as a call to | |
247 | hv_store_ent, and does not create a temporary SV for the key, so if your | |
248 | key data is not already in SV form then use hv_store in preference to | |
249 | hv_store_ent. | |
250 | ||
251 | See L<perlguts/"Understanding the Magic of Tied Hashes and Arrays"> for more | |
252 | information on how to use this function on tied hashes. | |
253 | ||
34a6f7b4 NC |
254 | =for apidoc hv_store_ent |
255 | ||
256 | Stores C<val> in a hash. The hash key is specified as C<key>. The C<hash> | |
257 | parameter is the precomputed hash value; if it is zero then Perl will | |
258 | compute it. The return value is the new hash entry so created. It will be | |
259 | NULL if the operation failed or if the value did not need to be actually | |
260 | stored within the hash (as in the case of tied hashes). Otherwise the | |
261 | contents of the return value can be accessed using the C<He?> macros | |
262 | described here. Note that the caller is responsible for suitably | |
263 | incrementing the reference count of C<val> before the call, and | |
264 | decrementing it if the function returned NULL. Effectively a successful | |
265 | hv_store_ent takes ownership of one reference to C<val>. This is | |
266 | usually what you want; a newly created SV has a reference count of one, so | |
267 | if all your code does is create SVs then store them in a hash, hv_store | |
268 | will own the only reference to the new SV, and your code doesn't need to do | |
269 | anything further to tidy up. Note that hv_store_ent only reads the C<key>; | |
270 | unlike C<val> it does not take ownership of it, so maintaining the correct | |
271 | reference count on C<key> is entirely the caller's responsibility. hv_store | |
272 | is not implemented as a call to hv_store_ent, and does not create a temporary | |
273 | SV for the key, so if your key data is not already in SV form then use | |
274 | hv_store in preference to hv_store_ent. | |
275 | ||
276 | See L<perlguts/"Understanding the Magic of Tied Hashes and Arrays"> for more | |
277 | information on how to use this function on tied hashes. | |
278 | ||
34a6f7b4 NC |
279 | =for apidoc hv_exists |
280 | ||
281 | Returns a boolean indicating whether the specified hash key exists. The | |
282 | C<klen> is the length of the key. | |
283 | ||
954c1994 GS |
284 | =for apidoc hv_fetch |
285 | ||
286 | Returns the SV which corresponds to the specified key in the hash. The | |
287 | C<klen> is the length of the key. If C<lval> is set then the fetch will be | |
288 | part of a store. Check that the return value is non-null before | |
d1be9408 | 289 | dereferencing it to an C<SV*>. |
954c1994 | 290 | |
96f1132b | 291 | See L<perlguts/"Understanding the Magic of Tied Hashes and Arrays"> for more |
954c1994 GS |
292 | information on how to use this function on tied hashes. |
293 | ||
34a6f7b4 NC |
294 | =for apidoc hv_exists_ent |
295 | ||
296 | Returns a boolean indicating whether the specified hash key exists. C<hash> | |
297 | can be a valid precomputed hash value, or 0 to ask for it to be | |
298 | computed. | |
299 | ||
300 | =cut | |
301 | */ | |
302 | ||
d1be9408 | 303 | /* returns an HE * structure with the all fields set */ |
fde52b5c | 304 | /* note that hent_val will be a mortal sv for MAGICAL hashes */ |
954c1994 GS |
305 | /* |
306 | =for apidoc hv_fetch_ent | |
307 | ||
308 | Returns the hash entry which corresponds to the specified key in the hash. | |
309 | C<hash> must be a valid precomputed hash number for the given C<key>, or 0 | |
310 | if you want the function to compute it. IF C<lval> is set then the fetch | |
311 | will be part of a store. Make sure the return value is non-null before | |
312 | accessing it. The return value when C<tb> is a tied hash is a pointer to a | |
313 | static location, so be sure to make a copy of the structure if you need to | |
1c846c1f | 314 | store it somewhere. |
954c1994 | 315 | |
96f1132b | 316 | See L<perlguts/"Understanding the Magic of Tied Hashes and Arrays"> for more |
954c1994 GS |
317 | information on how to use this function on tied hashes. |
318 | ||
319 | =cut | |
320 | */ | |
321 | ||
a038e571 NC |
322 | /* Common code for hv_delete()/hv_exists()/hv_fetch()/hv_store() */ |
323 | void * | |
324 | Perl_hv_common_key_len(pTHX_ HV *hv, const char *key, I32 klen_i32, | |
325 | const int action, SV *val, const U32 hash) | |
326 | { | |
327 | STRLEN klen; | |
328 | int flags; | |
329 | ||
7918f24d NC |
330 | PERL_ARGS_ASSERT_HV_COMMON_KEY_LEN; |
331 | ||
a038e571 NC |
332 | if (klen_i32 < 0) { |
333 | klen = -klen_i32; | |
334 | flags = HVhek_UTF8; | |
335 | } else { | |
336 | klen = klen_i32; | |
337 | flags = 0; | |
338 | } | |
339 | return hv_common(hv, NULL, key, klen, flags, action, val, hash); | |
340 | } | |
341 | ||
63c89345 | 342 | void * |
d3ba3f5c NC |
343 | Perl_hv_common(pTHX_ HV *hv, SV *keysv, const char *key, STRLEN klen, |
344 | int flags, int action, SV *val, register U32 hash) | |
113738bb | 345 | { |
27da23d5 | 346 | dVAR; |
b2c64049 | 347 | XPVHV* xhv; |
b2c64049 NC |
348 | HE *entry; |
349 | HE **oentry; | |
fde52b5c | 350 | SV *sv; |
da58a35d | 351 | bool is_utf8; |
113738bb | 352 | int masked_flags; |
3c84c864 | 353 | const int return_svp = action & HV_FETCH_JUST_SV; |
fde52b5c | 354 | |
355 | if (!hv) | |
a4fc7abc | 356 | return NULL; |
8265e3d1 NC |
357 | if (SvTYPE(hv) == SVTYPEMASK) |
358 | return NULL; | |
359 | ||
360 | assert(SvTYPE(hv) == SVt_PVHV); | |
fde52b5c | 361 | |
bdee33e4 | 362 | if (SvSMAGICAL(hv) && SvGMAGICAL(hv) && !(action & HV_DISABLE_UVAR_XKEY)) { |
fda2d18a | 363 | MAGIC* mg; |
ad64d0ec | 364 | if ((mg = mg_find((const SV *)hv, PERL_MAGIC_uvar))) { |
fda2d18a NC |
365 | struct ufuncs * const uf = (struct ufuncs *)mg->mg_ptr; |
366 | if (uf->uf_set == NULL) { | |
367 | SV* obj = mg->mg_obj; | |
368 | ||
369 | if (!keysv) { | |
59cd0e26 NC |
370 | keysv = newSVpvn_flags(key, klen, SVs_TEMP | |
371 | ((flags & HVhek_UTF8) | |
372 | ? SVf_UTF8 : 0)); | |
fda2d18a NC |
373 | } |
374 | ||
375 | mg->mg_obj = keysv; /* pass key */ | |
376 | uf->uf_index = action; /* pass action */ | |
ad64d0ec | 377 | magic_getuvar(MUTABLE_SV(hv), mg); |
fda2d18a NC |
378 | keysv = mg->mg_obj; /* may have changed */ |
379 | mg->mg_obj = obj; | |
380 | ||
381 | /* If the key may have changed, then we need to invalidate | |
382 | any passed-in computed hash value. */ | |
383 | hash = 0; | |
384 | } | |
385 | } | |
bdee33e4 | 386 | } |
113738bb | 387 | if (keysv) { |
e593d2fe AE |
388 | if (flags & HVhek_FREEKEY) |
389 | Safefree(key); | |
5c144d81 | 390 | key = SvPV_const(keysv, klen); |
c1fe5510 | 391 | flags = 0; |
113738bb NC |
392 | is_utf8 = (SvUTF8(keysv) != 0); |
393 | } else { | |
c1fe5510 | 394 | is_utf8 = ((flags & HVhek_UTF8) ? TRUE : FALSE); |
113738bb | 395 | } |
113738bb | 396 | |
9dbc5603 | 397 | if (action & HV_DELETE) { |
3c84c864 NC |
398 | return (void *) hv_delete_common(hv, keysv, key, klen, |
399 | flags | (is_utf8 ? HVhek_UTF8 : 0), | |
400 | action, hash); | |
9dbc5603 NC |
401 | } |
402 | ||
b2c64049 | 403 | xhv = (XPVHV*)SvANY(hv); |
7f66fda2 | 404 | if (SvMAGICAL(hv)) { |
6136c704 | 405 | if (SvRMAGICAL(hv) && !(action & (HV_FETCH_ISSTORE|HV_FETCH_ISEXISTS))) { |
ad64d0ec NC |
406 | if (mg_find((const SV *)hv, PERL_MAGIC_tied) |
407 | || SvGMAGICAL((const SV *)hv)) | |
e62cc96a | 408 | { |
3c84c864 | 409 | /* FIXME should be able to skimp on the HE/HEK here when |
7f66fda2 | 410 | HV_FETCH_JUST_SV is true. */ |
7f66fda2 | 411 | if (!keysv) { |
740cce10 NC |
412 | keysv = newSVpvn_utf8(key, klen, is_utf8); |
413 | } else { | |
7f66fda2 | 414 | keysv = newSVsv(keysv); |
113738bb | 415 | } |
44a2ac75 | 416 | sv = sv_newmortal(); |
ad64d0ec | 417 | mg_copy(MUTABLE_SV(hv), sv, (char *)keysv, HEf_SVKEY); |
7f66fda2 NC |
418 | |
419 | /* grab a fake HE/HEK pair from the pool or make a new one */ | |
420 | entry = PL_hv_fetch_ent_mh; | |
421 | if (entry) | |
422 | PL_hv_fetch_ent_mh = HeNEXT(entry); | |
423 | else { | |
424 | char *k; | |
425 | entry = new_HE(); | |
ad64d0ec | 426 | Newx(k, HEK_BASESIZE + sizeof(const SV *), char); |
7f66fda2 NC |
427 | HeKEY_hek(entry) = (HEK*)k; |
428 | } | |
4608196e | 429 | HeNEXT(entry) = NULL; |
7f66fda2 NC |
430 | HeSVKEY_set(entry, keysv); |
431 | HeVAL(entry) = sv; | |
432 | sv_upgrade(sv, SVt_PVLV); | |
433 | LvTYPE(sv) = 'T'; | |
434 | /* so we can free entry when freeing sv */ | |
ad64d0ec | 435 | LvTARG(sv) = MUTABLE_SV(entry); |
7f66fda2 NC |
436 | |
437 | /* XXX remove at some point? */ | |
438 | if (flags & HVhek_FREEKEY) | |
439 | Safefree(key); | |
440 | ||
3c84c864 NC |
441 | if (return_svp) { |
442 | return entry ? (void *) &HeVAL(entry) : NULL; | |
443 | } | |
444 | return (void *) entry; | |
113738bb | 445 | } |
7f66fda2 | 446 | #ifdef ENV_IS_CASELESS |
ad64d0ec | 447 | else if (mg_find((const SV *)hv, PERL_MAGIC_env)) { |
7f66fda2 NC |
448 | U32 i; |
449 | for (i = 0; i < klen; ++i) | |
450 | if (isLOWER(key[i])) { | |
086cb327 NC |
451 | /* Would be nice if we had a routine to do the |
452 | copy and upercase in a single pass through. */ | |
0bd48802 | 453 | const char * const nkey = strupr(savepvn(key,klen)); |
086cb327 NC |
454 | /* Note that this fetch is for nkey (the uppercased |
455 | key) whereas the store is for key (the original) */ | |
63c89345 NC |
456 | void *result = hv_common(hv, NULL, nkey, klen, |
457 | HVhek_FREEKEY, /* free nkey */ | |
458 | 0 /* non-LVAL fetch */ | |
3c84c864 NC |
459 | | HV_DISABLE_UVAR_XKEY |
460 | | return_svp, | |
63c89345 NC |
461 | NULL /* no value */, |
462 | 0 /* compute hash */); | |
26488bcf | 463 | if (!result && (action & HV_FETCH_LVALUE)) { |
086cb327 NC |
464 | /* This call will free key if necessary. |
465 | Do it this way to encourage compiler to tail | |
466 | call optimise. */ | |
63c89345 NC |
467 | result = hv_common(hv, keysv, key, klen, flags, |
468 | HV_FETCH_ISSTORE | |
3c84c864 NC |
469 | | HV_DISABLE_UVAR_XKEY |
470 | | return_svp, | |
63c89345 | 471 | newSV(0), hash); |
086cb327 NC |
472 | } else { |
473 | if (flags & HVhek_FREEKEY) | |
474 | Safefree(key); | |
475 | } | |
63c89345 | 476 | return result; |
7f66fda2 | 477 | } |
902173a3 | 478 | } |
7f66fda2 NC |
479 | #endif |
480 | } /* ISFETCH */ | |
481 | else if (SvRMAGICAL(hv) && (action & HV_FETCH_ISEXISTS)) { | |
ad64d0ec NC |
482 | if (mg_find((const SV *)hv, PERL_MAGIC_tied) |
483 | || SvGMAGICAL((const SV *)hv)) { | |
b2c64049 NC |
484 | /* I don't understand why hv_exists_ent has svret and sv, |
485 | whereas hv_exists only had one. */ | |
9d4ba2ae | 486 | SV * const svret = sv_newmortal(); |
b2c64049 | 487 | sv = sv_newmortal(); |
7f66fda2 NC |
488 | |
489 | if (keysv || is_utf8) { | |
490 | if (!keysv) { | |
740cce10 | 491 | keysv = newSVpvn_utf8(key, klen, TRUE); |
7f66fda2 NC |
492 | } else { |
493 | keysv = newSVsv(keysv); | |
494 | } | |
ad64d0ec | 495 | mg_copy(MUTABLE_SV(hv), sv, (char *)sv_2mortal(keysv), HEf_SVKEY); |
b2c64049 | 496 | } else { |
ad64d0ec | 497 | mg_copy(MUTABLE_SV(hv), sv, key, klen); |
7f66fda2 | 498 | } |
b2c64049 NC |
499 | if (flags & HVhek_FREEKEY) |
500 | Safefree(key); | |
7f66fda2 NC |
501 | magic_existspack(svret, mg_find(sv, PERL_MAGIC_tiedelem)); |
502 | /* This cast somewhat evil, but I'm merely using NULL/ | |
503 | not NULL to return the boolean exists. | |
504 | And I know hv is not NULL. */ | |
3c84c864 | 505 | return SvTRUE(svret) ? (void *)hv : NULL; |
e7152ba2 | 506 | } |
7f66fda2 | 507 | #ifdef ENV_IS_CASELESS |
ad64d0ec | 508 | else if (mg_find((const SV *)hv, PERL_MAGIC_env)) { |
7f66fda2 | 509 | /* XXX This code isn't UTF8 clean. */ |
a15d23f8 | 510 | char * const keysave = (char * const)key; |
b2c64049 NC |
511 | /* Will need to free this, so set FREEKEY flag. */ |
512 | key = savepvn(key,klen); | |
513 | key = (const char*)strupr((char*)key); | |
6136c704 | 514 | is_utf8 = FALSE; |
7f66fda2 | 515 | hash = 0; |
8b4f7dd5 | 516 | keysv = 0; |
b2c64049 NC |
517 | |
518 | if (flags & HVhek_FREEKEY) { | |
519 | Safefree(keysave); | |
520 | } | |
521 | flags |= HVhek_FREEKEY; | |
7f66fda2 | 522 | } |
902173a3 | 523 | #endif |
7f66fda2 | 524 | } /* ISEXISTS */ |
b2c64049 NC |
525 | else if (action & HV_FETCH_ISSTORE) { |
526 | bool needs_copy; | |
527 | bool needs_store; | |
528 | hv_magic_check (hv, &needs_copy, &needs_store); | |
529 | if (needs_copy) { | |
a3b680e6 | 530 | const bool save_taint = PL_tainted; |
b2c64049 NC |
531 | if (keysv || is_utf8) { |
532 | if (!keysv) { | |
740cce10 | 533 | keysv = newSVpvn_utf8(key, klen, TRUE); |
b2c64049 NC |
534 | } |
535 | if (PL_tainting) | |
536 | PL_tainted = SvTAINTED(keysv); | |
537 | keysv = sv_2mortal(newSVsv(keysv)); | |
ad64d0ec | 538 | mg_copy(MUTABLE_SV(hv), val, (char*)keysv, HEf_SVKEY); |
b2c64049 | 539 | } else { |
ad64d0ec | 540 | mg_copy(MUTABLE_SV(hv), val, key, klen); |
b2c64049 NC |
541 | } |
542 | ||
543 | TAINT_IF(save_taint); | |
1baaf5d7 | 544 | if (!needs_store) { |
b2c64049 NC |
545 | if (flags & HVhek_FREEKEY) |
546 | Safefree(key); | |
4608196e | 547 | return NULL; |
b2c64049 NC |
548 | } |
549 | #ifdef ENV_IS_CASELESS | |
ad64d0ec | 550 | else if (mg_find((const SV *)hv, PERL_MAGIC_env)) { |
b2c64049 NC |
551 | /* XXX This code isn't UTF8 clean. */ |
552 | const char *keysave = key; | |
553 | /* Will need to free this, so set FREEKEY flag. */ | |
554 | key = savepvn(key,klen); | |
555 | key = (const char*)strupr((char*)key); | |
6136c704 | 556 | is_utf8 = FALSE; |
b2c64049 | 557 | hash = 0; |
8b4f7dd5 | 558 | keysv = 0; |
b2c64049 NC |
559 | |
560 | if (flags & HVhek_FREEKEY) { | |
561 | Safefree(keysave); | |
562 | } | |
563 | flags |= HVhek_FREEKEY; | |
564 | } | |
565 | #endif | |
566 | } | |
567 | } /* ISSTORE */ | |
7f66fda2 | 568 | } /* SvMAGICAL */ |
fde52b5c | 569 | |
7b2c381c | 570 | if (!HvARRAY(hv)) { |
b2c64049 | 571 | if ((action & (HV_FETCH_LVALUE | HV_FETCH_ISSTORE)) |
fde52b5c | 572 | #ifdef DYNAMIC_ENV_FETCH /* if it's an %ENV lookup, we may get it on the fly */ |
ad64d0ec NC |
573 | || (SvRMAGICAL((const SV *)hv) |
574 | && mg_find((const SV *)hv, PERL_MAGIC_env)) | |
fde52b5c | 575 | #endif |
d58e6666 NC |
576 | ) { |
577 | char *array; | |
a02a5408 | 578 | Newxz(array, |
cbec9347 | 579 | PERL_HV_ARRAY_ALLOC_BYTES(xhv->xhv_max+1 /* HvMAX(hv)+1 */), |
d58e6666 NC |
580 | char); |
581 | HvARRAY(hv) = (HE**)array; | |
582 | } | |
7f66fda2 NC |
583 | #ifdef DYNAMIC_ENV_FETCH |
584 | else if (action & HV_FETCH_ISEXISTS) { | |
585 | /* for an %ENV exists, if we do an insert it's by a recursive | |
586 | store call, so avoid creating HvARRAY(hv) right now. */ | |
587 | } | |
588 | #endif | |
113738bb NC |
589 | else { |
590 | /* XXX remove at some point? */ | |
591 | if (flags & HVhek_FREEKEY) | |
592 | Safefree(key); | |
593 | ||
3c84c864 | 594 | return NULL; |
113738bb | 595 | } |
fde52b5c | 596 | } |
597 | ||
19692e8d | 598 | if (is_utf8) { |
41d88b63 | 599 | char * const keysave = (char *)key; |
f9a63242 | 600 | key = (char*)bytes_from_utf8((U8*)key, &klen, &is_utf8); |
19692e8d | 601 | if (is_utf8) |
c1fe5510 NC |
602 | flags |= HVhek_UTF8; |
603 | else | |
604 | flags &= ~HVhek_UTF8; | |
7f66fda2 NC |
605 | if (key != keysave) { |
606 | if (flags & HVhek_FREEKEY) | |
607 | Safefree(keysave); | |
19692e8d | 608 | flags |= HVhek_WASUTF8 | HVhek_FREEKEY; |
7f66fda2 | 609 | } |
19692e8d | 610 | } |
f9a63242 | 611 | |
4b5190b5 NC |
612 | if (HvREHASH(hv)) { |
613 | PERL_HASH_INTERNAL(hash, key, klen); | |
b2c64049 NC |
614 | /* We don't have a pointer to the hv, so we have to replicate the |
615 | flag into every HEK, so that hv_iterkeysv can see it. */ | |
616 | /* And yes, you do need this even though you are not "storing" because | |
fdcd69b6 NC |
617 | you can flip the flags below if doing an lval lookup. (And that |
618 | was put in to give the semantics Andreas was expecting.) */ | |
619 | flags |= HVhek_REHASH; | |
4b5190b5 | 620 | } else if (!hash) { |
113738bb | 621 | if (keysv && (SvIsCOW_shared_hash(keysv))) { |
c158a4fd | 622 | hash = SvSHARED_HASH(keysv); |
46187eeb NC |
623 | } else { |
624 | PERL_HASH(hash, key, klen); | |
625 | } | |
626 | } | |
effa1e2d | 627 | |
113738bb NC |
628 | masked_flags = (flags & HVhek_MASK); |
629 | ||
7f66fda2 | 630 | #ifdef DYNAMIC_ENV_FETCH |
4608196e | 631 | if (!HvARRAY(hv)) entry = NULL; |
7f66fda2 NC |
632 | else |
633 | #endif | |
b2c64049 | 634 | { |
7b2c381c | 635 | entry = (HvARRAY(hv))[hash & (I32) HvMAX(hv)]; |
b2c64049 | 636 | } |
0298d7b9 | 637 | for (; entry; entry = HeNEXT(entry)) { |
fde52b5c | 638 | if (HeHASH(entry) != hash) /* strings can't be equal */ |
639 | continue; | |
eb160463 | 640 | if (HeKLEN(entry) != (I32)klen) |
fde52b5c | 641 | continue; |
1c846c1f | 642 | if (HeKEY(entry) != key && memNE(HeKEY(entry),key,klen)) /* is this it? */ |
fde52b5c | 643 | continue; |
113738bb | 644 | if ((HeKFLAGS(entry) ^ masked_flags) & HVhek_UTF8) |
c3654f1a | 645 | continue; |
b2c64049 NC |
646 | |
647 | if (action & (HV_FETCH_LVALUE|HV_FETCH_ISSTORE)) { | |
648 | if (HeKFLAGS(entry) != masked_flags) { | |
649 | /* We match if HVhek_UTF8 bit in our flags and hash key's | |
650 | match. But if entry was set previously with HVhek_WASUTF8 | |
651 | and key now doesn't (or vice versa) then we should change | |
652 | the key's flag, as this is assignment. */ | |
653 | if (HvSHAREKEYS(hv)) { | |
654 | /* Need to swap the key we have for a key with the flags we | |
655 | need. As keys are shared we can't just write to the | |
656 | flag, so we share the new one, unshare the old one. */ | |
6136c704 | 657 | HEK * const new_hek = share_hek_flags(key, klen, hash, |
6e838c70 | 658 | masked_flags); |
b2c64049 NC |
659 | unshare_hek (HeKEY_hek(entry)); |
660 | HeKEY_hek(entry) = new_hek; | |
661 | } | |
5d2b1485 NC |
662 | else if (hv == PL_strtab) { |
663 | /* PL_strtab is usually the only hash without HvSHAREKEYS, | |
664 | so putting this test here is cheap */ | |
665 | if (flags & HVhek_FREEKEY) | |
666 | Safefree(key); | |
667 | Perl_croak(aTHX_ S_strtab_error, | |
668 | action & HV_FETCH_LVALUE ? "fetch" : "store"); | |
669 | } | |
b2c64049 NC |
670 | else |
671 | HeKFLAGS(entry) = masked_flags; | |
672 | if (masked_flags & HVhek_ENABLEHVKFLAGS) | |
673 | HvHASKFLAGS_on(hv); | |
674 | } | |
675 | if (HeVAL(entry) == &PL_sv_placeholder) { | |
676 | /* yes, can store into placeholder slot */ | |
677 | if (action & HV_FETCH_LVALUE) { | |
678 | if (SvMAGICAL(hv)) { | |
679 | /* This preserves behaviour with the old hv_fetch | |
680 | implementation which at this point would bail out | |
681 | with a break; (at "if we find a placeholder, we | |
682 | pretend we haven't found anything") | |
683 | ||
684 | That break mean that if a placeholder were found, it | |
685 | caused a call into hv_store, which in turn would | |
686 | check magic, and if there is no magic end up pretty | |
687 | much back at this point (in hv_store's code). */ | |
688 | break; | |
689 | } | |
690 | /* LVAL fetch which actaully needs a store. */ | |
561b68a9 | 691 | val = newSV(0); |
ca732855 | 692 | HvPLACEHOLDERS(hv)--; |
b2c64049 NC |
693 | } else { |
694 | /* store */ | |
695 | if (val != &PL_sv_placeholder) | |
ca732855 | 696 | HvPLACEHOLDERS(hv)--; |
b2c64049 NC |
697 | } |
698 | HeVAL(entry) = val; | |
699 | } else if (action & HV_FETCH_ISSTORE) { | |
700 | SvREFCNT_dec(HeVAL(entry)); | |
701 | HeVAL(entry) = val; | |
702 | } | |
27bcc0a7 | 703 | } else if (HeVAL(entry) == &PL_sv_placeholder) { |
b2c64049 NC |
704 | /* if we find a placeholder, we pretend we haven't found |
705 | anything */ | |
8aacddc1 | 706 | break; |
b2c64049 | 707 | } |
113738bb NC |
708 | if (flags & HVhek_FREEKEY) |
709 | Safefree(key); | |
3c84c864 NC |
710 | if (return_svp) { |
711 | return entry ? (void *) &HeVAL(entry) : NULL; | |
712 | } | |
fde52b5c | 713 | return entry; |
714 | } | |
715 | #ifdef DYNAMIC_ENV_FETCH /* %ENV lookup? If so, try to fetch the value now */ | |
0ed29950 | 716 | if (!(action & HV_FETCH_ISSTORE) |
ad64d0ec NC |
717 | && SvRMAGICAL((const SV *)hv) |
718 | && mg_find((const SV *)hv, PERL_MAGIC_env)) { | |
a6c40364 | 719 | unsigned long len; |
9d4ba2ae | 720 | const char * const env = PerlEnv_ENVgetenv_len(key,&len); |
a6c40364 GS |
721 | if (env) { |
722 | sv = newSVpvn(env,len); | |
723 | SvTAINTED_on(sv); | |
d3ba3f5c | 724 | return hv_common(hv, keysv, key, klen, flags, |
3c84c864 NC |
725 | HV_FETCH_ISSTORE|HV_DISABLE_UVAR_XKEY|return_svp, |
726 | sv, hash); | |
a6c40364 | 727 | } |
fde52b5c | 728 | } |
729 | #endif | |
7f66fda2 NC |
730 | |
731 | if (!entry && SvREADONLY(hv) && !(action & HV_FETCH_ISEXISTS)) { | |
c445ea15 | 732 | hv_notallowed(flags, key, klen, |
c8cd6465 NC |
733 | "Attempt to access disallowed key '%"SVf"' in" |
734 | " a restricted hash"); | |
1b1f1335 | 735 | } |
b2c64049 NC |
736 | if (!(action & (HV_FETCH_LVALUE|HV_FETCH_ISSTORE))) { |
737 | /* Not doing some form of store, so return failure. */ | |
738 | if (flags & HVhek_FREEKEY) | |
739 | Safefree(key); | |
3c84c864 | 740 | return NULL; |
b2c64049 | 741 | } |
113738bb | 742 | if (action & HV_FETCH_LVALUE) { |
561b68a9 | 743 | val = newSV(0); |
b2c64049 NC |
744 | if (SvMAGICAL(hv)) { |
745 | /* At this point the old hv_fetch code would call to hv_store, | |
746 | which in turn might do some tied magic. So we need to make that | |
747 | magic check happen. */ | |
748 | /* gonna assign to this, so it better be there */ | |
fda2d18a NC |
749 | /* If a fetch-as-store fails on the fetch, then the action is to |
750 | recurse once into "hv_store". If we didn't do this, then that | |
751 | recursive call would call the key conversion routine again. | |
752 | However, as we replace the original key with the converted | |
753 | key, this would result in a double conversion, which would show | |
754 | up as a bug if the conversion routine is not idempotent. */ | |
d3ba3f5c | 755 | return hv_common(hv, keysv, key, klen, flags, |
3c84c864 NC |
756 | HV_FETCH_ISSTORE|HV_DISABLE_UVAR_XKEY|return_svp, |
757 | val, hash); | |
b2c64049 NC |
758 | /* XXX Surely that could leak if the fetch-was-store fails? |
759 | Just like the hv_fetch. */ | |
113738bb NC |
760 | } |
761 | } | |
762 | ||
b2c64049 NC |
763 | /* Welcome to hv_store... */ |
764 | ||
7b2c381c | 765 | if (!HvARRAY(hv)) { |
b2c64049 NC |
766 | /* Not sure if we can get here. I think the only case of oentry being |
767 | NULL is for %ENV with dynamic env fetch. But that should disappear | |
768 | with magic in the previous code. */ | |
d58e6666 | 769 | char *array; |
a02a5408 | 770 | Newxz(array, |
b2c64049 | 771 | PERL_HV_ARRAY_ALLOC_BYTES(xhv->xhv_max+1 /* HvMAX(hv)+1 */), |
d58e6666 NC |
772 | char); |
773 | HvARRAY(hv) = (HE**)array; | |
b2c64049 NC |
774 | } |
775 | ||
7b2c381c | 776 | oentry = &(HvARRAY(hv))[hash & (I32) xhv->xhv_max]; |
ab4af705 | 777 | |
b2c64049 NC |
778 | entry = new_HE(); |
779 | /* share_hek_flags will do the free for us. This might be considered | |
780 | bad API design. */ | |
781 | if (HvSHAREKEYS(hv)) | |
6e838c70 | 782 | HeKEY_hek(entry) = share_hek_flags(key, klen, hash, flags); |
5d2b1485 NC |
783 | else if (hv == PL_strtab) { |
784 | /* PL_strtab is usually the only hash without HvSHAREKEYS, so putting | |
785 | this test here is cheap */ | |
786 | if (flags & HVhek_FREEKEY) | |
787 | Safefree(key); | |
788 | Perl_croak(aTHX_ S_strtab_error, | |
789 | action & HV_FETCH_LVALUE ? "fetch" : "store"); | |
790 | } | |
b2c64049 NC |
791 | else /* gotta do the real thing */ |
792 | HeKEY_hek(entry) = save_hek_flags(key, klen, hash, flags); | |
793 | HeVAL(entry) = val; | |
794 | HeNEXT(entry) = *oentry; | |
795 | *oentry = entry; | |
796 | ||
797 | if (val == &PL_sv_placeholder) | |
ca732855 | 798 | HvPLACEHOLDERS(hv)++; |
b2c64049 NC |
799 | if (masked_flags & HVhek_ENABLEHVKFLAGS) |
800 | HvHASKFLAGS_on(hv); | |
801 | ||
0298d7b9 NC |
802 | { |
803 | const HE *counter = HeNEXT(entry); | |
804 | ||
4c7185a0 | 805 | xhv->xhv_keys++; /* HvTOTALKEYS(hv)++ */ |
0298d7b9 NC |
806 | if (!counter) { /* initial entry? */ |
807 | xhv->xhv_fill++; /* HvFILL(hv)++ */ | |
808 | } else if (xhv->xhv_keys > (IV)xhv->xhv_max) { | |
809 | hsplit(hv); | |
810 | } else if(!HvREHASH(hv)) { | |
811 | U32 n_links = 1; | |
812 | ||
813 | while ((counter = HeNEXT(counter))) | |
814 | n_links++; | |
815 | ||
816 | if (n_links > HV_MAX_LENGTH_BEFORE_SPLIT) { | |
817 | /* Use only the old HvKEYS(hv) > HvMAX(hv) condition to limit | |
818 | bucket splits on a rehashed hash, as we're not going to | |
819 | split it again, and if someone is lucky (evil) enough to | |
820 | get all the keys in one list they could exhaust our memory | |
821 | as we repeatedly double the number of buckets on every | |
822 | entry. Linear search feels a less worse thing to do. */ | |
823 | hsplit(hv); | |
824 | } | |
825 | } | |
fde52b5c | 826 | } |
b2c64049 | 827 | |
3c84c864 NC |
828 | if (return_svp) { |
829 | return entry ? (void *) &HeVAL(entry) : NULL; | |
830 | } | |
831 | return (void *) entry; | |
fde52b5c | 832 | } |
833 | ||
864dbfa3 | 834 | STATIC void |
b0e6ae5b | 835 | S_hv_magic_check(HV *hv, bool *needs_copy, bool *needs_store) |
d0066dc7 | 836 | { |
a3b680e6 | 837 | const MAGIC *mg = SvMAGIC(hv); |
7918f24d NC |
838 | |
839 | PERL_ARGS_ASSERT_HV_MAGIC_CHECK; | |
840 | ||
d0066dc7 OT |
841 | *needs_copy = FALSE; |
842 | *needs_store = TRUE; | |
843 | while (mg) { | |
844 | if (isUPPER(mg->mg_type)) { | |
845 | *needs_copy = TRUE; | |
d60c5a05 | 846 | if (mg->mg_type == PERL_MAGIC_tied) { |
d0066dc7 | 847 | *needs_store = FALSE; |
4ab2a30b | 848 | return; /* We've set all there is to set. */ |
d0066dc7 OT |
849 | } |
850 | } | |
851 | mg = mg->mg_moremagic; | |
852 | } | |
853 | } | |
854 | ||
954c1994 | 855 | /* |
a3bcc51e TP |
856 | =for apidoc hv_scalar |
857 | ||
858 | Evaluates the hash in scalar context and returns the result. Handles magic when the hash is tied. | |
859 | ||
860 | =cut | |
861 | */ | |
862 | ||
863 | SV * | |
864 | Perl_hv_scalar(pTHX_ HV *hv) | |
865 | { | |
a3bcc51e | 866 | SV *sv; |
823a54a3 | 867 | |
7918f24d NC |
868 | PERL_ARGS_ASSERT_HV_SCALAR; |
869 | ||
823a54a3 | 870 | if (SvRMAGICAL(hv)) { |
ad64d0ec | 871 | MAGIC * const mg = mg_find((const SV *)hv, PERL_MAGIC_tied); |
823a54a3 AL |
872 | if (mg) |
873 | return magic_scalarpack(hv, mg); | |
874 | } | |
a3bcc51e TP |
875 | |
876 | sv = sv_newmortal(); | |
85fbaab2 | 877 | if (HvFILL((const HV *)hv)) |
a3bcc51e TP |
878 | Perl_sv_setpvf(aTHX_ sv, "%ld/%ld", |
879 | (long)HvFILL(hv), (long)HvMAX(hv) + 1); | |
880 | else | |
881 | sv_setiv(sv, 0); | |
882 | ||
883 | return sv; | |
884 | } | |
885 | ||
886 | /* | |
954c1994 GS |
887 | =for apidoc hv_delete |
888 | ||
889 | Deletes a key/value pair in the hash. The value SV is removed from the | |
1c846c1f | 890 | hash and returned to the caller. The C<klen> is the length of the key. |
954c1994 GS |
891 | The C<flags> value will normally be zero; if set to G_DISCARD then NULL |
892 | will be returned. | |
893 | ||
954c1994 GS |
894 | =for apidoc hv_delete_ent |
895 | ||
896 | Deletes a key/value pair in the hash. The value SV is removed from the | |
897 | hash and returned to the caller. The C<flags> value will normally be zero; | |
898 | if set to G_DISCARD then NULL will be returned. C<hash> can be a valid | |
899 | precomputed hash value, or 0 to ask for it to be computed. | |
900 | ||
901 | =cut | |
902 | */ | |
903 | ||
8f8d40ab | 904 | STATIC SV * |
cd6d36ac NC |
905 | S_hv_delete_common(pTHX_ HV *hv, SV *keysv, const char *key, STRLEN klen, |
906 | int k_flags, I32 d_flags, U32 hash) | |
f1317c8d | 907 | { |
27da23d5 | 908 | dVAR; |
cbec9347 | 909 | register XPVHV* xhv; |
fde52b5c | 910 | register HE *entry; |
911 | register HE **oentry; | |
9e720f71 | 912 | HE *const *first_entry; |
9dbc5603 | 913 | bool is_utf8 = (k_flags & HVhek_UTF8) ? TRUE : FALSE; |
7a9669ca | 914 | int masked_flags; |
1c846c1f | 915 | |
fde52b5c | 916 | if (SvRMAGICAL(hv)) { |
0a0bb7c7 OT |
917 | bool needs_copy; |
918 | bool needs_store; | |
919 | hv_magic_check (hv, &needs_copy, &needs_store); | |
920 | ||
f1317c8d | 921 | if (needs_copy) { |
6136c704 | 922 | SV *sv; |
63c89345 NC |
923 | entry = (HE *) hv_common(hv, keysv, key, klen, |
924 | k_flags & ~HVhek_FREEKEY, | |
925 | HV_FETCH_LVALUE|HV_DISABLE_UVAR_XKEY, | |
926 | NULL, hash); | |
7a9669ca | 927 | sv = entry ? HeVAL(entry) : NULL; |
f1317c8d NC |
928 | if (sv) { |
929 | if (SvMAGICAL(sv)) { | |
930 | mg_clear(sv); | |
931 | } | |
932 | if (!needs_store) { | |
933 | if (mg_find(sv, PERL_MAGIC_tiedelem)) { | |
934 | /* No longer an element */ | |
935 | sv_unmagic(sv, PERL_MAGIC_tiedelem); | |
936 | return sv; | |
937 | } | |
a0714e2c | 938 | return NULL; /* element cannot be deleted */ |
f1317c8d | 939 | } |
902173a3 | 940 | #ifdef ENV_IS_CASELESS |
ad64d0ec | 941 | else if (mg_find((const SV *)hv, PERL_MAGIC_env)) { |
8167a60a | 942 | /* XXX This code isn't UTF8 clean. */ |
59cd0e26 | 943 | keysv = newSVpvn_flags(key, klen, SVs_TEMP); |
8167a60a NC |
944 | if (k_flags & HVhek_FREEKEY) { |
945 | Safefree(key); | |
946 | } | |
947 | key = strupr(SvPVX(keysv)); | |
948 | is_utf8 = 0; | |
949 | k_flags = 0; | |
950 | hash = 0; | |
7f66fda2 | 951 | } |
510ac311 | 952 | #endif |
2fd1c6b8 | 953 | } |
2fd1c6b8 | 954 | } |
fde52b5c | 955 | } |
cbec9347 | 956 | xhv = (XPVHV*)SvANY(hv); |
7b2c381c | 957 | if (!HvARRAY(hv)) |
a0714e2c | 958 | return NULL; |
fde52b5c | 959 | |
19692e8d | 960 | if (is_utf8) { |
c445ea15 | 961 | const char * const keysave = key; |
b464bac0 | 962 | key = (char*)bytes_from_utf8((U8*)key, &klen, &is_utf8); |
cd6d36ac | 963 | |
19692e8d | 964 | if (is_utf8) |
cd6d36ac NC |
965 | k_flags |= HVhek_UTF8; |
966 | else | |
967 | k_flags &= ~HVhek_UTF8; | |
7f66fda2 NC |
968 | if (key != keysave) { |
969 | if (k_flags & HVhek_FREEKEY) { | |
970 | /* This shouldn't happen if our caller does what we expect, | |
971 | but strictly the API allows it. */ | |
972 | Safefree(keysave); | |
973 | } | |
974 | k_flags |= HVhek_WASUTF8 | HVhek_FREEKEY; | |
975 | } | |
ad64d0ec | 976 | HvHASKFLAGS_on(MUTABLE_SV(hv)); |
19692e8d | 977 | } |
f9a63242 | 978 | |
4b5190b5 NC |
979 | if (HvREHASH(hv)) { |
980 | PERL_HASH_INTERNAL(hash, key, klen); | |
981 | } else if (!hash) { | |
7a9669ca | 982 | if (keysv && (SvIsCOW_shared_hash(keysv))) { |
c158a4fd | 983 | hash = SvSHARED_HASH(keysv); |
7a9669ca NC |
984 | } else { |
985 | PERL_HASH(hash, key, klen); | |
986 | } | |
4b5190b5 | 987 | } |
fde52b5c | 988 | |
7a9669ca NC |
989 | masked_flags = (k_flags & HVhek_MASK); |
990 | ||
9e720f71 | 991 | first_entry = oentry = &(HvARRAY(hv))[hash & (I32) HvMAX(hv)]; |
fde52b5c | 992 | entry = *oentry; |
9e720f71 | 993 | for (; entry; oentry = &HeNEXT(entry), entry = *oentry) { |
6136c704 | 994 | SV *sv; |
fde52b5c | 995 | if (HeHASH(entry) != hash) /* strings can't be equal */ |
996 | continue; | |
eb160463 | 997 | if (HeKLEN(entry) != (I32)klen) |
fde52b5c | 998 | continue; |
1c846c1f | 999 | if (HeKEY(entry) != key && memNE(HeKEY(entry),key,klen)) /* is this it? */ |
fde52b5c | 1000 | continue; |
7a9669ca | 1001 | if ((HeKFLAGS(entry) ^ masked_flags) & HVhek_UTF8) |
c3654f1a | 1002 | continue; |
8aacddc1 | 1003 | |
5d2b1485 NC |
1004 | if (hv == PL_strtab) { |
1005 | if (k_flags & HVhek_FREEKEY) | |
1006 | Safefree(key); | |
1007 | Perl_croak(aTHX_ S_strtab_error, "delete"); | |
1008 | } | |
1009 | ||
8aacddc1 | 1010 | /* if placeholder is here, it's already been deleted.... */ |
6136c704 AL |
1011 | if (HeVAL(entry) == &PL_sv_placeholder) { |
1012 | if (k_flags & HVhek_FREEKEY) | |
1013 | Safefree(key); | |
1014 | return NULL; | |
8aacddc1 | 1015 | } |
6136c704 | 1016 | if (SvREADONLY(hv) && HeVAL(entry) && SvREADONLY(HeVAL(entry))) { |
d4c19fe8 | 1017 | hv_notallowed(k_flags, key, klen, |
c8cd6465 NC |
1018 | "Attempt to delete readonly key '%"SVf"' from" |
1019 | " a restricted hash"); | |
8aacddc1 | 1020 | } |
b84d0860 NC |
1021 | if (k_flags & HVhek_FREEKEY) |
1022 | Safefree(key); | |
8aacddc1 | 1023 | |
cd6d36ac | 1024 | if (d_flags & G_DISCARD) |
a0714e2c | 1025 | sv = NULL; |
94f7643d | 1026 | else { |
79d01fbf | 1027 | sv = sv_2mortal(HeVAL(entry)); |
7996736c | 1028 | HeVAL(entry) = &PL_sv_placeholder; |
94f7643d | 1029 | } |
8aacddc1 NIS |
1030 | |
1031 | /* | |
1032 | * If a restricted hash, rather than really deleting the entry, put | |
1033 | * a placeholder there. This marks the key as being "approved", so | |
1034 | * we can still access via not-really-existing key without raising | |
1035 | * an error. | |
1036 | */ | |
1037 | if (SvREADONLY(hv)) { | |
754604c4 | 1038 | SvREFCNT_dec(HeVAL(entry)); |
7996736c | 1039 | HeVAL(entry) = &PL_sv_placeholder; |
8aacddc1 NIS |
1040 | /* We'll be saving this slot, so the number of allocated keys |
1041 | * doesn't go down, but the number placeholders goes up */ | |
ca732855 | 1042 | HvPLACEHOLDERS(hv)++; |
8aacddc1 | 1043 | } else { |
a26e96df | 1044 | *oentry = HeNEXT(entry); |
9e720f71 | 1045 | if(!*first_entry) { |
a26e96df | 1046 | xhv->xhv_fill--; /* HvFILL(hv)-- */ |
9e720f71 | 1047 | } |
b79f7545 | 1048 | if (SvOOK(hv) && entry == HvAUX(hv)->xhv_eiter /* HvEITER(hv) */) |
8aacddc1 NIS |
1049 | HvLAZYDEL_on(hv); |
1050 | else | |
1051 | hv_free_ent(hv, entry); | |
4c7185a0 | 1052 | xhv->xhv_keys--; /* HvTOTALKEYS(hv)-- */ |
574c8022 | 1053 | if (xhv->xhv_keys == 0) |
19692e8d | 1054 | HvHASKFLAGS_off(hv); |
8aacddc1 | 1055 | } |
79072805 LW |
1056 | return sv; |
1057 | } | |
8aacddc1 | 1058 | if (SvREADONLY(hv)) { |
d4c19fe8 | 1059 | hv_notallowed(k_flags, key, klen, |
c8cd6465 NC |
1060 | "Attempt to delete disallowed key '%"SVf"' from" |
1061 | " a restricted hash"); | |
8aacddc1 NIS |
1062 | } |
1063 | ||
19692e8d | 1064 | if (k_flags & HVhek_FREEKEY) |
f9a63242 | 1065 | Safefree(key); |
a0714e2c | 1066 | return NULL; |
79072805 LW |
1067 | } |
1068 | ||
76e3520e | 1069 | STATIC void |
cea2e8a9 | 1070 | S_hsplit(pTHX_ HV *hv) |
79072805 | 1071 | { |
97aff369 | 1072 | dVAR; |
1e05feb3 | 1073 | register XPVHV* const xhv = (XPVHV*)SvANY(hv); |
a3b680e6 | 1074 | const I32 oldsize = (I32) xhv->xhv_max+1; /* HvMAX(hv)+1 (sick) */ |
79072805 LW |
1075 | register I32 newsize = oldsize * 2; |
1076 | register I32 i; | |
7b2c381c | 1077 | char *a = (char*) HvARRAY(hv); |
72311751 | 1078 | register HE **aep; |
79072805 | 1079 | register HE **oentry; |
4b5190b5 NC |
1080 | int longest_chain = 0; |
1081 | int was_shared; | |
79072805 | 1082 | |
7918f24d NC |
1083 | PERL_ARGS_ASSERT_HSPLIT; |
1084 | ||
18026298 | 1085 | /*PerlIO_printf(PerlIO_stderr(), "hsplit called for %p which had %d\n", |
6c9570dc | 1086 | (void*)hv, (int) oldsize);*/ |
18026298 | 1087 | |
5d88ecd7 | 1088 | if (HvPLACEHOLDERS_get(hv) && !SvREADONLY(hv)) { |
18026298 NC |
1089 | /* Can make this clear any placeholders first for non-restricted hashes, |
1090 | even though Storable rebuilds restricted hashes by putting in all the | |
1091 | placeholders (first) before turning on the readonly flag, because | |
1092 | Storable always pre-splits the hash. */ | |
1093 | hv_clear_placeholders(hv); | |
1094 | } | |
1095 | ||
3280af22 | 1096 | PL_nomemok = TRUE; |
8d6dde3e | 1097 | #if defined(STRANGE_MALLOC) || defined(MYMALLOC) |
b79f7545 NC |
1098 | Renew(a, PERL_HV_ARRAY_ALLOC_BYTES(newsize) |
1099 | + (SvOOK(hv) ? sizeof(struct xpvhv_aux) : 0), char); | |
422a93e5 | 1100 | if (!a) { |
4a33f861 | 1101 | PL_nomemok = FALSE; |
422a93e5 GA |
1102 | return; |
1103 | } | |
b79f7545 | 1104 | if (SvOOK(hv)) { |
31f0e52e | 1105 | Move(&a[oldsize * sizeof(HE*)], &a[newsize * sizeof(HE*)], 1, struct xpvhv_aux); |
b79f7545 | 1106 | } |
4633a7c4 | 1107 | #else |
a02a5408 | 1108 | Newx(a, PERL_HV_ARRAY_ALLOC_BYTES(newsize) |
b79f7545 | 1109 | + (SvOOK(hv) ? sizeof(struct xpvhv_aux) : 0), char); |
422a93e5 | 1110 | if (!a) { |
3280af22 | 1111 | PL_nomemok = FALSE; |
422a93e5 GA |
1112 | return; |
1113 | } | |
7b2c381c | 1114 | Copy(HvARRAY(hv), a, oldsize * sizeof(HE*), char); |
b79f7545 NC |
1115 | if (SvOOK(hv)) { |
1116 | Copy(HvAUX(hv), &a[newsize * sizeof(HE*)], 1, struct xpvhv_aux); | |
1117 | } | |
fba3b22e | 1118 | if (oldsize >= 64) { |
7b2c381c | 1119 | offer_nice_chunk(HvARRAY(hv), |
b79f7545 NC |
1120 | PERL_HV_ARRAY_ALLOC_BYTES(oldsize) |
1121 | + (SvOOK(hv) ? sizeof(struct xpvhv_aux) : 0)); | |
4633a7c4 LW |
1122 | } |
1123 | else | |
7b2c381c | 1124 | Safefree(HvARRAY(hv)); |
4633a7c4 LW |
1125 | #endif |
1126 | ||
3280af22 | 1127 | PL_nomemok = FALSE; |
72311751 | 1128 | Zero(&a[oldsize * sizeof(HE*)], (newsize-oldsize) * sizeof(HE*), char); /* zero 2nd half*/ |
cbec9347 | 1129 | xhv->xhv_max = --newsize; /* HvMAX(hv) = --newsize */ |
7b2c381c | 1130 | HvARRAY(hv) = (HE**) a; |
72311751 | 1131 | aep = (HE**)a; |
79072805 | 1132 | |
72311751 | 1133 | for (i=0; i<oldsize; i++,aep++) { |
4b5190b5 NC |
1134 | int left_length = 0; |
1135 | int right_length = 0; | |
a3b680e6 AL |
1136 | register HE *entry; |
1137 | register HE **bep; | |
4b5190b5 | 1138 | |
72311751 | 1139 | if (!*aep) /* non-existent */ |
79072805 | 1140 | continue; |
72311751 GS |
1141 | bep = aep+oldsize; |
1142 | for (oentry = aep, entry = *aep; entry; entry = *oentry) { | |
eb160463 | 1143 | if ((HeHASH(entry) & newsize) != (U32)i) { |
fde52b5c | 1144 | *oentry = HeNEXT(entry); |
72311751 GS |
1145 | HeNEXT(entry) = *bep; |
1146 | if (!*bep) | |
cbec9347 | 1147 | xhv->xhv_fill++; /* HvFILL(hv)++ */ |
72311751 | 1148 | *bep = entry; |
4b5190b5 | 1149 | right_length++; |
79072805 LW |
1150 | continue; |
1151 | } | |
4b5190b5 | 1152 | else { |
fde52b5c | 1153 | oentry = &HeNEXT(entry); |
4b5190b5 NC |
1154 | left_length++; |
1155 | } | |
79072805 | 1156 | } |
72311751 | 1157 | if (!*aep) /* everything moved */ |
cbec9347 | 1158 | xhv->xhv_fill--; /* HvFILL(hv)-- */ |
4b5190b5 NC |
1159 | /* I think we don't actually need to keep track of the longest length, |
1160 | merely flag if anything is too long. But for the moment while | |
1161 | developing this code I'll track it. */ | |
1162 | if (left_length > longest_chain) | |
1163 | longest_chain = left_length; | |
1164 | if (right_length > longest_chain) | |
1165 | longest_chain = right_length; | |
1166 | } | |
1167 | ||
1168 | ||
1169 | /* Pick your policy for "hashing isn't working" here: */ | |
fdcd69b6 | 1170 | if (longest_chain <= HV_MAX_LENGTH_BEFORE_SPLIT /* split worked? */ |
4b5190b5 NC |
1171 | || HvREHASH(hv)) { |
1172 | return; | |
79072805 | 1173 | } |
4b5190b5 NC |
1174 | |
1175 | if (hv == PL_strtab) { | |
1176 | /* Urg. Someone is doing something nasty to the string table. | |
1177 | Can't win. */ | |
1178 | return; | |
1179 | } | |
1180 | ||
1181 | /* Awooga. Awooga. Pathological data. */ | |
6c9570dc | 1182 | /*PerlIO_printf(PerlIO_stderr(), "%p %d of %d with %d/%d buckets\n", (void*)hv, |
4b5190b5 NC |
1183 | longest_chain, HvTOTALKEYS(hv), HvFILL(hv), 1+HvMAX(hv));*/ |
1184 | ||
1185 | ++newsize; | |
a02a5408 | 1186 | Newxz(a, PERL_HV_ARRAY_ALLOC_BYTES(newsize) |
b79f7545 NC |
1187 | + (SvOOK(hv) ? sizeof(struct xpvhv_aux) : 0), char); |
1188 | if (SvOOK(hv)) { | |
1189 | Copy(HvAUX(hv), &a[newsize * sizeof(HE*)], 1, struct xpvhv_aux); | |
1190 | } | |
1191 | ||
4b5190b5 NC |
1192 | was_shared = HvSHAREKEYS(hv); |
1193 | ||
1194 | xhv->xhv_fill = 0; | |
1195 | HvSHAREKEYS_off(hv); | |
1196 | HvREHASH_on(hv); | |
1197 | ||
7b2c381c | 1198 | aep = HvARRAY(hv); |
4b5190b5 NC |
1199 | |
1200 | for (i=0; i<newsize; i++,aep++) { | |
a3b680e6 | 1201 | register HE *entry = *aep; |
4b5190b5 NC |
1202 | while (entry) { |
1203 | /* We're going to trash this HE's next pointer when we chain it | |
1204 | into the new hash below, so store where we go next. */ | |
9d4ba2ae | 1205 | HE * const next = HeNEXT(entry); |
4b5190b5 | 1206 | UV hash; |
a3b680e6 | 1207 | HE **bep; |
4b5190b5 NC |
1208 | |
1209 | /* Rehash it */ | |
1210 | PERL_HASH_INTERNAL(hash, HeKEY(entry), HeKLEN(entry)); | |
1211 | ||
1212 | if (was_shared) { | |
1213 | /* Unshare it. */ | |
aec46f14 | 1214 | HEK * const new_hek |
4b5190b5 NC |
1215 | = save_hek_flags(HeKEY(entry), HeKLEN(entry), |
1216 | hash, HeKFLAGS(entry)); | |
1217 | unshare_hek (HeKEY_hek(entry)); | |
1218 | HeKEY_hek(entry) = new_hek; | |
1219 | } else { | |
1220 | /* Not shared, so simply write the new hash in. */ | |
1221 | HeHASH(entry) = hash; | |
1222 | } | |
1223 | /*PerlIO_printf(PerlIO_stderr(), "%d ", HeKFLAGS(entry));*/ | |
1224 | HEK_REHASH_on(HeKEY_hek(entry)); | |
1225 | /*PerlIO_printf(PerlIO_stderr(), "%d\n", HeKFLAGS(entry));*/ | |
1226 | ||
1227 | /* Copy oentry to the correct new chain. */ | |
1228 | bep = ((HE**)a) + (hash & (I32) xhv->xhv_max); | |
1229 | if (!*bep) | |
1230 | xhv->xhv_fill++; /* HvFILL(hv)++ */ | |
1231 | HeNEXT(entry) = *bep; | |
1232 | *bep = entry; | |
1233 | ||
1234 | entry = next; | |
1235 | } | |
1236 | } | |
7b2c381c NC |
1237 | Safefree (HvARRAY(hv)); |
1238 | HvARRAY(hv) = (HE **)a; | |
79072805 LW |
1239 | } |
1240 | ||
72940dca | 1241 | void |
864dbfa3 | 1242 | Perl_hv_ksplit(pTHX_ HV *hv, IV newmax) |
72940dca | 1243 | { |
97aff369 | 1244 | dVAR; |
cbec9347 | 1245 | register XPVHV* xhv = (XPVHV*)SvANY(hv); |
a3b680e6 | 1246 | const I32 oldsize = (I32) xhv->xhv_max+1; /* HvMAX(hv)+1 (sick) */ |
72940dca | 1247 | register I32 newsize; |
1248 | register I32 i; | |
72311751 GS |
1249 | register char *a; |
1250 | register HE **aep; | |
72940dca | 1251 | register HE *entry; |
1252 | register HE **oentry; | |
1253 | ||
7918f24d NC |
1254 | PERL_ARGS_ASSERT_HV_KSPLIT; |
1255 | ||
72940dca | 1256 | newsize = (I32) newmax; /* possible truncation here */ |
1257 | if (newsize != newmax || newmax <= oldsize) | |
1258 | return; | |
1259 | while ((newsize & (1 + ~newsize)) != newsize) { | |
1260 | newsize &= ~(newsize & (1 + ~newsize)); /* get proper power of 2 */ | |
1261 | } | |
1262 | if (newsize < newmax) | |
1263 | newsize *= 2; | |
1264 | if (newsize < newmax) | |
1265 | return; /* overflow detection */ | |
1266 | ||
7b2c381c | 1267 | a = (char *) HvARRAY(hv); |
72940dca | 1268 | if (a) { |
3280af22 | 1269 | PL_nomemok = TRUE; |
8d6dde3e | 1270 | #if defined(STRANGE_MALLOC) || defined(MYMALLOC) |
b79f7545 NC |
1271 | Renew(a, PERL_HV_ARRAY_ALLOC_BYTES(newsize) |
1272 | + (SvOOK(hv) ? sizeof(struct xpvhv_aux) : 0), char); | |
8aacddc1 | 1273 | if (!a) { |
4a33f861 | 1274 | PL_nomemok = FALSE; |
422a93e5 GA |
1275 | return; |
1276 | } | |
b79f7545 | 1277 | if (SvOOK(hv)) { |
7a9b70e9 | 1278 | Copy(&a[oldsize * sizeof(HE*)], &a[newsize * sizeof(HE*)], 1, struct xpvhv_aux); |
b79f7545 | 1279 | } |
72940dca | 1280 | #else |
a02a5408 | 1281 | Newx(a, PERL_HV_ARRAY_ALLOC_BYTES(newsize) |
b79f7545 | 1282 | + (SvOOK(hv) ? sizeof(struct xpvhv_aux) : 0), char); |
8aacddc1 | 1283 | if (!a) { |
3280af22 | 1284 | PL_nomemok = FALSE; |
422a93e5 GA |
1285 | return; |
1286 | } | |
7b2c381c | 1287 | Copy(HvARRAY(hv), a, oldsize * sizeof(HE*), char); |
b79f7545 NC |
1288 | if (SvOOK(hv)) { |
1289 | Copy(HvAUX(hv), &a[newsize * sizeof(HE*)], 1, struct xpvhv_aux); | |
1290 | } | |
fba3b22e | 1291 | if (oldsize >= 64) { |
7b2c381c | 1292 | offer_nice_chunk(HvARRAY(hv), |
b79f7545 NC |
1293 | PERL_HV_ARRAY_ALLOC_BYTES(oldsize) |
1294 | + (SvOOK(hv) ? sizeof(struct xpvhv_aux) : 0)); | |
72940dca | 1295 | } |
1296 | else | |
7b2c381c | 1297 | Safefree(HvARRAY(hv)); |
72940dca | 1298 | #endif |
3280af22 | 1299 | PL_nomemok = FALSE; |
72311751 | 1300 | Zero(&a[oldsize * sizeof(HE*)], (newsize-oldsize) * sizeof(HE*), char); /* zero 2nd half*/ |
72940dca | 1301 | } |
1302 | else { | |
a02a5408 | 1303 | Newxz(a, PERL_HV_ARRAY_ALLOC_BYTES(newsize), char); |
72940dca | 1304 | } |
cbec9347 | 1305 | xhv->xhv_max = --newsize; /* HvMAX(hv) = --newsize */ |
7b2c381c | 1306 | HvARRAY(hv) = (HE **) a; |
cbec9347 | 1307 | if (!xhv->xhv_fill /* !HvFILL(hv) */) /* skip rest if no entries */ |
72940dca | 1308 | return; |
1309 | ||
72311751 GS |
1310 | aep = (HE**)a; |
1311 | for (i=0; i<oldsize; i++,aep++) { | |
1312 | if (!*aep) /* non-existent */ | |
72940dca | 1313 | continue; |
72311751 | 1314 | for (oentry = aep, entry = *aep; entry; entry = *oentry) { |
6136c704 AL |
1315 | register I32 j = (HeHASH(entry) & newsize); |
1316 | ||
1317 | if (j != i) { | |
72940dca | 1318 | j -= i; |
1319 | *oentry = HeNEXT(entry); | |
72311751 | 1320 | if (!(HeNEXT(entry) = aep[j])) |
cbec9347 | 1321 | xhv->xhv_fill++; /* HvFILL(hv)++ */ |
72311751 | 1322 | aep[j] = entry; |
72940dca | 1323 | continue; |
1324 | } | |
1325 | else | |
1326 | oentry = &HeNEXT(entry); | |
1327 | } | |
72311751 | 1328 | if (!*aep) /* everything moved */ |
cbec9347 | 1329 | xhv->xhv_fill--; /* HvFILL(hv)-- */ |
72940dca | 1330 | } |
1331 | } | |
1332 | ||
b3ac6de7 | 1333 | HV * |
864dbfa3 | 1334 | Perl_newHVhv(pTHX_ HV *ohv) |
b3ac6de7 | 1335 | { |
9d4ba2ae | 1336 | HV * const hv = newHV(); |
4beac62f | 1337 | STRLEN hv_max, hv_fill; |
4beac62f AMS |
1338 | |
1339 | if (!ohv || (hv_fill = HvFILL(ohv)) == 0) | |
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(); |
b56ba0bf | 1369 | |
45dea987 | 1370 | HeVAL(ent) = newSVsv(HeVAL(oent)); |
19692e8d | 1371 | HeKEY_hek(ent) |
6e838c70 | 1372 | = shared ? share_hek_flags(key, len, hash, flags) |
19692e8d | 1373 | : save_hek_flags(key, len, hash, flags); |
b56ba0bf AMS |
1374 | if (prev) |
1375 | HeNEXT(prev) = ent; | |
1376 | else | |
1377 | ents[i] = ent; | |
1378 | prev = ent; | |
1379 | HeNEXT(ent) = NULL; | |
1380 | } | |
1381 | } | |
1382 | ||
1383 | HvMAX(hv) = hv_max; | |
1384 | HvFILL(hv) = hv_fill; | |
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); | |
b56ba0bf AMS |
1393 | |
1394 | /* Can we use fewer buckets? (hv_max is always 2^n-1) */ | |
1395 | while (hv_max && hv_max + 1 >= hv_fill * 2) | |
1396 | hv_max = hv_max / 2; | |
1397 | HvMAX(hv) = hv_max; | |
1398 | ||
4a76a316 | 1399 | hv_iterinit(ohv); |
e16e2ff8 | 1400 | while ((entry = hv_iternext_flags(ohv, 0))) { |
04fe65b0 RGS |
1401 | (void)hv_store_flags(hv, HeKEY(entry), HeKLEN(entry), |
1402 | newSVsv(HeVAL(entry)), HeHASH(entry), | |
1403 | HeKFLAGS(entry)); | |
b3ac6de7 | 1404 | } |
bfcb3514 NC |
1405 | HvRITER_set(ohv, riter); |
1406 | HvEITER_set(ohv, eiter); | |
b3ac6de7 | 1407 | } |
1c846c1f | 1408 | |
b3ac6de7 IZ |
1409 | return hv; |
1410 | } | |
1411 | ||
5b9c0671 NC |
1412 | /* A rather specialised version of newHVhv for copying %^H, ensuring all the |
1413 | magic stays on it. */ | |
1414 | HV * | |
1415 | Perl_hv_copy_hints_hv(pTHX_ HV *const ohv) | |
1416 | { | |
1417 | HV * const hv = newHV(); | |
1418 | STRLEN hv_fill; | |
1419 | ||
1420 | if (ohv && (hv_fill = HvFILL(ohv))) { | |
1421 | STRLEN hv_max = HvMAX(ohv); | |
1422 | HE *entry; | |
1423 | const I32 riter = HvRITER_get(ohv); | |
1424 | HE * const eiter = HvEITER_get(ohv); | |
1425 | ||
1426 | while (hv_max && hv_max + 1 >= hv_fill * 2) | |
1427 | hv_max = hv_max / 2; | |
1428 | HvMAX(hv) = hv_max; | |
1429 | ||
1430 | hv_iterinit(ohv); | |
1431 | while ((entry = hv_iternext_flags(ohv, 0))) { | |
1432 | SV *const sv = newSVsv(HeVAL(entry)); | |
1433 | sv_magic(sv, NULL, PERL_MAGIC_hintselem, | |
1434 | (char *)newSVhek (HeKEY_hek(entry)), HEf_SVKEY); | |
04fe65b0 RGS |
1435 | (void)hv_store_flags(hv, HeKEY(entry), HeKLEN(entry), |
1436 | sv, HeHASH(entry), HeKFLAGS(entry)); | |
5b9c0671 NC |
1437 | } |
1438 | HvRITER_set(ohv, riter); | |
1439 | HvEITER_set(ohv, eiter); | |
1440 | } | |
1441 | hv_magic(hv, NULL, PERL_MAGIC_hints); | |
1442 | return hv; | |
1443 | } | |
1444 | ||
79072805 | 1445 | void |
864dbfa3 | 1446 | Perl_hv_free_ent(pTHX_ HV *hv, register HE *entry) |
79072805 | 1447 | { |
97aff369 | 1448 | dVAR; |
16bdeea2 GS |
1449 | SV *val; |
1450 | ||
7918f24d NC |
1451 | PERL_ARGS_ASSERT_HV_FREE_ENT; |
1452 | ||
68dc0745 | 1453 | if (!entry) |
79072805 | 1454 | return; |
16bdeea2 | 1455 | val = HeVAL(entry); |
a5a709ec | 1456 | if (val && isGV(val) && isGV_with_GP(val) && GvCVu(val) && HvNAME_get(hv)) |
0fa56319 | 1457 | mro_method_changed_in(hv); /* deletion of method from stash */ |
16bdeea2 | 1458 | SvREFCNT_dec(val); |
68dc0745 | 1459 | if (HeKLEN(entry) == HEf_SVKEY) { |
1460 | SvREFCNT_dec(HeKEY_sv(entry)); | |
8aacddc1 | 1461 | Safefree(HeKEY_hek(entry)); |
44a8e56a | 1462 | } |
1463 | else if (HvSHAREKEYS(hv)) | |
68dc0745 | 1464 | unshare_hek(HeKEY_hek(entry)); |
fde52b5c | 1465 | else |
68dc0745 | 1466 | Safefree(HeKEY_hek(entry)); |
d33b2eba | 1467 | del_HE(entry); |
79072805 LW |
1468 | } |
1469 | ||
1470 | void | |
864dbfa3 | 1471 | Perl_hv_delayfree_ent(pTHX_ HV *hv, register HE *entry) |
79072805 | 1472 | { |
97aff369 | 1473 | dVAR; |
7918f24d NC |
1474 | |
1475 | PERL_ARGS_ASSERT_HV_DELAYFREE_ENT; | |
1476 | ||
68dc0745 | 1477 | if (!entry) |
79072805 | 1478 | return; |
bc4947fc NC |
1479 | /* SvREFCNT_inc to counter the SvREFCNT_dec in hv_free_ent */ |
1480 | sv_2mortal(SvREFCNT_inc(HeVAL(entry))); /* free between statements */ | |
68dc0745 | 1481 | if (HeKLEN(entry) == HEf_SVKEY) { |
bc4947fc | 1482 | sv_2mortal(SvREFCNT_inc(HeKEY_sv(entry))); |
44a8e56a | 1483 | } |
bc4947fc | 1484 | hv_free_ent(hv, entry); |
79072805 LW |
1485 | } |
1486 | ||
954c1994 GS |
1487 | /* |
1488 | =for apidoc hv_clear | |
1489 | ||
1490 | Clears a hash, making it empty. | |
1491 | ||
1492 | =cut | |
1493 | */ | |
1494 | ||
79072805 | 1495 | void |
864dbfa3 | 1496 | Perl_hv_clear(pTHX_ HV *hv) |
79072805 | 1497 | { |
27da23d5 | 1498 | dVAR; |
cbec9347 | 1499 | register XPVHV* xhv; |
79072805 LW |
1500 | if (!hv) |
1501 | return; | |
49293501 | 1502 | |
ecae49c0 NC |
1503 | DEBUG_A(Perl_hv_assert(aTHX_ hv)); |
1504 | ||
34c3c4e3 DM |
1505 | xhv = (XPVHV*)SvANY(hv); |
1506 | ||
7b2c381c | 1507 | if (SvREADONLY(hv) && HvARRAY(hv) != NULL) { |
34c3c4e3 | 1508 | /* restricted hash: convert all keys to placeholders */ |
b464bac0 AL |
1509 | STRLEN i; |
1510 | for (i = 0; i <= xhv->xhv_max; i++) { | |
7b2c381c | 1511 | HE *entry = (HvARRAY(hv))[i]; |
3a676441 JH |
1512 | for (; entry; entry = HeNEXT(entry)) { |
1513 | /* not already placeholder */ | |
7996736c | 1514 | if (HeVAL(entry) != &PL_sv_placeholder) { |
3a676441 | 1515 | if (HeVAL(entry) && SvREADONLY(HeVAL(entry))) { |
6136c704 | 1516 | SV* const keysv = hv_iterkeysv(entry); |
3a676441 | 1517 | Perl_croak(aTHX_ |
95b63a38 JH |
1518 | "Attempt to delete readonly key '%"SVf"' from a restricted hash", |
1519 | (void*)keysv); | |
3a676441 JH |
1520 | } |
1521 | SvREFCNT_dec(HeVAL(entry)); | |
7996736c | 1522 | HeVAL(entry) = &PL_sv_placeholder; |
ca732855 | 1523 | HvPLACEHOLDERS(hv)++; |
3a676441 | 1524 | } |
34c3c4e3 DM |
1525 | } |
1526 | } | |
df8c6964 | 1527 | goto reset; |
49293501 MS |
1528 | } |
1529 | ||
463ee0b2 | 1530 | hfreeentries(hv); |
ca732855 | 1531 | HvPLACEHOLDERS_set(hv, 0); |
7b2c381c | 1532 | if (HvARRAY(hv)) |
41f62432 | 1533 | Zero(HvARRAY(hv), xhv->xhv_max+1 /* HvMAX(hv)+1 */, HE*); |
a0d0e21e LW |
1534 | |
1535 | if (SvRMAGICAL(hv)) | |
ad64d0ec | 1536 | mg_clear(MUTABLE_SV(hv)); |
574c8022 | 1537 | |
19692e8d | 1538 | HvHASKFLAGS_off(hv); |
bb443f97 | 1539 | HvREHASH_off(hv); |
df8c6964 | 1540 | reset: |
b79f7545 | 1541 | if (SvOOK(hv)) { |
dd69841b BB |
1542 | if(HvNAME_get(hv)) |
1543 | mro_isa_changed_in(hv); | |
bfcb3514 NC |
1544 | HvEITER_set(hv, NULL); |
1545 | } | |
79072805 LW |
1546 | } |
1547 | ||
3540d4ce AB |
1548 | /* |
1549 | =for apidoc hv_clear_placeholders | |
1550 | ||
1551 | Clears any placeholders from a hash. If a restricted hash has any of its keys | |
1552 | marked as readonly and the key is subsequently deleted, the key is not actually | |
1553 | deleted but is marked by assigning it a value of &PL_sv_placeholder. This tags | |
1554 | it so it will be ignored by future operations such as iterating over the hash, | |
4cdaeff7 | 1555 | but will still allow the hash to have a value reassigned to the key at some |
3540d4ce AB |
1556 | future point. This function clears any such placeholder keys from the hash. |
1557 | See Hash::Util::lock_keys() for an example of its use. | |
1558 | ||
1559 | =cut | |
1560 | */ | |
1561 | ||
1562 | void | |
1563 | Perl_hv_clear_placeholders(pTHX_ HV *hv) | |
1564 | { | |
27da23d5 | 1565 | dVAR; |
b3ca2e83 NC |
1566 | const U32 items = (U32)HvPLACEHOLDERS_get(hv); |
1567 | ||
7918f24d NC |
1568 | PERL_ARGS_ASSERT_HV_CLEAR_PLACEHOLDERS; |
1569 | ||
b3ca2e83 NC |
1570 | if (items) |
1571 | clear_placeholders(hv, items); | |
1572 | } | |
1573 | ||
1574 | static void | |
1575 | S_clear_placeholders(pTHX_ HV *hv, U32 items) | |
1576 | { | |
1577 | dVAR; | |
b464bac0 | 1578 | I32 i; |
d3677389 | 1579 | |
7918f24d NC |
1580 | PERL_ARGS_ASSERT_CLEAR_PLACEHOLDERS; |
1581 | ||
d3677389 NC |
1582 | if (items == 0) |
1583 | return; | |
1584 | ||
b464bac0 | 1585 | i = HvMAX(hv); |
d3677389 NC |
1586 | do { |
1587 | /* Loop down the linked list heads */ | |
6136c704 | 1588 | bool first = TRUE; |
d3677389 | 1589 | HE **oentry = &(HvARRAY(hv))[i]; |
cf6db12b | 1590 | HE *entry; |
d3677389 | 1591 | |
cf6db12b | 1592 | while ((entry = *oentry)) { |
d3677389 NC |
1593 | if (HeVAL(entry) == &PL_sv_placeholder) { |
1594 | *oentry = HeNEXT(entry); | |
1595 | if (first && !*oentry) | |
1596 | HvFILL(hv)--; /* This linked list is now empty. */ | |
2e58978b | 1597 | if (entry == HvEITER_get(hv)) |
d3677389 NC |
1598 | HvLAZYDEL_on(hv); |
1599 | else | |
1600 | hv_free_ent(hv, entry); | |
1601 | ||
1602 | if (--items == 0) { | |
1603 | /* Finished. */ | |
5d88ecd7 | 1604 | HvTOTALKEYS(hv) -= (IV)HvPLACEHOLDERS_get(hv); |
d3677389 NC |
1605 | if (HvKEYS(hv) == 0) |
1606 | HvHASKFLAGS_off(hv); | |
5d88ecd7 | 1607 | HvPLACEHOLDERS_set(hv, 0); |
d3677389 NC |
1608 | return; |
1609 | } | |
213ce8b3 NC |
1610 | } else { |
1611 | oentry = &HeNEXT(entry); | |
6136c704 | 1612 | first = FALSE; |
d3677389 NC |
1613 | } |
1614 | } | |
1615 | } while (--i >= 0); | |
1616 | /* You can't get here, hence assertion should always fail. */ | |
1617 | assert (items == 0); | |
1618 | assert (0); | |
3540d4ce AB |
1619 | } |
1620 | ||
76e3520e | 1621 | STATIC void |
cea2e8a9 | 1622 | S_hfreeentries(pTHX_ HV *hv) |
79072805 | 1623 | { |
23976bdd | 1624 | /* This is the array that we're going to restore */ |
fd7de8a8 | 1625 | HE **const orig_array = HvARRAY(hv); |
23976bdd NC |
1626 | HEK *name; |
1627 | int attempts = 100; | |
3abe233e | 1628 | |
7918f24d NC |
1629 | PERL_ARGS_ASSERT_HFREEENTRIES; |
1630 | ||
fd7de8a8 | 1631 | if (!orig_array) |
79072805 | 1632 | return; |
a0d0e21e | 1633 | |
23976bdd NC |
1634 | if (SvOOK(hv)) { |
1635 | /* If the hash is actually a symbol table with a name, look after the | |
1636 | name. */ | |
1637 | struct xpvhv_aux *iter = HvAUX(hv); | |
1638 | ||
1639 | name = iter->xhv_name; | |
1640 | iter->xhv_name = NULL; | |
1641 | } else { | |
1642 | name = NULL; | |
1643 | } | |
1644 | ||
23976bdd NC |
1645 | /* orig_array remains unchanged throughout the loop. If after freeing all |
1646 | the entries it turns out that one of the little blighters has triggered | |
1647 | an action that has caused HvARRAY to be re-allocated, then we set | |
1648 | array to the new HvARRAY, and try again. */ | |
1649 | ||
1650 | while (1) { | |
1651 | /* This is the one we're going to try to empty. First time round | |
1652 | it's the original array. (Hopefully there will only be 1 time | |
1653 | round) */ | |
6136c704 | 1654 | HE ** const array = HvARRAY(hv); |
7440661e | 1655 | I32 i = HvMAX(hv); |
23976bdd NC |
1656 | |
1657 | /* Because we have taken xhv_name out, the only allocated pointer | |
1658 | in the aux structure that might exist is the backreference array. | |
1659 | */ | |
1660 | ||
1661 | if (SvOOK(hv)) { | |
7440661e | 1662 | HE *entry; |
e1a479c5 | 1663 | struct mro_meta *meta; |
23976bdd NC |
1664 | struct xpvhv_aux *iter = HvAUX(hv); |
1665 | /* If there are weak references to this HV, we need to avoid | |
1666 | freeing them up here. In particular we need to keep the AV | |
1667 | visible as what we're deleting might well have weak references | |
1668 | back to this HV, so the for loop below may well trigger | |
1669 | the removal of backreferences from this array. */ | |
1670 | ||
1671 | if (iter->xhv_backreferences) { | |
1672 | /* So donate them to regular backref magic to keep them safe. | |
1673 | The sv_magic will increase the reference count of the AV, | |
1674 | so we need to drop it first. */ | |
5b285ea4 | 1675 | SvREFCNT_dec(iter->xhv_backreferences); |
23976bdd NC |
1676 | if (AvFILLp(iter->xhv_backreferences) == -1) { |
1677 | /* Turns out that the array is empty. Just free it. */ | |
1678 | SvREFCNT_dec(iter->xhv_backreferences); | |
1b8791d1 | 1679 | |
23976bdd | 1680 | } else { |
ad64d0ec NC |
1681 | sv_magic(MUTABLE_SV(hv), |
1682 | MUTABLE_SV(iter->xhv_backreferences), | |
23976bdd NC |
1683 | PERL_MAGIC_backref, NULL, 0); |
1684 | } | |
1685 | iter->xhv_backreferences = NULL; | |
5b285ea4 | 1686 | } |
86f55936 | 1687 | |
23976bdd NC |
1688 | entry = iter->xhv_eiter; /* HvEITER(hv) */ |
1689 | if (entry && HvLAZYDEL(hv)) { /* was deleted earlier? */ | |
1690 | HvLAZYDEL_off(hv); | |
1691 | hv_free_ent(hv, entry); | |
1692 | } | |
1693 | iter->xhv_riter = -1; /* HvRITER(hv) = -1 */ | |
4608196e | 1694 | iter->xhv_eiter = NULL; /* HvEITER(hv) = NULL */ |
b79f7545 | 1695 | |
e1a479c5 BB |
1696 | if((meta = iter->xhv_mro_meta)) { |
1697 | if(meta->mro_linear_dfs) SvREFCNT_dec(meta->mro_linear_dfs); | |
1698 | if(meta->mro_linear_c3) SvREFCNT_dec(meta->mro_linear_c3); | |
e1a479c5 | 1699 | if(meta->mro_nextmethod) SvREFCNT_dec(meta->mro_nextmethod); |
a49ba3fc | 1700 | SvREFCNT_dec(meta->isa); |
e1a479c5 BB |
1701 | Safefree(meta); |
1702 | iter->xhv_mro_meta = NULL; | |
1703 | } | |
1704 | ||
23976bdd | 1705 | /* There are now no allocated pointers in the aux structure. */ |
2f86008e | 1706 | |
23976bdd NC |
1707 | SvFLAGS(hv) &= ~SVf_OOK; /* Goodbye, aux structure. */ |
1708 | /* What aux structure? */ | |
a0d0e21e | 1709 | } |
bfcb3514 | 1710 | |
23976bdd NC |
1711 | /* make everyone else think the array is empty, so that the destructors |
1712 | * called for freed entries can't recusively mess with us */ | |
1713 | HvARRAY(hv) = NULL; | |
1714 | HvFILL(hv) = 0; | |
1715 | ((XPVHV*) SvANY(hv))->xhv_keys = 0; | |
1716 | ||
7440661e NC |
1717 | |
1718 | do { | |
1719 | /* Loop down the linked list heads */ | |
1720 | HE *entry = array[i]; | |
1721 | ||
1722 | while (entry) { | |
23976bdd NC |
1723 | register HE * const oentry = entry; |
1724 | entry = HeNEXT(entry); | |
1725 | hv_free_ent(hv, oentry); | |
1726 | } | |
7440661e | 1727 | } while (--i >= 0); |
b79f7545 | 1728 | |
23976bdd NC |
1729 | /* As there are no allocated pointers in the aux structure, it's now |
1730 | safe to free the array we just cleaned up, if it's not the one we're | |
1731 | going to put back. */ | |
1732 | if (array != orig_array) { | |
1733 | Safefree(array); | |
1734 | } | |
b79f7545 | 1735 | |
23976bdd NC |
1736 | if (!HvARRAY(hv)) { |
1737 | /* Good. No-one added anything this time round. */ | |
1738 | break; | |
bfcb3514 | 1739 | } |
b79f7545 | 1740 | |
23976bdd NC |
1741 | if (SvOOK(hv)) { |
1742 | /* Someone attempted to iterate or set the hash name while we had | |
1743 | the array set to 0. We'll catch backferences on the next time | |
1744 | round the while loop. */ | |
1745 | assert(HvARRAY(hv)); | |
1b8791d1 | 1746 | |
23976bdd NC |
1747 | if (HvAUX(hv)->xhv_name) { |
1748 | unshare_hek_or_pvn(HvAUX(hv)->xhv_name, 0, 0, 0); | |
1749 | } | |
1750 | } | |
1751 | ||
1752 | if (--attempts == 0) { | |
1753 | Perl_die(aTHX_ "panic: hfreeentries failed to free hash - something is repeatedly re-creating entries"); | |
1754 | } | |
6136c704 | 1755 | } |
23976bdd NC |
1756 | |
1757 | HvARRAY(hv) = orig_array; | |
1758 | ||
1759 | /* If the hash was actually a symbol table, put the name back. */ | |
1760 | if (name) { | |
1761 | /* We have restored the original array. If name is non-NULL, then | |
1762 | the original array had an aux structure at the end. So this is | |
1763 | valid: */ | |
1764 | SvFLAGS(hv) |= SVf_OOK; | |
1765 | HvAUX(hv)->xhv_name = name; | |
1b8791d1 | 1766 | } |
79072805 LW |
1767 | } |
1768 | ||
954c1994 GS |
1769 | /* |
1770 | =for apidoc hv_undef | |
1771 | ||
1772 | Undefines the hash. | |
1773 | ||
1774 | =cut | |
1775 | */ | |
1776 | ||
79072805 | 1777 | void |
864dbfa3 | 1778 | Perl_hv_undef(pTHX_ HV *hv) |
79072805 | 1779 | { |
97aff369 | 1780 | dVAR; |
cbec9347 | 1781 | register XPVHV* xhv; |
bfcb3514 | 1782 | const char *name; |
86f55936 | 1783 | |
79072805 LW |
1784 | if (!hv) |
1785 | return; | |
ecae49c0 | 1786 | DEBUG_A(Perl_hv_assert(aTHX_ hv)); |
cbec9347 | 1787 | xhv = (XPVHV*)SvANY(hv); |
dd69841b | 1788 | |
0fa56319 | 1789 | if ((name = HvNAME_get(hv)) && !PL_dirty) |
dd69841b BB |
1790 | mro_isa_changed_in(hv); |
1791 | ||
463ee0b2 | 1792 | hfreeentries(hv); |
dd69841b | 1793 | if (name) { |
04fe65b0 RGS |
1794 | if (PL_stashcache) |
1795 | (void)hv_delete(PL_stashcache, name, HvNAMELEN_get(hv), G_DISCARD); | |
bd61b366 | 1796 | hv_name_set(hv, NULL, 0, 0); |
85e6fe83 | 1797 | } |
b79f7545 NC |
1798 | SvFLAGS(hv) &= ~SVf_OOK; |
1799 | Safefree(HvARRAY(hv)); | |
cbec9347 | 1800 | xhv->xhv_max = 7; /* HvMAX(hv) = 7 (it's a normal hash) */ |
7b2c381c | 1801 | HvARRAY(hv) = 0; |
ca732855 | 1802 | HvPLACEHOLDERS_set(hv, 0); |
a0d0e21e LW |
1803 | |
1804 | if (SvRMAGICAL(hv)) | |
ad64d0ec | 1805 | mg_clear(MUTABLE_SV(hv)); |
79072805 LW |
1806 | } |
1807 | ||
b464bac0 | 1808 | static struct xpvhv_aux* |
5f66b61c | 1809 | S_hv_auxinit(HV *hv) { |
bfcb3514 | 1810 | struct xpvhv_aux *iter; |
b79f7545 | 1811 | char *array; |
bfcb3514 | 1812 | |
7918f24d NC |
1813 | PERL_ARGS_ASSERT_HV_AUXINIT; |
1814 | ||
b79f7545 | 1815 | if (!HvARRAY(hv)) { |
a02a5408 | 1816 | Newxz(array, PERL_HV_ARRAY_ALLOC_BYTES(HvMAX(hv) + 1) |
b79f7545 NC |
1817 | + sizeof(struct xpvhv_aux), char); |
1818 | } else { | |
1819 | array = (char *) HvARRAY(hv); | |
1820 | Renew(array, PERL_HV_ARRAY_ALLOC_BYTES(HvMAX(hv) + 1) | |
1821 | + sizeof(struct xpvhv_aux), char); | |
1822 | } | |
1823 | HvARRAY(hv) = (HE**) array; | |
1824 | /* SvOOK_on(hv) attacks the IV flags. */ | |
1825 | SvFLAGS(hv) |= SVf_OOK; | |
1826 | iter = HvAUX(hv); | |
bfcb3514 NC |
1827 | |
1828 | iter->xhv_riter = -1; /* HvRITER(hv) = -1 */ | |
4608196e | 1829 | iter->xhv_eiter = NULL; /* HvEITER(hv) = NULL */ |
bfcb3514 | 1830 | iter->xhv_name = 0; |
86f55936 | 1831 | iter->xhv_backreferences = 0; |
e1a479c5 | 1832 | iter->xhv_mro_meta = NULL; |
bfcb3514 NC |
1833 | return iter; |
1834 | } | |
1835 | ||
954c1994 GS |
1836 | /* |
1837 | =for apidoc hv_iterinit | |
1838 | ||
1839 | Prepares a starting point to traverse a hash table. Returns the number of | |
1840 | keys in the hash (i.e. the same as C<HvKEYS(tb)>). The return value is | |
1c846c1f | 1841 | currently only meaningful for hashes without tie magic. |
954c1994 GS |
1842 | |
1843 | NOTE: Before version 5.004_65, C<hv_iterinit> used to return the number of | |
1844 | hash buckets that happen to be in use. If you still need that esoteric | |
1845 | value, you can get it through the macro C<HvFILL(tb)>. | |
1846 | ||
e16e2ff8 | 1847 | |
954c1994 GS |
1848 | =cut |
1849 | */ | |
1850 | ||
79072805 | 1851 | I32 |
864dbfa3 | 1852 | Perl_hv_iterinit(pTHX_ HV *hv) |
79072805 | 1853 | { |
7918f24d NC |
1854 | PERL_ARGS_ASSERT_HV_ITERINIT; |
1855 | ||
1856 | /* FIXME: Are we not NULL, or do we croak? Place bets now! */ | |
1857 | ||
aa689395 | 1858 | if (!hv) |
cea2e8a9 | 1859 | Perl_croak(aTHX_ "Bad hash"); |
bfcb3514 | 1860 | |
b79f7545 | 1861 | if (SvOOK(hv)) { |
6136c704 | 1862 | struct xpvhv_aux * const iter = HvAUX(hv); |
0bd48802 | 1863 | HE * const entry = iter->xhv_eiter; /* HvEITER(hv) */ |
bfcb3514 NC |
1864 | if (entry && HvLAZYDEL(hv)) { /* was deleted earlier? */ |
1865 | HvLAZYDEL_off(hv); | |
1866 | hv_free_ent(hv, entry); | |
1867 | } | |
1868 | iter->xhv_riter = -1; /* HvRITER(hv) = -1 */ | |
4608196e | 1869 | iter->xhv_eiter = NULL; /* HvEITER(hv) = NULL */ |
bfcb3514 | 1870 | } else { |
6136c704 | 1871 | hv_auxinit(hv); |
72940dca | 1872 | } |
44a2ac75 | 1873 | |
cbec9347 | 1874 | /* used to be xhv->xhv_fill before 5.004_65 */ |
5d88ecd7 | 1875 | return HvTOTALKEYS(hv); |
79072805 | 1876 | } |
bfcb3514 NC |
1877 | |
1878 | I32 * | |
1879 | Perl_hv_riter_p(pTHX_ HV *hv) { | |
1880 | struct xpvhv_aux *iter; | |
1881 | ||
7918f24d NC |
1882 | PERL_ARGS_ASSERT_HV_RITER_P; |
1883 | ||
bfcb3514 NC |
1884 | if (!hv) |
1885 | Perl_croak(aTHX_ "Bad hash"); | |
1886 | ||
6136c704 | 1887 | iter = SvOOK(hv) ? HvAUX(hv) : hv_auxinit(hv); |
bfcb3514 NC |
1888 | return &(iter->xhv_riter); |
1889 | } | |
1890 | ||
1891 | HE ** | |
1892 | Perl_hv_eiter_p(pTHX_ HV *hv) { | |
1893 | struct xpvhv_aux *iter; | |
1894 | ||
7918f24d NC |
1895 | PERL_ARGS_ASSERT_HV_EITER_P; |
1896 | ||
bfcb3514 NC |
1897 | if (!hv) |
1898 | Perl_croak(aTHX_ "Bad hash"); | |
1899 | ||
6136c704 | 1900 | iter = SvOOK(hv) ? HvAUX(hv) : hv_auxinit(hv); |
bfcb3514 NC |
1901 | return &(iter->xhv_eiter); |
1902 | } | |
1903 | ||
1904 | void | |
1905 | Perl_hv_riter_set(pTHX_ HV *hv, I32 riter) { | |
1906 | struct xpvhv_aux *iter; | |
1907 | ||
7918f24d NC |
1908 | PERL_ARGS_ASSERT_HV_RITER_SET; |
1909 | ||
bfcb3514 NC |
1910 | if (!hv) |
1911 | Perl_croak(aTHX_ "Bad hash"); | |
1912 | ||
b79f7545 NC |
1913 | if (SvOOK(hv)) { |
1914 | iter = HvAUX(hv); | |
1915 | } else { | |
bfcb3514 NC |
1916 | if (riter == -1) |
1917 | return; | |
1918 | ||
6136c704 | 1919 | iter = hv_auxinit(hv); |
bfcb3514 NC |
1920 | } |
1921 | iter->xhv_riter = riter; | |
1922 | } | |
1923 | ||
1924 | void | |
1925 | Perl_hv_eiter_set(pTHX_ HV *hv, HE *eiter) { | |
1926 | struct xpvhv_aux *iter; | |
1927 | ||
7918f24d NC |
1928 | PERL_ARGS_ASSERT_HV_EITER_SET; |
1929 | ||
bfcb3514 NC |
1930 | if (!hv) |
1931 | Perl_croak(aTHX_ "Bad hash"); | |
1932 | ||
b79f7545 NC |
1933 | if (SvOOK(hv)) { |
1934 | iter = HvAUX(hv); | |
1935 | } else { | |
bfcb3514 NC |
1936 | /* 0 is the default so don't go malloc()ing a new structure just to |
1937 | hold 0. */ | |
1938 | if (!eiter) | |
1939 | return; | |
1940 | ||
6136c704 | 1941 | iter = hv_auxinit(hv); |
bfcb3514 NC |
1942 | } |
1943 | iter->xhv_eiter = eiter; | |
1944 | } | |
1945 | ||
bfcb3514 | 1946 | void |
4164be69 | 1947 | Perl_hv_name_set(pTHX_ HV *hv, const char *name, U32 len, U32 flags) |
bfcb3514 | 1948 | { |
97aff369 | 1949 | dVAR; |
b79f7545 | 1950 | struct xpvhv_aux *iter; |
7423f6db | 1951 | U32 hash; |
46c461b5 | 1952 | |
7918f24d | 1953 | PERL_ARGS_ASSERT_HV_NAME_SET; |
46c461b5 | 1954 | PERL_UNUSED_ARG(flags); |
bfcb3514 | 1955 | |
4164be69 NC |
1956 | if (len > I32_MAX) |
1957 | Perl_croak(aTHX_ "panic: hv name too long (%"UVuf")", (UV) len); | |
1958 | ||
b79f7545 NC |
1959 | if (SvOOK(hv)) { |
1960 | iter = HvAUX(hv); | |
7423f6db NC |
1961 | if (iter->xhv_name) { |
1962 | unshare_hek_or_pvn(iter->xhv_name, 0, 0, 0); | |
1963 | } | |
16580ff5 | 1964 | } else { |
bfcb3514 NC |
1965 | if (name == 0) |
1966 | return; | |
1967 | ||
6136c704 | 1968 | iter = hv_auxinit(hv); |
bfcb3514 | 1969 | } |
7423f6db | 1970 | PERL_HASH(hash, name, len); |
adf4e37a | 1971 | iter->xhv_name = name ? share_hek(name, len, hash) : NULL; |
bfcb3514 NC |
1972 | } |
1973 | ||
86f55936 NC |
1974 | AV ** |
1975 | Perl_hv_backreferences_p(pTHX_ HV *hv) { | |
6136c704 | 1976 | struct xpvhv_aux * const iter = SvOOK(hv) ? HvAUX(hv) : hv_auxinit(hv); |
7918f24d NC |
1977 | |
1978 | PERL_ARGS_ASSERT_HV_BACKREFERENCES_P; | |
96a5add6 | 1979 | PERL_UNUSED_CONTEXT; |
7918f24d | 1980 | |
86f55936 NC |
1981 | return &(iter->xhv_backreferences); |
1982 | } | |
1983 | ||
1984 | void | |
1985 | Perl_hv_kill_backrefs(pTHX_ HV *hv) { | |
1986 | AV *av; | |
1987 | ||
7918f24d NC |
1988 | PERL_ARGS_ASSERT_HV_KILL_BACKREFS; |
1989 | ||
86f55936 NC |
1990 | if (!SvOOK(hv)) |
1991 | return; | |
1992 | ||
1993 | av = HvAUX(hv)->xhv_backreferences; | |
1994 | ||
1995 | if (av) { | |
1996 | HvAUX(hv)->xhv_backreferences = 0; | |
ad64d0ec | 1997 | Perl_sv_kill_backrefs(aTHX_ MUTABLE_SV(hv), av); |
b17f5ab7 | 1998 | SvREFCNT_dec(av); |
86f55936 NC |
1999 | } |
2000 | } | |
2001 | ||
954c1994 | 2002 | /* |
7a7b9979 NC |
2003 | hv_iternext is implemented as a macro in hv.h |
2004 | ||
954c1994 GS |
2005 | =for apidoc hv_iternext |
2006 | ||
2007 | Returns entries from a hash iterator. See C<hv_iterinit>. | |
2008 | ||
fe7bca90 NC |
2009 | You may call C<hv_delete> or C<hv_delete_ent> on the hash entry that the |
2010 | iterator currently points to, without losing your place or invalidating your | |
2011 | iterator. Note that in this case the current entry is deleted from the hash | |
2012 | with your iterator holding the last reference to it. Your iterator is flagged | |
2013 | to free the entry on the next call to C<hv_iternext>, so you must not discard | |
2014 | your iterator immediately else the entry will leak - call C<hv_iternext> to | |
2015 | trigger the resource deallocation. | |
2016 | ||
fe7bca90 NC |
2017 | =for apidoc hv_iternext_flags |
2018 | ||
2019 | Returns entries from a hash iterator. See C<hv_iterinit> and C<hv_iternext>. | |
2020 | The C<flags> value will normally be zero; if HV_ITERNEXT_WANTPLACEHOLDERS is | |
2021 | set the placeholders keys (for restricted hashes) will be returned in addition | |
2022 | to normal keys. By default placeholders are automatically skipped over. | |
7996736c MHM |
2023 | Currently a placeholder is implemented with a value that is |
2024 | C<&Perl_sv_placeholder>. Note that the implementation of placeholders and | |
fe7bca90 NC |
2025 | restricted hashes may change, and the implementation currently is |
2026 | insufficiently abstracted for any change to be tidy. | |
e16e2ff8 | 2027 | |
fe7bca90 | 2028 | =cut |
e16e2ff8 NC |
2029 | */ |
2030 | ||
2031 | HE * | |
2032 | Perl_hv_iternext_flags(pTHX_ HV *hv, I32 flags) | |
2033 | { | |
27da23d5 | 2034 | dVAR; |
cbec9347 | 2035 | register XPVHV* xhv; |
79072805 | 2036 | register HE *entry; |
a0d0e21e | 2037 | HE *oldentry; |
463ee0b2 | 2038 | MAGIC* mg; |
bfcb3514 | 2039 | struct xpvhv_aux *iter; |
79072805 | 2040 | |
7918f24d NC |
2041 | PERL_ARGS_ASSERT_HV_ITERNEXT_FLAGS; |
2042 | ||
79072805 | 2043 | if (!hv) |
cea2e8a9 | 2044 | Perl_croak(aTHX_ "Bad hash"); |
81714fb9 | 2045 | |
cbec9347 | 2046 | xhv = (XPVHV*)SvANY(hv); |
bfcb3514 | 2047 | |
b79f7545 | 2048 | if (!SvOOK(hv)) { |
bfcb3514 NC |
2049 | /* Too many things (well, pp_each at least) merrily assume that you can |
2050 | call iv_iternext without calling hv_iterinit, so we'll have to deal | |
2051 | with it. */ | |
2052 | hv_iterinit(hv); | |
bfcb3514 | 2053 | } |
b79f7545 | 2054 | iter = HvAUX(hv); |
bfcb3514 NC |
2055 | |
2056 | oldentry = entry = iter->xhv_eiter; /* HvEITER(hv) */ | |
e62cc96a | 2057 | if (SvMAGICAL(hv) && SvRMAGICAL(hv)) { |
ad64d0ec | 2058 | if ( ( mg = mg_find((const SV *)hv, PERL_MAGIC_tied) ) ) { |
e62cc96a YO |
2059 | SV * const key = sv_newmortal(); |
2060 | if (entry) { | |
2061 | sv_setsv(key, HeSVKEY_force(entry)); | |
2062 | SvREFCNT_dec(HeSVKEY(entry)); /* get rid of previous key */ | |
2063 | } | |
2064 | else { | |
2065 | char *k; | |
2066 | HEK *hek; | |
2067 | ||
2068 | /* one HE per MAGICAL hash */ | |
2069 | iter->xhv_eiter = entry = new_HE(); /* HvEITER(hv) = new_HE() */ | |
2070 | Zero(entry, 1, HE); | |
ad64d0ec | 2071 | Newxz(k, HEK_BASESIZE + sizeof(const SV *), char); |
e62cc96a YO |
2072 | hek = (HEK*)k; |
2073 | HeKEY_hek(entry) = hek; | |
2074 | HeKLEN(entry) = HEf_SVKEY; | |
2075 | } | |
ad64d0ec | 2076 | magic_nextpack(MUTABLE_SV(hv),mg,key); |
e62cc96a YO |
2077 | if (SvOK(key)) { |
2078 | /* force key to stay around until next time */ | |
2079 | HeSVKEY_set(entry, SvREFCNT_inc_simple_NN(key)); | |
2080 | return entry; /* beware, hent_val is not set */ | |
2081 | } | |
2082 | if (HeVAL(entry)) | |
2083 | SvREFCNT_dec(HeVAL(entry)); | |
2084 | Safefree(HeKEY_hek(entry)); | |
2085 | del_HE(entry); | |
2086 | iter->xhv_eiter = NULL; /* HvEITER(hv) = NULL */ | |
2087 | return NULL; | |
81714fb9 | 2088 | } |
79072805 | 2089 | } |
7ee146b1 | 2090 | #if defined(DYNAMIC_ENV_FETCH) && !defined(__riscos__) /* set up %ENV for iteration */ |
ad64d0ec NC |
2091 | if (!entry && SvRMAGICAL((const SV *)hv) |
2092 | && mg_find((const SV *)hv, PERL_MAGIC_env)) { | |
f675dbe5 | 2093 | prime_env_iter(); |
03026e68 JM |
2094 | #ifdef VMS |
2095 | /* The prime_env_iter() on VMS just loaded up new hash values | |
2096 | * so the iteration count needs to be reset back to the beginning | |
2097 | */ | |
2098 | hv_iterinit(hv); | |
2099 | iter = HvAUX(hv); | |
2100 | oldentry = entry = iter->xhv_eiter; /* HvEITER(hv) */ | |
2101 | #endif | |
2102 | } | |
f675dbe5 | 2103 | #endif |
463ee0b2 | 2104 | |
b79f7545 NC |
2105 | /* hv_iterint now ensures this. */ |
2106 | assert (HvARRAY(hv)); | |
2107 | ||
015a5f36 | 2108 | /* At start of hash, entry is NULL. */ |
fde52b5c | 2109 | if (entry) |
8aacddc1 | 2110 | { |
fde52b5c | 2111 | entry = HeNEXT(entry); |
e16e2ff8 NC |
2112 | if (!(flags & HV_ITERNEXT_WANTPLACEHOLDERS)) { |
2113 | /* | |
2114 | * Skip past any placeholders -- don't want to include them in | |
2115 | * any iteration. | |
2116 | */ | |
7996736c | 2117 | while (entry && HeVAL(entry) == &PL_sv_placeholder) { |
e16e2ff8 NC |
2118 | entry = HeNEXT(entry); |
2119 | } | |
8aacddc1 NIS |
2120 | } |
2121 | } | |
fde52b5c | 2122 | while (!entry) { |
015a5f36 NC |
2123 | /* OK. Come to the end of the current list. Grab the next one. */ |
2124 | ||
bfcb3514 NC |
2125 | iter->xhv_riter++; /* HvRITER(hv)++ */ |
2126 | if (iter->xhv_riter > (I32)xhv->xhv_max /* HvRITER(hv) > HvMAX(hv) */) { | |
015a5f36 | 2127 | /* There is no next one. End of the hash. */ |
bfcb3514 | 2128 | iter->xhv_riter = -1; /* HvRITER(hv) = -1 */ |
fde52b5c | 2129 | break; |
79072805 | 2130 | } |
7b2c381c | 2131 | entry = (HvARRAY(hv))[iter->xhv_riter]; |
8aacddc1 | 2132 | |
e16e2ff8 | 2133 | if (!(flags & HV_ITERNEXT_WANTPLACEHOLDERS)) { |
015a5f36 NC |
2134 | /* If we have an entry, but it's a placeholder, don't count it. |
2135 | Try the next. */ | |
7996736c | 2136 | while (entry && HeVAL(entry) == &PL_sv_placeholder) |
015a5f36 NC |
2137 | entry = HeNEXT(entry); |
2138 | } | |
2139 | /* Will loop again if this linked list starts NULL | |
2140 | (for HV_ITERNEXT_WANTPLACEHOLDERS) | |
2141 | or if we run through it and find only placeholders. */ | |
fde52b5c | 2142 | } |
79072805 | 2143 | |
72940dca | 2144 | if (oldentry && HvLAZYDEL(hv)) { /* was deleted earlier? */ |
2145 | HvLAZYDEL_off(hv); | |
68dc0745 | 2146 | hv_free_ent(hv, oldentry); |
72940dca | 2147 | } |
a0d0e21e | 2148 | |
fdcd69b6 | 2149 | /*if (HvREHASH(hv) && entry && !HeKREHASH(entry)) |
6c9570dc | 2150 | PerlIO_printf(PerlIO_stderr(), "Awooga %p %p\n", (void*)hv, (void*)entry);*/ |
fdcd69b6 | 2151 | |
bfcb3514 | 2152 | iter->xhv_eiter = entry; /* HvEITER(hv) = entry */ |
79072805 LW |
2153 | return entry; |
2154 | } | |
2155 | ||
954c1994 GS |
2156 | /* |
2157 | =for apidoc hv_iterkey | |
2158 | ||
2159 | Returns the key from the current position of the hash iterator. See | |
2160 | C<hv_iterinit>. | |
2161 | ||
2162 | =cut | |
2163 | */ | |
2164 | ||
79072805 | 2165 | char * |
864dbfa3 | 2166 | Perl_hv_iterkey(pTHX_ register HE *entry, I32 *retlen) |
79072805 | 2167 | { |
7918f24d NC |
2168 | PERL_ARGS_ASSERT_HV_ITERKEY; |
2169 | ||
fde52b5c | 2170 | if (HeKLEN(entry) == HEf_SVKEY) { |
fb73857a | 2171 | STRLEN len; |
0bd48802 | 2172 | char * const p = SvPV(HeKEY_sv(entry), len); |
fb73857a | 2173 | *retlen = len; |
2174 | return p; | |
fde52b5c | 2175 | } |
2176 | else { | |
2177 | *retlen = HeKLEN(entry); | |
2178 | return HeKEY(entry); | |
2179 | } | |
2180 | } | |
2181 | ||
2182 | /* unlike hv_iterval(), this always returns a mortal copy of the key */ | |
954c1994 GS |
2183 | /* |
2184 | =for apidoc hv_iterkeysv | |
2185 | ||
2186 | Returns the key as an C<SV*> from the current position of the hash | |
2187 | iterator. The return value will always be a mortal copy of the key. Also | |
2188 | see C<hv_iterinit>. | |
2189 | ||
2190 | =cut | |
2191 | */ | |
2192 | ||
fde52b5c | 2193 | SV * |
864dbfa3 | 2194 | Perl_hv_iterkeysv(pTHX_ register HE *entry) |
fde52b5c | 2195 | { |
7918f24d NC |
2196 | PERL_ARGS_ASSERT_HV_ITERKEYSV; |
2197 | ||
c1b02ed8 | 2198 | return sv_2mortal(newSVhek(HeKEY_hek(entry))); |
79072805 LW |
2199 | } |
2200 | ||
954c1994 GS |
2201 | /* |
2202 | =for apidoc hv_iterval | |
2203 | ||
2204 | Returns the value from the current position of the hash iterator. See | |
2205 | C<hv_iterkey>. | |
2206 | ||
2207 | =cut | |
2208 | */ | |
2209 | ||
79072805 | 2210 | SV * |
864dbfa3 | 2211 | Perl_hv_iterval(pTHX_ HV *hv, register HE *entry) |
79072805 | 2212 | { |
7918f24d NC |
2213 | PERL_ARGS_ASSERT_HV_ITERVAL; |
2214 | ||
8990e307 | 2215 | if (SvRMAGICAL(hv)) { |
ad64d0ec | 2216 | if (mg_find((const SV *)hv, PERL_MAGIC_tied)) { |
c4420975 | 2217 | SV* const sv = sv_newmortal(); |
bbce6d69 | 2218 | if (HeKLEN(entry) == HEf_SVKEY) |
ad64d0ec | 2219 | mg_copy(MUTABLE_SV(hv), sv, (char*)HeKEY_sv(entry), HEf_SVKEY); |
a3b680e6 | 2220 | else |
ad64d0ec | 2221 | mg_copy(MUTABLE_SV(hv), sv, HeKEY(entry), HeKLEN(entry)); |
463ee0b2 LW |
2222 | return sv; |
2223 | } | |
79072805 | 2224 | } |
fde52b5c | 2225 | return HeVAL(entry); |
79072805 LW |
2226 | } |
2227 | ||
954c1994 GS |
2228 | /* |
2229 | =for apidoc hv_iternextsv | |
2230 | ||
2231 | Performs an C<hv_iternext>, C<hv_iterkey>, and C<hv_iterval> in one | |
2232 | operation. | |
2233 | ||
2234 | =cut | |
2235 | */ | |
2236 | ||
a0d0e21e | 2237 | SV * |
864dbfa3 | 2238 | Perl_hv_iternextsv(pTHX_ HV *hv, char **key, I32 *retlen) |
a0d0e21e | 2239 | { |
0bd48802 AL |
2240 | HE * const he = hv_iternext_flags(hv, 0); |
2241 | ||
7918f24d NC |
2242 | PERL_ARGS_ASSERT_HV_ITERNEXTSV; |
2243 | ||
0bd48802 | 2244 | if (!he) |
a0d0e21e LW |
2245 | return NULL; |
2246 | *key = hv_iterkey(he, retlen); | |
2247 | return hv_iterval(hv, he); | |
2248 | } | |
2249 | ||
954c1994 | 2250 | /* |
bc5cdc23 NC |
2251 | |
2252 | Now a macro in hv.h | |
2253 | ||
954c1994 GS |
2254 | =for apidoc hv_magic |
2255 | ||
2256 | Adds magic to a hash. See C<sv_magic>. | |
2257 | ||
2258 | =cut | |
2259 | */ | |
2260 | ||
bbce6d69 | 2261 | /* possibly free a shared string if no one has access to it |
fde52b5c | 2262 | * len and hash must both be valid for str. |
2263 | */ | |
bbce6d69 | 2264 | void |
864dbfa3 | 2265 | Perl_unsharepvn(pTHX_ const char *str, I32 len, U32 hash) |
fde52b5c | 2266 | { |
19692e8d NC |
2267 | unshare_hek_or_pvn (NULL, str, len, hash); |
2268 | } | |
2269 | ||
2270 | ||
2271 | void | |
2272 | Perl_unshare_hek(pTHX_ HEK *hek) | |
2273 | { | |
bf11fd37 | 2274 | assert(hek); |
19692e8d NC |
2275 | unshare_hek_or_pvn(hek, NULL, 0, 0); |
2276 | } | |
2277 | ||
2278 | /* possibly free a shared string if no one has access to it | |
2279 | hek if non-NULL takes priority over the other 3, else str, len and hash | |
2280 | are used. If so, len and hash must both be valid for str. | |
2281 | */ | |
df132699 | 2282 | STATIC void |
97ddebaf | 2283 | S_unshare_hek_or_pvn(pTHX_ const HEK *hek, const char *str, I32 len, U32 hash) |
19692e8d | 2284 | { |
97aff369 | 2285 | dVAR; |
cbec9347 | 2286 | register XPVHV* xhv; |
20454177 | 2287 | HE *entry; |
fde52b5c | 2288 | register HE **oentry; |
45d1cc86 | 2289 | HE **first; |
c3654f1a | 2290 | bool is_utf8 = FALSE; |
19692e8d | 2291 | int k_flags = 0; |
aec46f14 | 2292 | const char * const save = str; |
cbbf8932 | 2293 | struct shared_he *he = NULL; |
c3654f1a | 2294 | |
19692e8d | 2295 | if (hek) { |
cbae3960 NC |
2296 | /* Find the shared he which is just before us in memory. */ |
2297 | he = (struct shared_he *)(((char *)hek) | |
2298 | - STRUCT_OFFSET(struct shared_he, | |
2299 | shared_he_hek)); | |
2300 | ||
2301 | /* Assert that the caller passed us a genuine (or at least consistent) | |
2302 | shared hek */ | |
2303 | assert (he->shared_he_he.hent_hek == hek); | |
29404ae0 NC |
2304 | |
2305 | LOCK_STRTAB_MUTEX; | |
de616631 NC |
2306 | if (he->shared_he_he.he_valu.hent_refcount - 1) { |
2307 | --he->shared_he_he.he_valu.hent_refcount; | |
29404ae0 NC |
2308 | UNLOCK_STRTAB_MUTEX; |
2309 | return; | |
2310 | } | |
2311 | UNLOCK_STRTAB_MUTEX; | |
2312 | ||
19692e8d NC |
2313 | hash = HEK_HASH(hek); |
2314 | } else if (len < 0) { | |
2315 | STRLEN tmplen = -len; | |
2316 | is_utf8 = TRUE; | |
2317 | /* See the note in hv_fetch(). --jhi */ | |
2318 | str = (char*)bytes_from_utf8((U8*)str, &tmplen, &is_utf8); | |
2319 | len = tmplen; | |
2320 | if (is_utf8) | |
2321 | k_flags = HVhek_UTF8; | |
2322 | if (str != save) | |
2323 | k_flags |= HVhek_WASUTF8 | HVhek_FREEKEY; | |
c3654f1a | 2324 | } |
1c846c1f | 2325 | |
de616631 | 2326 | /* what follows was the moral equivalent of: |
6b88bc9c | 2327 | if ((Svp = hv_fetch(PL_strtab, tmpsv, FALSE, hash))) { |
a0714e2c | 2328 | if (--*Svp == NULL) |
6b88bc9c | 2329 | hv_delete(PL_strtab, str, len, G_DISCARD, hash); |
bbce6d69 | 2330 | } */ |
cbec9347 | 2331 | xhv = (XPVHV*)SvANY(PL_strtab); |
fde52b5c | 2332 | /* assert(xhv_array != 0) */ |
5f08fbcd | 2333 | LOCK_STRTAB_MUTEX; |
45d1cc86 | 2334 | first = oentry = &(HvARRAY(PL_strtab))[hash & (I32) HvMAX(PL_strtab)]; |
6c1b96a1 NC |
2335 | if (he) { |
2336 | const HE *const he_he = &(he->shared_he_he); | |
45d1cc86 | 2337 | for (entry = *oentry; entry; oentry = &HeNEXT(entry), entry = *oentry) { |
35ab5632 NC |
2338 | if (entry == he_he) |
2339 | break; | |
19692e8d NC |
2340 | } |
2341 | } else { | |
35a4481c | 2342 | const int flags_masked = k_flags & HVhek_MASK; |
45d1cc86 | 2343 | for (entry = *oentry; entry; oentry = &HeNEXT(entry), entry = *oentry) { |
19692e8d NC |
2344 | if (HeHASH(entry) != hash) /* strings can't be equal */ |
2345 | continue; | |
2346 | if (HeKLEN(entry) != len) | |
2347 | continue; | |
2348 | if (HeKEY(entry) != str && memNE(HeKEY(entry),str,len)) /* is this it? */ | |
2349 | continue; | |
2350 | if (HeKFLAGS(entry) != flags_masked) | |
2351 | continue; | |
19692e8d NC |
2352 | break; |
2353 | } | |
2354 | } | |
2355 | ||
35ab5632 NC |
2356 | if (entry) { |
2357 | if (--entry->he_valu.hent_refcount == 0) { | |
19692e8d | 2358 | *oentry = HeNEXT(entry); |
45d1cc86 NC |
2359 | if (!*first) { |
2360 | /* There are now no entries in our slot. */ | |
19692e8d | 2361 | xhv->xhv_fill--; /* HvFILL(hv)-- */ |
45d1cc86 | 2362 | } |
cbae3960 | 2363 | Safefree(entry); |
4c7185a0 | 2364 | xhv->xhv_keys--; /* HvTOTALKEYS(hv)-- */ |
19692e8d | 2365 | } |
fde52b5c | 2366 | } |
19692e8d | 2367 | |
333f433b | 2368 | UNLOCK_STRTAB_MUTEX; |
35ab5632 | 2369 | if (!entry && ckWARN_d(WARN_INTERNAL)) |
19692e8d | 2370 | Perl_warner(aTHX_ packWARN(WARN_INTERNAL), |
472d47bc SB |
2371 | "Attempt to free non-existent shared string '%s'%s" |
2372 | pTHX__FORMAT, | |
19692e8d | 2373 | hek ? HEK_KEY(hek) : str, |
472d47bc | 2374 | ((k_flags & HVhek_UTF8) ? " (utf8)" : "") pTHX__VALUE); |
19692e8d NC |
2375 | if (k_flags & HVhek_FREEKEY) |
2376 | Safefree(str); | |
fde52b5c | 2377 | } |
2378 | ||
bbce6d69 | 2379 | /* get a (constant) string ptr from the global string table |
2380 | * string will get added if it is not already there. | |
fde52b5c | 2381 | * len and hash must both be valid for str. |
2382 | */ | |
bbce6d69 | 2383 | HEK * |
864dbfa3 | 2384 | Perl_share_hek(pTHX_ const char *str, I32 len, register U32 hash) |
fde52b5c | 2385 | { |
da58a35d | 2386 | bool is_utf8 = FALSE; |
19692e8d | 2387 | int flags = 0; |
aec46f14 | 2388 | const char * const save = str; |
da58a35d | 2389 | |
7918f24d NC |
2390 | PERL_ARGS_ASSERT_SHARE_HEK; |
2391 | ||
da58a35d | 2392 | if (len < 0) { |
77caf834 | 2393 | STRLEN tmplen = -len; |
da58a35d | 2394 | is_utf8 = TRUE; |
77caf834 JH |
2395 | /* See the note in hv_fetch(). --jhi */ |
2396 | str = (char*)bytes_from_utf8((U8*)str, &tmplen, &is_utf8); | |
2397 | len = tmplen; | |
19692e8d NC |
2398 | /* If we were able to downgrade here, then than means that we were passed |
2399 | in a key which only had chars 0-255, but was utf8 encoded. */ | |
2400 | if (is_utf8) | |
2401 | flags = HVhek_UTF8; | |
2402 | /* If we found we were able to downgrade the string to bytes, then | |
2403 | we should flag that it needs upgrading on keys or each. Also flag | |
2404 | that we need share_hek_flags to free the string. */ | |
2405 | if (str != save) | |
2406 | flags |= HVhek_WASUTF8 | HVhek_FREEKEY; | |
2407 | } | |
2408 | ||
6e838c70 | 2409 | return share_hek_flags (str, len, hash, flags); |
19692e8d NC |
2410 | } |
2411 | ||
6e838c70 | 2412 | STATIC HEK * |
19692e8d NC |
2413 | S_share_hek_flags(pTHX_ const char *str, I32 len, register U32 hash, int flags) |
2414 | { | |
97aff369 | 2415 | dVAR; |
19692e8d | 2416 | register HE *entry; |
35a4481c | 2417 | const int flags_masked = flags & HVhek_MASK; |
263cb4a6 | 2418 | const U32 hindex = hash & (I32) HvMAX(PL_strtab); |
7918f24d NC |
2419 | register XPVHV * const xhv = (XPVHV*)SvANY(PL_strtab); |
2420 | ||
2421 | PERL_ARGS_ASSERT_SHARE_HEK_FLAGS; | |
bbce6d69 | 2422 | |
fde52b5c | 2423 | /* what follows is the moral equivalent of: |
1c846c1f | 2424 | |
6b88bc9c | 2425 | if (!(Svp = hv_fetch(PL_strtab, str, len, FALSE))) |
a0714e2c | 2426 | hv_store(PL_strtab, str, len, NULL, hash); |
fdcd69b6 NC |
2427 | |
2428 | Can't rehash the shared string table, so not sure if it's worth | |
2429 | counting the number of entries in the linked list | |
bbce6d69 | 2430 | */ |
7918f24d | 2431 | |
fde52b5c | 2432 | /* assert(xhv_array != 0) */ |
5f08fbcd | 2433 | LOCK_STRTAB_MUTEX; |
263cb4a6 NC |
2434 | entry = (HvARRAY(PL_strtab))[hindex]; |
2435 | for (;entry; entry = HeNEXT(entry)) { | |
fde52b5c | 2436 | if (HeHASH(entry) != hash) /* strings can't be equal */ |
2437 | continue; | |
2438 | if (HeKLEN(entry) != len) | |
2439 | continue; | |
1c846c1f | 2440 | if (HeKEY(entry) != str && memNE(HeKEY(entry),str,len)) /* is this it? */ |
fde52b5c | 2441 | continue; |
19692e8d | 2442 | if (HeKFLAGS(entry) != flags_masked) |
c3654f1a | 2443 | continue; |
fde52b5c | 2444 | break; |
2445 | } | |
263cb4a6 NC |
2446 | |
2447 | if (!entry) { | |
45d1cc86 NC |
2448 | /* What used to be head of the list. |
2449 | If this is NULL, then we're the first entry for this slot, which | |
2450 | means we need to increate fill. */ | |
cbae3960 NC |
2451 | struct shared_he *new_entry; |
2452 | HEK *hek; | |
2453 | char *k; | |
263cb4a6 NC |
2454 | HE **const head = &HvARRAY(PL_strtab)[hindex]; |
2455 | HE *const next = *head; | |
cbae3960 NC |
2456 | |
2457 | /* We don't actually store a HE from the arena and a regular HEK. | |
2458 | Instead we allocate one chunk of memory big enough for both, | |
2459 | and put the HEK straight after the HE. This way we can find the | |
2460 | HEK directly from the HE. | |
2461 | */ | |
2462 | ||
a02a5408 | 2463 | Newx(k, STRUCT_OFFSET(struct shared_he, |
cbae3960 NC |
2464 | shared_he_hek.hek_key[0]) + len + 2, char); |
2465 | new_entry = (struct shared_he *)k; | |
2466 | entry = &(new_entry->shared_he_he); | |
2467 | hek = &(new_entry->shared_he_hek); | |
2468 | ||
2469 | Copy(str, HEK_KEY(hek), len, char); | |
2470 | HEK_KEY(hek)[len] = 0; | |
2471 | HEK_LEN(hek) = len; | |
2472 | HEK_HASH(hek) = hash; | |
2473 | HEK_FLAGS(hek) = (unsigned char)flags_masked; | |
2474 | ||
2475 | /* Still "point" to the HEK, so that other code need not know what | |
2476 | we're up to. */ | |
2477 | HeKEY_hek(entry) = hek; | |
de616631 | 2478 | entry->he_valu.hent_refcount = 0; |
263cb4a6 NC |
2479 | HeNEXT(entry) = next; |
2480 | *head = entry; | |
cbae3960 | 2481 | |
4c7185a0 | 2482 | xhv->xhv_keys++; /* HvTOTALKEYS(hv)++ */ |
263cb4a6 | 2483 | if (!next) { /* initial entry? */ |
cbec9347 | 2484 | xhv->xhv_fill++; /* HvFILL(hv)++ */ |
4c9cc595 | 2485 | } else if (xhv->xhv_keys > (IV)xhv->xhv_max /* HvKEYS(hv) > HvMAX(hv) */) { |
cbec9347 | 2486 | hsplit(PL_strtab); |
bbce6d69 | 2487 | } |
2488 | } | |
2489 | ||
de616631 | 2490 | ++entry->he_valu.hent_refcount; |
5f08fbcd | 2491 | UNLOCK_STRTAB_MUTEX; |
19692e8d NC |
2492 | |
2493 | if (flags & HVhek_FREEKEY) | |
f9a63242 | 2494 | Safefree(str); |
19692e8d | 2495 | |
6e838c70 | 2496 | return HeKEY_hek(entry); |
fde52b5c | 2497 | } |
ecae49c0 | 2498 | |
ca732855 NC |
2499 | I32 * |
2500 | Perl_hv_placeholders_p(pTHX_ HV *hv) | |
2501 | { | |
2502 | dVAR; | |
ad64d0ec | 2503 | MAGIC *mg = mg_find((const SV *)hv, PERL_MAGIC_rhash); |
ca732855 | 2504 | |
7918f24d NC |
2505 | PERL_ARGS_ASSERT_HV_PLACEHOLDERS_P; |
2506 | ||
ca732855 | 2507 | if (!mg) { |
ad64d0ec | 2508 | mg = sv_magicext(MUTABLE_SV(hv), 0, PERL_MAGIC_rhash, 0, 0, 0); |
ca732855 NC |
2509 | |
2510 | if (!mg) { | |
2511 | Perl_die(aTHX_ "panic: hv_placeholders_p"); | |
2512 | } | |
2513 | } | |
2514 | return &(mg->mg_len); | |
2515 | } | |
2516 | ||
2517 | ||
2518 | I32 | |
0c289d13 | 2519 | Perl_hv_placeholders_get(pTHX_ const HV *hv) |
ca732855 NC |
2520 | { |
2521 | dVAR; | |
0c289d13 | 2522 | MAGIC * const mg = mg_find((const SV *)hv, PERL_MAGIC_rhash); |
ca732855 | 2523 | |
7918f24d NC |
2524 | PERL_ARGS_ASSERT_HV_PLACEHOLDERS_GET; |
2525 | ||
ca732855 NC |
2526 | return mg ? mg->mg_len : 0; |
2527 | } | |
2528 | ||
2529 | void | |
ac1e784a | 2530 | Perl_hv_placeholders_set(pTHX_ HV *hv, I32 ph) |
ca732855 NC |
2531 | { |
2532 | dVAR; | |
ad64d0ec | 2533 | MAGIC * const mg = mg_find((const SV *)hv, PERL_MAGIC_rhash); |
ca732855 | 2534 | |
7918f24d NC |
2535 | PERL_ARGS_ASSERT_HV_PLACEHOLDERS_SET; |
2536 | ||
ca732855 NC |
2537 | if (mg) { |
2538 | mg->mg_len = ph; | |
2539 | } else if (ph) { | |
ad64d0ec | 2540 | if (!sv_magicext(MUTABLE_SV(hv), 0, PERL_MAGIC_rhash, 0, 0, ph)) |
ca732855 NC |
2541 | Perl_die(aTHX_ "panic: hv_placeholders_set"); |
2542 | } | |
2543 | /* else we don't need to add magic to record 0 placeholders. */ | |
2544 | } | |
ecae49c0 | 2545 | |
2a49f0f5 | 2546 | STATIC SV * |
7b0bddfa NC |
2547 | S_refcounted_he_value(pTHX_ const struct refcounted_he *he) |
2548 | { | |
0b2d3faa | 2549 | dVAR; |
7b0bddfa | 2550 | SV *value; |
7918f24d NC |
2551 | |
2552 | PERL_ARGS_ASSERT_REFCOUNTED_HE_VALUE; | |
2553 | ||
7b0bddfa NC |
2554 | switch(he->refcounted_he_data[0] & HVrhek_typemask) { |
2555 | case HVrhek_undef: | |
2556 | value = newSV(0); | |
2557 | break; | |
2558 | case HVrhek_delete: | |
2559 | value = &PL_sv_placeholder; | |
2560 | break; | |
2561 | case HVrhek_IV: | |
44ebaf21 NC |
2562 | value = newSViv(he->refcounted_he_val.refcounted_he_u_iv); |
2563 | break; | |
2564 | case HVrhek_UV: | |
2565 | value = newSVuv(he->refcounted_he_val.refcounted_he_u_uv); | |
7b0bddfa NC |
2566 | break; |
2567 | case HVrhek_PV: | |
44ebaf21 | 2568 | case HVrhek_PV_UTF8: |
7b0bddfa NC |
2569 | /* Create a string SV that directly points to the bytes in our |
2570 | structure. */ | |
b9f83d2f | 2571 | value = newSV_type(SVt_PV); |
7b0bddfa NC |
2572 | SvPV_set(value, (char *) he->refcounted_he_data + 1); |
2573 | SvCUR_set(value, he->refcounted_he_val.refcounted_he_u_len); | |
2574 | /* This stops anything trying to free it */ | |
2575 | SvLEN_set(value, 0); | |
2576 | SvPOK_on(value); | |
2577 | SvREADONLY_on(value); | |
44ebaf21 | 2578 | if ((he->refcounted_he_data[0] & HVrhek_typemask) == HVrhek_PV_UTF8) |
7b0bddfa NC |
2579 | SvUTF8_on(value); |
2580 | break; | |
2581 | default: | |
2582 | Perl_croak(aTHX_ "panic: refcounted_he_value bad flags %x", | |
2583 | he->refcounted_he_data[0]); | |
2584 | } | |
2585 | return value; | |
2586 | } | |
2587 | ||
ecae49c0 | 2588 | /* |
b3ca2e83 NC |
2589 | =for apidoc refcounted_he_chain_2hv |
2590 | ||
abc25d8c | 2591 | Generates and returns a C<HV *> by walking up the tree starting at the passed |
b3ca2e83 NC |
2592 | in C<struct refcounted_he *>. |
2593 | ||
2594 | =cut | |
2595 | */ | |
2596 | HV * | |
2597 | Perl_refcounted_he_chain_2hv(pTHX_ const struct refcounted_he *chain) | |
2598 | { | |
7a89be66 | 2599 | dVAR; |
b3ca2e83 NC |
2600 | HV *hv = newHV(); |
2601 | U32 placeholders = 0; | |
2602 | /* We could chase the chain once to get an idea of the number of keys, | |
2603 | and call ksplit. But for now we'll make a potentially inefficient | |
2604 | hash with only 8 entries in its array. */ | |
2605 | const U32 max = HvMAX(hv); | |
2606 | ||
2607 | if (!HvARRAY(hv)) { | |
2608 | char *array; | |
2609 | Newxz(array, PERL_HV_ARRAY_ALLOC_BYTES(max + 1), char); | |
2610 | HvARRAY(hv) = (HE**)array; | |
2611 | } | |
2612 | ||
2613 | while (chain) { | |
cbb1fbea | 2614 | #ifdef USE_ITHREADS |
b6bbf3fa | 2615 | U32 hash = chain->refcounted_he_hash; |
cbb1fbea NC |
2616 | #else |
2617 | U32 hash = HEK_HASH(chain->refcounted_he_hek); | |
2618 | #endif | |
b3ca2e83 NC |
2619 | HE **oentry = &((HvARRAY(hv))[hash & max]); |
2620 | HE *entry = *oentry; | |
b6bbf3fa | 2621 | SV *value; |
cbb1fbea | 2622 | |
b3ca2e83 NC |
2623 | for (; entry; entry = HeNEXT(entry)) { |
2624 | if (HeHASH(entry) == hash) { | |
9f769845 NC |
2625 | /* We might have a duplicate key here. If so, entry is older |
2626 | than the key we've already put in the hash, so if they are | |
2627 | the same, skip adding entry. */ | |
2628 | #ifdef USE_ITHREADS | |
2629 | const STRLEN klen = HeKLEN(entry); | |
2630 | const char *const key = HeKEY(entry); | |
2631 | if (klen == chain->refcounted_he_keylen | |
2632 | && (!!HeKUTF8(entry) | |
2633 | == !!(chain->refcounted_he_data[0] & HVhek_UTF8)) | |
2634 | && memEQ(key, REF_HE_KEY(chain), klen)) | |
2635 | goto next_please; | |
2636 | #else | |
2637 | if (HeKEY_hek(entry) == chain->refcounted_he_hek) | |
2638 | goto next_please; | |
2639 | if (HeKLEN(entry) == HEK_LEN(chain->refcounted_he_hek) | |
2640 | && HeKUTF8(entry) == HEK_UTF8(chain->refcounted_he_hek) | |
2641 | && memEQ(HeKEY(entry), HEK_KEY(chain->refcounted_he_hek), | |
2642 | HeKLEN(entry))) | |
2643 | goto next_please; | |
2644 | #endif | |
b3ca2e83 NC |
2645 | } |
2646 | } | |
2647 | assert (!entry); | |
2648 | entry = new_HE(); | |
2649 | ||
cbb1fbea NC |
2650 | #ifdef USE_ITHREADS |
2651 | HeKEY_hek(entry) | |
7b0bddfa | 2652 | = share_hek_flags(REF_HE_KEY(chain), |
b6bbf3fa NC |
2653 | chain->refcounted_he_keylen, |
2654 | chain->refcounted_he_hash, | |
2655 | (chain->refcounted_he_data[0] | |
2656 | & (HVhek_UTF8|HVhek_WASUTF8))); | |
cbb1fbea | 2657 | #else |
71ad1b0c | 2658 | HeKEY_hek(entry) = share_hek_hek(chain->refcounted_he_hek); |
cbb1fbea | 2659 | #endif |
7b0bddfa NC |
2660 | value = refcounted_he_value(chain); |
2661 | if (value == &PL_sv_placeholder) | |
b3ca2e83 | 2662 | placeholders++; |
b6bbf3fa | 2663 | HeVAL(entry) = value; |
b3ca2e83 NC |
2664 | |
2665 | /* Link it into the chain. */ | |
2666 | HeNEXT(entry) = *oentry; | |
2667 | if (!HeNEXT(entry)) { | |
2668 | /* initial entry. */ | |
2669 | HvFILL(hv)++; | |
2670 | } | |
2671 | *oentry = entry; | |
2672 | ||
2673 | HvTOTALKEYS(hv)++; | |
2674 | ||
2675 | next_please: | |
71ad1b0c | 2676 | chain = chain->refcounted_he_next; |
b3ca2e83 NC |
2677 | } |
2678 | ||
2679 | if (placeholders) { | |
2680 | clear_placeholders(hv, placeholders); | |
2681 | HvTOTALKEYS(hv) -= placeholders; | |
2682 | } | |
2683 | ||
2684 | /* We could check in the loop to see if we encounter any keys with key | |
2685 | flags, but it's probably not worth it, as this per-hash flag is only | |
2686 | really meant as an optimisation for things like Storable. */ | |
2687 | HvHASKFLAGS_on(hv); | |
def9038f | 2688 | DEBUG_A(Perl_hv_assert(aTHX_ hv)); |
b3ca2e83 NC |
2689 | |
2690 | return hv; | |
2691 | } | |
2692 | ||
7b0bddfa NC |
2693 | SV * |
2694 | Perl_refcounted_he_fetch(pTHX_ const struct refcounted_he *chain, SV *keysv, | |
2695 | const char *key, STRLEN klen, int flags, U32 hash) | |
2696 | { | |
0b2d3faa | 2697 | dVAR; |
7b0bddfa NC |
2698 | /* Just to be awkward, if you're using this interface the UTF-8-or-not-ness |
2699 | of your key has to exactly match that which is stored. */ | |
2700 | SV *value = &PL_sv_placeholder; | |
7b0bddfa | 2701 | |
cd1d2f8a NC |
2702 | if (chain) { |
2703 | /* No point in doing any of this if there's nothing to find. */ | |
2704 | bool is_utf8; | |
7b0bddfa | 2705 | |
cd1d2f8a NC |
2706 | if (keysv) { |
2707 | if (flags & HVhek_FREEKEY) | |
2708 | Safefree(key); | |
2709 | key = SvPV_const(keysv, klen); | |
2710 | flags = 0; | |
2711 | is_utf8 = (SvUTF8(keysv) != 0); | |
2712 | } else { | |
2713 | is_utf8 = ((flags & HVhek_UTF8) ? TRUE : FALSE); | |
2714 | } | |
2715 | ||
2716 | if (!hash) { | |
2717 | if (keysv && (SvIsCOW_shared_hash(keysv))) { | |
2718 | hash = SvSHARED_HASH(keysv); | |
2719 | } else { | |
2720 | PERL_HASH(hash, key, klen); | |
2721 | } | |
2722 | } | |
7b0bddfa | 2723 | |
cd1d2f8a | 2724 | for (; chain; chain = chain->refcounted_he_next) { |
7b0bddfa | 2725 | #ifdef USE_ITHREADS |
cd1d2f8a NC |
2726 | if (hash != chain->refcounted_he_hash) |
2727 | continue; | |
2728 | if (klen != chain->refcounted_he_keylen) | |
2729 | continue; | |
2730 | if (memNE(REF_HE_KEY(chain),key,klen)) | |
2731 | continue; | |
2732 | if (!!is_utf8 != !!(chain->refcounted_he_data[0] & HVhek_UTF8)) | |
2733 | continue; | |
7b0bddfa | 2734 | #else |
cd1d2f8a NC |
2735 | if (hash != HEK_HASH(chain->refcounted_he_hek)) |
2736 | continue; | |
2737 | if (klen != (STRLEN)HEK_LEN(chain->refcounted_he_hek)) | |
2738 | continue; | |
2739 | if (memNE(HEK_KEY(chain->refcounted_he_hek),key,klen)) | |
2740 | continue; | |
2741 | if (!!is_utf8 != !!HEK_UTF8(chain->refcounted_he_hek)) | |
2742 | continue; | |
7b0bddfa NC |
2743 | #endif |
2744 | ||
cd1d2f8a NC |
2745 | value = sv_2mortal(refcounted_he_value(chain)); |
2746 | break; | |
2747 | } | |
7b0bddfa NC |
2748 | } |
2749 | ||
2750 | if (flags & HVhek_FREEKEY) | |
2751 | Safefree(key); | |
2752 | ||
2753 | return value; | |
2754 | } | |
2755 | ||
b3ca2e83 NC |
2756 | /* |
2757 | =for apidoc refcounted_he_new | |
2758 | ||
ec2a1de7 NC |
2759 | Creates a new C<struct refcounted_he>. As S<key> is copied, and value is |
2760 | stored in a compact form, all references remain the property of the caller. | |
2761 | The C<struct refcounted_he> is returned with a reference count of 1. | |
b3ca2e83 NC |
2762 | |
2763 | =cut | |
2764 | */ | |
2765 | ||
2766 | struct refcounted_he * | |
2767 | Perl_refcounted_he_new(pTHX_ struct refcounted_he *const parent, | |
2768 | SV *const key, SV *const value) { | |
7a89be66 | 2769 | dVAR; |
b6bbf3fa NC |
2770 | STRLEN key_len; |
2771 | const char *key_p = SvPV_const(key, key_len); | |
2772 | STRLEN value_len = 0; | |
95b63a38 | 2773 | const char *value_p = NULL; |
b6bbf3fa NC |
2774 | char value_type; |
2775 | char flags; | |
d8c5b3c5 | 2776 | bool is_utf8 = SvUTF8(key) ? TRUE : FALSE; |
b6bbf3fa NC |
2777 | |
2778 | if (SvPOK(value)) { | |
2779 | value_type = HVrhek_PV; | |
2780 | } else if (SvIOK(value)) { | |
ad64d0ec | 2781 | value_type = SvUOK((const SV *)value) ? HVrhek_UV : HVrhek_IV; |
b6bbf3fa NC |
2782 | } else if (value == &PL_sv_placeholder) { |
2783 | value_type = HVrhek_delete; | |
2784 | } else if (!SvOK(value)) { | |
2785 | value_type = HVrhek_undef; | |
2786 | } else { | |
2787 | value_type = HVrhek_PV; | |
2788 | } | |
b3ca2e83 | 2789 | |
b6bbf3fa | 2790 | if (value_type == HVrhek_PV) { |
012da8e5 NC |
2791 | /* Do it this way so that the SvUTF8() test is after the SvPV, in case |
2792 | the value is overloaded, and doesn't yet have the UTF-8flag set. */ | |
b6bbf3fa | 2793 | value_p = SvPV_const(value, value_len); |
012da8e5 NC |
2794 | if (SvUTF8(value)) |
2795 | value_type = HVrhek_PV_UTF8; | |
b6bbf3fa | 2796 | } |
012da8e5 NC |
2797 | flags = value_type; |
2798 | ||
2799 | if (is_utf8) { | |
2800 | /* Hash keys are always stored normalised to (yes) ISO-8859-1. | |
2801 | As we're going to be building hash keys from this value in future, | |
2802 | normalise it now. */ | |
2803 | key_p = (char*)bytes_from_utf8((const U8*)key_p, &key_len, &is_utf8); | |
2804 | flags |= is_utf8 ? HVhek_UTF8 : HVhek_WASUTF8; | |
2805 | } | |
2806 | ||
2807 | return refcounted_he_new_common(parent, key_p, key_len, flags, value_type, | |
2808 | ((value_type == HVrhek_PV | |
2809 | || value_type == HVrhek_PV_UTF8) ? | |
2810 | (void *)value_p : (void *)value), | |
2811 | value_len); | |
2812 | } | |
2813 | ||
1da83c39 | 2814 | static struct refcounted_he * |
012da8e5 NC |
2815 | S_refcounted_he_new_common(pTHX_ struct refcounted_he *const parent, |
2816 | const char *const key_p, const STRLEN key_len, | |
2817 | const char flags, char value_type, | |
2818 | const void *value, const STRLEN value_len) { | |
2819 | dVAR; | |
2820 | struct refcounted_he *he; | |
2821 | U32 hash; | |
2822 | const bool is_pv = value_type == HVrhek_PV || value_type == HVrhek_PV_UTF8; | |
2823 | STRLEN key_offset = is_pv ? value_len + 2 : 1; | |
2824 | ||
2825 | PERL_ARGS_ASSERT_REFCOUNTED_HE_NEW_COMMON; | |
b6bbf3fa | 2826 | |
b6bbf3fa | 2827 | #ifdef USE_ITHREADS |
10edeb5d JH |
2828 | he = (struct refcounted_he*) |
2829 | PerlMemShared_malloc(sizeof(struct refcounted_he) - 1 | |
2830 | + key_len | |
2831 | + key_offset); | |
6cef672b | 2832 | #else |
10edeb5d JH |
2833 | he = (struct refcounted_he*) |
2834 | PerlMemShared_malloc(sizeof(struct refcounted_he) - 1 | |
2835 | + key_offset); | |
6cef672b | 2836 | #endif |
b3ca2e83 | 2837 | |
71ad1b0c | 2838 | he->refcounted_he_next = parent; |
b6bbf3fa | 2839 | |
012da8e5 NC |
2840 | if (is_pv) { |
2841 | Copy((char *)value, he->refcounted_he_data + 1, value_len + 1, char); | |
b6bbf3fa | 2842 | he->refcounted_he_val.refcounted_he_u_len = value_len; |
b6bbf3fa | 2843 | } else if (value_type == HVrhek_IV) { |
ad64d0ec | 2844 | he->refcounted_he_val.refcounted_he_u_iv = SvIVX((const SV *)value); |
012da8e5 | 2845 | } else if (value_type == HVrhek_UV) { |
ad64d0ec | 2846 | he->refcounted_he_val.refcounted_he_u_uv = SvUVX((const SV *)value); |
b6bbf3fa NC |
2847 | } |
2848 | ||
b6bbf3fa NC |
2849 | PERL_HASH(hash, key_p, key_len); |
2850 | ||
cbb1fbea | 2851 | #ifdef USE_ITHREADS |
b6bbf3fa NC |
2852 | he->refcounted_he_hash = hash; |
2853 | he->refcounted_he_keylen = key_len; | |
2854 | Copy(key_p, he->refcounted_he_data + key_offset, key_len, char); | |
cbb1fbea | 2855 | #else |
b6bbf3fa | 2856 | he->refcounted_he_hek = share_hek_flags(key_p, key_len, hash, flags); |
cbb1fbea | 2857 | #endif |
b6bbf3fa NC |
2858 | |
2859 | if (flags & HVhek_WASUTF8) { | |
2860 | /* If it was downgraded from UTF-8, then the pointer returned from | |
2861 | bytes_from_utf8 is an allocated pointer that we must free. */ | |
2862 | Safefree(key_p); | |
2863 | } | |
2864 | ||
2865 | he->refcounted_he_data[0] = flags; | |
b3ca2e83 NC |
2866 | he->refcounted_he_refcnt = 1; |
2867 | ||
2868 | return he; | |
2869 | } | |
2870 | ||
2871 | /* | |
2872 | =for apidoc refcounted_he_free | |
2873 | ||
2874 | Decrements the reference count of the passed in C<struct refcounted_he *> | |
2875 | by one. If the reference count reaches zero the structure's memory is freed, | |
2876 | and C<refcounted_he_free> iterates onto the parent node. | |
2877 | ||
2878 | =cut | |
2879 | */ | |
2880 | ||
2881 | void | |
2882 | Perl_refcounted_he_free(pTHX_ struct refcounted_he *he) { | |
53d44271 | 2883 | dVAR; |
57ca3b03 AL |
2884 | PERL_UNUSED_CONTEXT; |
2885 | ||
b3ca2e83 NC |
2886 | while (he) { |
2887 | struct refcounted_he *copy; | |
cbb1fbea | 2888 | U32 new_count; |
b3ca2e83 | 2889 | |
cbb1fbea NC |
2890 | HINTS_REFCNT_LOCK; |
2891 | new_count = --he->refcounted_he_refcnt; | |
2892 | HINTS_REFCNT_UNLOCK; | |
2893 | ||
2894 | if (new_count) { | |
b3ca2e83 | 2895 | return; |
cbb1fbea | 2896 | } |
b3ca2e83 | 2897 | |
b6bbf3fa | 2898 | #ifndef USE_ITHREADS |
71ad1b0c | 2899 | unshare_hek_or_pvn (he->refcounted_he_hek, 0, 0, 0); |
cbb1fbea | 2900 | #endif |
b3ca2e83 | 2901 | copy = he; |
71ad1b0c | 2902 | he = he->refcounted_he_next; |
b6bbf3fa | 2903 | PerlMemShared_free(copy); |
b3ca2e83 NC |
2904 | } |
2905 | } | |
2906 | ||
dca6062a NC |
2907 | const char * |
2908 | Perl_fetch_cop_label(pTHX_ struct refcounted_he *const chain, STRLEN *len, | |
2909 | U32 *flags) { | |
2910 | if (!chain) | |
2911 | return NULL; | |
2912 | #ifdef USE_ITHREADS | |
2913 | if (chain->refcounted_he_keylen != 1) | |
2914 | return NULL; | |
2915 | if (*REF_HE_KEY(chain) != ':') | |
2916 | return NULL; | |
2917 | #else | |
2918 | if ((STRLEN)HEK_LEN(chain->refcounted_he_hek) != 1) | |
2919 | return NULL; | |
2920 | if (*HEK_KEY(chain->refcounted_he_hek) != ':') | |
2921 | return NULL; | |
2922 | #endif | |
012da8e5 NC |
2923 | /* Stop anyone trying to really mess us up by adding their own value for |
2924 | ':' into %^H */ | |
2925 | if ((chain->refcounted_he_data[0] & HVrhek_typemask) != HVrhek_PV | |
2926 | && (chain->refcounted_he_data[0] & HVrhek_typemask) != HVrhek_PV_UTF8) | |
2927 | return NULL; | |
2928 | ||
dca6062a NC |
2929 | if (len) |
2930 | *len = chain->refcounted_he_val.refcounted_he_u_len; | |
2931 | if (flags) { | |
2932 | *flags = ((chain->refcounted_he_data[0] & HVrhek_typemask) | |
2933 | == HVrhek_PV_UTF8) ? SVf_UTF8 : 0; | |
2934 | } | |
2935 | return chain->refcounted_he_data + 1; | |
2936 | } | |
2937 | ||
012da8e5 NC |
2938 | /* As newSTATEOP currently gets passed plain char* labels, we will only provide |
2939 | that interface. Once it works out how to pass in length and UTF-8 ness, this | |
2940 | function will need superseding. */ | |
2941 | struct refcounted_he * | |
2942 | Perl_store_cop_label(pTHX_ struct refcounted_he *const chain, const char *label) | |
2943 | { | |
547bb267 NC |
2944 | PERL_ARGS_ASSERT_STORE_COP_LABEL; |
2945 | ||
012da8e5 NC |
2946 | return refcounted_he_new_common(chain, ":", 1, HVrhek_PV, HVrhek_PV, |
2947 | label, strlen(label)); | |
2948 | } | |
2949 | ||
b3ca2e83 | 2950 | /* |
ecae49c0 NC |
2951 | =for apidoc hv_assert |
2952 | ||
2953 | Check that a hash is in an internally consistent state. | |
2954 | ||
2955 | =cut | |
2956 | */ | |
2957 | ||
943795c2 NC |
2958 | #ifdef DEBUGGING |
2959 | ||
ecae49c0 NC |
2960 | void |
2961 | Perl_hv_assert(pTHX_ HV *hv) | |
2962 | { | |
57ca3b03 AL |
2963 | dVAR; |
2964 | HE* entry; | |
2965 | int withflags = 0; | |
2966 | int placeholders = 0; | |
2967 | int real = 0; | |
2968 | int bad = 0; | |
2969 | const I32 riter = HvRITER_get(hv); | |
2970 | HE *eiter = HvEITER_get(hv); | |
2971 | ||
7918f24d NC |
2972 | PERL_ARGS_ASSERT_HV_ASSERT; |
2973 | ||
57ca3b03 AL |
2974 | (void)hv_iterinit(hv); |
2975 | ||
2976 | while ((entry = hv_iternext_flags(hv, HV_ITERNEXT_WANTPLACEHOLDERS))) { | |
2977 | /* sanity check the values */ | |
2978 | if (HeVAL(entry) == &PL_sv_placeholder) | |
2979 | placeholders++; | |
2980 | else | |
2981 | real++; | |
2982 | /* sanity check the keys */ | |
2983 | if (HeSVKEY(entry)) { | |
6f207bd3 | 2984 | NOOP; /* Don't know what to check on SV keys. */ |
57ca3b03 AL |
2985 | } else if (HeKUTF8(entry)) { |
2986 | withflags++; | |
2987 | if (HeKWASUTF8(entry)) { | |
2988 | PerlIO_printf(Perl_debug_log, | |
d2a455e7 | 2989 | "hash key has both WASUTF8 and UTF8: '%.*s'\n", |
57ca3b03 AL |
2990 | (int) HeKLEN(entry), HeKEY(entry)); |
2991 | bad = 1; | |
2992 | } | |
2993 | } else if (HeKWASUTF8(entry)) | |
2994 | withflags++; | |
2995 | } | |
ad64d0ec | 2996 | if (!SvTIED_mg((const SV *)hv, PERL_MAGIC_tied)) { |
57ca3b03 AL |
2997 | static const char bad_count[] = "Count %d %s(s), but hash reports %d\n"; |
2998 | const int nhashkeys = HvUSEDKEYS(hv); | |
2999 | const int nhashplaceholders = HvPLACEHOLDERS_get(hv); | |
3000 | ||
3001 | if (nhashkeys != real) { | |
3002 | PerlIO_printf(Perl_debug_log, bad_count, real, "keys", nhashkeys ); | |
3003 | bad = 1; | |
3004 | } | |
3005 | if (nhashplaceholders != placeholders) { | |
3006 | PerlIO_printf(Perl_debug_log, bad_count, placeholders, "placeholder", nhashplaceholders ); | |
3007 | bad = 1; | |
3008 | } | |
3009 | } | |
3010 | if (withflags && ! HvHASKFLAGS(hv)) { | |
3011 | PerlIO_printf(Perl_debug_log, | |
3012 | "Hash has HASKFLAGS off but I count %d key(s) with flags\n", | |
3013 | withflags); | |
3014 | bad = 1; | |
3015 | } | |
3016 | if (bad) { | |
ad64d0ec | 3017 | sv_dump(MUTABLE_SV(hv)); |
57ca3b03 AL |
3018 | } |
3019 | HvRITER_set(hv, riter); /* Restore hash iterator state */ | |
3020 | HvEITER_set(hv, eiter); | |
ecae49c0 | 3021 | } |
af3babe4 | 3022 | |
943795c2 NC |
3023 | #endif |
3024 | ||
af3babe4 NC |
3025 | /* |
3026 | * Local variables: | |
3027 | * c-indentation-style: bsd | |
3028 | * c-basic-offset: 4 | |
3029 | * indent-tabs-mode: t | |
3030 | * End: | |
3031 | * | |
37442d52 RGS |
3032 | * ex: set ts=8 sts=4 sw=4 noet: |
3033 | */ |