| 1 | The changes to generator.c were greatly simplified, making the patch |
| 2 | easier to maintain and fixing the failing test in the testsuite. |
| 3 | Lightly tested. |
| 4 | |
| 5 | Be sure to run "make proto" before "make". |
| 6 | |
| 7 | --- orig/generator.c 2005-02-03 02:04:20 |
| 8 | +++ generator.c 2005-02-03 02:11:10 |
| 9 | @@ -47,6 +47,7 @@ extern int size_only; |
| 10 | extern OFF_T max_size; |
| 11 | extern int io_timeout; |
| 12 | extern int protocol_version; |
| 13 | +extern int fuzzy_basis; |
| 14 | extern int always_checksum; |
| 15 | extern char *partial_dir; |
| 16 | extern char *basis_dir[]; |
| 17 | @@ -227,6 +228,88 @@ static void generate_and_send_sums(int f |
| 18 | unmap_file(mapbuf); |
| 19 | } |
| 20 | |
| 21 | +/* Try to find a filename in the same dir as "fname" with a similar name. |
| 22 | + * |
| 23 | + * TODO: |
| 24 | + * - We should be using a cache of names for the current dir, not |
| 25 | + * re-reading the destination directory for every file. |
| 26 | + * - We must not return an rsync tempfile from the current transfer. |
| 27 | + * - If the highest-rated name is not a normal file, we should fall- |
| 28 | + * back to the next highest-rated file. |
| 29 | + * - We must not return a destination file that is being updated |
| 30 | + * during the current transfer, even if we already processed it |
| 31 | + * (since the receiver may not be done with it yet). |
| 32 | + * - We must weed out any names that a daemon's config has excluded. |
| 33 | + */ |
| 34 | +static int find_fuzzy(const char *fname, char *buf, STRUCT_STAT *st_ptr) |
| 35 | +{ |
| 36 | + DIR *d; |
| 37 | + struct dirent *di; |
| 38 | + char *basename, *dirname, *slash; |
| 39 | + char bestname[MAXPATHLEN]; |
| 40 | + int suf_len, basename_len; |
| 41 | + uint32 lowest_dist = 0x7FFFFFFF; |
| 42 | + const char *suf; |
| 43 | + |
| 44 | + strlcpy(buf, fname, MAXPATHLEN); |
| 45 | + if ((slash = strrchr(buf, '/')) != NULL) { |
| 46 | + dirname = buf; |
| 47 | + *slash = '\0'; |
| 48 | + basename = slash + 1; |
| 49 | + } else { |
| 50 | + basename = buf; |
| 51 | + dirname = "."; |
| 52 | + } |
| 53 | + basename_len = strlen(basename); |
| 54 | + |
| 55 | + if (!(d = opendir(dirname))) { |
| 56 | + rsyserr(FERROR, errno, "recv_generator opendir(%s)", dirname); |
| 57 | + return -1; |
| 58 | + } |
| 59 | + if (slash) |
| 60 | + *slash = '/'; |
| 61 | + |
| 62 | + suf = find_filename_suffix(basename, basename_len, &suf_len); |
| 63 | + |
| 64 | + bestname[0] = '\0'; |
| 65 | + while ((di = readdir(d)) != NULL) { |
| 66 | + const char *dname_suf, *dname = d_name(di); |
| 67 | + uint32 dist; |
| 68 | + int dname_len, dname_suf_len; |
| 69 | + |
| 70 | + if (dname[0] == '.' && (dname[1] == '\0' |
| 71 | + || (dname[1] == '.' && dname[2] == '\0'))) |
| 72 | + continue; |
| 73 | + |
| 74 | + dname_len = strlen(dname); |
| 75 | + dname_suf = find_filename_suffix(dname, dname_len, &dname_suf_len); |
| 76 | + |
| 77 | + dist = fuzzy_distance(dname, dname_len, basename, basename_len); |
| 78 | + /* Add some extra weight to how well the suffixes match. */ |
| 79 | + dist += fuzzy_distance(dname_suf, dname_suf_len, suf, suf_len) * 10; |
| 80 | + if (verbose > 4) { |
| 81 | + rprintf(FINFO, "fuzzy distance for %s = %d (%d)\n", |
| 82 | + dname, (int)(dist>>16), (int)(dist&0xFFFF)); |
| 83 | + } |
| 84 | + if (dist <= lowest_dist) { |
| 85 | + strlcpy(bestname, dname, sizeof bestname); |
| 86 | + lowest_dist = dist; |
| 87 | + } |
| 88 | + } |
| 89 | + closedir(d); |
| 90 | + |
| 91 | + /* Found a candidate. */ |
| 92 | + if (bestname[0] != '\0') { |
| 93 | + strlcpy(basename, bestname, MAXPATHLEN - (basename - buf)); |
| 94 | + if (verbose > 2) { |
| 95 | + rprintf(FINFO, "fuzzy match %s->%s\n", |
| 96 | + safe_fname(fname), buf); |
| 97 | + } |
| 98 | + return link_stat(buf, st_ptr, 0); |
| 99 | + } |
| 100 | + return -1; |
| 101 | +} |
| 102 | + |
| 103 | |
| 104 | /* Acts on flist->file's ndx'th item, whose name is fname. If a directory, |
| 105 | * make sure it exists, and has the right permissions/timestamp info. For |
| 106 | @@ -477,6 +560,15 @@ static void recv_generator(char *fname, |
| 107 | } else |
| 108 | partialptr = NULL; |
| 109 | |
| 110 | + if (statret == -1 && fuzzy_basis) { |
| 111 | + if (find_fuzzy(fname, fnamecmpbuf, &st) == 0 |
| 112 | + && S_ISREG(st.st_mode)) { |
| 113 | + statret = 0; |
| 114 | + fnamecmp = fnamecmpbuf; |
| 115 | + fnamecmp_type = FNAMECMP_FUZZY; |
| 116 | + } |
| 117 | + } |
| 118 | + |
| 119 | if (statret == -1) { |
| 120 | if (preserve_hard_links && hard_link_check(file, HL_SKIP)) |
| 121 | return; |
| 122 | @@ -505,6 +597,8 @@ static void recv_generator(char *fname, |
| 123 | |
| 124 | if (!compare_dest && fnamecmp_type <= FNAMECMP_BASIS_DIR_HIGH) |
| 125 | ; |
| 126 | + else if (fnamecmp_type == FNAMECMP_FUZZY) |
| 127 | + ; |
| 128 | else if (unchanged_file(fnamecmp, file, &st)) { |
| 129 | if (fnamecmp_type == FNAMECMP_FNAME) |
| 130 | set_perms(fname, file, &st, PERMS_REPORT); |
| 131 | @@ -579,8 +673,24 @@ notify_others: |
| 132 | write_int(f_out, ndx); |
| 133 | if (protocol_version >= 29 && inplace && !read_batch) |
| 134 | write_byte(f_out, fnamecmp_type); |
| 135 | - if (f_out_name >= 0) |
| 136 | + if (f_out_name >= 0) { |
| 137 | write_byte(f_out_name, fnamecmp_type); |
| 138 | + if (fnamecmp_type == FNAMECMP_FUZZY) { |
| 139 | + uchar lenbuf[3], *lb = lenbuf; |
| 140 | + int len = strlen(fnamecmpbuf); |
| 141 | + if (len > 0x7F) { |
| 142 | +#if MAXPATHLEN > 0x7FFF |
| 143 | + *lb++ = len / 0x10000 + 0x80; |
| 144 | + *lb++ = len / 0x100; |
| 145 | +#else |
| 146 | + *lb++ = len / 0x100 + 0x80; |
| 147 | +#endif |
| 148 | + } |
| 149 | + *lb = len; |
| 150 | + write_buf(f_out_name, lenbuf, lb - lenbuf + 1); |
| 151 | + write_buf(f_out_name, fnamecmpbuf, len); |
| 152 | + } |
| 153 | + } |
| 154 | |
| 155 | if (dry_run || read_batch) |
| 156 | return; |
| 157 | --- orig/main.c 2005-01-30 10:07:21 |
| 158 | +++ main.c 2005-01-14 18:33:15 |
| 159 | @@ -48,6 +48,7 @@ extern int keep_dirlinks; |
| 160 | extern int preserve_hard_links; |
| 161 | extern int protocol_version; |
| 162 | extern int recurse; |
| 163 | +extern int fuzzy_basis; |
| 164 | extern int relative_paths; |
| 165 | extern int rsync_port; |
| 166 | extern int whole_file; |
| 167 | @@ -491,7 +492,8 @@ static int do_recv(int f_in,int f_out,st |
| 168 | int pid; |
| 169 | int status = 0; |
| 170 | int error_pipe[2], name_pipe[2]; |
| 171 | - BOOL need_name_pipe = (basis_dir[0] || partial_dir) && !dry_run; |
| 172 | + BOOL need_name_pipe = (basis_dir[0] || partial_dir || fuzzy_basis) |
| 173 | + && !dry_run; |
| 174 | |
| 175 | /* The receiving side mustn't obey this, or an existing symlink that |
| 176 | * points to an identical file won't be replaced by the referent. */ |
| 177 | --- orig/options.c 2005-02-01 10:39:22 |
| 178 | +++ options.c 2005-01-28 19:31:20 |
| 179 | @@ -90,6 +90,7 @@ int copy_unsafe_links = 0; |
| 180 | int size_only = 0; |
| 181 | int daemon_bwlimit = 0; |
| 182 | int bwlimit = 0; |
| 183 | +int fuzzy_basis = 0; |
| 184 | size_t bwlimit_writemax = 0; |
| 185 | int only_existing = 0; |
| 186 | int opt_ignore_existing = 0; |
| 187 | @@ -303,6 +304,7 @@ void usage(enum logcode F) |
| 188 | rprintf(F," --compare-dest=DIR also compare destination files relative to DIR\n"); |
| 189 | rprintf(F," --copy-dest=DIR ... and include copies of unchanged files\n"); |
| 190 | rprintf(F," --link-dest=DIR hardlink to files in DIR when unchanged\n"); |
| 191 | + rprintf(F," --fuzzy find similar file for basis when no dest file\n"); |
| 192 | rprintf(F," -z, --compress compress file data\n"); |
| 193 | rprintf(F," -C, --cvs-exclude auto-ignore files the same way CVS does\n"); |
| 194 | rprintf(F," -f, --filter=RULE add a file-filtering RULE\n"); |
| 195 | @@ -408,6 +410,7 @@ static struct poptOption long_options[] |
| 196 | {"compare-dest", 0, POPT_ARG_STRING, 0, OPT_COMPARE_DEST, 0, 0 }, |
| 197 | {"copy-dest", 0, POPT_ARG_STRING, 0, OPT_COPY_DEST, 0, 0 }, |
| 198 | {"link-dest", 0, POPT_ARG_STRING, 0, OPT_LINK_DEST, 0, 0 }, |
| 199 | + {"fuzzy", 0, POPT_ARG_NONE, &fuzzy_basis, 0, 0, 0 }, |
| 200 | /* TODO: Should this take an optional int giving the compression level? */ |
| 201 | {"compress", 'z', POPT_ARG_NONE, &do_compression, 0, 0, 0 }, |
| 202 | {"stats", 0, POPT_ARG_NONE, &do_stats, 0, 0, 0 }, |
| 203 | @@ -1321,6 +1324,9 @@ void server_options(char **args,int *arg |
| 204 | if (!implied_dirs && !am_sender) |
| 205 | args[ac++] = "--no-implied-dirs"; |
| 206 | |
| 207 | + if (fuzzy_basis && am_sender) |
| 208 | + args[ac++] = "--fuzzy"; |
| 209 | + |
| 210 | *argc = ac; |
| 211 | return; |
| 212 | |
| 213 | --- orig/receiver.c 2005-02-03 02:04:20 |
| 214 | +++ receiver.c 2005-01-15 21:21:02 |
| 215 | @@ -256,6 +256,27 @@ static int receive_data(int f_in, char * |
| 216 | } |
| 217 | |
| 218 | |
| 219 | +static void read_gen_name(int fd, char *buf) |
| 220 | +{ |
| 221 | + int len = read_byte(fd); |
| 222 | + if (len & 0x80) { |
| 223 | +#if MAXPATHLEN > 32767 |
| 224 | + uchar lenbuf[2]; |
| 225 | + read_buf(fd, (char *)lenbuf, 2); |
| 226 | + len = (len & ~0x80) * 0x10000 + lenbuf[0] * 0x100 + lenbuf[1]; |
| 227 | +#else |
| 228 | + len = (len & ~0x80) * 0x100 + read_byte(fd); |
| 229 | +#endif |
| 230 | + } |
| 231 | + if (len >= MAXPATHLEN) { |
| 232 | + rprintf(FERROR, "bogus data on generator name pipe\n"); |
| 233 | + exit_cleanup(RERR_PROTOCOL); |
| 234 | + } |
| 235 | + |
| 236 | + read_sbuf(fd, buf, len); |
| 237 | +} |
| 238 | + |
| 239 | + |
| 240 | static void discard_receive_data(int f_in, OFF_T length) |
| 241 | { |
| 242 | receive_data(f_in, NULL, -1, 0, NULL, -1, length); |
| 243 | @@ -395,6 +416,10 @@ int recv_files(int f_in, struct file_lis |
| 244 | case FNAMECMP_BACKUP: |
| 245 | fnamecmp = get_backup_name(fname); |
| 246 | break; |
| 247 | + case FNAMECMP_FUZZY: |
| 248 | + read_gen_name(f_in_name, fnamecmpbuf); |
| 249 | + fnamecmp = fnamecmpbuf; |
| 250 | + break; |
| 251 | default: |
| 252 | if (j >= basis_dir_cnt) { |
| 253 | rprintf(FERROR, |
| 254 | --- orig/rsync.h 2005-02-03 02:04:20 |
| 255 | +++ rsync.h 2005-01-19 18:36:47 |
| 256 | @@ -131,6 +131,7 @@ |
| 257 | #define FNAMECMP_FNAME 0x80 |
| 258 | #define FNAMECMP_PARTIAL_DIR 0x81 |
| 259 | #define FNAMECMP_BACKUP 0x82 |
| 260 | +#define FNAMECMP_FUZZY 0x83 |
| 261 | |
| 262 | /* For calling delete_file() */ |
| 263 | #define DEL_DIR (1<<0) |
| 264 | --- orig/rsync.yo 2005-02-01 10:39:23 |
| 265 | +++ rsync.yo 2005-01-28 19:31:36 |
| 266 | @@ -354,6 +354,7 @@ to the detailed description below for a |
| 267 | --compare-dest=DIR also compare received files relative to DIR |
| 268 | --copy-dest=DIR ... and include copies of unchanged files |
| 269 | --link-dest=DIR hardlink to files in DIR when unchanged |
| 270 | + --fuzzy find similar file for basis when no dest |
| 271 | -z, --compress compress file data |
| 272 | -C, --cvs-exclude auto-ignore files in the same way CVS does |
| 273 | -f, --filter=RULE add a file-filtering RULE |
| 274 | @@ -937,6 +938,14 @@ bf(--link-dest) from working properly fo |
| 275 | (or implied by bf(-a)). You can work-around this bug by avoiding the bf(-o) option |
| 276 | when sending to an old rsync. |
| 277 | |
| 278 | +dit(bf(--fuzzy)) This option tells rsync that it should look around for a |
| 279 | +basis file for any destination file that is missing. The current algorithm |
| 280 | +looks for a similarly-named file in the same directory as the destination |
| 281 | +file, and, if found, uses that to try to speed up the transfer. Note that |
| 282 | +the use of the --delete option might get rid of any potential fuzzy-match |
| 283 | +files, so either use --delete-after or filename exclusions if you need to |
| 284 | +prevent this. |
| 285 | + |
| 286 | dit(bf(-z, --compress)) With this option, rsync compresses any data from |
| 287 | the files that it sends to the destination machine. This |
| 288 | option is useful on slow connections. The compression method used is the |
| 289 | --- orig/util.c 2005-01-28 19:08:20 |
| 290 | +++ util.c 2005-01-19 17:30:51 |
| 291 | @@ -1213,3 +1213,108 @@ void *_realloc_array(void *ptr, unsigned |
| 292 | return malloc(size * num); |
| 293 | return realloc(ptr, size * num); |
| 294 | } |
| 295 | + |
| 296 | +/* Take a filename and filename length and return the most significant |
| 297 | + * filename suffix we can find. This ignores suffixes such as "~", |
| 298 | + * ".bak", ".orig", ".~1~", etc. */ |
| 299 | +const char *find_filename_suffix(const char *fn, int fn_len, int *len_ptr) |
| 300 | +{ |
| 301 | + const char *suf, *s; |
| 302 | + BOOL had_tilde; |
| 303 | + int s_len; |
| 304 | + |
| 305 | + /* One or more dots at the start aren't a suffix. */ |
| 306 | + while (fn_len && *fn == '.') fn++, fn_len--; |
| 307 | + |
| 308 | + /* Ignore the ~ in a "foo~" filename. */ |
| 309 | + if (fn_len > 1 && fn[fn_len-1] == '~') |
| 310 | + fn_len--, had_tilde = True; |
| 311 | + else |
| 312 | + had_tilde = False; |
| 313 | + |
| 314 | + /* Assume we don't find an suffix. */ |
| 315 | + suf = ""; |
| 316 | + *len_ptr = 0; |
| 317 | + |
| 318 | + /* Find the last significant suffix. */ |
| 319 | + for (s = fn + fn_len - 1; fn_len > 1; ) { |
| 320 | + while (*s != '.' && s != fn) s--; |
| 321 | + if (s == fn) |
| 322 | + break; |
| 323 | + s_len = fn_len - (s - fn); |
| 324 | + fn_len = s - fn; |
| 325 | + if (s_len == 3) { |
| 326 | + if (strcmp(s+1, "bak") == 0 |
| 327 | + || strcmp(s+1, "old") == 0) |
| 328 | + continue; |
| 329 | + } else if (s_len == 4) { |
| 330 | + if (strcmp(s+1, "orig") == 0) |
| 331 | + continue; |
| 332 | + } else if (s_len > 2 && had_tilde |
| 333 | + && s[1] == '~' && isdigit(s[2])) |
| 334 | + continue; |
| 335 | + *len_ptr = s_len; |
| 336 | + suf = s; |
| 337 | + /* Determine if the suffix is all digits. */ |
| 338 | + for (s++, s_len--; s_len > 0; s++, s_len--) { |
| 339 | + if (!isdigit(*s)) |
| 340 | + return suf; |
| 341 | + } |
| 342 | + /* An all-digit suffix may not be that signficant. */ |
| 343 | + continue; |
| 344 | + } |
| 345 | + |
| 346 | + return suf; |
| 347 | +} |
| 348 | + |
| 349 | +/* This is an implementation of the Levenshtein distance algorithm. It |
| 350 | + * was implemented to avoid needing a two-dimensional matrix (to save |
| 351 | + * memory). It was also tweaked to try to factor in the ASCII distance |
| 352 | + * between changed characters as a minor distance quantity. The normal |
| 353 | + * Levenshtein units of distance (each signifying a single change between |
| 354 | + * the two strings) are defined as a "UNIT". */ |
| 355 | + |
| 356 | +#define UNIT (1 << 16) |
| 357 | + |
| 358 | +uint32 fuzzy_distance(const char *s1, int len1, const char *s2, int len2) |
| 359 | +{ |
| 360 | + uint32 a[MAXPATHLEN], diag, above, left, diag_inc, above_inc, left_inc; |
| 361 | + int32 cost; |
| 362 | + int i1, i2; |
| 363 | + |
| 364 | + if (!len1 || !len2) { |
| 365 | + if (!len1) { |
| 366 | + s1 = s2; |
| 367 | + len1 = len2; |
| 368 | + } |
| 369 | + for (i1 = 0, cost = 0; i1 < len1; i1++) |
| 370 | + cost += s1[i1]; |
| 371 | + return (int32)len1 * UNIT + cost; |
| 372 | + } |
| 373 | + |
| 374 | + for (i2 = 0; i2 < len2; i2++) |
| 375 | + a[i2] = (i2+1) * UNIT; |
| 376 | + |
| 377 | + for (i1 = 0; i1 < len1; i1++) { |
| 378 | + diag = i1 * UNIT; |
| 379 | + above = (i1+1) * UNIT; |
| 380 | + for (i2 = 0; i2 < len2; i2++) { |
| 381 | + left = a[i2]; |
| 382 | + if ((cost = *((uchar*)s1+i1) - *((uchar*)s2+i2)) != 0) { |
| 383 | + if (cost < 0) |
| 384 | + cost = UNIT - cost; |
| 385 | + else |
| 386 | + cost = UNIT + cost; |
| 387 | + } |
| 388 | + diag_inc = diag + cost; |
| 389 | + left_inc = left + UNIT + *((uchar*)s1+i1); |
| 390 | + above_inc = above + UNIT + *((uchar*)s2+i2); |
| 391 | + a[i2] = above = left < above |
| 392 | + ? (left_inc < diag_inc ? left_inc : diag_inc) |
| 393 | + : (above_inc < diag_inc ? above_inc : diag_inc); |
| 394 | + diag = left; |
| 395 | + } |
| 396 | + } |
| 397 | + |
| 398 | + return a[len2-1]; |
| 399 | +} |