Make --inplace work with --compare-dest, and efficient too.
[rsync/rsync-patches.git] / fuzzy.diff
CommitLineData
824abc86 1Depends-On-Patch: g2r-basis-filename.diff
8c5b8235 2
824abc86 3The changes to generator.c were greatly simplified, making the patch
8c5b8235 4easier to maintain and fixing the failing test in the testsuite.
58118c25 5Very lightly tested.
241013b4 6
824abc86
WD
7Be sure to run "make proto" before "make".
8
c21bf430 9--- orig/generator.c 2004-10-09 04:06:49
dd39e657 10+++ generator.c 2004-09-23 15:33:08
f48a237e 11@@ -43,6 +43,7 @@ extern int ignore_times;
58118c25
WD
12 extern int size_only;
13 extern int io_timeout;
14 extern int protocol_version;
09fb2223 15+extern int fuzzy_basis;
58118c25 16 extern int always_checksum;
b952a177 17 extern char *partial_dir;
58118c25 18 extern char *compare_dest;
dd39e657 19@@ -244,6 +245,92 @@ static void generate_and_send_sums(int f
58118c25
WD
20 }
21
22
47dd7a31
WD
23+static void split_names(char *fname, char **dirname, char **basename)
24+{
25+ char *slash = strrchr(fname, '/');
26+ if (slash) {
27+ *dirname = fname;
28+ *slash = '\0';
29+ *basename = slash+1;
30+ } else {
31+ *basename = fname;
32+ *dirname = ".";
33+ }
34+}
35+
58118c25 36+
47dd7a31
WD
37+static unsigned int measure_name(const char *name, const char *basename,
38+ const char *ext)
39+{
40+ int namelen = strlen(name);
41+ int extlen = strlen(ext);
42+ unsigned int score = 0;
43+
44+ /* Extensions must match */
45+ if (namelen <= extlen || strcmp(name + namelen - extlen, ext) != 0)
46+ return 0;
47+
48+ /* Now score depends on similarity of prefix */
49+ for (; *name == *basename && *name; name++, basename++)
50+ score++;
51+ return score;
52+}
53+
58118c25 54+
09fb2223 55+static int find_fuzzy(const char *fname, char *buf, STRUCT_STAT *st_ptr)
47dd7a31
WD
56+{
57+ DIR *d;
58+ struct dirent *di;
59+ char *basename, *dirname;
60+ char mangled_name[MAXPATHLEN];
61+ char bestname[MAXPATHLEN];
62+ unsigned int bestscore = 0;
63+ const char *ext;
64+
09fb2223 65+ strlcpy(mangled_name, fname, sizeof mangled_name);
47dd7a31
WD
66+
67+ split_names(mangled_name, &dirname, &basename);
68+ if (!(d = opendir(dirname))) {
69+ rsyserr(FERROR, errno, "recv_generator opendir(%s)", dirname);
70+ return -1;
71+ }
72+
73+ /* Get final extension, eg. .gz; never full basename though. */
dd39e657 74+ if (!(ext = strrchr(basename + 1, '.')))
47dd7a31
WD
75+ ext = basename + strlen(basename); /* ext = "" */
76+
77+ while ((di = readdir(d)) != NULL) {
78+ const char *dname = d_name(di);
79+ unsigned int score;
80+
81+ if (dname[0] == '.' && (dname[1] == '\0'
82+ || (dname[1] == '.' && dname[2] == '\0')))
83+ continue;
84+
85+ score = measure_name(dname, basename, ext);
86+ if (verbose > 4) {
8c5b8235
WD
87+ rprintf(FINFO, "[%s] fuzzy score for %s = %u\n",
88+ who_am_i(), dname, score);
47dd7a31
WD
89+ }
90+ if (score > bestscore) {
8c5b8235 91+ strlcpy(bestname, dname, sizeof bestname);
47dd7a31
WD
92+ bestscore = score;
93+ }
94+ }
95+ closedir(d);
96+
97+ /* Found a candidate. */
98+ if (bestscore != 0) {
8c5b8235 99+ pathjoin(buf, MAXPATHLEN, dirname, bestname);
47dd7a31 100+ if (verbose > 2) {
8c5b8235 101+ rprintf(FINFO, "[%s] fuzzy match %s->%s\n",
09fb2223 102+ who_am_i(), fname, buf);
47dd7a31 103+ }
58118c25 104+ return link_stat(buf, st_ptr, 0);
47dd7a31
WD
105+ }
106+ return -1;
107+}
58118c25
WD
108+
109
110 /*
111 * Acts on file number @p i from @p flist, whose name is @p fname.
c21bf430 112@@ -466,6 +553,15 @@ static void recv_generator(char *fname,
f48a237e
WD
113 } else
114 partialptr = NULL;
824abc86 115
09fb2223
WD
116+ if (statret == -1 && fuzzy_basis) {
117+ if (find_fuzzy(fname, fnamecmpbuf, &st) == 0
118+ && S_ISREG(st.st_mode)) {
119+ statret = 0;
120+ fnamecmp = fnamecmpbuf;
121+ fnamecmp_type = G2R_FUZZY;
122+ }
824abc86
WD
123+ }
124+
09fb2223
WD
125 if (statret == -1) {
126 if (preserve_hard_links && hard_link_check(file, HL_SKIP))
7628f156 127 return;
c21bf430 128@@ -492,7 +588,7 @@ static void recv_generator(char *fname,
241013b4
MP
129 return;
130 }
131
ec8473c6 132- if (skip_file(fnamecmp, file, &st)) {
09fb2223
WD
133+ if (fnamecmp_type != G2R_FUZZY && skip_file(fnamecmp, file, &st)) {
134 if (fnamecmp_type == G2R_FNAME)
8c5b8235
WD
135 set_perms(fname, file, &st, PERMS_REPORT);
136 return;
c21bf430 137@@ -564,6 +660,21 @@ prepare_to_open:
09fb2223
WD
138 notify_others:
139 if (f_out_name >= 0) {
140 write_byte(f_out_name, fnamecmp_type);
141+ if (fnamecmp_type == G2R_FUZZY) {
142+ uchar lenbuf[3], *lb = lenbuf;
143+ int len = strlen(fnamecmpbuf);
144+ if (len > 0x7F) {
145+#if MAXPATHLEN > 0x7FFF
146+ *lb++ = len / 0x10000 + 0x80;
147+ *lb++ = len / 0x100;
148+#else
149+ *lb++ = len / 0x100 + 0x80;
150+#endif
151+ }
152+ *lb = len;
153+ write_buf(f_out_name, lenbuf, lb - lenbuf + 1);
154+ write_buf(f_out_name, fnamecmpbuf, len);
155+ }
156 io_flush(NORMAL_FLUSH); /* XXX make this more efficient! */
157 }
158
c21bf430 159--- orig/main.c 2004-10-09 03:25:43
2f5fa77e 160+++ main.c 2004-07-22 00:32:31
d5753a22 161@@ -48,6 +48,7 @@ extern int keep_dirlinks;
495f1899
WD
162 extern int preserve_hard_links;
163 extern int protocol_version;
164 extern int recurse;
09fb2223 165+extern int fuzzy_basis;
495f1899
WD
166 extern int relative_paths;
167 extern int rsync_port;
168 extern int whole_file;
f635ed27 169@@ -463,7 +464,7 @@ static int do_recv(int f_in,int f_out,st
495f1899
WD
170 int pid;
171 int status = 0;
172 int error_pipe[2], name_pipe[2];
c2530f70 173- BOOL need_name_pipe = compare_dest && !dry_run;
09fb2223 174+ BOOL need_name_pipe = (compare_dest || fuzzy_basis) && !dry_run;
495f1899 175
d5753a22
WD
176 /* The receiving side mustn't obey this, or an existing symlink that
177 * points to an identical file won't be replaced by the referent. */
5388f859
WD
178--- orig/options.c 2004-10-14 17:11:40
179+++ options.c 2004-10-14 17:21:51
f635ed27 180@@ -85,6 +85,7 @@ int safe_symlinks = 0;
f6c3b300
WD
181 int copy_unsafe_links = 0;
182 int size_only = 0;
183 int bwlimit = 0;
09fb2223 184+int fuzzy_basis = 0;
f6c3b300
WD
185 size_t bwlimit_writemax = 0;
186 int delete_after = 0;
187 int only_existing = 0;
f635ed27 188@@ -279,6 +280,7 @@ void usage(enum logcode F)
c2530f70 189 rprintf(F," -T, --temp-dir=DIR create temporary files in directory DIR\n");
f0533c4c
WD
190 rprintf(F," --compare-dest=DIR also compare destination files relative to DIR\n");
191 rprintf(F," --link-dest=DIR create hardlinks to DIR for unchanged files\n");
192+ rprintf(F," --fuzzy use similar file as basis if basis doesn't exist\n");
193 rprintf(F," -P equivalent to --partial --progress\n");
194 rprintf(F," -z, --compress compress file data\n");
195 rprintf(F," -C, --cvs-exclude auto ignore files in the same way CVS does\n");
5388f859 196@@ -371,6 +373,7 @@ static struct poptOption long_options[]
f0533c4c
WD
197 {"temp-dir", 'T', POPT_ARG_STRING, &tmpdir, 0, 0, 0 },
198 {"compare-dest", 0, POPT_ARG_STRING, &compare_dest, 0, 0, 0 },
5388f859 199 {"link-dest", 0, POPT_ARG_STRING, &compare_dest, OPT_LINK_DEST, 0, 0 },
09fb2223 200+ {"fuzzy", 0, POPT_ARG_NONE, &fuzzy_basis, 0, 0, 0 },
f0533c4c
WD
201 /* TODO: Should this take an optional int giving the compression level? */
202 {"compress", 'z', POPT_ARG_NONE, &do_compression, 0, 0, 0 },
5388f859
WD
203 {"stats", 0, POPT_ARG_NONE, &do_stats, 0, 0, 0 },
204@@ -890,6 +893,11 @@ int parse_arguments(int *argc, const cha
dd39e657
WD
205 link_dest ? "--link-dest" : "--compare-dest");
206 return 0;
207 }
208+ if (compare_dest) {
209+ snprintf(err_buf, sizeof err_buf,
210+ "--inplace does not yet work with --fuzzy\n");
211+ return 0;
212+ }
213 } else {
214 if (keep_partial && !partial_dir)
215 partial_dir = getenv("RSYNC_PARTIAL_DIR");
5388f859 216@@ -1156,6 +1164,9 @@ void server_options(char **args,int *arg
f74d2272 217 }
241013b4 218 }
7b675ff5 219
09fb2223 220+ if (fuzzy_basis && am_sender)
241013b4 221+ args[ac++] = "--fuzzy";
7b675ff5 222+
241013b4 223 *argc = ac;
f74d2272 224 return;
7b675ff5 225
f48a237e 226--- orig/receiver.c 2004-09-07 21:57:20
09fb2223 227+++ receiver.c 2004-07-30 18:21:38
a6587818 228@@ -319,6 +319,27 @@ static int receive_data(int f_in, char *
09fb2223
WD
229 }
230
231
232+static void read_gen_name(int fd, char *buf)
233+{
234+ int len = read_byte(fd);
235+ if (len & 0x80) {
236+#if MAXPATHLEN > 32767
237+ uchar lenbuf[2];
238+ read_buf(fd, (char *)lenbuf, 2);
239+ len = (len & ~0x80) * 0x10000 + lenbuf[0] * 0x100 + lenbuf[1];
240+#else
241+ len = (len & ~0x80) * 0x100 + read_byte(fd);
242+#endif
243+ }
244+ if (len >= MAXPATHLEN) {
245+ rprintf(FERROR, "bogus data on generator name pipe\n");
246+ exit_cleanup(RERR_PROTOCOL);
247+ }
248+
249+ read_sbuf(fd, buf, len);
250+}
251+
252+
253 static void discard_receive_data(int f_in, OFF_T length)
254 {
255 receive_data(f_in, NULL, -1, 0, NULL, -1, length);
a6587818 256@@ -448,6 +469,10 @@ int recv_files(int f_in, struct file_lis
f48a237e
WD
257 case G2R_BACKUP:
258 fnamecmp = get_backup_name(fname);
09fb2223
WD
259 break;
260+ case G2R_FUZZY:
261+ read_gen_name(f_in_name, fnamecmpbuf);
262+ fnamecmp = fnamecmpbuf;
263+ break;
264 default:
265 pathjoin(fnamecmpbuf, sizeof fnamecmpbuf,
266 compare_dest, fname);
f48a237e
WD
267--- orig/rsync.h 2004-09-07 21:52:22
268+++ rsync.h 2004-09-07 22:02:37
dd39e657 269@@ -122,6 +122,7 @@
09fb2223
WD
270 #define G2R_FNAME 0x80
271 #define G2R_PARTIAL_DIR 0x81
f48a237e
WD
272 #define G2R_BACKUP 0x82
273+#define G2R_FUZZY 0x83
09fb2223
WD
274
275
276 /* Log-message categories. FLOG is only used on the daemon side to
c21bf430 277--- orig/rsync.yo 2004-10-06 00:12:16
824abc86 278+++ rsync.yo 2004-07-03 19:27:25
67c08134 279@@ -356,6 +356,7 @@ verb(
f0533c4c
WD
280 -T --temp-dir=DIR create temporary files in directory DIR
281 --compare-dest=DIR also compare received files relative to DIR
282 --link-dest=DIR create hardlinks to DIR for unchanged files
283+ --fuzzy use similar file as basis if basis is gone
284 -P equivalent to --partial --progress
285 -z, --compress compress file data
286 -C, --cvs-exclude auto ignore files in the same way CVS does