Commit | Line | Data |
a0d0e21e |
1 | /* hv.h |
79072805 |
2 | * |
4bb101f2 |
3 | * Copyright (C) 1991, 1992, 1993, 1996, 1997, 1998, 1999, |
4 | * 2000, 2001, 2002, by Larry Wall and others |
79072805 |
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 |
9 | */ |
10 | |
5cbe4eec |
11 | /* typedefs to eliminate some typing */ |
79072805 |
12 | typedef struct he HE; |
ff68c719 |
13 | typedef struct hek HEK; |
79072805 |
14 | |
5cbe4eec |
15 | /* entry in hash value chain */ |
79072805 |
16 | struct he { |
5cbe4eec |
17 | HE *hent_next; /* next entry in chain */ |
18 | HEK *hent_hek; /* hash key */ |
19 | SV *hent_val; /* scalar value that was hashed */ |
bbce6d69 |
20 | }; |
21 | |
5cbe4eec |
22 | /* hash key -- defined separately for use as shared pointer */ |
ff68c719 |
23 | struct hek { |
5cbe4eec |
24 | U32 hek_hash; /* hash of key */ |
25 | I32 hek_len; /* length of hash key */ |
26 | char hek_key[1]; /* variable-length hash key */ |
e05949c7 |
27 | /* the hash-key is \0-terminated */ |
1e54db1a |
28 | /* after the \0 there is a byte for flags, such as whether the key |
29 | is UTF-8 */ |
79072805 |
30 | }; |
31 | |
5cbe4eec |
32 | /* hash structure: */ |
6ee623d5 |
33 | /* This structure must match the beginning of struct xpvmg in sv.h. */ |
79072805 |
34 | struct xpvhv { |
463ee0b2 |
35 | char * xhv_array; /* pointer to malloced string */ |
36 | STRLEN xhv_fill; /* how full xhv_array currently is */ |
37 | STRLEN xhv_max; /* subscript of last element of xhv_array */ |
6ee623d5 |
38 | IV xhv_keys; /* how many elements in the array */ |
65202027 |
39 | NV xnv_nv; /* numeric value, if any */ |
8aacddc1 |
40 | #define xhv_placeholders xnv_nv |
79072805 |
41 | MAGIC* xmg_magic; /* magic for scalar array */ |
42 | HV* xmg_stash; /* class package */ |
43 | |
79072805 |
44 | I32 xhv_riter; /* current root of iterator */ |
45 | HE *xhv_eiter; /* current entry of iterator */ |
46 | PMOP *xhv_pmroot; /* list of pm's for this package */ |
47 | char *xhv_name; /* name, if a symbol table */ |
79072805 |
48 | }; |
49 | |
5cbe4eec |
50 | /* hash a key */ |
5afd6d42 |
51 | /* FYI: This is the "One-at-a-Time" algorithm by Bob Jenkins |
52 | * from requirements by Colin Plumb. |
53 | * (http://burtleburtle.net/bob/hash/doobs.html) */ |
54 | /* The use of a temporary pointer and the casting games |
55 | * is needed to serve the dual purposes of |
56 | * (a) the hashed data being interpreted as "unsigned char" (new since 5.8, |
57 | * a "char" can be either signed or signed, depending on the compiler) |
58 | * (b) catering for old code that uses a "char" |
830b38bd |
59 | * |
504f80c1 |
60 | * The "hash seed" feature was added in Perl 5.8.1 to perturb the results |
61 | * to avoid "algorithmic complexity attacks". |
830b38bd |
62 | * |
63 | * If USE_HASH_SEED is defined, hash randomisation is done by default |
64 | * If USE_HASH_SEED_EXPLICIT is defined, hash randomisation is done |
65 | * only if the environment variable PERL_HASH_SEED is set. |
66 | * For maximal control, one can define PERL_HASH_SEED. |
67 | * (see also erl.c:perl_parse()). |
5afd6d42 |
68 | */ |
830b38bd |
69 | #ifndef PERL_HASH_SEED |
70 | # if defined(USE_HASH_SEED) || defined(USE_HASH_SEED_EXPLICIT) |
71 | # define PERL_HASH_SEED PL_hash_seed |
72 | # else |
73 | # define PERL_HASH_SEED 0 |
74 | # endif |
504f80c1 |
75 | #endif |
bf6bd887 |
76 | #define PERL_HASH(hash,str,len) \ |
77 | STMT_START { \ |
5afd6d42 |
78 | register const char *s_PeRlHaSh_tmp = str; \ |
79 | register const unsigned char *s_PeRlHaSh = (const unsigned char *)s_PeRlHaSh_tmp; \ |
bf6bd887 |
80 | register I32 i_PeRlHaSh = len; \ |
504f80c1 |
81 | register U32 hash_PeRlHaSh = PERL_HASH_SEED; \ |
a6fe520e |
82 | while (i_PeRlHaSh--) { \ |
83 | hash_PeRlHaSh += *s_PeRlHaSh++; \ |
84 | hash_PeRlHaSh += (hash_PeRlHaSh << 10); \ |
85 | hash_PeRlHaSh ^= (hash_PeRlHaSh >> 6); \ |
86 | } \ |
87 | hash_PeRlHaSh += (hash_PeRlHaSh << 3); \ |
88 | hash_PeRlHaSh ^= (hash_PeRlHaSh >> 11); \ |
4fee5c71 |
89 | (hash) = (hash_PeRlHaSh + (hash_PeRlHaSh << 15)); \ |
bf6bd887 |
90 | } STMT_END |
91 | |
22718ec5 |
92 | #if defined (PERL_IN_HV_C) || defined (MOD_PERL) |
4b5190b5 |
93 | #define PERL_HASH_INTERNAL(hash,str,len) \ |
94 | STMT_START { \ |
95 | register const char *s_PeRlHaSh_tmp = str; \ |
96 | register const unsigned char *s_PeRlHaSh = (const unsigned char *)s_PeRlHaSh_tmp; \ |
97 | register I32 i_PeRlHaSh = len; \ |
98 | register U32 hash_PeRlHaSh = PL_new_hash_seed; \ |
99 | while (i_PeRlHaSh--) { \ |
100 | hash_PeRlHaSh += *s_PeRlHaSh++; \ |
101 | hash_PeRlHaSh += (hash_PeRlHaSh << 10); \ |
102 | hash_PeRlHaSh ^= (hash_PeRlHaSh >> 6); \ |
103 | } \ |
104 | hash_PeRlHaSh += (hash_PeRlHaSh << 3); \ |
105 | hash_PeRlHaSh ^= (hash_PeRlHaSh >> 11); \ |
106 | (hash) = (hash_PeRlHaSh + (hash_PeRlHaSh << 15)); \ |
107 | } STMT_END |
108 | #endif |
109 | |
954c1994 |
110 | /* |
ccfc67b7 |
111 | =head1 Hash Manipulation Functions |
112 | |
954c1994 |
113 | =for apidoc AmU||HEf_SVKEY |
114 | This flag, used in the length slot of hash entries and magic structures, |
d1be9408 |
115 | specifies the structure contains an C<SV*> pointer where a C<char*> pointer |
954c1994 |
116 | is to be expected. (For information only--not to be used). |
117 | |
ccfc67b7 |
118 | =head1 Handy Values |
119 | |
954c1994 |
120 | =for apidoc AmU||Nullhv |
121 | Null HV pointer. |
122 | |
ccfc67b7 |
123 | =head1 Hash Manipulation Functions |
124 | |
954c1994 |
125 | =for apidoc Am|char*|HvNAME|HV* stash |
126 | Returns the package name of a stash. See C<SvSTASH>, C<CvSTASH>. |
127 | |
128 | =for apidoc Am|void*|HeKEY|HE* he |
129 | Returns the actual pointer stored in the key slot of the hash entry. The |
130 | pointer may be either C<char*> or C<SV*>, depending on the value of |
131 | C<HeKLEN()>. Can be assigned to. The C<HePV()> or C<HeSVKEY()> macros are |
132 | usually preferable for finding the value of a key. |
133 | |
134 | =for apidoc Am|STRLEN|HeKLEN|HE* he |
135 | If this is negative, and amounts to C<HEf_SVKEY>, it indicates the entry |
136 | holds an C<SV*> key. Otherwise, holds the actual length of the key. Can |
137 | be assigned to. The C<HePV()> macro is usually preferable for finding key |
138 | lengths. |
139 | |
140 | =for apidoc Am|SV*|HeVAL|HE* he |
141 | Returns the value slot (type C<SV*>) stored in the hash entry. |
142 | |
143 | =for apidoc Am|U32|HeHASH|HE* he |
144 | Returns the computed hash stored in the hash entry. |
145 | |
146 | =for apidoc Am|char*|HePV|HE* he|STRLEN len |
147 | Returns the key slot of the hash entry as a C<char*> value, doing any |
148 | necessary dereferencing of possibly C<SV*> keys. The length of the string |
149 | is placed in C<len> (this is a macro, so do I<not> use C<&len>). If you do |
150 | not care about what the length of the key is, you may use the global |
151 | variable C<PL_na>, though this is rather less efficient than using a local |
152 | variable. Remember though, that hash keys in perl are free to contain |
153 | embedded nulls, so using C<strlen()> or similar is not a good way to find |
154 | the length of hash keys. This is very similar to the C<SvPV()> macro |
155 | described elsewhere in this document. |
156 | |
157 | =for apidoc Am|SV*|HeSVKEY|HE* he |
158 | Returns the key as an C<SV*>, or C<Nullsv> if the hash entry does not |
159 | contain an C<SV*> key. |
160 | |
161 | =for apidoc Am|SV*|HeSVKEY_force|HE* he |
162 | Returns the key as an C<SV*>. Will create and return a temporary mortal |
163 | C<SV*> if the hash entry contains only a C<char*> key. |
164 | |
165 | =for apidoc Am|SV*|HeSVKEY_set|HE* he|SV* sv |
166 | Sets the key to a given C<SV*>, taking care to set the appropriate flags to |
167 | indicate the presence of an C<SV*> key, and returns the same |
168 | C<SV*>. |
169 | |
170 | =cut |
171 | */ |
bf6bd887 |
172 | |
bf5b86ae |
173 | /* these hash entry flags ride on hent_klen (for use only in magic/tied HVs) */ |
d1be9408 |
174 | #define HEf_SVKEY -2 /* hent_key is an SV* */ |
bf6bd887 |
175 | |
176 | |
79072805 |
177 | #define Nullhv Null(HV*) |
4efcf9a2 |
178 | #define HvARRAY(hv) (*(HE***)&((XPVHV*) SvANY(hv))->xhv_array) |
79072805 |
179 | #define HvFILL(hv) ((XPVHV*) SvANY(hv))->xhv_fill |
463ee0b2 |
180 | #define HvMAX(hv) ((XPVHV*) SvANY(hv))->xhv_max |
79072805 |
181 | #define HvRITER(hv) ((XPVHV*) SvANY(hv))->xhv_riter |
182 | #define HvEITER(hv) ((XPVHV*) SvANY(hv))->xhv_eiter |
183 | #define HvPMROOT(hv) ((XPVHV*) SvANY(hv))->xhv_pmroot |
184 | #define HvNAME(hv) ((XPVHV*) SvANY(hv))->xhv_name |
a0d0e21e |
185 | |
8aacddc1 |
186 | /* the number of keys (including any placeholers) */ |
187 | #define XHvTOTALKEYS(xhv) ((xhv)->xhv_keys) |
188 | |
189 | /* The number of placeholders in the enumerated-keys hash */ |
205c8ad3 |
190 | #define XHvPLACEHOLDERS(xhv) ((xhv)->xhv_placeholders) |
8aacddc1 |
191 | |
dd2710a1 |
192 | /* the number of keys that exist() (i.e. excluding placeholders) */ |
193 | #define XHvUSEDKEYS(xhv) (XHvTOTALKEYS(xhv) - (IV)XHvPLACEHOLDERS(xhv)) |
8aacddc1 |
194 | |
195 | /* |
196 | * HvKEYS gets the number of keys that actually exist(), and is provided |
197 | * for backwards compatibility with old XS code. The core uses HvUSEDKEYS |
198 | * (keys, excluding placeholdes) and HvTOTALKEYS (including placeholders) |
199 | */ |
200 | #define HvKEYS(hv) XHvUSEDKEYS((XPVHV*) SvANY(hv)) |
201 | #define HvUSEDKEYS(hv) XHvUSEDKEYS((XPVHV*) SvANY(hv)) |
202 | #define HvTOTALKEYS(hv) XHvTOTALKEYS((XPVHV*) SvANY(hv)) |
203 | #define HvPLACEHOLDERS(hv) XHvPLACEHOLDERS((XPVHV*) SvANY(hv)) |
204 | |
bf6bd887 |
205 | #define HvSHAREKEYS(hv) (SvFLAGS(hv) & SVphv_SHAREKEYS) |
206 | #define HvSHAREKEYS_on(hv) (SvFLAGS(hv) |= SVphv_SHAREKEYS) |
207 | #define HvSHAREKEYS_off(hv) (SvFLAGS(hv) &= ~SVphv_SHAREKEYS) |
208 | |
19692e8d |
209 | /* This is an optimisation flag. It won't be set if all hash keys have a 0 |
210 | * flag. Currently the only flags relate to utf8. |
211 | * Hence it won't be set if all keys are 8 bit only. It will be set if any key |
212 | * is utf8 (including 8 bit keys that were entered as utf8, and need upgrading |
213 | * when retrieved during iteration. It may still be set when there are no longer |
214 | * any utf8 keys. |
fdcd69b6 |
215 | * See HVhek_ENABLEHVKFLAGS for the trigger. |
19692e8d |
216 | */ |
217 | #define HvHASKFLAGS(hv) (SvFLAGS(hv) & SVphv_HASKFLAGS) |
218 | #define HvHASKFLAGS_on(hv) (SvFLAGS(hv) |= SVphv_HASKFLAGS) |
219 | #define HvHASKFLAGS_off(hv) (SvFLAGS(hv) &= ~SVphv_HASKFLAGS) |
574c8022 |
220 | |
bf5b86ae |
221 | #define HvLAZYDEL(hv) (SvFLAGS(hv) & SVphv_LAZYDEL) |
222 | #define HvLAZYDEL_on(hv) (SvFLAGS(hv) |= SVphv_LAZYDEL) |
223 | #define HvLAZYDEL_off(hv) (SvFLAGS(hv) &= ~SVphv_LAZYDEL) |
224 | |
4b5190b5 |
225 | #define HvREHASH(hv) (SvFLAGS(hv) & SVphv_REHASH) |
226 | #define HvREHASH_on(hv) (SvFLAGS(hv) |= SVphv_REHASH) |
227 | #define HvREHASH_off(hv) (SvFLAGS(hv) &= ~SVphv_REHASH) |
228 | |
a0d0e21e |
229 | /* Maybe amagical: */ |
230 | /* #define HV_AMAGICmb(hv) (SvFLAGS(hv) & (SVpgv_badAM | SVpgv_AM)) */ |
231 | |
232 | #define HV_AMAGIC(hv) (SvFLAGS(hv) & SVpgv_AM) |
233 | #define HV_AMAGIC_on(hv) (SvFLAGS(hv) |= SVpgv_AM) |
234 | #define HV_AMAGIC_off(hv) (SvFLAGS(hv) &= ~SVpgv_AM) |
235 | |
236 | /* |
237 | #define HV_AMAGICbad(hv) (SvFLAGS(hv) & SVpgv_badAM) |
238 | #define HV_badAMAGIC_on(hv) (SvFLAGS(hv) |= SVpgv_badAM) |
239 | #define HV_badAMAGIC_off(hv) (SvFLAGS(hv) &= ~SVpgv_badAM) |
240 | */ |
bf6bd887 |
241 | |
242 | #define Nullhe Null(HE*) |
243 | #define HeNEXT(he) (he)->hent_next |
ff68c719 |
244 | #define HeKEY_hek(he) (he)->hent_hek |
245 | #define HeKEY(he) HEK_KEY(HeKEY_hek(he)) |
bbce6d69 |
246 | #define HeKEY_sv(he) (*(SV**)HeKEY(he)) |
ff68c719 |
247 | #define HeKLEN(he) HEK_LEN(HeKEY_hek(he)) |
da58a35d |
248 | #define HeKUTF8(he) HEK_UTF8(HeKEY_hek(he)) |
19692e8d |
249 | #define HeKWASUTF8(he) HEK_WASUTF8(HeKEY_hek(he)) |
4b5190b5 |
250 | #define HeKREHASH(he) HEK_REHASH(HeKEY_hek(he)) |
da58a35d |
251 | #define HeKLEN_UTF8(he) (HeKUTF8(he) ? -HeKLEN(he) : HeKLEN(he)) |
19692e8d |
252 | #define HeKFLAGS(he) HEK_FLAGS(HeKEY_hek(he)) |
bf6bd887 |
253 | #define HeVAL(he) (he)->hent_val |
ff68c719 |
254 | #define HeHASH(he) HEK_HASH(HeKEY_hek(he)) |
1e422769 |
255 | #define HePV(he,lp) ((HeKLEN(he) == HEf_SVKEY) ? \ |
256 | SvPV(HeKEY_sv(he),lp) : \ |
257 | (((lp = HeKLEN(he)) >= 0) ? \ |
258 | HeKEY(he) : Nullch)) |
259 | |
bbce6d69 |
260 | #define HeSVKEY(he) ((HeKEY(he) && \ |
261 | HeKLEN(he) == HEf_SVKEY) ? \ |
262 | HeKEY_sv(he) : Nullsv) |
263 | |
264 | #define HeSVKEY_force(he) (HeKEY(he) ? \ |
265 | ((HeKLEN(he) == HEf_SVKEY) ? \ |
266 | HeKEY_sv(he) : \ |
79cb57f6 |
267 | sv_2mortal(newSVpvn(HeKEY(he), \ |
bbce6d69 |
268 | HeKLEN(he)))) : \ |
3280af22 |
269 | &PL_sv_undef) |
1e422769 |
270 | #define HeSVKEY_set(he,sv) ((HeKLEN(he) = HEf_SVKEY), (HeKEY_sv(he) = sv)) |
bbce6d69 |
271 | |
ff68c719 |
272 | #define Nullhek Null(HEK*) |
71be2cbc |
273 | #define HEK_BASESIZE STRUCT_OFFSET(HEK, hek_key[0]) |
ff68c719 |
274 | #define HEK_HASH(hek) (hek)->hek_hash |
275 | #define HEK_LEN(hek) (hek)->hek_len |
276 | #define HEK_KEY(hek) (hek)->hek_key |
19692e8d |
277 | #define HEK_FLAGS(hek) (*((unsigned char *)(HEK_KEY(hek))+HEK_LEN(hek)+1)) |
278 | |
279 | #define HVhek_UTF8 0x01 /* Key is utf8 encoded. */ |
280 | #define HVhek_WASUTF8 0x02 /* Key is bytes here, but was supplied as utf8. */ |
4b5190b5 |
281 | #define HVhek_REHASH 0x04 /* This key is in an hv using a custom HASH . */ |
19692e8d |
282 | #define HVhek_FREEKEY 0x100 /* Internal flag to say key is malloc()ed. */ |
e16e2ff8 |
283 | #define HVhek_PLACEHOLD 0x200 /* Internal flag to create placeholder. |
284 | * (may change, but Storable is a core module) */ |
19692e8d |
285 | #define HVhek_MASK 0xFF |
286 | |
fdcd69b6 |
287 | /* Which flags enable HvHASKFLAGS? Somewhat a hack on a hack, as |
288 | HVhek_REHASH is only needed because the rehash flag has to be duplicated |
289 | into all keys as hv_iternext has no access to the hash flags. At this |
290 | point Storable's tests get upset, because sometimes hashes are "keyed" |
291 | and sometimes not, depending on the order of data insertion, and whether |
292 | it triggered rehashing. So currently HVhek_REHAS is exempt. |
293 | */ |
294 | |
295 | #define HVhek_ENABLEHVKFLAGS (HVhek_MASK - HVhek_REHASH) |
296 | |
19692e8d |
297 | #define HEK_UTF8(hek) (HEK_FLAGS(hek) & HVhek_UTF8) |
298 | #define HEK_UTF8_on(hek) (HEK_FLAGS(hek) |= HVhek_UTF8) |
299 | #define HEK_UTF8_off(hek) (HEK_FLAGS(hek) &= ~HVhek_UTF8) |
300 | #define HEK_WASUTF8(hek) (HEK_FLAGS(hek) & HVhek_WASUTF8) |
301 | #define HEK_WASUTF8_on(hek) (HEK_FLAGS(hek) |= HVhek_WASUTF8) |
302 | #define HEK_WASUTF8_off(hek) (HEK_FLAGS(hek) &= ~HVhek_WASUTF8) |
4b5190b5 |
303 | #define HEK_REHASH(hek) (HEK_FLAGS(hek) & HVhek_REHASH) |
304 | #define HEK_REHASH_on(hek) (HEK_FLAGS(hek) |= HVhek_REHASH) |
d18c6117 |
305 | |
5cbe4eec |
306 | /* calculate HV array allocation */ |
d18c6117 |
307 | #if defined(STRANGE_MALLOC) || defined(MYMALLOC) |
308 | # define PERL_HV_ARRAY_ALLOC_BYTES(size) ((size) * sizeof(HE*)) |
309 | #else |
310 | # define MALLOC_OVERHEAD 16 |
311 | # define PERL_HV_ARRAY_ALLOC_BYTES(size) \ |
312 | (((size) < 64) \ |
313 | ? (size) * sizeof(HE*) \ |
314 | : (size) * sizeof(HE*) * 2 - MALLOC_OVERHEAD) |
315 | #endif |
37d85e3a |
316 | |
e16e2ff8 |
317 | /* Flags for hv_iternext_flags. */ |
318 | #define HV_ITERNEXT_WANTPLACEHOLDERS 0x01 /* Don't skip placeholders. */ |
319 | |
37d85e3a |
320 | /* available as a function in hv.c */ |
321 | #define Perl_sharepvn(sv, len, hash) HEK_KEY(share_hek(sv, len, hash)) |
322 | #define sharepvn(sv, len, hash) Perl_sharepvn(sv, len, hash) |