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 long xdl_guess_lines(mmfile_t *mf, long sample) {
 126         long nl = 0, size, tsize = 0;
 127         char const *data, *cur, *top;
 128
 129         if ((cur = data = xdl_mmfile_first(mf, &size)) != NULL) {
 130                 for (top = data + size; nl < sample && cur < top; ) {
 131                         nl++;
 132                         if (!(cur = memchr(cur, '\n', top - cur)))
 133                                 cur = top;
 134                         else
 135                                 cur++;
 136                 }
 137                 tsize += (long) (cur - data);
 138         }
 139
 140         if (nl && tsize)
 141                 nl = xdl_mmfile_size(mf) / (tsize / nl);
 142
 143         return nl + 1;
 144 }
 145
 146 int xdl_blankline(const char *line, long size, long flags)
 147 {
 148         long i;
 149
 150         if (!(flags & XDF_WHITESPACE_FLAGS))
 151                 return (size <= 1);
 152
 153         for (i = 0; i < size && XDL_ISSPACE(line[i]); i++)
 154                 ;
 155
 156         return (i == size);
 157 }
 158
 159 int xdl_recmatch(const char *l1, long s1, const char *l2, long s2, long flags)
 160 {
 161         int i1, i2;
 162
 163         if (s1 == s2 && !memcmp(l1, l2, s1))
 164                 return 1;
 165         if (!(flags & XDF_WHITESPACE_FLAGS))
 166                 return 0;
 167
 168         i1 = 0;
 169         i2 = 0;
 170
 171         /*
 172          * -w matches everything that matches with -b, and -b in turn
 173          * matches everything that matches with --ignore-space-at-eol.
 174          *
 175          * Each flavor of ignoring needs different logic to skip whitespaces
 176          * while we have both sides to compare.
 177          */
 178         if (flags & XDF_IGNORE_WHITESPACE) {
 179                 goto skip_ws;
 180                 while (i1 < s1 && i2 < s2) {
 181                         if (l1[i1++] != l2[i2++])
 182                                 return 0;
 183                 skip_ws:
 184                         while (i1 < s1 && XDL_ISSPACE(l1[i1]))
 185                                 i1++;
 186                         while (i2 < s2 && XDL_ISSPACE(l2[i2]))
 187                                 i2++;
 188                 }
 189         } else if (flags & XDF_IGNORE_WHITESPACE_CHANGE) {
 190                 while (i1 < s1 && i2 < s2) {
 191                         if (XDL_ISSPACE(l1[i1]) && XDL_ISSPACE(l2[i2])) {
 192                                 /* Skip matching spaces and try again */
 193                                 while (i1 < s1 && XDL_ISSPACE(l1[i1]))
 194                                         i1++;
 195                                 while (i2 < s2 && XDL_ISSPACE(l2[i2]))
 196                                         i2++;
 197                                 continue;
 198                         }
 199                         if (l1[i1++] != l2[i2++])
 200                                 return 0;
 201                 }
 202         } else if (flags & XDF_IGNORE_WHITESPACE_AT_EOL) {
 203                 while (i1 < s1 && i2 < s2 && l1[i1] == l2[i2]) {
 204                         i1++;
 205                         i2++;
 206                 }
 207         }
 208
 209         /*
 210          * After running out of one side, the remaining side must have
 211          * nothing but whitespace for the lines to match.  Note that
 212          * ignore-whitespace-at-eol case may break out of the loop
 213          * while there still are characters remaining on both lines.
 214          */
 215         if (i1 < s1) {
 216                 while (i1 < s1 && XDL_ISSPACE(l1[i1]))
 217                         i1++;
 218                 if (s1 != i1)
 219                         return 0;
 220         }
 221         if (i2 < s2) {
 222                 while (i2 < s2 && XDL_ISSPACE(l2[i2]))
 223                         i2++;
 224                 return (s2 == i2);
 225         }
 226         return 1;
 227 }
 228
 229 static unsigned long xdl_hash_record_with_whitespace(char const **data,
 230                 char const *top, long flags) {
 231         unsigned long ha = 5381;
 232         char const *ptr = *data;
 233
 234         for (; ptr < top && *ptr != '\n'; ptr++) {
 235                 if (XDL_ISSPACE(*ptr)) {
 236                         const char *ptr2 = ptr;
 237                         int at_eol;
 238                         while (ptr + 1 < top && XDL_ISSPACE(ptr[1])
 239                                         && ptr[1] != '\n')
 240                                 ptr++;
 241                         at_eol = (top <= ptr + 1 || ptr[1] == '\n');
 242                         if (flags & XDF_IGNORE_WHITESPACE)
 243                                 ; /* already handled */
 244                         else if (flags & XDF_IGNORE_WHITESPACE_CHANGE
 245                                  && !at_eol) {
 246                                 ha += (ha << 5);
 247                                 ha ^= (unsigned long) ' ';
 248                         }
 249                         else if (flags & XDF_IGNORE_WHITESPACE_AT_EOL
 250                                  && !at_eol) {
 251                                 while (ptr2 != ptr + 1) {
 252                                         ha += (ha << 5);
 253                                         ha ^= (unsigned long) *ptr2;
 254                                         ptr2++;
 255                                 }
 256                         }
 257                         continue;
 258                 }
 259                 ha += (ha << 5);
 260                 ha ^= (unsigned long) *ptr;
 261         }
 262         *data = ptr < top ? ptr + 1: ptr;
 263
 264         return ha;
 265 }
 266
 267 #ifdef XDL_FAST_HASH
 268
 269 #define REPEAT_BYTE(x)  ((~0ul / 0xff) * (x))
 270
 271 #define ONEBYTES        REPEAT_BYTE(0x01)
 272 #define NEWLINEBYTES    REPEAT_BYTE(0x0a)
 273 #define HIGHBITS        REPEAT_BYTE(0x80)
 274
 275 /* Return the high bit set in the first byte that is a zero */
 276 static inline unsigned long has_zero(unsigned long a)
 277 {
 278         return ((a - ONEBYTES) & ~a) & HIGHBITS;
 279 }
 280
 281 static inline long count_masked_bytes(unsigned long mask)
 282 {
 283         if (sizeof(long) == 8) {
 284                 /*
 285                  * Jan Achrenius on G+: microoptimized version of
 286                  * the simpler "(mask & ONEBYTES) * ONEBYTES >> 56"
 287                  * that works for the bytemasks without having to
 288                  * mask them first.
 289                  */
 290                 /*
 291                  * return mask * 0x0001020304050608 >> 56;
 292                  *
 293                  * Doing it like this avoids warnings on 32-bit machines.
 294                  */
 295                 long a = (REPEAT_BYTE(0x01) / 0xff + 1);
 296                 return mask * a >> (sizeof(long) * 7);
 297         } else {
 298                 /* Carl Chatfield / Jan Achrenius G+ version for 32-bit */
 299                 /* (000000 0000ff 00ffff ffffff) -> ( 1 1 2 3 ) */
 300                 long a = (0x0ff0001 + mask) >> 23;
 301                 /* Fix the 1 for 00 case */
 302                 return a & mask;
 303         }
 304 }
 305
 306 unsigned long xdl_hash_record(char const **data, char const *top, long flags)
 307 {
 308         unsigned long hash = 5381;
 309         unsigned long a = 0, mask = 0;
 310         char const *ptr = *data;
 311         char const *end = top - sizeof(unsigned long) + 1;
 312
 313         if (flags & XDF_WHITESPACE_FLAGS)
 314                 return xdl_hash_record_with_whitespace(data, top, flags);
 315
 316         ptr -= sizeof(unsigned long);
 317         do {
 318                 hash += hash << 5;
 319                 hash ^= a;
 320                 ptr += sizeof(unsigned long);
 321                 if (ptr >= end)
 322                         break;
 323                 a = *(unsigned long *)ptr;
 324                 /* Do we have any '\n' bytes in this word? */
 325                 mask = has_zero(a ^ NEWLINEBYTES);
 326         } while (!mask);
 327
 328         if (ptr >= end) {
 329                 /*
 330                  * There is only a partial word left at the end of the
 331                  * buffer. Because we may work with a memory mapping,
 332                  * we have to grab the rest byte by byte instead of
 333                  * blindly reading it.
 334                  *
 335                  * To avoid problems with masking in a signed value,
 336                  * we use an unsigned char here.
 337                  */
 338                 const char *p;
 339                 for (p = top - 1; p >= ptr; p--)
 340                         a = (a << 8) + *((const unsigned char *)p);
 341                 mask = has_zero(a ^ NEWLINEBYTES);
 342                 if (!mask)
 343                         /*
 344                          * No '\n' found in the partial word.  Make a
 345                          * mask that matches what we read.
 346                          */
 347                         mask = 1UL << (8 * (top - ptr) + 7);
 348         }
 349
 350         /* The mask *below* the first high bit set */
 351         mask = (mask - 1) & ~mask;
 352         mask >>= 7;
 353         hash += hash << 5;
 354         hash ^= a & mask;
 355
 356         /* Advance past the last (possibly partial) word */
 357         ptr += count_masked_bytes(mask);
 358
 359         if (ptr < top) {
 360                 assert(*ptr == '\n');
 361                 ptr++;
 362         }
 363
 364         *data = ptr;
 365
 366         return hash;
 367 }
 368
 369 #else /* XDL_FAST_HASH */
 370
 371 unsigned long xdl_hash_record(char const **data, char const *top, long flags) {
 372         unsigned long ha = 5381;
 373         char const *ptr = *data;
 374
 375         if (flags & XDF_WHITESPACE_FLAGS)
 376                 return xdl_hash_record_with_whitespace(data, top, flags);
 377
 378         for (; ptr < top && *ptr != '\n'; ptr++) {
 379                 ha += (ha << 5);
 380                 ha ^= (unsigned long) *ptr;
 381         }
 382         *data = ptr < top ? ptr + 1: ptr;
 383
 384         return ha;
 385 }
 386
 387 #endif /* XDL_FAST_HASH */
 388
 389 unsigned int xdl_hashbits(unsigned int size) {
 390         unsigned int val = 1, bits = 0;
 391
 392         for (; val < size && bits < CHAR_BIT * sizeof(unsigned int); val <<= 1, bits++);
 393         return bits ? bits: 1;
 394 }
 395
 396
 397 int xdl_num_out(char *out, long val) {
 398         char *ptr, *str = out;
 399         char buf[32];
 400
 401         ptr = buf + sizeof(buf) - 1;
 402         *ptr = '\0';
 403         if (val < 0) {
 404                 *--ptr = '-';
 405                 val = -val;
 406         }
 407         for (; val && ptr > buf; val /= 10)
 408                 *--ptr = "0123456789"[val % 10];
 409         if (*ptr)
 410                 for (; *ptr; ptr++, str++)
 411                         *str = *ptr;
 412         else
 413                 *str++ = '0';
 414         *str = '\0';
 415
 416         return str - out;
 417 }
 418
 419 int xdl_emit_hunk_hdr(long s1, long c1, long s2, long c2,
 420                       const char *func, long funclen, xdemitcb_t *ecb) {
 421         int nb = 0;
 422         mmbuffer_t mb;
 423         char buf[128];
 424
 425         memcpy(buf, "@@ -", 4);
 426         nb += 4;
 427
 428         nb += xdl_num_out(buf + nb, c1 ? s1: s1 - 1);
 429
 430         if (c1 != 1) {
 431                 memcpy(buf + nb, ",", 1);
 432                 nb += 1;
 433
 434                 nb += xdl_num_out(buf + nb, c1);
 435         }
 436
 437         memcpy(buf + nb, " +", 2);
 438         nb += 2;
 439
 440         nb += xdl_num_out(buf + nb, c2 ? s2: s2 - 1);
 441
 442         if (c2 != 1) {
 443                 memcpy(buf + nb, ",", 1);
 444                 nb += 1;
 445
 446                 nb += xdl_num_out(buf + nb, c2);
 447         }
 448
 449         memcpy(buf + nb, " @@", 3);
 450         nb += 3;
 451         if (func && funclen) {
 452                 buf[nb++] = ' ';
 453                 if (funclen > sizeof(buf) - nb - 1)
 454                         funclen = sizeof(buf) - nb - 1;
 455                 memcpy(buf + nb, func, funclen);
 456                 nb += funclen;
 457         }
 458         buf[nb++] = '\n';
 459
 460         mb.ptr = buf;
 461         mb.size = nb;
 462         if (ecb->outf(ecb->priv, &mb, 1) < 0)
 463                 return -1;
 464
 465         return 0;
 466 }
 467
 468 int xdl_fall_back_diff(xdfenv_t *diff_env, xpparam_t const *xpp,
 469                 int line1, int count1, int line2, int count2)
 470 {
 471         /*
 472          * This probably does not work outside Git, since
 473          * we have a very simple mmfile structure.
 474          *
 475          * Note: ideally, we would reuse the prepared environment, but
 476          * the libxdiff interface does not (yet) allow for diffing only
 477          * ranges of lines instead of the whole files.
 478          */
 479         mmfile_t subfile1, subfile2;
 480         xdfenv_t env;
 481
 482         subfile1.ptr = (char *)diff_env->xdf1.recs[line1 - 1]->ptr;
 483         subfile1.size = diff_env->xdf1.recs[line1 + count1 - 2]->ptr +
 484                 diff_env->xdf1.recs[line1 + count1 - 2]->size - subfile1.ptr;
 485         subfile2.ptr = (char *)diff_env->xdf2.recs[line2 - 1]->ptr;
 486         subfile2.size = diff_env->xdf2.recs[line2 + count2 - 2]->ptr +
 487                 diff_env->xdf2.recs[line2 + count2 - 2]->size - subfile2.ptr;
 488         if (xdl_do_diff(&subfile1, &subfile2, xpp, &env) < 0)
 489                 return -1;
 490
 491         memcpy(diff_env->xdf1.rchg + line1 - 1, env.xdf1.rchg, count1);
 492         memcpy(diff_env->xdf2.rchg + line2 - 1, env.xdf2.rchg, count2);
 493
 494         xdl_free_env(&env);
 495
 496         return 0;
 497 }