Applied to trunk (with significant changes).
[rsync/rsync-patches.git] / fuzzy.diff
CommitLineData
824abc86 1The changes to generator.c were greatly simplified, making the patch
8c5b8235 2easier to maintain and fixing the failing test in the testsuite.
58118c25 3Very lightly tested.
241013b4 4
824abc86
WD
5Be sure to run "make proto" before "make".
6
0edc7d7f
WD
7--- orig/generator.c 2004-11-03 20:30:45
8+++ generator.c 2004-11-03 22:52:53
f48a237e 9@@ -43,6 +43,7 @@ extern int ignore_times;
58118c25
WD
10 extern int size_only;
11 extern int io_timeout;
12 extern int protocol_version;
09fb2223 13+extern int fuzzy_basis;
58118c25 14 extern int always_checksum;
b952a177 15 extern char *partial_dir;
58118c25 16 extern char *compare_dest;
dd39e657 17@@ -244,6 +245,92 @@ static void generate_and_send_sums(int f
58118c25
WD
18 }
19
20
47dd7a31
WD
21+static void split_names(char *fname, char **dirname, char **basename)
22+{
23+ char *slash = strrchr(fname, '/');
24+ if (slash) {
25+ *dirname = fname;
26+ *slash = '\0';
27+ *basename = slash+1;
28+ } else {
29+ *basename = fname;
30+ *dirname = ".";
31+ }
32+}
33+
58118c25 34+
47dd7a31
WD
35+static unsigned int measure_name(const char *name, const char *basename,
36+ const char *ext)
37+{
38+ int namelen = strlen(name);
39+ int extlen = strlen(ext);
40+ unsigned int score = 0;
41+
42+ /* Extensions must match */
43+ if (namelen <= extlen || strcmp(name + namelen - extlen, ext) != 0)
44+ return 0;
45+
46+ /* Now score depends on similarity of prefix */
47+ for (; *name == *basename && *name; name++, basename++)
48+ score++;
49+ return score;
50+}
51+
58118c25 52+
09fb2223 53+static int find_fuzzy(const char *fname, char *buf, STRUCT_STAT *st_ptr)
47dd7a31
WD
54+{
55+ DIR *d;
56+ struct dirent *di;
57+ char *basename, *dirname;
58+ char mangled_name[MAXPATHLEN];
59+ char bestname[MAXPATHLEN];
60+ unsigned int bestscore = 0;
61+ const char *ext;
62+
09fb2223 63+ strlcpy(mangled_name, fname, sizeof mangled_name);
47dd7a31
WD
64+
65+ split_names(mangled_name, &dirname, &basename);
66+ if (!(d = opendir(dirname))) {
67+ rsyserr(FERROR, errno, "recv_generator opendir(%s)", dirname);
68+ return -1;
69+ }
70+
71+ /* Get final extension, eg. .gz; never full basename though. */
dd39e657 72+ if (!(ext = strrchr(basename + 1, '.')))
47dd7a31
WD
73+ ext = basename + strlen(basename); /* ext = "" */
74+
75+ while ((di = readdir(d)) != NULL) {
76+ const char *dname = d_name(di);
77+ unsigned int score;
78+
79+ if (dname[0] == '.' && (dname[1] == '\0'
80+ || (dname[1] == '.' && dname[2] == '\0')))
81+ continue;
82+
83+ score = measure_name(dname, basename, ext);
84+ if (verbose > 4) {
8c5b8235
WD
85+ rprintf(FINFO, "[%s] fuzzy score for %s = %u\n",
86+ who_am_i(), dname, score);
47dd7a31
WD
87+ }
88+ if (score > bestscore) {
8c5b8235 89+ strlcpy(bestname, dname, sizeof bestname);
47dd7a31
WD
90+ bestscore = score;
91+ }
92+ }
93+ closedir(d);
94+
95+ /* Found a candidate. */
96+ if (bestscore != 0) {
8c5b8235 97+ pathjoin(buf, MAXPATHLEN, dirname, bestname);
47dd7a31 98+ if (verbose > 2) {
8c5b8235 99+ rprintf(FINFO, "[%s] fuzzy match %s->%s\n",
09fb2223 100+ who_am_i(), fname, buf);
47dd7a31 101+ }
58118c25 102+ return link_stat(buf, st_ptr, 0);
47dd7a31
WD
103+ }
104+ return -1;
105+}
58118c25
WD
106+
107
108 /*
109 * Acts on file number @p i from @p flist, whose name is @p fname.
c21bf430 110@@ -466,6 +553,15 @@ static void recv_generator(char *fname,
f48a237e
WD
111 } else
112 partialptr = NULL;
824abc86 113
09fb2223
WD
114+ if (statret == -1 && fuzzy_basis) {
115+ if (find_fuzzy(fname, fnamecmpbuf, &st) == 0
116+ && S_ISREG(st.st_mode)) {
117+ statret = 0;
118+ fnamecmp = fnamecmpbuf;
0edc7d7f 119+ fnamecmp_type = FNAMECMP_FUZZY;
09fb2223 120+ }
824abc86
WD
121+ }
122+
09fb2223
WD
123 if (statret == -1) {
124 if (preserve_hard_links && hard_link_check(file, HL_SKIP))
7628f156 125 return;
c21bf430 126@@ -492,7 +588,7 @@ static void recv_generator(char *fname,
241013b4
MP
127 return;
128 }
129
ec8473c6 130- if (skip_file(fnamecmp, file, &st)) {
0edc7d7f
WD
131+ if (fnamecmp_type != FNAMECMP_FUZZY && skip_file(fnamecmp, file, &st)) {
132 if (fnamecmp_type == FNAMECMP_FNAME)
8c5b8235
WD
133 set_perms(fname, file, &st, PERMS_REPORT);
134 return;
0edc7d7f
WD
135@@ -563,8 +659,24 @@ prepare_to_open:
136
09fb2223 137 notify_others:
0edc7d7f
WD
138 write_int(f_out, i);
139- if (f_out_name >= 0)
140+ if (f_out_name >= 0) {
09fb2223 141 write_byte(f_out_name, fnamecmp_type);
0edc7d7f 142+ if (fnamecmp_type == FNAMECMP_FUZZY) {
09fb2223
WD
143+ uchar lenbuf[3], *lb = lenbuf;
144+ int len = strlen(fnamecmpbuf);
145+ if (len > 0x7F) {
146+#if MAXPATHLEN > 0x7FFF
147+ *lb++ = len / 0x10000 + 0x80;
148+ *lb++ = len / 0x100;
149+#else
150+ *lb++ = len / 0x100 + 0x80;
151+#endif
152+ }
153+ *lb = len;
154+ write_buf(f_out_name, lenbuf, lb - lenbuf + 1);
155+ write_buf(f_out_name, fnamecmpbuf, len);
156+ }
0edc7d7f 157+ }
09fb2223 158
0edc7d7f
WD
159 if (dry_run || read_batch)
160 return;
161--- orig/main.c 2004-11-03 20:30:45
2f5fa77e 162+++ main.c 2004-07-22 00:32:31
d5753a22 163@@ -48,6 +48,7 @@ extern int keep_dirlinks;
495f1899
WD
164 extern int preserve_hard_links;
165 extern int protocol_version;
166 extern int recurse;
09fb2223 167+extern int fuzzy_basis;
495f1899
WD
168 extern int relative_paths;
169 extern int rsync_port;
170 extern int whole_file;
f635ed27 171@@ -463,7 +464,7 @@ static int do_recv(int f_in,int f_out,st
495f1899
WD
172 int pid;
173 int status = 0;
174 int error_pipe[2], name_pipe[2];
c2530f70 175- BOOL need_name_pipe = compare_dest && !dry_run;
09fb2223 176+ BOOL need_name_pipe = (compare_dest || fuzzy_basis) && !dry_run;
495f1899 177
d5753a22
WD
178 /* The receiving side mustn't obey this, or an existing symlink that
179 * points to an identical file won't be replaced by the referent. */
5388f859
WD
180--- orig/options.c 2004-10-14 17:11:40
181+++ options.c 2004-10-14 17:21:51
f635ed27 182@@ -85,6 +85,7 @@ int safe_symlinks = 0;
f6c3b300
WD
183 int copy_unsafe_links = 0;
184 int size_only = 0;
185 int bwlimit = 0;
09fb2223 186+int fuzzy_basis = 0;
f6c3b300
WD
187 size_t bwlimit_writemax = 0;
188 int delete_after = 0;
189 int only_existing = 0;
f635ed27 190@@ -279,6 +280,7 @@ void usage(enum logcode F)
c2530f70 191 rprintf(F," -T, --temp-dir=DIR create temporary files in directory DIR\n");
f0533c4c
WD
192 rprintf(F," --compare-dest=DIR also compare destination files relative to DIR\n");
193 rprintf(F," --link-dest=DIR create hardlinks to DIR for unchanged files\n");
194+ rprintf(F," --fuzzy use similar file as basis if basis doesn't exist\n");
195 rprintf(F," -P equivalent to --partial --progress\n");
196 rprintf(F," -z, --compress compress file data\n");
197 rprintf(F," -C, --cvs-exclude auto ignore files in the same way CVS does\n");
5388f859 198@@ -371,6 +373,7 @@ static struct poptOption long_options[]
f0533c4c
WD
199 {"temp-dir", 'T', POPT_ARG_STRING, &tmpdir, 0, 0, 0 },
200 {"compare-dest", 0, POPT_ARG_STRING, &compare_dest, 0, 0, 0 },
5388f859 201 {"link-dest", 0, POPT_ARG_STRING, &compare_dest, OPT_LINK_DEST, 0, 0 },
09fb2223 202+ {"fuzzy", 0, POPT_ARG_NONE, &fuzzy_basis, 0, 0, 0 },
f0533c4c
WD
203 /* TODO: Should this take an optional int giving the compression level? */
204 {"compress", 'z', POPT_ARG_NONE, &do_compression, 0, 0, 0 },
5388f859
WD
205 {"stats", 0, POPT_ARG_NONE, &do_stats, 0, 0, 0 },
206@@ -890,6 +893,11 @@ int parse_arguments(int *argc, const cha
dd39e657
WD
207 link_dest ? "--link-dest" : "--compare-dest");
208 return 0;
209 }
210+ if (compare_dest) {
211+ snprintf(err_buf, sizeof err_buf,
212+ "--inplace does not yet work with --fuzzy\n");
213+ return 0;
214+ }
215 } else {
216 if (keep_partial && !partial_dir)
217 partial_dir = getenv("RSYNC_PARTIAL_DIR");
5388f859 218@@ -1156,6 +1164,9 @@ void server_options(char **args,int *arg
f74d2272 219 }
241013b4 220 }
7b675ff5 221
09fb2223 222+ if (fuzzy_basis && am_sender)
241013b4 223+ args[ac++] = "--fuzzy";
7b675ff5 224+
241013b4 225 *argc = ac;
f74d2272 226 return;
7b675ff5 227
0edc7d7f
WD
228--- orig/receiver.c 2004-11-03 20:30:45
229+++ receiver.c 2004-11-03 22:54:05
230@@ -320,6 +320,27 @@ static int receive_data(int f_in, char *
09fb2223
WD
231 }
232
233
234+static void read_gen_name(int fd, char *buf)
235+{
236+ int len = read_byte(fd);
237+ if (len & 0x80) {
238+#if MAXPATHLEN > 32767
239+ uchar lenbuf[2];
240+ read_buf(fd, (char *)lenbuf, 2);
241+ len = (len & ~0x80) * 0x10000 + lenbuf[0] * 0x100 + lenbuf[1];
242+#else
243+ len = (len & ~0x80) * 0x100 + read_byte(fd);
244+#endif
245+ }
246+ if (len >= MAXPATHLEN) {
247+ rprintf(FERROR, "bogus data on generator name pipe\n");
248+ exit_cleanup(RERR_PROTOCOL);
249+ }
250+
251+ read_sbuf(fd, buf, len);
252+}
253+
254+
255 static void discard_receive_data(int f_in, OFF_T length)
256 {
257 receive_data(f_in, NULL, -1, 0, NULL, -1, length);
0edc7d7f
WD
258@@ -449,6 +470,10 @@ int recv_files(int f_in, struct file_lis
259 case FNAMECMP_BACKUP:
f48a237e 260 fnamecmp = get_backup_name(fname);
09fb2223 261 break;
0edc7d7f 262+ case FNAMECMP_FUZZY:
09fb2223
WD
263+ read_gen_name(f_in_name, fnamecmpbuf);
264+ fnamecmp = fnamecmpbuf;
265+ break;
0edc7d7f 266 case FNAMECMP_CMPDEST:
09fb2223
WD
267 default:
268 pathjoin(fnamecmpbuf, sizeof fnamecmpbuf,
0edc7d7f
WD
269--- orig/rsync.h 2004-11-03 20:30:45
270+++ rsync.h 2004-11-03 22:53:09
271@@ -123,6 +123,7 @@
272 #define FNAMECMP_FNAME 0x80
273 #define FNAMECMP_PARTIAL_DIR 0x81
274 #define FNAMECMP_BACKUP 0x82
275+#define FNAMECMP_FUZZY 0x83
09fb2223
WD
276
277
278 /* Log-message categories. FLOG is only used on the daemon side to
c21bf430 279--- orig/rsync.yo 2004-10-06 00:12:16
824abc86 280+++ rsync.yo 2004-07-03 19:27:25
67c08134 281@@ -356,6 +356,7 @@ verb(
f0533c4c
WD
282 -T --temp-dir=DIR create temporary files in directory DIR
283 --compare-dest=DIR also compare received files relative to DIR
284 --link-dest=DIR create hardlinks to DIR for unchanged files
285+ --fuzzy use similar file as basis if basis is gone
286 -P equivalent to --partial --progress
287 -z, --compress compress file data
288 -C, --cvs-exclude auto ignore files in the same way CVS does