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;
125 long xdl_guess_lines(mmfile_t *mf, long sample) {
126 long nl = 0, size, tsize = 0;
127 char const *data, *cur, *top;
129 if ((cur = data = xdl_mmfile_first(mf, &size)) != NULL) {
130 for (top = data + size; nl < sample && cur < top; ) {
132 if (!(cur = memchr(cur, '\n', top - cur)))
137 tsize += (long) (cur - data);
141 nl = xdl_mmfile_size(mf) / (tsize / nl);
146 int xdl_blankline(const char *line, long size, long flags)
150 if (!(flags & XDF_WHITESPACE_FLAGS))
153 for (i = 0; i < size && XDL_ISSPACE(line[i]); i++)
159 int xdl_recmatch(const char *l1, long s1, const char *l2, long s2, long flags)
163 if (s1 == s2 && !memcmp(l1, l2, s1))
165 if (!(flags & XDF_WHITESPACE_FLAGS))
172 * -w matches everything that matches with -b, and -b in turn
173 * matches everything that matches with --ignore-space-at-eol.
175 * Each flavor of ignoring needs different logic to skip whitespaces
176 * while we have both sides to compare.
178 if (flags & XDF_IGNORE_WHITESPACE) {
180 while (i1 < s1 && i2 < s2) {
181 if (l1[i1++] != l2[i2++])
184 while (i1 < s1 && XDL_ISSPACE(l1[i1]))
186 while (i2 < s2 && XDL_ISSPACE(l2[i2]))
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]))
195 while (i2 < s2 && XDL_ISSPACE(l2[i2]))
199 if (l1[i1++] != l2[i2++])
202 } else if (flags & XDF_IGNORE_WHITESPACE_AT_EOL) {
203 while (i1 < s1 && i2 < s2 && l1[i1] == l2[i2]) {
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.
216 while (i1 < s1 && XDL_ISSPACE(l1[i1]))
222 while (i2 < s2 && XDL_ISSPACE(l2[i2]))
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;
234 for (; ptr < top && *ptr != '\n'; ptr++) {
235 if (XDL_ISSPACE(*ptr)) {
236 const char *ptr2 = ptr;
238 while (ptr + 1 < top && XDL_ISSPACE(ptr[1])
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
247 ha ^= (unsigned long) ' ';
249 else if (flags & XDF_IGNORE_WHITESPACE_AT_EOL
251 while (ptr2 != ptr + 1) {
253 ha ^= (unsigned long) *ptr2;
260 ha ^= (unsigned long) *ptr;
262 *data = ptr < top ? ptr + 1: ptr;
269 #define REPEAT_BYTE(x) ((~0ul / 0xff) * (x))
271 #define ONEBYTES REPEAT_BYTE(0x01)
272 #define NEWLINEBYTES REPEAT_BYTE(0x0a)
273 #define HIGHBITS REPEAT_BYTE(0x80)
275 /* Return the high bit set in the first byte that is a zero */
276 static inline unsigned long has_zero(unsigned long a)
278 return ((a - ONEBYTES) & ~a) & HIGHBITS;
281 static inline long count_masked_bytes(unsigned long mask)
283 if (sizeof(long) == 8) {
285 * Jan Achrenius on G+: microoptimized version of
286 * the simpler "(mask & ONEBYTES) * ONEBYTES >> 56"
287 * that works for the bytemasks without having to
291 * return mask * 0x0001020304050608 >> 56;
293 * Doing it like this avoids warnings on 32-bit machines.
295 long a = (REPEAT_BYTE(0x01) / 0xff + 1);
296 return mask * a >> (sizeof(long) * 7);
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 */
306 unsigned long xdl_hash_record(char const **data, char const *top, long flags)
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;
313 if (flags & XDF_WHITESPACE_FLAGS)
314 return xdl_hash_record_with_whitespace(data, top, flags);
316 ptr -= sizeof(unsigned long);
320 ptr += sizeof(unsigned long);
323 a = *(unsigned long *)ptr;
324 /* Do we have any '\n' bytes in this word? */
325 mask = has_zero(a ^ NEWLINEBYTES);
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.
335 * To avoid problems with masking in a signed value,
336 * we use an unsigned char here.
339 for (p = top - 1; p >= ptr; p--)
340 a = (a << 8) + *((const unsigned char *)p);
341 mask = has_zero(a ^ NEWLINEBYTES);
344 * No '\n' found in the partial word. Make a
345 * mask that matches what we read.
347 mask = 1UL << (8 * (top - ptr) + 7);
350 /* The mask *below* the first high bit set */
351 mask = (mask - 1) & ~mask;
356 /* Advance past the last (possibly partial) word */
357 ptr += count_masked_bytes(mask);
360 assert(*ptr == '\n');
369 #else /* XDL_FAST_HASH */
371 unsigned long xdl_hash_record(char const **data, char const *top, long flags) {
372 unsigned long ha = 5381;
373 char const *ptr = *data;
375 if (flags & XDF_WHITESPACE_FLAGS)
376 return xdl_hash_record_with_whitespace(data, top, flags);
378 for (; ptr < top && *ptr != '\n'; ptr++) {
380 ha ^= (unsigned long) *ptr;
382 *data = ptr < top ? ptr + 1: ptr;
387 #endif /* XDL_FAST_HASH */
389 unsigned int xdl_hashbits(unsigned int size) {
390 unsigned int val = 1, bits = 0;
392 for (; val < size && bits < CHAR_BIT * sizeof(unsigned int); val <<= 1, bits++);
393 return bits ? bits: 1;
397 int xdl_num_out(char *out, long val) {
398 char *ptr, *str = out;
401 ptr = buf + sizeof(buf) - 1;
407 for (; val && ptr > buf; val /= 10)
408 *--ptr = "0123456789"[val % 10];
410 for (; *ptr; ptr++, str++)
419 int xdl_emit_hunk_hdr(long s1, long c1, long s2, long c2,
420 const char *func, long funclen, xdemitcb_t *ecb) {
425 memcpy(buf, "@@ -", 4);
428 nb += xdl_num_out(buf + nb, c1 ? s1: s1 - 1);
431 memcpy(buf + nb, ",", 1);
434 nb += xdl_num_out(buf + nb, c1);
437 memcpy(buf + nb, " +", 2);
440 nb += xdl_num_out(buf + nb, c2 ? s2: s2 - 1);
443 memcpy(buf + nb, ",", 1);
446 nb += xdl_num_out(buf + nb, c2);
449 memcpy(buf + nb, " @@", 3);
451 if (func && funclen) {
453 if (funclen > sizeof(buf) - nb - 1)
454 funclen = sizeof(buf) - nb - 1;
455 memcpy(buf + nb, func, funclen);
462 if (ecb->outf(ecb->priv, &mb, 1) < 0)
468 int xdl_fall_back_diff(xdfenv_t *diff_env, xpparam_t const *xpp,
469 int line1, int count1, int line2, int count2)
472 * This probably does not work outside Git, since
473 * we have a very simple mmfile structure.
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.
479 mmfile_t subfile1, subfile2;
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)
491 memcpy(diff_env->xdf1.rchg + line1 - 1, env.xdf1.rchg, count1);
492 memcpy(diff_env->xdf2.rchg + line2 - 1, env.xdf2.rchg, count2);