Use strbuf API in cache-tree.c
[git] / cache-tree.c
1 #include "cache.h"
2 #include "strbuf.h"
3 #include "tree.h"
4 #include "cache-tree.h"
5
6 #ifndef DEBUG
7 #define DEBUG 0
8 #endif
9
10 struct cache_tree *cache_tree(void)
11 {
12         struct cache_tree *it = xcalloc(1, sizeof(struct cache_tree));
13         it->entry_count = -1;
14         return it;
15 }
16
17 void cache_tree_free(struct cache_tree **it_p)
18 {
19         int i;
20         struct cache_tree *it = *it_p;
21
22         if (!it)
23                 return;
24         for (i = 0; i < it->subtree_nr; i++)
25                 if (it->down[i])
26                         cache_tree_free(&it->down[i]->cache_tree);
27         free(it->down);
28         free(it);
29         *it_p = NULL;
30 }
31
32 static int subtree_name_cmp(const char *one, int onelen,
33                             const char *two, int twolen)
34 {
35         if (onelen < twolen)
36                 return -1;
37         if (twolen < onelen)
38                 return 1;
39         return memcmp(one, two, onelen);
40 }
41
42 static int subtree_pos(struct cache_tree *it, const char *path, int pathlen)
43 {
44         struct cache_tree_sub **down = it->down;
45         int lo, hi;
46         lo = 0;
47         hi = it->subtree_nr;
48         while (lo < hi) {
49                 int mi = (lo + hi) / 2;
50                 struct cache_tree_sub *mdl = down[mi];
51                 int cmp = subtree_name_cmp(path, pathlen,
52                                            mdl->name, mdl->namelen);
53                 if (!cmp)
54                         return mi;
55                 if (cmp < 0)
56                         hi = mi;
57                 else
58                         lo = mi + 1;
59         }
60         return -lo-1;
61 }
62
63 static struct cache_tree_sub *find_subtree(struct cache_tree *it,
64                                            const char *path,
65                                            int pathlen,
66                                            int create)
67 {
68         struct cache_tree_sub *down;
69         int pos = subtree_pos(it, path, pathlen);
70         if (0 <= pos)
71                 return it->down[pos];
72         if (!create)
73                 return NULL;
74
75         pos = -pos-1;
76         if (it->subtree_alloc <= it->subtree_nr) {
77                 it->subtree_alloc = alloc_nr(it->subtree_alloc);
78                 it->down = xrealloc(it->down, it->subtree_alloc *
79                                     sizeof(*it->down));
80         }
81         it->subtree_nr++;
82
83         down = xmalloc(sizeof(*down) + pathlen + 1);
84         down->cache_tree = NULL;
85         down->namelen = pathlen;
86         memcpy(down->name, path, pathlen);
87         down->name[pathlen] = 0;
88
89         if (pos < it->subtree_nr)
90                 memmove(it->down + pos + 1,
91                         it->down + pos,
92                         sizeof(down) * (it->subtree_nr - pos - 1));
93         it->down[pos] = down;
94         return down;
95 }
96
97 struct cache_tree_sub *cache_tree_sub(struct cache_tree *it, const char *path)
98 {
99         int pathlen = strlen(path);
100         return find_subtree(it, path, pathlen, 1);
101 }
102
103 void cache_tree_invalidate_path(struct cache_tree *it, const char *path)
104 {
105         /* a/b/c
106          * ==> invalidate self
107          * ==> find "a", have it invalidate "b/c"
108          * a
109          * ==> invalidate self
110          * ==> if "a" exists as a subtree, remove it.
111          */
112         const char *slash;
113         int namelen;
114         struct cache_tree_sub *down;
115
116 #if DEBUG
117         fprintf(stderr, "cache-tree invalidate <%s>\n", path);
118 #endif
119
120         if (!it)
121                 return;
122         slash = strchr(path, '/');
123         it->entry_count = -1;
124         if (!slash) {
125                 int pos;
126                 namelen = strlen(path);
127                 pos = subtree_pos(it, path, namelen);
128                 if (0 <= pos) {
129                         cache_tree_free(&it->down[pos]->cache_tree);
130                         free(it->down[pos]);
131                         /* 0 1 2 3 4 5
132                          *       ^     ^subtree_nr = 6
133                          *       pos
134                          * move 4 and 5 up one place (2 entries)
135                          * 2 = 6 - 3 - 1 = subtree_nr - pos - 1
136                          */
137                         memmove(it->down+pos, it->down+pos+1,
138                                 sizeof(struct cache_tree_sub *) *
139                                 (it->subtree_nr - pos - 1));
140                         it->subtree_nr--;
141                 }
142                 return;
143         }
144         namelen = slash - path;
145         down = find_subtree(it, path, namelen, 0);
146         if (down)
147                 cache_tree_invalidate_path(down->cache_tree, slash + 1);
148 }
149
150 static int verify_cache(struct cache_entry **cache,
151                         int entries)
152 {
153         int i, funny;
154
155         /* Verify that the tree is merged */
156         funny = 0;
157         for (i = 0; i < entries; i++) {
158                 struct cache_entry *ce = cache[i];
159                 if (ce_stage(ce)) {
160                         if (10 < ++funny) {
161                                 fprintf(stderr, "...\n");
162                                 break;
163                         }
164                         fprintf(stderr, "%s: unmerged (%s)\n",
165                                 ce->name, sha1_to_hex(ce->sha1));
166                 }
167         }
168         if (funny)
169                 return -1;
170
171         /* Also verify that the cache does not have path and path/file
172          * at the same time.  At this point we know the cache has only
173          * stage 0 entries.
174          */
175         funny = 0;
176         for (i = 0; i < entries - 1; i++) {
177                 /* path/file always comes after path because of the way
178                  * the cache is sorted.  Also path can appear only once,
179                  * which means conflicting one would immediately follow.
180                  */
181                 const char *this_name = cache[i]->name;
182                 const char *next_name = cache[i+1]->name;
183                 int this_len = strlen(this_name);
184                 if (this_len < strlen(next_name) &&
185                     strncmp(this_name, next_name, this_len) == 0 &&
186                     next_name[this_len] == '/') {
187                         if (10 < ++funny) {
188                                 fprintf(stderr, "...\n");
189                                 break;
190                         }
191                         fprintf(stderr, "You have both %s and %s\n",
192                                 this_name, next_name);
193                 }
194         }
195         if (funny)
196                 return -1;
197         return 0;
198 }
199
200 static void discard_unused_subtrees(struct cache_tree *it)
201 {
202         struct cache_tree_sub **down = it->down;
203         int nr = it->subtree_nr;
204         int dst, src;
205         for (dst = src = 0; src < nr; src++) {
206                 struct cache_tree_sub *s = down[src];
207                 if (s->used)
208                         down[dst++] = s;
209                 else {
210                         cache_tree_free(&s->cache_tree);
211                         free(s);
212                         it->subtree_nr--;
213                 }
214         }
215 }
216
217 int cache_tree_fully_valid(struct cache_tree *it)
218 {
219         int i;
220         if (!it)
221                 return 0;
222         if (it->entry_count < 0 || !has_sha1_file(it->sha1))
223                 return 0;
224         for (i = 0; i < it->subtree_nr; i++) {
225                 if (!cache_tree_fully_valid(it->down[i]->cache_tree))
226                         return 0;
227         }
228         return 1;
229 }
230
231 static int update_one(struct cache_tree *it,
232                       struct cache_entry **cache,
233                       int entries,
234                       const char *base,
235                       int baselen,
236                       int missing_ok,
237                       int dryrun)
238 {
239         struct strbuf buffer;
240         int i;
241
242         if (0 <= it->entry_count && has_sha1_file(it->sha1))
243                 return it->entry_count;
244
245         /*
246          * We first scan for subtrees and update them; we start by
247          * marking existing subtrees -- the ones that are unmarked
248          * should not be in the result.
249          */
250         for (i = 0; i < it->subtree_nr; i++)
251                 it->down[i]->used = 0;
252
253         /*
254          * Find the subtrees and update them.
255          */
256         for (i = 0; i < entries; i++) {
257                 struct cache_entry *ce = cache[i];
258                 struct cache_tree_sub *sub;
259                 const char *path, *slash;
260                 int pathlen, sublen, subcnt;
261
262                 path = ce->name;
263                 pathlen = ce_namelen(ce);
264                 if (pathlen <= baselen || memcmp(base, path, baselen))
265                         break; /* at the end of this level */
266
267                 slash = strchr(path + baselen, '/');
268                 if (!slash)
269                         continue;
270                 /*
271                  * a/bbb/c (base = a/, slash = /c)
272                  * ==>
273                  * path+baselen = bbb/c, sublen = 3
274                  */
275                 sublen = slash - (path + baselen);
276                 sub = find_subtree(it, path + baselen, sublen, 1);
277                 if (!sub->cache_tree)
278                         sub->cache_tree = cache_tree();
279                 subcnt = update_one(sub->cache_tree,
280                                     cache + i, entries - i,
281                                     path,
282                                     baselen + sublen + 1,
283                                     missing_ok,
284                                     dryrun);
285                 if (subcnt < 0)
286                         return subcnt;
287                 i += subcnt - 1;
288                 sub->used = 1;
289         }
290
291         discard_unused_subtrees(it);
292
293         /*
294          * Then write out the tree object for this level.
295          */
296         strbuf_init(&buffer);
297         strbuf_grow(&buffer, 8192);
298
299         for (i = 0; i < entries; i++) {
300                 struct cache_entry *ce = cache[i];
301                 struct cache_tree_sub *sub;
302                 const char *path, *slash;
303                 int pathlen, entlen;
304                 const unsigned char *sha1;
305                 unsigned mode;
306
307                 path = ce->name;
308                 pathlen = ce_namelen(ce);
309                 if (pathlen <= baselen || memcmp(base, path, baselen))
310                         break; /* at the end of this level */
311
312                 slash = strchr(path + baselen, '/');
313                 if (slash) {
314                         entlen = slash - (path + baselen);
315                         sub = find_subtree(it, path + baselen, entlen, 0);
316                         if (!sub)
317                                 die("cache-tree.c: '%.*s' in '%s' not found",
318                                     entlen, path + baselen, path);
319                         i += sub->cache_tree->entry_count - 1;
320                         sha1 = sub->cache_tree->sha1;
321                         mode = S_IFDIR;
322                 }
323                 else {
324                         sha1 = ce->sha1;
325                         mode = ntohl(ce->ce_mode);
326                         entlen = pathlen - baselen;
327                 }
328                 if (mode != S_IFGITLINK && !missing_ok && !has_sha1_file(sha1))
329                         return error("invalid object %s", sha1_to_hex(sha1));
330
331                 if (!ce->ce_mode)
332                         continue; /* entry being removed */
333
334                 strbuf_grow(&buffer, entlen + 100);
335                 strbuf_addf(&buffer, "%o %.*s%c", mode, entlen, path + baselen, '\0');
336                 strbuf_add(&buffer, sha1, 20);
337
338 #if DEBUG
339                 fprintf(stderr, "cache-tree update-one %o %.*s\n",
340                         mode, entlen, path + baselen);
341 #endif
342         }
343
344         if (dryrun)
345                 hash_sha1_file(buffer.buf, buffer.len, tree_type, it->sha1);
346         else
347                 write_sha1_file(buffer.buf, buffer.len, tree_type, it->sha1);
348         strbuf_release(&buffer);
349         it->entry_count = i;
350 #if DEBUG
351         fprintf(stderr, "cache-tree update-one (%d ent, %d subtree) %s\n",
352                 it->entry_count, it->subtree_nr,
353                 sha1_to_hex(it->sha1));
354 #endif
355         return i;
356 }
357
358 int cache_tree_update(struct cache_tree *it,
359                       struct cache_entry **cache,
360                       int entries,
361                       int missing_ok,
362                       int dryrun)
363 {
364         int i;
365         i = verify_cache(cache, entries);
366         if (i)
367                 return i;
368         i = update_one(it, cache, entries, "", 0, missing_ok, dryrun);
369         if (i < 0)
370                 return i;
371         return 0;
372 }
373
374 static void write_one(struct cache_tree *it,
375                        char *path,
376                        int pathlen,
377                            struct strbuf *buffer)
378 {
379         int i;
380
381         /* One "cache-tree" entry consists of the following:
382          * path (NUL terminated)
383          * entry_count, subtree_nr ("%d %d\n")
384          * tree-sha1 (missing if invalid)
385          * subtree_nr "cache-tree" entries for subtrees.
386          */
387         strbuf_grow(buffer, pathlen + 100);
388         strbuf_add(buffer, path, pathlen);
389         strbuf_addf(buffer, "%c%d %d\n", 0, it->entry_count, it->subtree_nr);
390
391 #if DEBUG
392         if (0 <= it->entry_count)
393                 fprintf(stderr, "cache-tree <%.*s> (%d ent, %d subtree) %s\n",
394                         pathlen, path, it->entry_count, it->subtree_nr,
395                         sha1_to_hex(it->sha1));
396         else
397                 fprintf(stderr, "cache-tree <%.*s> (%d subtree) invalid\n",
398                         pathlen, path, it->subtree_nr);
399 #endif
400
401         if (0 <= it->entry_count) {
402                 strbuf_add(buffer, it->sha1, 20);
403         }
404         for (i = 0; i < it->subtree_nr; i++) {
405                 struct cache_tree_sub *down = it->down[i];
406                 if (i) {
407                         struct cache_tree_sub *prev = it->down[i-1];
408                         if (subtree_name_cmp(down->name, down->namelen,
409                                              prev->name, prev->namelen) <= 0)
410                                 die("fatal - unsorted cache subtree");
411                 }
412                 write_one(down->cache_tree, down->name, down->namelen, buffer);
413         }
414 }
415
416 void *cache_tree_write(struct cache_tree *root, unsigned long *size_p)
417 {
418         char path[PATH_MAX];
419         struct strbuf buffer;
420
421         path[0] = 0;
422         strbuf_init(&buffer);
423         write_one(root, path, 0, &buffer);
424         *size_p = buffer.len;
425         return strbuf_detach(&buffer);
426 }
427
428 static struct cache_tree *read_one(const char **buffer, unsigned long *size_p)
429 {
430         const char *buf = *buffer;
431         unsigned long size = *size_p;
432         const char *cp;
433         char *ep;
434         struct cache_tree *it;
435         int i, subtree_nr;
436
437         it = NULL;
438         /* skip name, but make sure name exists */
439         while (size && *buf) {
440                 size--;
441                 buf++;
442         }
443         if (!size)
444                 goto free_return;
445         buf++; size--;
446         it = cache_tree();
447
448         cp = buf;
449         it->entry_count = strtol(cp, &ep, 10);
450         if (cp == ep)
451                 goto free_return;
452         cp = ep;
453         subtree_nr = strtol(cp, &ep, 10);
454         if (cp == ep)
455                 goto free_return;
456         while (size && *buf && *buf != '\n') {
457                 size--;
458                 buf++;
459         }
460         if (!size)
461                 goto free_return;
462         buf++; size--;
463         if (0 <= it->entry_count) {
464                 if (size < 20)
465                         goto free_return;
466                 hashcpy(it->sha1, (const unsigned char*)buf);
467                 buf += 20;
468                 size -= 20;
469         }
470
471 #if DEBUG
472         if (0 <= it->entry_count)
473                 fprintf(stderr, "cache-tree <%s> (%d ent, %d subtree) %s\n",
474                         *buffer, it->entry_count, subtree_nr,
475                         sha1_to_hex(it->sha1));
476         else
477                 fprintf(stderr, "cache-tree <%s> (%d subtrees) invalid\n",
478                         *buffer, subtree_nr);
479 #endif
480
481         /*
482          * Just a heuristic -- we do not add directories that often but
483          * we do not want to have to extend it immediately when we do,
484          * hence +2.
485          */
486         it->subtree_alloc = subtree_nr + 2;
487         it->down = xcalloc(it->subtree_alloc, sizeof(struct cache_tree_sub *));
488         for (i = 0; i < subtree_nr; i++) {
489                 /* read each subtree */
490                 struct cache_tree *sub;
491                 struct cache_tree_sub *subtree;
492                 const char *name = buf;
493
494                 sub = read_one(&buf, &size);
495                 if (!sub)
496                         goto free_return;
497                 subtree = cache_tree_sub(it, name);
498                 subtree->cache_tree = sub;
499         }
500         if (subtree_nr != it->subtree_nr)
501                 die("cache-tree: internal error");
502         *buffer = buf;
503         *size_p = size;
504         return it;
505
506  free_return:
507         cache_tree_free(&it);
508         return NULL;
509 }
510
511 struct cache_tree *cache_tree_read(const char *buffer, unsigned long size)
512 {
513         if (buffer[0])
514                 return NULL; /* not the whole tree */
515         return read_one(&buffer, &size);
516 }
517
518 struct cache_tree *cache_tree_find(struct cache_tree *it, const char *path)
519 {
520         while (*path) {
521                 const char *slash;
522                 struct cache_tree_sub *sub;
523
524                 slash = strchr(path, '/');
525                 if (!slash)
526                         slash = path + strlen(path);
527                 /* between path and slash is the name of the
528                  * subtree to look for.
529                  */
530                 sub = find_subtree(it, path, slash - path, 0);
531                 if (!sub)
532                         return NULL;
533                 it = sub->cache_tree;
534                 if (slash)
535                         while (*slash && *slash == '/')
536                                 slash++;
537                 if (!slash || !*slash)
538                         return it; /* prefix ended with slashes */
539                 path = slash;
540         }
541         return it;
542 }