Merge commit 'git/master'
[git/git-bigfiles.git] / tree-diff.c
blob9f67af6c1fbb9130962cd373d8e2ebecf543c640
1 /*
2 * Helper functions for tree diff generation
3 */
4 #include "cache.h"
5 #include "diff.h"
6 #include "diffcore.h"
7 #include "tree.h"
9 static char *malloc_base(const char *base, int baselen, const char *path, int pathlen)
11 char *newbase = xmalloc(baselen + pathlen + 2);
12 memcpy(newbase, base, baselen);
13 memcpy(newbase + baselen, path, pathlen);
14 memcpy(newbase + baselen + pathlen, "/", 2);
15 return newbase;
18 static char *malloc_fullname(const char *base, int baselen, const char *path, int pathlen)
20 char *fullname = xmalloc(baselen + pathlen + 1);
21 memcpy(fullname, base, baselen);
22 memcpy(fullname + baselen, path, pathlen);
23 fullname[baselen + pathlen] = 0;
24 return fullname;
27 static void show_entry(struct diff_options *opt, const char *prefix, struct tree_desc *desc,
28 const char *base, int baselen);
30 static int compare_tree_entry(struct tree_desc *t1, struct tree_desc *t2, const char *base, int baselen, struct diff_options *opt)
32 unsigned mode1, mode2;
33 const char *path1, *path2;
34 const unsigned char *sha1, *sha2;
35 int cmp, pathlen1, pathlen2;
36 char *fullname;
38 sha1 = tree_entry_extract(t1, &path1, &mode1);
39 sha2 = tree_entry_extract(t2, &path2, &mode2);
41 pathlen1 = tree_entry_len(path1, sha1);
42 pathlen2 = tree_entry_len(path2, sha2);
43 cmp = base_name_compare(path1, pathlen1, mode1, path2, pathlen2, mode2);
44 if (cmp < 0) {
45 show_entry(opt, "-", t1, base, baselen);
46 return -1;
48 if (cmp > 0) {
49 show_entry(opt, "+", t2, base, baselen);
50 return 1;
52 if (!DIFF_OPT_TST(opt, FIND_COPIES_HARDER) && !hashcmp(sha1, sha2) && mode1 == mode2)
53 return 0;
56 * If the filemode has changed to/from a directory from/to a regular
57 * file, we need to consider it a remove and an add.
59 if (S_ISDIR(mode1) != S_ISDIR(mode2)) {
60 show_entry(opt, "-", t1, base, baselen);
61 show_entry(opt, "+", t2, base, baselen);
62 return 0;
65 if (DIFF_OPT_TST(opt, RECURSIVE) && S_ISDIR(mode1)) {
66 int retval;
67 char *newbase = malloc_base(base, baselen, path1, pathlen1);
68 if (DIFF_OPT_TST(opt, TREE_IN_RECURSIVE)) {
69 newbase[baselen + pathlen1] = 0;
70 opt->change(opt, mode1, mode2,
71 sha1, sha2, newbase);
72 newbase[baselen + pathlen1] = '/';
74 retval = diff_tree_sha1(sha1, sha2, newbase, opt);
75 free(newbase);
76 return retval;
79 fullname = malloc_fullname(base, baselen, path1, pathlen1);
80 opt->change(opt, mode1, mode2, sha1, sha2, fullname);
81 free(fullname);
82 return 0;
86 * Is a tree entry interesting given the pathspec we have?
88 * Return:
89 * - 2 for "yes, and all subsequent entries will be"
90 * - 1 for yes
91 * - zero for no
92 * - negative for "no, and no subsequent entries will be either"
94 static int tree_entry_interesting(struct tree_desc *desc, const char *base, int baselen, struct diff_options *opt)
96 const char *path;
97 const unsigned char *sha1;
98 unsigned mode;
99 int i;
100 int pathlen;
101 int never_interesting = -1;
103 if (!opt->nr_paths)
104 return 1;
106 sha1 = tree_entry_extract(desc, &path, &mode);
108 pathlen = tree_entry_len(path, sha1);
110 for (i = 0; i < opt->nr_paths; i++) {
111 const char *match = opt->paths[i];
112 int matchlen = opt->pathlens[i];
113 int m = -1; /* signals that we haven't called strncmp() */
115 if (baselen >= matchlen) {
116 /* If it doesn't match, move along... */
117 if (strncmp(base, match, matchlen))
118 continue;
121 * The base is a subdirectory of a path which
122 * was specified, so all of them are interesting.
124 return 2;
127 /* Does the base match? */
128 if (strncmp(base, match, baselen))
129 continue;
131 match += baselen;
132 matchlen -= baselen;
134 if (never_interesting) {
136 * We have not seen any match that sorts later
137 * than the current path.
141 * Does match sort strictly earlier than path
142 * with their common parts?
144 m = strncmp(match, path,
145 (matchlen < pathlen) ? matchlen : pathlen);
146 if (m < 0)
147 continue;
150 * If we come here even once, that means there is at
151 * least one pathspec that would sort equal to or
152 * later than the path we are currently looking at.
153 * In other words, if we have never reached this point
154 * after iterating all pathspecs, it means all
155 * pathspecs are either outside of base, or inside the
156 * base but sorts strictly earlier than the current
157 * one. In either case, they will never match the
158 * subsequent entries. In such a case, we initialized
159 * the variable to -1 and that is what will be
160 * returned, allowing the caller to terminate early.
162 never_interesting = 0;
165 if (pathlen > matchlen)
166 continue;
168 if (matchlen > pathlen) {
169 if (match[pathlen] != '/')
170 continue;
171 if (!S_ISDIR(mode))
172 continue;
175 if (m == -1)
177 * we cheated and did not do strncmp(), so we do
178 * that here.
180 m = strncmp(match, path, pathlen);
183 * If common part matched earlier then it is a hit,
184 * because we rejected the case where path is not a
185 * leading directory and is shorter than match.
187 if (!m)
188 return 1;
190 return never_interesting; /* No matches */
193 /* A whole sub-tree went away or appeared */
194 static void show_tree(struct diff_options *opt, const char *prefix, struct tree_desc *desc, const char *base, int baselen)
196 int all_interesting = 0;
197 while (desc->size) {
198 int show;
200 if (all_interesting)
201 show = 1;
202 else {
203 show = tree_entry_interesting(desc, base, baselen,
204 opt);
205 if (show == 2)
206 all_interesting = 1;
208 if (show < 0)
209 break;
210 if (show)
211 show_entry(opt, prefix, desc, base, baselen);
212 update_tree_entry(desc);
216 /* A file entry went away or appeared */
217 static void show_entry(struct diff_options *opt, const char *prefix, struct tree_desc *desc,
218 const char *base, int baselen)
220 unsigned mode;
221 const char *path;
222 const unsigned char *sha1 = tree_entry_extract(desc, &path, &mode);
223 int pathlen = tree_entry_len(path, sha1);
225 if (DIFF_OPT_TST(opt, RECURSIVE) && S_ISDIR(mode)) {
226 enum object_type type;
227 char *newbase = malloc_base(base, baselen, path, pathlen);
228 struct tree_desc inner;
229 void *tree;
230 unsigned long size;
232 tree = read_sha1_file(sha1, &type, &size);
233 if (!tree || type != OBJ_TREE)
234 die("corrupt tree sha %s", sha1_to_hex(sha1));
236 init_tree_desc(&inner, tree, size);
237 show_tree(opt, prefix, &inner, newbase, baselen + 1 + pathlen);
239 free(tree);
240 free(newbase);
241 } else {
242 char *fullname = malloc_fullname(base, baselen, path, pathlen);
243 opt->add_remove(opt, prefix[0], mode, sha1, fullname);
244 free(fullname);
248 static void skip_uninteresting(struct tree_desc *t, const char *base, int baselen, struct diff_options *opt)
250 int all_interesting = 0;
251 while (t->size) {
252 int show;
254 if (all_interesting)
255 show = 1;
256 else {
257 show = tree_entry_interesting(t, base, baselen, opt);
258 if (show == 2)
259 all_interesting = 1;
261 if (!show) {
262 update_tree_entry(t);
263 continue;
265 /* Skip it all? */
266 if (show < 0)
267 t->size = 0;
268 return;
272 int diff_tree(struct tree_desc *t1, struct tree_desc *t2, const char *base, struct diff_options *opt)
274 int baselen = strlen(base);
276 for (;;) {
277 if (DIFF_OPT_TST(opt, QUIET) && DIFF_OPT_TST(opt, HAS_CHANGES))
278 break;
279 if (opt->nr_paths) {
280 skip_uninteresting(t1, base, baselen, opt);
281 skip_uninteresting(t2, base, baselen, opt);
283 if (!t1->size) {
284 if (!t2->size)
285 break;
286 show_entry(opt, "+", t2, base, baselen);
287 update_tree_entry(t2);
288 continue;
290 if (!t2->size) {
291 show_entry(opt, "-", t1, base, baselen);
292 update_tree_entry(t1);
293 continue;
295 switch (compare_tree_entry(t1, t2, base, baselen, opt)) {
296 case -1:
297 update_tree_entry(t1);
298 continue;
299 case 0:
300 update_tree_entry(t1);
301 /* Fallthrough */
302 case 1:
303 update_tree_entry(t2);
304 continue;
306 die("git diff-tree: internal error");
308 return 0;
312 * Does it look like the resulting diff might be due to a rename?
313 * - single entry
314 * - not a valid previous file
316 static inline int diff_might_be_rename(void)
318 return diff_queued_diff.nr == 1 &&
319 !DIFF_FILE_VALID(diff_queued_diff.queue[0]->one);
322 static void try_to_follow_renames(struct tree_desc *t1, struct tree_desc *t2, const char *base, struct diff_options *opt)
324 struct diff_options diff_opts;
325 struct diff_queue_struct *q = &diff_queued_diff;
326 struct diff_filepair *choice;
327 const char *paths[1];
328 int i;
330 /* Remove the file creation entry from the diff queue, and remember it */
331 choice = q->queue[0];
332 q->nr = 0;
334 diff_setup(&diff_opts);
335 DIFF_OPT_SET(&diff_opts, RECURSIVE);
336 diff_opts.detect_rename = DIFF_DETECT_RENAME;
337 diff_opts.output_format = DIFF_FORMAT_NO_OUTPUT;
338 diff_opts.single_follow = opt->paths[0];
339 diff_opts.break_opt = opt->break_opt;
340 paths[0] = NULL;
341 diff_tree_setup_paths(paths, &diff_opts);
342 if (diff_setup_done(&diff_opts) < 0)
343 die("unable to set up diff options to follow renames");
344 diff_tree(t1, t2, base, &diff_opts);
345 diffcore_std(&diff_opts);
346 diff_tree_release_paths(&diff_opts);
348 /* Go through the new set of filepairing, and see if we find a more interesting one */
349 for (i = 0; i < q->nr; i++) {
350 struct diff_filepair *p = q->queue[i];
353 * Found a source? Not only do we use that for the new
354 * diff_queued_diff, we will also use that as the path in
355 * the future!
357 if ((p->status == 'R' || p->status == 'C') && !strcmp(p->two->path, opt->paths[0])) {
358 /* Switch the file-pairs around */
359 q->queue[i] = choice;
360 choice = p;
362 /* Update the path we use from now on.. */
363 diff_tree_release_paths(opt);
364 opt->paths[0] = xstrdup(p->one->path);
365 diff_tree_setup_paths(opt->paths, opt);
366 break;
371 * Then, discard all the non-relevane file pairs...
373 for (i = 0; i < q->nr; i++) {
374 struct diff_filepair *p = q->queue[i];
375 diff_free_filepair(p);
379 * .. and re-instate the one we want (which might be either the
380 * original one, or the rename/copy we found)
382 q->queue[0] = choice;
383 q->nr = 1;
386 int diff_tree_sha1(const unsigned char *old, const unsigned char *new, const char *base, struct diff_options *opt)
388 void *tree1, *tree2;
389 struct tree_desc t1, t2;
390 unsigned long size1, size2;
391 int retval;
393 tree1 = read_object_with_reference(old, tree_type, &size1, NULL);
394 if (!tree1)
395 die("unable to read source tree (%s)", sha1_to_hex(old));
396 tree2 = read_object_with_reference(new, tree_type, &size2, NULL);
397 if (!tree2)
398 die("unable to read destination tree (%s)", sha1_to_hex(new));
399 init_tree_desc(&t1, tree1, size1);
400 init_tree_desc(&t2, tree2, size2);
401 retval = diff_tree(&t1, &t2, base, opt);
402 if (DIFF_OPT_TST(opt, FOLLOW_RENAMES) && diff_might_be_rename()) {
403 init_tree_desc(&t1, tree1, size1);
404 init_tree_desc(&t2, tree2, size2);
405 try_to_follow_renames(&t1, &t2, base, opt);
407 free(tree1);
408 free(tree2);
409 return retval;
412 int diff_root_tree_sha1(const unsigned char *new, const char *base, struct diff_options *opt)
414 int retval;
415 void *tree;
416 unsigned long size;
417 struct tree_desc empty, real;
419 tree = read_object_with_reference(new, tree_type, &size, NULL);
420 if (!tree)
421 die("unable to read root tree (%s)", sha1_to_hex(new));
422 init_tree_desc(&real, tree, size);
424 init_tree_desc(&empty, "", 0);
425 retval = diff_tree(&empty, &real, base, opt);
426 free(tree);
427 return retval;
430 static int count_paths(const char **paths)
432 int i = 0;
433 while (*paths++)
434 i++;
435 return i;
438 void diff_tree_release_paths(struct diff_options *opt)
440 free(opt->pathlens);
443 void diff_tree_setup_paths(const char **p, struct diff_options *opt)
445 opt->nr_paths = 0;
446 opt->pathlens = NULL;
447 opt->paths = NULL;
449 if (p) {
450 int i;
452 opt->paths = p;
453 opt->nr_paths = count_paths(p);
454 if (opt->nr_paths == 0) {
455 opt->pathlens = NULL;
456 return;
458 opt->pathlens = xmalloc(opt->nr_paths * sizeof(int));
459 for (i=0; i < opt->nr_paths; i++)
460 opt->pathlens[i] = strlen(p[i]);