Commit | Line | Data |
---|---|---|
a0d0e21e | 1 | /* hv.h |
79072805 | 2 | * |
4bb101f2 | 3 | * Copyright (C) 1991, 1992, 1993, 1996, 1997, 1998, 1999, |
62a1a1ef | 4 | * 2000, 2001, 2002, 2003, 2005, 2006, 2007, 2008, by Larry Wall and others |
79072805 LW |
5 | * |
6 | * You may distribute under the terms of either the GNU General Public | |
7 | * License or the Artistic License, as specified in the README file. | |
8 | * | |
79072805 LW |
9 | */ |
10 | ||
5cbe4eec | 11 | /* entry in hash value chain */ |
79072805 | 12 | struct he { |
6f51351d NC |
13 | /* Keep hent_next first in this structure, because sv_free_arenas take |
14 | advantage of this to share code between the he arenas and the SV | |
15 | body arenas */ | |
5cbe4eec MLF |
16 | HE *hent_next; /* next entry in chain */ |
17 | HEK *hent_hek; /* hash key */ | |
de616631 NC |
18 | union { |
19 | SV *hent_val; /* scalar value that was hashed */ | |
20 | Size_t hent_refcount; /* references for this shared hash key */ | |
21 | } he_valu; | |
bbce6d69 | 22 | }; |
23 | ||
5cbe4eec | 24 | /* hash key -- defined separately for use as shared pointer */ |
ff68c719 | 25 | struct hek { |
5cbe4eec MLF |
26 | U32 hek_hash; /* hash of key */ |
27 | I32 hek_len; /* length of hash key */ | |
28 | char hek_key[1]; /* variable-length hash key */ | |
e05949c7 | 29 | /* the hash-key is \0-terminated */ |
1e54db1a JH |
30 | /* after the \0 there is a byte for flags, such as whether the key |
31 | is UTF-8 */ | |
79072805 LW |
32 | }; |
33 | ||
cbae3960 NC |
34 | struct shared_he { |
35 | struct he shared_he_he; | |
36 | struct hek shared_he_hek; | |
37 | }; | |
bfcb3514 NC |
38 | |
39 | /* Subject to change. | |
40 | Don't access this directly. | |
e1a479c5 | 41 | Use the funcs in mro.c |
bfcb3514 | 42 | */ |
e1a479c5 | 43 | |
b2685f0c NC |
44 | struct mro_alg { |
45 | AV *(*resolve)(pTHX_ HV* stash, U32 level); | |
46 | const char *name; | |
47 | U16 length; | |
48 | U16 kflags; /* For the hash API - set HVhek_UTF8 if name is UTF-8 */ | |
49 | U32 hash; /* or 0 */ | |
50 | }; | |
e1a479c5 BB |
51 | |
52 | struct mro_meta { | |
9953ff72 NC |
53 | /* a hash holding the different MROs private data. */ |
54 | HV *mro_linear_all; | |
3a6fa573 NC |
55 | /* a pointer directly to the current MROs private data. If mro_linear_all |
56 | is NULL, this owns the SV reference, else it is just a pointer to a | |
57 | value stored in and owned by mro_linear_all. */ | |
58 | SV *mro_linear_current; | |
dd69841b BB |
59 | HV *mro_nextmethod; /* next::method caching */ |
60 | U32 cache_gen; /* Bumping this invalidates our method cache */ | |
70cd14a1 | 61 | U32 pkg_gen; /* Bumps when local methods/@ISA change */ |
3d76853f | 62 | const struct mro_alg *mro_which; /* which mro alg is in use? */ |
a49ba3fc | 63 | HV *isa; /* Everything this class @ISA */ |
e1a479c5 BB |
64 | }; |
65 | ||
123892d9 NC |
66 | #define MRO_GET_PRIVATE_DATA(smeta, which) \ |
67 | (((smeta)->mro_which && (which) == (smeta)->mro_which) \ | |
3a6fa573 | 68 | ? (smeta)->mro_linear_current \ |
123892d9 NC |
69 | : Perl_mro_get_private_data(aTHX_ (smeta), (which))) |
70 | ||
e1a479c5 BB |
71 | /* Subject to change. |
72 | Don't access this directly. | |
73 | */ | |
74 | ||
15d9236d NC |
75 | union _xhvnameu { |
76 | HEK *xhvnameu_name; /* When xhv_name_count is 0 */ | |
77 | HEK **xhvnameu_names; /* When xhv_name_count is non-0 */ | |
78 | }; | |
79 | ||
bfcb3514 | 80 | struct xpvhv_aux { |
15d9236d | 81 | union _xhvnameu xhv_name_u; /* name, if a symbol table */ |
86f55936 | 82 | AV *xhv_backreferences; /* back references for weak references */ |
bfcb3514 NC |
83 | HE *xhv_eiter; /* current entry of iterator */ |
84 | I32 xhv_riter; /* current root of iterator */ | |
7dc86639 | 85 | |
15d9236d | 86 | /* Concerning xhv_name_count: When non-zero, xhv_name_u contains a pointer |
78b79c77 FC |
87 | * to an array of HEK pointers, this being the length. The first element is |
88 | * the name of the stash, which may be NULL. If xhv_name_count is positive, | |
89 | * then *xhv_name is one of the effective names. If xhv_name_count is nega- | |
15d9236d | 90 | * tive, then xhv_name_u.xhvnameu_names[1] is the first effective name. |
78b79c77 FC |
91 | */ |
92 | I32 xhv_name_count; | |
bc85b3a1 | 93 | struct mro_meta *xhv_mro_meta; |
aae43805 | 94 | HV * xhv_super; /* SUPER method cache */ |
78b79c77 | 95 | }; |
bfcb3514 | 96 | |
5cbe4eec | 97 | /* hash structure: */ |
6ee623d5 | 98 | /* This structure must match the beginning of struct xpvmg in sv.h. */ |
79072805 | 99 | struct xpvhv { |
6e128786 NC |
100 | HV* xmg_stash; /* class package */ |
101 | union _xmgu xmg_u; | |
359164a0 | 102 | STRLEN xhv_keys; /* total keys, including placeholders */ |
c8e503bf | 103 | STRLEN xhv_max; /* subscript of last element of xhv_array */ |
79072805 LW |
104 | }; |
105 | ||
5cbe4eec | 106 | /* hash a key */ |
5afd6d42 JH |
107 | /* The use of a temporary pointer and the casting games |
108 | * is needed to serve the dual purposes of | |
109 | * (a) the hashed data being interpreted as "unsigned char" (new since 5.8, | |
df805c3e | 110 | * a "char" can be either signed or unsigned, depending on the compiler) |
5afd6d42 | 111 | * (b) catering for old code that uses a "char" |
830b38bd | 112 | * |
504f80c1 JH |
113 | * The "hash seed" feature was added in Perl 5.8.1 to perturb the results |
114 | * to avoid "algorithmic complexity attacks". | |
830b38bd JH |
115 | * |
116 | * If USE_HASH_SEED is defined, hash randomisation is done by default | |
117 | * If USE_HASH_SEED_EXPLICIT is defined, hash randomisation is done | |
118 | * only if the environment variable PERL_HASH_SEED is set. | |
7dc86639 | 119 | * (see also perl.c:perl_parse() and S_init_tls_and_interp() and util.c:get_hash_seed()) |
5afd6d42 | 120 | */ |
830b38bd JH |
121 | #ifndef PERL_HASH_SEED |
122 | # if defined(USE_HASH_SEED) || defined(USE_HASH_SEED_EXPLICIT) | |
7dc86639 | 123 | # define PERL_HASH_SEED PL_hash_seed |
830b38bd | 124 | # else |
7dc86639 | 125 | # define PERL_HASH_SEED "PeRlHaShhAcKpErl" |
830b38bd | 126 | # endif |
504f80c1 | 127 | #endif |
f8d50d94 | 128 | |
7dc86639 YO |
129 | #define PERL_HASH_SEED_U32 *((U32*)PERL_HASH_SEED) |
130 | #define PERL_HASH_SEED_U64_1 (((U64*)PERL_HASH_SEED)[0]) | |
131 | #define PERL_HASH_SEED_U64_2 (((U64*)PERL_HASH_SEED)[1]) | |
4886dc4f | 132 | #define PERL_HASH_SEED_U16_x(idx) (((U16*)PERL_HASH_SEED)[idx]) |
bf6bd887 | 133 | |
7dc86639 | 134 | /* legacy - only mod_perl should be doing this. */ |
3d78eb94 | 135 | #ifdef PERL_HASH_INTERNAL_ACCESS |
7dc86639 YO |
136 | #define PERL_HASH_INTERNAL(hash,str,len) PERL_HASH(hash,str,len) |
137 | #endif | |
138 | ||
139 | /* Uncomment one of the following lines to use an alternative hash algorithm. | |
140 | #define PERL_HASH_FUNC_SDBM | |
141 | #define PERL_HASH_FUNC_DJB2 | |
142 | #define PERL_HASH_FUNC_SUPERFAST | |
143 | #define PERL_HASH_FUNC_MURMUR3 | |
144 | #define PERL_HASH_FUNC_SIPHASH | |
145 | #define PERL_HASH_FUNC_ONE_AT_A_TIME | |
8c9e24c7 | 146 | #define PERL_HASH_FUNC_ONE_AT_A_TIME_OLD |
4886dc4f | 147 | #define PERL_HASH_FUNC_BUZZHASH16 |
7dc86639 YO |
148 | */ |
149 | ||
8c9e24c7 YO |
150 | #if !( 0 \ |
151 | || defined(PERL_HASH_FUNC_SDBM) \ | |
152 | || defined(PERL_HASH_FUNC_DJB2) \ | |
153 | || defined(PERL_HASH_FUNC_SUPERFAST) \ | |
154 | || defined(PERL_HASH_FUNC_MURMUR3) \ | |
155 | || defined(PERL_HASH_FUNC_ONE_AT_A_TIME) \ | |
156 | || defined(PERL_HASH_FUNC_ONE_AT_A_TIME_OLD) \ | |
157 | || defined(PERL_HASH_FUNC_BUZZHASH16) \ | |
158 | ) | |
3db6cbfc YO |
159 | #ifdef U64 |
160 | #define PERL_HASH_FUNC_SIPHASH | |
161 | #else | |
162 | #define PERL_HASH_FUNC_ONE_AT_A_TIME | |
163 | #endif | |
7dc86639 YO |
164 | #endif |
165 | ||
4886dc4f YO |
166 | #if defined(PERL_HASH_FUNC_BUZZHASH16) |
167 | /* "BUZZHASH16" | |
168 | * | |
169 | * I whacked this together while just playing around. | |
170 | * | |
171 | * The idea is that instead of hashing the actual string input we use the | |
172 | * bytes of the string as an index into a table of randomly generated | |
173 | * 16 bit values. | |
174 | * | |
175 | * A left rotate is used to "mix" in previous bits as we go, and I borrowed | |
176 | * the avalanche function from one-at-a-time for the final step. A lookup | |
177 | * into the table based on the lower 8 bits of the length combined with | |
178 | * the length itself is used as an itializer. | |
179 | * | |
180 | * The resulting hash value has no actual bits fed in from the string so | |
181 | * I would guess it is pretty secure, although I am not a cryptographer | |
182 | * and have no idea for sure. Nor has it been rigorously tested. On the | |
183 | * other hand it is reasonably fast, and seems to produce reasonable | |
184 | * distributions. | |
185 | * | |
186 | * Yves Orton | |
187 | */ | |
188 | ||
189 | ||
190 | #define PERL_HASH_FUNC "BUZZHASH16" | |
191 | #define PERL_HASH_SEED_BYTES 512 /* 2 bytes per octet value, 2 * 256 */ | |
192 | /* Find best way to ROTL32 */ | |
193 | #if defined(_MSC_VER) | |
194 | #include <stdlib.h> /* Microsoft put _rotl declaration in here */ | |
195 | #define BUZZHASH_ROTL32(x,r) _rotl(x,r) | |
196 | #else | |
197 | /* gcc recognises this code and generates a rotate instruction for CPUs with one */ | |
198 | #define BUZZHASH_ROTL32(x,r) (((U32)x << r) | ((U32)x >> (32 - r))) | |
199 | #endif | |
200 | ||
201 | #define PERL_HASH(hash,str,len) \ | |
202 | STMT_START { \ | |
5aaab254 KW |
203 | const char * const s_PeRlHaSh_tmp = (str); \ |
204 | const unsigned char *s_PeRlHaSh = (const unsigned char *)s_PeRlHaSh_tmp; \ | |
205 | const unsigned char *end_PeRlHaSh = (const unsigned char *)s_PeRlHaSh + len; \ | |
206 | U32 hash_PeRlHaSh = (PERL_HASH_SEED_U16_x(len & 0xff) << 16) + len; \ | |
4886dc4f YO |
207 | while (s_PeRlHaSh < end_PeRlHaSh) { \ |
208 | hash_PeRlHaSh ^= PERL_HASH_SEED_U16_x((U8)*s_PeRlHaSh++); \ | |
209 | hash_PeRlHaSh += BUZZHASH_ROTL32(hash_PeRlHaSh,11); \ | |
210 | } \ | |
211 | hash_PeRlHaSh += (hash_PeRlHaSh << 3); \ | |
212 | hash_PeRlHaSh ^= (hash_PeRlHaSh >> 11); \ | |
213 | (hash) = (hash_PeRlHaSh + (hash_PeRlHaSh << 15)); \ | |
214 | } STMT_END | |
215 | ||
216 | #elif defined(PERL_HASH_FUNC_SIPHASH) | |
7dc86639 YO |
217 | #define PERL_HASH_FUNC "SIPHASH" |
218 | #define PERL_HASH_SEED_BYTES 16 | |
219 | ||
220 | /* This is SipHash by Jean-Philippe Aumasson and Daniel J. Bernstein. | |
221 | * The authors claim it is relatively secure compared to the alternatives | |
222 | * and that performance wise it is a suitable hash for languages like Perl. | |
223 | * See: | |
224 | * | |
225 | * https://www.131002.net/siphash/ | |
226 | * | |
227 | * This implementation seems to perform slightly slower than one-at-a-time for | |
228 | * short keys, but degrades slower for longer keys. Murmur Hash outperforms it | |
229 | * regardless of keys size. | |
230 | * | |
231 | * It is 64 bit only. | |
232 | */ | |
233 | ||
234 | #define PERL_HASH_NEEDS_TWO_SEEDS | |
235 | ||
236 | #ifndef U64 | |
237 | #define U64 uint64_t | |
238 | #endif | |
239 | ||
240 | #define ROTL(x,b) (U64)( ((x) << (b)) | ( (x) >> (64 - (b))) ) | |
241 | ||
242 | #define U32TO8_LE(p, v) \ | |
243 | (p)[0] = (U8)((v) ); (p)[1] = (U8)((v) >> 8); \ | |
244 | (p)[2] = (U8)((v) >> 16); (p)[3] = (U8)((v) >> 24); | |
245 | ||
246 | #define U64TO8_LE(p, v) \ | |
247 | U32TO8_LE((p), (U32)((v) )); \ | |
248 | U32TO8_LE((p) + 4, (U32)((v) >> 32)); | |
249 | ||
250 | #define U8TO64_LE(p) \ | |
251 | (((U64)((p)[0]) ) | \ | |
252 | ((U64)((p)[1]) << 8) | \ | |
253 | ((U64)((p)[2]) << 16) | \ | |
254 | ((U64)((p)[3]) << 24) | \ | |
255 | ((U64)((p)[4]) << 32) | \ | |
256 | ((U64)((p)[5]) << 40) | \ | |
257 | ((U64)((p)[6]) << 48) | \ | |
258 | ((U64)((p)[7]) << 56)) | |
259 | ||
260 | #define SIPROUND \ | |
261 | do { \ | |
262 | v0_PeRlHaSh += v1_PeRlHaSh; v1_PeRlHaSh=ROTL(v1_PeRlHaSh,13); v1_PeRlHaSh ^= v0_PeRlHaSh; v0_PeRlHaSh=ROTL(v0_PeRlHaSh,32); \ | |
263 | v2_PeRlHaSh += v3_PeRlHaSh; v3_PeRlHaSh=ROTL(v3_PeRlHaSh,16); v3_PeRlHaSh ^= v2_PeRlHaSh; \ | |
264 | v0_PeRlHaSh += v3_PeRlHaSh; v3_PeRlHaSh=ROTL(v3_PeRlHaSh,21); v3_PeRlHaSh ^= v0_PeRlHaSh; \ | |
265 | v2_PeRlHaSh += v1_PeRlHaSh; v1_PeRlHaSh=ROTL(v1_PeRlHaSh,17); v1_PeRlHaSh ^= v2_PeRlHaSh; v2_PeRlHaSh=ROTL(v2_PeRlHaSh,32); \ | |
266 | } while(0) | |
267 | ||
268 | /* SipHash-2-4 */ | |
269 | #define PERL_HASH(hash,str,len) STMT_START { \ | |
270 | const char * const strtmp_PeRlHaSh = (str); \ | |
271 | const unsigned char *in_PeRlHaSh = (const unsigned char *)strtmp_PeRlHaSh; \ | |
272 | const U32 inlen_PeRlHaSh = (len); \ | |
273 | /* "somepseudorandomlygeneratedbytes" */ \ | |
274 | U64 v0_PeRlHaSh = 0x736f6d6570736575ULL; \ | |
275 | U64 v1_PeRlHaSh = 0x646f72616e646f6dULL; \ | |
276 | U64 v2_PeRlHaSh = 0x6c7967656e657261ULL; \ | |
277 | U64 v3_PeRlHaSh = 0x7465646279746573ULL; \ | |
278 | \ | |
279 | U64 b_PeRlHaSh; \ | |
280 | U64 k0_PeRlHaSh = PERL_HASH_SEED_U64_1; \ | |
281 | U64 k1_PeRlHaSh = PERL_HASH_SEED_U64_2; \ | |
282 | U64 m_PeRlHaSh; \ | |
283 | const int left_PeRlHaSh = inlen_PeRlHaSh & 7; \ | |
284 | const U8 *end_PeRlHaSh = in_PeRlHaSh + inlen_PeRlHaSh - left_PeRlHaSh; \ | |
285 | \ | |
286 | b_PeRlHaSh = ( ( U64 )(len) ) << 56; \ | |
287 | v3_PeRlHaSh ^= k1_PeRlHaSh; \ | |
288 | v2_PeRlHaSh ^= k0_PeRlHaSh; \ | |
289 | v1_PeRlHaSh ^= k1_PeRlHaSh; \ | |
290 | v0_PeRlHaSh ^= k0_PeRlHaSh; \ | |
291 | \ | |
292 | for ( ; in_PeRlHaSh != end_PeRlHaSh; in_PeRlHaSh += 8 ) \ | |
293 | { \ | |
294 | m_PeRlHaSh = U8TO64_LE( in_PeRlHaSh ); \ | |
295 | v3_PeRlHaSh ^= m_PeRlHaSh; \ | |
296 | SIPROUND; \ | |
297 | SIPROUND; \ | |
298 | v0_PeRlHaSh ^= m_PeRlHaSh; \ | |
299 | } \ | |
300 | \ | |
301 | switch( left_PeRlHaSh ) \ | |
302 | { \ | |
303 | case 7: b_PeRlHaSh |= ( ( U64 )in_PeRlHaSh[ 6] ) << 48; \ | |
304 | case 6: b_PeRlHaSh |= ( ( U64 )in_PeRlHaSh[ 5] ) << 40; \ | |
305 | case 5: b_PeRlHaSh |= ( ( U64 )in_PeRlHaSh[ 4] ) << 32; \ | |
306 | case 4: b_PeRlHaSh |= ( ( U64 )in_PeRlHaSh[ 3] ) << 24; \ | |
307 | case 3: b_PeRlHaSh |= ( ( U64 )in_PeRlHaSh[ 2] ) << 16; \ | |
308 | case 2: b_PeRlHaSh |= ( ( U64 )in_PeRlHaSh[ 1] ) << 8; \ | |
309 | case 1: b_PeRlHaSh |= ( ( U64 )in_PeRlHaSh[ 0] ); break; \ | |
310 | case 0: break; \ | |
311 | } \ | |
312 | \ | |
313 | v3_PeRlHaSh ^= b_PeRlHaSh; \ | |
314 | SIPROUND; \ | |
315 | SIPROUND; \ | |
316 | v0_PeRlHaSh ^= b_PeRlHaSh; \ | |
317 | \ | |
318 | v2_PeRlHaSh ^= 0xff; \ | |
319 | SIPROUND; \ | |
320 | SIPROUND; \ | |
321 | SIPROUND; \ | |
322 | SIPROUND; \ | |
323 | b_PeRlHaSh = v0_PeRlHaSh ^ v1_PeRlHaSh ^ v2_PeRlHaSh ^ v3_PeRlHaSh; \ | |
324 | (hash)= (U32)(b_PeRlHaSh & U32_MAX); \ | |
325 | } STMT_END | |
326 | ||
327 | #elif defined(PERL_HASH_FUNC_SUPERFAST) | |
328 | #define PERL_HASH_FUNC "SUPERFAST" | |
c3a88658 | 329 | #define PERL_HASH_SEED_BYTES 4 |
7dc86639 YO |
330 | /* FYI: This is the "Super-Fast" algorithm mentioned by Bob Jenkins in |
331 | * (http://burtleburtle.net/bob/hash/doobs.html) | |
332 | * It is by Paul Hsieh (c) 2004 and is analysed here | |
333 | * http://www.azillionmonkeys.com/qed/hash.html | |
334 | * license terms are here: | |
335 | * http://www.azillionmonkeys.com/qed/weblicense.html | |
336 | */ | |
337 | #undef get16bits | |
338 | #if (defined(__GNUC__) && defined(__i386__)) || defined(__WATCOMC__) \ | |
339 | || defined(_MSC_VER) || defined (__BORLANDC__) || defined (__TURBOC__) | |
340 | #define get16bits(d) (*((const U16 *) (d))) | |
f8d50d94 DM |
341 | #endif |
342 | ||
7dc86639 YO |
343 | #if !defined (get16bits) |
344 | #define get16bits(d) ((((const U8 *)(d))[1] << UINT32_C(8))\ | |
345 | +((const U8 *)(d))[0]) | |
346 | #endif | |
347 | #define PERL_HASH(hash,str,len) \ | |
348 | STMT_START { \ | |
5aaab254 KW |
349 | const char * const strtmp_PeRlHaSh = (str); \ |
350 | const unsigned char *str_PeRlHaSh = (const unsigned char *)strtmp_PeRlHaSh; \ | |
351 | U32 len_PeRlHaSh = (len); \ | |
352 | U32 hash_PeRlHaSh = PERL_HASH_SEED_U32 ^ len; \ | |
353 | U32 tmp_PeRlHaSh; \ | |
354 | int rem_PeRlHaSh= len_PeRlHaSh & 3; \ | |
7dc86639 YO |
355 | len_PeRlHaSh >>= 2; \ |
356 | \ | |
357 | for (;len_PeRlHaSh > 0; len_PeRlHaSh--) { \ | |
358 | hash_PeRlHaSh += get16bits (str_PeRlHaSh); \ | |
359 | tmp_PeRlHaSh = (get16bits (str_PeRlHaSh+2) << 11) ^ hash_PeRlHaSh; \ | |
360 | hash_PeRlHaSh = (hash_PeRlHaSh << 16) ^ tmp_PeRlHaSh; \ | |
361 | str_PeRlHaSh += 2 * sizeof (U16); \ | |
362 | hash_PeRlHaSh += hash_PeRlHaSh >> 11; \ | |
363 | } \ | |
364 | \ | |
365 | /* Handle end cases */ \ | |
366 | switch (rem_PeRlHaSh) { \ | |
367 | case 3: hash_PeRlHaSh += get16bits (str_PeRlHaSh); \ | |
368 | hash_PeRlHaSh ^= hash_PeRlHaSh << 16; \ | |
369 | hash_PeRlHaSh ^= str_PeRlHaSh[sizeof (U16)] << 18; \ | |
370 | hash_PeRlHaSh += hash_PeRlHaSh >> 11; \ | |
371 | break; \ | |
372 | case 2: hash_PeRlHaSh += get16bits (str_PeRlHaSh); \ | |
373 | hash_PeRlHaSh ^= hash_PeRlHaSh << 11; \ | |
374 | hash_PeRlHaSh += hash_PeRlHaSh >> 17; \ | |
375 | break; \ | |
376 | case 1: hash_PeRlHaSh += *str_PeRlHaSh; \ | |
377 | hash_PeRlHaSh ^= hash_PeRlHaSh << 10; \ | |
378 | hash_PeRlHaSh += hash_PeRlHaSh >> 1; \ | |
379 | } \ | |
380 | \ | |
381 | /* Force "avalanching" of final 127 bits */ \ | |
382 | hash_PeRlHaSh ^= hash_PeRlHaSh << 3; \ | |
383 | hash_PeRlHaSh += hash_PeRlHaSh >> 5; \ | |
384 | hash_PeRlHaSh ^= hash_PeRlHaSh << 4; \ | |
385 | hash_PeRlHaSh += hash_PeRlHaSh >> 17; \ | |
386 | hash_PeRlHaSh ^= hash_PeRlHaSh << 25; \ | |
387 | (hash) = (hash_PeRlHaSh + (hash_PeRlHaSh >> 6)); \ | |
388 | } STMT_END | |
389 | ||
390 | #elif defined(PERL_HASH_FUNC_MURMUR3) | |
391 | #define PERL_HASH_FUNC "MURMUR3" | |
392 | #define PERL_HASH_SEED_BYTES 4 | |
393 | ||
394 | /*----------------------------------------------------------------------------- | |
395 | * MurmurHash3 was written by Austin Appleby, and is placed in the public | |
396 | * domain. | |
397 | * | |
398 | * This implementation was originally written by Shane Day, and is also public domain, | |
399 | * and was modified to function as a macro similar to other perl hash functions by | |
400 | * Yves Orton. | |
401 | * | |
402 | * This is a portable ANSI C implementation of MurmurHash3_x86_32 (Murmur3A) | |
403 | * with support for progressive processing. | |
404 | * | |
405 | * If you want to understand the MurmurHash algorithm you would be much better | |
406 | * off reading the original source. Just point your browser at: | |
407 | * http://code.google.com/p/smhasher/source/browse/trunk/MurmurHash3.cpp | |
408 | * | |
409 | * How does it work? | |
410 | * | |
411 | * We can only process entire 32 bit chunks of input, except for the very end | |
412 | * that may be shorter. | |
413 | * | |
414 | * To handle endianess I simply use a macro that reads a U32 and define | |
415 | * that macro to be a direct read on little endian machines, a read and swap | |
416 | * on big endian machines, or a byte-by-byte read if the endianess is unknown. | |
417 | */ | |
418 | ||
419 | ||
420 | /*----------------------------------------------------------------------------- | |
421 | * Endianess, misalignment capabilities and util macros | |
422 | * | |
423 | * The following 3 macros are defined in this section. The other macros defined | |
424 | * are only needed to help derive these 3. | |
425 | * | |
426 | * MURMUR_READ_UINT32(x) Read a little endian unsigned 32-bit int | |
427 | * MURMUR_UNALIGNED_SAFE Defined if READ_UINT32 works on non-word boundaries | |
428 | * MURMUR_ROTL32(x,r) Rotate x left by r bits | |
429 | */ | |
f8d50d94 | 430 | |
7dc86639 | 431 | /* Now find best way we can to READ_UINT32 */ |
b5a2311a | 432 | #if (BYTEORDER == 0x1234 || BYTEORDER == 0x12345678) && U32SIZE == 4 |
7dc86639 YO |
433 | /* CPU endian matches murmurhash algorithm, so read 32-bit word directly */ |
434 | #define MURMUR_READ_UINT32(ptr) (*((U32*)(ptr))) | |
b5a2311a | 435 | #elif BYTEORDER == 0x4321 || BYTEORDER == 0x87654321 |
7dc86639 YO |
436 | /* TODO: Add additional cases below where a compiler provided bswap32 is available */ |
437 | #if defined(__GNUC__) && (__GNUC__>4 || (__GNUC__==4 && __GNUC_MINOR__>=3)) | |
438 | #define MURMUR_READ_UINT32(ptr) (__builtin_bswap32(*((U32*)(ptr)))) | |
439 | #else | |
440 | /* Without a known fast bswap32 we're just as well off doing this */ | |
441 | #define MURMUR_READ_UINT32(ptr) (ptr[0]|ptr[1]<<8|ptr[2]<<16|ptr[3]<<24) | |
442 | #define MURMUR_UNALIGNED_SAFE | |
443 | #endif | |
444 | #else | |
445 | /* Unknown endianess so last resort is to read individual bytes */ | |
446 | #define MURMUR_READ_UINT32(ptr) (ptr[0]|ptr[1]<<8|ptr[2]<<16|ptr[3]<<24) | |
447 | ||
448 | /* Since we're not doing word-reads we can skip the messing about with realignment */ | |
449 | #define MURMUR_UNALIGNED_SAFE | |
450 | #endif | |
451 | ||
452 | /* Find best way to ROTL32 */ | |
453 | #if defined(_MSC_VER) | |
454 | #include <stdlib.h> /* Microsoft put _rotl declaration in here */ | |
455 | #define MURMUR_ROTL32(x,r) _rotl(x,r) | |
456 | #else | |
457 | /* gcc recognises this code and generates a rotate instruction for CPUs with one */ | |
458 | #define MURMUR_ROTL32(x,r) (((U32)x << r) | ((U32)x >> (32 - r))) | |
459 | #endif | |
460 | ||
461 | ||
462 | /*----------------------------------------------------------------------------- | |
463 | * Core murmurhash algorithm macros */ | |
464 | ||
465 | #define MURMUR_C1 (0xcc9e2d51) | |
466 | #define MURMUR_C2 (0x1b873593) | |
467 | #define MURMUR_C3 (0xe6546b64) | |
468 | #define MURMUR_C4 (0x85ebca6b) | |
469 | #define MURMUR_C5 (0xc2b2ae35) | |
470 | ||
471 | /* This is the main processing body of the algorithm. It operates | |
472 | * on each full 32-bits of input. */ | |
473 | #define MURMUR_DOBLOCK(h1, k1) STMT_START { \ | |
474 | k1 *= MURMUR_C1; \ | |
475 | k1 = MURMUR_ROTL32(k1,15); \ | |
476 | k1 *= MURMUR_C2; \ | |
477 | \ | |
478 | h1 ^= k1; \ | |
479 | h1 = MURMUR_ROTL32(h1,13); \ | |
480 | h1 = h1 * 5 + MURMUR_C3; \ | |
481 | } STMT_END | |
482 | ||
483 | ||
484 | /* Append unaligned bytes to carry, forcing hash churn if we have 4 bytes */ | |
485 | /* cnt=bytes to process, h1=name of h1 var, c=carry, n=bytes in c, ptr/len=payload */ | |
486 | #define MURMUR_DOBYTES(cnt, h1, c, n, ptr, len) STMT_START { \ | |
487 | int MURMUR_DOBYTES_i = cnt; \ | |
488 | while(MURMUR_DOBYTES_i--) { \ | |
489 | c = c>>8 | *ptr++<<24; \ | |
490 | n++; len--; \ | |
491 | if(n==4) { \ | |
492 | MURMUR_DOBLOCK(h1, c); \ | |
493 | n = 0; \ | |
494 | } \ | |
495 | } \ | |
496 | } STMT_END | |
497 | ||
498 | /* process the last 1..3 bytes and finalize */ | |
499 | #define MURMUR_FINALIZE(hash, PeRlHaSh_len, PeRlHaSh_k1, PeRlHaSh_h1, PeRlHaSh_carry, PeRlHaSh_bytes_in_carry, PeRlHaSh_ptr, PeRlHaSh_total_length) STMT_START { \ | |
500 | /* Advance over whole 32-bit chunks, possibly leaving 1..3 bytes */\ | |
501 | PeRlHaSh_len -= PeRlHaSh_len/4*4; \ | |
502 | \ | |
503 | /* Append any remaining bytes into carry */ \ | |
504 | MURMUR_DOBYTES(PeRlHaSh_len, PeRlHaSh_h1, PeRlHaSh_carry, PeRlHaSh_bytes_in_carry, PeRlHaSh_ptr, PeRlHaSh_len); \ | |
505 | \ | |
506 | if (PeRlHaSh_bytes_in_carry) { \ | |
507 | PeRlHaSh_k1 = PeRlHaSh_carry >> ( 4 - PeRlHaSh_bytes_in_carry ) * 8; \ | |
508 | PeRlHaSh_k1 *= MURMUR_C1; \ | |
509 | PeRlHaSh_k1 = MURMUR_ROTL32(PeRlHaSh_k1,15); \ | |
510 | PeRlHaSh_k1 *= MURMUR_C2; \ | |
511 | PeRlHaSh_h1 ^= PeRlHaSh_k1; \ | |
512 | } \ | |
513 | PeRlHaSh_h1 ^= PeRlHaSh_total_length; \ | |
514 | \ | |
515 | /* fmix */ \ | |
516 | PeRlHaSh_h1 ^= PeRlHaSh_h1 >> 16; \ | |
517 | PeRlHaSh_h1 *= MURMUR_C4; \ | |
518 | PeRlHaSh_h1 ^= PeRlHaSh_h1 >> 13; \ | |
519 | PeRlHaSh_h1 *= MURMUR_C5; \ | |
520 | PeRlHaSh_h1 ^= PeRlHaSh_h1 >> 16; \ | |
521 | (hash)= PeRlHaSh_h1; \ | |
522 | } STMT_END | |
523 | ||
524 | /* now we create the hash function */ | |
525 | ||
526 | #if defined(UNALIGNED_SAFE) | |
527 | #define PERL_HASH(hash,str,len) STMT_START { \ | |
5aaab254 KW |
528 | const char * const s_PeRlHaSh_tmp = (str); \ |
529 | const unsigned char *PeRlHaSh_ptr = (const unsigned char *)s_PeRlHaSh_tmp; \ | |
530 | I32 PeRlHaSh_len = len; \ | |
7dc86639 YO |
531 | \ |
532 | U32 PeRlHaSh_h1 = PERL_HASH_SEED_U32; \ | |
533 | U32 PeRlHaSh_k1; \ | |
534 | U32 PeRlHaSh_carry = 0; \ | |
535 | \ | |
536 | const unsigned char *PeRlHaSh_end; \ | |
537 | \ | |
538 | int PeRlHaSh_bytes_in_carry = 0; /* bytes in carry */ \ | |
539 | I32 PeRlHaSh_total_length= PeRlHaSh_len; \ | |
540 | \ | |
541 | /* This CPU handles unaligned word access */ \ | |
542 | /* Process 32-bit chunks */ \ | |
543 | PeRlHaSh_end = PeRlHaSh_ptr + PeRlHaSh_len/4*4; \ | |
544 | for( ; PeRlHaSh_ptr < PeRlHaSh_end ; PeRlHaSh_ptr+=4) { \ | |
545 | PeRlHaSh_k1 = MURMUR_READ_UINT32(PeRlHaSh_ptr); \ | |
546 | MURMUR_DOBLOCK(PeRlHaSh_h1, PeRlHaSh_k1); \ | |
547 | } \ | |
548 | \ | |
549 | MURMUR_FINALIZE(hash, PeRlHaSh_len, PeRlHaSh_k1, PeRlHaSh_h1, PeRlHaSh_carry, PeRlHaSh_bytes_in_carry, PeRlHaSh_ptr, PeRlHaSh_total_length);\ | |
550 | } STMT_END | |
551 | #else | |
552 | #define PERL_HASH(hash,str,len) STMT_START { \ | |
5aaab254 KW |
553 | const char * const s_PeRlHaSh_tmp = (str); \ |
554 | const unsigned char *PeRlHaSh_ptr = (const unsigned char *)s_PeRlHaSh_tmp; \ | |
555 | I32 PeRlHaSh_len = len; \ | |
7dc86639 YO |
556 | \ |
557 | U32 PeRlHaSh_h1 = PERL_HASH_SEED_U32; \ | |
558 | U32 PeRlHaSh_k1; \ | |
559 | U32 PeRlHaSh_carry = 0; \ | |
560 | \ | |
561 | const unsigned char *PeRlHaSh_end; \ | |
562 | \ | |
563 | int PeRlHaSh_bytes_in_carry = 0; /* bytes in carry */ \ | |
564 | I32 PeRlHaSh_total_length= PeRlHaSh_len; \ | |
565 | \ | |
566 | /* This CPU does not handle unaligned word access */ \ | |
567 | \ | |
568 | /* Consume enough so that the next data byte is word aligned */ \ | |
569 | int PeRlHaSh_i = -(long)PeRlHaSh_ptr & 3; \ | |
570 | if(PeRlHaSh_i && PeRlHaSh_i <= PeRlHaSh_len) { \ | |
571 | MURMUR_DOBYTES(PeRlHaSh_i, PeRlHaSh_h1, PeRlHaSh_carry, PeRlHaSh_bytes_in_carry, PeRlHaSh_ptr, PeRlHaSh_len);\ | |
572 | } \ | |
573 | \ | |
574 | /* We're now aligned. Process in aligned blocks. Specialise for each possible carry count */ \ | |
575 | PeRlHaSh_end = PeRlHaSh_ptr + PeRlHaSh_len/4*4; \ | |
576 | switch(PeRlHaSh_bytes_in_carry) { /* how many bytes in carry */ \ | |
577 | case 0: /* c=[----] w=[3210] b=[3210]=w c'=[----] */ \ | |
578 | for( ; PeRlHaSh_ptr < PeRlHaSh_end ; PeRlHaSh_ptr+=4) { \ | |
579 | PeRlHaSh_k1 = MURMUR_READ_UINT32(PeRlHaSh_ptr); \ | |
580 | MURMUR_DOBLOCK(PeRlHaSh_h1, PeRlHaSh_k1); \ | |
581 | } \ | |
582 | break; \ | |
583 | case 1: /* c=[0---] w=[4321] b=[3210]=c>>24|w<<8 c'=[4---] */ \ | |
584 | for( ; PeRlHaSh_ptr < PeRlHaSh_end ; PeRlHaSh_ptr+=4) { \ | |
585 | PeRlHaSh_k1 = PeRlHaSh_carry>>24; \ | |
586 | PeRlHaSh_carry = MURMUR_READ_UINT32(PeRlHaSh_ptr); \ | |
587 | PeRlHaSh_k1 |= PeRlHaSh_carry<<8; \ | |
588 | MURMUR_DOBLOCK(PeRlHaSh_h1, PeRlHaSh_k1); \ | |
589 | } \ | |
590 | break; \ | |
591 | case 2: /* c=[10--] w=[5432] b=[3210]=c>>16|w<<16 c'=[54--] */ \ | |
592 | for( ; PeRlHaSh_ptr < PeRlHaSh_end ; PeRlHaSh_ptr+=4) { \ | |
593 | PeRlHaSh_k1 = PeRlHaSh_carry>>16; \ | |
594 | PeRlHaSh_carry = MURMUR_READ_UINT32(PeRlHaSh_ptr); \ | |
595 | PeRlHaSh_k1 |= PeRlHaSh_carry<<16; \ | |
596 | MURMUR_DOBLOCK(PeRlHaSh_h1, PeRlHaSh_k1); \ | |
597 | } \ | |
598 | break; \ | |
599 | case 3: /* c=[210-] w=[6543] b=[3210]=c>>8|w<<24 c'=[654-] */ \ | |
600 | for( ; PeRlHaSh_ptr < PeRlHaSh_end ; PeRlHaSh_ptr+=4) { \ | |
601 | PeRlHaSh_k1 = PeRlHaSh_carry>>8; \ | |
602 | PeRlHaSh_carry = MURMUR_READ_UINT32(PeRlHaSh_ptr); \ | |
603 | PeRlHaSh_k1 |= PeRlHaSh_carry<<24; \ | |
604 | MURMUR_DOBLOCK(PeRlHaSh_h1, PeRlHaSh_k1); \ | |
605 | } \ | |
606 | } \ | |
607 | \ | |
608 | MURMUR_FINALIZE(hash, PeRlHaSh_len, PeRlHaSh_k1, PeRlHaSh_h1, PeRlHaSh_carry, PeRlHaSh_bytes_in_carry, PeRlHaSh_ptr, PeRlHaSh_total_length);\ | |
609 | } STMT_END | |
610 | #endif | |
611 | ||
612 | #elif defined(PERL_HASH_FUNC_DJB2) | |
613 | #define PERL_HASH_FUNC "DJB2" | |
614 | #define PERL_HASH_SEED_BYTES 4 | |
615 | #define PERL_HASH(hash,str,len) \ | |
616 | STMT_START { \ | |
5aaab254 KW |
617 | const char * const s_PeRlHaSh_tmp = (str); \ |
618 | const unsigned char *s_PeRlHaSh = (const unsigned char *)s_PeRlHaSh_tmp; \ | |
619 | I32 i_PeRlHaSh = len; \ | |
620 | U32 hash_PeRlHaSh = PERL_HASH_SEED_U32 ^ len; \ | |
7dc86639 YO |
621 | while (i_PeRlHaSh--) { \ |
622 | hash_PeRlHaSh = ((hash_PeRlHaSh << 5) + hash_PeRlHaSh) + *s_PeRlHaSh++; \ | |
623 | } \ | |
624 | (hash) = hash_PeRlHaSh;\ | |
625 | } STMT_END | |
626 | ||
627 | #elif defined(PERL_HASH_FUNC_SDBM) | |
628 | #define PERL_HASH_FUNC "SDBM" | |
629 | #define PERL_HASH_SEED_BYTES 4 | |
630 | #define PERL_HASH(hash,str,len) \ | |
631 | STMT_START { \ | |
5aaab254 KW |
632 | const char * const s_PeRlHaSh_tmp = (str); \ |
633 | const unsigned char *s_PeRlHaSh = (const unsigned char *)s_PeRlHaSh_tmp; \ | |
634 | I32 i_PeRlHaSh = len; \ | |
635 | U32 hash_PeRlHaSh = PERL_HASH_SEED_U32 ^ len; \ | |
7dc86639 YO |
636 | while (i_PeRlHaSh--) { \ |
637 | hash_PeRlHaSh = (hash_PeRlHaSh << 6) + (hash_PeRlHaSh << 16) - hash_PeRlHaSh + *s_PeRlHaSh++; \ | |
638 | } \ | |
639 | (hash) = hash_PeRlHaSh;\ | |
640 | } STMT_END | |
641 | ||
8c9e24c7 YO |
642 | #elif defined(PERL_HASH_FUNC_ONE_AT_A_TIME) || defined(PERL_HASH_FUNC_ONE_AT_A_TIME_OLD) |
643 | ||
7dc86639 YO |
644 | #define PERL_HASH_SEED_BYTES 4 |
645 | ||
8c9e24c7 YO |
646 | #ifdef PERL_HASH_FUNC_ONE_AT_A_TIME |
647 | /* new version, add the length to the seed so that adding characters changes the "seed" being used. */ | |
648 | #define PERL_HASH_FUNC "ONE_AT_A_TIME" | |
649 | #define MIX_SEED_AND_LEN(seed,len) (seed + len) | |
650 | #else | |
651 | /* old version, just use the seed. - not recommended */ | |
652 | #define PERL_HASH_FUNC "ONE_AT_A_TIME_OLD" | |
653 | #define MIX_SEED_AND_LEN(seed,len) (seed) | |
654 | #endif | |
655 | ||
7dc86639 YO |
656 | /* FYI: This is the "One-at-a-Time" algorithm by Bob Jenkins |
657 | * from requirements by Colin Plumb. | |
658 | * (http://burtleburtle.net/bob/hash/doobs.html) */ | |
659 | #define PERL_HASH(hash,str,len) \ | |
4b5190b5 | 660 | STMT_START { \ |
5aaab254 KW |
661 | const char * const s_PeRlHaSh_tmp = (str); \ |
662 | const unsigned char *s_PeRlHaSh = (const unsigned char *)s_PeRlHaSh_tmp; \ | |
8c9e24c7 YO |
663 | const unsigned char *end_PeRlHaSh = (const unsigned char *)s_PeRlHaSh_tmp + (len); \ |
664 | U32 hash_PeRlHaSh = MIX_SEED_AND_LEN(PERL_HASH_SEED_U32, len); \ | |
665 | while (s_PeRlHaSh < end_PeRlHaSh) { \ | |
7dc86639 | 666 | hash_PeRlHaSh += (U8)*s_PeRlHaSh++; \ |
4b5190b5 NC |
667 | hash_PeRlHaSh += (hash_PeRlHaSh << 10); \ |
668 | hash_PeRlHaSh ^= (hash_PeRlHaSh >> 6); \ | |
669 | } \ | |
670 | hash_PeRlHaSh += (hash_PeRlHaSh << 3); \ | |
671 | hash_PeRlHaSh ^= (hash_PeRlHaSh >> 11); \ | |
672 | (hash) = (hash_PeRlHaSh + (hash_PeRlHaSh << 15)); \ | |
673 | } STMT_END | |
7dc86639 YO |
674 | #endif |
675 | #ifndef PERL_HASH | |
676 | #error "No hash function defined!" | |
677 | #endif | |
954c1994 | 678 | /* |
ccfc67b7 JH |
679 | =head1 Hash Manipulation Functions |
680 | ||
954c1994 GS |
681 | =for apidoc AmU||HEf_SVKEY |
682 | This flag, used in the length slot of hash entries and magic structures, | |
d1be9408 | 683 | specifies the structure contains an C<SV*> pointer where a C<char*> pointer |
954c1994 GS |
684 | is to be expected. (For information only--not to be used). |
685 | ||
ccfc67b7 JH |
686 | =head1 Handy Values |
687 | ||
954c1994 GS |
688 | =for apidoc AmU||Nullhv |
689 | Null HV pointer. | |
690 | ||
3ae1b226 NC |
691 | (deprecated - use C<(HV *)NULL> instead) |
692 | ||
ccfc67b7 JH |
693 | =head1 Hash Manipulation Functions |
694 | ||
954c1994 | 695 | =for apidoc Am|char*|HvNAME|HV* stash |
9282b5fd SH |
696 | Returns the package name of a stash, or NULL if C<stash> isn't a stash. |
697 | See C<SvSTASH>, C<CvSTASH>. | |
954c1994 | 698 | |
d9021235 FC |
699 | =for apidoc Am|STRLEN|HvNAMELEN|HV *stash |
700 | Returns the length of the stash's name. | |
701 | ||
702 | =for apidoc Am|unsigned char|HvNAMEUTF8|HV *stash | |
703 | Returns true if the name is in UTF8 encoding. | |
704 | ||
bc56db2a FC |
705 | =for apidoc Am|char*|HvENAME|HV* stash |
706 | Returns the effective name of a stash, or NULL if there is none. The | |
707 | effective name represents a location in the symbol table where this stash | |
708 | resides. It is updated automatically when packages are aliased or deleted. | |
709 | A stash that is no longer in the symbol table has no effective name. This | |
710 | name is preferable to C<HvNAME> for use in MRO linearisations and isa | |
711 | caches. | |
712 | ||
d9021235 FC |
713 | =for apidoc Am|STRLEN|HvENAMELEN|HV *stash |
714 | Returns the length of the stash's effective name. | |
715 | ||
716 | =for apidoc Am|unsigned char|HvENAMEUTF8|HV *stash | |
717 | Returns true if the effective name is in UTF8 encoding. | |
718 | ||
954c1994 GS |
719 | =for apidoc Am|void*|HeKEY|HE* he |
720 | Returns the actual pointer stored in the key slot of the hash entry. The | |
721 | pointer may be either C<char*> or C<SV*>, depending on the value of | |
722 | C<HeKLEN()>. Can be assigned to. The C<HePV()> or C<HeSVKEY()> macros are | |
723 | usually preferable for finding the value of a key. | |
724 | ||
725 | =for apidoc Am|STRLEN|HeKLEN|HE* he | |
726 | If this is negative, and amounts to C<HEf_SVKEY>, it indicates the entry | |
727 | holds an C<SV*> key. Otherwise, holds the actual length of the key. Can | |
728 | be assigned to. The C<HePV()> macro is usually preferable for finding key | |
729 | lengths. | |
730 | ||
731 | =for apidoc Am|SV*|HeVAL|HE* he | |
566a4718 YO |
732 | Returns the value slot (type C<SV*>) stored in the hash entry. Can be assigned |
733 | to. | |
734 | ||
735 | SV *foo= HeVAL(hv); | |
736 | HeVAL(hv)= sv; | |
737 | ||
954c1994 GS |
738 | |
739 | =for apidoc Am|U32|HeHASH|HE* he | |
740 | Returns the computed hash stored in the hash entry. | |
741 | ||
742 | =for apidoc Am|char*|HePV|HE* he|STRLEN len | |
743 | Returns the key slot of the hash entry as a C<char*> value, doing any | |
744 | necessary dereferencing of possibly C<SV*> keys. The length of the string | |
745 | is placed in C<len> (this is a macro, so do I<not> use C<&len>). If you do | |
746 | not care about what the length of the key is, you may use the global | |
747 | variable C<PL_na>, though this is rather less efficient than using a local | |
748 | variable. Remember though, that hash keys in perl are free to contain | |
749 | embedded nulls, so using C<strlen()> or similar is not a good way to find | |
750 | the length of hash keys. This is very similar to the C<SvPV()> macro | |
289d3c6a NC |
751 | described elsewhere in this document. See also C<HeUTF8>. |
752 | ||
753 | If you are using C<HePV> to get values to pass to C<newSVpvn()> to create a | |
754 | new SV, you should consider using C<newSVhek(HeKEY_hek(he))> as it is more | |
755 | efficient. | |
756 | ||
cca4e9fa | 757 | =for apidoc Am|char*|HeUTF8|HE* he |
289d3c6a NC |
758 | Returns whether the C<char *> value returned by C<HePV> is encoded in UTF-8, |
759 | doing any necessary dereferencing of possibly C<SV*> keys. The value returned | |
62a1a1ef | 760 | will be 0 or non-0, not necessarily 1 (or even a value with any low bits set), |
289d3c6a NC |
761 | so B<do not> blindly assign this to a C<bool> variable, as C<bool> may be a |
762 | typedef for C<char>. | |
954c1994 GS |
763 | |
764 | =for apidoc Am|SV*|HeSVKEY|HE* he | |
a0714e2c | 765 | Returns the key as an C<SV*>, or C<NULL> if the hash entry does not |
954c1994 GS |
766 | contain an C<SV*> key. |
767 | ||
768 | =for apidoc Am|SV*|HeSVKEY_force|HE* he | |
769 | Returns the key as an C<SV*>. Will create and return a temporary mortal | |
770 | C<SV*> if the hash entry contains only a C<char*> key. | |
771 | ||
772 | =for apidoc Am|SV*|HeSVKEY_set|HE* he|SV* sv | |
773 | Sets the key to a given C<SV*>, taking care to set the appropriate flags to | |
774 | indicate the presence of an C<SV*> key, and returns the same | |
775 | C<SV*>. | |
776 | ||
777 | =cut | |
778 | */ | |
bf6bd887 | 779 | |
bf5b86ae | 780 | /* these hash entry flags ride on hent_klen (for use only in magic/tied HVs) */ |
d1be9408 | 781 | #define HEf_SVKEY -2 /* hent_key is an SV* */ |
bf6bd887 | 782 | |
3ae1b226 NC |
783 | #ifndef PERL_CORE |
784 | # define Nullhv Null(HV*) | |
785 | #endif | |
43e6e717 | 786 | #define HvARRAY(hv) ((hv)->sv_u.svu_hash) |
4d0fbddd | 787 | #define HvFILL(hv) Perl_hv_fill(aTHX_ (const HV *)(hv)) |
463ee0b2 | 788 | #define HvMAX(hv) ((XPVHV*) SvANY(hv))->xhv_max |
b79f7545 NC |
789 | /* This quite intentionally does no flag checking first. That's your |
790 | responsibility. */ | |
791 | #define HvAUX(hv) ((struct xpvhv_aux*)&(HvARRAY(hv)[HvMAX(hv)+1])) | |
dbebbdb4 NC |
792 | #define HvRITER(hv) (*Perl_hv_riter_p(aTHX_ MUTABLE_HV(hv))) |
793 | #define HvEITER(hv) (*Perl_hv_eiter_p(aTHX_ MUTABLE_HV(hv))) | |
794 | #define HvRITER_set(hv,r) Perl_hv_riter_set(aTHX_ MUTABLE_HV(hv), r) | |
795 | #define HvEITER_set(hv,e) Perl_hv_eiter_set(aTHX_ MUTABLE_HV(hv), e) | |
b79f7545 | 796 | #define HvRITER_get(hv) (SvOOK(hv) ? HvAUX(hv)->xhv_riter : -1) |
cef6ea9d | 797 | #define HvEITER_get(hv) (SvOOK(hv) ? HvAUX(hv)->xhv_eiter : NULL) |
7423f6db | 798 | #define HvNAME(hv) HvNAME_get(hv) |
f2462604 | 799 | #define HvNAMELEN(hv) HvNAMELEN_get(hv) |
bc56db2a | 800 | #define HvENAME(hv) HvENAME_get(hv) |
f2462604 | 801 | #define HvENAMELEN(hv) HvENAMELEN_get(hv) |
2c7f4b87 BB |
802 | |
803 | /* Checking that hv is a valid package stash is the | |
804 | caller's responsibility */ | |
805 | #define HvMROMETA(hv) (HvAUX(hv)->xhv_mro_meta \ | |
806 | ? HvAUX(hv)->xhv_mro_meta \ | |
c60bad7b | 807 | : Perl_mro_meta_init(aTHX_ hv)) |
2c7f4b87 | 808 | |
15d9236d NC |
809 | #define HvNAME_HEK_NN(hv) \ |
810 | ( \ | |
811 | HvAUX(hv)->xhv_name_count \ | |
812 | ? *HvAUX(hv)->xhv_name_u.xhvnameu_names \ | |
813 | : HvAUX(hv)->xhv_name_u.xhvnameu_name \ | |
b7247a80 | 814 | ) |
9b9d0b15 | 815 | /* This macro may go away without notice. */ |
b7247a80 | 816 | #define HvNAME_HEK(hv) \ |
15d9236d | 817 | (SvOOK(hv) && HvAUX(hv)->xhv_name_u.xhvnameu_name ? HvNAME_HEK_NN(hv) : NULL) |
78b79c77 | 818 | #define HvNAME_get(hv) \ |
15d9236d | 819 | ((SvOOK(hv) && HvAUX(hv)->xhv_name_u.xhvnameu_name && HvNAME_HEK_NN(hv)) \ |
b7247a80 | 820 | ? HEK_KEY(HvNAME_HEK_NN(hv)) : NULL) |
78b79c77 | 821 | #define HvNAMELEN_get(hv) \ |
15d9236d | 822 | ((SvOOK(hv) && HvAUX(hv)->xhv_name_u.xhvnameu_name && HvNAME_HEK_NN(hv)) \ |
b7247a80 | 823 | ? HEK_LEN(HvNAME_HEK_NN(hv)) : 0) |
b2c03ebd | 824 | #define HvNAMEUTF8(hv) \ |
f2462604 BF |
825 | ((SvOOK(hv) && HvAUX(hv)->xhv_name_u.xhvnameu_name && HvNAME_HEK_NN(hv)) \ |
826 | ? HEK_UTF8(HvNAME_HEK_NN(hv)) : 0) | |
c4842d1d | 827 | #define HvENAME_HEK_NN(hv) \ |
78b79c77 | 828 | ( \ |
15d9236d NC |
829 | HvAUX(hv)->xhv_name_count > 0 ? HvAUX(hv)->xhv_name_u.xhvnameu_names[0] : \ |
830 | HvAUX(hv)->xhv_name_count < -1 ? HvAUX(hv)->xhv_name_u.xhvnameu_names[1] : \ | |
78b79c77 | 831 | HvAUX(hv)->xhv_name_count == -1 ? NULL : \ |
15d9236d | 832 | HvAUX(hv)->xhv_name_u.xhvnameu_name \ |
78b79c77 | 833 | ) |
c4842d1d | 834 | #define HvENAME_HEK(hv) \ |
15d9236d | 835 | (SvOOK(hv) && HvAUX(hv)->xhv_name_u.xhvnameu_name ? HvENAME_HEK_NN(hv) : NULL) |
c4842d1d | 836 | #define HvENAME_get(hv) \ |
c49a809b | 837 | ((SvOOK(hv) && HvAUX(hv)->xhv_name_u.xhvnameu_name && HvAUX(hv)->xhv_name_count != -1) \ |
78b79c77 | 838 | ? HEK_KEY(HvENAME_HEK_NN(hv)) : NULL) |
c4842d1d | 839 | #define HvENAMELEN_get(hv) \ |
c49a809b | 840 | ((SvOOK(hv) && HvAUX(hv)->xhv_name_u.xhvnameu_name && HvAUX(hv)->xhv_name_count != -1) \ |
78b79c77 | 841 | ? HEK_LEN(HvENAME_HEK_NN(hv)) : 0) |
b2c03ebd | 842 | #define HvENAMEUTF8(hv) \ |
c49a809b | 843 | ((SvOOK(hv) && HvAUX(hv)->xhv_name_u.xhvnameu_name && HvAUX(hv)->xhv_name_count != -1) \ |
f2462604 | 844 | ? HEK_UTF8(HvENAME_HEK_NN(hv)) : 0) |
a0d0e21e | 845 | |
486ec47a | 846 | /* the number of keys (including any placeholders) */ |
8aacddc1 NIS |
847 | #define XHvTOTALKEYS(xhv) ((xhv)->xhv_keys) |
848 | ||
8aacddc1 NIS |
849 | /* |
850 | * HvKEYS gets the number of keys that actually exist(), and is provided | |
851 | * for backwards compatibility with old XS code. The core uses HvUSEDKEYS | |
486ec47a | 852 | * (keys, excluding placeholders) and HvTOTALKEYS (including placeholders) |
8aacddc1 | 853 | */ |
ca732855 NC |
854 | #define HvKEYS(hv) HvUSEDKEYS(hv) |
855 | #define HvUSEDKEYS(hv) (HvTOTALKEYS(hv) - HvPLACEHOLDERS_get(hv)) | |
8aacddc1 | 856 | #define HvTOTALKEYS(hv) XHvTOTALKEYS((XPVHV*) SvANY(hv)) |
dbebbdb4 NC |
857 | #define HvPLACEHOLDERS(hv) (*Perl_hv_placeholders_p(aTHX_ MUTABLE_HV(hv))) |
858 | #define HvPLACEHOLDERS_get(hv) (SvMAGIC(hv) ? Perl_hv_placeholders_get(aTHX_ (const HV *)hv) : 0) | |
859 | #define HvPLACEHOLDERS_set(hv,p) Perl_hv_placeholders_set(aTHX_ MUTABLE_HV(hv), p) | |
8aacddc1 | 860 | |
bf6bd887 | 861 | #define HvSHAREKEYS(hv) (SvFLAGS(hv) & SVphv_SHAREKEYS) |
862 | #define HvSHAREKEYS_on(hv) (SvFLAGS(hv) |= SVphv_SHAREKEYS) | |
863 | #define HvSHAREKEYS_off(hv) (SvFLAGS(hv) &= ~SVphv_SHAREKEYS) | |
864 | ||
19692e8d NC |
865 | /* This is an optimisation flag. It won't be set if all hash keys have a 0 |
866 | * flag. Currently the only flags relate to utf8. | |
867 | * Hence it won't be set if all keys are 8 bit only. It will be set if any key | |
868 | * is utf8 (including 8 bit keys that were entered as utf8, and need upgrading | |
869 | * when retrieved during iteration. It may still be set when there are no longer | |
870 | * any utf8 keys. | |
fdcd69b6 | 871 | * See HVhek_ENABLEHVKFLAGS for the trigger. |
19692e8d NC |
872 | */ |
873 | #define HvHASKFLAGS(hv) (SvFLAGS(hv) & SVphv_HASKFLAGS) | |
874 | #define HvHASKFLAGS_on(hv) (SvFLAGS(hv) |= SVphv_HASKFLAGS) | |
875 | #define HvHASKFLAGS_off(hv) (SvFLAGS(hv) &= ~SVphv_HASKFLAGS) | |
574c8022 | 876 | |
bf5b86ae GS |
877 | #define HvLAZYDEL(hv) (SvFLAGS(hv) & SVphv_LAZYDEL) |
878 | #define HvLAZYDEL_on(hv) (SvFLAGS(hv) |= SVphv_LAZYDEL) | |
879 | #define HvLAZYDEL_off(hv) (SvFLAGS(hv) &= ~SVphv_LAZYDEL) | |
880 | ||
3ae1b226 NC |
881 | #ifndef PERL_CORE |
882 | # define Nullhe Null(HE*) | |
883 | #endif | |
bf6bd887 | 884 | #define HeNEXT(he) (he)->hent_next |
ff68c719 | 885 | #define HeKEY_hek(he) (he)->hent_hek |
886 | #define HeKEY(he) HEK_KEY(HeKEY_hek(he)) | |
bbce6d69 | 887 | #define HeKEY_sv(he) (*(SV**)HeKEY(he)) |
ff68c719 | 888 | #define HeKLEN(he) HEK_LEN(HeKEY_hek(he)) |
da58a35d | 889 | #define HeKUTF8(he) HEK_UTF8(HeKEY_hek(he)) |
19692e8d | 890 | #define HeKWASUTF8(he) HEK_WASUTF8(HeKEY_hek(he)) |
da58a35d | 891 | #define HeKLEN_UTF8(he) (HeKUTF8(he) ? -HeKLEN(he) : HeKLEN(he)) |
19692e8d | 892 | #define HeKFLAGS(he) HEK_FLAGS(HeKEY_hek(he)) |
de616631 | 893 | #define HeVAL(he) (he)->he_valu.hent_val |
ff68c719 | 894 | #define HeHASH(he) HEK_HASH(HeKEY_hek(he)) |
1e422769 | 895 | #define HePV(he,lp) ((HeKLEN(he) == HEf_SVKEY) ? \ |
896 | SvPV(HeKEY_sv(he),lp) : \ | |
402f0539 | 897 | ((lp = HeKLEN(he)), HeKEY(he))) |
289d3c6a NC |
898 | #define HeUTF8(he) ((HeKLEN(he) == HEf_SVKEY) ? \ |
899 | SvUTF8(HeKEY_sv(he)) : \ | |
900 | (U32)HeKUTF8(he)) | |
1e422769 | 901 | |
bbce6d69 | 902 | #define HeSVKEY(he) ((HeKEY(he) && \ |
903 | HeKLEN(he) == HEf_SVKEY) ? \ | |
a0714e2c | 904 | HeKEY_sv(he) : NULL) |
bbce6d69 | 905 | |
906 | #define HeSVKEY_force(he) (HeKEY(he) ? \ | |
907 | ((HeKLEN(he) == HEf_SVKEY) ? \ | |
908 | HeKEY_sv(he) : \ | |
ad25789c NC |
909 | newSVpvn_flags(HeKEY(he), \ |
910 | HeKLEN(he), SVs_TEMP)) : \ | |
3280af22 | 911 | &PL_sv_undef) |
1e422769 | 912 | #define HeSVKEY_set(he,sv) ((HeKLEN(he) = HEf_SVKEY), (HeKEY_sv(he) = sv)) |
bbce6d69 | 913 | |
3ae1b226 NC |
914 | #ifndef PERL_CORE |
915 | # define Nullhek Null(HEK*) | |
916 | #endif | |
71be2cbc | 917 | #define HEK_BASESIZE STRUCT_OFFSET(HEK, hek_key[0]) |
ff68c719 | 918 | #define HEK_HASH(hek) (hek)->hek_hash |
919 | #define HEK_LEN(hek) (hek)->hek_len | |
920 | #define HEK_KEY(hek) (hek)->hek_key | |
19692e8d NC |
921 | #define HEK_FLAGS(hek) (*((unsigned char *)(HEK_KEY(hek))+HEK_LEN(hek)+1)) |
922 | ||
923 | #define HVhek_UTF8 0x01 /* Key is utf8 encoded. */ | |
924 | #define HVhek_WASUTF8 0x02 /* Key is bytes here, but was supplied as utf8. */ | |
45e34800 | 925 | #define HVhek_UNSHARED 0x08 /* This key isn't a shared hash key. */ |
19692e8d | 926 | #define HVhek_FREEKEY 0x100 /* Internal flag to say key is malloc()ed. */ |
e16e2ff8 NC |
927 | #define HVhek_PLACEHOLD 0x200 /* Internal flag to create placeholder. |
928 | * (may change, but Storable is a core module) */ | |
44b87b50 NC |
929 | #define HVhek_KEYCANONICAL 0x400 /* Internal flag - key is in canonical form. |
930 | If the string is UTF-8, it cannot be | |
931 | converted to bytes. */ | |
19692e8d NC |
932 | #define HVhek_MASK 0xFF |
933 | ||
7dc86639 | 934 | #define HVhek_ENABLEHVKFLAGS (HVhek_MASK & ~(HVhek_UNSHARED)) |
fdcd69b6 | 935 | |
19692e8d NC |
936 | #define HEK_UTF8(hek) (HEK_FLAGS(hek) & HVhek_UTF8) |
937 | #define HEK_UTF8_on(hek) (HEK_FLAGS(hek) |= HVhek_UTF8) | |
938 | #define HEK_UTF8_off(hek) (HEK_FLAGS(hek) &= ~HVhek_UTF8) | |
939 | #define HEK_WASUTF8(hek) (HEK_FLAGS(hek) & HVhek_WASUTF8) | |
940 | #define HEK_WASUTF8_on(hek) (HEK_FLAGS(hek) |= HVhek_WASUTF8) | |
941 | #define HEK_WASUTF8_off(hek) (HEK_FLAGS(hek) &= ~HVhek_WASUTF8) | |
d18c6117 | 942 | |
5cbe4eec | 943 | /* calculate HV array allocation */ |
36768cf4 NC |
944 | #ifndef PERL_USE_LARGE_HV_ALLOC |
945 | /* Default to allocating the correct size - default to assuming that malloc() | |
946 | is not broken and is efficient at allocating blocks sized at powers-of-two. | |
947 | */ | |
d18c6117 GS |
948 | # define PERL_HV_ARRAY_ALLOC_BYTES(size) ((size) * sizeof(HE*)) |
949 | #else | |
950 | # define MALLOC_OVERHEAD 16 | |
951 | # define PERL_HV_ARRAY_ALLOC_BYTES(size) \ | |
952 | (((size) < 64) \ | |
953 | ? (size) * sizeof(HE*) \ | |
954 | : (size) * sizeof(HE*) * 2 - MALLOC_OVERHEAD) | |
955 | #endif | |
37d85e3a | 956 | |
e16e2ff8 NC |
957 | /* Flags for hv_iternext_flags. */ |
958 | #define HV_ITERNEXT_WANTPLACEHOLDERS 0x01 /* Don't skip placeholders. */ | |
959 | ||
7a7b9979 | 960 | #define hv_iternext(hv) hv_iternext_flags(hv, 0) |
b1bc3f34 | 961 | #define hv_magic(hv, gv, how) sv_magic(MUTABLE_SV(hv), MUTABLE_SV(gv), how, NULL, 0) |
8581adba | 962 | #define hv_undef(hv) Perl_hv_undef_flags(aTHX_ hv, 0) |
7a7b9979 | 963 | |
bbd42945 FC |
964 | #define Perl_sharepvn(pv, len, hash) HEK_KEY(share_hek(pv, len, hash)) |
965 | #define sharepvn(pv, len, hash) Perl_sharepvn(pv, len, hash) | |
bfcb3514 | 966 | |
d1db91c6 NC |
967 | #define share_hek_hek(hek) \ |
968 | (++(((struct shared_he *)(((char *)hek) \ | |
969 | - STRUCT_OFFSET(struct shared_he, \ | |
970 | shared_he_hek))) \ | |
de616631 | 971 | ->shared_he_he.he_valu.hent_refcount), \ |
d1db91c6 NC |
972 | hek) |
973 | ||
99978bb4 NC |
974 | #define hv_store_ent(hv, keysv, val, hash) \ |
975 | ((HE *) hv_common((hv), (keysv), NULL, 0, 0, HV_FETCH_ISSTORE, \ | |
976 | (val), (hash))) | |
4c2df08c | 977 | |
99978bb4 NC |
978 | #define hv_exists_ent(hv, keysv, hash) \ |
979 | (hv_common((hv), (keysv), NULL, 0, 0, HV_FETCH_ISEXISTS, 0, (hash)) \ | |
4c2df08c | 980 | ? TRUE : FALSE) |
99978bb4 NC |
981 | #define hv_fetch_ent(hv, keysv, lval, hash) \ |
982 | ((HE *) hv_common((hv), (keysv), NULL, 0, 0, \ | |
983 | ((lval) ? HV_FETCH_LVALUE : 0), NULL, (hash))) | |
984 | #define hv_delete_ent(hv, key, flags, hash) \ | |
b1bc3f34 NC |
985 | (MUTABLE_SV(hv_common((hv), (key), NULL, 0, 0, (flags) | HV_DELETE, \ |
986 | NULL, (hash)))) | |
99978bb4 NC |
987 | |
988 | #define hv_store_flags(hv, key, klen, val, hash, flags) \ | |
989 | ((SV**) hv_common((hv), NULL, (key), (klen), (flags), \ | |
990 | (HV_FETCH_ISSTORE|HV_FETCH_JUST_SV), (val), \ | |
991 | (hash))) | |
992 | ||
993 | #define hv_store(hv, key, klen, val, hash) \ | |
994 | ((SV**) hv_common_key_len((hv), (key), (klen), \ | |
a038e571 | 995 | (HV_FETCH_ISSTORE|HV_FETCH_JUST_SV), \ |
99978bb4 | 996 | (val), (hash))) |
a038e571 | 997 | |
99978bb4 NC |
998 | #define hv_exists(hv, key, klen) \ |
999 | (hv_common_key_len((hv), (key), (klen), HV_FETCH_ISEXISTS, NULL, 0) \ | |
a038e571 NC |
1000 | ? TRUE : FALSE) |
1001 | ||
99978bb4 NC |
1002 | #define hv_fetch(hv, key, klen, lval) \ |
1003 | ((SV**) hv_common_key_len((hv), (key), (klen), (lval) \ | |
a038e571 NC |
1004 | ? (HV_FETCH_JUST_SV | HV_FETCH_LVALUE) \ |
1005 | : HV_FETCH_JUST_SV, NULL, 0)) | |
1006 | ||
99978bb4 | 1007 | #define hv_delete(hv, key, klen, flags) \ |
b1bc3f34 NC |
1008 | (MUTABLE_SV(hv_common_key_len((hv), (key), (klen), \ |
1009 | (flags) | HV_DELETE, NULL, 0))) | |
a038e571 | 1010 | |
71ad1b0c NC |
1011 | /* This refcounted he structure is used for storing the hints used for lexical |
1012 | pragmas. Without threads, it's basically struct he + refcount. | |
1013 | With threads, life gets more complex as the structure needs to be shared | |
1014 | between threads (because it hangs from OPs, which are shared), hence the | |
1015 | alternate definition and mutex. */ | |
1016 | ||
44ebaf21 NC |
1017 | struct refcounted_he; |
1018 | ||
20439bc7 Z |
1019 | /* flags for the refcounted_he API */ |
1020 | #define REFCOUNTED_HE_KEY_UTF8 0x00000001 | |
94250aee FC |
1021 | #ifdef PERL_CORE |
1022 | # define REFCOUNTED_HE_EXISTS 0x00000002 | |
1023 | #endif | |
20439bc7 | 1024 | |
71ad1b0c NC |
1025 | #ifdef PERL_CORE |
1026 | ||
b6bbf3fa | 1027 | /* Gosh. This really isn't a good name any longer. */ |
71ad1b0c NC |
1028 | struct refcounted_he { |
1029 | struct refcounted_he *refcounted_he_next; /* next entry in chain */ | |
cbb1fbea | 1030 | #ifdef USE_ITHREADS |
b6bbf3fa NC |
1031 | U32 refcounted_he_hash; |
1032 | U32 refcounted_he_keylen; | |
cbb1fbea | 1033 | #else |
71ad1b0c | 1034 | HEK *refcounted_he_hek; /* hint key */ |
cbb1fbea | 1035 | #endif |
b6bbf3fa NC |
1036 | union { |
1037 | IV refcounted_he_u_iv; | |
1038 | UV refcounted_he_u_uv; | |
1039 | STRLEN refcounted_he_u_len; | |
44ebaf21 | 1040 | void *refcounted_he_u_ptr; /* Might be useful in future */ |
b6bbf3fa | 1041 | } refcounted_he_val; |
9bd87817 | 1042 | U32 refcounted_he_refcnt; /* reference count */ |
b6bbf3fa NC |
1043 | /* First byte is flags. Then NUL-terminated value. Then for ithreads, |
1044 | non-NUL terminated key. */ | |
1045 | char refcounted_he_data[1]; | |
71ad1b0c NC |
1046 | }; |
1047 | ||
20439bc7 Z |
1048 | /* |
1049 | =for apidoc m|SV *|refcounted_he_fetch_pvs|const struct refcounted_he *chain|const char *key|U32 flags | |
1050 | ||
1051 | Like L</refcounted_he_fetch_pvn>, but takes a literal string instead of | |
1052 | a string/length pair, and no precomputed hash. | |
1053 | ||
1054 | =cut | |
1055 | */ | |
1056 | ||
1057 | #define refcounted_he_fetch_pvs(chain, key, flags) \ | |
1058 | Perl_refcounted_he_fetch_pvn(aTHX_ chain, STR_WITH_LEN(key), 0, flags) | |
1059 | ||
1060 | /* | |
1061 | =for apidoc m|struct refcounted_he *|refcounted_he_new_pvs|struct refcounted_he *parent|const char *key|SV *value|U32 flags | |
1062 | ||
1063 | Like L</refcounted_he_new_pvn>, but takes a literal string instead of | |
1064 | a string/length pair, and no precomputed hash. | |
1065 | ||
1066 | =cut | |
1067 | */ | |
1068 | ||
1069 | #define refcounted_he_new_pvs(parent, key, value, flags) \ | |
1070 | Perl_refcounted_he_new_pvn(aTHX_ parent, STR_WITH_LEN(key), 0, value, flags) | |
1071 | ||
b6bbf3fa NC |
1072 | /* Flag bits are HVhek_UTF8, HVhek_WASUTF8, then */ |
1073 | #define HVrhek_undef 0x00 /* Value is undef. */ | |
44ebaf21 NC |
1074 | #define HVrhek_delete 0x10 /* Value is placeholder - signifies delete. */ |
1075 | #define HVrhek_IV 0x20 /* Value is IV. */ | |
1076 | #define HVrhek_UV 0x30 /* Value is UV. */ | |
1077 | #define HVrhek_PV 0x40 /* Value is a (byte) string. */ | |
1078 | #define HVrhek_PV_UTF8 0x50 /* Value is a (utf8) string. */ | |
1079 | /* Two spare. As these have to live in the optree, you can't store anything | |
1080 | interpreter specific, such as SVs. :-( */ | |
1081 | #define HVrhek_typemask 0x70 | |
1082 | ||
1083 | #ifdef USE_ITHREADS | |
1084 | /* A big expression to find the key offset */ | |
1085 | #define REF_HE_KEY(chain) \ | |
1086 | ((((chain->refcounted_he_data[0] & 0x60) == 0x40) \ | |
1087 | ? chain->refcounted_he_val.refcounted_he_u_len + 1 : 0) \ | |
1088 | + 1 + chain->refcounted_he_data) | |
1089 | #endif | |
b6bbf3fa | 1090 | |
71ad1b0c NC |
1091 | # ifdef USE_ITHREADS |
1092 | # define HINTS_REFCNT_LOCK MUTEX_LOCK(&PL_hints_mutex) | |
1093 | # define HINTS_REFCNT_UNLOCK MUTEX_UNLOCK(&PL_hints_mutex) | |
1094 | # else | |
1095 | # define HINTS_REFCNT_LOCK NOOP | |
1096 | # define HINTS_REFCNT_UNLOCK NOOP | |
1097 | # endif | |
1098 | #endif | |
1099 | ||
1100 | #ifdef USE_ITHREADS | |
1101 | # define HINTS_REFCNT_INIT MUTEX_INIT(&PL_hints_mutex) | |
1102 | # define HINTS_REFCNT_TERM MUTEX_DESTROY(&PL_hints_mutex) | |
1103 | #else | |
1104 | # define HINTS_REFCNT_INIT NOOP | |
1105 | # define HINTS_REFCNT_TERM NOOP | |
1106 | #endif | |
1107 | ||
324a0d18 JH |
1108 | /* Hash actions |
1109 | * Passed in PERL_MAGIC_uvar calls | |
1110 | */ | |
b54b4831 NC |
1111 | #define HV_DISABLE_UVAR_XKEY 0x01 |
1112 | /* We need to ensure that these don't clash with G_DISCARD, which is 2, as it | |
1113 | is documented as being passed to hv_delete(). */ | |
1114 | #define HV_FETCH_ISSTORE 0x04 | |
1115 | #define HV_FETCH_ISEXISTS 0x08 | |
1116 | #define HV_FETCH_LVALUE 0x10 | |
1117 | #define HV_FETCH_JUST_SV 0x20 | |
9dbc5603 | 1118 | #define HV_DELETE 0x40 |
df5f182b | 1119 | #define HV_FETCH_EMPTY_HE 0x80 /* Leave HeVAL null. */ |
324a0d18 | 1120 | |
745edda6 FC |
1121 | /* Must not conflict with HVhek_UTF8 */ |
1122 | #define HV_NAME_SETALL 0x02 | |
1123 | ||
bfcb3514 | 1124 | /* |
78ac7dd9 NC |
1125 | =for apidoc newHV |
1126 | ||
1127 | Creates a new HV. The reference count is set to 1. | |
1128 | ||
1129 | =cut | |
1130 | */ | |
1131 | ||
dbebbdb4 | 1132 | #define newHV() MUTABLE_HV(newSV_type(SVt_PVHV)) |
78ac7dd9 NC |
1133 | |
1134 | /* | |
bfcb3514 NC |
1135 | * Local variables: |
1136 | * c-indentation-style: bsd | |
1137 | * c-basic-offset: 4 | |
14d04a33 | 1138 | * indent-tabs-mode: nil |
bfcb3514 NC |
1139 | * End: |
1140 | * | |
14d04a33 | 1141 | * ex: set ts=8 sts=4 sw=4 et: |
bfcb3514 | 1142 | */ |