2 * LibXDiff by Davide Libenzi ( File Differential Library )
3 * Copyright (C) 2003 Davide Libenzi
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.
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.
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
19 * Davide Libenzi <davidel@xmailserver.org>
30 long xdl_bogosqrt(long n) {
34 * Classical integer square root approximation using shifts.
36 for (i = 1; n > 0; n >>= 2)
43 int xdl_emit_diffrec(char const *rec, long size, char const *pre, long psize,
48 mb[0].ptr = (char *) pre;
50 mb[1].ptr = (char *) rec;
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);
57 if (ecb->outf(ecb->priv, mb, i) < 0) {
65 void *xdl_mmfile_first(mmfile_t *mmf, long *size)
72 long xdl_mmfile_size(mmfile_t *mmf)
78 int xdl_cha_init(chastore_t *cha, long isize, long icount) {
80 cha->head = cha->tail = NULL;
82 cha->nsize = icount * isize;
83 cha->ancur = cha->sncur = NULL;
90 void xdl_cha_free(chastore_t *cha) {
93 for (cur = cha->head; (tmp = cur) != NULL;) {
100 void *xdl_cha_alloc(chastore_t *cha) {
104 if (!(ancur = cha->ancur) || ancur->icurr == cha->nsize) {
105 if (!(ancur = (chanode_t *) xdl_malloc(sizeof(chanode_t) + cha->nsize))) {
112 cha->tail->next = ancur;
119 data = (char *) ancur + sizeof(chanode_t) + ancur->icurr;
120 ancur->icurr += cha->isize;
126 void *xdl_cha_first(chastore_t *cha) {
129 if (!(cha->sncur = sncur = cha->head))
134 return (char *) sncur + sizeof(chanode_t) + cha->scurr;
138 void *xdl_cha_next(chastore_t *cha) {
141 if (!(sncur = cha->sncur))
143 cha->scurr += cha->isize;
144 if (cha->scurr == sncur->icurr) {
145 if (!(sncur = cha->sncur = sncur->next))
150 return (char *) sncur + sizeof(chanode_t) + cha->scurr;
154 long xdl_guess_lines(mmfile_t *mf, long sample) {
155 long nl = 0, size, tsize = 0;
156 char const *data, *cur, *top;
158 if ((cur = data = xdl_mmfile_first(mf, &size)) != NULL) {
159 for (top = data + size; nl < sample && cur < top; ) {
161 if (!(cur = memchr(cur, '\n', top - cur)))
166 tsize += (long) (cur - data);
170 nl = xdl_mmfile_size(mf) / (tsize / nl);
175 int xdl_recmatch(const char *l1, long s1, const char *l2, long s2, long flags)
179 if (s1 == s2 && !memcmp(l1, l2, s1))
181 if (!(flags & XDF_WHITESPACE_FLAGS))
188 * -w matches everything that matches with -b, and -b in turn
189 * matches everything that matches with --ignore-space-at-eol.
191 * Each flavor of ignoring needs different logic to skip whitespaces
192 * while we have both sides to compare.
194 if (flags & XDF_IGNORE_WHITESPACE) {
196 while (i1 < s1 && i2 < s2) {
197 if (l1[i1++] != l2[i2++])
200 while (i1 < s1 && XDL_ISSPACE(l1[i1]))
202 while (i2 < s2 && XDL_ISSPACE(l2[i2]))
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]))
211 while (i2 < s2 && XDL_ISSPACE(l2[i2]))
215 if (l1[i1++] != l2[i2++])
218 } else if (flags & XDF_IGNORE_WHITESPACE_AT_EOL) {
219 while (i1 < s1 && i2 < s2 && l1[i1++] == l2[i2++])
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.
230 while (i1 < s1 && XDL_ISSPACE(l1[i1]))
236 while (i2 < s2 && XDL_ISSPACE(l2[i2]))
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;
248 for (; ptr < top && *ptr != '\n'; ptr++) {
249 if (XDL_ISSPACE(*ptr)) {
250 const char *ptr2 = ptr;
252 while (ptr + 1 < top && XDL_ISSPACE(ptr[1])
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
261 ha ^= (unsigned long) ' ';
263 else if (flags & XDF_IGNORE_WHITESPACE_AT_EOL
265 while (ptr2 != ptr + 1) {
267 ha ^= (unsigned long) *ptr2;
274 ha ^= (unsigned long) *ptr;
276 *data = ptr < top ? ptr + 1: ptr;
283 #define REPEAT_BYTE(x) ((~0ul / 0xff) * (x))
285 #define ONEBYTES REPEAT_BYTE(0x01)
286 #define NEWLINEBYTES REPEAT_BYTE(0x0a)
287 #define HIGHBITS REPEAT_BYTE(0x80)
289 /* Return the high bit set in the first byte that is a zero */
290 static inline unsigned long has_zero(unsigned long a)
292 return ((a - ONEBYTES) & ~a) & HIGHBITS;
295 static inline long count_masked_bytes(unsigned long mask)
297 if (sizeof(long) == 8) {
299 * Jan Achrenius on G+: microoptimized version of
300 * the simpler "(mask & ONEBYTES) * ONEBYTES >> 56"
301 * that works for the bytemasks without having to
304 return mask * 0x0001020304050608 >> 56;
307 * Modified Carl Chatfield G+ version for 32-bit *
310 * -1 (0, ff), 0 (ffff) or 1 (ffffff)
312 * 0 for 0, 1 for (ff ffff ffffff)
314 * correct 0-3 bytemask count result
316 long a = (mask - 256) >> 23;
322 unsigned long xdl_hash_record(char const **data, char const *top, long flags)
324 unsigned long hash = 5381;
325 unsigned long a = 0, mask = 0;
326 char const *ptr = *data;
327 char const *end = top - sizeof(unsigned long) + 1;
329 if (flags & XDF_WHITESPACE_FLAGS)
330 return xdl_hash_record_with_whitespace(data, top, flags);
332 ptr -= sizeof(unsigned long);
336 ptr += sizeof(unsigned long);
339 a = *(unsigned long *)ptr;
340 /* Do we have any '\n' bytes in this word? */
341 mask = has_zero(a ^ NEWLINEBYTES);
346 * There is only a partial word left at the end of the
347 * buffer. Because we may work with a memory mapping,
348 * we have to grab the rest byte by byte instead of
349 * blindly reading it.
351 * To avoid problems with masking in a signed value,
352 * we use an unsigned char here.
355 for (p = top - 1; p >= ptr; p--)
356 a = (a << 8) + *((const unsigned char *)p);
357 mask = has_zero(a ^ NEWLINEBYTES);
360 * No '\n' found in the partial word. Make a
361 * mask that matches what we read.
363 mask = 1UL << (8 * (top - ptr) + 7);
366 /* The mask *below* the first high bit set */
367 mask = (mask - 1) & ~mask;
372 /* Advance past the last (possibly partial) word */
373 ptr += count_masked_bytes(mask);
376 assert(*ptr == '\n');
385 #else /* XDL_FAST_HASH */
387 unsigned long xdl_hash_record(char const **data, char const *top, long flags) {
388 unsigned long ha = 5381;
389 char const *ptr = *data;
391 if (flags & XDF_WHITESPACE_FLAGS)
392 return xdl_hash_record_with_whitespace(data, top, flags);
394 for (; ptr < top && *ptr != '\n'; ptr++) {
396 ha ^= (unsigned long) *ptr;
398 *data = ptr < top ? ptr + 1: ptr;
403 #endif /* XDL_FAST_HASH */
405 unsigned int xdl_hashbits(unsigned int size) {
406 unsigned int val = 1, bits = 0;
408 for (; val < size && bits < CHAR_BIT * sizeof(unsigned int); val <<= 1, bits++);
409 return bits ? bits: 1;
413 int xdl_num_out(char *out, long val) {
414 char *ptr, *str = out;
417 ptr = buf + sizeof(buf) - 1;
423 for (; val && ptr > buf; val /= 10)
424 *--ptr = "0123456789"[val % 10];
426 for (; *ptr; ptr++, str++)
436 long xdl_atol(char const *str, char const **next) {
440 for (top = str; XDL_ISDIGIT(*top); top++);
443 for (val = 0, base = 1, top--; top >= str; top--, base *= 10)
444 val += base * (long)(*top - '0');
449 int xdl_emit_hunk_hdr(long s1, long c1, long s2, long c2,
450 const char *func, long funclen, xdemitcb_t *ecb) {
455 memcpy(buf, "@@ -", 4);
458 nb += xdl_num_out(buf + nb, c1 ? s1: s1 - 1);
461 memcpy(buf + nb, ",", 1);
464 nb += xdl_num_out(buf + nb, c1);
467 memcpy(buf + nb, " +", 2);
470 nb += xdl_num_out(buf + nb, c2 ? s2: s2 - 1);
473 memcpy(buf + nb, ",", 1);
476 nb += xdl_num_out(buf + nb, c2);
479 memcpy(buf + nb, " @@", 3);
481 if (func && funclen) {
483 if (funclen > sizeof(buf) - nb - 1)
484 funclen = sizeof(buf) - nb - 1;
485 memcpy(buf + nb, func, funclen);
492 if (ecb->outf(ecb->priv, &mb, 1) < 0)
498 int xdl_fall_back_diff(xdfenv_t *diff_env, xpparam_t const *xpp,
499 int line1, int count1, int line2, int count2)
502 * This probably does not work outside Git, since
503 * we have a very simple mmfile structure.
505 * Note: ideally, we would reuse the prepared environment, but
506 * the libxdiff interface does not (yet) allow for diffing only
507 * ranges of lines instead of the whole files.
509 mmfile_t subfile1, subfile2;
512 subfile1.ptr = (char *)diff_env->xdf1.recs[line1 - 1]->ptr;
513 subfile1.size = diff_env->xdf1.recs[line1 + count1 - 2]->ptr +
514 diff_env->xdf1.recs[line1 + count1 - 2]->size - subfile1.ptr;
515 subfile2.ptr = (char *)diff_env->xdf2.recs[line2 - 1]->ptr;
516 subfile2.size = diff_env->xdf2.recs[line2 + count2 - 2]->ptr +
517 diff_env->xdf2.recs[line2 + count2 - 2]->size - subfile2.ptr;
518 if (xdl_do_diff(&subfile1, &subfile2, xpp, &env) < 0)
521 memcpy(diff_env->xdf1.rchg + line1 - 1, env.xdf1.rchg, count1);
522 memcpy(diff_env->xdf2.rchg + line2 - 1, env.xdf2.rchg, count2);