xdiff/xutils.c

   1 /*
   2  *  LibXDiff by Davide Libenzi ( File Differential Library )
   3  *  Copyright (C) 2003  Davide Libenzi
   4  *
   5  *  This library is free software; you can redistribute it and/or
   6  *  modify it under the terms of the GNU Lesser General Public
   7  *  License as published by the Free Software Foundation; either
   8  *  version 2.1 of the License, or (at your option) any later version.
   9  *
  10  *  This library is distributed in the hope that it will be useful,
  11  *  but WITHOUT ANY WARRANTY; without even the implied warranty of
  12  *  MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  13  *  Lesser General Public License for more details.
  14  *
  15  *  You should have received a copy of the GNU Lesser General Public
  16  *  License along with this library; if not, write to the Free Software
  17  *  Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  18  *
  19  *  Davide Libenzi <davidel@xmailserver.org>
  20  *
  21  */
  22
  23 #include <limits.h>
  24 #include <assert.h>
  25 #include "xinclude.h"
  26
  27
  28
  29
  30 long xdl_bogosqrt(long n) {
  31         long i;
  32
  33         /*
  34          * Classical integer square root approximation using shifts.
  35          */
  36         for (i = 1; n > 0; n >>= 2)
  37                 i <<= 1;
  38
  39         return i;
  40 }
  41
  42
  43 int xdl_emit_diffrec(char const *rec, long size, char const *pre, long psize,
  44                      xdemitcb_t *ecb) {
  45         int i = 2;
  46         mmbuffer_t mb[3];
  47
  48         mb[0].ptr = (char *) pre;
  49         mb[0].size = psize;
  50         mb[1].ptr = (char *) rec;
  51         mb[1].size = size;
  52         if (size > 0 && rec[size - 1] != '\n') {
  53                 mb[2].ptr = (char *) "\n\\ No newline at end of file\n";
  54                 mb[2].size = strlen(mb[2].ptr);
  55                 i++;
  56         }
  57         if (ecb->outf(ecb->priv, mb, i) < 0) {
  58
  59                 return -1;
  60         }
  61
  62         return 0;
  63 }
  64
  65 void *xdl_mmfile_first(mmfile_t *mmf, long *size)
  66 {
  67         *size = mmf->size;
  68         return mmf->ptr;
  69 }
  70
  71
  72 long xdl_mmfile_size(mmfile_t *mmf)
  73 {
  74         return mmf->size;
  75 }
  76
  77
  78 int xdl_cha_init(chastore_t *cha, long isize, long icount) {
  79
  80         cha->head = cha->tail = NULL;
  81         cha->isize = isize;
  82         cha->nsize = icount * isize;
  83         cha->ancur = cha->sncur = NULL;
  84         cha->scurr = 0;
  85
  86         return 0;
  87 }
  88
  89
  90 void xdl_cha_free(chastore_t *cha) {
  91         chanode_t *cur, *tmp;
  92
  93         for (cur = cha->head; (tmp = cur) != NULL;) {
  94                 cur = cur->next;
  95                 xdl_free(tmp);
  96         }
  97 }
  98
  99
 100 void *xdl_cha_alloc(chastore_t *cha) {
 101         chanode_t *ancur;
 102         void *data;
 103
 104         if (!(ancur = cha->ancur) || ancur->icurr == cha->nsize) {
 105                 if (!(ancur = (chanode_t *) xdl_malloc(sizeof(chanode_t) + cha->nsize))) {
 106
 107                         return NULL;
 108                 }
 109                 ancur->icurr = 0;
 110                 ancur->next = NULL;
 111                 if (cha->tail)
 112                         cha->tail->next = ancur;
 113                 if (!cha->head)
 114                         cha->head = ancur;
 115                 cha->tail = ancur;
 116                 cha->ancur = ancur;
 117         }
 118
 119         data = (char *) ancur + sizeof(chanode_t) + ancur->icurr;
 120         ancur->icurr += cha->isize;
 121
 122         return data;
 123 }
 124
 125
 126 void *xdl_cha_first(chastore_t *cha) {
 127         chanode_t *sncur;
 128
 129         if (!(cha->sncur = sncur = cha->head))
 130                 return NULL;
 131
 132         cha->scurr = 0;
 133
 134         return (char *) sncur + sizeof(chanode_t) + cha->scurr;
 135 }
 136
 137
 138 void *xdl_cha_next(chastore_t *cha) {
 139         chanode_t *sncur;
 140
 141         if (!(sncur = cha->sncur))
 142                 return NULL;
 143         cha->scurr += cha->isize;
 144         if (cha->scurr == sncur->icurr) {
 145                 if (!(sncur = cha->sncur = sncur->next))
 146                         return NULL;
 147                 cha->scurr = 0;
 148         }
 149
 150         return (char *) sncur + sizeof(chanode_t) + cha->scurr;
 151 }
 152
 153
 154 long xdl_guess_lines(mmfile_t *mf, long sample) {
 155         long nl = 0, size, tsize = 0;
 156         char const *data, *cur, *top;
 157
 158         if ((cur = data = xdl_mmfile_first(mf, &size)) != NULL) {
 159                 for (top = data + size; nl < sample && cur < top; ) {
 160                         nl++;
 161                         if (!(cur = memchr(cur, '\n', top - cur)))
 162                                 cur = top;
 163                         else
 164                                 cur++;
 165                 }
 166                 tsize += (long) (cur - data);
 167         }
 168
 169         if (nl && tsize)
 170                 nl = xdl_mmfile_size(mf) / (tsize / nl);
 171
 172         return nl + 1;
 173 }
 174
 175 int xdl_recmatch(const char *l1, long s1, const char *l2, long s2, long flags)
 176 {
 177         int i1, i2;
 178
 179         if (s1 == s2 && !memcmp(l1, l2, s1))
 180                 return 1;
 181         if (!(flags & XDF_WHITESPACE_FLAGS))
 182                 return 0;
 183
 184         i1 = 0;
 185         i2 = 0;
 186
 187         /*
 188          * -w matches everything that matches with -b, and -b in turn
 189          * matches everything that matches with --ignore-space-at-eol.
 190          *
 191          * Each flavor of ignoring needs different logic to skip whitespaces
 192          * while we have both sides to compare.
 193          */
 194         if (flags & XDF_IGNORE_WHITESPACE) {
 195                 goto skip_ws;
 196                 while (i1 < s1 && i2 < s2) {
 197                         if (l1[i1++] != l2[i2++])
 198                                 return 0;
 199                 skip_ws:
 200                         while (i1 < s1 && XDL_ISSPACE(l1[i1]))
 201                                 i1++;
 202                         while (i2 < s2 && XDL_ISSPACE(l2[i2]))
 203                                 i2++;
 204                 }
 205         } else if (flags & XDF_IGNORE_WHITESPACE_CHANGE) {
 206                 while (i1 < s1 && i2 < s2) {
 207                         if (XDL_ISSPACE(l1[i1]) && XDL_ISSPACE(l2[i2])) {
 208                                 /* Skip matching spaces and try again */
 209                                 while (i1 < s1 && XDL_ISSPACE(l1[i1]))
 210                                         i1++;
 211                                 while (i2 < s2 && XDL_ISSPACE(l2[i2]))
 212                                         i2++;
 213                                 continue;
 214                         }
 215                         if (l1[i1++] != l2[i2++])
 216                                 return 0;
 217                 }
 218         } else if (flags & XDF_IGNORE_WHITESPACE_AT_EOL) {
 219                 while (i1 < s1 && i2 < s2 && l1[i1++] == l2[i2++])
 220                         ; /* keep going */
 221         }
 222
 223         /*
 224          * After running out of one side, the remaining side must have
 225          * nothing but whitespace for the lines to match.  Note that
 226          * ignore-whitespace-at-eol case may break out of the loop
 227          * while there still are characters remaining on both lines.
 228          */
 229         if (i1 < s1) {
 230                 while (i1 < s1 && XDL_ISSPACE(l1[i1]))
 231                         i1++;
 232                 if (s1 != i1)
 233                         return 0;
 234         }
 235         if (i2 < s2) {
 236                 while (i2 < s2 && XDL_ISSPACE(l2[i2]))
 237                         i2++;
 238                 return (s2 == i2);
 239         }
 240         return 1;
 241 }
 242
 243 static unsigned long xdl_hash_record_with_whitespace(char const **data,
 244                 char const *top, long flags) {
 245         unsigned long ha = 5381;
 246         char const *ptr = *data;
 247
 248         for (; ptr < top && *ptr != '\n'; ptr++) {
 249                 if (XDL_ISSPACE(*ptr)) {
 250                         const char *ptr2 = ptr;
 251                         int at_eol;
 252                         while (ptr + 1 < top && XDL_ISSPACE(ptr[1])
 253                                         && ptr[1] != '\n')
 254                                 ptr++;
 255                         at_eol = (top <= ptr + 1 || ptr[1] == '\n');
 256                         if (flags & XDF_IGNORE_WHITESPACE)
 257                                 ; /* already handled */
 258                         else if (flags & XDF_IGNORE_WHITESPACE_CHANGE
 259                                  && !at_eol) {
 260                                 ha += (ha << 5);
 261                                 ha ^= (unsigned long) ' ';
 262                         }
 263                         else if (flags & XDF_IGNORE_WHITESPACE_AT_EOL
 264                                  && !at_eol) {
 265                                 while (ptr2 != ptr + 1) {
 266                                         ha += (ha << 5);
 267                                         ha ^= (unsigned long) *ptr2;
 268                                         ptr2++;
 269                                 }
 270                         }
 271                         continue;
 272                 }
 273                 ha += (ha << 5);
 274                 ha ^= (unsigned long) *ptr;
 275         }
 276         *data = ptr < top ? ptr + 1: ptr;
 277
 278         return ha;
 279 }
 280
 281 #ifdef XDL_FAST_HASH
 282
 283 #define ONEBYTES        0x0101010101010101ul
 284 #define NEWLINEBYTES    0x0a0a0a0a0a0a0a0aul
 285 #define HIGHBITS        0x8080808080808080ul
 286
 287 /* Return the high bit set in the first byte that is a zero */
 288 static inline unsigned long has_zero(unsigned long a)
 289 {
 290         return ((a - ONEBYTES) & ~a) & HIGHBITS;
 291 }
 292
 293 static inline long count_masked_bytes(unsigned long mask)
 294 {
 295         if (sizeof(long) == 8) {
 296                 /*
 297                  * Jan Achrenius on G+: microoptimized version of
 298                  * the simpler "(mask & ONEBYTES) * ONEBYTES >> 56"
 299                  * that works for the bytemasks without having to
 300                  * mask them first.
 301                  */
 302                 return mask * 0x0001020304050608 >> 56;
 303         } else {
 304                 /*
 305                  * Modified Carl Chatfield G+ version for 32-bit *
 306                  *
 307                  * (a) gives us
 308                  *   -1 (0, ff), 0 (ffff) or 1 (ffffff)
 309                  * (b) gives us
 310                  *   0 for 0, 1 for (ff ffff ffffff)
 311                  * (a+b+1) gives us
 312                  *   correct 0-3 bytemask count result
 313                  */
 314                 long a = (mask - 256) >> 23;
 315                 long b = mask & 1;
 316                 return a + b + 1;
 317         }
 318 }
 319
 320 unsigned long xdl_hash_record(char const **data, char const *top, long flags)
 321 {
 322         unsigned long hash = 5381;
 323         unsigned long a = 0, mask = 0;
 324         char const *ptr = *data;
 325         char const *end = top - sizeof(unsigned long) + 1;
 326
 327         if (flags & XDF_WHITESPACE_FLAGS)
 328                 return xdl_hash_record_with_whitespace(data, top, flags);
 329
 330         ptr -= sizeof(unsigned long);
 331         do {
 332                 hash += hash << 5;
 333                 hash ^= a;
 334                 ptr += sizeof(unsigned long);
 335                 if (ptr >= end)
 336                         break;
 337                 a = *(unsigned long *)ptr;
 338                 /* Do we have any '\n' bytes in this word? */
 339                 mask = has_zero(a ^ NEWLINEBYTES);
 340         } while (!mask);
 341
 342         if (ptr >= end) {
 343                 /*
 344                  * There is only a partial word left at the end of the
 345                  * buffer. Because we may work with a memory mapping,
 346                  * we have to grab the rest byte by byte instead of
 347                  * blindly reading it.
 348                  *
 349                  * To avoid problems with masking in a signed value,
 350                  * we use an unsigned char here.
 351                  */
 352                 const char *p;
 353                 for (p = top - 1; p >= ptr; p--)
 354                         a = (a << 8) + *((const unsigned char *)p);
 355                 mask = has_zero(a ^ NEWLINEBYTES);
 356                 if (!mask)
 357                         /*
 358                          * No '\n' found in the partial word.  Make a
 359                          * mask that matches what we read.
 360                          */
 361                         mask = 1UL << (8 * (top - ptr) + 7);
 362         }
 363
 364         /* The mask *below* the first high bit set */
 365         mask = (mask - 1) & ~mask;
 366         mask >>= 7;
 367         hash += hash << 5;
 368         hash ^= a & mask;
 369
 370         /* Advance past the last (possibly partial) word */
 371         ptr += count_masked_bytes(mask);
 372
 373         if (ptr < top) {
 374                 assert(*ptr == '\n');
 375                 ptr++;
 376         }
 377
 378         *data = ptr;
 379
 380         return hash;
 381 }
 382
 383 #else /* XDL_FAST_HASH */
 384
 385 unsigned long xdl_hash_record(char const **data, char const *top, long flags) {
 386         unsigned long ha = 5381;
 387         char const *ptr = *data;
 388
 389         if (flags & XDF_WHITESPACE_FLAGS)
 390                 return xdl_hash_record_with_whitespace(data, top, flags);
 391
 392         for (; ptr < top && *ptr != '\n'; ptr++) {
 393                 ha += (ha << 5);
 394                 ha ^= (unsigned long) *ptr;
 395         }
 396         *data = ptr < top ? ptr + 1: ptr;
 397
 398         return ha;
 399 }
 400
 401 #endif /* XDL_FAST_HASH */
 402
 403 unsigned int xdl_hashbits(unsigned int size) {
 404         unsigned int val = 1, bits = 0;
 405
 406         for (; val < size && bits < CHAR_BIT * sizeof(unsigned int); val <<= 1, bits++);
 407         return bits ? bits: 1;
 408 }
 409
 410
 411 int xdl_num_out(char *out, long val) {
 412         char *ptr, *str = out;
 413         char buf[32];
 414
 415         ptr = buf + sizeof(buf) - 1;
 416         *ptr = '\0';
 417         if (val < 0) {
 418                 *--ptr = '-';
 419                 val = -val;
 420         }
 421         for (; val && ptr > buf; val /= 10)
 422                 *--ptr = "0123456789"[val % 10];
 423         if (*ptr)
 424                 for (; *ptr; ptr++, str++)
 425                         *str = *ptr;
 426         else
 427                 *str++ = '0';
 428         *str = '\0';
 429
 430         return str - out;
 431 }
 432
 433
 434 long xdl_atol(char const *str, char const **next) {
 435         long val, base;
 436         char const *top;
 437
 438         for (top = str; XDL_ISDIGIT(*top); top++);
 439         if (next)
 440                 *next = top;
 441         for (val = 0, base = 1, top--; top >= str; top--, base *= 10)
 442                 val += base * (long)(*top - '0');
 443         return val;
 444 }
 445
 446
 447 int xdl_emit_hunk_hdr(long s1, long c1, long s2, long c2,
 448                       const char *func, long funclen, xdemitcb_t *ecb) {
 449         int nb = 0;
 450         mmbuffer_t mb;
 451         char buf[128];
 452
 453         memcpy(buf, "@@ -", 4);
 454         nb += 4;
 455
 456         nb += xdl_num_out(buf + nb, c1 ? s1: s1 - 1);
 457
 458         if (c1 != 1) {
 459                 memcpy(buf + nb, ",", 1);
 460                 nb += 1;
 461
 462                 nb += xdl_num_out(buf + nb, c1);
 463         }
 464
 465         memcpy(buf + nb, " +", 2);
 466         nb += 2;
 467
 468         nb += xdl_num_out(buf + nb, c2 ? s2: s2 - 1);
 469
 470         if (c2 != 1) {
 471                 memcpy(buf + nb, ",", 1);
 472                 nb += 1;
 473
 474                 nb += xdl_num_out(buf + nb, c2);
 475         }
 476
 477         memcpy(buf + nb, " @@", 3);
 478         nb += 3;
 479         if (func && funclen) {
 480                 buf[nb++] = ' ';
 481                 if (funclen > sizeof(buf) - nb - 1)
 482                         funclen = sizeof(buf) - nb - 1;
 483                 memcpy(buf + nb, func, funclen);
 484                 nb += funclen;
 485         }
 486         buf[nb++] = '\n';
 487
 488         mb.ptr = buf;
 489         mb.size = nb;
 490         if (ecb->outf(ecb->priv, &mb, 1) < 0)
 491                 return -1;
 492
 493         return 0;
 494 }
 495
 496 int xdl_fall_back_diff(xdfenv_t *diff_env, xpparam_t const *xpp,
 497                 int line1, int count1, int line2, int count2)
 498 {
 499         /*
 500          * This probably does not work outside Git, since
 501          * we have a very simple mmfile structure.
 502          *
 503          * Note: ideally, we would reuse the prepared environment, but
 504          * the libxdiff interface does not (yet) allow for diffing only
 505          * ranges of lines instead of the whole files.
 506          */
 507         mmfile_t subfile1, subfile2;
 508         xdfenv_t env;
 509
 510         subfile1.ptr = (char *)diff_env->xdf1.recs[line1 - 1]->ptr;
 511         subfile1.size = diff_env->xdf1.recs[line1 + count1 - 2]->ptr +
 512                 diff_env->xdf1.recs[line1 + count1 - 2]->size - subfile1.ptr;
 513         subfile2.ptr = (char *)diff_env->xdf2.recs[line2 - 1]->ptr;
 514         subfile2.size = diff_env->xdf2.recs[line2 + count2 - 2]->ptr +
 515                 diff_env->xdf2.recs[line2 + count2 - 2]->size - subfile2.ptr;
 516         if (xdl_do_diff(&subfile1, &subfile2, xpp, &env) < 0)
 517                 return -1;
 518
 519         memcpy(diff_env->xdf1.rchg + line1 - 1, env.xdf1.rchg, count1);
 520         memcpy(diff_env->xdf2.rchg + line2 - 1, env.xdf2.rchg, count2);
 521
 522         xdl_free_env(&env);
 523
 524         return 0;
 525 }