Commit | Line | Data |
---|---|---|
a0d0e21e | 1 | /* hv.c |
79072805 | 2 | * |
4bb101f2 | 3 | * Copyright (C) 1991, 1992, 1993, 1994, 1995, 1996, 1997, 1998, 1999, |
b94e2f88 | 4 | * 2000, 2001, 2002, 2003, 2004, 2005, 2006, 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 | /* | |
12 | * "I sit beside the fire and think of all that I have seen." --Bilbo | |
79072805 LW |
13 | */ |
14 | ||
d5afce77 RB |
15 | /* |
16 | =head1 Hash Manipulation Functions | |
166f8a29 DM |
17 | |
18 | A HV structure represents a Perl hash. It consists mainly of an array | |
19 | of pointers, each of which points to a linked list of HE structures. The | |
20 | array is indexed by the hash function of the key, so each linked list | |
21 | represents all the hash entries with the same hash value. Each HE contains | |
22 | a pointer to the actual value, plus a pointer to a HEK structure which | |
23 | holds the key and hash value. | |
24 | ||
25 | =cut | |
26 | ||
d5afce77 RB |
27 | */ |
28 | ||
79072805 | 29 | #include "EXTERN.h" |
864dbfa3 | 30 | #define PERL_IN_HV_C |
3d78eb94 | 31 | #define PERL_HASH_INTERNAL_ACCESS |
79072805 LW |
32 | #include "perl.h" |
33 | ||
d8012aaf | 34 | #define HV_MAX_LENGTH_BEFORE_SPLIT 14 |
fdcd69b6 | 35 | |
d75ce684 | 36 | static const char S_strtab_error[] |
5d2b1485 NC |
37 | = "Cannot modify shared string table in hv_%s"; |
38 | ||
cac9b346 NC |
39 | STATIC void |
40 | S_more_he(pTHX) | |
41 | { | |
97aff369 | 42 | dVAR; |
cac9b346 NC |
43 | HE* he; |
44 | HE* heend; | |
5e258f8c JC |
45 | |
46 | he = (HE*) Perl_get_arena(aTHX_ PERL_ARENA_SIZE); | |
cac9b346 NC |
47 | |
48 | heend = &he[PERL_ARENA_SIZE / sizeof(HE) - 1]; | |
d2a0f284 | 49 | PL_body_roots[HE_SVSLOT] = he; |
cac9b346 NC |
50 | while (he < heend) { |
51 | HeNEXT(he) = (HE*)(he + 1); | |
52 | he++; | |
53 | } | |
54 | HeNEXT(he) = 0; | |
55 | } | |
56 | ||
c941fb51 NC |
57 | #ifdef PURIFY |
58 | ||
59 | #define new_HE() (HE*)safemalloc(sizeof(HE)) | |
60 | #define del_HE(p) safefree((char*)p) | |
61 | ||
62 | #else | |
63 | ||
76e3520e | 64 | STATIC HE* |
cea2e8a9 | 65 | S_new_he(pTHX) |
4633a7c4 | 66 | { |
97aff369 | 67 | dVAR; |
4633a7c4 | 68 | HE* he; |
0bd48802 | 69 | void ** const root = &PL_body_roots[HE_SVSLOT]; |
6a93a7e5 | 70 | |
333f433b | 71 | LOCK_SV_MUTEX; |
6a93a7e5 | 72 | if (!*root) |
cac9b346 | 73 | S_more_he(aTHX); |
6a93a7e5 NC |
74 | he = *root; |
75 | *root = HeNEXT(he); | |
333f433b DG |
76 | UNLOCK_SV_MUTEX; |
77 | return he; | |
4633a7c4 LW |
78 | } |
79 | ||
c941fb51 NC |
80 | #define new_HE() new_he() |
81 | #define del_HE(p) \ | |
82 | STMT_START { \ | |
83 | LOCK_SV_MUTEX; \ | |
6a93a7e5 NC |
84 | HeNEXT(p) = (HE*)(PL_body_roots[HE_SVSLOT]); \ |
85 | PL_body_roots[HE_SVSLOT] = p; \ | |
c941fb51 NC |
86 | UNLOCK_SV_MUTEX; \ |
87 | } STMT_END | |
d33b2eba | 88 | |
d33b2eba | 89 | |
d33b2eba GS |
90 | |
91 | #endif | |
92 | ||
76e3520e | 93 | STATIC HEK * |
5f66b61c | 94 | S_save_hek_flags(const char *str, I32 len, U32 hash, int flags) |
bbce6d69 | 95 | { |
35a4481c | 96 | const int flags_masked = flags & HVhek_MASK; |
bbce6d69 | 97 | char *k; |
98 | register HEK *hek; | |
1c846c1f | 99 | |
a02a5408 | 100 | Newx(k, HEK_BASESIZE + len + 2, char); |
bbce6d69 | 101 | hek = (HEK*)k; |
ff68c719 | 102 | Copy(str, HEK_KEY(hek), len, char); |
e05949c7 | 103 | HEK_KEY(hek)[len] = 0; |
ff68c719 | 104 | HEK_LEN(hek) = len; |
105 | HEK_HASH(hek) = hash; | |
dcf933a4 NC |
106 | HEK_FLAGS(hek) = (unsigned char)flags_masked; |
107 | ||
108 | if (flags & HVhek_FREEKEY) | |
109 | Safefree(str); | |
bbce6d69 | 110 | return hek; |
111 | } | |
112 | ||
4a31713e | 113 | /* free the pool of temporary HE/HEK pairs returned by hv_fetch_ent |
dd28f7bb DM |
114 | * for tied hashes */ |
115 | ||
116 | void | |
117 | Perl_free_tied_hv_pool(pTHX) | |
118 | { | |
97aff369 | 119 | dVAR; |
dd28f7bb DM |
120 | HE *he = PL_hv_fetch_ent_mh; |
121 | while (he) { | |
9d4ba2ae | 122 | HE * const ohe = he; |
dd28f7bb | 123 | Safefree(HeKEY_hek(he)); |
dd28f7bb DM |
124 | he = HeNEXT(he); |
125 | del_HE(ohe); | |
126 | } | |
4608196e | 127 | PL_hv_fetch_ent_mh = NULL; |
dd28f7bb DM |
128 | } |
129 | ||
d18c6117 | 130 | #if defined(USE_ITHREADS) |
0bff533c NC |
131 | HEK * |
132 | Perl_hek_dup(pTHX_ HEK *source, CLONE_PARAMS* param) | |
133 | { | |
658b4a4a | 134 | HEK *shared = (HEK*)ptr_table_fetch(PL_ptr_table, source); |
9d4ba2ae AL |
135 | |
136 | PERL_UNUSED_ARG(param); | |
0bff533c NC |
137 | |
138 | if (shared) { | |
139 | /* We already shared this hash key. */ | |
454f1e26 | 140 | (void)share_hek_hek(shared); |
0bff533c NC |
141 | } |
142 | else { | |
658b4a4a | 143 | shared |
6e838c70 NC |
144 | = share_hek_flags(HEK_KEY(source), HEK_LEN(source), |
145 | HEK_HASH(source), HEK_FLAGS(source)); | |
658b4a4a | 146 | ptr_table_store(PL_ptr_table, source, shared); |
0bff533c | 147 | } |
658b4a4a | 148 | return shared; |
0bff533c NC |
149 | } |
150 | ||
d18c6117 | 151 | HE * |
5c4138a0 | 152 | Perl_he_dup(pTHX_ const HE *e, bool shared, CLONE_PARAMS* param) |
d18c6117 GS |
153 | { |
154 | HE *ret; | |
155 | ||
156 | if (!e) | |
4608196e | 157 | return NULL; |
7766f137 GS |
158 | /* look for it in the table first */ |
159 | ret = (HE*)ptr_table_fetch(PL_ptr_table, e); | |
160 | if (ret) | |
161 | return ret; | |
162 | ||
163 | /* create anew and remember what it is */ | |
d33b2eba | 164 | ret = new_HE(); |
7766f137 GS |
165 | ptr_table_store(PL_ptr_table, e, ret); |
166 | ||
d2d73c3e | 167 | HeNEXT(ret) = he_dup(HeNEXT(e),shared, param); |
dd28f7bb DM |
168 | if (HeKLEN(e) == HEf_SVKEY) { |
169 | char *k; | |
a02a5408 | 170 | Newx(k, HEK_BASESIZE + sizeof(SV*), char); |
dd28f7bb | 171 | HeKEY_hek(ret) = (HEK*)k; |
d2d73c3e | 172 | HeKEY_sv(ret) = SvREFCNT_inc(sv_dup(HeKEY_sv(e), param)); |
dd28f7bb | 173 | } |
c21d1a0f | 174 | else if (shared) { |
0bff533c NC |
175 | /* This is hek_dup inlined, which seems to be important for speed |
176 | reasons. */ | |
1b6737cc | 177 | HEK * const source = HeKEY_hek(e); |
658b4a4a | 178 | HEK *shared = (HEK*)ptr_table_fetch(PL_ptr_table, source); |
c21d1a0f NC |
179 | |
180 | if (shared) { | |
181 | /* We already shared this hash key. */ | |
454f1e26 | 182 | (void)share_hek_hek(shared); |
c21d1a0f NC |
183 | } |
184 | else { | |
658b4a4a | 185 | shared |
6e838c70 NC |
186 | = share_hek_flags(HEK_KEY(source), HEK_LEN(source), |
187 | HEK_HASH(source), HEK_FLAGS(source)); | |
658b4a4a | 188 | ptr_table_store(PL_ptr_table, source, shared); |
c21d1a0f | 189 | } |
658b4a4a | 190 | HeKEY_hek(ret) = shared; |
c21d1a0f | 191 | } |
d18c6117 | 192 | else |
19692e8d NC |
193 | HeKEY_hek(ret) = save_hek_flags(HeKEY(e), HeKLEN(e), HeHASH(e), |
194 | HeKFLAGS(e)); | |
d2d73c3e | 195 | HeVAL(ret) = SvREFCNT_inc(sv_dup(HeVAL(e), param)); |
d18c6117 GS |
196 | return ret; |
197 | } | |
198 | #endif /* USE_ITHREADS */ | |
199 | ||
1b1f1335 | 200 | static void |
2393f1b9 JH |
201 | S_hv_notallowed(pTHX_ int flags, const char *key, I32 klen, |
202 | const char *msg) | |
1b1f1335 | 203 | { |
1b6737cc | 204 | SV * const sv = sv_newmortal(); |
19692e8d | 205 | if (!(flags & HVhek_FREEKEY)) { |
1b1f1335 NIS |
206 | sv_setpvn(sv, key, klen); |
207 | } | |
208 | else { | |
209 | /* Need to free saved eventually assign to mortal SV */ | |
34c3c4e3 | 210 | /* XXX is this line an error ???: SV *sv = sv_newmortal(); */ |
1b1f1335 NIS |
211 | sv_usepvn(sv, (char *) key, klen); |
212 | } | |
19692e8d | 213 | if (flags & HVhek_UTF8) { |
1b1f1335 NIS |
214 | SvUTF8_on(sv); |
215 | } | |
c8cd6465 | 216 | Perl_croak(aTHX_ msg, sv); |
1b1f1335 NIS |
217 | } |
218 | ||
fde52b5c | 219 | /* (klen == HEf_SVKEY) is special for MAGICAL hv entries, meaning key slot |
220 | * contains an SV* */ | |
221 | ||
34a6f7b4 NC |
222 | #define HV_FETCH_ISSTORE 0x01 |
223 | #define HV_FETCH_ISEXISTS 0x02 | |
224 | #define HV_FETCH_LVALUE 0x04 | |
225 | #define HV_FETCH_JUST_SV 0x08 | |
226 | ||
227 | /* | |
228 | =for apidoc hv_store | |
229 | ||
230 | Stores an SV in a hash. The hash key is specified as C<key> and C<klen> is | |
231 | the length of the key. The C<hash> parameter is the precomputed hash | |
232 | value; if it is zero then Perl will compute it. The return value will be | |
233 | NULL if the operation failed or if the value did not need to be actually | |
234 | stored within the hash (as in the case of tied hashes). Otherwise it can | |
235 | be dereferenced to get the original C<SV*>. Note that the caller is | |
236 | responsible for suitably incrementing the reference count of C<val> before | |
237 | the call, and decrementing it if the function returned NULL. Effectively | |
238 | a successful hv_store takes ownership of one reference to C<val>. This is | |
239 | usually what you want; a newly created SV has a reference count of one, so | |
240 | if all your code does is create SVs then store them in a hash, hv_store | |
241 | will own the only reference to the new SV, and your code doesn't need to do | |
242 | anything further to tidy up. hv_store is not implemented as a call to | |
243 | hv_store_ent, and does not create a temporary SV for the key, so if your | |
244 | key data is not already in SV form then use hv_store in preference to | |
245 | hv_store_ent. | |
246 | ||
247 | See L<perlguts/"Understanding the Magic of Tied Hashes and Arrays"> for more | |
248 | information on how to use this function on tied hashes. | |
249 | ||
250 | =cut | |
251 | */ | |
252 | ||
253 | SV** | |
254 | Perl_hv_store(pTHX_ HV *hv, const char *key, I32 klen_i32, SV *val, U32 hash) | |
255 | { | |
256 | HE *hek; | |
257 | STRLEN klen; | |
258 | int flags; | |
259 | ||
260 | if (klen_i32 < 0) { | |
261 | klen = -klen_i32; | |
262 | flags = HVhek_UTF8; | |
263 | } else { | |
264 | klen = klen_i32; | |
265 | flags = 0; | |
266 | } | |
267 | hek = hv_fetch_common (hv, NULL, key, klen, flags, | |
52d01cc2 | 268 | (HV_FETCH_ISSTORE|HV_FETCH_JUST_SV), val, hash); |
34a6f7b4 NC |
269 | return hek ? &HeVAL(hek) : NULL; |
270 | } | |
271 | ||
fabdb6c0 | 272 | /* XXX This looks like an ideal candidate to inline */ |
34a6f7b4 NC |
273 | SV** |
274 | Perl_hv_store_flags(pTHX_ HV *hv, const char *key, I32 klen, SV *val, | |
275 | register U32 hash, int flags) | |
276 | { | |
9d4ba2ae | 277 | HE * const hek = hv_fetch_common (hv, NULL, key, klen, flags, |
34a6f7b4 NC |
278 | (HV_FETCH_ISSTORE|HV_FETCH_JUST_SV), val, hash); |
279 | return hek ? &HeVAL(hek) : NULL; | |
280 | } | |
281 | ||
282 | /* | |
283 | =for apidoc hv_store_ent | |
284 | ||
285 | Stores C<val> in a hash. The hash key is specified as C<key>. The C<hash> | |
286 | parameter is the precomputed hash value; if it is zero then Perl will | |
287 | compute it. The return value is the new hash entry so created. It will be | |
288 | NULL if the operation failed or if the value did not need to be actually | |
289 | stored within the hash (as in the case of tied hashes). Otherwise the | |
290 | contents of the return value can be accessed using the C<He?> macros | |
291 | described here. Note that the caller is responsible for suitably | |
292 | incrementing the reference count of C<val> before the call, and | |
293 | decrementing it if the function returned NULL. Effectively a successful | |
294 | hv_store_ent takes ownership of one reference to C<val>. This is | |
295 | usually what you want; a newly created SV has a reference count of one, so | |
296 | if all your code does is create SVs then store them in a hash, hv_store | |
297 | will own the only reference to the new SV, and your code doesn't need to do | |
298 | anything further to tidy up. Note that hv_store_ent only reads the C<key>; | |
299 | unlike C<val> it does not take ownership of it, so maintaining the correct | |
300 | reference count on C<key> is entirely the caller's responsibility. hv_store | |
301 | is not implemented as a call to hv_store_ent, and does not create a temporary | |
302 | SV for the key, so if your key data is not already in SV form then use | |
303 | hv_store in preference to hv_store_ent. | |
304 | ||
305 | See L<perlguts/"Understanding the Magic of Tied Hashes and Arrays"> for more | |
306 | information on how to use this function on tied hashes. | |
307 | ||
308 | =cut | |
309 | */ | |
310 | ||
fabdb6c0 | 311 | /* XXX This looks like an ideal candidate to inline */ |
34a6f7b4 NC |
312 | HE * |
313 | Perl_hv_store_ent(pTHX_ HV *hv, SV *keysv, SV *val, U32 hash) | |
314 | { | |
315 | return hv_fetch_common(hv, keysv, NULL, 0, 0, HV_FETCH_ISSTORE, val, hash); | |
316 | } | |
317 | ||
318 | /* | |
319 | =for apidoc hv_exists | |
320 | ||
321 | Returns a boolean indicating whether the specified hash key exists. The | |
322 | C<klen> is the length of the key. | |
323 | ||
324 | =cut | |
325 | */ | |
326 | ||
327 | bool | |
328 | Perl_hv_exists(pTHX_ HV *hv, const char *key, I32 klen_i32) | |
329 | { | |
330 | STRLEN klen; | |
331 | int flags; | |
332 | ||
333 | if (klen_i32 < 0) { | |
334 | klen = -klen_i32; | |
335 | flags = HVhek_UTF8; | |
336 | } else { | |
337 | klen = klen_i32; | |
338 | flags = 0; | |
339 | } | |
340 | return hv_fetch_common(hv, NULL, key, klen, flags, HV_FETCH_ISEXISTS, 0, 0) | |
341 | ? TRUE : FALSE; | |
342 | } | |
343 | ||
954c1994 GS |
344 | /* |
345 | =for apidoc hv_fetch | |
346 | ||
347 | Returns the SV which corresponds to the specified key in the hash. The | |
348 | C<klen> is the length of the key. If C<lval> is set then the fetch will be | |
349 | part of a store. Check that the return value is non-null before | |
d1be9408 | 350 | dereferencing it to an C<SV*>. |
954c1994 | 351 | |
96f1132b | 352 | See L<perlguts/"Understanding the Magic of Tied Hashes and Arrays"> for more |
954c1994 GS |
353 | information on how to use this function on tied hashes. |
354 | ||
355 | =cut | |
356 | */ | |
357 | ||
79072805 | 358 | SV** |
c1fe5510 | 359 | Perl_hv_fetch(pTHX_ HV *hv, const char *key, I32 klen_i32, I32 lval) |
79072805 | 360 | { |
c1fe5510 NC |
361 | HE *hek; |
362 | STRLEN klen; | |
363 | int flags; | |
364 | ||
365 | if (klen_i32 < 0) { | |
366 | klen = -klen_i32; | |
367 | flags = HVhek_UTF8; | |
368 | } else { | |
369 | klen = klen_i32; | |
370 | flags = 0; | |
371 | } | |
372 | hek = hv_fetch_common (hv, NULL, key, klen, flags, | |
c445ea15 | 373 | lval ? (HV_FETCH_JUST_SV | HV_FETCH_LVALUE) : HV_FETCH_JUST_SV, |
a0714e2c | 374 | NULL, 0); |
113738bb | 375 | return hek ? &HeVAL(hek) : NULL; |
79072805 LW |
376 | } |
377 | ||
34a6f7b4 NC |
378 | /* |
379 | =for apidoc hv_exists_ent | |
380 | ||
381 | Returns a boolean indicating whether the specified hash key exists. C<hash> | |
382 | can be a valid precomputed hash value, or 0 to ask for it to be | |
383 | computed. | |
384 | ||
385 | =cut | |
386 | */ | |
387 | ||
fabdb6c0 | 388 | /* XXX This looks like an ideal candidate to inline */ |
34a6f7b4 NC |
389 | bool |
390 | Perl_hv_exists_ent(pTHX_ HV *hv, SV *keysv, U32 hash) | |
391 | { | |
392 | return hv_fetch_common(hv, keysv, NULL, 0, 0, HV_FETCH_ISEXISTS, 0, hash) | |
393 | ? TRUE : FALSE; | |
394 | } | |
395 | ||
d1be9408 | 396 | /* returns an HE * structure with the all fields set */ |
fde52b5c | 397 | /* note that hent_val will be a mortal sv for MAGICAL hashes */ |
954c1994 GS |
398 | /* |
399 | =for apidoc hv_fetch_ent | |
400 | ||
401 | Returns the hash entry which corresponds to the specified key in the hash. | |
402 | C<hash> must be a valid precomputed hash number for the given C<key>, or 0 | |
403 | if you want the function to compute it. IF C<lval> is set then the fetch | |
404 | will be part of a store. Make sure the return value is non-null before | |
405 | accessing it. The return value when C<tb> is a tied hash is a pointer to a | |
406 | static location, so be sure to make a copy of the structure if you need to | |
1c846c1f | 407 | store it somewhere. |
954c1994 | 408 | |
96f1132b | 409 | See L<perlguts/"Understanding the Magic of Tied Hashes and Arrays"> for more |
954c1994 GS |
410 | information on how to use this function on tied hashes. |
411 | ||
412 | =cut | |
413 | */ | |
414 | ||
fde52b5c | 415 | HE * |
864dbfa3 | 416 | Perl_hv_fetch_ent(pTHX_ HV *hv, SV *keysv, I32 lval, register U32 hash) |
fde52b5c | 417 | { |
7f66fda2 | 418 | return hv_fetch_common(hv, keysv, NULL, 0, 0, |
a0714e2c | 419 | (lval ? HV_FETCH_LVALUE : 0), NULL, hash); |
113738bb NC |
420 | } |
421 | ||
8f8d40ab | 422 | STATIC HE * |
c1fe5510 | 423 | S_hv_fetch_common(pTHX_ HV *hv, SV *keysv, const char *key, STRLEN klen, |
b2c64049 | 424 | int flags, int action, SV *val, register U32 hash) |
113738bb | 425 | { |
27da23d5 | 426 | dVAR; |
b2c64049 | 427 | XPVHV* xhv; |
b2c64049 NC |
428 | HE *entry; |
429 | HE **oentry; | |
fde52b5c | 430 | SV *sv; |
da58a35d | 431 | bool is_utf8; |
113738bb | 432 | int masked_flags; |
fde52b5c | 433 | |
434 | if (!hv) | |
a4fc7abc | 435 | return NULL; |
fde52b5c | 436 | |
113738bb | 437 | if (keysv) { |
e593d2fe AE |
438 | if (flags & HVhek_FREEKEY) |
439 | Safefree(key); | |
5c144d81 | 440 | key = SvPV_const(keysv, klen); |
c1fe5510 | 441 | flags = 0; |
113738bb NC |
442 | is_utf8 = (SvUTF8(keysv) != 0); |
443 | } else { | |
c1fe5510 | 444 | is_utf8 = ((flags & HVhek_UTF8) ? TRUE : FALSE); |
113738bb | 445 | } |
113738bb | 446 | |
b2c64049 | 447 | xhv = (XPVHV*)SvANY(hv); |
7f66fda2 | 448 | if (SvMAGICAL(hv)) { |
6136c704 | 449 | if (SvRMAGICAL(hv) && !(action & (HV_FETCH_ISSTORE|HV_FETCH_ISEXISTS))) { |
7f66fda2 NC |
450 | if (mg_find((SV*)hv, PERL_MAGIC_tied) || SvGMAGICAL((SV*)hv)) { |
451 | sv = sv_newmortal(); | |
113738bb | 452 | |
7f66fda2 NC |
453 | /* XXX should be able to skimp on the HE/HEK here when |
454 | HV_FETCH_JUST_SV is true. */ | |
113738bb | 455 | |
7f66fda2 NC |
456 | if (!keysv) { |
457 | keysv = newSVpvn(key, klen); | |
458 | if (is_utf8) { | |
459 | SvUTF8_on(keysv); | |
460 | } | |
461 | } else { | |
462 | keysv = newSVsv(keysv); | |
113738bb | 463 | } |
7f66fda2 NC |
464 | mg_copy((SV*)hv, sv, (char *)keysv, HEf_SVKEY); |
465 | ||
466 | /* grab a fake HE/HEK pair from the pool or make a new one */ | |
467 | entry = PL_hv_fetch_ent_mh; | |
468 | if (entry) | |
469 | PL_hv_fetch_ent_mh = HeNEXT(entry); | |
470 | else { | |
471 | char *k; | |
472 | entry = new_HE(); | |
a02a5408 | 473 | Newx(k, HEK_BASESIZE + sizeof(SV*), char); |
7f66fda2 NC |
474 | HeKEY_hek(entry) = (HEK*)k; |
475 | } | |
4608196e | 476 | HeNEXT(entry) = NULL; |
7f66fda2 NC |
477 | HeSVKEY_set(entry, keysv); |
478 | HeVAL(entry) = sv; | |
479 | sv_upgrade(sv, SVt_PVLV); | |
480 | LvTYPE(sv) = 'T'; | |
481 | /* so we can free entry when freeing sv */ | |
482 | LvTARG(sv) = (SV*)entry; | |
483 | ||
484 | /* XXX remove at some point? */ | |
485 | if (flags & HVhek_FREEKEY) | |
486 | Safefree(key); | |
487 | ||
488 | return entry; | |
113738bb | 489 | } |
7f66fda2 NC |
490 | #ifdef ENV_IS_CASELESS |
491 | else if (mg_find((SV*)hv, PERL_MAGIC_env)) { | |
492 | U32 i; | |
493 | for (i = 0; i < klen; ++i) | |
494 | if (isLOWER(key[i])) { | |
086cb327 NC |
495 | /* Would be nice if we had a routine to do the |
496 | copy and upercase in a single pass through. */ | |
0bd48802 | 497 | const char * const nkey = strupr(savepvn(key,klen)); |
086cb327 NC |
498 | /* Note that this fetch is for nkey (the uppercased |
499 | key) whereas the store is for key (the original) */ | |
a0714e2c | 500 | entry = hv_fetch_common(hv, NULL, nkey, klen, |
086cb327 NC |
501 | HVhek_FREEKEY, /* free nkey */ |
502 | 0 /* non-LVAL fetch */, | |
a0714e2c | 503 | NULL /* no value */, |
086cb327 NC |
504 | 0 /* compute hash */); |
505 | if (!entry && (action & HV_FETCH_LVALUE)) { | |
506 | /* This call will free key if necessary. | |
507 | Do it this way to encourage compiler to tail | |
508 | call optimise. */ | |
509 | entry = hv_fetch_common(hv, keysv, key, klen, | |
510 | flags, HV_FETCH_ISSTORE, | |
561b68a9 | 511 | newSV(0), hash); |
086cb327 NC |
512 | } else { |
513 | if (flags & HVhek_FREEKEY) | |
514 | Safefree(key); | |
515 | } | |
516 | return entry; | |
7f66fda2 | 517 | } |
902173a3 | 518 | } |
7f66fda2 NC |
519 | #endif |
520 | } /* ISFETCH */ | |
521 | else if (SvRMAGICAL(hv) && (action & HV_FETCH_ISEXISTS)) { | |
522 | if (mg_find((SV*)hv, PERL_MAGIC_tied) || SvGMAGICAL((SV*)hv)) { | |
b2c64049 NC |
523 | /* I don't understand why hv_exists_ent has svret and sv, |
524 | whereas hv_exists only had one. */ | |
9d4ba2ae | 525 | SV * const svret = sv_newmortal(); |
b2c64049 | 526 | sv = sv_newmortal(); |
7f66fda2 NC |
527 | |
528 | if (keysv || is_utf8) { | |
529 | if (!keysv) { | |
530 | keysv = newSVpvn(key, klen); | |
531 | SvUTF8_on(keysv); | |
532 | } else { | |
533 | keysv = newSVsv(keysv); | |
534 | } | |
b2c64049 NC |
535 | mg_copy((SV*)hv, sv, (char *)sv_2mortal(keysv), HEf_SVKEY); |
536 | } else { | |
537 | mg_copy((SV*)hv, sv, key, klen); | |
7f66fda2 | 538 | } |
b2c64049 NC |
539 | if (flags & HVhek_FREEKEY) |
540 | Safefree(key); | |
7f66fda2 NC |
541 | magic_existspack(svret, mg_find(sv, PERL_MAGIC_tiedelem)); |
542 | /* This cast somewhat evil, but I'm merely using NULL/ | |
543 | not NULL to return the boolean exists. | |
544 | And I know hv is not NULL. */ | |
545 | return SvTRUE(svret) ? (HE *)hv : NULL; | |
e7152ba2 | 546 | } |
7f66fda2 NC |
547 | #ifdef ENV_IS_CASELESS |
548 | else if (mg_find((SV*)hv, PERL_MAGIC_env)) { | |
549 | /* XXX This code isn't UTF8 clean. */ | |
a15d23f8 | 550 | char * const keysave = (char * const)key; |
b2c64049 NC |
551 | /* Will need to free this, so set FREEKEY flag. */ |
552 | key = savepvn(key,klen); | |
553 | key = (const char*)strupr((char*)key); | |
6136c704 | 554 | is_utf8 = FALSE; |
7f66fda2 | 555 | hash = 0; |
8b4f7dd5 | 556 | keysv = 0; |
b2c64049 NC |
557 | |
558 | if (flags & HVhek_FREEKEY) { | |
559 | Safefree(keysave); | |
560 | } | |
561 | flags |= HVhek_FREEKEY; | |
7f66fda2 | 562 | } |
902173a3 | 563 | #endif |
7f66fda2 | 564 | } /* ISEXISTS */ |
b2c64049 NC |
565 | else if (action & HV_FETCH_ISSTORE) { |
566 | bool needs_copy; | |
567 | bool needs_store; | |
568 | hv_magic_check (hv, &needs_copy, &needs_store); | |
569 | if (needs_copy) { | |
a3b680e6 | 570 | const bool save_taint = PL_tainted; |
b2c64049 NC |
571 | if (keysv || is_utf8) { |
572 | if (!keysv) { | |
573 | keysv = newSVpvn(key, klen); | |
574 | SvUTF8_on(keysv); | |
575 | } | |
576 | if (PL_tainting) | |
577 | PL_tainted = SvTAINTED(keysv); | |
578 | keysv = sv_2mortal(newSVsv(keysv)); | |
579 | mg_copy((SV*)hv, val, (char*)keysv, HEf_SVKEY); | |
580 | } else { | |
581 | mg_copy((SV*)hv, val, key, klen); | |
582 | } | |
583 | ||
584 | TAINT_IF(save_taint); | |
7b2c381c | 585 | if (!HvARRAY(hv) && !needs_store) { |
b2c64049 NC |
586 | if (flags & HVhek_FREEKEY) |
587 | Safefree(key); | |
4608196e | 588 | return NULL; |
b2c64049 NC |
589 | } |
590 | #ifdef ENV_IS_CASELESS | |
591 | else if (mg_find((SV*)hv, PERL_MAGIC_env)) { | |
592 | /* XXX This code isn't UTF8 clean. */ | |
593 | const char *keysave = key; | |
594 | /* Will need to free this, so set FREEKEY flag. */ | |
595 | key = savepvn(key,klen); | |
596 | key = (const char*)strupr((char*)key); | |
6136c704 | 597 | is_utf8 = FALSE; |
b2c64049 | 598 | hash = 0; |
8b4f7dd5 | 599 | keysv = 0; |
b2c64049 NC |
600 | |
601 | if (flags & HVhek_FREEKEY) { | |
602 | Safefree(keysave); | |
603 | } | |
604 | flags |= HVhek_FREEKEY; | |
605 | } | |
606 | #endif | |
607 | } | |
608 | } /* ISSTORE */ | |
7f66fda2 | 609 | } /* SvMAGICAL */ |
fde52b5c | 610 | |
7b2c381c | 611 | if (!HvARRAY(hv)) { |
b2c64049 | 612 | if ((action & (HV_FETCH_LVALUE | HV_FETCH_ISSTORE)) |
fde52b5c | 613 | #ifdef DYNAMIC_ENV_FETCH /* if it's an %ENV lookup, we may get it on the fly */ |
8aacddc1 | 614 | || (SvRMAGICAL((SV*)hv) && mg_find((SV*)hv, PERL_MAGIC_env)) |
fde52b5c | 615 | #endif |
d58e6666 NC |
616 | ) { |
617 | char *array; | |
a02a5408 | 618 | Newxz(array, |
cbec9347 | 619 | PERL_HV_ARRAY_ALLOC_BYTES(xhv->xhv_max+1 /* HvMAX(hv)+1 */), |
d58e6666 NC |
620 | char); |
621 | HvARRAY(hv) = (HE**)array; | |
622 | } | |
7f66fda2 NC |
623 | #ifdef DYNAMIC_ENV_FETCH |
624 | else if (action & HV_FETCH_ISEXISTS) { | |
625 | /* for an %ENV exists, if we do an insert it's by a recursive | |
626 | store call, so avoid creating HvARRAY(hv) right now. */ | |
627 | } | |
628 | #endif | |
113738bb NC |
629 | else { |
630 | /* XXX remove at some point? */ | |
631 | if (flags & HVhek_FREEKEY) | |
632 | Safefree(key); | |
633 | ||
fde52b5c | 634 | return 0; |
113738bb | 635 | } |
fde52b5c | 636 | } |
637 | ||
19692e8d | 638 | if (is_utf8) { |
41d88b63 | 639 | char * const keysave = (char *)key; |
f9a63242 | 640 | key = (char*)bytes_from_utf8((U8*)key, &klen, &is_utf8); |
19692e8d | 641 | if (is_utf8) |
c1fe5510 NC |
642 | flags |= HVhek_UTF8; |
643 | else | |
644 | flags &= ~HVhek_UTF8; | |
7f66fda2 NC |
645 | if (key != keysave) { |
646 | if (flags & HVhek_FREEKEY) | |
647 | Safefree(keysave); | |
19692e8d | 648 | flags |= HVhek_WASUTF8 | HVhek_FREEKEY; |
7f66fda2 | 649 | } |
19692e8d | 650 | } |
f9a63242 | 651 | |
4b5190b5 NC |
652 | if (HvREHASH(hv)) { |
653 | PERL_HASH_INTERNAL(hash, key, klen); | |
b2c64049 NC |
654 | /* We don't have a pointer to the hv, so we have to replicate the |
655 | flag into every HEK, so that hv_iterkeysv can see it. */ | |
656 | /* And yes, you do need this even though you are not "storing" because | |
fdcd69b6 NC |
657 | you can flip the flags below if doing an lval lookup. (And that |
658 | was put in to give the semantics Andreas was expecting.) */ | |
659 | flags |= HVhek_REHASH; | |
4b5190b5 | 660 | } else if (!hash) { |
113738bb | 661 | if (keysv && (SvIsCOW_shared_hash(keysv))) { |
c158a4fd | 662 | hash = SvSHARED_HASH(keysv); |
46187eeb NC |
663 | } else { |
664 | PERL_HASH(hash, key, klen); | |
665 | } | |
666 | } | |
effa1e2d | 667 | |
113738bb NC |
668 | masked_flags = (flags & HVhek_MASK); |
669 | ||
7f66fda2 | 670 | #ifdef DYNAMIC_ENV_FETCH |
4608196e | 671 | if (!HvARRAY(hv)) entry = NULL; |
7f66fda2 NC |
672 | else |
673 | #endif | |
b2c64049 | 674 | { |
7b2c381c | 675 | entry = (HvARRAY(hv))[hash & (I32) HvMAX(hv)]; |
b2c64049 | 676 | } |
0298d7b9 | 677 | for (; entry; entry = HeNEXT(entry)) { |
fde52b5c | 678 | if (HeHASH(entry) != hash) /* strings can't be equal */ |
679 | continue; | |
eb160463 | 680 | if (HeKLEN(entry) != (I32)klen) |
fde52b5c | 681 | continue; |
1c846c1f | 682 | if (HeKEY(entry) != key && memNE(HeKEY(entry),key,klen)) /* is this it? */ |
fde52b5c | 683 | continue; |
113738bb | 684 | if ((HeKFLAGS(entry) ^ masked_flags) & HVhek_UTF8) |
c3654f1a | 685 | continue; |
b2c64049 NC |
686 | |
687 | if (action & (HV_FETCH_LVALUE|HV_FETCH_ISSTORE)) { | |
688 | if (HeKFLAGS(entry) != masked_flags) { | |
689 | /* We match if HVhek_UTF8 bit in our flags and hash key's | |
690 | match. But if entry was set previously with HVhek_WASUTF8 | |
691 | and key now doesn't (or vice versa) then we should change | |
692 | the key's flag, as this is assignment. */ | |
693 | if (HvSHAREKEYS(hv)) { | |
694 | /* Need to swap the key we have for a key with the flags we | |
695 | need. As keys are shared we can't just write to the | |
696 | flag, so we share the new one, unshare the old one. */ | |
6136c704 | 697 | HEK * const new_hek = share_hek_flags(key, klen, hash, |
6e838c70 | 698 | masked_flags); |
b2c64049 NC |
699 | unshare_hek (HeKEY_hek(entry)); |
700 | HeKEY_hek(entry) = new_hek; | |
701 | } | |
5d2b1485 NC |
702 | else if (hv == PL_strtab) { |
703 | /* PL_strtab is usually the only hash without HvSHAREKEYS, | |
704 | so putting this test here is cheap */ | |
705 | if (flags & HVhek_FREEKEY) | |
706 | Safefree(key); | |
707 | Perl_croak(aTHX_ S_strtab_error, | |
708 | action & HV_FETCH_LVALUE ? "fetch" : "store"); | |
709 | } | |
b2c64049 NC |
710 | else |
711 | HeKFLAGS(entry) = masked_flags; | |
712 | if (masked_flags & HVhek_ENABLEHVKFLAGS) | |
713 | HvHASKFLAGS_on(hv); | |
714 | } | |
715 | if (HeVAL(entry) == &PL_sv_placeholder) { | |
716 | /* yes, can store into placeholder slot */ | |
717 | if (action & HV_FETCH_LVALUE) { | |
718 | if (SvMAGICAL(hv)) { | |
719 | /* This preserves behaviour with the old hv_fetch | |
720 | implementation which at this point would bail out | |
721 | with a break; (at "if we find a placeholder, we | |
722 | pretend we haven't found anything") | |
723 | ||
724 | That break mean that if a placeholder were found, it | |
725 | caused a call into hv_store, which in turn would | |
726 | check magic, and if there is no magic end up pretty | |
727 | much back at this point (in hv_store's code). */ | |
728 | break; | |
729 | } | |
730 | /* LVAL fetch which actaully needs a store. */ | |
561b68a9 | 731 | val = newSV(0); |
ca732855 | 732 | HvPLACEHOLDERS(hv)--; |
b2c64049 NC |
733 | } else { |
734 | /* store */ | |
735 | if (val != &PL_sv_placeholder) | |
ca732855 | 736 | HvPLACEHOLDERS(hv)--; |
b2c64049 NC |
737 | } |
738 | HeVAL(entry) = val; | |
739 | } else if (action & HV_FETCH_ISSTORE) { | |
740 | SvREFCNT_dec(HeVAL(entry)); | |
741 | HeVAL(entry) = val; | |
742 | } | |
27bcc0a7 | 743 | } else if (HeVAL(entry) == &PL_sv_placeholder) { |
b2c64049 NC |
744 | /* if we find a placeholder, we pretend we haven't found |
745 | anything */ | |
8aacddc1 | 746 | break; |
b2c64049 | 747 | } |
113738bb NC |
748 | if (flags & HVhek_FREEKEY) |
749 | Safefree(key); | |
fde52b5c | 750 | return entry; |
751 | } | |
752 | #ifdef DYNAMIC_ENV_FETCH /* %ENV lookup? If so, try to fetch the value now */ | |
0ed29950 NC |
753 | if (!(action & HV_FETCH_ISSTORE) |
754 | && SvRMAGICAL((SV*)hv) && mg_find((SV*)hv, PERL_MAGIC_env)) { | |
a6c40364 | 755 | unsigned long len; |
9d4ba2ae | 756 | const char * const env = PerlEnv_ENVgetenv_len(key,&len); |
a6c40364 GS |
757 | if (env) { |
758 | sv = newSVpvn(env,len); | |
759 | SvTAINTED_on(sv); | |
7fd3d16e | 760 | return hv_fetch_common(hv,keysv,key,klen,flags,HV_FETCH_ISSTORE,sv, |
b2c64049 | 761 | hash); |
a6c40364 | 762 | } |
fde52b5c | 763 | } |
764 | #endif | |
7f66fda2 NC |
765 | |
766 | if (!entry && SvREADONLY(hv) && !(action & HV_FETCH_ISEXISTS)) { | |
c445ea15 | 767 | hv_notallowed(flags, key, klen, |
c8cd6465 NC |
768 | "Attempt to access disallowed key '%"SVf"' in" |
769 | " a restricted hash"); | |
1b1f1335 | 770 | } |
b2c64049 NC |
771 | if (!(action & (HV_FETCH_LVALUE|HV_FETCH_ISSTORE))) { |
772 | /* Not doing some form of store, so return failure. */ | |
773 | if (flags & HVhek_FREEKEY) | |
774 | Safefree(key); | |
775 | return 0; | |
776 | } | |
113738bb | 777 | if (action & HV_FETCH_LVALUE) { |
561b68a9 | 778 | val = newSV(0); |
b2c64049 NC |
779 | if (SvMAGICAL(hv)) { |
780 | /* At this point the old hv_fetch code would call to hv_store, | |
781 | which in turn might do some tied magic. So we need to make that | |
782 | magic check happen. */ | |
783 | /* gonna assign to this, so it better be there */ | |
784 | return hv_fetch_common(hv, keysv, key, klen, flags, | |
785 | HV_FETCH_ISSTORE, val, hash); | |
786 | /* XXX Surely that could leak if the fetch-was-store fails? | |
787 | Just like the hv_fetch. */ | |
113738bb NC |
788 | } |
789 | } | |
790 | ||
b2c64049 NC |
791 | /* Welcome to hv_store... */ |
792 | ||
7b2c381c | 793 | if (!HvARRAY(hv)) { |
b2c64049 NC |
794 | /* Not sure if we can get here. I think the only case of oentry being |
795 | NULL is for %ENV with dynamic env fetch. But that should disappear | |
796 | with magic in the previous code. */ | |
d58e6666 | 797 | char *array; |
a02a5408 | 798 | Newxz(array, |
b2c64049 | 799 | PERL_HV_ARRAY_ALLOC_BYTES(xhv->xhv_max+1 /* HvMAX(hv)+1 */), |
d58e6666 NC |
800 | char); |
801 | HvARRAY(hv) = (HE**)array; | |
b2c64049 NC |
802 | } |
803 | ||
7b2c381c | 804 | oentry = &(HvARRAY(hv))[hash & (I32) xhv->xhv_max]; |
ab4af705 | 805 | |
b2c64049 NC |
806 | entry = new_HE(); |
807 | /* share_hek_flags will do the free for us. This might be considered | |
808 | bad API design. */ | |
809 | if (HvSHAREKEYS(hv)) | |
6e838c70 | 810 | HeKEY_hek(entry) = share_hek_flags(key, klen, hash, flags); |
5d2b1485 NC |
811 | else if (hv == PL_strtab) { |
812 | /* PL_strtab is usually the only hash without HvSHAREKEYS, so putting | |
813 | this test here is cheap */ | |
814 | if (flags & HVhek_FREEKEY) | |
815 | Safefree(key); | |
816 | Perl_croak(aTHX_ S_strtab_error, | |
817 | action & HV_FETCH_LVALUE ? "fetch" : "store"); | |
818 | } | |
b2c64049 NC |
819 | else /* gotta do the real thing */ |
820 | HeKEY_hek(entry) = save_hek_flags(key, klen, hash, flags); | |
821 | HeVAL(entry) = val; | |
822 | HeNEXT(entry) = *oentry; | |
823 | *oentry = entry; | |
824 | ||
825 | if (val == &PL_sv_placeholder) | |
ca732855 | 826 | HvPLACEHOLDERS(hv)++; |
b2c64049 NC |
827 | if (masked_flags & HVhek_ENABLEHVKFLAGS) |
828 | HvHASKFLAGS_on(hv); | |
829 | ||
0298d7b9 NC |
830 | { |
831 | const HE *counter = HeNEXT(entry); | |
832 | ||
833 | xhv->xhv_keys++; /* HvKEYS(hv)++ */ | |
834 | if (!counter) { /* initial entry? */ | |
835 | xhv->xhv_fill++; /* HvFILL(hv)++ */ | |
836 | } else if (xhv->xhv_keys > (IV)xhv->xhv_max) { | |
837 | hsplit(hv); | |
838 | } else if(!HvREHASH(hv)) { | |
839 | U32 n_links = 1; | |
840 | ||
841 | while ((counter = HeNEXT(counter))) | |
842 | n_links++; | |
843 | ||
844 | if (n_links > HV_MAX_LENGTH_BEFORE_SPLIT) { | |
845 | /* Use only the old HvKEYS(hv) > HvMAX(hv) condition to limit | |
846 | bucket splits on a rehashed hash, as we're not going to | |
847 | split it again, and if someone is lucky (evil) enough to | |
848 | get all the keys in one list they could exhaust our memory | |
849 | as we repeatedly double the number of buckets on every | |
850 | entry. Linear search feels a less worse thing to do. */ | |
851 | hsplit(hv); | |
852 | } | |
853 | } | |
fde52b5c | 854 | } |
b2c64049 NC |
855 | |
856 | return entry; | |
fde52b5c | 857 | } |
858 | ||
864dbfa3 | 859 | STATIC void |
b0e6ae5b | 860 | S_hv_magic_check(HV *hv, bool *needs_copy, bool *needs_store) |
d0066dc7 | 861 | { |
a3b680e6 | 862 | const MAGIC *mg = SvMAGIC(hv); |
d0066dc7 OT |
863 | *needs_copy = FALSE; |
864 | *needs_store = TRUE; | |
865 | while (mg) { | |
866 | if (isUPPER(mg->mg_type)) { | |
867 | *needs_copy = TRUE; | |
d60c5a05 | 868 | if (mg->mg_type == PERL_MAGIC_tied) { |
d0066dc7 | 869 | *needs_store = FALSE; |
4ab2a30b | 870 | return; /* We've set all there is to set. */ |
d0066dc7 OT |
871 | } |
872 | } | |
873 | mg = mg->mg_moremagic; | |
874 | } | |
875 | } | |
876 | ||
954c1994 | 877 | /* |
a3bcc51e TP |
878 | =for apidoc hv_scalar |
879 | ||
880 | Evaluates the hash in scalar context and returns the result. Handles magic when the hash is tied. | |
881 | ||
882 | =cut | |
883 | */ | |
884 | ||
885 | SV * | |
886 | Perl_hv_scalar(pTHX_ HV *hv) | |
887 | { | |
a3bcc51e | 888 | SV *sv; |
823a54a3 AL |
889 | |
890 | if (SvRMAGICAL(hv)) { | |
891 | MAGIC * const mg = mg_find((SV*)hv, PERL_MAGIC_tied); | |
892 | if (mg) | |
893 | return magic_scalarpack(hv, mg); | |
894 | } | |
a3bcc51e TP |
895 | |
896 | sv = sv_newmortal(); | |
897 | if (HvFILL((HV*)hv)) | |
898 | Perl_sv_setpvf(aTHX_ sv, "%ld/%ld", | |
899 | (long)HvFILL(hv), (long)HvMAX(hv) + 1); | |
900 | else | |
901 | sv_setiv(sv, 0); | |
902 | ||
903 | return sv; | |
904 | } | |
905 | ||
906 | /* | |
954c1994 GS |
907 | =for apidoc hv_delete |
908 | ||
909 | Deletes a key/value pair in the hash. The value SV is removed from the | |
1c846c1f | 910 | hash and returned to the caller. The C<klen> is the length of the key. |
954c1994 GS |
911 | The C<flags> value will normally be zero; if set to G_DISCARD then NULL |
912 | will be returned. | |
913 | ||
914 | =cut | |
915 | */ | |
916 | ||
79072805 | 917 | SV * |
cd6d36ac | 918 | Perl_hv_delete(pTHX_ HV *hv, const char *key, I32 klen_i32, I32 flags) |
79072805 | 919 | { |
cd6d36ac | 920 | STRLEN klen; |
6136c704 | 921 | int k_flags; |
cd6d36ac NC |
922 | |
923 | if (klen_i32 < 0) { | |
924 | klen = -klen_i32; | |
6136c704 | 925 | k_flags = HVhek_UTF8; |
cd6d36ac NC |
926 | } else { |
927 | klen = klen_i32; | |
6136c704 | 928 | k_flags = 0; |
cd6d36ac NC |
929 | } |
930 | return hv_delete_common(hv, NULL, key, klen, k_flags, flags, 0); | |
fde52b5c | 931 | } |
932 | ||
954c1994 GS |
933 | /* |
934 | =for apidoc hv_delete_ent | |
935 | ||
936 | Deletes a key/value pair in the hash. The value SV is removed from the | |
937 | hash and returned to the caller. The C<flags> value will normally be zero; | |
938 | if set to G_DISCARD then NULL will be returned. C<hash> can be a valid | |
939 | precomputed hash value, or 0 to ask for it to be computed. | |
940 | ||
941 | =cut | |
942 | */ | |
943 | ||
fabdb6c0 | 944 | /* XXX This looks like an ideal candidate to inline */ |
fde52b5c | 945 | SV * |
864dbfa3 | 946 | Perl_hv_delete_ent(pTHX_ HV *hv, SV *keysv, I32 flags, U32 hash) |
fde52b5c | 947 | { |
cd6d36ac | 948 | return hv_delete_common(hv, keysv, NULL, 0, 0, flags, hash); |
f1317c8d NC |
949 | } |
950 | ||
8f8d40ab | 951 | STATIC SV * |
cd6d36ac NC |
952 | S_hv_delete_common(pTHX_ HV *hv, SV *keysv, const char *key, STRLEN klen, |
953 | int k_flags, I32 d_flags, U32 hash) | |
f1317c8d | 954 | { |
27da23d5 | 955 | dVAR; |
cbec9347 | 956 | register XPVHV* xhv; |
fde52b5c | 957 | register HE *entry; |
958 | register HE **oentry; | |
9e720f71 | 959 | HE *const *first_entry; |
da58a35d | 960 | bool is_utf8; |
7a9669ca | 961 | int masked_flags; |
1c846c1f | 962 | |
fde52b5c | 963 | if (!hv) |
a0714e2c | 964 | return NULL; |
f1317c8d NC |
965 | |
966 | if (keysv) { | |
e593d2fe AE |
967 | if (k_flags & HVhek_FREEKEY) |
968 | Safefree(key); | |
5c144d81 | 969 | key = SvPV_const(keysv, klen); |
cd6d36ac | 970 | k_flags = 0; |
f1317c8d NC |
971 | is_utf8 = (SvUTF8(keysv) != 0); |
972 | } else { | |
cd6d36ac | 973 | is_utf8 = ((k_flags & HVhek_UTF8) ? TRUE : FALSE); |
f1317c8d | 974 | } |
f1317c8d | 975 | |
fde52b5c | 976 | if (SvRMAGICAL(hv)) { |
0a0bb7c7 OT |
977 | bool needs_copy; |
978 | bool needs_store; | |
979 | hv_magic_check (hv, &needs_copy, &needs_store); | |
980 | ||
f1317c8d | 981 | if (needs_copy) { |
6136c704 | 982 | SV *sv; |
7a9669ca NC |
983 | entry = hv_fetch_common(hv, keysv, key, klen, |
984 | k_flags & ~HVhek_FREEKEY, HV_FETCH_LVALUE, | |
a0714e2c | 985 | NULL, hash); |
7a9669ca | 986 | sv = entry ? HeVAL(entry) : NULL; |
f1317c8d NC |
987 | if (sv) { |
988 | if (SvMAGICAL(sv)) { | |
989 | mg_clear(sv); | |
990 | } | |
991 | if (!needs_store) { | |
992 | if (mg_find(sv, PERL_MAGIC_tiedelem)) { | |
993 | /* No longer an element */ | |
994 | sv_unmagic(sv, PERL_MAGIC_tiedelem); | |
995 | return sv; | |
996 | } | |
a0714e2c | 997 | return NULL; /* element cannot be deleted */ |
f1317c8d | 998 | } |
902173a3 | 999 | #ifdef ENV_IS_CASELESS |
8167a60a NC |
1000 | else if (mg_find((SV*)hv, PERL_MAGIC_env)) { |
1001 | /* XXX This code isn't UTF8 clean. */ | |
1002 | keysv = sv_2mortal(newSVpvn(key,klen)); | |
1003 | if (k_flags & HVhek_FREEKEY) { | |
1004 | Safefree(key); | |
1005 | } | |
1006 | key = strupr(SvPVX(keysv)); | |
1007 | is_utf8 = 0; | |
1008 | k_flags = 0; | |
1009 | hash = 0; | |
7f66fda2 | 1010 | } |
510ac311 | 1011 | #endif |
2fd1c6b8 | 1012 | } |
2fd1c6b8 | 1013 | } |
fde52b5c | 1014 | } |
cbec9347 | 1015 | xhv = (XPVHV*)SvANY(hv); |
7b2c381c | 1016 | if (!HvARRAY(hv)) |
a0714e2c | 1017 | return NULL; |
fde52b5c | 1018 | |
19692e8d | 1019 | if (is_utf8) { |
c445ea15 | 1020 | const char * const keysave = key; |
b464bac0 | 1021 | key = (char*)bytes_from_utf8((U8*)key, &klen, &is_utf8); |
cd6d36ac | 1022 | |
19692e8d | 1023 | if (is_utf8) |
cd6d36ac NC |
1024 | k_flags |= HVhek_UTF8; |
1025 | else | |
1026 | k_flags &= ~HVhek_UTF8; | |
7f66fda2 NC |
1027 | if (key != keysave) { |
1028 | if (k_flags & HVhek_FREEKEY) { | |
1029 | /* This shouldn't happen if our caller does what we expect, | |
1030 | but strictly the API allows it. */ | |
1031 | Safefree(keysave); | |
1032 | } | |
1033 | k_flags |= HVhek_WASUTF8 | HVhek_FREEKEY; | |
1034 | } | |
cd6d36ac | 1035 | HvHASKFLAGS_on((SV*)hv); |
19692e8d | 1036 | } |
f9a63242 | 1037 | |
4b5190b5 NC |
1038 | if (HvREHASH(hv)) { |
1039 | PERL_HASH_INTERNAL(hash, key, klen); | |
1040 | } else if (!hash) { | |
7a9669ca | 1041 | if (keysv && (SvIsCOW_shared_hash(keysv))) { |
c158a4fd | 1042 | hash = SvSHARED_HASH(keysv); |
7a9669ca NC |
1043 | } else { |
1044 | PERL_HASH(hash, key, klen); | |
1045 | } | |
4b5190b5 | 1046 | } |
fde52b5c | 1047 | |
7a9669ca NC |
1048 | masked_flags = (k_flags & HVhek_MASK); |
1049 | ||
9e720f71 | 1050 | first_entry = oentry = &(HvARRAY(hv))[hash & (I32) HvMAX(hv)]; |
fde52b5c | 1051 | entry = *oentry; |
9e720f71 | 1052 | for (; entry; oentry = &HeNEXT(entry), entry = *oentry) { |
6136c704 | 1053 | SV *sv; |
fde52b5c | 1054 | if (HeHASH(entry) != hash) /* strings can't be equal */ |
1055 | continue; | |
eb160463 | 1056 | if (HeKLEN(entry) != (I32)klen) |
fde52b5c | 1057 | continue; |
1c846c1f | 1058 | if (HeKEY(entry) != key && memNE(HeKEY(entry),key,klen)) /* is this it? */ |
fde52b5c | 1059 | continue; |
7a9669ca | 1060 | if ((HeKFLAGS(entry) ^ masked_flags) & HVhek_UTF8) |
c3654f1a | 1061 | continue; |
8aacddc1 | 1062 | |
5d2b1485 NC |
1063 | if (hv == PL_strtab) { |
1064 | if (k_flags & HVhek_FREEKEY) | |
1065 | Safefree(key); | |
1066 | Perl_croak(aTHX_ S_strtab_error, "delete"); | |
1067 | } | |
1068 | ||
8aacddc1 | 1069 | /* if placeholder is here, it's already been deleted.... */ |
6136c704 AL |
1070 | if (HeVAL(entry) == &PL_sv_placeholder) { |
1071 | if (k_flags & HVhek_FREEKEY) | |
1072 | Safefree(key); | |
1073 | return NULL; | |
8aacddc1 | 1074 | } |
6136c704 | 1075 | if (SvREADONLY(hv) && HeVAL(entry) && SvREADONLY(HeVAL(entry))) { |
2393f1b9 | 1076 | S_hv_notallowed(aTHX_ k_flags, key, klen, |
c8cd6465 NC |
1077 | "Attempt to delete readonly key '%"SVf"' from" |
1078 | " a restricted hash"); | |
8aacddc1 | 1079 | } |
b84d0860 NC |
1080 | if (k_flags & HVhek_FREEKEY) |
1081 | Safefree(key); | |
8aacddc1 | 1082 | |
cd6d36ac | 1083 | if (d_flags & G_DISCARD) |
a0714e2c | 1084 | sv = NULL; |
94f7643d | 1085 | else { |
79d01fbf | 1086 | sv = sv_2mortal(HeVAL(entry)); |
7996736c | 1087 | HeVAL(entry) = &PL_sv_placeholder; |
94f7643d | 1088 | } |
8aacddc1 NIS |
1089 | |
1090 | /* | |
1091 | * If a restricted hash, rather than really deleting the entry, put | |
1092 | * a placeholder there. This marks the key as being "approved", so | |
1093 | * we can still access via not-really-existing key without raising | |
1094 | * an error. | |
1095 | */ | |
1096 | if (SvREADONLY(hv)) { | |
754604c4 | 1097 | SvREFCNT_dec(HeVAL(entry)); |
7996736c | 1098 | HeVAL(entry) = &PL_sv_placeholder; |
8aacddc1 NIS |
1099 | /* We'll be saving this slot, so the number of allocated keys |
1100 | * doesn't go down, but the number placeholders goes up */ | |
ca732855 | 1101 | HvPLACEHOLDERS(hv)++; |
8aacddc1 | 1102 | } else { |
a26e96df | 1103 | *oentry = HeNEXT(entry); |
9e720f71 | 1104 | if(!*first_entry) { |
a26e96df | 1105 | xhv->xhv_fill--; /* HvFILL(hv)-- */ |
9e720f71 | 1106 | } |
b79f7545 | 1107 | if (SvOOK(hv) && entry == HvAUX(hv)->xhv_eiter /* HvEITER(hv) */) |
8aacddc1 NIS |
1108 | HvLAZYDEL_on(hv); |
1109 | else | |
1110 | hv_free_ent(hv, entry); | |
1111 | xhv->xhv_keys--; /* HvKEYS(hv)-- */ | |
574c8022 | 1112 | if (xhv->xhv_keys == 0) |
19692e8d | 1113 | HvHASKFLAGS_off(hv); |
8aacddc1 | 1114 | } |
79072805 LW |
1115 | return sv; |
1116 | } | |
8aacddc1 | 1117 | if (SvREADONLY(hv)) { |
2393f1b9 | 1118 | S_hv_notallowed(aTHX_ k_flags, key, klen, |
c8cd6465 NC |
1119 | "Attempt to delete disallowed key '%"SVf"' from" |
1120 | " a restricted hash"); | |
8aacddc1 NIS |
1121 | } |
1122 | ||
19692e8d | 1123 | if (k_flags & HVhek_FREEKEY) |
f9a63242 | 1124 | Safefree(key); |
a0714e2c | 1125 | return NULL; |
79072805 LW |
1126 | } |
1127 | ||
76e3520e | 1128 | STATIC void |
cea2e8a9 | 1129 | S_hsplit(pTHX_ HV *hv) |
79072805 | 1130 | { |
97aff369 | 1131 | dVAR; |
cbec9347 | 1132 | register XPVHV* xhv = (XPVHV*)SvANY(hv); |
a3b680e6 | 1133 | const I32 oldsize = (I32) xhv->xhv_max+1; /* HvMAX(hv)+1 (sick) */ |
79072805 LW |
1134 | register I32 newsize = oldsize * 2; |
1135 | register I32 i; | |
7b2c381c | 1136 | char *a = (char*) HvARRAY(hv); |
72311751 | 1137 | register HE **aep; |
79072805 | 1138 | register HE **oentry; |
4b5190b5 NC |
1139 | int longest_chain = 0; |
1140 | int was_shared; | |
79072805 | 1141 | |
18026298 NC |
1142 | /*PerlIO_printf(PerlIO_stderr(), "hsplit called for %p which had %d\n", |
1143 | hv, (int) oldsize);*/ | |
1144 | ||
5d88ecd7 | 1145 | if (HvPLACEHOLDERS_get(hv) && !SvREADONLY(hv)) { |
18026298 NC |
1146 | /* Can make this clear any placeholders first for non-restricted hashes, |
1147 | even though Storable rebuilds restricted hashes by putting in all the | |
1148 | placeholders (first) before turning on the readonly flag, because | |
1149 | Storable always pre-splits the hash. */ | |
1150 | hv_clear_placeholders(hv); | |
1151 | } | |
1152 | ||
3280af22 | 1153 | PL_nomemok = TRUE; |
8d6dde3e | 1154 | #if defined(STRANGE_MALLOC) || defined(MYMALLOC) |
b79f7545 NC |
1155 | Renew(a, PERL_HV_ARRAY_ALLOC_BYTES(newsize) |
1156 | + (SvOOK(hv) ? sizeof(struct xpvhv_aux) : 0), char); | |
422a93e5 | 1157 | if (!a) { |
4a33f861 | 1158 | PL_nomemok = FALSE; |
422a93e5 GA |
1159 | return; |
1160 | } | |
b79f7545 | 1161 | if (SvOOK(hv)) { |
7a9b70e9 | 1162 | Copy(&a[oldsize * sizeof(HE*)], &a[newsize * sizeof(HE*)], 1, struct xpvhv_aux); |
b79f7545 | 1163 | } |
4633a7c4 | 1164 | #else |
a02a5408 | 1165 | Newx(a, PERL_HV_ARRAY_ALLOC_BYTES(newsize) |
b79f7545 | 1166 | + (SvOOK(hv) ? sizeof(struct xpvhv_aux) : 0), char); |
422a93e5 | 1167 | if (!a) { |
3280af22 | 1168 | PL_nomemok = FALSE; |
422a93e5 GA |
1169 | return; |
1170 | } | |
7b2c381c | 1171 | Copy(HvARRAY(hv), a, oldsize * sizeof(HE*), char); |
b79f7545 NC |
1172 | if (SvOOK(hv)) { |
1173 | Copy(HvAUX(hv), &a[newsize * sizeof(HE*)], 1, struct xpvhv_aux); | |
1174 | } | |
fba3b22e | 1175 | if (oldsize >= 64) { |
7b2c381c | 1176 | offer_nice_chunk(HvARRAY(hv), |
b79f7545 NC |
1177 | PERL_HV_ARRAY_ALLOC_BYTES(oldsize) |
1178 | + (SvOOK(hv) ? sizeof(struct xpvhv_aux) : 0)); | |
4633a7c4 LW |
1179 | } |
1180 | else | |
7b2c381c | 1181 | Safefree(HvARRAY(hv)); |
4633a7c4 LW |
1182 | #endif |
1183 | ||
3280af22 | 1184 | PL_nomemok = FALSE; |
72311751 | 1185 | Zero(&a[oldsize * sizeof(HE*)], (newsize-oldsize) * sizeof(HE*), char); /* zero 2nd half*/ |
cbec9347 | 1186 | xhv->xhv_max = --newsize; /* HvMAX(hv) = --newsize */ |
7b2c381c | 1187 | HvARRAY(hv) = (HE**) a; |
72311751 | 1188 | aep = (HE**)a; |
79072805 | 1189 | |
72311751 | 1190 | for (i=0; i<oldsize; i++,aep++) { |
4b5190b5 NC |
1191 | int left_length = 0; |
1192 | int right_length = 0; | |
a3b680e6 AL |
1193 | register HE *entry; |
1194 | register HE **bep; | |
4b5190b5 | 1195 | |
72311751 | 1196 | if (!*aep) /* non-existent */ |
79072805 | 1197 | continue; |
72311751 GS |
1198 | bep = aep+oldsize; |
1199 | for (oentry = aep, entry = *aep; entry; entry = *oentry) { | |
eb160463 | 1200 | if ((HeHASH(entry) & newsize) != (U32)i) { |
fde52b5c | 1201 | *oentry = HeNEXT(entry); |
72311751 GS |
1202 | HeNEXT(entry) = *bep; |
1203 | if (!*bep) | |
cbec9347 | 1204 | xhv->xhv_fill++; /* HvFILL(hv)++ */ |
72311751 | 1205 | *bep = entry; |
4b5190b5 | 1206 | right_length++; |
79072805 LW |
1207 | continue; |
1208 | } | |
4b5190b5 | 1209 | else { |
fde52b5c | 1210 | oentry = &HeNEXT(entry); |
4b5190b5 NC |
1211 | left_length++; |
1212 | } | |
79072805 | 1213 | } |
72311751 | 1214 | if (!*aep) /* everything moved */ |
cbec9347 | 1215 | xhv->xhv_fill--; /* HvFILL(hv)-- */ |
4b5190b5 NC |
1216 | /* I think we don't actually need to keep track of the longest length, |
1217 | merely flag if anything is too long. But for the moment while | |
1218 | developing this code I'll track it. */ | |
1219 | if (left_length > longest_chain) | |
1220 | longest_chain = left_length; | |
1221 | if (right_length > longest_chain) | |
1222 | longest_chain = right_length; | |
1223 | } | |
1224 | ||
1225 | ||
1226 | /* Pick your policy for "hashing isn't working" here: */ | |
fdcd69b6 | 1227 | if (longest_chain <= HV_MAX_LENGTH_BEFORE_SPLIT /* split worked? */ |
4b5190b5 NC |
1228 | || HvREHASH(hv)) { |
1229 | return; | |
79072805 | 1230 | } |
4b5190b5 NC |
1231 | |
1232 | if (hv == PL_strtab) { | |
1233 | /* Urg. Someone is doing something nasty to the string table. | |
1234 | Can't win. */ | |
1235 | return; | |
1236 | } | |
1237 | ||
1238 | /* Awooga. Awooga. Pathological data. */ | |
fdcd69b6 | 1239 | /*PerlIO_printf(PerlIO_stderr(), "%p %d of %d with %d/%d buckets\n", hv, |
4b5190b5 NC |
1240 | longest_chain, HvTOTALKEYS(hv), HvFILL(hv), 1+HvMAX(hv));*/ |
1241 | ||
1242 | ++newsize; | |
a02a5408 | 1243 | Newxz(a, PERL_HV_ARRAY_ALLOC_BYTES(newsize) |
b79f7545 NC |
1244 | + (SvOOK(hv) ? sizeof(struct xpvhv_aux) : 0), char); |
1245 | if (SvOOK(hv)) { | |
1246 | Copy(HvAUX(hv), &a[newsize * sizeof(HE*)], 1, struct xpvhv_aux); | |
1247 | } | |
1248 | ||
4b5190b5 NC |
1249 | was_shared = HvSHAREKEYS(hv); |
1250 | ||
1251 | xhv->xhv_fill = 0; | |
1252 | HvSHAREKEYS_off(hv); | |
1253 | HvREHASH_on(hv); | |
1254 | ||
7b2c381c | 1255 | aep = HvARRAY(hv); |
4b5190b5 NC |
1256 | |
1257 | for (i=0; i<newsize; i++,aep++) { | |
a3b680e6 | 1258 | register HE *entry = *aep; |
4b5190b5 NC |
1259 | while (entry) { |
1260 | /* We're going to trash this HE's next pointer when we chain it | |
1261 | into the new hash below, so store where we go next. */ | |
9d4ba2ae | 1262 | HE * const next = HeNEXT(entry); |
4b5190b5 | 1263 | UV hash; |
a3b680e6 | 1264 | HE **bep; |
4b5190b5 NC |
1265 | |
1266 | /* Rehash it */ | |
1267 | PERL_HASH_INTERNAL(hash, HeKEY(entry), HeKLEN(entry)); | |
1268 | ||
1269 | if (was_shared) { | |
1270 | /* Unshare it. */ | |
aec46f14 | 1271 | HEK * const new_hek |
4b5190b5 NC |
1272 | = save_hek_flags(HeKEY(entry), HeKLEN(entry), |
1273 | hash, HeKFLAGS(entry)); | |
1274 | unshare_hek (HeKEY_hek(entry)); | |
1275 | HeKEY_hek(entry) = new_hek; | |
1276 | } else { | |
1277 | /* Not shared, so simply write the new hash in. */ | |
1278 | HeHASH(entry) = hash; | |
1279 | } | |
1280 | /*PerlIO_printf(PerlIO_stderr(), "%d ", HeKFLAGS(entry));*/ | |
1281 | HEK_REHASH_on(HeKEY_hek(entry)); | |
1282 | /*PerlIO_printf(PerlIO_stderr(), "%d\n", HeKFLAGS(entry));*/ | |
1283 | ||
1284 | /* Copy oentry to the correct new chain. */ | |
1285 | bep = ((HE**)a) + (hash & (I32) xhv->xhv_max); | |
1286 | if (!*bep) | |
1287 | xhv->xhv_fill++; /* HvFILL(hv)++ */ | |
1288 | HeNEXT(entry) = *bep; | |
1289 | *bep = entry; | |
1290 | ||
1291 | entry = next; | |
1292 | } | |
1293 | } | |
7b2c381c NC |
1294 | Safefree (HvARRAY(hv)); |
1295 | HvARRAY(hv) = (HE **)a; | |
79072805 LW |
1296 | } |
1297 | ||
72940dca | 1298 | void |
864dbfa3 | 1299 | Perl_hv_ksplit(pTHX_ HV *hv, IV newmax) |
72940dca | 1300 | { |
97aff369 | 1301 | dVAR; |
cbec9347 | 1302 | register XPVHV* xhv = (XPVHV*)SvANY(hv); |
a3b680e6 | 1303 | const I32 oldsize = (I32) xhv->xhv_max+1; /* HvMAX(hv)+1 (sick) */ |
72940dca | 1304 | register I32 newsize; |
1305 | register I32 i; | |
72311751 GS |
1306 | register char *a; |
1307 | register HE **aep; | |
72940dca | 1308 | register HE *entry; |
1309 | register HE **oentry; | |
1310 | ||
1311 | newsize = (I32) newmax; /* possible truncation here */ | |
1312 | if (newsize != newmax || newmax <= oldsize) | |
1313 | return; | |
1314 | while ((newsize & (1 + ~newsize)) != newsize) { | |
1315 | newsize &= ~(newsize & (1 + ~newsize)); /* get proper power of 2 */ | |
1316 | } | |
1317 | if (newsize < newmax) | |
1318 | newsize *= 2; | |
1319 | if (newsize < newmax) | |
1320 | return; /* overflow detection */ | |
1321 | ||
7b2c381c | 1322 | a = (char *) HvARRAY(hv); |
72940dca | 1323 | if (a) { |
3280af22 | 1324 | PL_nomemok = TRUE; |
8d6dde3e | 1325 | #if defined(STRANGE_MALLOC) || defined(MYMALLOC) |
b79f7545 NC |
1326 | Renew(a, PERL_HV_ARRAY_ALLOC_BYTES(newsize) |
1327 | + (SvOOK(hv) ? sizeof(struct xpvhv_aux) : 0), char); | |
8aacddc1 | 1328 | if (!a) { |
4a33f861 | 1329 | PL_nomemok = FALSE; |
422a93e5 GA |
1330 | return; |
1331 | } | |
b79f7545 | 1332 | if (SvOOK(hv)) { |
7a9b70e9 | 1333 | Copy(&a[oldsize * sizeof(HE*)], &a[newsize * sizeof(HE*)], 1, struct xpvhv_aux); |
b79f7545 | 1334 | } |
72940dca | 1335 | #else |
a02a5408 | 1336 | Newx(a, PERL_HV_ARRAY_ALLOC_BYTES(newsize) |
b79f7545 | 1337 | + (SvOOK(hv) ? sizeof(struct xpvhv_aux) : 0), char); |
8aacddc1 | 1338 | if (!a) { |
3280af22 | 1339 | PL_nomemok = FALSE; |
422a93e5 GA |
1340 | return; |
1341 | } | |
7b2c381c | 1342 | Copy(HvARRAY(hv), a, oldsize * sizeof(HE*), char); |
b79f7545 NC |
1343 | if (SvOOK(hv)) { |
1344 | Copy(HvAUX(hv), &a[newsize * sizeof(HE*)], 1, struct xpvhv_aux); | |
1345 | } | |
fba3b22e | 1346 | if (oldsize >= 64) { |
7b2c381c | 1347 | offer_nice_chunk(HvARRAY(hv), |
b79f7545 NC |
1348 | PERL_HV_ARRAY_ALLOC_BYTES(oldsize) |
1349 | + (SvOOK(hv) ? sizeof(struct xpvhv_aux) : 0)); | |
72940dca | 1350 | } |
1351 | else | |
7b2c381c | 1352 | Safefree(HvARRAY(hv)); |
72940dca | 1353 | #endif |
3280af22 | 1354 | PL_nomemok = FALSE; |
72311751 | 1355 | Zero(&a[oldsize * sizeof(HE*)], (newsize-oldsize) * sizeof(HE*), char); /* zero 2nd half*/ |
72940dca | 1356 | } |
1357 | else { | |
a02a5408 | 1358 | Newxz(a, PERL_HV_ARRAY_ALLOC_BYTES(newsize), char); |
72940dca | 1359 | } |
cbec9347 | 1360 | xhv->xhv_max = --newsize; /* HvMAX(hv) = --newsize */ |
7b2c381c | 1361 | HvARRAY(hv) = (HE **) a; |
cbec9347 | 1362 | if (!xhv->xhv_fill /* !HvFILL(hv) */) /* skip rest if no entries */ |
72940dca | 1363 | return; |
1364 | ||
72311751 GS |
1365 | aep = (HE**)a; |
1366 | for (i=0; i<oldsize; i++,aep++) { | |
1367 | if (!*aep) /* non-existent */ | |
72940dca | 1368 | continue; |
72311751 | 1369 | for (oentry = aep, entry = *aep; entry; entry = *oentry) { |
6136c704 AL |
1370 | register I32 j = (HeHASH(entry) & newsize); |
1371 | ||
1372 | if (j != i) { | |
72940dca | 1373 | j -= i; |
1374 | *oentry = HeNEXT(entry); | |
72311751 | 1375 | if (!(HeNEXT(entry) = aep[j])) |
cbec9347 | 1376 | xhv->xhv_fill++; /* HvFILL(hv)++ */ |
72311751 | 1377 | aep[j] = entry; |
72940dca | 1378 | continue; |
1379 | } | |
1380 | else | |
1381 | oentry = &HeNEXT(entry); | |
1382 | } | |
72311751 | 1383 | if (!*aep) /* everything moved */ |
cbec9347 | 1384 | xhv->xhv_fill--; /* HvFILL(hv)-- */ |
72940dca | 1385 | } |
1386 | } | |
1387 | ||
954c1994 GS |
1388 | /* |
1389 | =for apidoc newHV | |
1390 | ||
1391 | Creates a new HV. The reference count is set to 1. | |
1392 | ||
1393 | =cut | |
1394 | */ | |
1395 | ||
79072805 | 1396 | HV * |
864dbfa3 | 1397 | Perl_newHV(pTHX) |
79072805 | 1398 | { |
cbec9347 | 1399 | register XPVHV* xhv; |
561b68a9 | 1400 | HV * const hv = (HV*)newSV(0); |
79072805 | 1401 | |
a0d0e21e | 1402 | sv_upgrade((SV *)hv, SVt_PVHV); |
cbec9347 | 1403 | xhv = (XPVHV*)SvANY(hv); |
79072805 LW |
1404 | SvPOK_off(hv); |
1405 | SvNOK_off(hv); | |
1c846c1f | 1406 | #ifndef NODEFAULT_SHAREKEYS |
fde52b5c | 1407 | HvSHAREKEYS_on(hv); /* key-sharing on by default */ |
1c846c1f | 1408 | #endif |
4b5190b5 | 1409 | |
cbec9347 JH |
1410 | xhv->xhv_max = 7; /* HvMAX(hv) = 7 (start with 8 buckets) */ |
1411 | xhv->xhv_fill = 0; /* HvFILL(hv) = 0 */ | |
79072805 LW |
1412 | return hv; |
1413 | } | |
1414 | ||
b3ac6de7 | 1415 | HV * |
864dbfa3 | 1416 | Perl_newHVhv(pTHX_ HV *ohv) |
b3ac6de7 | 1417 | { |
9d4ba2ae | 1418 | HV * const hv = newHV(); |
4beac62f | 1419 | STRLEN hv_max, hv_fill; |
4beac62f AMS |
1420 | |
1421 | if (!ohv || (hv_fill = HvFILL(ohv)) == 0) | |
1422 | return hv; | |
4beac62f | 1423 | hv_max = HvMAX(ohv); |
b3ac6de7 | 1424 | |
b56ba0bf AMS |
1425 | if (!SvMAGICAL((SV *)ohv)) { |
1426 | /* It's an ordinary hash, so copy it fast. AMS 20010804 */ | |
eb160463 | 1427 | STRLEN i; |
a3b680e6 | 1428 | const bool shared = !!HvSHAREKEYS(ohv); |
aec46f14 | 1429 | HE **ents, ** const oents = (HE **)HvARRAY(ohv); |
ff875642 | 1430 | char *a; |
a02a5408 | 1431 | Newx(a, PERL_HV_ARRAY_ALLOC_BYTES(hv_max+1), char); |
ff875642 | 1432 | ents = (HE**)a; |
b56ba0bf AMS |
1433 | |
1434 | /* In each bucket... */ | |
1435 | for (i = 0; i <= hv_max; i++) { | |
6136c704 | 1436 | HE *prev = NULL; |
aec46f14 | 1437 | HE *oent = oents[i]; |
b56ba0bf AMS |
1438 | |
1439 | if (!oent) { | |
1440 | ents[i] = NULL; | |
1441 | continue; | |
1442 | } | |
1443 | ||
1444 | /* Copy the linked list of entries. */ | |
aec46f14 | 1445 | for (; oent; oent = HeNEXT(oent)) { |
a3b680e6 AL |
1446 | const U32 hash = HeHASH(oent); |
1447 | const char * const key = HeKEY(oent); | |
1448 | const STRLEN len = HeKLEN(oent); | |
1449 | const int flags = HeKFLAGS(oent); | |
6136c704 | 1450 | HE * const ent = new_HE(); |
b56ba0bf | 1451 | |
45dea987 | 1452 | HeVAL(ent) = newSVsv(HeVAL(oent)); |
19692e8d | 1453 | HeKEY_hek(ent) |
6e838c70 | 1454 | = shared ? share_hek_flags(key, len, hash, flags) |
19692e8d | 1455 | : save_hek_flags(key, len, hash, flags); |
b56ba0bf AMS |
1456 | if (prev) |
1457 | HeNEXT(prev) = ent; | |
1458 | else | |
1459 | ents[i] = ent; | |
1460 | prev = ent; | |
1461 | HeNEXT(ent) = NULL; | |
1462 | } | |
1463 | } | |
1464 | ||
1465 | HvMAX(hv) = hv_max; | |
1466 | HvFILL(hv) = hv_fill; | |
8aacddc1 | 1467 | HvTOTALKEYS(hv) = HvTOTALKEYS(ohv); |
b56ba0bf | 1468 | HvARRAY(hv) = ents; |
aec46f14 | 1469 | } /* not magical */ |
b56ba0bf AMS |
1470 | else { |
1471 | /* Iterate over ohv, copying keys and values one at a time. */ | |
b3ac6de7 | 1472 | HE *entry; |
bfcb3514 NC |
1473 | const I32 riter = HvRITER_get(ohv); |
1474 | HE * const eiter = HvEITER_get(ohv); | |
b56ba0bf AMS |
1475 | |
1476 | /* Can we use fewer buckets? (hv_max is always 2^n-1) */ | |
1477 | while (hv_max && hv_max + 1 >= hv_fill * 2) | |
1478 | hv_max = hv_max / 2; | |
1479 | HvMAX(hv) = hv_max; | |
1480 | ||
4a76a316 | 1481 | hv_iterinit(ohv); |
e16e2ff8 | 1482 | while ((entry = hv_iternext_flags(ohv, 0))) { |
19692e8d NC |
1483 | hv_store_flags(hv, HeKEY(entry), HeKLEN(entry), |
1484 | newSVsv(HeVAL(entry)), HeHASH(entry), | |
1485 | HeKFLAGS(entry)); | |
b3ac6de7 | 1486 | } |
bfcb3514 NC |
1487 | HvRITER_set(ohv, riter); |
1488 | HvEITER_set(ohv, eiter); | |
b3ac6de7 | 1489 | } |
1c846c1f | 1490 | |
b3ac6de7 IZ |
1491 | return hv; |
1492 | } | |
1493 | ||
79072805 | 1494 | void |
864dbfa3 | 1495 | Perl_hv_free_ent(pTHX_ HV *hv, register HE *entry) |
79072805 | 1496 | { |
97aff369 | 1497 | dVAR; |
16bdeea2 GS |
1498 | SV *val; |
1499 | ||
68dc0745 | 1500 | if (!entry) |
79072805 | 1501 | return; |
16bdeea2 | 1502 | val = HeVAL(entry); |
bfcb3514 | 1503 | if (val && isGV(val) && GvCVu(val) && HvNAME_get(hv)) |
3280af22 | 1504 | PL_sub_generation++; /* may be deletion of method from stash */ |
16bdeea2 | 1505 | SvREFCNT_dec(val); |
68dc0745 | 1506 | if (HeKLEN(entry) == HEf_SVKEY) { |
1507 | SvREFCNT_dec(HeKEY_sv(entry)); | |
8aacddc1 | 1508 | Safefree(HeKEY_hek(entry)); |
44a8e56a | 1509 | } |
1510 | else if (HvSHAREKEYS(hv)) | |
68dc0745 | 1511 | unshare_hek(HeKEY_hek(entry)); |
fde52b5c | 1512 | else |
68dc0745 | 1513 | Safefree(HeKEY_hek(entry)); |
d33b2eba | 1514 | del_HE(entry); |
79072805 LW |
1515 | } |
1516 | ||
1517 | void | |
864dbfa3 | 1518 | Perl_hv_delayfree_ent(pTHX_ HV *hv, register HE *entry) |
79072805 | 1519 | { |
97aff369 | 1520 | dVAR; |
68dc0745 | 1521 | if (!entry) |
79072805 | 1522 | return; |
bc4947fc NC |
1523 | /* SvREFCNT_inc to counter the SvREFCNT_dec in hv_free_ent */ |
1524 | sv_2mortal(SvREFCNT_inc(HeVAL(entry))); /* free between statements */ | |
68dc0745 | 1525 | if (HeKLEN(entry) == HEf_SVKEY) { |
bc4947fc | 1526 | sv_2mortal(SvREFCNT_inc(HeKEY_sv(entry))); |
44a8e56a | 1527 | } |
bc4947fc | 1528 | hv_free_ent(hv, entry); |
79072805 LW |
1529 | } |
1530 | ||
954c1994 GS |
1531 | /* |
1532 | =for apidoc hv_clear | |
1533 | ||
1534 | Clears a hash, making it empty. | |
1535 | ||
1536 | =cut | |
1537 | */ | |
1538 | ||
79072805 | 1539 | void |
864dbfa3 | 1540 | Perl_hv_clear(pTHX_ HV *hv) |
79072805 | 1541 | { |
27da23d5 | 1542 | dVAR; |
cbec9347 | 1543 | register XPVHV* xhv; |
79072805 LW |
1544 | if (!hv) |
1545 | return; | |
49293501 | 1546 | |
ecae49c0 NC |
1547 | DEBUG_A(Perl_hv_assert(aTHX_ hv)); |
1548 | ||
34c3c4e3 DM |
1549 | xhv = (XPVHV*)SvANY(hv); |
1550 | ||
7b2c381c | 1551 | if (SvREADONLY(hv) && HvARRAY(hv) != NULL) { |
34c3c4e3 | 1552 | /* restricted hash: convert all keys to placeholders */ |
b464bac0 AL |
1553 | STRLEN i; |
1554 | for (i = 0; i <= xhv->xhv_max; i++) { | |
7b2c381c | 1555 | HE *entry = (HvARRAY(hv))[i]; |
3a676441 JH |
1556 | for (; entry; entry = HeNEXT(entry)) { |
1557 | /* not already placeholder */ | |
7996736c | 1558 | if (HeVAL(entry) != &PL_sv_placeholder) { |
3a676441 | 1559 | if (HeVAL(entry) && SvREADONLY(HeVAL(entry))) { |
6136c704 | 1560 | SV* const keysv = hv_iterkeysv(entry); |
3a676441 JH |
1561 | Perl_croak(aTHX_ |
1562 | "Attempt to delete readonly key '%"SVf"' from a restricted hash", | |
1563 | keysv); | |
1564 | } | |
1565 | SvREFCNT_dec(HeVAL(entry)); | |
7996736c | 1566 | HeVAL(entry) = &PL_sv_placeholder; |
ca732855 | 1567 | HvPLACEHOLDERS(hv)++; |
3a676441 | 1568 | } |
34c3c4e3 DM |
1569 | } |
1570 | } | |
df8c6964 | 1571 | goto reset; |
49293501 MS |
1572 | } |
1573 | ||
463ee0b2 | 1574 | hfreeentries(hv); |
ca732855 | 1575 | HvPLACEHOLDERS_set(hv, 0); |
7b2c381c NC |
1576 | if (HvARRAY(hv)) |
1577 | (void)memzero(HvARRAY(hv), | |
cbec9347 | 1578 | (xhv->xhv_max+1 /* HvMAX(hv)+1 */) * sizeof(HE*)); |
a0d0e21e LW |
1579 | |
1580 | if (SvRMAGICAL(hv)) | |
1c846c1f | 1581 | mg_clear((SV*)hv); |
574c8022 | 1582 | |
19692e8d | 1583 | HvHASKFLAGS_off(hv); |
bb443f97 | 1584 | HvREHASH_off(hv); |
df8c6964 | 1585 | reset: |
b79f7545 | 1586 | if (SvOOK(hv)) { |
bfcb3514 NC |
1587 | HvEITER_set(hv, NULL); |
1588 | } | |
79072805 LW |
1589 | } |
1590 | ||
3540d4ce AB |
1591 | /* |
1592 | =for apidoc hv_clear_placeholders | |
1593 | ||
1594 | Clears any placeholders from a hash. If a restricted hash has any of its keys | |
1595 | marked as readonly and the key is subsequently deleted, the key is not actually | |
1596 | deleted but is marked by assigning it a value of &PL_sv_placeholder. This tags | |
1597 | it so it will be ignored by future operations such as iterating over the hash, | |
4cdaeff7 | 1598 | but will still allow the hash to have a value reassigned to the key at some |
3540d4ce AB |
1599 | future point. This function clears any such placeholder keys from the hash. |
1600 | See Hash::Util::lock_keys() for an example of its use. | |
1601 | ||
1602 | =cut | |
1603 | */ | |
1604 | ||
1605 | void | |
1606 | Perl_hv_clear_placeholders(pTHX_ HV *hv) | |
1607 | { | |
27da23d5 | 1608 | dVAR; |
5d88ecd7 | 1609 | I32 items = (I32)HvPLACEHOLDERS_get(hv); |
b464bac0 | 1610 | I32 i; |
d3677389 NC |
1611 | |
1612 | if (items == 0) | |
1613 | return; | |
1614 | ||
b464bac0 | 1615 | i = HvMAX(hv); |
d3677389 NC |
1616 | do { |
1617 | /* Loop down the linked list heads */ | |
6136c704 | 1618 | bool first = TRUE; |
d3677389 | 1619 | HE **oentry = &(HvARRAY(hv))[i]; |
cf6db12b | 1620 | HE *entry; |
d3677389 | 1621 | |
cf6db12b | 1622 | while ((entry = *oentry)) { |
d3677389 NC |
1623 | if (HeVAL(entry) == &PL_sv_placeholder) { |
1624 | *oentry = HeNEXT(entry); | |
1625 | if (first && !*oentry) | |
1626 | HvFILL(hv)--; /* This linked list is now empty. */ | |
2e58978b | 1627 | if (entry == HvEITER_get(hv)) |
d3677389 NC |
1628 | HvLAZYDEL_on(hv); |
1629 | else | |
1630 | hv_free_ent(hv, entry); | |
1631 | ||
1632 | if (--items == 0) { | |
1633 | /* Finished. */ | |
5d88ecd7 | 1634 | HvTOTALKEYS(hv) -= (IV)HvPLACEHOLDERS_get(hv); |
d3677389 NC |
1635 | if (HvKEYS(hv) == 0) |
1636 | HvHASKFLAGS_off(hv); | |
5d88ecd7 | 1637 | HvPLACEHOLDERS_set(hv, 0); |
d3677389 NC |
1638 | return; |
1639 | } | |
213ce8b3 NC |
1640 | } else { |
1641 | oentry = &HeNEXT(entry); | |
6136c704 | 1642 | first = FALSE; |
d3677389 NC |
1643 | } |
1644 | } | |
1645 | } while (--i >= 0); | |
1646 | /* You can't get here, hence assertion should always fail. */ | |
1647 | assert (items == 0); | |
1648 | assert (0); | |
3540d4ce AB |
1649 | } |
1650 | ||
76e3520e | 1651 | STATIC void |
cea2e8a9 | 1652 | S_hfreeentries(pTHX_ HV *hv) |
79072805 | 1653 | { |
23976bdd NC |
1654 | /* This is the array that we're going to restore */ |
1655 | HE **orig_array; | |
1656 | HEK *name; | |
1657 | int attempts = 100; | |
3abe233e | 1658 | |
a0d0e21e | 1659 | if (!HvARRAY(hv)) |
79072805 | 1660 | return; |
a0d0e21e | 1661 | |
23976bdd NC |
1662 | if (SvOOK(hv)) { |
1663 | /* If the hash is actually a symbol table with a name, look after the | |
1664 | name. */ | |
1665 | struct xpvhv_aux *iter = HvAUX(hv); | |
1666 | ||
1667 | name = iter->xhv_name; | |
1668 | iter->xhv_name = NULL; | |
1669 | } else { | |
1670 | name = NULL; | |
1671 | } | |
1672 | ||
1673 | orig_array = HvARRAY(hv); | |
1674 | /* orig_array remains unchanged throughout the loop. If after freeing all | |
1675 | the entries it turns out that one of the little blighters has triggered | |
1676 | an action that has caused HvARRAY to be re-allocated, then we set | |
1677 | array to the new HvARRAY, and try again. */ | |
1678 | ||
1679 | while (1) { | |
1680 | /* This is the one we're going to try to empty. First time round | |
1681 | it's the original array. (Hopefully there will only be 1 time | |
1682 | round) */ | |
6136c704 | 1683 | HE ** const array = HvARRAY(hv); |
7440661e | 1684 | I32 i = HvMAX(hv); |
23976bdd NC |
1685 | |
1686 | /* Because we have taken xhv_name out, the only allocated pointer | |
1687 | in the aux structure that might exist is the backreference array. | |
1688 | */ | |
1689 | ||
1690 | if (SvOOK(hv)) { | |
7440661e | 1691 | HE *entry; |
23976bdd NC |
1692 | struct xpvhv_aux *iter = HvAUX(hv); |
1693 | /* If there are weak references to this HV, we need to avoid | |
1694 | freeing them up here. In particular we need to keep the AV | |
1695 | visible as what we're deleting might well have weak references | |
1696 | back to this HV, so the for loop below may well trigger | |
1697 | the removal of backreferences from this array. */ | |
1698 | ||
1699 | if (iter->xhv_backreferences) { | |
1700 | /* So donate them to regular backref magic to keep them safe. | |
1701 | The sv_magic will increase the reference count of the AV, | |
1702 | so we need to drop it first. */ | |
5b285ea4 | 1703 | SvREFCNT_dec(iter->xhv_backreferences); |
23976bdd NC |
1704 | if (AvFILLp(iter->xhv_backreferences) == -1) { |
1705 | /* Turns out that the array is empty. Just free it. */ | |
1706 | SvREFCNT_dec(iter->xhv_backreferences); | |
1b8791d1 | 1707 | |
23976bdd NC |
1708 | } else { |
1709 | sv_magic((SV*)hv, (SV*)iter->xhv_backreferences, | |
1710 | PERL_MAGIC_backref, NULL, 0); | |
1711 | } | |
1712 | iter->xhv_backreferences = NULL; | |
5b285ea4 | 1713 | } |
86f55936 | 1714 | |
23976bdd NC |
1715 | entry = iter->xhv_eiter; /* HvEITER(hv) */ |
1716 | if (entry && HvLAZYDEL(hv)) { /* was deleted earlier? */ | |
1717 | HvLAZYDEL_off(hv); | |
1718 | hv_free_ent(hv, entry); | |
1719 | } | |
1720 | iter->xhv_riter = -1; /* HvRITER(hv) = -1 */ | |
4608196e | 1721 | iter->xhv_eiter = NULL; /* HvEITER(hv) = NULL */ |
b79f7545 | 1722 | |
23976bdd | 1723 | /* There are now no allocated pointers in the aux structure. */ |
2f86008e | 1724 | |
23976bdd NC |
1725 | SvFLAGS(hv) &= ~SVf_OOK; /* Goodbye, aux structure. */ |
1726 | /* What aux structure? */ | |
a0d0e21e | 1727 | } |
bfcb3514 | 1728 | |
23976bdd NC |
1729 | /* make everyone else think the array is empty, so that the destructors |
1730 | * called for freed entries can't recusively mess with us */ | |
1731 | HvARRAY(hv) = NULL; | |
1732 | HvFILL(hv) = 0; | |
1733 | ((XPVHV*) SvANY(hv))->xhv_keys = 0; | |
1734 | ||
7440661e NC |
1735 | |
1736 | do { | |
1737 | /* Loop down the linked list heads */ | |
1738 | HE *entry = array[i]; | |
1739 | ||
1740 | while (entry) { | |
23976bdd NC |
1741 | register HE * const oentry = entry; |
1742 | entry = HeNEXT(entry); | |
1743 | hv_free_ent(hv, oentry); | |
1744 | } | |
7440661e | 1745 | } while (--i >= 0); |
b79f7545 | 1746 | |
23976bdd NC |
1747 | /* As there are no allocated pointers in the aux structure, it's now |
1748 | safe to free the array we just cleaned up, if it's not the one we're | |
1749 | going to put back. */ | |
1750 | if (array != orig_array) { | |
1751 | Safefree(array); | |
1752 | } | |
b79f7545 | 1753 | |
23976bdd NC |
1754 | if (!HvARRAY(hv)) { |
1755 | /* Good. No-one added anything this time round. */ | |
1756 | break; | |
bfcb3514 | 1757 | } |
b79f7545 | 1758 | |
23976bdd NC |
1759 | if (SvOOK(hv)) { |
1760 | /* Someone attempted to iterate or set the hash name while we had | |
1761 | the array set to 0. We'll catch backferences on the next time | |
1762 | round the while loop. */ | |
1763 | assert(HvARRAY(hv)); | |
1b8791d1 | 1764 | |
23976bdd NC |
1765 | if (HvAUX(hv)->xhv_name) { |
1766 | unshare_hek_or_pvn(HvAUX(hv)->xhv_name, 0, 0, 0); | |
1767 | } | |
1768 | } | |
1769 | ||
1770 | if (--attempts == 0) { | |
1771 | Perl_die(aTHX_ "panic: hfreeentries failed to free hash - something is repeatedly re-creating entries"); | |
1772 | } | |
6136c704 | 1773 | } |
23976bdd NC |
1774 | |
1775 | HvARRAY(hv) = orig_array; | |
1776 | ||
1777 | /* If the hash was actually a symbol table, put the name back. */ | |
1778 | if (name) { | |
1779 | /* We have restored the original array. If name is non-NULL, then | |
1780 | the original array had an aux structure at the end. So this is | |
1781 | valid: */ | |
1782 | SvFLAGS(hv) |= SVf_OOK; | |
1783 | HvAUX(hv)->xhv_name = name; | |
1b8791d1 | 1784 | } |
79072805 LW |
1785 | } |
1786 | ||
954c1994 GS |
1787 | /* |
1788 | =for apidoc hv_undef | |
1789 | ||
1790 | Undefines the hash. | |
1791 | ||
1792 | =cut | |
1793 | */ | |
1794 | ||
79072805 | 1795 | void |
864dbfa3 | 1796 | Perl_hv_undef(pTHX_ HV *hv) |
79072805 | 1797 | { |
97aff369 | 1798 | dVAR; |
cbec9347 | 1799 | register XPVHV* xhv; |
bfcb3514 | 1800 | const char *name; |
86f55936 | 1801 | |
79072805 LW |
1802 | if (!hv) |
1803 | return; | |
ecae49c0 | 1804 | DEBUG_A(Perl_hv_assert(aTHX_ hv)); |
cbec9347 | 1805 | xhv = (XPVHV*)SvANY(hv); |
463ee0b2 | 1806 | hfreeentries(hv); |
bfcb3514 | 1807 | if ((name = HvNAME_get(hv))) { |
7e8961ec | 1808 | if(PL_stashcache) |
7423f6db | 1809 | hv_delete(PL_stashcache, name, HvNAMELEN_get(hv), G_DISCARD); |
bd61b366 | 1810 | hv_name_set(hv, NULL, 0, 0); |
85e6fe83 | 1811 | } |
b79f7545 NC |
1812 | SvFLAGS(hv) &= ~SVf_OOK; |
1813 | Safefree(HvARRAY(hv)); | |
cbec9347 | 1814 | xhv->xhv_max = 7; /* HvMAX(hv) = 7 (it's a normal hash) */ |
7b2c381c | 1815 | HvARRAY(hv) = 0; |
ca732855 | 1816 | HvPLACEHOLDERS_set(hv, 0); |
a0d0e21e LW |
1817 | |
1818 | if (SvRMAGICAL(hv)) | |
1c846c1f | 1819 | mg_clear((SV*)hv); |
79072805 LW |
1820 | } |
1821 | ||
b464bac0 | 1822 | static struct xpvhv_aux* |
5f66b61c | 1823 | S_hv_auxinit(HV *hv) { |
bfcb3514 | 1824 | struct xpvhv_aux *iter; |
b79f7545 | 1825 | char *array; |
bfcb3514 | 1826 | |
b79f7545 | 1827 | if (!HvARRAY(hv)) { |
a02a5408 | 1828 | Newxz(array, PERL_HV_ARRAY_ALLOC_BYTES(HvMAX(hv) + 1) |
b79f7545 NC |
1829 | + sizeof(struct xpvhv_aux), char); |
1830 | } else { | |
1831 | array = (char *) HvARRAY(hv); | |
1832 | Renew(array, PERL_HV_ARRAY_ALLOC_BYTES(HvMAX(hv) + 1) | |
1833 | + sizeof(struct xpvhv_aux), char); | |
1834 | } | |
1835 | HvARRAY(hv) = (HE**) array; | |
1836 | /* SvOOK_on(hv) attacks the IV flags. */ | |
1837 | SvFLAGS(hv) |= SVf_OOK; | |
1838 | iter = HvAUX(hv); | |
bfcb3514 NC |
1839 | |
1840 | iter->xhv_riter = -1; /* HvRITER(hv) = -1 */ | |
4608196e | 1841 | iter->xhv_eiter = NULL; /* HvEITER(hv) = NULL */ |
bfcb3514 | 1842 | iter->xhv_name = 0; |
86f55936 | 1843 | iter->xhv_backreferences = 0; |
bfcb3514 NC |
1844 | return iter; |
1845 | } | |
1846 | ||
954c1994 GS |
1847 | /* |
1848 | =for apidoc hv_iterinit | |
1849 | ||
1850 | Prepares a starting point to traverse a hash table. Returns the number of | |
1851 | keys in the hash (i.e. the same as C<HvKEYS(tb)>). The return value is | |
1c846c1f | 1852 | currently only meaningful for hashes without tie magic. |
954c1994 GS |
1853 | |
1854 | NOTE: Before version 5.004_65, C<hv_iterinit> used to return the number of | |
1855 | hash buckets that happen to be in use. If you still need that esoteric | |
1856 | value, you can get it through the macro C<HvFILL(tb)>. | |
1857 | ||
e16e2ff8 | 1858 | |
954c1994 GS |
1859 | =cut |
1860 | */ | |
1861 | ||
79072805 | 1862 | I32 |
864dbfa3 | 1863 | Perl_hv_iterinit(pTHX_ HV *hv) |
79072805 | 1864 | { |
aa689395 | 1865 | if (!hv) |
cea2e8a9 | 1866 | Perl_croak(aTHX_ "Bad hash"); |
bfcb3514 | 1867 | |
b79f7545 | 1868 | if (SvOOK(hv)) { |
6136c704 | 1869 | struct xpvhv_aux * const iter = HvAUX(hv); |
0bd48802 | 1870 | HE * const entry = iter->xhv_eiter; /* HvEITER(hv) */ |
bfcb3514 NC |
1871 | if (entry && HvLAZYDEL(hv)) { /* was deleted earlier? */ |
1872 | HvLAZYDEL_off(hv); | |
1873 | hv_free_ent(hv, entry); | |
1874 | } | |
1875 | iter->xhv_riter = -1; /* HvRITER(hv) = -1 */ | |
4608196e | 1876 | iter->xhv_eiter = NULL; /* HvEITER(hv) = NULL */ |
bfcb3514 | 1877 | } else { |
6136c704 | 1878 | hv_auxinit(hv); |
72940dca | 1879 | } |
bfcb3514 | 1880 | |
cbec9347 | 1881 | /* used to be xhv->xhv_fill before 5.004_65 */ |
5d88ecd7 | 1882 | return HvTOTALKEYS(hv); |
79072805 | 1883 | } |
bfcb3514 NC |
1884 | |
1885 | I32 * | |
1886 | Perl_hv_riter_p(pTHX_ HV *hv) { | |
1887 | struct xpvhv_aux *iter; | |
1888 | ||
1889 | if (!hv) | |
1890 | Perl_croak(aTHX_ "Bad hash"); | |
1891 | ||
6136c704 | 1892 | iter = SvOOK(hv) ? HvAUX(hv) : hv_auxinit(hv); |
bfcb3514 NC |
1893 | return &(iter->xhv_riter); |
1894 | } | |
1895 | ||
1896 | HE ** | |
1897 | Perl_hv_eiter_p(pTHX_ HV *hv) { | |
1898 | struct xpvhv_aux *iter; | |
1899 | ||
1900 | if (!hv) | |
1901 | Perl_croak(aTHX_ "Bad hash"); | |
1902 | ||
6136c704 | 1903 | iter = SvOOK(hv) ? HvAUX(hv) : hv_auxinit(hv); |
bfcb3514 NC |
1904 | return &(iter->xhv_eiter); |
1905 | } | |
1906 | ||
1907 | void | |
1908 | Perl_hv_riter_set(pTHX_ HV *hv, I32 riter) { | |
1909 | struct xpvhv_aux *iter; | |
1910 | ||
1911 | if (!hv) | |
1912 | Perl_croak(aTHX_ "Bad hash"); | |
1913 | ||
b79f7545 NC |
1914 | if (SvOOK(hv)) { |
1915 | iter = HvAUX(hv); | |
1916 | } else { | |
bfcb3514 NC |
1917 | if (riter == -1) |
1918 | return; | |
1919 | ||
6136c704 | 1920 | iter = hv_auxinit(hv); |
bfcb3514 NC |
1921 | } |
1922 | iter->xhv_riter = riter; | |
1923 | } | |
1924 | ||
1925 | void | |
1926 | Perl_hv_eiter_set(pTHX_ HV *hv, HE *eiter) { | |
1927 | struct xpvhv_aux *iter; | |
1928 | ||
1929 | if (!hv) | |
1930 | Perl_croak(aTHX_ "Bad hash"); | |
1931 | ||
b79f7545 NC |
1932 | if (SvOOK(hv)) { |
1933 | iter = HvAUX(hv); | |
1934 | } else { | |
bfcb3514 NC |
1935 | /* 0 is the default so don't go malloc()ing a new structure just to |
1936 | hold 0. */ | |
1937 | if (!eiter) | |
1938 | return; | |
1939 | ||
6136c704 | 1940 | iter = hv_auxinit(hv); |
bfcb3514 NC |
1941 | } |
1942 | iter->xhv_eiter = eiter; | |
1943 | } | |
1944 | ||
bfcb3514 | 1945 | void |
4164be69 | 1946 | Perl_hv_name_set(pTHX_ HV *hv, const char *name, U32 len, U32 flags) |
bfcb3514 | 1947 | { |
97aff369 | 1948 | dVAR; |
b79f7545 | 1949 | struct xpvhv_aux *iter; |
7423f6db | 1950 | U32 hash; |
46c461b5 AL |
1951 | |
1952 | PERL_UNUSED_ARG(flags); | |
bfcb3514 | 1953 | |
4164be69 NC |
1954 | if (len > I32_MAX) |
1955 | Perl_croak(aTHX_ "panic: hv name too long (%"UVuf")", (UV) len); | |
1956 | ||
b79f7545 NC |
1957 | if (SvOOK(hv)) { |
1958 | iter = HvAUX(hv); | |
7423f6db NC |
1959 | if (iter->xhv_name) { |
1960 | unshare_hek_or_pvn(iter->xhv_name, 0, 0, 0); | |
1961 | } | |
16580ff5 | 1962 | } else { |
bfcb3514 NC |
1963 | if (name == 0) |
1964 | return; | |
1965 | ||
6136c704 | 1966 | iter = hv_auxinit(hv); |
bfcb3514 | 1967 | } |
7423f6db NC |
1968 | PERL_HASH(hash, name, len); |
1969 | iter->xhv_name = name ? share_hek(name, len, hash) : 0; | |
bfcb3514 NC |
1970 | } |
1971 | ||
86f55936 NC |
1972 | AV ** |
1973 | Perl_hv_backreferences_p(pTHX_ HV *hv) { | |
6136c704 | 1974 | struct xpvhv_aux * const iter = SvOOK(hv) ? HvAUX(hv) : hv_auxinit(hv); |
96a5add6 | 1975 | PERL_UNUSED_CONTEXT; |
86f55936 NC |
1976 | return &(iter->xhv_backreferences); |
1977 | } | |
1978 | ||
1979 | void | |
1980 | Perl_hv_kill_backrefs(pTHX_ HV *hv) { | |
1981 | AV *av; | |
1982 | ||
1983 | if (!SvOOK(hv)) | |
1984 | return; | |
1985 | ||
1986 | av = HvAUX(hv)->xhv_backreferences; | |
1987 | ||
1988 | if (av) { | |
1989 | HvAUX(hv)->xhv_backreferences = 0; | |
1990 | Perl_sv_kill_backrefs(aTHX_ (SV*) hv, av); | |
1991 | } | |
1992 | } | |
1993 | ||
954c1994 | 1994 | /* |
7a7b9979 NC |
1995 | hv_iternext is implemented as a macro in hv.h |
1996 | ||
954c1994 GS |
1997 | =for apidoc hv_iternext |
1998 | ||
1999 | Returns entries from a hash iterator. See C<hv_iterinit>. | |
2000 | ||
fe7bca90 NC |
2001 | You may call C<hv_delete> or C<hv_delete_ent> on the hash entry that the |
2002 | iterator currently points to, without losing your place or invalidating your | |
2003 | iterator. Note that in this case the current entry is deleted from the hash | |
2004 | with your iterator holding the last reference to it. Your iterator is flagged | |
2005 | to free the entry on the next call to C<hv_iternext>, so you must not discard | |
2006 | your iterator immediately else the entry will leak - call C<hv_iternext> to | |
2007 | trigger the resource deallocation. | |
2008 | ||
fe7bca90 NC |
2009 | =for apidoc hv_iternext_flags |
2010 | ||
2011 | Returns entries from a hash iterator. See C<hv_iterinit> and C<hv_iternext>. | |
2012 | The C<flags> value will normally be zero; if HV_ITERNEXT_WANTPLACEHOLDERS is | |
2013 | set the placeholders keys (for restricted hashes) will be returned in addition | |
2014 | to normal keys. By default placeholders are automatically skipped over. | |
7996736c MHM |
2015 | Currently a placeholder is implemented with a value that is |
2016 | C<&Perl_sv_placeholder>. Note that the implementation of placeholders and | |
fe7bca90 NC |
2017 | restricted hashes may change, and the implementation currently is |
2018 | insufficiently abstracted for any change to be tidy. | |
e16e2ff8 | 2019 | |
fe7bca90 | 2020 | =cut |
e16e2ff8 NC |
2021 | */ |
2022 | ||
2023 | HE * | |
2024 | Perl_hv_iternext_flags(pTHX_ HV *hv, I32 flags) | |
2025 | { | |
27da23d5 | 2026 | dVAR; |
cbec9347 | 2027 | register XPVHV* xhv; |
79072805 | 2028 | register HE *entry; |
a0d0e21e | 2029 | HE *oldentry; |
463ee0b2 | 2030 | MAGIC* mg; |
bfcb3514 | 2031 | struct xpvhv_aux *iter; |
79072805 LW |
2032 | |
2033 | if (!hv) | |
cea2e8a9 | 2034 | Perl_croak(aTHX_ "Bad hash"); |
cbec9347 | 2035 | xhv = (XPVHV*)SvANY(hv); |
bfcb3514 | 2036 | |
b79f7545 | 2037 | if (!SvOOK(hv)) { |
bfcb3514 NC |
2038 | /* Too many things (well, pp_each at least) merrily assume that you can |
2039 | call iv_iternext without calling hv_iterinit, so we'll have to deal | |
2040 | with it. */ | |
2041 | hv_iterinit(hv); | |
bfcb3514 | 2042 | } |
b79f7545 | 2043 | iter = HvAUX(hv); |
bfcb3514 NC |
2044 | |
2045 | oldentry = entry = iter->xhv_eiter; /* HvEITER(hv) */ | |
463ee0b2 | 2046 | |
14befaf4 | 2047 | if ((mg = SvTIED_mg((SV*)hv, PERL_MAGIC_tied))) { |
c4420975 | 2048 | SV * const key = sv_newmortal(); |
cd1469e6 | 2049 | if (entry) { |
fde52b5c | 2050 | sv_setsv(key, HeSVKEY_force(entry)); |
cd1469e6 | 2051 | SvREFCNT_dec(HeSVKEY(entry)); /* get rid of previous key */ |
2052 | } | |
a0d0e21e | 2053 | else { |
ff68c719 | 2054 | char *k; |
bbce6d69 | 2055 | HEK *hek; |
ff68c719 | 2056 | |
cbec9347 | 2057 | /* one HE per MAGICAL hash */ |
bfcb3514 | 2058 | iter->xhv_eiter = entry = new_HE(); /* HvEITER(hv) = new_HE() */ |
4633a7c4 | 2059 | Zero(entry, 1, HE); |
a02a5408 | 2060 | Newxz(k, HEK_BASESIZE + sizeof(SV*), char); |
ff68c719 | 2061 | hek = (HEK*)k; |
2062 | HeKEY_hek(entry) = hek; | |
fde52b5c | 2063 | HeKLEN(entry) = HEf_SVKEY; |
a0d0e21e LW |
2064 | } |
2065 | magic_nextpack((SV*) hv,mg,key); | |
8aacddc1 | 2066 | if (SvOK(key)) { |
cd1469e6 | 2067 | /* force key to stay around until next time */ |
b37c2d43 | 2068 | HeSVKEY_set(entry, SvREFCNT_inc_simple_NN(key)); |
bbce6d69 | 2069 | return entry; /* beware, hent_val is not set */ |
8aacddc1 | 2070 | } |
fde52b5c | 2071 | if (HeVAL(entry)) |
2072 | SvREFCNT_dec(HeVAL(entry)); | |
ff68c719 | 2073 | Safefree(HeKEY_hek(entry)); |
d33b2eba | 2074 | del_HE(entry); |
4608196e RGS |
2075 | iter->xhv_eiter = NULL; /* HvEITER(hv) = NULL */ |
2076 | return NULL; | |
79072805 | 2077 | } |
f675dbe5 | 2078 | #ifdef DYNAMIC_ENV_FETCH /* set up %ENV for iteration */ |
03026e68 | 2079 | if (!entry && SvRMAGICAL((SV*)hv) && mg_find((SV*)hv, PERL_MAGIC_env)) { |
f675dbe5 | 2080 | prime_env_iter(); |
03026e68 JM |
2081 | #ifdef VMS |
2082 | /* The prime_env_iter() on VMS just loaded up new hash values | |
2083 | * so the iteration count needs to be reset back to the beginning | |
2084 | */ | |
2085 | hv_iterinit(hv); | |
2086 | iter = HvAUX(hv); | |
2087 | oldentry = entry = iter->xhv_eiter; /* HvEITER(hv) */ | |
2088 | #endif | |
2089 | } | |
f675dbe5 | 2090 | #endif |
463ee0b2 | 2091 | |
b79f7545 NC |
2092 | /* hv_iterint now ensures this. */ |
2093 | assert (HvARRAY(hv)); | |
2094 | ||
015a5f36 | 2095 | /* At start of hash, entry is NULL. */ |
fde52b5c | 2096 | if (entry) |
8aacddc1 | 2097 | { |
fde52b5c | 2098 | entry = HeNEXT(entry); |
e16e2ff8 NC |
2099 | if (!(flags & HV_ITERNEXT_WANTPLACEHOLDERS)) { |
2100 | /* | |
2101 | * Skip past any placeholders -- don't want to include them in | |
2102 | * any iteration. | |
2103 | */ | |
7996736c | 2104 | while (entry && HeVAL(entry) == &PL_sv_placeholder) { |
e16e2ff8 NC |
2105 | entry = HeNEXT(entry); |
2106 | } | |
8aacddc1 NIS |
2107 | } |
2108 | } | |
fde52b5c | 2109 | while (!entry) { |
015a5f36 NC |
2110 | /* OK. Come to the end of the current list. Grab the next one. */ |
2111 | ||
bfcb3514 NC |
2112 | iter->xhv_riter++; /* HvRITER(hv)++ */ |
2113 | if (iter->xhv_riter > (I32)xhv->xhv_max /* HvRITER(hv) > HvMAX(hv) */) { | |
015a5f36 | 2114 | /* There is no next one. End of the hash. */ |
bfcb3514 | 2115 | iter->xhv_riter = -1; /* HvRITER(hv) = -1 */ |
fde52b5c | 2116 | break; |
79072805 | 2117 | } |
7b2c381c | 2118 | entry = (HvARRAY(hv))[iter->xhv_riter]; |
8aacddc1 | 2119 | |
e16e2ff8 | 2120 | if (!(flags & HV_ITERNEXT_WANTPLACEHOLDERS)) { |
015a5f36 NC |
2121 | /* If we have an entry, but it's a placeholder, don't count it. |
2122 | Try the next. */ | |
7996736c | 2123 | while (entry && HeVAL(entry) == &PL_sv_placeholder) |
015a5f36 NC |
2124 | entry = HeNEXT(entry); |
2125 | } | |
2126 | /* Will loop again if this linked list starts NULL | |
2127 | (for HV_ITERNEXT_WANTPLACEHOLDERS) | |
2128 | or if we run through it and find only placeholders. */ | |
fde52b5c | 2129 | } |
79072805 | 2130 | |
72940dca | 2131 | if (oldentry && HvLAZYDEL(hv)) { /* was deleted earlier? */ |
2132 | HvLAZYDEL_off(hv); | |
68dc0745 | 2133 | hv_free_ent(hv, oldentry); |
72940dca | 2134 | } |
a0d0e21e | 2135 | |
fdcd69b6 NC |
2136 | /*if (HvREHASH(hv) && entry && !HeKREHASH(entry)) |
2137 | PerlIO_printf(PerlIO_stderr(), "Awooga %p %p\n", hv, entry);*/ | |
2138 | ||
bfcb3514 | 2139 | iter->xhv_eiter = entry; /* HvEITER(hv) = entry */ |
79072805 LW |
2140 | return entry; |
2141 | } | |
2142 | ||
954c1994 GS |
2143 | /* |
2144 | =for apidoc hv_iterkey | |
2145 | ||
2146 | Returns the key from the current position of the hash iterator. See | |
2147 | C<hv_iterinit>. | |
2148 | ||
2149 | =cut | |
2150 | */ | |
2151 | ||
79072805 | 2152 | char * |
864dbfa3 | 2153 | Perl_hv_iterkey(pTHX_ register HE *entry, I32 *retlen) |
79072805 | 2154 | { |
fde52b5c | 2155 | if (HeKLEN(entry) == HEf_SVKEY) { |
fb73857a | 2156 | STRLEN len; |
0bd48802 | 2157 | char * const p = SvPV(HeKEY_sv(entry), len); |
fb73857a | 2158 | *retlen = len; |
2159 | return p; | |
fde52b5c | 2160 | } |
2161 | else { | |
2162 | *retlen = HeKLEN(entry); | |
2163 | return HeKEY(entry); | |
2164 | } | |
2165 | } | |
2166 | ||
2167 | /* unlike hv_iterval(), this always returns a mortal copy of the key */ | |
954c1994 GS |
2168 | /* |
2169 | =for apidoc hv_iterkeysv | |
2170 | ||
2171 | Returns the key as an C<SV*> from the current position of the hash | |
2172 | iterator. The return value will always be a mortal copy of the key. Also | |
2173 | see C<hv_iterinit>. | |
2174 | ||
2175 | =cut | |
2176 | */ | |
2177 | ||
fde52b5c | 2178 | SV * |
864dbfa3 | 2179 | Perl_hv_iterkeysv(pTHX_ register HE *entry) |
fde52b5c | 2180 | { |
c1b02ed8 | 2181 | return sv_2mortal(newSVhek(HeKEY_hek(entry))); |
79072805 LW |
2182 | } |
2183 | ||
954c1994 GS |
2184 | /* |
2185 | =for apidoc hv_iterval | |
2186 | ||
2187 | Returns the value from the current position of the hash iterator. See | |
2188 | C<hv_iterkey>. | |
2189 | ||
2190 | =cut | |
2191 | */ | |
2192 | ||
79072805 | 2193 | SV * |
864dbfa3 | 2194 | Perl_hv_iterval(pTHX_ HV *hv, register HE *entry) |
79072805 | 2195 | { |
8990e307 | 2196 | if (SvRMAGICAL(hv)) { |
14befaf4 | 2197 | if (mg_find((SV*)hv, PERL_MAGIC_tied)) { |
c4420975 | 2198 | SV* const sv = sv_newmortal(); |
bbce6d69 | 2199 | if (HeKLEN(entry) == HEf_SVKEY) |
2200 | mg_copy((SV*)hv, sv, (char*)HeKEY_sv(entry), HEf_SVKEY); | |
a3b680e6 AL |
2201 | else |
2202 | mg_copy((SV*)hv, sv, HeKEY(entry), HeKLEN(entry)); | |
463ee0b2 LW |
2203 | return sv; |
2204 | } | |
79072805 | 2205 | } |
fde52b5c | 2206 | return HeVAL(entry); |
79072805 LW |
2207 | } |
2208 | ||
954c1994 GS |
2209 | /* |
2210 | =for apidoc hv_iternextsv | |
2211 | ||
2212 | Performs an C<hv_iternext>, C<hv_iterkey>, and C<hv_iterval> in one | |
2213 | operation. | |
2214 | ||
2215 | =cut | |
2216 | */ | |
2217 | ||
a0d0e21e | 2218 | SV * |
864dbfa3 | 2219 | Perl_hv_iternextsv(pTHX_ HV *hv, char **key, I32 *retlen) |
a0d0e21e | 2220 | { |
0bd48802 AL |
2221 | HE * const he = hv_iternext_flags(hv, 0); |
2222 | ||
2223 | if (!he) | |
a0d0e21e LW |
2224 | return NULL; |
2225 | *key = hv_iterkey(he, retlen); | |
2226 | return hv_iterval(hv, he); | |
2227 | } | |
2228 | ||
954c1994 | 2229 | /* |
bc5cdc23 NC |
2230 | |
2231 | Now a macro in hv.h | |
2232 | ||
954c1994 GS |
2233 | =for apidoc hv_magic |
2234 | ||
2235 | Adds magic to a hash. See C<sv_magic>. | |
2236 | ||
2237 | =cut | |
2238 | */ | |
2239 | ||
bbce6d69 | 2240 | /* possibly free a shared string if no one has access to it |
fde52b5c | 2241 | * len and hash must both be valid for str. |
2242 | */ | |
bbce6d69 | 2243 | void |
864dbfa3 | 2244 | Perl_unsharepvn(pTHX_ const char *str, I32 len, U32 hash) |
fde52b5c | 2245 | { |
19692e8d NC |
2246 | unshare_hek_or_pvn (NULL, str, len, hash); |
2247 | } | |
2248 | ||
2249 | ||
2250 | void | |
2251 | Perl_unshare_hek(pTHX_ HEK *hek) | |
2252 | { | |
2253 | unshare_hek_or_pvn(hek, NULL, 0, 0); | |
2254 | } | |
2255 | ||
2256 | /* possibly free a shared string if no one has access to it | |
2257 | hek if non-NULL takes priority over the other 3, else str, len and hash | |
2258 | are used. If so, len and hash must both be valid for str. | |
2259 | */ | |
df132699 | 2260 | STATIC void |
97ddebaf | 2261 | S_unshare_hek_or_pvn(pTHX_ const HEK *hek, const char *str, I32 len, U32 hash) |
19692e8d | 2262 | { |
97aff369 | 2263 | dVAR; |
cbec9347 | 2264 | register XPVHV* xhv; |
20454177 | 2265 | HE *entry; |
fde52b5c | 2266 | register HE **oentry; |
45d1cc86 | 2267 | HE **first; |
a3b680e6 | 2268 | bool found = 0; |
c3654f1a | 2269 | bool is_utf8 = FALSE; |
19692e8d | 2270 | int k_flags = 0; |
aec46f14 | 2271 | const char * const save = str; |
cbbf8932 | 2272 | struct shared_he *he = NULL; |
c3654f1a | 2273 | |
19692e8d | 2274 | if (hek) { |
cbae3960 NC |
2275 | /* Find the shared he which is just before us in memory. */ |
2276 | he = (struct shared_he *)(((char *)hek) | |
2277 | - STRUCT_OFFSET(struct shared_he, | |
2278 | shared_he_hek)); | |
2279 | ||
2280 | /* Assert that the caller passed us a genuine (or at least consistent) | |
2281 | shared hek */ | |
2282 | assert (he->shared_he_he.hent_hek == hek); | |
29404ae0 NC |
2283 | |
2284 | LOCK_STRTAB_MUTEX; | |
de616631 NC |
2285 | if (he->shared_he_he.he_valu.hent_refcount - 1) { |
2286 | --he->shared_he_he.he_valu.hent_refcount; | |
29404ae0 NC |
2287 | UNLOCK_STRTAB_MUTEX; |
2288 | return; | |
2289 | } | |
2290 | UNLOCK_STRTAB_MUTEX; | |
2291 | ||
19692e8d NC |
2292 | hash = HEK_HASH(hek); |
2293 | } else if (len < 0) { | |
2294 | STRLEN tmplen = -len; | |
2295 | is_utf8 = TRUE; | |
2296 | /* See the note in hv_fetch(). --jhi */ | |
2297 | str = (char*)bytes_from_utf8((U8*)str, &tmplen, &is_utf8); | |
2298 | len = tmplen; | |
2299 | if (is_utf8) | |
2300 | k_flags = HVhek_UTF8; | |
2301 | if (str != save) | |
2302 | k_flags |= HVhek_WASUTF8 | HVhek_FREEKEY; | |
c3654f1a | 2303 | } |
1c846c1f | 2304 | |
de616631 | 2305 | /* what follows was the moral equivalent of: |
6b88bc9c | 2306 | if ((Svp = hv_fetch(PL_strtab, tmpsv, FALSE, hash))) { |
a0714e2c | 2307 | if (--*Svp == NULL) |
6b88bc9c | 2308 | hv_delete(PL_strtab, str, len, G_DISCARD, hash); |
bbce6d69 | 2309 | } */ |
cbec9347 | 2310 | xhv = (XPVHV*)SvANY(PL_strtab); |
fde52b5c | 2311 | /* assert(xhv_array != 0) */ |
5f08fbcd | 2312 | LOCK_STRTAB_MUTEX; |
45d1cc86 | 2313 | first = oentry = &(HvARRAY(PL_strtab))[hash & (I32) HvMAX(PL_strtab)]; |
6c1b96a1 NC |
2314 | if (he) { |
2315 | const HE *const he_he = &(he->shared_he_he); | |
45d1cc86 | 2316 | for (entry = *oentry; entry; oentry = &HeNEXT(entry), entry = *oentry) { |
6c1b96a1 | 2317 | if (entry != he_he) |
19692e8d NC |
2318 | continue; |
2319 | found = 1; | |
2320 | break; | |
2321 | } | |
2322 | } else { | |
35a4481c | 2323 | const int flags_masked = k_flags & HVhek_MASK; |
45d1cc86 | 2324 | for (entry = *oentry; entry; oentry = &HeNEXT(entry), entry = *oentry) { |
19692e8d NC |
2325 | if (HeHASH(entry) != hash) /* strings can't be equal */ |
2326 | continue; | |
2327 | if (HeKLEN(entry) != len) | |
2328 | continue; | |
2329 | if (HeKEY(entry) != str && memNE(HeKEY(entry),str,len)) /* is this it? */ | |
2330 | continue; | |
2331 | if (HeKFLAGS(entry) != flags_masked) | |
2332 | continue; | |
2333 | found = 1; | |
2334 | break; | |
2335 | } | |
2336 | } | |
2337 | ||
2338 | if (found) { | |
de616631 | 2339 | if (--he->shared_he_he.he_valu.hent_refcount == 0) { |
19692e8d | 2340 | *oentry = HeNEXT(entry); |
45d1cc86 NC |
2341 | if (!*first) { |
2342 | /* There are now no entries in our slot. */ | |
19692e8d | 2343 | xhv->xhv_fill--; /* HvFILL(hv)-- */ |
45d1cc86 | 2344 | } |
cbae3960 | 2345 | Safefree(entry); |
19692e8d NC |
2346 | xhv->xhv_keys--; /* HvKEYS(hv)-- */ |
2347 | } | |
fde52b5c | 2348 | } |
19692e8d | 2349 | |
333f433b | 2350 | UNLOCK_STRTAB_MUTEX; |
411caa50 | 2351 | if (!found && ckWARN_d(WARN_INTERNAL)) |
19692e8d | 2352 | Perl_warner(aTHX_ packWARN(WARN_INTERNAL), |
472d47bc SB |
2353 | "Attempt to free non-existent shared string '%s'%s" |
2354 | pTHX__FORMAT, | |
19692e8d | 2355 | hek ? HEK_KEY(hek) : str, |
472d47bc | 2356 | ((k_flags & HVhek_UTF8) ? " (utf8)" : "") pTHX__VALUE); |
19692e8d NC |
2357 | if (k_flags & HVhek_FREEKEY) |
2358 | Safefree(str); | |
fde52b5c | 2359 | } |
2360 | ||
bbce6d69 | 2361 | /* get a (constant) string ptr from the global string table |
2362 | * string will get added if it is not already there. | |
fde52b5c | 2363 | * len and hash must both be valid for str. |
2364 | */ | |
bbce6d69 | 2365 | HEK * |
864dbfa3 | 2366 | Perl_share_hek(pTHX_ const char *str, I32 len, register U32 hash) |
fde52b5c | 2367 | { |
da58a35d | 2368 | bool is_utf8 = FALSE; |
19692e8d | 2369 | int flags = 0; |
aec46f14 | 2370 | const char * const save = str; |
da58a35d JH |
2371 | |
2372 | if (len < 0) { | |
77caf834 | 2373 | STRLEN tmplen = -len; |
da58a35d | 2374 | is_utf8 = TRUE; |
77caf834 JH |
2375 | /* See the note in hv_fetch(). --jhi */ |
2376 | str = (char*)bytes_from_utf8((U8*)str, &tmplen, &is_utf8); | |
2377 | len = tmplen; | |
19692e8d NC |
2378 | /* If we were able to downgrade here, then than means that we were passed |
2379 | in a key which only had chars 0-255, but was utf8 encoded. */ | |
2380 | if (is_utf8) | |
2381 | flags = HVhek_UTF8; | |
2382 | /* If we found we were able to downgrade the string to bytes, then | |
2383 | we should flag that it needs upgrading on keys or each. Also flag | |
2384 | that we need share_hek_flags to free the string. */ | |
2385 | if (str != save) | |
2386 | flags |= HVhek_WASUTF8 | HVhek_FREEKEY; | |
2387 | } | |
2388 | ||
6e838c70 | 2389 | return share_hek_flags (str, len, hash, flags); |
19692e8d NC |
2390 | } |
2391 | ||
6e838c70 | 2392 | STATIC HEK * |
19692e8d NC |
2393 | S_share_hek_flags(pTHX_ const char *str, I32 len, register U32 hash, int flags) |
2394 | { | |
97aff369 | 2395 | dVAR; |
19692e8d | 2396 | register HE *entry; |
35a4481c | 2397 | const int flags_masked = flags & HVhek_MASK; |
263cb4a6 | 2398 | const U32 hindex = hash & (I32) HvMAX(PL_strtab); |
bbce6d69 | 2399 | |
fde52b5c | 2400 | /* what follows is the moral equivalent of: |
1c846c1f | 2401 | |
6b88bc9c | 2402 | if (!(Svp = hv_fetch(PL_strtab, str, len, FALSE))) |
a0714e2c | 2403 | hv_store(PL_strtab, str, len, NULL, hash); |
fdcd69b6 NC |
2404 | |
2405 | Can't rehash the shared string table, so not sure if it's worth | |
2406 | counting the number of entries in the linked list | |
bbce6d69 | 2407 | */ |
1b6737cc | 2408 | register XPVHV * const xhv = (XPVHV*)SvANY(PL_strtab); |
fde52b5c | 2409 | /* assert(xhv_array != 0) */ |
5f08fbcd | 2410 | LOCK_STRTAB_MUTEX; |
263cb4a6 NC |
2411 | entry = (HvARRAY(PL_strtab))[hindex]; |
2412 | for (;entry; entry = HeNEXT(entry)) { | |
fde52b5c | 2413 | if (HeHASH(entry) != hash) /* strings can't be equal */ |
2414 | continue; | |
2415 | if (HeKLEN(entry) != len) | |
2416 | continue; | |
1c846c1f | 2417 | if (HeKEY(entry) != str && memNE(HeKEY(entry),str,len)) /* is this it? */ |
fde52b5c | 2418 | continue; |
19692e8d | 2419 | if (HeKFLAGS(entry) != flags_masked) |
c3654f1a | 2420 | continue; |
fde52b5c | 2421 | break; |
2422 | } | |
263cb4a6 NC |
2423 | |
2424 | if (!entry) { | |
45d1cc86 NC |
2425 | /* What used to be head of the list. |
2426 | If this is NULL, then we're the first entry for this slot, which | |
2427 | means we need to increate fill. */ | |
cbae3960 NC |
2428 | struct shared_he *new_entry; |
2429 | HEK *hek; | |
2430 | char *k; | |
263cb4a6 NC |
2431 | HE **const head = &HvARRAY(PL_strtab)[hindex]; |
2432 | HE *const next = *head; | |
cbae3960 NC |
2433 | |
2434 | /* We don't actually store a HE from the arena and a regular HEK. | |
2435 | Instead we allocate one chunk of memory big enough for both, | |
2436 | and put the HEK straight after the HE. This way we can find the | |
2437 | HEK directly from the HE. | |
2438 | */ | |
2439 | ||
a02a5408 | 2440 | Newx(k, STRUCT_OFFSET(struct shared_he, |
cbae3960 NC |
2441 | shared_he_hek.hek_key[0]) + len + 2, char); |
2442 | new_entry = (struct shared_he *)k; | |
2443 | entry = &(new_entry->shared_he_he); | |
2444 | hek = &(new_entry->shared_he_hek); | |
2445 | ||
2446 | Copy(str, HEK_KEY(hek), len, char); | |
2447 | HEK_KEY(hek)[len] = 0; | |
2448 | HEK_LEN(hek) = len; | |
2449 | HEK_HASH(hek) = hash; | |
2450 | HEK_FLAGS(hek) = (unsigned char)flags_masked; | |
2451 | ||
2452 | /* Still "point" to the HEK, so that other code need not know what | |
2453 | we're up to. */ | |
2454 | HeKEY_hek(entry) = hek; | |
de616631 | 2455 | entry->he_valu.hent_refcount = 0; |
263cb4a6 NC |
2456 | HeNEXT(entry) = next; |
2457 | *head = entry; | |
cbae3960 | 2458 | |
cbec9347 | 2459 | xhv->xhv_keys++; /* HvKEYS(hv)++ */ |
263cb4a6 | 2460 | if (!next) { /* initial entry? */ |
cbec9347 | 2461 | xhv->xhv_fill++; /* HvFILL(hv)++ */ |
4c9cc595 | 2462 | } else if (xhv->xhv_keys > (IV)xhv->xhv_max /* HvKEYS(hv) > HvMAX(hv) */) { |
cbec9347 | 2463 | hsplit(PL_strtab); |
bbce6d69 | 2464 | } |
2465 | } | |
2466 | ||
de616631 | 2467 | ++entry->he_valu.hent_refcount; |
5f08fbcd | 2468 | UNLOCK_STRTAB_MUTEX; |
19692e8d NC |
2469 | |
2470 | if (flags & HVhek_FREEKEY) | |
f9a63242 | 2471 | Safefree(str); |
19692e8d | 2472 | |
6e838c70 | 2473 | return HeKEY_hek(entry); |
fde52b5c | 2474 | } |
ecae49c0 | 2475 | |
ca732855 NC |
2476 | I32 * |
2477 | Perl_hv_placeholders_p(pTHX_ HV *hv) | |
2478 | { | |
2479 | dVAR; | |
2480 | MAGIC *mg = mg_find((SV*)hv, PERL_MAGIC_rhash); | |
2481 | ||
2482 | if (!mg) { | |
2483 | mg = sv_magicext((SV*)hv, 0, PERL_MAGIC_rhash, 0, 0, 0); | |
2484 | ||
2485 | if (!mg) { | |
2486 | Perl_die(aTHX_ "panic: hv_placeholders_p"); | |
2487 | } | |
2488 | } | |
2489 | return &(mg->mg_len); | |
2490 | } | |
2491 | ||
2492 | ||
2493 | I32 | |
2494 | Perl_hv_placeholders_get(pTHX_ HV *hv) | |
2495 | { | |
2496 | dVAR; | |
b464bac0 | 2497 | MAGIC * const mg = mg_find((SV*)hv, PERL_MAGIC_rhash); |
ca732855 NC |
2498 | |
2499 | return mg ? mg->mg_len : 0; | |
2500 | } | |
2501 | ||
2502 | void | |
ac1e784a | 2503 | Perl_hv_placeholders_set(pTHX_ HV *hv, I32 ph) |
ca732855 NC |
2504 | { |
2505 | dVAR; | |
b464bac0 | 2506 | MAGIC * const mg = mg_find((SV*)hv, PERL_MAGIC_rhash); |
ca732855 NC |
2507 | |
2508 | if (mg) { | |
2509 | mg->mg_len = ph; | |
2510 | } else if (ph) { | |
2511 | if (!sv_magicext((SV*)hv, 0, PERL_MAGIC_rhash, 0, 0, ph)) | |
2512 | Perl_die(aTHX_ "panic: hv_placeholders_set"); | |
2513 | } | |
2514 | /* else we don't need to add magic to record 0 placeholders. */ | |
2515 | } | |
ecae49c0 NC |
2516 | |
2517 | /* | |
2518 | =for apidoc hv_assert | |
2519 | ||
2520 | Check that a hash is in an internally consistent state. | |
2521 | ||
2522 | =cut | |
2523 | */ | |
2524 | ||
2525 | void | |
2526 | Perl_hv_assert(pTHX_ HV *hv) | |
2527 | { | |
27da23d5 | 2528 | dVAR; |
ecae49c0 NC |
2529 | HE* entry; |
2530 | int withflags = 0; | |
2531 | int placeholders = 0; | |
2532 | int real = 0; | |
2533 | int bad = 0; | |
bfcb3514 NC |
2534 | const I32 riter = HvRITER_get(hv); |
2535 | HE *eiter = HvEITER_get(hv); | |
ecae49c0 NC |
2536 | |
2537 | (void)hv_iterinit(hv); | |
2538 | ||
2539 | while ((entry = hv_iternext_flags(hv, HV_ITERNEXT_WANTPLACEHOLDERS))) { | |
2540 | /* sanity check the values */ | |
2541 | if (HeVAL(entry) == &PL_sv_placeholder) { | |
2542 | placeholders++; | |
2543 | } else { | |
2544 | real++; | |
2545 | } | |
2546 | /* sanity check the keys */ | |
2547 | if (HeSVKEY(entry)) { | |
bb263b4e | 2548 | /*EMPTY*/ /* Don't know what to check on SV keys. */ |
ecae49c0 NC |
2549 | } else if (HeKUTF8(entry)) { |
2550 | withflags++; | |
2551 | if (HeKWASUTF8(entry)) { | |
2552 | PerlIO_printf(Perl_debug_log, | |
2553 | "hash key has both WASUFT8 and UTF8: '%.*s'\n", | |
2554 | (int) HeKLEN(entry), HeKEY(entry)); | |
2555 | bad = 1; | |
2556 | } | |
2557 | } else if (HeKWASUTF8(entry)) { | |
2558 | withflags++; | |
2559 | } | |
2560 | } | |
2561 | if (!SvTIED_mg((SV*)hv, PERL_MAGIC_tied)) { | |
2562 | if (HvUSEDKEYS(hv) != real) { | |
2563 | PerlIO_printf(Perl_debug_log, "Count %d key(s), but hash reports %d\n", | |
2564 | (int) real, (int) HvUSEDKEYS(hv)); | |
2565 | bad = 1; | |
2566 | } | |
5d88ecd7 | 2567 | if (HvPLACEHOLDERS_get(hv) != placeholders) { |
ecae49c0 NC |
2568 | PerlIO_printf(Perl_debug_log, |
2569 | "Count %d placeholder(s), but hash reports %d\n", | |
5d88ecd7 | 2570 | (int) placeholders, (int) HvPLACEHOLDERS_get(hv)); |
ecae49c0 NC |
2571 | bad = 1; |
2572 | } | |
2573 | } | |
2574 | if (withflags && ! HvHASKFLAGS(hv)) { | |
2575 | PerlIO_printf(Perl_debug_log, | |
2576 | "Hash has HASKFLAGS off but I count %d key(s) with flags\n", | |
2577 | withflags); | |
2578 | bad = 1; | |
2579 | } | |
2580 | if (bad) { | |
2581 | sv_dump((SV *)hv); | |
2582 | } | |
bfcb3514 NC |
2583 | HvRITER_set(hv, riter); /* Restore hash iterator state */ |
2584 | HvEITER_set(hv, eiter); | |
ecae49c0 | 2585 | } |
af3babe4 NC |
2586 | |
2587 | /* | |
2588 | * Local variables: | |
2589 | * c-indentation-style: bsd | |
2590 | * c-basic-offset: 4 | |
2591 | * indent-tabs-mode: t | |
2592 | * End: | |
2593 | * | |
37442d52 RGS |
2594 | * ex: set ts=8 sts=4 sw=4 noet: |
2595 | */ |