French translation: copy -> copie.
[git/dscho.git] / tree-walk.c
blob492c7cd744d5f0739c8393ab67705c0be1254783
1 #include "cache.h"
2 #include "tree-walk.h"
3 #include "unpack-trees.h"
4 #include "dir.h"
5 #include "tree.h"
7 static const char *get_mode(const char *str, unsigned int *modep)
9 unsigned char c;
10 unsigned int mode = 0;
12 if (*str == ' ')
13 return NULL;
15 while ((c = *str++) != ' ') {
16 if (c < '0' || c > '7')
17 return NULL;
18 mode = (mode << 3) + (c - '0');
20 *modep = mode;
21 return str;
24 static void decode_tree_entry(struct tree_desc *desc, const char *buf, unsigned long size)
26 const char *path;
27 unsigned int mode, len;
29 if (size < 24 || buf[size - 21])
30 die("corrupt tree file");
32 path = get_mode(buf, &mode);
33 if (!path || !*path)
34 die("corrupt tree file");
35 len = strlen(path) + 1;
37 /* Initialize the descriptor entry */
38 desc->entry.path = path;
39 desc->entry.mode = mode;
40 desc->entry.sha1 = (const unsigned char *)(path + len);
43 void init_tree_desc(struct tree_desc *desc, const void *buffer, unsigned long size)
45 desc->buffer = buffer;
46 desc->size = size;
47 if (size)
48 decode_tree_entry(desc, buffer, size);
51 void *fill_tree_descriptor(struct tree_desc *desc, const unsigned char *sha1)
53 unsigned long size = 0;
54 void *buf = NULL;
56 if (sha1) {
57 buf = read_object_with_reference(sha1, tree_type, &size, NULL);
58 if (!buf)
59 die("unable to read tree %s", sha1_to_hex(sha1));
61 init_tree_desc(desc, buf, size);
62 return buf;
65 static void entry_clear(struct name_entry *a)
67 memset(a, 0, sizeof(*a));
70 static void entry_extract(struct tree_desc *t, struct name_entry *a)
72 *a = t->entry;
75 void update_tree_entry(struct tree_desc *desc)
77 const void *buf = desc->buffer;
78 const unsigned char *end = desc->entry.sha1 + 20;
79 unsigned long size = desc->size;
80 unsigned long len = end - (const unsigned char *)buf;
82 if (size < len)
83 die("corrupt tree file");
84 buf = end;
85 size -= len;
86 desc->buffer = buf;
87 desc->size = size;
88 if (size)
89 decode_tree_entry(desc, buf, size);
92 int tree_entry(struct tree_desc *desc, struct name_entry *entry)
94 if (!desc->size)
95 return 0;
97 *entry = desc->entry;
98 update_tree_entry(desc);
99 return 1;
102 void setup_traverse_info(struct traverse_info *info, const char *base)
104 int pathlen = strlen(base);
105 static struct traverse_info dummy;
107 memset(info, 0, sizeof(*info));
108 if (pathlen && base[pathlen-1] == '/')
109 pathlen--;
110 info->pathlen = pathlen ? pathlen + 1 : 0;
111 info->name.path = base;
112 info->name.sha1 = (void *)(base + pathlen + 1);
113 if (pathlen)
114 info->prev = &dummy;
117 char *make_traverse_path(char *path, const struct traverse_info *info, const struct name_entry *n)
119 int len = tree_entry_len(n);
120 int pathlen = info->pathlen;
122 path[pathlen + len] = 0;
123 for (;;) {
124 memcpy(path + pathlen, n->path, len);
125 if (!pathlen)
126 break;
127 path[--pathlen] = '/';
128 n = &info->name;
129 len = tree_entry_len(n);
130 info = info->prev;
131 pathlen -= len;
133 return path;
136 struct tree_desc_skip {
137 struct tree_desc_skip *prev;
138 const void *ptr;
141 struct tree_desc_x {
142 struct tree_desc d;
143 struct tree_desc_skip *skip;
146 static int name_compare(const char *a, int a_len,
147 const char *b, int b_len)
149 int len = (a_len < b_len) ? a_len : b_len;
150 int cmp = memcmp(a, b, len);
151 if (cmp)
152 return cmp;
153 return (a_len - b_len);
156 static int check_entry_match(const char *a, int a_len, const char *b, int b_len)
159 * The caller wants to pick *a* from a tree or nothing.
160 * We are looking at *b* in a tree.
162 * (0) If a and b are the same name, we are trivially happy.
164 * There are three possibilities where *a* could be hiding
165 * behind *b*.
167 * (1) *a* == "t", *b* == "ab" i.e. *b* sorts earlier than *a* no
168 * matter what.
169 * (2) *a* == "t", *b* == "t-2" and "t" is a subtree in the tree;
170 * (3) *a* == "t-2", *b* == "t" and "t-2" is a blob in the tree.
172 * Otherwise we know *a* won't appear in the tree without
173 * scanning further.
176 int cmp = name_compare(a, a_len, b, b_len);
178 /* Most common case first -- reading sync'd trees */
179 if (!cmp)
180 return cmp;
182 if (0 < cmp) {
183 /* a comes after b; it does not matter if it is case (3)
184 if (b_len < a_len && !memcmp(a, b, b_len) && a[b_len] < '/')
185 return 1;
187 return 1; /* keep looking */
190 /* b comes after a; are we looking at case (2)? */
191 if (a_len < b_len && !memcmp(a, b, a_len) && b[a_len] < '/')
192 return 1; /* keep looking */
194 return -1; /* a cannot appear in the tree */
198 * From the extended tree_desc, extract the first name entry, while
199 * paying attention to the candidate "first" name. Most importantly,
200 * when looking for an entry, if there are entries that sorts earlier
201 * in the tree object representation than that name, skip them and
202 * process the named entry first. We will remember that we haven't
203 * processed the first entry yet, and in the later call skip the
204 * entry we processed early when update_extended_entry() is called.
206 * E.g. if the underlying tree object has these entries:
208 * blob "t-1"
209 * blob "t-2"
210 * tree "t"
211 * blob "t=1"
213 * and the "first" asks for "t", remember that we still need to
214 * process "t-1" and "t-2" but extract "t". After processing the
215 * entry "t" from this call, the caller will let us know by calling
216 * update_extended_entry() that we can remember "t" has been processed
217 * already.
220 static void extended_entry_extract(struct tree_desc_x *t,
221 struct name_entry *a,
222 const char *first,
223 int first_len)
225 const char *path;
226 int len;
227 struct tree_desc probe;
228 struct tree_desc_skip *skip;
231 * Extract the first entry from the tree_desc, but skip the
232 * ones that we already returned in earlier rounds.
234 while (1) {
235 if (!t->d.size) {
236 entry_clear(a);
237 break; /* not found */
239 entry_extract(&t->d, a);
240 for (skip = t->skip; skip; skip = skip->prev)
241 if (a->path == skip->ptr)
242 break; /* found */
243 if (!skip)
244 break;
245 /* We have processed this entry already. */
246 update_tree_entry(&t->d);
249 if (!first || !a->path)
250 return;
253 * The caller wants "first" from this tree, or nothing.
255 path = a->path;
256 len = tree_entry_len(a);
257 switch (check_entry_match(first, first_len, path, len)) {
258 case -1:
259 entry_clear(a);
260 case 0:
261 return;
262 default:
263 break;
267 * We need to look-ahead -- we suspect that a subtree whose
268 * name is "first" may be hiding behind the current entry "path".
270 probe = t->d;
271 while (probe.size) {
272 entry_extract(&probe, a);
273 path = a->path;
274 len = tree_entry_len(a);
275 switch (check_entry_match(first, first_len, path, len)) {
276 case -1:
277 entry_clear(a);
278 case 0:
279 return;
280 default:
281 update_tree_entry(&probe);
282 break;
284 /* keep looking */
286 entry_clear(a);
289 static void update_extended_entry(struct tree_desc_x *t, struct name_entry *a)
291 if (t->d.entry.path == a->path) {
292 update_tree_entry(&t->d);
293 } else {
294 /* we have returned this entry early */
295 struct tree_desc_skip *skip = xmalloc(sizeof(*skip));
296 skip->ptr = a->path;
297 skip->prev = t->skip;
298 t->skip = skip;
302 static void free_extended_entry(struct tree_desc_x *t)
304 struct tree_desc_skip *p, *s;
306 for (s = t->skip; s; s = p) {
307 p = s->prev;
308 free(s);
312 static inline int prune_traversal(struct name_entry *e,
313 struct traverse_info *info,
314 struct strbuf *base,
315 int still_interesting)
317 if (!info->pathspec || still_interesting == 2)
318 return 2;
319 if (still_interesting < 0)
320 return still_interesting;
321 return tree_entry_interesting(e, base, 0, info->pathspec);
324 int traverse_trees(int n, struct tree_desc *t, struct traverse_info *info)
326 int ret = 0;
327 int error = 0;
328 struct name_entry *entry = xmalloc(n*sizeof(*entry));
329 int i;
330 struct tree_desc_x *tx = xcalloc(n, sizeof(*tx));
331 struct strbuf base = STRBUF_INIT;
332 int interesting = 1;
334 for (i = 0; i < n; i++)
335 tx[i].d = t[i];
337 if (info->prev) {
338 strbuf_grow(&base, info->pathlen);
339 make_traverse_path(base.buf, info->prev, &info->name);
340 base.buf[info->pathlen-1] = '/';
341 strbuf_setlen(&base, info->pathlen);
343 for (;;) {
344 unsigned long mask, dirmask;
345 const char *first = NULL;
346 int first_len = 0;
347 struct name_entry *e = NULL;
348 int len;
350 for (i = 0; i < n; i++) {
351 e = entry + i;
352 extended_entry_extract(tx + i, e, NULL, 0);
356 * A tree may have "t-2" at the current location even
357 * though it may have "t" that is a subtree behind it,
358 * and another tree may return "t". We want to grab
359 * all "t" from all trees to match in such a case.
361 for (i = 0; i < n; i++) {
362 e = entry + i;
363 if (!e->path)
364 continue;
365 len = tree_entry_len(e);
366 if (!first) {
367 first = e->path;
368 first_len = len;
369 continue;
371 if (name_compare(e->path, len, first, first_len) < 0) {
372 first = e->path;
373 first_len = len;
377 if (first) {
378 for (i = 0; i < n; i++) {
379 e = entry + i;
380 extended_entry_extract(tx + i, e, first, first_len);
381 /* Cull the ones that are not the earliest */
382 if (!e->path)
383 continue;
384 len = tree_entry_len(e);
385 if (name_compare(e->path, len, first, first_len))
386 entry_clear(e);
390 /* Now we have in entry[i] the earliest name from the trees */
391 mask = 0;
392 dirmask = 0;
393 for (i = 0; i < n; i++) {
394 if (!entry[i].path)
395 continue;
396 mask |= 1ul << i;
397 if (S_ISDIR(entry[i].mode))
398 dirmask |= 1ul << i;
399 e = &entry[i];
401 if (!mask)
402 break;
403 interesting = prune_traversal(e, info, &base, interesting);
404 if (interesting < 0)
405 break;
406 if (interesting) {
407 ret = info->fn(n, mask, dirmask, entry, info);
408 if (ret < 0) {
409 error = ret;
410 if (!info->show_all_errors)
411 break;
413 mask &= ret;
415 ret = 0;
416 for (i = 0; i < n; i++)
417 if (mask & (1ul << i))
418 update_extended_entry(tx + i, entry + i);
420 free(entry);
421 for (i = 0; i < n; i++)
422 free_extended_entry(tx + i);
423 free(tx);
424 strbuf_release(&base);
425 return error;
428 static int find_tree_entry(struct tree_desc *t, const char *name, unsigned char *result, unsigned *mode)
430 int namelen = strlen(name);
431 while (t->size) {
432 const char *entry;
433 const unsigned char *sha1;
434 int entrylen, cmp;
436 sha1 = tree_entry_extract(t, &entry, mode);
437 entrylen = tree_entry_len(&t->entry);
438 update_tree_entry(t);
439 if (entrylen > namelen)
440 continue;
441 cmp = memcmp(name, entry, entrylen);
442 if (cmp > 0)
443 continue;
444 if (cmp < 0)
445 break;
446 if (entrylen == namelen) {
447 hashcpy(result, sha1);
448 return 0;
450 if (name[entrylen] != '/')
451 continue;
452 if (!S_ISDIR(*mode))
453 break;
454 if (++entrylen == namelen) {
455 hashcpy(result, sha1);
456 return 0;
458 return get_tree_entry(sha1, name + entrylen, result, mode);
460 return -1;
463 int get_tree_entry(const unsigned char *tree_sha1, const char *name, unsigned char *sha1, unsigned *mode)
465 int retval;
466 void *tree;
467 unsigned long size;
468 unsigned char root[20];
470 tree = read_object_with_reference(tree_sha1, tree_type, &size, root);
471 if (!tree)
472 return -1;
474 if (name[0] == '\0') {
475 hashcpy(sha1, root);
476 free(tree);
477 return 0;
480 if (!size) {
481 retval = -1;
482 } else {
483 struct tree_desc t;
484 init_tree_desc(&t, tree, size);
485 retval = find_tree_entry(&t, name, sha1, mode);
487 free(tree);
488 return retval;
491 static int match_entry(const struct name_entry *entry, int pathlen,
492 const char *match, int matchlen,
493 int *never_interesting)
495 int m = -1; /* signals that we haven't called strncmp() */
497 if (*never_interesting) {
499 * We have not seen any match that sorts later
500 * than the current path.
504 * Does match sort strictly earlier than path
505 * with their common parts?
507 m = strncmp(match, entry->path,
508 (matchlen < pathlen) ? matchlen : pathlen);
509 if (m < 0)
510 return 0;
513 * If we come here even once, that means there is at
514 * least one pathspec that would sort equal to or
515 * later than the path we are currently looking at.
516 * In other words, if we have never reached this point
517 * after iterating all pathspecs, it means all
518 * pathspecs are either outside of base, or inside the
519 * base but sorts strictly earlier than the current
520 * one. In either case, they will never match the
521 * subsequent entries. In such a case, we initialized
522 * the variable to -1 and that is what will be
523 * returned, allowing the caller to terminate early.
525 *never_interesting = 0;
528 if (pathlen > matchlen)
529 return 0;
531 if (matchlen > pathlen) {
532 if (match[pathlen] != '/')
533 return 0;
534 if (!S_ISDIR(entry->mode))
535 return 0;
538 if (m == -1)
540 * we cheated and did not do strncmp(), so we do
541 * that here.
543 m = strncmp(match, entry->path, pathlen);
546 * If common part matched earlier then it is a hit,
547 * because we rejected the case where path is not a
548 * leading directory and is shorter than match.
550 if (!m)
551 return 1;
553 return 0;
556 static int match_dir_prefix(const char *base,
557 const char *match, int matchlen)
559 if (strncmp(base, match, matchlen))
560 return 0;
563 * If the base is a subdirectory of a path which
564 * was specified, all of them are interesting.
566 if (!matchlen ||
567 base[matchlen] == '/' ||
568 match[matchlen - 1] == '/')
569 return 1;
571 /* Just a random prefix match */
572 return 0;
576 * Is a tree entry interesting given the pathspec we have?
578 * Pre-condition: either baselen == base_offset (i.e. empty path)
579 * or base[baselen-1] == '/' (i.e. with trailing slash).
581 enum interesting tree_entry_interesting(const struct name_entry *entry,
582 struct strbuf *base, int base_offset,
583 const struct pathspec *ps)
585 int i;
586 int pathlen, baselen = base->len - base_offset;
587 int never_interesting = ps->has_wildcard ?
588 entry_not_interesting : all_entries_not_interesting;
590 if (!ps->nr) {
591 if (!ps->recursive || ps->max_depth == -1)
592 return all_entries_interesting;
593 return within_depth(base->buf + base_offset, baselen,
594 !!S_ISDIR(entry->mode),
595 ps->max_depth) ?
596 entry_interesting : entry_not_interesting;
599 pathlen = tree_entry_len(entry);
601 for (i = ps->nr - 1; i >= 0; i--) {
602 const struct pathspec_item *item = ps->items+i;
603 const char *match = item->match;
604 const char *base_str = base->buf + base_offset;
605 int matchlen = item->len;
607 if (baselen >= matchlen) {
608 /* If it doesn't match, move along... */
609 if (!match_dir_prefix(base_str, match, matchlen))
610 goto match_wildcards;
612 if (!ps->recursive || ps->max_depth == -1)
613 return all_entries_interesting;
615 return within_depth(base_str + matchlen + 1,
616 baselen - matchlen - 1,
617 !!S_ISDIR(entry->mode),
618 ps->max_depth) ?
619 entry_interesting : entry_not_interesting;
622 /* Either there must be no base, or the base must match. */
623 if (baselen == 0 || !strncmp(base_str, match, baselen)) {
624 if (match_entry(entry, pathlen,
625 match + baselen, matchlen - baselen,
626 &never_interesting))
627 return entry_interesting;
629 if (item->use_wildcard) {
630 if (!fnmatch(match + baselen, entry->path, 0))
631 return entry_interesting;
634 * Match all directories. We'll try to
635 * match files later on.
637 if (ps->recursive && S_ISDIR(entry->mode))
638 return entry_interesting;
641 continue;
644 match_wildcards:
645 if (!item->use_wildcard)
646 continue;
649 * Concatenate base and entry->path into one and do
650 * fnmatch() on it.
653 strbuf_add(base, entry->path, pathlen);
655 if (!fnmatch(match, base->buf + base_offset, 0)) {
656 strbuf_setlen(base, base_offset + baselen);
657 return entry_interesting;
659 strbuf_setlen(base, base_offset + baselen);
662 * Match all directories. We'll try to match files
663 * later on.
664 * max_depth is ignored but we may consider support it
665 * in future, see
666 * http://thread.gmane.org/gmane.comp.version-control.git/163757/focus=163840
668 if (ps->recursive && S_ISDIR(entry->mode))
669 return entry_interesting;
671 return never_interesting; /* No matches */