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 | 74 | he = *root; |
ce3e5c45 | 75 | assert(he); |
6a93a7e5 | 76 | *root = HeNEXT(he); |
333f433b DG |
77 | UNLOCK_SV_MUTEX; |
78 | return he; | |
4633a7c4 LW |
79 | } |
80 | ||
c941fb51 NC |
81 | #define new_HE() new_he() |
82 | #define del_HE(p) \ | |
83 | STMT_START { \ | |
84 | LOCK_SV_MUTEX; \ | |
6a93a7e5 NC |
85 | HeNEXT(p) = (HE*)(PL_body_roots[HE_SVSLOT]); \ |
86 | PL_body_roots[HE_SVSLOT] = p; \ | |
c941fb51 NC |
87 | UNLOCK_SV_MUTEX; \ |
88 | } STMT_END | |
d33b2eba | 89 | |
d33b2eba | 90 | |
d33b2eba GS |
91 | |
92 | #endif | |
93 | ||
76e3520e | 94 | STATIC HEK * |
5f66b61c | 95 | S_save_hek_flags(const char *str, I32 len, U32 hash, int flags) |
bbce6d69 | 96 | { |
35a4481c | 97 | const int flags_masked = flags & HVhek_MASK; |
bbce6d69 | 98 | char *k; |
99 | register HEK *hek; | |
1c846c1f | 100 | |
a02a5408 | 101 | Newx(k, HEK_BASESIZE + len + 2, char); |
bbce6d69 | 102 | hek = (HEK*)k; |
ff68c719 | 103 | Copy(str, HEK_KEY(hek), len, char); |
e05949c7 | 104 | HEK_KEY(hek)[len] = 0; |
ff68c719 | 105 | HEK_LEN(hek) = len; |
106 | HEK_HASH(hek) = hash; | |
45e34800 | 107 | HEK_FLAGS(hek) = (unsigned char)flags_masked | HVhek_UNSHARED; |
dcf933a4 NC |
108 | |
109 | if (flags & HVhek_FREEKEY) | |
110 | Safefree(str); | |
bbce6d69 | 111 | return hek; |
112 | } | |
113 | ||
4a31713e | 114 | /* free the pool of temporary HE/HEK pairs returned by hv_fetch_ent |
dd28f7bb DM |
115 | * for tied hashes */ |
116 | ||
117 | void | |
118 | Perl_free_tied_hv_pool(pTHX) | |
119 | { | |
97aff369 | 120 | dVAR; |
dd28f7bb DM |
121 | HE *he = PL_hv_fetch_ent_mh; |
122 | while (he) { | |
9d4ba2ae | 123 | HE * const ohe = he; |
dd28f7bb | 124 | Safefree(HeKEY_hek(he)); |
dd28f7bb DM |
125 | he = HeNEXT(he); |
126 | del_HE(ohe); | |
127 | } | |
4608196e | 128 | PL_hv_fetch_ent_mh = NULL; |
dd28f7bb DM |
129 | } |
130 | ||
d18c6117 | 131 | #if defined(USE_ITHREADS) |
0bff533c NC |
132 | HEK * |
133 | Perl_hek_dup(pTHX_ HEK *source, CLONE_PARAMS* param) | |
134 | { | |
658b4a4a | 135 | HEK *shared = (HEK*)ptr_table_fetch(PL_ptr_table, source); |
9d4ba2ae AL |
136 | |
137 | PERL_UNUSED_ARG(param); | |
0bff533c NC |
138 | |
139 | if (shared) { | |
140 | /* We already shared this hash key. */ | |
454f1e26 | 141 | (void)share_hek_hek(shared); |
0bff533c NC |
142 | } |
143 | else { | |
658b4a4a | 144 | shared |
6e838c70 NC |
145 | = share_hek_flags(HEK_KEY(source), HEK_LEN(source), |
146 | HEK_HASH(source), HEK_FLAGS(source)); | |
658b4a4a | 147 | ptr_table_store(PL_ptr_table, source, shared); |
0bff533c | 148 | } |
658b4a4a | 149 | return shared; |
0bff533c NC |
150 | } |
151 | ||
d18c6117 | 152 | HE * |
5c4138a0 | 153 | Perl_he_dup(pTHX_ const HE *e, bool shared, CLONE_PARAMS* param) |
d18c6117 GS |
154 | { |
155 | HE *ret; | |
156 | ||
157 | if (!e) | |
4608196e | 158 | return NULL; |
7766f137 GS |
159 | /* look for it in the table first */ |
160 | ret = (HE*)ptr_table_fetch(PL_ptr_table, e); | |
161 | if (ret) | |
162 | return ret; | |
163 | ||
164 | /* create anew and remember what it is */ | |
d33b2eba | 165 | ret = new_HE(); |
7766f137 GS |
166 | ptr_table_store(PL_ptr_table, e, ret); |
167 | ||
d2d73c3e | 168 | HeNEXT(ret) = he_dup(HeNEXT(e),shared, param); |
dd28f7bb DM |
169 | if (HeKLEN(e) == HEf_SVKEY) { |
170 | char *k; | |
a02a5408 | 171 | Newx(k, HEK_BASESIZE + sizeof(SV*), char); |
dd28f7bb | 172 | HeKEY_hek(ret) = (HEK*)k; |
d2d73c3e | 173 | HeKEY_sv(ret) = SvREFCNT_inc(sv_dup(HeKEY_sv(e), param)); |
dd28f7bb | 174 | } |
c21d1a0f | 175 | else if (shared) { |
0bff533c NC |
176 | /* This is hek_dup inlined, which seems to be important for speed |
177 | reasons. */ | |
1b6737cc | 178 | HEK * const source = HeKEY_hek(e); |
658b4a4a | 179 | HEK *shared = (HEK*)ptr_table_fetch(PL_ptr_table, source); |
c21d1a0f NC |
180 | |
181 | if (shared) { | |
182 | /* We already shared this hash key. */ | |
454f1e26 | 183 | (void)share_hek_hek(shared); |
c21d1a0f NC |
184 | } |
185 | else { | |
658b4a4a | 186 | shared |
6e838c70 NC |
187 | = share_hek_flags(HEK_KEY(source), HEK_LEN(source), |
188 | HEK_HASH(source), HEK_FLAGS(source)); | |
658b4a4a | 189 | ptr_table_store(PL_ptr_table, source, shared); |
c21d1a0f | 190 | } |
658b4a4a | 191 | HeKEY_hek(ret) = shared; |
c21d1a0f | 192 | } |
d18c6117 | 193 | else |
19692e8d NC |
194 | HeKEY_hek(ret) = save_hek_flags(HeKEY(e), HeKLEN(e), HeHASH(e), |
195 | HeKFLAGS(e)); | |
d2d73c3e | 196 | HeVAL(ret) = SvREFCNT_inc(sv_dup(HeVAL(e), param)); |
d18c6117 GS |
197 | return ret; |
198 | } | |
199 | #endif /* USE_ITHREADS */ | |
200 | ||
1b1f1335 | 201 | static void |
2393f1b9 JH |
202 | S_hv_notallowed(pTHX_ int flags, const char *key, I32 klen, |
203 | const char *msg) | |
1b1f1335 | 204 | { |
1b6737cc | 205 | SV * const sv = sv_newmortal(); |
19692e8d | 206 | if (!(flags & HVhek_FREEKEY)) { |
1b1f1335 NIS |
207 | sv_setpvn(sv, key, klen); |
208 | } | |
209 | else { | |
210 | /* Need to free saved eventually assign to mortal SV */ | |
34c3c4e3 | 211 | /* XXX is this line an error ???: SV *sv = sv_newmortal(); */ |
1b1f1335 NIS |
212 | sv_usepvn(sv, (char *) key, klen); |
213 | } | |
19692e8d | 214 | if (flags & HVhek_UTF8) { |
1b1f1335 NIS |
215 | SvUTF8_on(sv); |
216 | } | |
c8cd6465 | 217 | Perl_croak(aTHX_ msg, sv); |
1b1f1335 NIS |
218 | } |
219 | ||
fde52b5c | 220 | /* (klen == HEf_SVKEY) is special for MAGICAL hv entries, meaning key slot |
221 | * contains an SV* */ | |
222 | ||
34a6f7b4 NC |
223 | #define HV_FETCH_ISSTORE 0x01 |
224 | #define HV_FETCH_ISEXISTS 0x02 | |
225 | #define HV_FETCH_LVALUE 0x04 | |
226 | #define HV_FETCH_JUST_SV 0x08 | |
227 | ||
228 | /* | |
229 | =for apidoc hv_store | |
230 | ||
231 | Stores an SV in a hash. The hash key is specified as C<key> and C<klen> is | |
232 | the length of the key. The C<hash> parameter is the precomputed hash | |
233 | value; if it is zero then Perl will compute it. The return value will be | |
234 | NULL if the operation failed or if the value did not need to be actually | |
235 | stored within the hash (as in the case of tied hashes). Otherwise it can | |
236 | be dereferenced to get the original C<SV*>. Note that the caller is | |
237 | responsible for suitably incrementing the reference count of C<val> before | |
238 | the call, and decrementing it if the function returned NULL. Effectively | |
239 | a successful hv_store takes ownership of one reference to C<val>. This is | |
240 | usually what you want; a newly created SV has a reference count of one, so | |
241 | if all your code does is create SVs then store them in a hash, hv_store | |
242 | will own the only reference to the new SV, and your code doesn't need to do | |
243 | anything further to tidy up. hv_store is not implemented as a call to | |
244 | hv_store_ent, and does not create a temporary SV for the key, so if your | |
245 | key data is not already in SV form then use hv_store in preference to | |
246 | hv_store_ent. | |
247 | ||
248 | See L<perlguts/"Understanding the Magic of Tied Hashes and Arrays"> for more | |
249 | information on how to use this function on tied hashes. | |
250 | ||
251 | =cut | |
252 | */ | |
253 | ||
254 | SV** | |
255 | Perl_hv_store(pTHX_ HV *hv, const char *key, I32 klen_i32, SV *val, U32 hash) | |
256 | { | |
257 | HE *hek; | |
258 | STRLEN klen; | |
259 | int flags; | |
260 | ||
261 | if (klen_i32 < 0) { | |
262 | klen = -klen_i32; | |
263 | flags = HVhek_UTF8; | |
264 | } else { | |
265 | klen = klen_i32; | |
266 | flags = 0; | |
267 | } | |
268 | hek = hv_fetch_common (hv, NULL, key, klen, flags, | |
52d01cc2 | 269 | (HV_FETCH_ISSTORE|HV_FETCH_JUST_SV), val, hash); |
34a6f7b4 NC |
270 | return hek ? &HeVAL(hek) : NULL; |
271 | } | |
272 | ||
fabdb6c0 | 273 | /* XXX This looks like an ideal candidate to inline */ |
34a6f7b4 NC |
274 | SV** |
275 | Perl_hv_store_flags(pTHX_ HV *hv, const char *key, I32 klen, SV *val, | |
276 | register U32 hash, int flags) | |
277 | { | |
9d4ba2ae | 278 | HE * const hek = hv_fetch_common (hv, NULL, key, klen, flags, |
34a6f7b4 NC |
279 | (HV_FETCH_ISSTORE|HV_FETCH_JUST_SV), val, hash); |
280 | return hek ? &HeVAL(hek) : NULL; | |
281 | } | |
282 | ||
283 | /* | |
284 | =for apidoc hv_store_ent | |
285 | ||
286 | Stores C<val> in a hash. The hash key is specified as C<key>. The C<hash> | |
287 | parameter is the precomputed hash value; if it is zero then Perl will | |
288 | compute it. The return value is the new hash entry so created. It will be | |
289 | NULL if the operation failed or if the value did not need to be actually | |
290 | stored within the hash (as in the case of tied hashes). Otherwise the | |
291 | contents of the return value can be accessed using the C<He?> macros | |
292 | described here. Note that the caller is responsible for suitably | |
293 | incrementing the reference count of C<val> before the call, and | |
294 | decrementing it if the function returned NULL. Effectively a successful | |
295 | hv_store_ent takes ownership of one reference to C<val>. This is | |
296 | usually what you want; a newly created SV has a reference count of one, so | |
297 | if all your code does is create SVs then store them in a hash, hv_store | |
298 | will own the only reference to the new SV, and your code doesn't need to do | |
299 | anything further to tidy up. Note that hv_store_ent only reads the C<key>; | |
300 | unlike C<val> it does not take ownership of it, so maintaining the correct | |
301 | reference count on C<key> is entirely the caller's responsibility. hv_store | |
302 | is not implemented as a call to hv_store_ent, and does not create a temporary | |
303 | SV for the key, so if your key data is not already in SV form then use | |
304 | hv_store in preference to hv_store_ent. | |
305 | ||
306 | See L<perlguts/"Understanding the Magic of Tied Hashes and Arrays"> for more | |
307 | information on how to use this function on tied hashes. | |
308 | ||
309 | =cut | |
310 | */ | |
311 | ||
fabdb6c0 | 312 | /* XXX This looks like an ideal candidate to inline */ |
34a6f7b4 NC |
313 | HE * |
314 | Perl_hv_store_ent(pTHX_ HV *hv, SV *keysv, SV *val, U32 hash) | |
315 | { | |
316 | return hv_fetch_common(hv, keysv, NULL, 0, 0, HV_FETCH_ISSTORE, val, hash); | |
317 | } | |
318 | ||
319 | /* | |
320 | =for apidoc hv_exists | |
321 | ||
322 | Returns a boolean indicating whether the specified hash key exists. The | |
323 | C<klen> is the length of the key. | |
324 | ||
325 | =cut | |
326 | */ | |
327 | ||
328 | bool | |
329 | Perl_hv_exists(pTHX_ HV *hv, const char *key, I32 klen_i32) | |
330 | { | |
331 | STRLEN klen; | |
332 | int flags; | |
333 | ||
334 | if (klen_i32 < 0) { | |
335 | klen = -klen_i32; | |
336 | flags = HVhek_UTF8; | |
337 | } else { | |
338 | klen = klen_i32; | |
339 | flags = 0; | |
340 | } | |
341 | return hv_fetch_common(hv, NULL, key, klen, flags, HV_FETCH_ISEXISTS, 0, 0) | |
342 | ? TRUE : FALSE; | |
343 | } | |
344 | ||
954c1994 GS |
345 | /* |
346 | =for apidoc hv_fetch | |
347 | ||
348 | Returns the SV which corresponds to the specified key in the hash. The | |
349 | C<klen> is the length of the key. If C<lval> is set then the fetch will be | |
350 | part of a store. Check that the return value is non-null before | |
d1be9408 | 351 | dereferencing it to an C<SV*>. |
954c1994 | 352 | |
96f1132b | 353 | See L<perlguts/"Understanding the Magic of Tied Hashes and Arrays"> for more |
954c1994 GS |
354 | information on how to use this function on tied hashes. |
355 | ||
356 | =cut | |
357 | */ | |
358 | ||
79072805 | 359 | SV** |
c1fe5510 | 360 | Perl_hv_fetch(pTHX_ HV *hv, const char *key, I32 klen_i32, I32 lval) |
79072805 | 361 | { |
c1fe5510 NC |
362 | HE *hek; |
363 | STRLEN klen; | |
364 | int flags; | |
365 | ||
366 | if (klen_i32 < 0) { | |
367 | klen = -klen_i32; | |
368 | flags = HVhek_UTF8; | |
369 | } else { | |
370 | klen = klen_i32; | |
371 | flags = 0; | |
372 | } | |
373 | hek = hv_fetch_common (hv, NULL, key, klen, flags, | |
c445ea15 | 374 | lval ? (HV_FETCH_JUST_SV | HV_FETCH_LVALUE) : HV_FETCH_JUST_SV, |
a0714e2c | 375 | NULL, 0); |
113738bb | 376 | return hek ? &HeVAL(hek) : NULL; |
79072805 LW |
377 | } |
378 | ||
34a6f7b4 NC |
379 | /* |
380 | =for apidoc hv_exists_ent | |
381 | ||
382 | Returns a boolean indicating whether the specified hash key exists. C<hash> | |
383 | can be a valid precomputed hash value, or 0 to ask for it to be | |
384 | computed. | |
385 | ||
386 | =cut | |
387 | */ | |
388 | ||
fabdb6c0 | 389 | /* XXX This looks like an ideal candidate to inline */ |
34a6f7b4 NC |
390 | bool |
391 | Perl_hv_exists_ent(pTHX_ HV *hv, SV *keysv, U32 hash) | |
392 | { | |
393 | return hv_fetch_common(hv, keysv, NULL, 0, 0, HV_FETCH_ISEXISTS, 0, hash) | |
394 | ? TRUE : FALSE; | |
395 | } | |
396 | ||
d1be9408 | 397 | /* returns an HE * structure with the all fields set */ |
fde52b5c | 398 | /* note that hent_val will be a mortal sv for MAGICAL hashes */ |
954c1994 GS |
399 | /* |
400 | =for apidoc hv_fetch_ent | |
401 | ||
402 | Returns the hash entry which corresponds to the specified key in the hash. | |
403 | C<hash> must be a valid precomputed hash number for the given C<key>, or 0 | |
404 | if you want the function to compute it. IF C<lval> is set then the fetch | |
405 | will be part of a store. Make sure the return value is non-null before | |
406 | accessing it. The return value when C<tb> is a tied hash is a pointer to a | |
407 | static location, so be sure to make a copy of the structure if you need to | |
1c846c1f | 408 | store it somewhere. |
954c1994 | 409 | |
96f1132b | 410 | See L<perlguts/"Understanding the Magic of Tied Hashes and Arrays"> for more |
954c1994 GS |
411 | information on how to use this function on tied hashes. |
412 | ||
413 | =cut | |
414 | */ | |
415 | ||
fde52b5c | 416 | HE * |
864dbfa3 | 417 | Perl_hv_fetch_ent(pTHX_ HV *hv, SV *keysv, I32 lval, register U32 hash) |
fde52b5c | 418 | { |
7f66fda2 | 419 | return hv_fetch_common(hv, keysv, NULL, 0, 0, |
a0714e2c | 420 | (lval ? HV_FETCH_LVALUE : 0), NULL, hash); |
113738bb NC |
421 | } |
422 | ||
8f8d40ab | 423 | STATIC HE * |
c1fe5510 | 424 | S_hv_fetch_common(pTHX_ HV *hv, SV *keysv, const char *key, STRLEN klen, |
b2c64049 | 425 | int flags, int action, SV *val, register U32 hash) |
113738bb | 426 | { |
27da23d5 | 427 | dVAR; |
b2c64049 | 428 | XPVHV* xhv; |
b2c64049 NC |
429 | HE *entry; |
430 | HE **oentry; | |
fde52b5c | 431 | SV *sv; |
da58a35d | 432 | bool is_utf8; |
113738bb | 433 | int masked_flags; |
fde52b5c | 434 | |
435 | if (!hv) | |
a4fc7abc | 436 | return NULL; |
fde52b5c | 437 | |
113738bb | 438 | if (keysv) { |
e593d2fe AE |
439 | if (flags & HVhek_FREEKEY) |
440 | Safefree(key); | |
5c144d81 | 441 | key = SvPV_const(keysv, klen); |
c1fe5510 | 442 | flags = 0; |
113738bb NC |
443 | is_utf8 = (SvUTF8(keysv) != 0); |
444 | } else { | |
c1fe5510 | 445 | is_utf8 = ((flags & HVhek_UTF8) ? TRUE : FALSE); |
113738bb | 446 | } |
113738bb | 447 | |
b2c64049 | 448 | xhv = (XPVHV*)SvANY(hv); |
7f66fda2 | 449 | if (SvMAGICAL(hv)) { |
6136c704 | 450 | if (SvRMAGICAL(hv) && !(action & (HV_FETCH_ISSTORE|HV_FETCH_ISEXISTS))) { |
7f66fda2 NC |
451 | if (mg_find((SV*)hv, PERL_MAGIC_tied) || SvGMAGICAL((SV*)hv)) { |
452 | sv = sv_newmortal(); | |
113738bb | 453 | |
7f66fda2 NC |
454 | /* XXX should be able to skimp on the HE/HEK here when |
455 | HV_FETCH_JUST_SV is true. */ | |
113738bb | 456 | |
7f66fda2 NC |
457 | if (!keysv) { |
458 | keysv = newSVpvn(key, klen); | |
459 | if (is_utf8) { | |
460 | SvUTF8_on(keysv); | |
461 | } | |
462 | } else { | |
463 | keysv = newSVsv(keysv); | |
113738bb | 464 | } |
7f66fda2 NC |
465 | mg_copy((SV*)hv, sv, (char *)keysv, HEf_SVKEY); |
466 | ||
467 | /* grab a fake HE/HEK pair from the pool or make a new one */ | |
468 | entry = PL_hv_fetch_ent_mh; | |
469 | if (entry) | |
470 | PL_hv_fetch_ent_mh = HeNEXT(entry); | |
471 | else { | |
472 | char *k; | |
473 | entry = new_HE(); | |
a02a5408 | 474 | Newx(k, HEK_BASESIZE + sizeof(SV*), char); |
7f66fda2 NC |
475 | HeKEY_hek(entry) = (HEK*)k; |
476 | } | |
4608196e | 477 | HeNEXT(entry) = NULL; |
7f66fda2 NC |
478 | HeSVKEY_set(entry, keysv); |
479 | HeVAL(entry) = sv; | |
480 | sv_upgrade(sv, SVt_PVLV); | |
481 | LvTYPE(sv) = 'T'; | |
482 | /* so we can free entry when freeing sv */ | |
483 | LvTARG(sv) = (SV*)entry; | |
484 | ||
485 | /* XXX remove at some point? */ | |
486 | if (flags & HVhek_FREEKEY) | |
487 | Safefree(key); | |
488 | ||
489 | return entry; | |
113738bb | 490 | } |
7f66fda2 NC |
491 | #ifdef ENV_IS_CASELESS |
492 | else if (mg_find((SV*)hv, PERL_MAGIC_env)) { | |
493 | U32 i; | |
494 | for (i = 0; i < klen; ++i) | |
495 | if (isLOWER(key[i])) { | |
086cb327 NC |
496 | /* Would be nice if we had a routine to do the |
497 | copy and upercase in a single pass through. */ | |
0bd48802 | 498 | const char * const nkey = strupr(savepvn(key,klen)); |
086cb327 NC |
499 | /* Note that this fetch is for nkey (the uppercased |
500 | key) whereas the store is for key (the original) */ | |
a0714e2c | 501 | entry = hv_fetch_common(hv, NULL, nkey, klen, |
086cb327 NC |
502 | HVhek_FREEKEY, /* free nkey */ |
503 | 0 /* non-LVAL fetch */, | |
a0714e2c | 504 | NULL /* no value */, |
086cb327 NC |
505 | 0 /* compute hash */); |
506 | if (!entry && (action & HV_FETCH_LVALUE)) { | |
507 | /* This call will free key if necessary. | |
508 | Do it this way to encourage compiler to tail | |
509 | call optimise. */ | |
510 | entry = hv_fetch_common(hv, keysv, key, klen, | |
511 | flags, HV_FETCH_ISSTORE, | |
561b68a9 | 512 | newSV(0), hash); |
086cb327 NC |
513 | } else { |
514 | if (flags & HVhek_FREEKEY) | |
515 | Safefree(key); | |
516 | } | |
517 | return entry; | |
7f66fda2 | 518 | } |
902173a3 | 519 | } |
7f66fda2 NC |
520 | #endif |
521 | } /* ISFETCH */ | |
522 | else if (SvRMAGICAL(hv) && (action & HV_FETCH_ISEXISTS)) { | |
523 | if (mg_find((SV*)hv, PERL_MAGIC_tied) || SvGMAGICAL((SV*)hv)) { | |
b2c64049 NC |
524 | /* I don't understand why hv_exists_ent has svret and sv, |
525 | whereas hv_exists only had one. */ | |
9d4ba2ae | 526 | SV * const svret = sv_newmortal(); |
b2c64049 | 527 | sv = sv_newmortal(); |
7f66fda2 NC |
528 | |
529 | if (keysv || is_utf8) { | |
530 | if (!keysv) { | |
531 | keysv = newSVpvn(key, klen); | |
532 | SvUTF8_on(keysv); | |
533 | } else { | |
534 | keysv = newSVsv(keysv); | |
535 | } | |
b2c64049 NC |
536 | mg_copy((SV*)hv, sv, (char *)sv_2mortal(keysv), HEf_SVKEY); |
537 | } else { | |
538 | mg_copy((SV*)hv, sv, key, klen); | |
7f66fda2 | 539 | } |
b2c64049 NC |
540 | if (flags & HVhek_FREEKEY) |
541 | Safefree(key); | |
7f66fda2 NC |
542 | magic_existspack(svret, mg_find(sv, PERL_MAGIC_tiedelem)); |
543 | /* This cast somewhat evil, but I'm merely using NULL/ | |
544 | not NULL to return the boolean exists. | |
545 | And I know hv is not NULL. */ | |
546 | return SvTRUE(svret) ? (HE *)hv : NULL; | |
e7152ba2 | 547 | } |
7f66fda2 NC |
548 | #ifdef ENV_IS_CASELESS |
549 | else if (mg_find((SV*)hv, PERL_MAGIC_env)) { | |
550 | /* XXX This code isn't UTF8 clean. */ | |
a15d23f8 | 551 | char * const keysave = (char * const)key; |
b2c64049 NC |
552 | /* Will need to free this, so set FREEKEY flag. */ |
553 | key = savepvn(key,klen); | |
554 | key = (const char*)strupr((char*)key); | |
6136c704 | 555 | is_utf8 = FALSE; |
7f66fda2 | 556 | hash = 0; |
8b4f7dd5 | 557 | keysv = 0; |
b2c64049 NC |
558 | |
559 | if (flags & HVhek_FREEKEY) { | |
560 | Safefree(keysave); | |
561 | } | |
562 | flags |= HVhek_FREEKEY; | |
7f66fda2 | 563 | } |
902173a3 | 564 | #endif |
7f66fda2 | 565 | } /* ISEXISTS */ |
b2c64049 NC |
566 | else if (action & HV_FETCH_ISSTORE) { |
567 | bool needs_copy; | |
568 | bool needs_store; | |
569 | hv_magic_check (hv, &needs_copy, &needs_store); | |
570 | if (needs_copy) { | |
a3b680e6 | 571 | const bool save_taint = PL_tainted; |
b2c64049 NC |
572 | if (keysv || is_utf8) { |
573 | if (!keysv) { | |
574 | keysv = newSVpvn(key, klen); | |
575 | SvUTF8_on(keysv); | |
576 | } | |
577 | if (PL_tainting) | |
578 | PL_tainted = SvTAINTED(keysv); | |
579 | keysv = sv_2mortal(newSVsv(keysv)); | |
580 | mg_copy((SV*)hv, val, (char*)keysv, HEf_SVKEY); | |
581 | } else { | |
582 | mg_copy((SV*)hv, val, key, klen); | |
583 | } | |
584 | ||
585 | TAINT_IF(save_taint); | |
1baaf5d7 | 586 | if (!needs_store) { |
b2c64049 NC |
587 | if (flags & HVhek_FREEKEY) |
588 | Safefree(key); | |
4608196e | 589 | return NULL; |
b2c64049 NC |
590 | } |
591 | #ifdef ENV_IS_CASELESS | |
592 | else if (mg_find((SV*)hv, PERL_MAGIC_env)) { | |
593 | /* XXX This code isn't UTF8 clean. */ | |
594 | const char *keysave = key; | |
595 | /* Will need to free this, so set FREEKEY flag. */ | |
596 | key = savepvn(key,klen); | |
597 | key = (const char*)strupr((char*)key); | |
6136c704 | 598 | is_utf8 = FALSE; |
b2c64049 | 599 | hash = 0; |
8b4f7dd5 | 600 | keysv = 0; |
b2c64049 NC |
601 | |
602 | if (flags & HVhek_FREEKEY) { | |
603 | Safefree(keysave); | |
604 | } | |
605 | flags |= HVhek_FREEKEY; | |
606 | } | |
607 | #endif | |
608 | } | |
609 | } /* ISSTORE */ | |
7f66fda2 | 610 | } /* SvMAGICAL */ |
fde52b5c | 611 | |
7b2c381c | 612 | if (!HvARRAY(hv)) { |
b2c64049 | 613 | if ((action & (HV_FETCH_LVALUE | HV_FETCH_ISSTORE)) |
fde52b5c | 614 | #ifdef DYNAMIC_ENV_FETCH /* if it's an %ENV lookup, we may get it on the fly */ |
8aacddc1 | 615 | || (SvRMAGICAL((SV*)hv) && mg_find((SV*)hv, PERL_MAGIC_env)) |
fde52b5c | 616 | #endif |
d58e6666 NC |
617 | ) { |
618 | char *array; | |
a02a5408 | 619 | Newxz(array, |
cbec9347 | 620 | PERL_HV_ARRAY_ALLOC_BYTES(xhv->xhv_max+1 /* HvMAX(hv)+1 */), |
d58e6666 NC |
621 | char); |
622 | HvARRAY(hv) = (HE**)array; | |
623 | } | |
7f66fda2 NC |
624 | #ifdef DYNAMIC_ENV_FETCH |
625 | else if (action & HV_FETCH_ISEXISTS) { | |
626 | /* for an %ENV exists, if we do an insert it's by a recursive | |
627 | store call, so avoid creating HvARRAY(hv) right now. */ | |
628 | } | |
629 | #endif | |
113738bb NC |
630 | else { |
631 | /* XXX remove at some point? */ | |
632 | if (flags & HVhek_FREEKEY) | |
633 | Safefree(key); | |
634 | ||
fde52b5c | 635 | return 0; |
113738bb | 636 | } |
fde52b5c | 637 | } |
638 | ||
19692e8d | 639 | if (is_utf8) { |
41d88b63 | 640 | char * const keysave = (char *)key; |
f9a63242 | 641 | key = (char*)bytes_from_utf8((U8*)key, &klen, &is_utf8); |
19692e8d | 642 | if (is_utf8) |
c1fe5510 NC |
643 | flags |= HVhek_UTF8; |
644 | else | |
645 | flags &= ~HVhek_UTF8; | |
7f66fda2 NC |
646 | if (key != keysave) { |
647 | if (flags & HVhek_FREEKEY) | |
648 | Safefree(keysave); | |
19692e8d | 649 | flags |= HVhek_WASUTF8 | HVhek_FREEKEY; |
7f66fda2 | 650 | } |
19692e8d | 651 | } |
f9a63242 | 652 | |
4b5190b5 NC |
653 | if (HvREHASH(hv)) { |
654 | PERL_HASH_INTERNAL(hash, key, klen); | |
b2c64049 NC |
655 | /* We don't have a pointer to the hv, so we have to replicate the |
656 | flag into every HEK, so that hv_iterkeysv can see it. */ | |
657 | /* And yes, you do need this even though you are not "storing" because | |
fdcd69b6 NC |
658 | you can flip the flags below if doing an lval lookup. (And that |
659 | was put in to give the semantics Andreas was expecting.) */ | |
660 | flags |= HVhek_REHASH; | |
4b5190b5 | 661 | } else if (!hash) { |
113738bb | 662 | if (keysv && (SvIsCOW_shared_hash(keysv))) { |
c158a4fd | 663 | hash = SvSHARED_HASH(keysv); |
46187eeb NC |
664 | } else { |
665 | PERL_HASH(hash, key, klen); | |
666 | } | |
667 | } | |
effa1e2d | 668 | |
113738bb NC |
669 | masked_flags = (flags & HVhek_MASK); |
670 | ||
7f66fda2 | 671 | #ifdef DYNAMIC_ENV_FETCH |
4608196e | 672 | if (!HvARRAY(hv)) entry = NULL; |
7f66fda2 NC |
673 | else |
674 | #endif | |
b2c64049 | 675 | { |
7b2c381c | 676 | entry = (HvARRAY(hv))[hash & (I32) HvMAX(hv)]; |
b2c64049 | 677 | } |
0298d7b9 | 678 | for (; entry; entry = HeNEXT(entry)) { |
fde52b5c | 679 | if (HeHASH(entry) != hash) /* strings can't be equal */ |
680 | continue; | |
eb160463 | 681 | if (HeKLEN(entry) != (I32)klen) |
fde52b5c | 682 | continue; |
1c846c1f | 683 | if (HeKEY(entry) != key && memNE(HeKEY(entry),key,klen)) /* is this it? */ |
fde52b5c | 684 | continue; |
113738bb | 685 | if ((HeKFLAGS(entry) ^ masked_flags) & HVhek_UTF8) |
c3654f1a | 686 | continue; |
b2c64049 NC |
687 | |
688 | if (action & (HV_FETCH_LVALUE|HV_FETCH_ISSTORE)) { | |
689 | if (HeKFLAGS(entry) != masked_flags) { | |
690 | /* We match if HVhek_UTF8 bit in our flags and hash key's | |
691 | match. But if entry was set previously with HVhek_WASUTF8 | |
692 | and key now doesn't (or vice versa) then we should change | |
693 | the key's flag, as this is assignment. */ | |
694 | if (HvSHAREKEYS(hv)) { | |
695 | /* Need to swap the key we have for a key with the flags we | |
696 | need. As keys are shared we can't just write to the | |
697 | flag, so we share the new one, unshare the old one. */ | |
6136c704 | 698 | HEK * const new_hek = share_hek_flags(key, klen, hash, |
6e838c70 | 699 | masked_flags); |
b2c64049 NC |
700 | unshare_hek (HeKEY_hek(entry)); |
701 | HeKEY_hek(entry) = new_hek; | |
702 | } | |
5d2b1485 NC |
703 | else if (hv == PL_strtab) { |
704 | /* PL_strtab is usually the only hash without HvSHAREKEYS, | |
705 | so putting this test here is cheap */ | |
706 | if (flags & HVhek_FREEKEY) | |
707 | Safefree(key); | |
708 | Perl_croak(aTHX_ S_strtab_error, | |
709 | action & HV_FETCH_LVALUE ? "fetch" : "store"); | |
710 | } | |
b2c64049 NC |
711 | else |
712 | HeKFLAGS(entry) = masked_flags; | |
713 | if (masked_flags & HVhek_ENABLEHVKFLAGS) | |
714 | HvHASKFLAGS_on(hv); | |
715 | } | |
716 | if (HeVAL(entry) == &PL_sv_placeholder) { | |
717 | /* yes, can store into placeholder slot */ | |
718 | if (action & HV_FETCH_LVALUE) { | |
719 | if (SvMAGICAL(hv)) { | |
720 | /* This preserves behaviour with the old hv_fetch | |
721 | implementation which at this point would bail out | |
722 | with a break; (at "if we find a placeholder, we | |
723 | pretend we haven't found anything") | |
724 | ||
725 | That break mean that if a placeholder were found, it | |
726 | caused a call into hv_store, which in turn would | |
727 | check magic, and if there is no magic end up pretty | |
728 | much back at this point (in hv_store's code). */ | |
729 | break; | |
730 | } | |
731 | /* LVAL fetch which actaully needs a store. */ | |
561b68a9 | 732 | val = newSV(0); |
ca732855 | 733 | HvPLACEHOLDERS(hv)--; |
b2c64049 NC |
734 | } else { |
735 | /* store */ | |
736 | if (val != &PL_sv_placeholder) | |
ca732855 | 737 | HvPLACEHOLDERS(hv)--; |
b2c64049 NC |
738 | } |
739 | HeVAL(entry) = val; | |
740 | } else if (action & HV_FETCH_ISSTORE) { | |
741 | SvREFCNT_dec(HeVAL(entry)); | |
742 | HeVAL(entry) = val; | |
743 | } | |
27bcc0a7 | 744 | } else if (HeVAL(entry) == &PL_sv_placeholder) { |
b2c64049 NC |
745 | /* if we find a placeholder, we pretend we haven't found |
746 | anything */ | |
8aacddc1 | 747 | break; |
b2c64049 | 748 | } |
113738bb NC |
749 | if (flags & HVhek_FREEKEY) |
750 | Safefree(key); | |
fde52b5c | 751 | return entry; |
752 | } | |
753 | #ifdef DYNAMIC_ENV_FETCH /* %ENV lookup? If so, try to fetch the value now */ | |
0ed29950 NC |
754 | if (!(action & HV_FETCH_ISSTORE) |
755 | && SvRMAGICAL((SV*)hv) && mg_find((SV*)hv, PERL_MAGIC_env)) { | |
a6c40364 | 756 | unsigned long len; |
9d4ba2ae | 757 | const char * const env = PerlEnv_ENVgetenv_len(key,&len); |
a6c40364 GS |
758 | if (env) { |
759 | sv = newSVpvn(env,len); | |
760 | SvTAINTED_on(sv); | |
7fd3d16e | 761 | return hv_fetch_common(hv,keysv,key,klen,flags,HV_FETCH_ISSTORE,sv, |
b2c64049 | 762 | hash); |
a6c40364 | 763 | } |
fde52b5c | 764 | } |
765 | #endif | |
7f66fda2 NC |
766 | |
767 | if (!entry && SvREADONLY(hv) && !(action & HV_FETCH_ISEXISTS)) { | |
c445ea15 | 768 | hv_notallowed(flags, key, klen, |
c8cd6465 NC |
769 | "Attempt to access disallowed key '%"SVf"' in" |
770 | " a restricted hash"); | |
1b1f1335 | 771 | } |
b2c64049 NC |
772 | if (!(action & (HV_FETCH_LVALUE|HV_FETCH_ISSTORE))) { |
773 | /* Not doing some form of store, so return failure. */ | |
774 | if (flags & HVhek_FREEKEY) | |
775 | Safefree(key); | |
776 | return 0; | |
777 | } | |
113738bb | 778 | if (action & HV_FETCH_LVALUE) { |
561b68a9 | 779 | val = newSV(0); |
b2c64049 NC |
780 | if (SvMAGICAL(hv)) { |
781 | /* At this point the old hv_fetch code would call to hv_store, | |
782 | which in turn might do some tied magic. So we need to make that | |
783 | magic check happen. */ | |
784 | /* gonna assign to this, so it better be there */ | |
785 | return hv_fetch_common(hv, keysv, key, klen, flags, | |
786 | HV_FETCH_ISSTORE, val, hash); | |
787 | /* XXX Surely that could leak if the fetch-was-store fails? | |
788 | Just like the hv_fetch. */ | |
113738bb NC |
789 | } |
790 | } | |
791 | ||
b2c64049 NC |
792 | /* Welcome to hv_store... */ |
793 | ||
7b2c381c | 794 | if (!HvARRAY(hv)) { |
b2c64049 NC |
795 | /* Not sure if we can get here. I think the only case of oentry being |
796 | NULL is for %ENV with dynamic env fetch. But that should disappear | |
797 | with magic in the previous code. */ | |
d58e6666 | 798 | char *array; |
a02a5408 | 799 | Newxz(array, |
b2c64049 | 800 | PERL_HV_ARRAY_ALLOC_BYTES(xhv->xhv_max+1 /* HvMAX(hv)+1 */), |
d58e6666 NC |
801 | char); |
802 | HvARRAY(hv) = (HE**)array; | |
b2c64049 NC |
803 | } |
804 | ||
7b2c381c | 805 | oentry = &(HvARRAY(hv))[hash & (I32) xhv->xhv_max]; |
ab4af705 | 806 | |
b2c64049 NC |
807 | entry = new_HE(); |
808 | /* share_hek_flags will do the free for us. This might be considered | |
809 | bad API design. */ | |
810 | if (HvSHAREKEYS(hv)) | |
6e838c70 | 811 | HeKEY_hek(entry) = share_hek_flags(key, klen, hash, flags); |
5d2b1485 NC |
812 | else if (hv == PL_strtab) { |
813 | /* PL_strtab is usually the only hash without HvSHAREKEYS, so putting | |
814 | this test here is cheap */ | |
815 | if (flags & HVhek_FREEKEY) | |
816 | Safefree(key); | |
817 | Perl_croak(aTHX_ S_strtab_error, | |
818 | action & HV_FETCH_LVALUE ? "fetch" : "store"); | |
819 | } | |
b2c64049 NC |
820 | else /* gotta do the real thing */ |
821 | HeKEY_hek(entry) = save_hek_flags(key, klen, hash, flags); | |
822 | HeVAL(entry) = val; | |
823 | HeNEXT(entry) = *oentry; | |
824 | *oentry = entry; | |
825 | ||
826 | if (val == &PL_sv_placeholder) | |
ca732855 | 827 | HvPLACEHOLDERS(hv)++; |
b2c64049 NC |
828 | if (masked_flags & HVhek_ENABLEHVKFLAGS) |
829 | HvHASKFLAGS_on(hv); | |
830 | ||
0298d7b9 NC |
831 | { |
832 | const HE *counter = HeNEXT(entry); | |
833 | ||
4c7185a0 | 834 | xhv->xhv_keys++; /* HvTOTALKEYS(hv)++ */ |
0298d7b9 NC |
835 | if (!counter) { /* initial entry? */ |
836 | xhv->xhv_fill++; /* HvFILL(hv)++ */ | |
837 | } else if (xhv->xhv_keys > (IV)xhv->xhv_max) { | |
838 | hsplit(hv); | |
839 | } else if(!HvREHASH(hv)) { | |
840 | U32 n_links = 1; | |
841 | ||
842 | while ((counter = HeNEXT(counter))) | |
843 | n_links++; | |
844 | ||
845 | if (n_links > HV_MAX_LENGTH_BEFORE_SPLIT) { | |
846 | /* Use only the old HvKEYS(hv) > HvMAX(hv) condition to limit | |
847 | bucket splits on a rehashed hash, as we're not going to | |
848 | split it again, and if someone is lucky (evil) enough to | |
849 | get all the keys in one list they could exhaust our memory | |
850 | as we repeatedly double the number of buckets on every | |
851 | entry. Linear search feels a less worse thing to do. */ | |
852 | hsplit(hv); | |
853 | } | |
854 | } | |
fde52b5c | 855 | } |
b2c64049 NC |
856 | |
857 | return entry; | |
fde52b5c | 858 | } |
859 | ||
864dbfa3 | 860 | STATIC void |
b0e6ae5b | 861 | S_hv_magic_check(HV *hv, bool *needs_copy, bool *needs_store) |
d0066dc7 | 862 | { |
a3b680e6 | 863 | const MAGIC *mg = SvMAGIC(hv); |
d0066dc7 OT |
864 | *needs_copy = FALSE; |
865 | *needs_store = TRUE; | |
866 | while (mg) { | |
867 | if (isUPPER(mg->mg_type)) { | |
868 | *needs_copy = TRUE; | |
d60c5a05 | 869 | if (mg->mg_type == PERL_MAGIC_tied) { |
d0066dc7 | 870 | *needs_store = FALSE; |
4ab2a30b | 871 | return; /* We've set all there is to set. */ |
d0066dc7 OT |
872 | } |
873 | } | |
874 | mg = mg->mg_moremagic; | |
875 | } | |
876 | } | |
877 | ||
954c1994 | 878 | /* |
a3bcc51e TP |
879 | =for apidoc hv_scalar |
880 | ||
881 | Evaluates the hash in scalar context and returns the result. Handles magic when the hash is tied. | |
882 | ||
883 | =cut | |
884 | */ | |
885 | ||
886 | SV * | |
887 | Perl_hv_scalar(pTHX_ HV *hv) | |
888 | { | |
a3bcc51e | 889 | SV *sv; |
823a54a3 AL |
890 | |
891 | if (SvRMAGICAL(hv)) { | |
892 | MAGIC * const mg = mg_find((SV*)hv, PERL_MAGIC_tied); | |
893 | if (mg) | |
894 | return magic_scalarpack(hv, mg); | |
895 | } | |
a3bcc51e TP |
896 | |
897 | sv = sv_newmortal(); | |
898 | if (HvFILL((HV*)hv)) | |
899 | Perl_sv_setpvf(aTHX_ sv, "%ld/%ld", | |
900 | (long)HvFILL(hv), (long)HvMAX(hv) + 1); | |
901 | else | |
902 | sv_setiv(sv, 0); | |
903 | ||
904 | return sv; | |
905 | } | |
906 | ||
907 | /* | |
954c1994 GS |
908 | =for apidoc hv_delete |
909 | ||
910 | Deletes a key/value pair in the hash. The value SV is removed from the | |
1c846c1f | 911 | hash and returned to the caller. The C<klen> is the length of the key. |
954c1994 GS |
912 | The C<flags> value will normally be zero; if set to G_DISCARD then NULL |
913 | will be returned. | |
914 | ||
915 | =cut | |
916 | */ | |
917 | ||
79072805 | 918 | SV * |
cd6d36ac | 919 | Perl_hv_delete(pTHX_ HV *hv, const char *key, I32 klen_i32, I32 flags) |
79072805 | 920 | { |
cd6d36ac | 921 | STRLEN klen; |
6136c704 | 922 | int k_flags; |
cd6d36ac NC |
923 | |
924 | if (klen_i32 < 0) { | |
925 | klen = -klen_i32; | |
6136c704 | 926 | k_flags = HVhek_UTF8; |
cd6d36ac NC |
927 | } else { |
928 | klen = klen_i32; | |
6136c704 | 929 | k_flags = 0; |
cd6d36ac NC |
930 | } |
931 | return hv_delete_common(hv, NULL, key, klen, k_flags, flags, 0); | |
fde52b5c | 932 | } |
933 | ||
954c1994 GS |
934 | /* |
935 | =for apidoc hv_delete_ent | |
936 | ||
937 | Deletes a key/value pair in the hash. The value SV is removed from the | |
938 | hash and returned to the caller. The C<flags> value will normally be zero; | |
939 | if set to G_DISCARD then NULL will be returned. C<hash> can be a valid | |
940 | precomputed hash value, or 0 to ask for it to be computed. | |
941 | ||
942 | =cut | |
943 | */ | |
944 | ||
fabdb6c0 | 945 | /* XXX This looks like an ideal candidate to inline */ |
fde52b5c | 946 | SV * |
864dbfa3 | 947 | Perl_hv_delete_ent(pTHX_ HV *hv, SV *keysv, I32 flags, U32 hash) |
fde52b5c | 948 | { |
cd6d36ac | 949 | return hv_delete_common(hv, keysv, NULL, 0, 0, flags, hash); |
f1317c8d NC |
950 | } |
951 | ||
8f8d40ab | 952 | STATIC SV * |
cd6d36ac NC |
953 | S_hv_delete_common(pTHX_ HV *hv, SV *keysv, const char *key, STRLEN klen, |
954 | int k_flags, I32 d_flags, U32 hash) | |
f1317c8d | 955 | { |
27da23d5 | 956 | dVAR; |
cbec9347 | 957 | register XPVHV* xhv; |
fde52b5c | 958 | register HE *entry; |
959 | register HE **oentry; | |
9e720f71 | 960 | HE *const *first_entry; |
da58a35d | 961 | bool is_utf8; |
7a9669ca | 962 | int masked_flags; |
1c846c1f | 963 | |
fde52b5c | 964 | if (!hv) |
a0714e2c | 965 | return NULL; |
f1317c8d NC |
966 | |
967 | if (keysv) { | |
e593d2fe AE |
968 | if (k_flags & HVhek_FREEKEY) |
969 | Safefree(key); | |
5c144d81 | 970 | key = SvPV_const(keysv, klen); |
cd6d36ac | 971 | k_flags = 0; |
f1317c8d NC |
972 | is_utf8 = (SvUTF8(keysv) != 0); |
973 | } else { | |
cd6d36ac | 974 | is_utf8 = ((k_flags & HVhek_UTF8) ? TRUE : FALSE); |
f1317c8d | 975 | } |
f1317c8d | 976 | |
fde52b5c | 977 | if (SvRMAGICAL(hv)) { |
0a0bb7c7 OT |
978 | bool needs_copy; |
979 | bool needs_store; | |
980 | hv_magic_check (hv, &needs_copy, &needs_store); | |
981 | ||
f1317c8d | 982 | if (needs_copy) { |
6136c704 | 983 | SV *sv; |
7a9669ca NC |
984 | entry = hv_fetch_common(hv, keysv, key, klen, |
985 | k_flags & ~HVhek_FREEKEY, HV_FETCH_LVALUE, | |
a0714e2c | 986 | NULL, hash); |
7a9669ca | 987 | sv = entry ? HeVAL(entry) : NULL; |
f1317c8d NC |
988 | if (sv) { |
989 | if (SvMAGICAL(sv)) { | |
990 | mg_clear(sv); | |
991 | } | |
992 | if (!needs_store) { | |
993 | if (mg_find(sv, PERL_MAGIC_tiedelem)) { | |
994 | /* No longer an element */ | |
995 | sv_unmagic(sv, PERL_MAGIC_tiedelem); | |
996 | return sv; | |
997 | } | |
a0714e2c | 998 | return NULL; /* element cannot be deleted */ |
f1317c8d | 999 | } |
902173a3 | 1000 | #ifdef ENV_IS_CASELESS |
8167a60a NC |
1001 | else if (mg_find((SV*)hv, PERL_MAGIC_env)) { |
1002 | /* XXX This code isn't UTF8 clean. */ | |
1003 | keysv = sv_2mortal(newSVpvn(key,klen)); | |
1004 | if (k_flags & HVhek_FREEKEY) { | |
1005 | Safefree(key); | |
1006 | } | |
1007 | key = strupr(SvPVX(keysv)); | |
1008 | is_utf8 = 0; | |
1009 | k_flags = 0; | |
1010 | hash = 0; | |
7f66fda2 | 1011 | } |
510ac311 | 1012 | #endif |
2fd1c6b8 | 1013 | } |
2fd1c6b8 | 1014 | } |
fde52b5c | 1015 | } |
cbec9347 | 1016 | xhv = (XPVHV*)SvANY(hv); |
7b2c381c | 1017 | if (!HvARRAY(hv)) |
a0714e2c | 1018 | return NULL; |
fde52b5c | 1019 | |
19692e8d | 1020 | if (is_utf8) { |
c445ea15 | 1021 | const char * const keysave = key; |
b464bac0 | 1022 | key = (char*)bytes_from_utf8((U8*)key, &klen, &is_utf8); |
cd6d36ac | 1023 | |
19692e8d | 1024 | if (is_utf8) |
cd6d36ac NC |
1025 | k_flags |= HVhek_UTF8; |
1026 | else | |
1027 | k_flags &= ~HVhek_UTF8; | |
7f66fda2 NC |
1028 | if (key != keysave) { |
1029 | if (k_flags & HVhek_FREEKEY) { | |
1030 | /* This shouldn't happen if our caller does what we expect, | |
1031 | but strictly the API allows it. */ | |
1032 | Safefree(keysave); | |
1033 | } | |
1034 | k_flags |= HVhek_WASUTF8 | HVhek_FREEKEY; | |
1035 | } | |
cd6d36ac | 1036 | HvHASKFLAGS_on((SV*)hv); |
19692e8d | 1037 | } |
f9a63242 | 1038 | |
4b5190b5 NC |
1039 | if (HvREHASH(hv)) { |
1040 | PERL_HASH_INTERNAL(hash, key, klen); | |
1041 | } else if (!hash) { | |
7a9669ca | 1042 | if (keysv && (SvIsCOW_shared_hash(keysv))) { |
c158a4fd | 1043 | hash = SvSHARED_HASH(keysv); |
7a9669ca NC |
1044 | } else { |
1045 | PERL_HASH(hash, key, klen); | |
1046 | } | |
4b5190b5 | 1047 | } |
fde52b5c | 1048 | |
7a9669ca NC |
1049 | masked_flags = (k_flags & HVhek_MASK); |
1050 | ||
9e720f71 | 1051 | first_entry = oentry = &(HvARRAY(hv))[hash & (I32) HvMAX(hv)]; |
fde52b5c | 1052 | entry = *oentry; |
9e720f71 | 1053 | for (; entry; oentry = &HeNEXT(entry), entry = *oentry) { |
6136c704 | 1054 | SV *sv; |
fde52b5c | 1055 | if (HeHASH(entry) != hash) /* strings can't be equal */ |
1056 | continue; | |
eb160463 | 1057 | if (HeKLEN(entry) != (I32)klen) |
fde52b5c | 1058 | continue; |
1c846c1f | 1059 | if (HeKEY(entry) != key && memNE(HeKEY(entry),key,klen)) /* is this it? */ |
fde52b5c | 1060 | continue; |
7a9669ca | 1061 | if ((HeKFLAGS(entry) ^ masked_flags) & HVhek_UTF8) |
c3654f1a | 1062 | continue; |
8aacddc1 | 1063 | |
5d2b1485 NC |
1064 | if (hv == PL_strtab) { |
1065 | if (k_flags & HVhek_FREEKEY) | |
1066 | Safefree(key); | |
1067 | Perl_croak(aTHX_ S_strtab_error, "delete"); | |
1068 | } | |
1069 | ||
8aacddc1 | 1070 | /* if placeholder is here, it's already been deleted.... */ |
6136c704 AL |
1071 | if (HeVAL(entry) == &PL_sv_placeholder) { |
1072 | if (k_flags & HVhek_FREEKEY) | |
1073 | Safefree(key); | |
1074 | return NULL; | |
8aacddc1 | 1075 | } |
6136c704 | 1076 | if (SvREADONLY(hv) && HeVAL(entry) && SvREADONLY(HeVAL(entry))) { |
d4c19fe8 | 1077 | hv_notallowed(k_flags, key, klen, |
c8cd6465 NC |
1078 | "Attempt to delete readonly key '%"SVf"' from" |
1079 | " a restricted hash"); | |
8aacddc1 | 1080 | } |
b84d0860 NC |
1081 | if (k_flags & HVhek_FREEKEY) |
1082 | Safefree(key); | |
8aacddc1 | 1083 | |
cd6d36ac | 1084 | if (d_flags & G_DISCARD) |
a0714e2c | 1085 | sv = NULL; |
94f7643d | 1086 | else { |
79d01fbf | 1087 | sv = sv_2mortal(HeVAL(entry)); |
7996736c | 1088 | HeVAL(entry) = &PL_sv_placeholder; |
94f7643d | 1089 | } |
8aacddc1 NIS |
1090 | |
1091 | /* | |
1092 | * If a restricted hash, rather than really deleting the entry, put | |
1093 | * a placeholder there. This marks the key as being "approved", so | |
1094 | * we can still access via not-really-existing key without raising | |
1095 | * an error. | |
1096 | */ | |
1097 | if (SvREADONLY(hv)) { | |
754604c4 | 1098 | SvREFCNT_dec(HeVAL(entry)); |
7996736c | 1099 | HeVAL(entry) = &PL_sv_placeholder; |
8aacddc1 NIS |
1100 | /* We'll be saving this slot, so the number of allocated keys |
1101 | * doesn't go down, but the number placeholders goes up */ | |
ca732855 | 1102 | HvPLACEHOLDERS(hv)++; |
8aacddc1 | 1103 | } else { |
a26e96df | 1104 | *oentry = HeNEXT(entry); |
9e720f71 | 1105 | if(!*first_entry) { |
a26e96df | 1106 | xhv->xhv_fill--; /* HvFILL(hv)-- */ |
9e720f71 | 1107 | } |
b79f7545 | 1108 | if (SvOOK(hv) && entry == HvAUX(hv)->xhv_eiter /* HvEITER(hv) */) |
8aacddc1 NIS |
1109 | HvLAZYDEL_on(hv); |
1110 | else | |
1111 | hv_free_ent(hv, entry); | |
4c7185a0 | 1112 | xhv->xhv_keys--; /* HvTOTALKEYS(hv)-- */ |
574c8022 | 1113 | if (xhv->xhv_keys == 0) |
19692e8d | 1114 | HvHASKFLAGS_off(hv); |
8aacddc1 | 1115 | } |
79072805 LW |
1116 | return sv; |
1117 | } | |
8aacddc1 | 1118 | if (SvREADONLY(hv)) { |
d4c19fe8 | 1119 | hv_notallowed(k_flags, key, klen, |
c8cd6465 NC |
1120 | "Attempt to delete disallowed key '%"SVf"' from" |
1121 | " a restricted hash"); | |
8aacddc1 NIS |
1122 | } |
1123 | ||
19692e8d | 1124 | if (k_flags & HVhek_FREEKEY) |
f9a63242 | 1125 | Safefree(key); |
a0714e2c | 1126 | return NULL; |
79072805 LW |
1127 | } |
1128 | ||
76e3520e | 1129 | STATIC void |
cea2e8a9 | 1130 | S_hsplit(pTHX_ HV *hv) |
79072805 | 1131 | { |
97aff369 | 1132 | dVAR; |
cbec9347 | 1133 | register XPVHV* xhv = (XPVHV*)SvANY(hv); |
a3b680e6 | 1134 | const I32 oldsize = (I32) xhv->xhv_max+1; /* HvMAX(hv)+1 (sick) */ |
79072805 LW |
1135 | register I32 newsize = oldsize * 2; |
1136 | register I32 i; | |
7b2c381c | 1137 | char *a = (char*) HvARRAY(hv); |
72311751 | 1138 | register HE **aep; |
79072805 | 1139 | register HE **oentry; |
4b5190b5 NC |
1140 | int longest_chain = 0; |
1141 | int was_shared; | |
79072805 | 1142 | |
18026298 NC |
1143 | /*PerlIO_printf(PerlIO_stderr(), "hsplit called for %p which had %d\n", |
1144 | hv, (int) oldsize);*/ | |
1145 | ||
5d88ecd7 | 1146 | if (HvPLACEHOLDERS_get(hv) && !SvREADONLY(hv)) { |
18026298 NC |
1147 | /* Can make this clear any placeholders first for non-restricted hashes, |
1148 | even though Storable rebuilds restricted hashes by putting in all the | |
1149 | placeholders (first) before turning on the readonly flag, because | |
1150 | Storable always pre-splits the hash. */ | |
1151 | hv_clear_placeholders(hv); | |
1152 | } | |
1153 | ||
3280af22 | 1154 | PL_nomemok = TRUE; |
8d6dde3e | 1155 | #if defined(STRANGE_MALLOC) || defined(MYMALLOC) |
b79f7545 NC |
1156 | Renew(a, PERL_HV_ARRAY_ALLOC_BYTES(newsize) |
1157 | + (SvOOK(hv) ? sizeof(struct xpvhv_aux) : 0), char); | |
422a93e5 | 1158 | if (!a) { |
4a33f861 | 1159 | PL_nomemok = FALSE; |
422a93e5 GA |
1160 | return; |
1161 | } | |
b79f7545 | 1162 | if (SvOOK(hv)) { |
7a9b70e9 | 1163 | Copy(&a[oldsize * sizeof(HE*)], &a[newsize * sizeof(HE*)], 1, struct xpvhv_aux); |
b79f7545 | 1164 | } |
4633a7c4 | 1165 | #else |
a02a5408 | 1166 | Newx(a, PERL_HV_ARRAY_ALLOC_BYTES(newsize) |
b79f7545 | 1167 | + (SvOOK(hv) ? sizeof(struct xpvhv_aux) : 0), char); |
422a93e5 | 1168 | if (!a) { |
3280af22 | 1169 | PL_nomemok = FALSE; |
422a93e5 GA |
1170 | return; |
1171 | } | |
7b2c381c | 1172 | Copy(HvARRAY(hv), a, oldsize * sizeof(HE*), char); |
b79f7545 NC |
1173 | if (SvOOK(hv)) { |
1174 | Copy(HvAUX(hv), &a[newsize * sizeof(HE*)], 1, struct xpvhv_aux); | |
1175 | } | |
fba3b22e | 1176 | if (oldsize >= 64) { |
7b2c381c | 1177 | offer_nice_chunk(HvARRAY(hv), |
b79f7545 NC |
1178 | PERL_HV_ARRAY_ALLOC_BYTES(oldsize) |
1179 | + (SvOOK(hv) ? sizeof(struct xpvhv_aux) : 0)); | |
4633a7c4 LW |
1180 | } |
1181 | else | |
7b2c381c | 1182 | Safefree(HvARRAY(hv)); |
4633a7c4 LW |
1183 | #endif |
1184 | ||
3280af22 | 1185 | PL_nomemok = FALSE; |
72311751 | 1186 | Zero(&a[oldsize * sizeof(HE*)], (newsize-oldsize) * sizeof(HE*), char); /* zero 2nd half*/ |
cbec9347 | 1187 | xhv->xhv_max = --newsize; /* HvMAX(hv) = --newsize */ |
7b2c381c | 1188 | HvARRAY(hv) = (HE**) a; |
72311751 | 1189 | aep = (HE**)a; |
79072805 | 1190 | |
72311751 | 1191 | for (i=0; i<oldsize; i++,aep++) { |
4b5190b5 NC |
1192 | int left_length = 0; |
1193 | int right_length = 0; | |
a3b680e6 AL |
1194 | register HE *entry; |
1195 | register HE **bep; | |
4b5190b5 | 1196 | |
72311751 | 1197 | if (!*aep) /* non-existent */ |
79072805 | 1198 | continue; |
72311751 GS |
1199 | bep = aep+oldsize; |
1200 | for (oentry = aep, entry = *aep; entry; entry = *oentry) { | |
eb160463 | 1201 | if ((HeHASH(entry) & newsize) != (U32)i) { |
fde52b5c | 1202 | *oentry = HeNEXT(entry); |
72311751 GS |
1203 | HeNEXT(entry) = *bep; |
1204 | if (!*bep) | |
cbec9347 | 1205 | xhv->xhv_fill++; /* HvFILL(hv)++ */ |
72311751 | 1206 | *bep = entry; |
4b5190b5 | 1207 | right_length++; |
79072805 LW |
1208 | continue; |
1209 | } | |
4b5190b5 | 1210 | else { |
fde52b5c | 1211 | oentry = &HeNEXT(entry); |
4b5190b5 NC |
1212 | left_length++; |
1213 | } | |
79072805 | 1214 | } |
72311751 | 1215 | if (!*aep) /* everything moved */ |
cbec9347 | 1216 | xhv->xhv_fill--; /* HvFILL(hv)-- */ |
4b5190b5 NC |
1217 | /* I think we don't actually need to keep track of the longest length, |
1218 | merely flag if anything is too long. But for the moment while | |
1219 | developing this code I'll track it. */ | |
1220 | if (left_length > longest_chain) | |
1221 | longest_chain = left_length; | |
1222 | if (right_length > longest_chain) | |
1223 | longest_chain = right_length; | |
1224 | } | |
1225 | ||
1226 | ||
1227 | /* Pick your policy for "hashing isn't working" here: */ | |
fdcd69b6 | 1228 | if (longest_chain <= HV_MAX_LENGTH_BEFORE_SPLIT /* split worked? */ |
4b5190b5 NC |
1229 | || HvREHASH(hv)) { |
1230 | return; | |
79072805 | 1231 | } |
4b5190b5 NC |
1232 | |
1233 | if (hv == PL_strtab) { | |
1234 | /* Urg. Someone is doing something nasty to the string table. | |
1235 | Can't win. */ | |
1236 | return; | |
1237 | } | |
1238 | ||
1239 | /* Awooga. Awooga. Pathological data. */ | |
fdcd69b6 | 1240 | /*PerlIO_printf(PerlIO_stderr(), "%p %d of %d with %d/%d buckets\n", hv, |
4b5190b5 NC |
1241 | longest_chain, HvTOTALKEYS(hv), HvFILL(hv), 1+HvMAX(hv));*/ |
1242 | ||
1243 | ++newsize; | |
a02a5408 | 1244 | Newxz(a, PERL_HV_ARRAY_ALLOC_BYTES(newsize) |
b79f7545 NC |
1245 | + (SvOOK(hv) ? sizeof(struct xpvhv_aux) : 0), char); |
1246 | if (SvOOK(hv)) { | |
1247 | Copy(HvAUX(hv), &a[newsize * sizeof(HE*)], 1, struct xpvhv_aux); | |
1248 | } | |
1249 | ||
4b5190b5 NC |
1250 | was_shared = HvSHAREKEYS(hv); |
1251 | ||
1252 | xhv->xhv_fill = 0; | |
1253 | HvSHAREKEYS_off(hv); | |
1254 | HvREHASH_on(hv); | |
1255 | ||
7b2c381c | 1256 | aep = HvARRAY(hv); |
4b5190b5 NC |
1257 | |
1258 | for (i=0; i<newsize; i++,aep++) { | |
a3b680e6 | 1259 | register HE *entry = *aep; |
4b5190b5 NC |
1260 | while (entry) { |
1261 | /* We're going to trash this HE's next pointer when we chain it | |
1262 | into the new hash below, so store where we go next. */ | |
9d4ba2ae | 1263 | HE * const next = HeNEXT(entry); |
4b5190b5 | 1264 | UV hash; |
a3b680e6 | 1265 | HE **bep; |
4b5190b5 NC |
1266 | |
1267 | /* Rehash it */ | |
1268 | PERL_HASH_INTERNAL(hash, HeKEY(entry), HeKLEN(entry)); | |
1269 | ||
1270 | if (was_shared) { | |
1271 | /* Unshare it. */ | |
aec46f14 | 1272 | HEK * const new_hek |
4b5190b5 NC |
1273 | = save_hek_flags(HeKEY(entry), HeKLEN(entry), |
1274 | hash, HeKFLAGS(entry)); | |
1275 | unshare_hek (HeKEY_hek(entry)); | |
1276 | HeKEY_hek(entry) = new_hek; | |
1277 | } else { | |
1278 | /* Not shared, so simply write the new hash in. */ | |
1279 | HeHASH(entry) = hash; | |
1280 | } | |
1281 | /*PerlIO_printf(PerlIO_stderr(), "%d ", HeKFLAGS(entry));*/ | |
1282 | HEK_REHASH_on(HeKEY_hek(entry)); | |
1283 | /*PerlIO_printf(PerlIO_stderr(), "%d\n", HeKFLAGS(entry));*/ | |
1284 | ||
1285 | /* Copy oentry to the correct new chain. */ | |
1286 | bep = ((HE**)a) + (hash & (I32) xhv->xhv_max); | |
1287 | if (!*bep) | |
1288 | xhv->xhv_fill++; /* HvFILL(hv)++ */ | |
1289 | HeNEXT(entry) = *bep; | |
1290 | *bep = entry; | |
1291 | ||
1292 | entry = next; | |
1293 | } | |
1294 | } | |
7b2c381c NC |
1295 | Safefree (HvARRAY(hv)); |
1296 | HvARRAY(hv) = (HE **)a; | |
79072805 LW |
1297 | } |
1298 | ||
72940dca | 1299 | void |
864dbfa3 | 1300 | Perl_hv_ksplit(pTHX_ HV *hv, IV newmax) |
72940dca | 1301 | { |
97aff369 | 1302 | dVAR; |
cbec9347 | 1303 | register XPVHV* xhv = (XPVHV*)SvANY(hv); |
a3b680e6 | 1304 | const I32 oldsize = (I32) xhv->xhv_max+1; /* HvMAX(hv)+1 (sick) */ |
72940dca | 1305 | register I32 newsize; |
1306 | register I32 i; | |
72311751 GS |
1307 | register char *a; |
1308 | register HE **aep; | |
72940dca | 1309 | register HE *entry; |
1310 | register HE **oentry; | |
1311 | ||
1312 | newsize = (I32) newmax; /* possible truncation here */ | |
1313 | if (newsize != newmax || newmax <= oldsize) | |
1314 | return; | |
1315 | while ((newsize & (1 + ~newsize)) != newsize) { | |
1316 | newsize &= ~(newsize & (1 + ~newsize)); /* get proper power of 2 */ | |
1317 | } | |
1318 | if (newsize < newmax) | |
1319 | newsize *= 2; | |
1320 | if (newsize < newmax) | |
1321 | return; /* overflow detection */ | |
1322 | ||
7b2c381c | 1323 | a = (char *) HvARRAY(hv); |
72940dca | 1324 | if (a) { |
3280af22 | 1325 | PL_nomemok = TRUE; |
8d6dde3e | 1326 | #if defined(STRANGE_MALLOC) || defined(MYMALLOC) |
b79f7545 NC |
1327 | Renew(a, PERL_HV_ARRAY_ALLOC_BYTES(newsize) |
1328 | + (SvOOK(hv) ? sizeof(struct xpvhv_aux) : 0), char); | |
8aacddc1 | 1329 | if (!a) { |
4a33f861 | 1330 | PL_nomemok = FALSE; |
422a93e5 GA |
1331 | return; |
1332 | } | |
b79f7545 | 1333 | if (SvOOK(hv)) { |
7a9b70e9 | 1334 | Copy(&a[oldsize * sizeof(HE*)], &a[newsize * sizeof(HE*)], 1, struct xpvhv_aux); |
b79f7545 | 1335 | } |
72940dca | 1336 | #else |
a02a5408 | 1337 | Newx(a, PERL_HV_ARRAY_ALLOC_BYTES(newsize) |
b79f7545 | 1338 | + (SvOOK(hv) ? sizeof(struct xpvhv_aux) : 0), char); |
8aacddc1 | 1339 | if (!a) { |
3280af22 | 1340 | PL_nomemok = FALSE; |
422a93e5 GA |
1341 | return; |
1342 | } | |
7b2c381c | 1343 | Copy(HvARRAY(hv), a, oldsize * sizeof(HE*), char); |
b79f7545 NC |
1344 | if (SvOOK(hv)) { |
1345 | Copy(HvAUX(hv), &a[newsize * sizeof(HE*)], 1, struct xpvhv_aux); | |
1346 | } | |
fba3b22e | 1347 | if (oldsize >= 64) { |
7b2c381c | 1348 | offer_nice_chunk(HvARRAY(hv), |
b79f7545 NC |
1349 | PERL_HV_ARRAY_ALLOC_BYTES(oldsize) |
1350 | + (SvOOK(hv) ? sizeof(struct xpvhv_aux) : 0)); | |
72940dca | 1351 | } |
1352 | else | |
7b2c381c | 1353 | Safefree(HvARRAY(hv)); |
72940dca | 1354 | #endif |
3280af22 | 1355 | PL_nomemok = FALSE; |
72311751 | 1356 | Zero(&a[oldsize * sizeof(HE*)], (newsize-oldsize) * sizeof(HE*), char); /* zero 2nd half*/ |
72940dca | 1357 | } |
1358 | else { | |
a02a5408 | 1359 | Newxz(a, PERL_HV_ARRAY_ALLOC_BYTES(newsize), char); |
72940dca | 1360 | } |
cbec9347 | 1361 | xhv->xhv_max = --newsize; /* HvMAX(hv) = --newsize */ |
7b2c381c | 1362 | HvARRAY(hv) = (HE **) a; |
cbec9347 | 1363 | if (!xhv->xhv_fill /* !HvFILL(hv) */) /* skip rest if no entries */ |
72940dca | 1364 | return; |
1365 | ||
72311751 GS |
1366 | aep = (HE**)a; |
1367 | for (i=0; i<oldsize; i++,aep++) { | |
1368 | if (!*aep) /* non-existent */ | |
72940dca | 1369 | continue; |
72311751 | 1370 | for (oentry = aep, entry = *aep; entry; entry = *oentry) { |
6136c704 AL |
1371 | register I32 j = (HeHASH(entry) & newsize); |
1372 | ||
1373 | if (j != i) { | |
72940dca | 1374 | j -= i; |
1375 | *oentry = HeNEXT(entry); | |
72311751 | 1376 | if (!(HeNEXT(entry) = aep[j])) |
cbec9347 | 1377 | xhv->xhv_fill++; /* HvFILL(hv)++ */ |
72311751 | 1378 | aep[j] = entry; |
72940dca | 1379 | continue; |
1380 | } | |
1381 | else | |
1382 | oentry = &HeNEXT(entry); | |
1383 | } | |
72311751 | 1384 | if (!*aep) /* everything moved */ |
cbec9347 | 1385 | xhv->xhv_fill--; /* HvFILL(hv)-- */ |
72940dca | 1386 | } |
1387 | } | |
1388 | ||
954c1994 GS |
1389 | /* |
1390 | =for apidoc newHV | |
1391 | ||
1392 | Creates a new HV. The reference count is set to 1. | |
1393 | ||
1394 | =cut | |
1395 | */ | |
1396 | ||
79072805 | 1397 | HV * |
864dbfa3 | 1398 | Perl_newHV(pTHX) |
79072805 | 1399 | { |
cbec9347 | 1400 | register XPVHV* xhv; |
561b68a9 | 1401 | HV * const hv = (HV*)newSV(0); |
79072805 | 1402 | |
a0d0e21e | 1403 | sv_upgrade((SV *)hv, SVt_PVHV); |
cbec9347 | 1404 | xhv = (XPVHV*)SvANY(hv); |
79072805 LW |
1405 | SvPOK_off(hv); |
1406 | SvNOK_off(hv); | |
1c846c1f | 1407 | #ifndef NODEFAULT_SHAREKEYS |
fde52b5c | 1408 | HvSHAREKEYS_on(hv); /* key-sharing on by default */ |
1c846c1f | 1409 | #endif |
4b5190b5 | 1410 | |
cbec9347 JH |
1411 | xhv->xhv_max = 7; /* HvMAX(hv) = 7 (start with 8 buckets) */ |
1412 | xhv->xhv_fill = 0; /* HvFILL(hv) = 0 */ | |
79072805 LW |
1413 | return hv; |
1414 | } | |
1415 | ||
b3ac6de7 | 1416 | HV * |
864dbfa3 | 1417 | Perl_newHVhv(pTHX_ HV *ohv) |
b3ac6de7 | 1418 | { |
9d4ba2ae | 1419 | HV * const hv = newHV(); |
4beac62f | 1420 | STRLEN hv_max, hv_fill; |
4beac62f AMS |
1421 | |
1422 | if (!ohv || (hv_fill = HvFILL(ohv)) == 0) | |
1423 | return hv; | |
4beac62f | 1424 | hv_max = HvMAX(ohv); |
b3ac6de7 | 1425 | |
b56ba0bf AMS |
1426 | if (!SvMAGICAL((SV *)ohv)) { |
1427 | /* It's an ordinary hash, so copy it fast. AMS 20010804 */ | |
eb160463 | 1428 | STRLEN i; |
a3b680e6 | 1429 | const bool shared = !!HvSHAREKEYS(ohv); |
aec46f14 | 1430 | HE **ents, ** const oents = (HE **)HvARRAY(ohv); |
ff875642 | 1431 | char *a; |
a02a5408 | 1432 | Newx(a, PERL_HV_ARRAY_ALLOC_BYTES(hv_max+1), char); |
ff875642 | 1433 | ents = (HE**)a; |
b56ba0bf AMS |
1434 | |
1435 | /* In each bucket... */ | |
1436 | for (i = 0; i <= hv_max; i++) { | |
6136c704 | 1437 | HE *prev = NULL; |
aec46f14 | 1438 | HE *oent = oents[i]; |
b56ba0bf AMS |
1439 | |
1440 | if (!oent) { | |
1441 | ents[i] = NULL; | |
1442 | continue; | |
1443 | } | |
1444 | ||
1445 | /* Copy the linked list of entries. */ | |
aec46f14 | 1446 | for (; oent; oent = HeNEXT(oent)) { |
a3b680e6 AL |
1447 | const U32 hash = HeHASH(oent); |
1448 | const char * const key = HeKEY(oent); | |
1449 | const STRLEN len = HeKLEN(oent); | |
1450 | const int flags = HeKFLAGS(oent); | |
6136c704 | 1451 | HE * const ent = new_HE(); |
b56ba0bf | 1452 | |
45dea987 | 1453 | HeVAL(ent) = newSVsv(HeVAL(oent)); |
19692e8d | 1454 | HeKEY_hek(ent) |
6e838c70 | 1455 | = shared ? share_hek_flags(key, len, hash, flags) |
19692e8d | 1456 | : save_hek_flags(key, len, hash, flags); |
b56ba0bf AMS |
1457 | if (prev) |
1458 | HeNEXT(prev) = ent; | |
1459 | else | |
1460 | ents[i] = ent; | |
1461 | prev = ent; | |
1462 | HeNEXT(ent) = NULL; | |
1463 | } | |
1464 | } | |
1465 | ||
1466 | HvMAX(hv) = hv_max; | |
1467 | HvFILL(hv) = hv_fill; | |
8aacddc1 | 1468 | HvTOTALKEYS(hv) = HvTOTALKEYS(ohv); |
b56ba0bf | 1469 | HvARRAY(hv) = ents; |
aec46f14 | 1470 | } /* not magical */ |
b56ba0bf AMS |
1471 | else { |
1472 | /* Iterate over ohv, copying keys and values one at a time. */ | |
b3ac6de7 | 1473 | HE *entry; |
bfcb3514 NC |
1474 | const I32 riter = HvRITER_get(ohv); |
1475 | HE * const eiter = HvEITER_get(ohv); | |
b56ba0bf AMS |
1476 | |
1477 | /* Can we use fewer buckets? (hv_max is always 2^n-1) */ | |
1478 | while (hv_max && hv_max + 1 >= hv_fill * 2) | |
1479 | hv_max = hv_max / 2; | |
1480 | HvMAX(hv) = hv_max; | |
1481 | ||
4a76a316 | 1482 | hv_iterinit(ohv); |
e16e2ff8 | 1483 | while ((entry = hv_iternext_flags(ohv, 0))) { |
19692e8d NC |
1484 | hv_store_flags(hv, HeKEY(entry), HeKLEN(entry), |
1485 | newSVsv(HeVAL(entry)), HeHASH(entry), | |
1486 | HeKFLAGS(entry)); | |
b3ac6de7 | 1487 | } |
bfcb3514 NC |
1488 | HvRITER_set(ohv, riter); |
1489 | HvEITER_set(ohv, eiter); | |
b3ac6de7 | 1490 | } |
1c846c1f | 1491 | |
b3ac6de7 IZ |
1492 | return hv; |
1493 | } | |
1494 | ||
5b9c0671 NC |
1495 | /* A rather specialised version of newHVhv for copying %^H, ensuring all the |
1496 | magic stays on it. */ | |
1497 | HV * | |
1498 | Perl_hv_copy_hints_hv(pTHX_ HV *const ohv) | |
1499 | { | |
1500 | HV * const hv = newHV(); | |
1501 | STRLEN hv_fill; | |
1502 | ||
1503 | if (ohv && (hv_fill = HvFILL(ohv))) { | |
1504 | STRLEN hv_max = HvMAX(ohv); | |
1505 | HE *entry; | |
1506 | const I32 riter = HvRITER_get(ohv); | |
1507 | HE * const eiter = HvEITER_get(ohv); | |
1508 | ||
1509 | while (hv_max && hv_max + 1 >= hv_fill * 2) | |
1510 | hv_max = hv_max / 2; | |
1511 | HvMAX(hv) = hv_max; | |
1512 | ||
1513 | hv_iterinit(ohv); | |
1514 | while ((entry = hv_iternext_flags(ohv, 0))) { | |
1515 | SV *const sv = newSVsv(HeVAL(entry)); | |
1516 | sv_magic(sv, NULL, PERL_MAGIC_hintselem, | |
1517 | (char *)newSVhek (HeKEY_hek(entry)), HEf_SVKEY); | |
1518 | hv_store_flags(hv, HeKEY(entry), HeKLEN(entry), | |
1519 | sv, HeHASH(entry), HeKFLAGS(entry)); | |
1520 | } | |
1521 | HvRITER_set(ohv, riter); | |
1522 | HvEITER_set(ohv, eiter); | |
1523 | } | |
1524 | hv_magic(hv, NULL, PERL_MAGIC_hints); | |
1525 | return hv; | |
1526 | } | |
1527 | ||
79072805 | 1528 | void |
864dbfa3 | 1529 | Perl_hv_free_ent(pTHX_ HV *hv, register HE *entry) |
79072805 | 1530 | { |
97aff369 | 1531 | dVAR; |
16bdeea2 GS |
1532 | SV *val; |
1533 | ||
68dc0745 | 1534 | if (!entry) |
79072805 | 1535 | return; |
16bdeea2 | 1536 | val = HeVAL(entry); |
bfcb3514 | 1537 | if (val && isGV(val) && GvCVu(val) && HvNAME_get(hv)) |
3280af22 | 1538 | PL_sub_generation++; /* may be deletion of method from stash */ |
16bdeea2 | 1539 | SvREFCNT_dec(val); |
68dc0745 | 1540 | if (HeKLEN(entry) == HEf_SVKEY) { |
1541 | SvREFCNT_dec(HeKEY_sv(entry)); | |
8aacddc1 | 1542 | Safefree(HeKEY_hek(entry)); |
44a8e56a | 1543 | } |
1544 | else if (HvSHAREKEYS(hv)) | |
68dc0745 | 1545 | unshare_hek(HeKEY_hek(entry)); |
fde52b5c | 1546 | else |
68dc0745 | 1547 | Safefree(HeKEY_hek(entry)); |
d33b2eba | 1548 | del_HE(entry); |
79072805 LW |
1549 | } |
1550 | ||
1551 | void | |
864dbfa3 | 1552 | Perl_hv_delayfree_ent(pTHX_ HV *hv, register HE *entry) |
79072805 | 1553 | { |
97aff369 | 1554 | dVAR; |
68dc0745 | 1555 | if (!entry) |
79072805 | 1556 | return; |
bc4947fc NC |
1557 | /* SvREFCNT_inc to counter the SvREFCNT_dec in hv_free_ent */ |
1558 | sv_2mortal(SvREFCNT_inc(HeVAL(entry))); /* free between statements */ | |
68dc0745 | 1559 | if (HeKLEN(entry) == HEf_SVKEY) { |
bc4947fc | 1560 | sv_2mortal(SvREFCNT_inc(HeKEY_sv(entry))); |
44a8e56a | 1561 | } |
bc4947fc | 1562 | hv_free_ent(hv, entry); |
79072805 LW |
1563 | } |
1564 | ||
954c1994 GS |
1565 | /* |
1566 | =for apidoc hv_clear | |
1567 | ||
1568 | Clears a hash, making it empty. | |
1569 | ||
1570 | =cut | |
1571 | */ | |
1572 | ||
79072805 | 1573 | void |
864dbfa3 | 1574 | Perl_hv_clear(pTHX_ HV *hv) |
79072805 | 1575 | { |
27da23d5 | 1576 | dVAR; |
cbec9347 | 1577 | register XPVHV* xhv; |
79072805 LW |
1578 | if (!hv) |
1579 | return; | |
49293501 | 1580 | |
ecae49c0 NC |
1581 | DEBUG_A(Perl_hv_assert(aTHX_ hv)); |
1582 | ||
34c3c4e3 DM |
1583 | xhv = (XPVHV*)SvANY(hv); |
1584 | ||
7b2c381c | 1585 | if (SvREADONLY(hv) && HvARRAY(hv) != NULL) { |
34c3c4e3 | 1586 | /* restricted hash: convert all keys to placeholders */ |
b464bac0 AL |
1587 | STRLEN i; |
1588 | for (i = 0; i <= xhv->xhv_max; i++) { | |
7b2c381c | 1589 | HE *entry = (HvARRAY(hv))[i]; |
3a676441 JH |
1590 | for (; entry; entry = HeNEXT(entry)) { |
1591 | /* not already placeholder */ | |
7996736c | 1592 | if (HeVAL(entry) != &PL_sv_placeholder) { |
3a676441 | 1593 | if (HeVAL(entry) && SvREADONLY(HeVAL(entry))) { |
6136c704 | 1594 | SV* const keysv = hv_iterkeysv(entry); |
3a676441 | 1595 | Perl_croak(aTHX_ |
95b63a38 JH |
1596 | "Attempt to delete readonly key '%"SVf"' from a restricted hash", |
1597 | (void*)keysv); | |
3a676441 JH |
1598 | } |
1599 | SvREFCNT_dec(HeVAL(entry)); | |
7996736c | 1600 | HeVAL(entry) = &PL_sv_placeholder; |
ca732855 | 1601 | HvPLACEHOLDERS(hv)++; |
3a676441 | 1602 | } |
34c3c4e3 DM |
1603 | } |
1604 | } | |
df8c6964 | 1605 | goto reset; |
49293501 MS |
1606 | } |
1607 | ||
463ee0b2 | 1608 | hfreeentries(hv); |
ca732855 | 1609 | HvPLACEHOLDERS_set(hv, 0); |
7b2c381c | 1610 | if (HvARRAY(hv)) |
41f62432 | 1611 | Zero(HvARRAY(hv), xhv->xhv_max+1 /* HvMAX(hv)+1 */, HE*); |
a0d0e21e LW |
1612 | |
1613 | if (SvRMAGICAL(hv)) | |
1c846c1f | 1614 | mg_clear((SV*)hv); |
574c8022 | 1615 | |
19692e8d | 1616 | HvHASKFLAGS_off(hv); |
bb443f97 | 1617 | HvREHASH_off(hv); |
df8c6964 | 1618 | reset: |
b79f7545 | 1619 | if (SvOOK(hv)) { |
bfcb3514 NC |
1620 | HvEITER_set(hv, NULL); |
1621 | } | |
79072805 LW |
1622 | } |
1623 | ||
3540d4ce AB |
1624 | /* |
1625 | =for apidoc hv_clear_placeholders | |
1626 | ||
1627 | Clears any placeholders from a hash. If a restricted hash has any of its keys | |
1628 | marked as readonly and the key is subsequently deleted, the key is not actually | |
1629 | deleted but is marked by assigning it a value of &PL_sv_placeholder. This tags | |
1630 | it so it will be ignored by future operations such as iterating over the hash, | |
4cdaeff7 | 1631 | but will still allow the hash to have a value reassigned to the key at some |
3540d4ce AB |
1632 | future point. This function clears any such placeholder keys from the hash. |
1633 | See Hash::Util::lock_keys() for an example of its use. | |
1634 | ||
1635 | =cut | |
1636 | */ | |
1637 | ||
1638 | void | |
1639 | Perl_hv_clear_placeholders(pTHX_ HV *hv) | |
1640 | { | |
27da23d5 | 1641 | dVAR; |
b3ca2e83 NC |
1642 | const U32 items = (U32)HvPLACEHOLDERS_get(hv); |
1643 | ||
1644 | if (items) | |
1645 | clear_placeholders(hv, items); | |
1646 | } | |
1647 | ||
1648 | static void | |
1649 | S_clear_placeholders(pTHX_ HV *hv, U32 items) | |
1650 | { | |
1651 | dVAR; | |
b464bac0 | 1652 | I32 i; |
d3677389 NC |
1653 | |
1654 | if (items == 0) | |
1655 | return; | |
1656 | ||
b464bac0 | 1657 | i = HvMAX(hv); |
d3677389 NC |
1658 | do { |
1659 | /* Loop down the linked list heads */ | |
6136c704 | 1660 | bool first = TRUE; |
d3677389 | 1661 | HE **oentry = &(HvARRAY(hv))[i]; |
cf6db12b | 1662 | HE *entry; |
d3677389 | 1663 | |
cf6db12b | 1664 | while ((entry = *oentry)) { |
d3677389 NC |
1665 | if (HeVAL(entry) == &PL_sv_placeholder) { |
1666 | *oentry = HeNEXT(entry); | |
1667 | if (first && !*oentry) | |
1668 | HvFILL(hv)--; /* This linked list is now empty. */ | |
2e58978b | 1669 | if (entry == HvEITER_get(hv)) |
d3677389 NC |
1670 | HvLAZYDEL_on(hv); |
1671 | else | |
1672 | hv_free_ent(hv, entry); | |
1673 | ||
1674 | if (--items == 0) { | |
1675 | /* Finished. */ | |
5d88ecd7 | 1676 | HvTOTALKEYS(hv) -= (IV)HvPLACEHOLDERS_get(hv); |
d3677389 NC |
1677 | if (HvKEYS(hv) == 0) |
1678 | HvHASKFLAGS_off(hv); | |
5d88ecd7 | 1679 | HvPLACEHOLDERS_set(hv, 0); |
d3677389 NC |
1680 | return; |
1681 | } | |
213ce8b3 NC |
1682 | } else { |
1683 | oentry = &HeNEXT(entry); | |
6136c704 | 1684 | first = FALSE; |
d3677389 NC |
1685 | } |
1686 | } | |
1687 | } while (--i >= 0); | |
1688 | /* You can't get here, hence assertion should always fail. */ | |
1689 | assert (items == 0); | |
1690 | assert (0); | |
3540d4ce AB |
1691 | } |
1692 | ||
76e3520e | 1693 | STATIC void |
cea2e8a9 | 1694 | S_hfreeentries(pTHX_ HV *hv) |
79072805 | 1695 | { |
23976bdd NC |
1696 | /* This is the array that we're going to restore */ |
1697 | HE **orig_array; | |
1698 | HEK *name; | |
1699 | int attempts = 100; | |
3abe233e | 1700 | |
a0d0e21e | 1701 | if (!HvARRAY(hv)) |
79072805 | 1702 | return; |
a0d0e21e | 1703 | |
23976bdd NC |
1704 | if (SvOOK(hv)) { |
1705 | /* If the hash is actually a symbol table with a name, look after the | |
1706 | name. */ | |
1707 | struct xpvhv_aux *iter = HvAUX(hv); | |
1708 | ||
1709 | name = iter->xhv_name; | |
1710 | iter->xhv_name = NULL; | |
1711 | } else { | |
1712 | name = NULL; | |
1713 | } | |
1714 | ||
1715 | orig_array = HvARRAY(hv); | |
1716 | /* orig_array remains unchanged throughout the loop. If after freeing all | |
1717 | the entries it turns out that one of the little blighters has triggered | |
1718 | an action that has caused HvARRAY to be re-allocated, then we set | |
1719 | array to the new HvARRAY, and try again. */ | |
1720 | ||
1721 | while (1) { | |
1722 | /* This is the one we're going to try to empty. First time round | |
1723 | it's the original array. (Hopefully there will only be 1 time | |
1724 | round) */ | |
6136c704 | 1725 | HE ** const array = HvARRAY(hv); |
7440661e | 1726 | I32 i = HvMAX(hv); |
23976bdd NC |
1727 | |
1728 | /* Because we have taken xhv_name out, the only allocated pointer | |
1729 | in the aux structure that might exist is the backreference array. | |
1730 | */ | |
1731 | ||
1732 | if (SvOOK(hv)) { | |
7440661e | 1733 | HE *entry; |
23976bdd NC |
1734 | struct xpvhv_aux *iter = HvAUX(hv); |
1735 | /* If there are weak references to this HV, we need to avoid | |
1736 | freeing them up here. In particular we need to keep the AV | |
1737 | visible as what we're deleting might well have weak references | |
1738 | back to this HV, so the for loop below may well trigger | |
1739 | the removal of backreferences from this array. */ | |
1740 | ||
1741 | if (iter->xhv_backreferences) { | |
1742 | /* So donate them to regular backref magic to keep them safe. | |
1743 | The sv_magic will increase the reference count of the AV, | |
1744 | so we need to drop it first. */ | |
5b285ea4 | 1745 | SvREFCNT_dec(iter->xhv_backreferences); |
23976bdd NC |
1746 | if (AvFILLp(iter->xhv_backreferences) == -1) { |
1747 | /* Turns out that the array is empty. Just free it. */ | |
1748 | SvREFCNT_dec(iter->xhv_backreferences); | |
1b8791d1 | 1749 | |
23976bdd NC |
1750 | } else { |
1751 | sv_magic((SV*)hv, (SV*)iter->xhv_backreferences, | |
1752 | PERL_MAGIC_backref, NULL, 0); | |
1753 | } | |
1754 | iter->xhv_backreferences = NULL; | |
5b285ea4 | 1755 | } |
86f55936 | 1756 | |
23976bdd NC |
1757 | entry = iter->xhv_eiter; /* HvEITER(hv) */ |
1758 | if (entry && HvLAZYDEL(hv)) { /* was deleted earlier? */ | |
1759 | HvLAZYDEL_off(hv); | |
1760 | hv_free_ent(hv, entry); | |
1761 | } | |
1762 | iter->xhv_riter = -1; /* HvRITER(hv) = -1 */ | |
4608196e | 1763 | iter->xhv_eiter = NULL; /* HvEITER(hv) = NULL */ |
b79f7545 | 1764 | |
23976bdd | 1765 | /* There are now no allocated pointers in the aux structure. */ |
2f86008e | 1766 | |
23976bdd NC |
1767 | SvFLAGS(hv) &= ~SVf_OOK; /* Goodbye, aux structure. */ |
1768 | /* What aux structure? */ | |
a0d0e21e | 1769 | } |
bfcb3514 | 1770 | |
23976bdd NC |
1771 | /* make everyone else think the array is empty, so that the destructors |
1772 | * called for freed entries can't recusively mess with us */ | |
1773 | HvARRAY(hv) = NULL; | |
1774 | HvFILL(hv) = 0; | |
1775 | ((XPVHV*) SvANY(hv))->xhv_keys = 0; | |
1776 | ||
7440661e NC |
1777 | |
1778 | do { | |
1779 | /* Loop down the linked list heads */ | |
1780 | HE *entry = array[i]; | |
1781 | ||
1782 | while (entry) { | |
23976bdd NC |
1783 | register HE * const oentry = entry; |
1784 | entry = HeNEXT(entry); | |
1785 | hv_free_ent(hv, oentry); | |
1786 | } | |
7440661e | 1787 | } while (--i >= 0); |
b79f7545 | 1788 | |
23976bdd NC |
1789 | /* As there are no allocated pointers in the aux structure, it's now |
1790 | safe to free the array we just cleaned up, if it's not the one we're | |
1791 | going to put back. */ | |
1792 | if (array != orig_array) { | |
1793 | Safefree(array); | |
1794 | } | |
b79f7545 | 1795 | |
23976bdd NC |
1796 | if (!HvARRAY(hv)) { |
1797 | /* Good. No-one added anything this time round. */ | |
1798 | break; | |
bfcb3514 | 1799 | } |
b79f7545 | 1800 | |
23976bdd NC |
1801 | if (SvOOK(hv)) { |
1802 | /* Someone attempted to iterate or set the hash name while we had | |
1803 | the array set to 0. We'll catch backferences on the next time | |
1804 | round the while loop. */ | |
1805 | assert(HvARRAY(hv)); | |
1b8791d1 | 1806 | |
23976bdd NC |
1807 | if (HvAUX(hv)->xhv_name) { |
1808 | unshare_hek_or_pvn(HvAUX(hv)->xhv_name, 0, 0, 0); | |
1809 | } | |
1810 | } | |
1811 | ||
1812 | if (--attempts == 0) { | |
1813 | Perl_die(aTHX_ "panic: hfreeentries failed to free hash - something is repeatedly re-creating entries"); | |
1814 | } | |
6136c704 | 1815 | } |
23976bdd NC |
1816 | |
1817 | HvARRAY(hv) = orig_array; | |
1818 | ||
1819 | /* If the hash was actually a symbol table, put the name back. */ | |
1820 | if (name) { | |
1821 | /* We have restored the original array. If name is non-NULL, then | |
1822 | the original array had an aux structure at the end. So this is | |
1823 | valid: */ | |
1824 | SvFLAGS(hv) |= SVf_OOK; | |
1825 | HvAUX(hv)->xhv_name = name; | |
1b8791d1 | 1826 | } |
79072805 LW |
1827 | } |
1828 | ||
954c1994 GS |
1829 | /* |
1830 | =for apidoc hv_undef | |
1831 | ||
1832 | Undefines the hash. | |
1833 | ||
1834 | =cut | |
1835 | */ | |
1836 | ||
79072805 | 1837 | void |
864dbfa3 | 1838 | Perl_hv_undef(pTHX_ HV *hv) |
79072805 | 1839 | { |
97aff369 | 1840 | dVAR; |
cbec9347 | 1841 | register XPVHV* xhv; |
bfcb3514 | 1842 | const char *name; |
86f55936 | 1843 | |
79072805 LW |
1844 | if (!hv) |
1845 | return; | |
ecae49c0 | 1846 | DEBUG_A(Perl_hv_assert(aTHX_ hv)); |
cbec9347 | 1847 | xhv = (XPVHV*)SvANY(hv); |
463ee0b2 | 1848 | hfreeentries(hv); |
bfcb3514 | 1849 | if ((name = HvNAME_get(hv))) { |
7e8961ec | 1850 | if(PL_stashcache) |
7423f6db | 1851 | hv_delete(PL_stashcache, name, HvNAMELEN_get(hv), G_DISCARD); |
bd61b366 | 1852 | hv_name_set(hv, NULL, 0, 0); |
85e6fe83 | 1853 | } |
b79f7545 NC |
1854 | SvFLAGS(hv) &= ~SVf_OOK; |
1855 | Safefree(HvARRAY(hv)); | |
cbec9347 | 1856 | xhv->xhv_max = 7; /* HvMAX(hv) = 7 (it's a normal hash) */ |
7b2c381c | 1857 | HvARRAY(hv) = 0; |
ca732855 | 1858 | HvPLACEHOLDERS_set(hv, 0); |
a0d0e21e LW |
1859 | |
1860 | if (SvRMAGICAL(hv)) | |
1c846c1f | 1861 | mg_clear((SV*)hv); |
79072805 LW |
1862 | } |
1863 | ||
b464bac0 | 1864 | static struct xpvhv_aux* |
5f66b61c | 1865 | S_hv_auxinit(HV *hv) { |
bfcb3514 | 1866 | struct xpvhv_aux *iter; |
b79f7545 | 1867 | char *array; |
bfcb3514 | 1868 | |
b79f7545 | 1869 | if (!HvARRAY(hv)) { |
a02a5408 | 1870 | Newxz(array, PERL_HV_ARRAY_ALLOC_BYTES(HvMAX(hv) + 1) |
b79f7545 NC |
1871 | + sizeof(struct xpvhv_aux), char); |
1872 | } else { | |
1873 | array = (char *) HvARRAY(hv); | |
1874 | Renew(array, PERL_HV_ARRAY_ALLOC_BYTES(HvMAX(hv) + 1) | |
1875 | + sizeof(struct xpvhv_aux), char); | |
1876 | } | |
1877 | HvARRAY(hv) = (HE**) array; | |
1878 | /* SvOOK_on(hv) attacks the IV flags. */ | |
1879 | SvFLAGS(hv) |= SVf_OOK; | |
1880 | iter = HvAUX(hv); | |
bfcb3514 NC |
1881 | |
1882 | iter->xhv_riter = -1; /* HvRITER(hv) = -1 */ | |
4608196e | 1883 | iter->xhv_eiter = NULL; /* HvEITER(hv) = NULL */ |
bfcb3514 | 1884 | iter->xhv_name = 0; |
86f55936 | 1885 | iter->xhv_backreferences = 0; |
bfcb3514 NC |
1886 | return iter; |
1887 | } | |
1888 | ||
954c1994 GS |
1889 | /* |
1890 | =for apidoc hv_iterinit | |
1891 | ||
1892 | Prepares a starting point to traverse a hash table. Returns the number of | |
1893 | keys in the hash (i.e. the same as C<HvKEYS(tb)>). The return value is | |
1c846c1f | 1894 | currently only meaningful for hashes without tie magic. |
954c1994 GS |
1895 | |
1896 | NOTE: Before version 5.004_65, C<hv_iterinit> used to return the number of | |
1897 | hash buckets that happen to be in use. If you still need that esoteric | |
1898 | value, you can get it through the macro C<HvFILL(tb)>. | |
1899 | ||
e16e2ff8 | 1900 | |
954c1994 GS |
1901 | =cut |
1902 | */ | |
1903 | ||
79072805 | 1904 | I32 |
864dbfa3 | 1905 | Perl_hv_iterinit(pTHX_ HV *hv) |
79072805 | 1906 | { |
aa689395 | 1907 | if (!hv) |
cea2e8a9 | 1908 | Perl_croak(aTHX_ "Bad hash"); |
bfcb3514 | 1909 | |
b79f7545 | 1910 | if (SvOOK(hv)) { |
6136c704 | 1911 | struct xpvhv_aux * const iter = HvAUX(hv); |
0bd48802 | 1912 | HE * const entry = iter->xhv_eiter; /* HvEITER(hv) */ |
bfcb3514 NC |
1913 | if (entry && HvLAZYDEL(hv)) { /* was deleted earlier? */ |
1914 | HvLAZYDEL_off(hv); | |
1915 | hv_free_ent(hv, entry); | |
1916 | } | |
1917 | iter->xhv_riter = -1; /* HvRITER(hv) = -1 */ | |
4608196e | 1918 | iter->xhv_eiter = NULL; /* HvEITER(hv) = NULL */ |
bfcb3514 | 1919 | } else { |
6136c704 | 1920 | hv_auxinit(hv); |
72940dca | 1921 | } |
bfcb3514 | 1922 | |
cbec9347 | 1923 | /* used to be xhv->xhv_fill before 5.004_65 */ |
5d88ecd7 | 1924 | return HvTOTALKEYS(hv); |
79072805 | 1925 | } |
bfcb3514 NC |
1926 | |
1927 | I32 * | |
1928 | Perl_hv_riter_p(pTHX_ HV *hv) { | |
1929 | struct xpvhv_aux *iter; | |
1930 | ||
1931 | if (!hv) | |
1932 | Perl_croak(aTHX_ "Bad hash"); | |
1933 | ||
6136c704 | 1934 | iter = SvOOK(hv) ? HvAUX(hv) : hv_auxinit(hv); |
bfcb3514 NC |
1935 | return &(iter->xhv_riter); |
1936 | } | |
1937 | ||
1938 | HE ** | |
1939 | Perl_hv_eiter_p(pTHX_ HV *hv) { | |
1940 | struct xpvhv_aux *iter; | |
1941 | ||
1942 | if (!hv) | |
1943 | Perl_croak(aTHX_ "Bad hash"); | |
1944 | ||
6136c704 | 1945 | iter = SvOOK(hv) ? HvAUX(hv) : hv_auxinit(hv); |
bfcb3514 NC |
1946 | return &(iter->xhv_eiter); |
1947 | } | |
1948 | ||
1949 | void | |
1950 | Perl_hv_riter_set(pTHX_ HV *hv, I32 riter) { | |
1951 | struct xpvhv_aux *iter; | |
1952 | ||
1953 | if (!hv) | |
1954 | Perl_croak(aTHX_ "Bad hash"); | |
1955 | ||
b79f7545 NC |
1956 | if (SvOOK(hv)) { |
1957 | iter = HvAUX(hv); | |
1958 | } else { | |
bfcb3514 NC |
1959 | if (riter == -1) |
1960 | return; | |
1961 | ||
6136c704 | 1962 | iter = hv_auxinit(hv); |
bfcb3514 NC |
1963 | } |
1964 | iter->xhv_riter = riter; | |
1965 | } | |
1966 | ||
1967 | void | |
1968 | Perl_hv_eiter_set(pTHX_ HV *hv, HE *eiter) { | |
1969 | struct xpvhv_aux *iter; | |
1970 | ||
1971 | if (!hv) | |
1972 | Perl_croak(aTHX_ "Bad hash"); | |
1973 | ||
b79f7545 NC |
1974 | if (SvOOK(hv)) { |
1975 | iter = HvAUX(hv); | |
1976 | } else { | |
bfcb3514 NC |
1977 | /* 0 is the default so don't go malloc()ing a new structure just to |
1978 | hold 0. */ | |
1979 | if (!eiter) | |
1980 | return; | |
1981 | ||
6136c704 | 1982 | iter = hv_auxinit(hv); |
bfcb3514 NC |
1983 | } |
1984 | iter->xhv_eiter = eiter; | |
1985 | } | |
1986 | ||
bfcb3514 | 1987 | void |
4164be69 | 1988 | Perl_hv_name_set(pTHX_ HV *hv, const char *name, U32 len, U32 flags) |
bfcb3514 | 1989 | { |
97aff369 | 1990 | dVAR; |
b79f7545 | 1991 | struct xpvhv_aux *iter; |
7423f6db | 1992 | U32 hash; |
46c461b5 AL |
1993 | |
1994 | PERL_UNUSED_ARG(flags); | |
bfcb3514 | 1995 | |
4164be69 NC |
1996 | if (len > I32_MAX) |
1997 | Perl_croak(aTHX_ "panic: hv name too long (%"UVuf")", (UV) len); | |
1998 | ||
b79f7545 NC |
1999 | if (SvOOK(hv)) { |
2000 | iter = HvAUX(hv); | |
7423f6db NC |
2001 | if (iter->xhv_name) { |
2002 | unshare_hek_or_pvn(iter->xhv_name, 0, 0, 0); | |
2003 | } | |
16580ff5 | 2004 | } else { |
bfcb3514 NC |
2005 | if (name == 0) |
2006 | return; | |
2007 | ||
6136c704 | 2008 | iter = hv_auxinit(hv); |
bfcb3514 | 2009 | } |
7423f6db NC |
2010 | PERL_HASH(hash, name, len); |
2011 | iter->xhv_name = name ? share_hek(name, len, hash) : 0; | |
bfcb3514 NC |
2012 | } |
2013 | ||
86f55936 NC |
2014 | AV ** |
2015 | Perl_hv_backreferences_p(pTHX_ HV *hv) { | |
6136c704 | 2016 | struct xpvhv_aux * const iter = SvOOK(hv) ? HvAUX(hv) : hv_auxinit(hv); |
96a5add6 | 2017 | PERL_UNUSED_CONTEXT; |
86f55936 NC |
2018 | return &(iter->xhv_backreferences); |
2019 | } | |
2020 | ||
2021 | void | |
2022 | Perl_hv_kill_backrefs(pTHX_ HV *hv) { | |
2023 | AV *av; | |
2024 | ||
2025 | if (!SvOOK(hv)) | |
2026 | return; | |
2027 | ||
2028 | av = HvAUX(hv)->xhv_backreferences; | |
2029 | ||
2030 | if (av) { | |
2031 | HvAUX(hv)->xhv_backreferences = 0; | |
2032 | Perl_sv_kill_backrefs(aTHX_ (SV*) hv, av); | |
2033 | } | |
2034 | } | |
2035 | ||
954c1994 | 2036 | /* |
7a7b9979 NC |
2037 | hv_iternext is implemented as a macro in hv.h |
2038 | ||
954c1994 GS |
2039 | =for apidoc hv_iternext |
2040 | ||
2041 | Returns entries from a hash iterator. See C<hv_iterinit>. | |
2042 | ||
fe7bca90 NC |
2043 | You may call C<hv_delete> or C<hv_delete_ent> on the hash entry that the |
2044 | iterator currently points to, without losing your place or invalidating your | |
2045 | iterator. Note that in this case the current entry is deleted from the hash | |
2046 | with your iterator holding the last reference to it. Your iterator is flagged | |
2047 | to free the entry on the next call to C<hv_iternext>, so you must not discard | |
2048 | your iterator immediately else the entry will leak - call C<hv_iternext> to | |
2049 | trigger the resource deallocation. | |
2050 | ||
fe7bca90 NC |
2051 | =for apidoc hv_iternext_flags |
2052 | ||
2053 | Returns entries from a hash iterator. See C<hv_iterinit> and C<hv_iternext>. | |
2054 | The C<flags> value will normally be zero; if HV_ITERNEXT_WANTPLACEHOLDERS is | |
2055 | set the placeholders keys (for restricted hashes) will be returned in addition | |
2056 | to normal keys. By default placeholders are automatically skipped over. | |
7996736c MHM |
2057 | Currently a placeholder is implemented with a value that is |
2058 | C<&Perl_sv_placeholder>. Note that the implementation of placeholders and | |
fe7bca90 NC |
2059 | restricted hashes may change, and the implementation currently is |
2060 | insufficiently abstracted for any change to be tidy. | |
e16e2ff8 | 2061 | |
fe7bca90 | 2062 | =cut |
e16e2ff8 NC |
2063 | */ |
2064 | ||
2065 | HE * | |
2066 | Perl_hv_iternext_flags(pTHX_ HV *hv, I32 flags) | |
2067 | { | |
27da23d5 | 2068 | dVAR; |
cbec9347 | 2069 | register XPVHV* xhv; |
79072805 | 2070 | register HE *entry; |
a0d0e21e | 2071 | HE *oldentry; |
463ee0b2 | 2072 | MAGIC* mg; |
bfcb3514 | 2073 | struct xpvhv_aux *iter; |
79072805 LW |
2074 | |
2075 | if (!hv) | |
cea2e8a9 | 2076 | Perl_croak(aTHX_ "Bad hash"); |
cbec9347 | 2077 | xhv = (XPVHV*)SvANY(hv); |
bfcb3514 | 2078 | |
b79f7545 | 2079 | if (!SvOOK(hv)) { |
bfcb3514 NC |
2080 | /* Too many things (well, pp_each at least) merrily assume that you can |
2081 | call iv_iternext without calling hv_iterinit, so we'll have to deal | |
2082 | with it. */ | |
2083 | hv_iterinit(hv); | |
bfcb3514 | 2084 | } |
b79f7545 | 2085 | iter = HvAUX(hv); |
bfcb3514 NC |
2086 | |
2087 | oldentry = entry = iter->xhv_eiter; /* HvEITER(hv) */ | |
463ee0b2 | 2088 | |
14befaf4 | 2089 | if ((mg = SvTIED_mg((SV*)hv, PERL_MAGIC_tied))) { |
c4420975 | 2090 | SV * const key = sv_newmortal(); |
cd1469e6 | 2091 | if (entry) { |
fde52b5c | 2092 | sv_setsv(key, HeSVKEY_force(entry)); |
cd1469e6 | 2093 | SvREFCNT_dec(HeSVKEY(entry)); /* get rid of previous key */ |
2094 | } | |
a0d0e21e | 2095 | else { |
ff68c719 | 2096 | char *k; |
bbce6d69 | 2097 | HEK *hek; |
ff68c719 | 2098 | |
cbec9347 | 2099 | /* one HE per MAGICAL hash */ |
bfcb3514 | 2100 | iter->xhv_eiter = entry = new_HE(); /* HvEITER(hv) = new_HE() */ |
4633a7c4 | 2101 | Zero(entry, 1, HE); |
a02a5408 | 2102 | Newxz(k, HEK_BASESIZE + sizeof(SV*), char); |
ff68c719 | 2103 | hek = (HEK*)k; |
2104 | HeKEY_hek(entry) = hek; | |
fde52b5c | 2105 | HeKLEN(entry) = HEf_SVKEY; |
a0d0e21e LW |
2106 | } |
2107 | magic_nextpack((SV*) hv,mg,key); | |
8aacddc1 | 2108 | if (SvOK(key)) { |
cd1469e6 | 2109 | /* force key to stay around until next time */ |
b37c2d43 | 2110 | HeSVKEY_set(entry, SvREFCNT_inc_simple_NN(key)); |
bbce6d69 | 2111 | return entry; /* beware, hent_val is not set */ |
8aacddc1 | 2112 | } |
fde52b5c | 2113 | if (HeVAL(entry)) |
2114 | SvREFCNT_dec(HeVAL(entry)); | |
ff68c719 | 2115 | Safefree(HeKEY_hek(entry)); |
d33b2eba | 2116 | del_HE(entry); |
4608196e RGS |
2117 | iter->xhv_eiter = NULL; /* HvEITER(hv) = NULL */ |
2118 | return NULL; | |
79072805 | 2119 | } |
7ee146b1 | 2120 | #if defined(DYNAMIC_ENV_FETCH) && !defined(__riscos__) /* set up %ENV for iteration */ |
03026e68 | 2121 | if (!entry && SvRMAGICAL((SV*)hv) && mg_find((SV*)hv, PERL_MAGIC_env)) { |
f675dbe5 | 2122 | prime_env_iter(); |
03026e68 JM |
2123 | #ifdef VMS |
2124 | /* The prime_env_iter() on VMS just loaded up new hash values | |
2125 | * so the iteration count needs to be reset back to the beginning | |
2126 | */ | |
2127 | hv_iterinit(hv); | |
2128 | iter = HvAUX(hv); | |
2129 | oldentry = entry = iter->xhv_eiter; /* HvEITER(hv) */ | |
2130 | #endif | |
2131 | } | |
f675dbe5 | 2132 | #endif |
463ee0b2 | 2133 | |
b79f7545 NC |
2134 | /* hv_iterint now ensures this. */ |
2135 | assert (HvARRAY(hv)); | |
2136 | ||
015a5f36 | 2137 | /* At start of hash, entry is NULL. */ |
fde52b5c | 2138 | if (entry) |
8aacddc1 | 2139 | { |
fde52b5c | 2140 | entry = HeNEXT(entry); |
e16e2ff8 NC |
2141 | if (!(flags & HV_ITERNEXT_WANTPLACEHOLDERS)) { |
2142 | /* | |
2143 | * Skip past any placeholders -- don't want to include them in | |
2144 | * any iteration. | |
2145 | */ | |
7996736c | 2146 | while (entry && HeVAL(entry) == &PL_sv_placeholder) { |
e16e2ff8 NC |
2147 | entry = HeNEXT(entry); |
2148 | } | |
8aacddc1 NIS |
2149 | } |
2150 | } | |
fde52b5c | 2151 | while (!entry) { |
015a5f36 NC |
2152 | /* OK. Come to the end of the current list. Grab the next one. */ |
2153 | ||
bfcb3514 NC |
2154 | iter->xhv_riter++; /* HvRITER(hv)++ */ |
2155 | if (iter->xhv_riter > (I32)xhv->xhv_max /* HvRITER(hv) > HvMAX(hv) */) { | |
015a5f36 | 2156 | /* There is no next one. End of the hash. */ |
bfcb3514 | 2157 | iter->xhv_riter = -1; /* HvRITER(hv) = -1 */ |
fde52b5c | 2158 | break; |
79072805 | 2159 | } |
7b2c381c | 2160 | entry = (HvARRAY(hv))[iter->xhv_riter]; |
8aacddc1 | 2161 | |
e16e2ff8 | 2162 | if (!(flags & HV_ITERNEXT_WANTPLACEHOLDERS)) { |
015a5f36 NC |
2163 | /* If we have an entry, but it's a placeholder, don't count it. |
2164 | Try the next. */ | |
7996736c | 2165 | while (entry && HeVAL(entry) == &PL_sv_placeholder) |
015a5f36 NC |
2166 | entry = HeNEXT(entry); |
2167 | } | |
2168 | /* Will loop again if this linked list starts NULL | |
2169 | (for HV_ITERNEXT_WANTPLACEHOLDERS) | |
2170 | or if we run through it and find only placeholders. */ | |
fde52b5c | 2171 | } |
79072805 | 2172 | |
72940dca | 2173 | if (oldentry && HvLAZYDEL(hv)) { /* was deleted earlier? */ |
2174 | HvLAZYDEL_off(hv); | |
68dc0745 | 2175 | hv_free_ent(hv, oldentry); |
72940dca | 2176 | } |
a0d0e21e | 2177 | |
fdcd69b6 NC |
2178 | /*if (HvREHASH(hv) && entry && !HeKREHASH(entry)) |
2179 | PerlIO_printf(PerlIO_stderr(), "Awooga %p %p\n", hv, entry);*/ | |
2180 | ||
bfcb3514 | 2181 | iter->xhv_eiter = entry; /* HvEITER(hv) = entry */ |
79072805 LW |
2182 | return entry; |
2183 | } | |
2184 | ||
954c1994 GS |
2185 | /* |
2186 | =for apidoc hv_iterkey | |
2187 | ||
2188 | Returns the key from the current position of the hash iterator. See | |
2189 | C<hv_iterinit>. | |
2190 | ||
2191 | =cut | |
2192 | */ | |
2193 | ||
79072805 | 2194 | char * |
864dbfa3 | 2195 | Perl_hv_iterkey(pTHX_ register HE *entry, I32 *retlen) |
79072805 | 2196 | { |
fde52b5c | 2197 | if (HeKLEN(entry) == HEf_SVKEY) { |
fb73857a | 2198 | STRLEN len; |
0bd48802 | 2199 | char * const p = SvPV(HeKEY_sv(entry), len); |
fb73857a | 2200 | *retlen = len; |
2201 | return p; | |
fde52b5c | 2202 | } |
2203 | else { | |
2204 | *retlen = HeKLEN(entry); | |
2205 | return HeKEY(entry); | |
2206 | } | |
2207 | } | |
2208 | ||
2209 | /* unlike hv_iterval(), this always returns a mortal copy of the key */ | |
954c1994 GS |
2210 | /* |
2211 | =for apidoc hv_iterkeysv | |
2212 | ||
2213 | Returns the key as an C<SV*> from the current position of the hash | |
2214 | iterator. The return value will always be a mortal copy of the key. Also | |
2215 | see C<hv_iterinit>. | |
2216 | ||
2217 | =cut | |
2218 | */ | |
2219 | ||
fde52b5c | 2220 | SV * |
864dbfa3 | 2221 | Perl_hv_iterkeysv(pTHX_ register HE *entry) |
fde52b5c | 2222 | { |
c1b02ed8 | 2223 | return sv_2mortal(newSVhek(HeKEY_hek(entry))); |
79072805 LW |
2224 | } |
2225 | ||
954c1994 GS |
2226 | /* |
2227 | =for apidoc hv_iterval | |
2228 | ||
2229 | Returns the value from the current position of the hash iterator. See | |
2230 | C<hv_iterkey>. | |
2231 | ||
2232 | =cut | |
2233 | */ | |
2234 | ||
79072805 | 2235 | SV * |
864dbfa3 | 2236 | Perl_hv_iterval(pTHX_ HV *hv, register HE *entry) |
79072805 | 2237 | { |
8990e307 | 2238 | if (SvRMAGICAL(hv)) { |
14befaf4 | 2239 | if (mg_find((SV*)hv, PERL_MAGIC_tied)) { |
c4420975 | 2240 | SV* const sv = sv_newmortal(); |
bbce6d69 | 2241 | if (HeKLEN(entry) == HEf_SVKEY) |
2242 | mg_copy((SV*)hv, sv, (char*)HeKEY_sv(entry), HEf_SVKEY); | |
a3b680e6 AL |
2243 | else |
2244 | mg_copy((SV*)hv, sv, HeKEY(entry), HeKLEN(entry)); | |
463ee0b2 LW |
2245 | return sv; |
2246 | } | |
79072805 | 2247 | } |
fde52b5c | 2248 | return HeVAL(entry); |
79072805 LW |
2249 | } |
2250 | ||
954c1994 GS |
2251 | /* |
2252 | =for apidoc hv_iternextsv | |
2253 | ||
2254 | Performs an C<hv_iternext>, C<hv_iterkey>, and C<hv_iterval> in one | |
2255 | operation. | |
2256 | ||
2257 | =cut | |
2258 | */ | |
2259 | ||
a0d0e21e | 2260 | SV * |
864dbfa3 | 2261 | Perl_hv_iternextsv(pTHX_ HV *hv, char **key, I32 *retlen) |
a0d0e21e | 2262 | { |
0bd48802 AL |
2263 | HE * const he = hv_iternext_flags(hv, 0); |
2264 | ||
2265 | if (!he) | |
a0d0e21e LW |
2266 | return NULL; |
2267 | *key = hv_iterkey(he, retlen); | |
2268 | return hv_iterval(hv, he); | |
2269 | } | |
2270 | ||
954c1994 | 2271 | /* |
bc5cdc23 NC |
2272 | |
2273 | Now a macro in hv.h | |
2274 | ||
954c1994 GS |
2275 | =for apidoc hv_magic |
2276 | ||
2277 | Adds magic to a hash. See C<sv_magic>. | |
2278 | ||
2279 | =cut | |
2280 | */ | |
2281 | ||
bbce6d69 | 2282 | /* possibly free a shared string if no one has access to it |
fde52b5c | 2283 | * len and hash must both be valid for str. |
2284 | */ | |
bbce6d69 | 2285 | void |
864dbfa3 | 2286 | Perl_unsharepvn(pTHX_ const char *str, I32 len, U32 hash) |
fde52b5c | 2287 | { |
19692e8d NC |
2288 | unshare_hek_or_pvn (NULL, str, len, hash); |
2289 | } | |
2290 | ||
2291 | ||
2292 | void | |
2293 | Perl_unshare_hek(pTHX_ HEK *hek) | |
2294 | { | |
2295 | unshare_hek_or_pvn(hek, NULL, 0, 0); | |
2296 | } | |
2297 | ||
2298 | /* possibly free a shared string if no one has access to it | |
2299 | hek if non-NULL takes priority over the other 3, else str, len and hash | |
2300 | are used. If so, len and hash must both be valid for str. | |
2301 | */ | |
df132699 | 2302 | STATIC void |
97ddebaf | 2303 | S_unshare_hek_or_pvn(pTHX_ const HEK *hek, const char *str, I32 len, U32 hash) |
19692e8d | 2304 | { |
97aff369 | 2305 | dVAR; |
cbec9347 | 2306 | register XPVHV* xhv; |
20454177 | 2307 | HE *entry; |
fde52b5c | 2308 | register HE **oentry; |
45d1cc86 | 2309 | HE **first; |
c3654f1a | 2310 | bool is_utf8 = FALSE; |
19692e8d | 2311 | int k_flags = 0; |
aec46f14 | 2312 | const char * const save = str; |
cbbf8932 | 2313 | struct shared_he *he = NULL; |
c3654f1a | 2314 | |
19692e8d | 2315 | if (hek) { |
cbae3960 NC |
2316 | /* Find the shared he which is just before us in memory. */ |
2317 | he = (struct shared_he *)(((char *)hek) | |
2318 | - STRUCT_OFFSET(struct shared_he, | |
2319 | shared_he_hek)); | |
2320 | ||
2321 | /* Assert that the caller passed us a genuine (or at least consistent) | |
2322 | shared hek */ | |
2323 | assert (he->shared_he_he.hent_hek == hek); | |
29404ae0 NC |
2324 | |
2325 | LOCK_STRTAB_MUTEX; | |
de616631 NC |
2326 | if (he->shared_he_he.he_valu.hent_refcount - 1) { |
2327 | --he->shared_he_he.he_valu.hent_refcount; | |
29404ae0 NC |
2328 | UNLOCK_STRTAB_MUTEX; |
2329 | return; | |
2330 | } | |
2331 | UNLOCK_STRTAB_MUTEX; | |
2332 | ||
19692e8d NC |
2333 | hash = HEK_HASH(hek); |
2334 | } else if (len < 0) { | |
2335 | STRLEN tmplen = -len; | |
2336 | is_utf8 = TRUE; | |
2337 | /* See the note in hv_fetch(). --jhi */ | |
2338 | str = (char*)bytes_from_utf8((U8*)str, &tmplen, &is_utf8); | |
2339 | len = tmplen; | |
2340 | if (is_utf8) | |
2341 | k_flags = HVhek_UTF8; | |
2342 | if (str != save) | |
2343 | k_flags |= HVhek_WASUTF8 | HVhek_FREEKEY; | |
c3654f1a | 2344 | } |
1c846c1f | 2345 | |
de616631 | 2346 | /* what follows was the moral equivalent of: |
6b88bc9c | 2347 | if ((Svp = hv_fetch(PL_strtab, tmpsv, FALSE, hash))) { |
a0714e2c | 2348 | if (--*Svp == NULL) |
6b88bc9c | 2349 | hv_delete(PL_strtab, str, len, G_DISCARD, hash); |
bbce6d69 | 2350 | } */ |
cbec9347 | 2351 | xhv = (XPVHV*)SvANY(PL_strtab); |
fde52b5c | 2352 | /* assert(xhv_array != 0) */ |
5f08fbcd | 2353 | LOCK_STRTAB_MUTEX; |
45d1cc86 | 2354 | first = oentry = &(HvARRAY(PL_strtab))[hash & (I32) HvMAX(PL_strtab)]; |
6c1b96a1 NC |
2355 | if (he) { |
2356 | const HE *const he_he = &(he->shared_he_he); | |
45d1cc86 | 2357 | for (entry = *oentry; entry; oentry = &HeNEXT(entry), entry = *oentry) { |
35ab5632 NC |
2358 | if (entry == he_he) |
2359 | break; | |
19692e8d NC |
2360 | } |
2361 | } else { | |
35a4481c | 2362 | const int flags_masked = k_flags & HVhek_MASK; |
45d1cc86 | 2363 | for (entry = *oentry; entry; oentry = &HeNEXT(entry), entry = *oentry) { |
19692e8d NC |
2364 | if (HeHASH(entry) != hash) /* strings can't be equal */ |
2365 | continue; | |
2366 | if (HeKLEN(entry) != len) | |
2367 | continue; | |
2368 | if (HeKEY(entry) != str && memNE(HeKEY(entry),str,len)) /* is this it? */ | |
2369 | continue; | |
2370 | if (HeKFLAGS(entry) != flags_masked) | |
2371 | continue; | |
19692e8d NC |
2372 | break; |
2373 | } | |
2374 | } | |
2375 | ||
35ab5632 NC |
2376 | if (entry) { |
2377 | if (--entry->he_valu.hent_refcount == 0) { | |
19692e8d | 2378 | *oentry = HeNEXT(entry); |
45d1cc86 NC |
2379 | if (!*first) { |
2380 | /* There are now no entries in our slot. */ | |
19692e8d | 2381 | xhv->xhv_fill--; /* HvFILL(hv)-- */ |
45d1cc86 | 2382 | } |
cbae3960 | 2383 | Safefree(entry); |
4c7185a0 | 2384 | xhv->xhv_keys--; /* HvTOTALKEYS(hv)-- */ |
19692e8d | 2385 | } |
fde52b5c | 2386 | } |
19692e8d | 2387 | |
333f433b | 2388 | UNLOCK_STRTAB_MUTEX; |
35ab5632 | 2389 | if (!entry && ckWARN_d(WARN_INTERNAL)) |
19692e8d | 2390 | Perl_warner(aTHX_ packWARN(WARN_INTERNAL), |
472d47bc SB |
2391 | "Attempt to free non-existent shared string '%s'%s" |
2392 | pTHX__FORMAT, | |
19692e8d | 2393 | hek ? HEK_KEY(hek) : str, |
472d47bc | 2394 | ((k_flags & HVhek_UTF8) ? " (utf8)" : "") pTHX__VALUE); |
19692e8d NC |
2395 | if (k_flags & HVhek_FREEKEY) |
2396 | Safefree(str); | |
fde52b5c | 2397 | } |
2398 | ||
bbce6d69 | 2399 | /* get a (constant) string ptr from the global string table |
2400 | * string will get added if it is not already there. | |
fde52b5c | 2401 | * len and hash must both be valid for str. |
2402 | */ | |
bbce6d69 | 2403 | HEK * |
864dbfa3 | 2404 | Perl_share_hek(pTHX_ const char *str, I32 len, register U32 hash) |
fde52b5c | 2405 | { |
da58a35d | 2406 | bool is_utf8 = FALSE; |
19692e8d | 2407 | int flags = 0; |
aec46f14 | 2408 | const char * const save = str; |
da58a35d JH |
2409 | |
2410 | if (len < 0) { | |
77caf834 | 2411 | STRLEN tmplen = -len; |
da58a35d | 2412 | is_utf8 = TRUE; |
77caf834 JH |
2413 | /* See the note in hv_fetch(). --jhi */ |
2414 | str = (char*)bytes_from_utf8((U8*)str, &tmplen, &is_utf8); | |
2415 | len = tmplen; | |
19692e8d NC |
2416 | /* If we were able to downgrade here, then than means that we were passed |
2417 | in a key which only had chars 0-255, but was utf8 encoded. */ | |
2418 | if (is_utf8) | |
2419 | flags = HVhek_UTF8; | |
2420 | /* If we found we were able to downgrade the string to bytes, then | |
2421 | we should flag that it needs upgrading on keys or each. Also flag | |
2422 | that we need share_hek_flags to free the string. */ | |
2423 | if (str != save) | |
2424 | flags |= HVhek_WASUTF8 | HVhek_FREEKEY; | |
2425 | } | |
2426 | ||
6e838c70 | 2427 | return share_hek_flags (str, len, hash, flags); |
19692e8d NC |
2428 | } |
2429 | ||
6e838c70 | 2430 | STATIC HEK * |
19692e8d NC |
2431 | S_share_hek_flags(pTHX_ const char *str, I32 len, register U32 hash, int flags) |
2432 | { | |
97aff369 | 2433 | dVAR; |
19692e8d | 2434 | register HE *entry; |
35a4481c | 2435 | const int flags_masked = flags & HVhek_MASK; |
263cb4a6 | 2436 | const U32 hindex = hash & (I32) HvMAX(PL_strtab); |
bbce6d69 | 2437 | |
fde52b5c | 2438 | /* what follows is the moral equivalent of: |
1c846c1f | 2439 | |
6b88bc9c | 2440 | if (!(Svp = hv_fetch(PL_strtab, str, len, FALSE))) |
a0714e2c | 2441 | hv_store(PL_strtab, str, len, NULL, hash); |
fdcd69b6 NC |
2442 | |
2443 | Can't rehash the shared string table, so not sure if it's worth | |
2444 | counting the number of entries in the linked list | |
bbce6d69 | 2445 | */ |
1b6737cc | 2446 | register XPVHV * const xhv = (XPVHV*)SvANY(PL_strtab); |
fde52b5c | 2447 | /* assert(xhv_array != 0) */ |
5f08fbcd | 2448 | LOCK_STRTAB_MUTEX; |
263cb4a6 NC |
2449 | entry = (HvARRAY(PL_strtab))[hindex]; |
2450 | for (;entry; entry = HeNEXT(entry)) { | |
fde52b5c | 2451 | if (HeHASH(entry) != hash) /* strings can't be equal */ |
2452 | continue; | |
2453 | if (HeKLEN(entry) != len) | |
2454 | continue; | |
1c846c1f | 2455 | if (HeKEY(entry) != str && memNE(HeKEY(entry),str,len)) /* is this it? */ |
fde52b5c | 2456 | continue; |
19692e8d | 2457 | if (HeKFLAGS(entry) != flags_masked) |
c3654f1a | 2458 | continue; |
fde52b5c | 2459 | break; |
2460 | } | |
263cb4a6 NC |
2461 | |
2462 | if (!entry) { | |
45d1cc86 NC |
2463 | /* What used to be head of the list. |
2464 | If this is NULL, then we're the first entry for this slot, which | |
2465 | means we need to increate fill. */ | |
cbae3960 NC |
2466 | struct shared_he *new_entry; |
2467 | HEK *hek; | |
2468 | char *k; | |
263cb4a6 NC |
2469 | HE **const head = &HvARRAY(PL_strtab)[hindex]; |
2470 | HE *const next = *head; | |
cbae3960 NC |
2471 | |
2472 | /* We don't actually store a HE from the arena and a regular HEK. | |
2473 | Instead we allocate one chunk of memory big enough for both, | |
2474 | and put the HEK straight after the HE. This way we can find the | |
2475 | HEK directly from the HE. | |
2476 | */ | |
2477 | ||
a02a5408 | 2478 | Newx(k, STRUCT_OFFSET(struct shared_he, |
cbae3960 NC |
2479 | shared_he_hek.hek_key[0]) + len + 2, char); |
2480 | new_entry = (struct shared_he *)k; | |
2481 | entry = &(new_entry->shared_he_he); | |
2482 | hek = &(new_entry->shared_he_hek); | |
2483 | ||
2484 | Copy(str, HEK_KEY(hek), len, char); | |
2485 | HEK_KEY(hek)[len] = 0; | |
2486 | HEK_LEN(hek) = len; | |
2487 | HEK_HASH(hek) = hash; | |
2488 | HEK_FLAGS(hek) = (unsigned char)flags_masked; | |
2489 | ||
2490 | /* Still "point" to the HEK, so that other code need not know what | |
2491 | we're up to. */ | |
2492 | HeKEY_hek(entry) = hek; | |
de616631 | 2493 | entry->he_valu.hent_refcount = 0; |
263cb4a6 NC |
2494 | HeNEXT(entry) = next; |
2495 | *head = entry; | |
cbae3960 | 2496 | |
4c7185a0 | 2497 | xhv->xhv_keys++; /* HvTOTALKEYS(hv)++ */ |
263cb4a6 | 2498 | if (!next) { /* initial entry? */ |
cbec9347 | 2499 | xhv->xhv_fill++; /* HvFILL(hv)++ */ |
4c9cc595 | 2500 | } else if (xhv->xhv_keys > (IV)xhv->xhv_max /* HvKEYS(hv) > HvMAX(hv) */) { |
cbec9347 | 2501 | hsplit(PL_strtab); |
bbce6d69 | 2502 | } |
2503 | } | |
2504 | ||
de616631 | 2505 | ++entry->he_valu.hent_refcount; |
5f08fbcd | 2506 | UNLOCK_STRTAB_MUTEX; |
19692e8d NC |
2507 | |
2508 | if (flags & HVhek_FREEKEY) | |
f9a63242 | 2509 | Safefree(str); |
19692e8d | 2510 | |
6e838c70 | 2511 | return HeKEY_hek(entry); |
fde52b5c | 2512 | } |
ecae49c0 | 2513 | |
ca732855 NC |
2514 | I32 * |
2515 | Perl_hv_placeholders_p(pTHX_ HV *hv) | |
2516 | { | |
2517 | dVAR; | |
2518 | MAGIC *mg = mg_find((SV*)hv, PERL_MAGIC_rhash); | |
2519 | ||
2520 | if (!mg) { | |
2521 | mg = sv_magicext((SV*)hv, 0, PERL_MAGIC_rhash, 0, 0, 0); | |
2522 | ||
2523 | if (!mg) { | |
2524 | Perl_die(aTHX_ "panic: hv_placeholders_p"); | |
2525 | } | |
2526 | } | |
2527 | return &(mg->mg_len); | |
2528 | } | |
2529 | ||
2530 | ||
2531 | I32 | |
2532 | Perl_hv_placeholders_get(pTHX_ HV *hv) | |
2533 | { | |
2534 | dVAR; | |
b464bac0 | 2535 | MAGIC * const mg = mg_find((SV*)hv, PERL_MAGIC_rhash); |
ca732855 NC |
2536 | |
2537 | return mg ? mg->mg_len : 0; | |
2538 | } | |
2539 | ||
2540 | void | |
ac1e784a | 2541 | Perl_hv_placeholders_set(pTHX_ HV *hv, I32 ph) |
ca732855 NC |
2542 | { |
2543 | dVAR; | |
b464bac0 | 2544 | MAGIC * const mg = mg_find((SV*)hv, PERL_MAGIC_rhash); |
ca732855 NC |
2545 | |
2546 | if (mg) { | |
2547 | mg->mg_len = ph; | |
2548 | } else if (ph) { | |
2549 | if (!sv_magicext((SV*)hv, 0, PERL_MAGIC_rhash, 0, 0, ph)) | |
2550 | Perl_die(aTHX_ "panic: hv_placeholders_set"); | |
2551 | } | |
2552 | /* else we don't need to add magic to record 0 placeholders. */ | |
2553 | } | |
ecae49c0 | 2554 | |
7b0bddfa NC |
2555 | SV * |
2556 | S_refcounted_he_value(pTHX_ const struct refcounted_he *he) | |
2557 | { | |
2558 | SV *value; | |
2559 | switch(he->refcounted_he_data[0] & HVrhek_typemask) { | |
2560 | case HVrhek_undef: | |
2561 | value = newSV(0); | |
2562 | break; | |
2563 | case HVrhek_delete: | |
2564 | value = &PL_sv_placeholder; | |
2565 | break; | |
2566 | case HVrhek_IV: | |
2567 | value = (he->refcounted_he_data[0] & HVrhek_UV) | |
2568 | ? newSVuv(he->refcounted_he_val.refcounted_he_u_iv) | |
2569 | : newSViv(he->refcounted_he_val.refcounted_he_u_uv); | |
2570 | break; | |
2571 | case HVrhek_PV: | |
2572 | /* Create a string SV that directly points to the bytes in our | |
2573 | structure. */ | |
2574 | value = newSV(0); | |
2575 | sv_upgrade(value, SVt_PV); | |
2576 | SvPV_set(value, (char *) he->refcounted_he_data + 1); | |
2577 | SvCUR_set(value, he->refcounted_he_val.refcounted_he_u_len); | |
2578 | /* This stops anything trying to free it */ | |
2579 | SvLEN_set(value, 0); | |
2580 | SvPOK_on(value); | |
2581 | SvREADONLY_on(value); | |
2582 | if (he->refcounted_he_data[0] & HVrhek_UTF8) | |
2583 | SvUTF8_on(value); | |
2584 | break; | |
2585 | default: | |
2586 | Perl_croak(aTHX_ "panic: refcounted_he_value bad flags %x", | |
2587 | he->refcounted_he_data[0]); | |
2588 | } | |
2589 | return value; | |
2590 | } | |
2591 | ||
2592 | #ifdef USE_ITHREADS | |
2593 | /* A big expression to find the key offset */ | |
2594 | #define REF_HE_KEY(chain) \ | |
2595 | ((((chain->refcounted_he_data[0] & HVrhek_typemask) == HVrhek_PV) \ | |
2596 | ? chain->refcounted_he_val.refcounted_he_u_len + 1 : 0) \ | |
2597 | + 1 + chain->refcounted_he_data) | |
2598 | #endif | |
2599 | ||
ecae49c0 | 2600 | /* |
b3ca2e83 NC |
2601 | =for apidoc refcounted_he_chain_2hv |
2602 | ||
2603 | Generates an returns a C<HV *> by walking up the tree starting at the passed | |
2604 | in C<struct refcounted_he *>. | |
2605 | ||
2606 | =cut | |
2607 | */ | |
2608 | HV * | |
2609 | Perl_refcounted_he_chain_2hv(pTHX_ const struct refcounted_he *chain) | |
2610 | { | |
7a89be66 | 2611 | dVAR; |
b3ca2e83 NC |
2612 | HV *hv = newHV(); |
2613 | U32 placeholders = 0; | |
2614 | /* We could chase the chain once to get an idea of the number of keys, | |
2615 | and call ksplit. But for now we'll make a potentially inefficient | |
2616 | hash with only 8 entries in its array. */ | |
2617 | const U32 max = HvMAX(hv); | |
2618 | ||
2619 | if (!HvARRAY(hv)) { | |
2620 | char *array; | |
2621 | Newxz(array, PERL_HV_ARRAY_ALLOC_BYTES(max + 1), char); | |
2622 | HvARRAY(hv) = (HE**)array; | |
2623 | } | |
2624 | ||
2625 | while (chain) { | |
cbb1fbea | 2626 | #ifdef USE_ITHREADS |
b6bbf3fa | 2627 | U32 hash = chain->refcounted_he_hash; |
cbb1fbea NC |
2628 | #else |
2629 | U32 hash = HEK_HASH(chain->refcounted_he_hek); | |
2630 | #endif | |
b3ca2e83 NC |
2631 | HE **oentry = &((HvARRAY(hv))[hash & max]); |
2632 | HE *entry = *oentry; | |
b6bbf3fa | 2633 | SV *value; |
cbb1fbea | 2634 | |
b3ca2e83 NC |
2635 | for (; entry; entry = HeNEXT(entry)) { |
2636 | if (HeHASH(entry) == hash) { | |
9f769845 NC |
2637 | /* We might have a duplicate key here. If so, entry is older |
2638 | than the key we've already put in the hash, so if they are | |
2639 | the same, skip adding entry. */ | |
2640 | #ifdef USE_ITHREADS | |
2641 | const STRLEN klen = HeKLEN(entry); | |
2642 | const char *const key = HeKEY(entry); | |
2643 | if (klen == chain->refcounted_he_keylen | |
2644 | && (!!HeKUTF8(entry) | |
2645 | == !!(chain->refcounted_he_data[0] & HVhek_UTF8)) | |
2646 | && memEQ(key, REF_HE_KEY(chain), klen)) | |
2647 | goto next_please; | |
2648 | #else | |
2649 | if (HeKEY_hek(entry) == chain->refcounted_he_hek) | |
2650 | goto next_please; | |
2651 | if (HeKLEN(entry) == HEK_LEN(chain->refcounted_he_hek) | |
2652 | && HeKUTF8(entry) == HEK_UTF8(chain->refcounted_he_hek) | |
2653 | && memEQ(HeKEY(entry), HEK_KEY(chain->refcounted_he_hek), | |
2654 | HeKLEN(entry))) | |
2655 | goto next_please; | |
2656 | #endif | |
b3ca2e83 NC |
2657 | } |
2658 | } | |
2659 | assert (!entry); | |
2660 | entry = new_HE(); | |
2661 | ||
cbb1fbea NC |
2662 | #ifdef USE_ITHREADS |
2663 | HeKEY_hek(entry) | |
7b0bddfa | 2664 | = share_hek_flags(REF_HE_KEY(chain), |
b6bbf3fa NC |
2665 | chain->refcounted_he_keylen, |
2666 | chain->refcounted_he_hash, | |
2667 | (chain->refcounted_he_data[0] | |
2668 | & (HVhek_UTF8|HVhek_WASUTF8))); | |
cbb1fbea | 2669 | #else |
71ad1b0c | 2670 | HeKEY_hek(entry) = share_hek_hek(chain->refcounted_he_hek); |
cbb1fbea | 2671 | #endif |
7b0bddfa NC |
2672 | value = refcounted_he_value(chain); |
2673 | if (value == &PL_sv_placeholder) | |
b3ca2e83 | 2674 | placeholders++; |
b6bbf3fa | 2675 | HeVAL(entry) = value; |
b3ca2e83 NC |
2676 | |
2677 | /* Link it into the chain. */ | |
2678 | HeNEXT(entry) = *oentry; | |
2679 | if (!HeNEXT(entry)) { | |
2680 | /* initial entry. */ | |
2681 | HvFILL(hv)++; | |
2682 | } | |
2683 | *oentry = entry; | |
2684 | ||
2685 | HvTOTALKEYS(hv)++; | |
2686 | ||
2687 | next_please: | |
71ad1b0c | 2688 | chain = chain->refcounted_he_next; |
b3ca2e83 NC |
2689 | } |
2690 | ||
2691 | if (placeholders) { | |
2692 | clear_placeholders(hv, placeholders); | |
2693 | HvTOTALKEYS(hv) -= placeholders; | |
2694 | } | |
2695 | ||
2696 | /* We could check in the loop to see if we encounter any keys with key | |
2697 | flags, but it's probably not worth it, as this per-hash flag is only | |
2698 | really meant as an optimisation for things like Storable. */ | |
2699 | HvHASKFLAGS_on(hv); | |
def9038f | 2700 | DEBUG_A(Perl_hv_assert(aTHX_ hv)); |
b3ca2e83 NC |
2701 | |
2702 | return hv; | |
2703 | } | |
2704 | ||
7b0bddfa NC |
2705 | SV * |
2706 | Perl_refcounted_he_fetch(pTHX_ const struct refcounted_he *chain, SV *keysv, | |
2707 | const char *key, STRLEN klen, int flags, U32 hash) | |
2708 | { | |
2709 | /* Just to be awkward, if you're using this interface the UTF-8-or-not-ness | |
2710 | of your key has to exactly match that which is stored. */ | |
2711 | SV *value = &PL_sv_placeholder; | |
d8c5b3c5 | 2712 | bool is_utf8; |
7b0bddfa NC |
2713 | |
2714 | if (keysv) { | |
2715 | if (flags & HVhek_FREEKEY) | |
2716 | Safefree(key); | |
2717 | key = SvPV_const(keysv, klen); | |
2718 | flags = 0; | |
d8c5b3c5 NC |
2719 | is_utf8 = (SvUTF8(keysv) != 0); |
2720 | } else { | |
2721 | is_utf8 = ((flags & HVhek_UTF8) ? TRUE : FALSE); | |
7b0bddfa NC |
2722 | } |
2723 | ||
2724 | if (!hash) { | |
2725 | if (keysv && (SvIsCOW_shared_hash(keysv))) { | |
2726 | hash = SvSHARED_HASH(keysv); | |
2727 | } else { | |
2728 | PERL_HASH(hash, key, klen); | |
2729 | } | |
2730 | } | |
2731 | ||
2732 | for (; chain; chain = chain->refcounted_he_next) { | |
2733 | #ifdef USE_ITHREADS | |
2734 | if (hash != chain->refcounted_he_hash) | |
2735 | continue; | |
2736 | if (klen != chain->refcounted_he_keylen) | |
2737 | continue; | |
2738 | if (memNE(REF_HE_KEY(chain),key,klen)) | |
2739 | continue; | |
d8c5b3c5 NC |
2740 | if (!!is_utf8 != !!(chain->refcounted_he_data[0] & HVhek_UTF8)) |
2741 | continue; | |
7b0bddfa NC |
2742 | #else |
2743 | if (hash != HEK_HASH(chain->refcounted_he_hek)) | |
2744 | continue; | |
2745 | if (klen != HEK_LEN(chain->refcounted_he_hek)) | |
2746 | continue; | |
2747 | if (memNE(HEK_KEY(chain->refcounted_he_hek),key,klen)) | |
2748 | continue; | |
d8c5b3c5 NC |
2749 | if (!!is_utf8 != !!HEK_UTF8(chain->refcounted_he_hek)) |
2750 | continue; | |
7b0bddfa NC |
2751 | #endif |
2752 | ||
2753 | value = sv_2mortal(refcounted_he_value(chain)); | |
2754 | break; | |
2755 | } | |
2756 | ||
2757 | if (flags & HVhek_FREEKEY) | |
2758 | Safefree(key); | |
2759 | ||
2760 | return value; | |
2761 | } | |
2762 | ||
b3ca2e83 NC |
2763 | /* |
2764 | =for apidoc refcounted_he_new | |
2765 | ||
ec2a1de7 NC |
2766 | Creates a new C<struct refcounted_he>. As S<key> is copied, and value is |
2767 | stored in a compact form, all references remain the property of the caller. | |
2768 | The C<struct refcounted_he> is returned with a reference count of 1. | |
b3ca2e83 NC |
2769 | |
2770 | =cut | |
2771 | */ | |
2772 | ||
2773 | struct refcounted_he * | |
2774 | Perl_refcounted_he_new(pTHX_ struct refcounted_he *const parent, | |
2775 | SV *const key, SV *const value) { | |
7a89be66 | 2776 | dVAR; |
b3ca2e83 | 2777 | struct refcounted_he *he; |
b6bbf3fa NC |
2778 | STRLEN key_len; |
2779 | const char *key_p = SvPV_const(key, key_len); | |
2780 | STRLEN value_len = 0; | |
95b63a38 | 2781 | const char *value_p = NULL; |
b6bbf3fa NC |
2782 | char value_type; |
2783 | char flags; | |
2784 | STRLEN key_offset; | |
b3ca2e83 | 2785 | U32 hash; |
d8c5b3c5 | 2786 | bool is_utf8 = SvUTF8(key) ? TRUE : FALSE; |
b6bbf3fa NC |
2787 | |
2788 | if (SvPOK(value)) { | |
2789 | value_type = HVrhek_PV; | |
2790 | } else if (SvIOK(value)) { | |
2791 | value_type = HVrhek_IV; | |
2792 | } else if (value == &PL_sv_placeholder) { | |
2793 | value_type = HVrhek_delete; | |
2794 | } else if (!SvOK(value)) { | |
2795 | value_type = HVrhek_undef; | |
2796 | } else { | |
2797 | value_type = HVrhek_PV; | |
2798 | } | |
b3ca2e83 | 2799 | |
b6bbf3fa NC |
2800 | if (value_type == HVrhek_PV) { |
2801 | value_p = SvPV_const(value, value_len); | |
2802 | key_offset = value_len + 2; | |
2803 | } else { | |
2804 | value_len = 0; | |
2805 | key_offset = 1; | |
2806 | } | |
2807 | flags = value_type; | |
2808 | ||
b6bbf3fa | 2809 | #ifdef USE_ITHREADS |
6cef672b | 2810 | he = PerlMemShared_malloc(sizeof(struct refcounted_he) - 1 |
b6bbf3fa | 2811 | + key_len |
b6bbf3fa | 2812 | + key_offset); |
6cef672b NC |
2813 | #else |
2814 | he = PerlMemShared_malloc(sizeof(struct refcounted_he) - 1 | |
2815 | + key_offset); | |
2816 | #endif | |
b3ca2e83 | 2817 | |
b3ca2e83 | 2818 | |
71ad1b0c | 2819 | he->refcounted_he_next = parent; |
b6bbf3fa NC |
2820 | |
2821 | if (value_type == HVrhek_PV) { | |
2822 | Copy(value_p, he->refcounted_he_data + 1, value_len + 1, char); | |
2823 | he->refcounted_he_val.refcounted_he_u_len = value_len; | |
2824 | if (SvUTF8(value)) { | |
2825 | flags |= HVrhek_UTF8; | |
2826 | } | |
2827 | } else if (value_type == HVrhek_IV) { | |
2828 | if (SvUOK(value)) { | |
2829 | he->refcounted_he_val.refcounted_he_u_uv = SvUVX(value); | |
2830 | flags |= HVrhek_UV; | |
2831 | } else { | |
2832 | he->refcounted_he_val.refcounted_he_u_iv = SvIVX(value); | |
2833 | } | |
2834 | } | |
2835 | ||
2836 | if (is_utf8) { | |
2837 | /* Hash keys are always stored normalised to (yes) ISO-8859-1. | |
2838 | As we're going to be building hash keys from this value in future, | |
2839 | normalise it now. */ | |
2840 | key_p = (char*)bytes_from_utf8((const U8*)key_p, &key_len, &is_utf8); | |
2841 | flags |= is_utf8 ? HVhek_UTF8 : HVhek_WASUTF8; | |
2842 | } | |
2843 | PERL_HASH(hash, key_p, key_len); | |
2844 | ||
cbb1fbea | 2845 | #ifdef USE_ITHREADS |
b6bbf3fa NC |
2846 | he->refcounted_he_hash = hash; |
2847 | he->refcounted_he_keylen = key_len; | |
2848 | Copy(key_p, he->refcounted_he_data + key_offset, key_len, char); | |
cbb1fbea | 2849 | #else |
b6bbf3fa | 2850 | he->refcounted_he_hek = share_hek_flags(key_p, key_len, hash, flags); |
cbb1fbea | 2851 | #endif |
b6bbf3fa NC |
2852 | |
2853 | if (flags & HVhek_WASUTF8) { | |
2854 | /* If it was downgraded from UTF-8, then the pointer returned from | |
2855 | bytes_from_utf8 is an allocated pointer that we must free. */ | |
2856 | Safefree(key_p); | |
2857 | } | |
2858 | ||
2859 | he->refcounted_he_data[0] = flags; | |
b3ca2e83 NC |
2860 | he->refcounted_he_refcnt = 1; |
2861 | ||
2862 | return he; | |
2863 | } | |
2864 | ||
2865 | /* | |
2866 | =for apidoc refcounted_he_free | |
2867 | ||
2868 | Decrements the reference count of the passed in C<struct refcounted_he *> | |
2869 | by one. If the reference count reaches zero the structure's memory is freed, | |
2870 | and C<refcounted_he_free> iterates onto the parent node. | |
2871 | ||
2872 | =cut | |
2873 | */ | |
2874 | ||
2875 | void | |
2876 | Perl_refcounted_he_free(pTHX_ struct refcounted_he *he) { | |
57ca3b03 AL |
2877 | PERL_UNUSED_CONTEXT; |
2878 | ||
b3ca2e83 NC |
2879 | while (he) { |
2880 | struct refcounted_he *copy; | |
cbb1fbea | 2881 | U32 new_count; |
b3ca2e83 | 2882 | |
cbb1fbea NC |
2883 | HINTS_REFCNT_LOCK; |
2884 | new_count = --he->refcounted_he_refcnt; | |
2885 | HINTS_REFCNT_UNLOCK; | |
2886 | ||
2887 | if (new_count) { | |
b3ca2e83 | 2888 | return; |
cbb1fbea | 2889 | } |
b3ca2e83 | 2890 | |
b6bbf3fa | 2891 | #ifndef USE_ITHREADS |
71ad1b0c | 2892 | unshare_hek_or_pvn (he->refcounted_he_hek, 0, 0, 0); |
cbb1fbea | 2893 | #endif |
b3ca2e83 | 2894 | copy = he; |
71ad1b0c | 2895 | he = he->refcounted_he_next; |
b6bbf3fa | 2896 | PerlMemShared_free(copy); |
b3ca2e83 NC |
2897 | } |
2898 | } | |
2899 | ||
b3ca2e83 | 2900 | /* |
ecae49c0 NC |
2901 | =for apidoc hv_assert |
2902 | ||
2903 | Check that a hash is in an internally consistent state. | |
2904 | ||
2905 | =cut | |
2906 | */ | |
2907 | ||
943795c2 NC |
2908 | #ifdef DEBUGGING |
2909 | ||
ecae49c0 NC |
2910 | void |
2911 | Perl_hv_assert(pTHX_ HV *hv) | |
2912 | { | |
57ca3b03 AL |
2913 | dVAR; |
2914 | HE* entry; | |
2915 | int withflags = 0; | |
2916 | int placeholders = 0; | |
2917 | int real = 0; | |
2918 | int bad = 0; | |
2919 | const I32 riter = HvRITER_get(hv); | |
2920 | HE *eiter = HvEITER_get(hv); | |
2921 | ||
2922 | (void)hv_iterinit(hv); | |
2923 | ||
2924 | while ((entry = hv_iternext_flags(hv, HV_ITERNEXT_WANTPLACEHOLDERS))) { | |
2925 | /* sanity check the values */ | |
2926 | if (HeVAL(entry) == &PL_sv_placeholder) | |
2927 | placeholders++; | |
2928 | else | |
2929 | real++; | |
2930 | /* sanity check the keys */ | |
2931 | if (HeSVKEY(entry)) { | |
6f207bd3 | 2932 | NOOP; /* Don't know what to check on SV keys. */ |
57ca3b03 AL |
2933 | } else if (HeKUTF8(entry)) { |
2934 | withflags++; | |
2935 | if (HeKWASUTF8(entry)) { | |
2936 | PerlIO_printf(Perl_debug_log, | |
2937 | "hash key has both WASUFT8 and UTF8: '%.*s'\n", | |
2938 | (int) HeKLEN(entry), HeKEY(entry)); | |
2939 | bad = 1; | |
2940 | } | |
2941 | } else if (HeKWASUTF8(entry)) | |
2942 | withflags++; | |
2943 | } | |
2944 | if (!SvTIED_mg((SV*)hv, PERL_MAGIC_tied)) { | |
2945 | static const char bad_count[] = "Count %d %s(s), but hash reports %d\n"; | |
2946 | const int nhashkeys = HvUSEDKEYS(hv); | |
2947 | const int nhashplaceholders = HvPLACEHOLDERS_get(hv); | |
2948 | ||
2949 | if (nhashkeys != real) { | |
2950 | PerlIO_printf(Perl_debug_log, bad_count, real, "keys", nhashkeys ); | |
2951 | bad = 1; | |
2952 | } | |
2953 | if (nhashplaceholders != placeholders) { | |
2954 | PerlIO_printf(Perl_debug_log, bad_count, placeholders, "placeholder", nhashplaceholders ); | |
2955 | bad = 1; | |
2956 | } | |
2957 | } | |
2958 | if (withflags && ! HvHASKFLAGS(hv)) { | |
2959 | PerlIO_printf(Perl_debug_log, | |
2960 | "Hash has HASKFLAGS off but I count %d key(s) with flags\n", | |
2961 | withflags); | |
2962 | bad = 1; | |
2963 | } | |
2964 | if (bad) { | |
2965 | sv_dump((SV *)hv); | |
2966 | } | |
2967 | HvRITER_set(hv, riter); /* Restore hash iterator state */ | |
2968 | HvEITER_set(hv, eiter); | |
ecae49c0 | 2969 | } |
af3babe4 | 2970 | |
943795c2 NC |
2971 | #endif |
2972 | ||
af3babe4 NC |
2973 | /* |
2974 | * Local variables: | |
2975 | * c-indentation-style: bsd | |
2976 | * c-basic-offset: 4 | |
2977 | * indent-tabs-mode: t | |
2978 | * End: | |
2979 | * | |
37442d52 RGS |
2980 | * ex: set ts=8 sts=4 sw=4 noet: |
2981 | */ |