| 1 |
|
| /* |
| 2 |
|
| * Digest::FP56x1xor.xs - C implementation of a 56bit pseudo additive fingerprint hash. |
| 3 |
|
| * |
| 4 |
|
| * (C) 2007-2008 jw@suse.de, Novell Inc. |
| 5 |
|
| * This module is free software. IT may be used |
| 6 |
|
| * and/or modified under the same terms as perl itself. |
| 7 |
|
| * |
| 8 |
|
| * See also |
| 9 |
|
| * perldoc perlxs |
| 10 |
|
| * perldoc perlguts |
| 11 |
|
| * perldoc perlxstut |
| 12 |
|
| * perldoc perlapi |
| 13 |
|
| * |
| 14 |
|
| * Currently gen_l() croaks, if no 64bit integers are available |
| 15 |
|
| * and gen() is suggested, which returns a hex string. |
| 16 |
|
| * If the compiler does not respond to 'long long' with 64bit, |
| 17 |
|
| * we die(). |
| 18 |
|
| * |
| 19 |
|
| */ |
| 20 |
|
| #include "EXTERN.h" |
| 21 |
|
| #include "perl.h" |
| 22 |
|
| #include "XSUB.h" |
| 23 |
|
| |
| 24 |
|
| #include "ppport.h" |
| 25 |
|
| |
| 26 |
|
| #define VERSION "0.15" // keep in sync with Changes, lib/Digest/FP56x1xor.pm |
| 27 |
|
| |
| 28 |
|
| #if !defined(_STDINT_H) && !defined(__WORDSIZE) |
| 29 |
|
| // stdint.h and bits/types.h also have these: |
| 30 |
|
| typedef unsigned long long uint64_t; |
| 31 |
|
| typedef unsigned int uint32_t; |
| 32 |
|
| #endif |
| 33 |
|
| |
| 34 |
|
| static uint64_t rand56[256] = { |
| 35 |
|
| #include "random56x256.h" |
| 36 |
|
| }; |
| 37 |
|
| |
| 38 |
|
| // rotate56 rotates the lower 56 bits in h by m bits. |
| 39 |
|
| // The high 8 bit of h are cleared and remain zero. |
| 40 |
5 |
| static uint64_t rotate56(uint64_t h, int m) |
| 41 |
|
| { |
| 42 |
5 |
| uint64_t x = (h << m) & 0x00ffffffffffffffLL; |
| 43 |
5 |
| h &= 0x00ffffffffffffffLL; |
| 44 |
5 |
| x |= (h >> (56-m)); |
| 45 |
5 |
| return x; |
| 46 |
|
| } |
| 47 |
|
| |
| 48 |
|
| #if 0 |
| 49 |
|
| // hmm, is this useful? how ould perl work with that |
| 50 |
|
| // string? I need to push it into |
| 51 |
|
| // postgresql and sqlite databases and have it compared there. |
| 52 |
|
| // the sql statements for this are always 7bit ascii, right? |
| 53 |
|
| // |
| 54 |
|
| static unsigned char *uint64tostr8(uint64_t h, unsigned char *buf) |
| 55 |
|
| { |
| 56 |
|
| unsigned char buf8[9]; |
| 57 |
|
| int i; |
| 58 |
|
| if (!buf) buf = buf8; |
| 59 |
|
| for (i = 0; i < 8; i++) |
| 60 |
|
| { |
| 61 |
|
| buf[i] = (unsigned char)((h>>(8*(7-i))) & 0xff); |
| 62 |
|
| } |
| 63 |
|
| return buf; |
| 64 |
|
| } |
| 65 |
|
| #endif |
| 66 |
|
| |
| 67 |
|
| // fp is a 64bit integer. |
| 68 |
|
| // highest 2 bit are always zero, next 6 bit are the modulo count. |
| 69 |
|
| // remaining lower 56 bit are fingerprint hash value. |
| 70 |
3 |
| static uint64_t gen_fp(const char *text, unsigned int len) |
| 71 |
|
| { |
| 72 |
3 |
| uint64_t h = 0; |
| 73 |
3 |
| unsigned int m = 0; // byte counter modulo 56 |
| 74 |
|
| unsigned int c; |
| 75 |
|
| |
| 76 |
|
| if (sizeof(uint64_t) != 8) |
| 77 |
|
| croak("sizeof(uint64_t) is %d, need 8\n", (int)sizeof(uint64_t)); |
| 78 |
|
| |
| 79 |
7 |
▶ | while (len-- > 0) |
Branch: 1 branch
| Branch | True | False | Total |
| while (len-- > 0) |
4 |
3 |
7 |
|
| 80 |
|
| { |
| 81 |
4 |
| c = (unsigned char)*text++; |
| 82 |
4 |
| uint64_t x = rotate56(rand56[c], m); |
| 83 |
4 |
| h ^= x; |
| 84 |
4 |
▶ | if (++m >= 56) m = 0; |
Branch: 1 branch
| Branch | True | False | Total |
| if (++m >= 56) m = 0; |
0 |
4 |
4 |
|
| 85 |
|
| } |
| 86 |
3 |
| h |= ((uint64_t)m) << 56; |
| 87 |
3 |
| return h; |
| 88 |
|
| } |
| 89 |
|
| |
| 90 |
1 |
| static uint64_t cat_fp(uint64_t h1, uint64_t h2) |
| 91 |
|
| { |
| 92 |
1 |
| unsigned int m1 = (h1 >> 56); |
| 93 |
1 |
| unsigned int m2 = (h2 >> 56); |
| 94 |
1 |
| uint64_t r = rotate56(h2, m1) ^ (h1 & 0x00ffffffffffffffLL); |
| 95 |
1 |
| m2 += m1; |
| 96 |
1 |
▶ | if (m2 >= 56) m2 -= 56; |
Branch: 1 branch
| Branch | True | False | Total |
| if (m2 >= 56) m2 -= 56; |
0 |
1 |
1 |
|
| 97 |
1 |
| r |= ((uint64_t)m2)<<56; |
| 98 |
1 |
| return r; |
| 99 |
|
| } |
| 100 |
|
| |
| 101 |
32 |
| static int xc2i(unsigned char c) |
| 102 |
|
| { |
| 103 |
|
| // ascii has numbers before upper before lower case. |
| 104 |
32 |
▶ | if (c <= '9') return c - '0'; |
Branch: 1 branch
| Branch | True | False | Total |
| if (c <= '9') return c - '0'; |
24 |
8 |
32 |
|
| 105 |
8 |
▶ | if (c <= 'F') return c - 'A' + 10; |
Branch: 1 branch
| Branch | True | False | Total |
| if (c <= 'F') return c - 'A' + 10; |
0 |
8 |
8 |
|
| 106 |
8 |
| return c - 'a' + 10; |
| 107 |
|
| } |
| 108 |
|
| |
| 109 |
2 |
| static uint64_t x2i64(unsigned char *t, int n) |
| 110 |
|
| { |
| 111 |
2 |
| uint64_t h = 0; |
| 112 |
34 |
▶ | while (n-- > 0) |
Branch: 1 branch
| Branch | True | False | Total |
| while (n-- > 0) |
32 |
2 |
34 |
|
| 113 |
32 |
| h = (h<<4) + xc2i(*t++); |
| 114 |
2 |
| return h; |
| 115 |
|
| } |
| 116 |
|
| |
| 117 |
4 |
| static char *ll2x16(uint64_t fp) |
| 118 |
|
| { |
| 119 |
|
| static char buf[18]; |
| 120 |
4 |
| sprintf(buf, "%08x%08x", (unsigned int)(fp>>32), (unsigned int)(fp & 0xffffffff)); |
| 121 |
4 |
| return buf; |
| 122 |
|
| } |
| 123 |
|
| |
| 124 |
|
| #if 0 |
| 125 |
|
| // same as ll2x16, but with 0x prefix |
| 126 |
|
| // may be used for human consumption. Databases don't like that. |
| 127 |
|
| static char *ll2x18(uint64_t fp) |
| 128 |
|
| { |
| 129 |
|
| static char buf[20]; |
| 130 |
|
| sprintf(buf, "0x%08x%08x", (unsigned int)(fp>>32), (unsigned int)(fp & 0xffffffff)); |
| 131 |
|
| return buf; |
| 132 |
|
| } |
| 133 |
|
| #endif |
| 134 |
|
| |
| 135 |
1 |
| static uint64_t cat_fp_x(const char *x1, const char *x2) |
| 136 |
|
| { |
| 137 |
1 |
| uint64_t h1 = x2i64((unsigned char *)x1, strlen(x1)); |
| 138 |
1 |
| uint64_t h2 = x2i64((unsigned char *)x2, strlen(x2)); |
| 139 |
1 |
| return cat_fp(h1, h2); |
| 140 |
|
| } |
| 141 |
|
| |
| 142 |
|
| // private version of \w, just like perl in ascii C locale. |
| 143 |
|
| // but also with quote chars, so that <div id="foo" name=bar> |
| 144 |
|
| // is preserverd during cooked() |
| 145 |
|
| |
| 146 |
0 |
| static int is_word_ch_q(int c) |
| 147 |
|
| { |
| 148 |
0 |
▶ | if ((c <= 'z' && c >= 'a') || |
Branch: 3 branches
| Branch | True | False | Total |
| if ((c <= 'z' && c >= 'a') || |
0 |
0 |
0 |
| if ((c <= 'z' && c >= 'a') || |
0 |
0 |
0 |
| if ((c <= 'z' && c >= 'a') || |
0 |
0 |
0 |
|
| 149 |
0 |
▶ | (c <= 'Z' && c >= 'A') || |
Branch: 2 branches
| Branch | True | False | Total |
| (c <= 'Z' && c >= 'A') || |
0 |
0 |
0 |
| (c <= 'Z' && c >= 'A') || |
0 |
0 |
0 |
|
| 150 |
0 |
▶ | (c <= '9' && c >= '0') || |
Branch: 2 branches
| Branch | True | False | Total |
| (c <= '9' && c >= '0') || |
0 |
0 |
0 |
| (c <= '9' && c >= '0') || |
0 |
0 |
0 |
|
| 151 |
0 |
▶ | c == '_' || c == '"' || c == '\'') |
Branch: 2 branches
| Branch | True | False | Total |
| c == '_' || c == '"' || c == '\'') |
0 |
0 |
0 |
| c == '_' || c == '"' || c == '\'') |
0 |
0 |
0 |
|
| 152 |
0 |
| return 1; |
| 153 |
0 |
| return 0; |
| 154 |
|
| } |
| 155 |
|
| |
| 156 |
|
| // called with len including the trailing \n or \f. |
| 157 |
|
| // returns the number of char written to outp |
| 158 |
|
| // it never writes more than len chars to out. |
| 159 |
0 |
| static int cook_text(char *outp, const char *text, int len) |
| 160 |
|
| { |
| 161 |
0 |
| char *p = outp; |
| 162 |
0 |
| char *s = p; |
| 163 |
0 |
| unsigned char lc = ' '; // last char we sent to output |
| 164 |
0 |
| int line_start = 1; // we are at start of a line. |
| 165 |
0 |
| int whitespace = 0; // we saw whitespace chars since lc. |
| 166 |
|
| |
| 167 |
0 |
▶ | while (len-- > 0) |
Branch: 1 branch
| Branch | True | False | Total |
| while (len-- > 0) |
0 |
0 |
0 |
|
| 168 |
|
| { |
| 169 |
0 |
| unsigned char c = *text++; |
| 170 |
0 |
▶ | if (!c) continue; // completly ignore \0 chars |
Branch: 1 branch
| Branch | True | False | Total |
| if (!c) continue; // completly ignore \0 chars |
0 |
0 |
0 |
|
| 171 |
|
| |
| 172 |
0 |
▶ | if (line_start && (c == '+' || c == '-' || c == '<' || c == '>')) |
Branch: 5 branches
| Branch | True | False | Total |
| if (line_start && (c == '+' || c == '-' || c == '<' || c == '>')) |
0 |
0 |
0 |
| if (line_start && (c == '+' || c == '-' || c == '<' || c == '>')) |
0 |
0 |
0 |
| if (line_start && (c == '+' || c == '-' || c == '<' || c == '>')) |
0 |
0 |
0 |
| if (line_start && (c == '+' || c == '-' || c == '<' || c == '>')) |
0 |
0 |
0 |
| if (line_start && (c == '+' || c == '-' || c == '<' || c == '>')) |
0 |
0 |
0 |
|
| 173 |
|
| { |
| 174 |
0 |
| line_start = 0; |
| 175 |
0 |
| continue; // ignore typ. diff chars at line start. |
| 176 |
|
| } |
| 177 |
0 |
| line_start = 0; |
| 178 |
|
| |
| 179 |
0 |
▶ | if (c <= ' ') // (Whitespace: all character codes <= 32) |
Branch: 1 branch
| Branch | True | False | Total |
| if (c <= ' ') // (Whitespace: all character codes <= 32) |
0 |
0 |
0 |
|
| 180 |
|
| { |
| 181 |
0 |
▶ | if (c == '\n' || c == '\r' || c == '\f' || c == '\v') |
Branch: 4 branches
| Branch | True | False | Total |
| if (c == '\n' || c == '\r' || c == '\f' || c == '\v') |
0 |
0 |
0 |
| if (c == '\n' || c == '\r' || c == '\f' || c == '\v') |
0 |
0 |
0 |
| if (c == '\n' || c == '\r' || c == '\f' || c == '\v') |
0 |
0 |
0 |
| if (c == '\n' || c == '\r' || c == '\f' || c == '\v') |
0 |
0 |
0 |
|
| 182 |
0 |
| line_start = 1; |
| 183 |
0 |
| whitespace = 1; |
| 184 |
0 |
| continue; |
| 185 |
|
| } |
| 186 |
|
| |
| 187 |
|
| // output a space *only* to prevent run together words. |
| 188 |
0 |
▶ | if (whitespace && is_word_ch_q(lc) && is_word_ch_q(c)) *p++ = ' '; |
Branch: 3 branches
| Branch | True | False | Total |
| if (whitespace && is_word_ch_q(lc) && is_word_ch_q(c)) *p++ = ' '; |
0 |
0 |
0 |
| if (whitespace && is_word_ch_q(lc) && is_word_ch_q(c)) *p++ = ' '; |
0 |
0 |
0 |
| if (whitespace && is_word_ch_q(lc) && is_word_ch_q(c)) *p++ = ' '; |
0 |
0 |
0 |
|
| 189 |
0 |
| whitespace = 0; |
| 190 |
|
| |
| 191 |
|
| // this ~-magic does not really fold utf8 to latin1. |
| 192 |
|
| // maybe it helps to map other encodings. |
| 193 |
0 |
▶ | if (c >= 128) |
Branch: 1 branch
| Branch | True | False | Total |
| if (c >= 128) |
0 |
0 |
0 |
|
| 194 |
|
| { |
| 195 |
0 |
▶ | if (lc < 128) *p++ = '~'; |
Branch: 1 branch
| Branch | True | False | Total |
| if (lc < 128) *p++ = '~'; |
0 |
0 |
0 |
|
| 196 |
0 |
| lc = c; |
| 197 |
|
| } |
| 198 |
|
| else |
| 199 |
|
| { |
| 200 |
0 |
| *p++ = lc = c; |
| 201 |
|
| } |
| 202 |
|
| } |
| 203 |
|
| |
| 204 |
0 |
| return p - s; |
| 205 |
|
| } |
| 206 |
|
| |
| 207 |
|
| |
| 208 |
|
| MODULE = Digest::FP56x1xor PACKAGE = Digest::FP56x1xor |
| 209 |
|
| PROTOTYPES: ENABLE |
| 210 |
|
| |
| 211 |
|
| long gen_l(text) |
| 212 |
|
| const char *text |
| 213 |
|
| CODE: |
| 214 |
|
| if (sizeof(uint64_t) != 8) |
| 215 |
|
| croak("sizeof(uint64_t) is %d, need 8. Please try the gen() method\n", (int)sizeof(uint64_t)); |
| 216 |
0 |
▶ | RETVAL = gen_fp(text, strlen(text)); |
Branch: 1 branch
| Branch | True | False | Total |
| RETVAL = gen_fp(text, strlen(text)); |
0 |
0 |
0 |
|
| 217 |
|
| OUTPUT: |
| 218 |
|
| RETVAL |
| 219 |
|
| |
| 220 |
|
| long cat_l(h1, h2) |
| 221 |
|
| long h1 |
| 222 |
|
| long h2 |
| 223 |
|
| CODE: |
| 224 |
0 |
▶ | RETVAL = cat_fp((uint64_t)h1, (uint64_t)h2); |
Branch: 1 branch
| Branch | True | False | Total |
| RETVAL = cat_fp((uint64_t)h1, (uint64_t)h2); |
0 |
0 |
0 |
|
| 225 |
|
| OUTPUT: |
| 226 |
|
| RETVAL |
| 227 |
|
| |
| 228 |
|
| SV *gen(sv_text, ...) |
| 229 |
|
| SV *sv_text |
| 230 |
|
| INIT: |
| 231 |
|
| STRLEN text_len; |
| 232 |
|
| IV start_off; |
| 233 |
|
| IV len; |
| 234 |
|
| const char *text; |
| 235 |
|
| |
| 236 |
3 |
| text = (const char *)SvPV(sv_text, text_len); |
| 237 |
3 |
▶ | if ( items > 1) |
Branch: 1 branch
| Branch | True | False | Total |
| if ( items > 1) |
0 |
3 |
3 |
|
| 238 |
|
| { |
| 239 |
0 |
| start_off = SvIV(ST(1)); |
| 240 |
0 |
▶ | if (start_off > text_len) |
Branch: 1 branch
| Branch | True | False | Total |
| if (start_off > text_len) |
0 |
0 |
0 |
|
| 241 |
0 |
| croak("specified offset=%d is beyond text_len=%d", |
| 242 |
|
| (int)start_off, (int)text_len); |
| 243 |
0 |
| text_len -= start_off; |
| 244 |
|
| } |
| 245 |
|
| else |
| 246 |
3 |
| start_off = 0; |
| 247 |
|
| |
| 248 |
3 |
▶ | if ( items > 2) |
Branch: 1 branch
| Branch | True | False | Total |
| if ( items > 2) |
0 |
3 |
3 |
|
| 249 |
|
| { |
| 250 |
0 |
| len = SvIV(ST(2)); |
| 251 |
0 |
▶ | if (len > text_len) |
Branch: 1 branch
| Branch | True | False | Total |
| if (len > text_len) |
0 |
0 |
0 |
|
| 252 |
0 |
| croak("specified length=%d exceeds text_len=%d", (int)len, (int)text_len); |
| 253 |
0 |
| text_len = len; |
| 254 |
|
| } |
| 255 |
|
| |
| 256 |
|
| CODE: |
| 257 |
3 |
| uint64_t fp = gen_fp(text+start_off, text_len); |
| 258 |
3 |
| RETVAL = newSVpvn(ll2x16(fp), 16); |
| 259 |
|
| OUTPUT: |
| 260 |
|
| RETVAL |
| 261 |
|
| |
| 262 |
|
| SV *cat(x1, x2) |
| 263 |
|
| const char *x1 |
| 264 |
|
| const char *x2 |
| 265 |
|
| CODE: |
| 266 |
1 |
| uint64_t fp = cat_fp_x(x1, x2); |
| 267 |
1 |
| RETVAL = newSVpvn(ll2x16(fp), 16); |
| 268 |
|
| OUTPUT: |
| 269 |
|
| RETVAL |
| 270 |
|
| |
| 271 |
|
| SV *ll2x(fp) |
| 272 |
|
| long fp; |
| 273 |
|
| CODE: |
| 274 |
0 |
| RETVAL = newSVpvn(ll2x16(fp), 16); |
| 275 |
|
| OUTPUT: |
| 276 |
|
| RETVAL |
| 277 |
|
| |
| 278 |
|
| long x2l(x) |
| 279 |
|
| const char *x |
| 280 |
|
| CODE: |
| 281 |
0 |
▶ | RETVAL = x2i64((unsigned char *)x, strlen(x)); |
Branch: 1 branch
| Branch | True | False | Total |
| RETVAL = x2i64((unsigned char *)x, strlen(x)); |
0 |
0 |
0 |
|
| 282 |
|
| OUTPUT: |
| 283 |
|
| RETVAL |
| 284 |
|
| |
| 285 |
|
| SV* |
| 286 |
|
| cooked(sv_text, ...) |
| 287 |
|
| SV *sv_text |
| 288 |
|
| INIT: |
| 289 |
|
| STRLEN text_len; // like strlen(text), but including '\0' bytes. STRLEN is unsigned! |
| 290 |
|
| IV start_off; |
| 291 |
|
| IV len; |
| 292 |
|
| const char *text; |
| 293 |
|
| char *cooked_text; |
| 294 |
|
| int cooked_len; |
| 295 |
|
| |
| 296 |
|
| SV *text_out; |
| 297 |
|
| |
| 298 |
0 |
| text = (const char *)SvPV(sv_text, text_len); |
| 299 |
0 |
▶ | if ( items > 1) |
Branch: 1 branch
| Branch | True | False | Total |
| if ( items > 1) |
0 |
0 |
0 |
|
| 300 |
|
| { |
| 301 |
0 |
| start_off = SvIV(ST(1)); |
| 302 |
0 |
▶ | if (start_off > text_len) |
Branch: 1 branch
| Branch | True | False | Total |
| if (start_off > text_len) |
0 |
0 |
0 |
|
| 303 |
0 |
| croak("specified offset=%d is beyond text_len=%d", |
| 304 |
|
| (int)start_off, (int)text_len); |
| 305 |
0 |
| text_len -= start_off; |
| 306 |
|
| } |
| 307 |
|
| else |
| 308 |
0 |
| start_off = 0; |
| 309 |
|
| |
| 310 |
0 |
▶ | if ( items > 2) |
Branch: 1 branch
| Branch | True | False | Total |
| if ( items > 2) |
0 |
0 |
0 |
|
| 311 |
|
| { |
| 312 |
0 |
| len = SvIV(ST(2)); |
| 313 |
0 |
▶ | if (len > text_len) |
Branch: 1 branch
| Branch | True | False | Total |
| if (len > text_len) |
0 |
0 |
0 |
|
| 314 |
0 |
| croak("specified length=%d exceeds text_len=%d", (int)len, (int)text_len); |
| 315 |
0 |
| text_len = len; |
| 316 |
|
| } |
| 317 |
|
| |
| 318 |
|
| CODE: |
| 319 |
0 |
| cooked_text = (char *)malloc(text_len); // never longer than raw text; |
| 320 |
0 |
| cooked_len = cook_text(cooked_text, text+start_off, text_len); |
| 321 |
|
| |
| 322 |
|
| // transfer output from C to perl |
| 323 |
0 |
| text_out = newSVpvn(cooked_text, cooked_len); |
| 324 |
|
| |
| 325 |
|
| // forget C |
| 326 |
0 |
| free((char *)cooked_text); |
| 327 |
|
| // do not free text, it points into perl proper. |
| 328 |
|
| // (Would cause an abort in PerlIOBuf_close) |
| 329 |
|
| |
| 330 |
0 |
| RETVAL = text_out; |
| 331 |
|
| OUTPUT: |
| 332 |
|
| RETVAL |
| 333 |
|
| |