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