gitweb: use new Git::Repo API, and add optional caching
[git/gitweb-caching.git] / cache-tree.c
blob5f8ee87bb1c446341b640c2f978a658d6bfcfcd0
1 #include "cache.h"
2 #include "tree.h"
3 #include "cache-tree.h"
5 #ifndef DEBUG
6 #define DEBUG 0
7 #endif
9 struct cache_tree *cache_tree(void)
11 struct cache_tree *it = xcalloc(1, sizeof(struct cache_tree));
12 it->entry_count = -1;
13 return it;
16 void cache_tree_free(struct cache_tree **it_p)
18 int i;
19 struct cache_tree *it = *it_p;
21 if (!it)
22 return;
23 for (i = 0; i < it->subtree_nr; i++)
24 if (it->down[i])
25 cache_tree_free(&it->down[i]->cache_tree);
26 free(it->down);
27 free(it);
28 *it_p = NULL;
31 static int subtree_name_cmp(const char *one, int onelen,
32 const char *two, int twolen)
34 if (onelen < twolen)
35 return -1;
36 if (twolen < onelen)
37 return 1;
38 return memcmp(one, two, onelen);
41 static int subtree_pos(struct cache_tree *it, const char *path, int pathlen)
43 struct cache_tree_sub **down = it->down;
44 int lo, hi;
45 lo = 0;
46 hi = it->subtree_nr;
47 while (lo < hi) {
48 int mi = (lo + hi) / 2;
49 struct cache_tree_sub *mdl = down[mi];
50 int cmp = subtree_name_cmp(path, pathlen,
51 mdl->name, mdl->namelen);
52 if (!cmp)
53 return mi;
54 if (cmp < 0)
55 hi = mi;
56 else
57 lo = mi + 1;
59 return -lo-1;
62 static struct cache_tree_sub *find_subtree(struct cache_tree *it,
63 const char *path,
64 int pathlen,
65 int create)
67 struct cache_tree_sub *down;
68 int pos = subtree_pos(it, path, pathlen);
69 if (0 <= pos)
70 return it->down[pos];
71 if (!create)
72 return NULL;
74 pos = -pos-1;
75 if (it->subtree_alloc <= it->subtree_nr) {
76 it->subtree_alloc = alloc_nr(it->subtree_alloc);
77 it->down = xrealloc(it->down, it->subtree_alloc *
78 sizeof(*it->down));
80 it->subtree_nr++;
82 down = xmalloc(sizeof(*down) + pathlen + 1);
83 down->cache_tree = NULL;
84 down->namelen = pathlen;
85 memcpy(down->name, path, pathlen);
86 down->name[pathlen] = 0;
88 if (pos < it->subtree_nr)
89 memmove(it->down + pos + 1,
90 it->down + pos,
91 sizeof(down) * (it->subtree_nr - pos - 1));
92 it->down[pos] = down;
93 return down;
96 struct cache_tree_sub *cache_tree_sub(struct cache_tree *it, const char *path)
98 int pathlen = strlen(path);
99 return find_subtree(it, path, pathlen, 1);
102 void cache_tree_invalidate_path(struct cache_tree *it, const char *path)
104 /* a/b/c
105 * ==> invalidate self
106 * ==> find "a", have it invalidate "b/c"
108 * ==> invalidate self
109 * ==> if "a" exists as a subtree, remove it.
111 const char *slash;
112 int namelen;
113 struct cache_tree_sub *down;
115 #if DEBUG
116 fprintf(stderr, "cache-tree invalidate <%s>\n", path);
117 #endif
119 if (!it)
120 return;
121 slash = strchr(path, '/');
122 it->entry_count = -1;
123 if (!slash) {
124 int pos;
125 namelen = strlen(path);
126 pos = subtree_pos(it, path, namelen);
127 if (0 <= pos) {
128 cache_tree_free(&it->down[pos]->cache_tree);
129 free(it->down[pos]);
130 /* 0 1 2 3 4 5
131 * ^ ^subtree_nr = 6
132 * pos
133 * move 4 and 5 up one place (2 entries)
134 * 2 = 6 - 3 - 1 = subtree_nr - pos - 1
136 memmove(it->down+pos, it->down+pos+1,
137 sizeof(struct cache_tree_sub *) *
138 (it->subtree_nr - pos - 1));
139 it->subtree_nr--;
141 return;
143 namelen = slash - path;
144 down = find_subtree(it, path, namelen, 0);
145 if (down)
146 cache_tree_invalidate_path(down->cache_tree, slash + 1);
149 static int verify_cache(struct cache_entry **cache,
150 int entries)
152 int i, funny;
154 /* Verify that the tree is merged */
155 funny = 0;
156 for (i = 0; i < entries; i++) {
157 struct cache_entry *ce = cache[i];
158 if (ce_stage(ce)) {
159 if (10 < ++funny) {
160 fprintf(stderr, "...\n");
161 break;
163 fprintf(stderr, "%s: unmerged (%s)\n",
164 ce->name, sha1_to_hex(ce->sha1));
167 if (funny)
168 return -1;
170 /* Also verify that the cache does not have path and path/file
171 * at the same time. At this point we know the cache has only
172 * stage 0 entries.
174 funny = 0;
175 for (i = 0; i < entries - 1; i++) {
176 /* path/file always comes after path because of the way
177 * the cache is sorted. Also path can appear only once,
178 * which means conflicting one would immediately follow.
180 const char *this_name = cache[i]->name;
181 const char *next_name = cache[i+1]->name;
182 int this_len = strlen(this_name);
183 if (this_len < strlen(next_name) &&
184 strncmp(this_name, next_name, this_len) == 0 &&
185 next_name[this_len] == '/') {
186 if (10 < ++funny) {
187 fprintf(stderr, "...\n");
188 break;
190 fprintf(stderr, "You have both %s and %s\n",
191 this_name, next_name);
194 if (funny)
195 return -1;
196 return 0;
199 static void discard_unused_subtrees(struct cache_tree *it)
201 struct cache_tree_sub **down = it->down;
202 int nr = it->subtree_nr;
203 int dst, src;
204 for (dst = src = 0; src < nr; src++) {
205 struct cache_tree_sub *s = down[src];
206 if (s->used)
207 down[dst++] = s;
208 else {
209 cache_tree_free(&s->cache_tree);
210 free(s);
211 it->subtree_nr--;
216 int cache_tree_fully_valid(struct cache_tree *it)
218 int i;
219 if (!it)
220 return 0;
221 if (it->entry_count < 0 || !has_sha1_file(it->sha1))
222 return 0;
223 for (i = 0; i < it->subtree_nr; i++) {
224 if (!cache_tree_fully_valid(it->down[i]->cache_tree))
225 return 0;
227 return 1;
230 static int update_one(struct cache_tree *it,
231 struct cache_entry **cache,
232 int entries,
233 const char *base,
234 int baselen,
235 int missing_ok,
236 int dryrun)
238 struct strbuf buffer;
239 int i;
241 if (0 <= it->entry_count && has_sha1_file(it->sha1))
242 return it->entry_count;
245 * We first scan for subtrees and update them; we start by
246 * marking existing subtrees -- the ones that are unmarked
247 * should not be in the result.
249 for (i = 0; i < it->subtree_nr; i++)
250 it->down[i]->used = 0;
253 * Find the subtrees and update them.
255 for (i = 0; i < entries; i++) {
256 struct cache_entry *ce = cache[i];
257 struct cache_tree_sub *sub;
258 const char *path, *slash;
259 int pathlen, sublen, subcnt;
261 path = ce->name;
262 pathlen = ce_namelen(ce);
263 if (pathlen <= baselen || memcmp(base, path, baselen))
264 break; /* at the end of this level */
266 slash = strchr(path + baselen, '/');
267 if (!slash)
268 continue;
270 * a/bbb/c (base = a/, slash = /c)
271 * ==>
272 * path+baselen = bbb/c, sublen = 3
274 sublen = slash - (path + baselen);
275 sub = find_subtree(it, path + baselen, sublen, 1);
276 if (!sub->cache_tree)
277 sub->cache_tree = cache_tree();
278 subcnt = update_one(sub->cache_tree,
279 cache + i, entries - i,
280 path,
281 baselen + sublen + 1,
282 missing_ok,
283 dryrun);
284 if (subcnt < 0)
285 return subcnt;
286 i += subcnt - 1;
287 sub->used = 1;
290 discard_unused_subtrees(it);
293 * Then write out the tree object for this level.
295 strbuf_init(&buffer, 8192);
297 for (i = 0; i < entries; i++) {
298 struct cache_entry *ce = cache[i];
299 struct cache_tree_sub *sub;
300 const char *path, *slash;
301 int pathlen, entlen;
302 const unsigned char *sha1;
303 unsigned mode;
305 path = ce->name;
306 pathlen = ce_namelen(ce);
307 if (pathlen <= baselen || memcmp(base, path, baselen))
308 break; /* at the end of this level */
310 slash = strchr(path + baselen, '/');
311 if (slash) {
312 entlen = slash - (path + baselen);
313 sub = find_subtree(it, path + baselen, entlen, 0);
314 if (!sub)
315 die("cache-tree.c: '%.*s' in '%s' not found",
316 entlen, path + baselen, path);
317 i += sub->cache_tree->entry_count - 1;
318 sha1 = sub->cache_tree->sha1;
319 mode = S_IFDIR;
321 else {
322 sha1 = ce->sha1;
323 mode = ce->ce_mode;
324 entlen = pathlen - baselen;
326 if (mode != S_IFGITLINK && !missing_ok && !has_sha1_file(sha1))
327 return error("invalid object %s", sha1_to_hex(sha1));
329 if (ce->ce_flags & CE_REMOVE)
330 continue; /* entry being removed */
332 strbuf_grow(&buffer, entlen + 100);
333 strbuf_addf(&buffer, "%o %.*s%c", mode, entlen, path + baselen, '\0');
334 strbuf_add(&buffer, sha1, 20);
336 #if DEBUG
337 fprintf(stderr, "cache-tree update-one %o %.*s\n",
338 mode, entlen, path + baselen);
339 #endif
342 if (dryrun)
343 hash_sha1_file(buffer.buf, buffer.len, tree_type, it->sha1);
344 else if (write_sha1_file(buffer.buf, buffer.len, tree_type, it->sha1)) {
345 strbuf_release(&buffer);
346 return -1;
349 strbuf_release(&buffer);
350 it->entry_count = i;
351 #if DEBUG
352 fprintf(stderr, "cache-tree update-one (%d ent, %d subtree) %s\n",
353 it->entry_count, it->subtree_nr,
354 sha1_to_hex(it->sha1));
355 #endif
356 return i;
359 int cache_tree_update(struct cache_tree *it,
360 struct cache_entry **cache,
361 int entries,
362 int missing_ok,
363 int dryrun)
365 int i;
366 i = verify_cache(cache, entries);
367 if (i)
368 return i;
369 i = update_one(it, cache, entries, "", 0, missing_ok, dryrun);
370 if (i < 0)
371 return i;
372 return 0;
375 static void write_one(struct strbuf *buffer, struct cache_tree *it,
376 const char *path, int pathlen)
378 int i;
380 /* One "cache-tree" entry consists of the following:
381 * path (NUL terminated)
382 * entry_count, subtree_nr ("%d %d\n")
383 * tree-sha1 (missing if invalid)
384 * subtree_nr "cache-tree" entries for subtrees.
386 strbuf_grow(buffer, pathlen + 100);
387 strbuf_add(buffer, path, pathlen);
388 strbuf_addf(buffer, "%c%d %d\n", 0, it->entry_count, it->subtree_nr);
390 #if DEBUG
391 if (0 <= it->entry_count)
392 fprintf(stderr, "cache-tree <%.*s> (%d ent, %d subtree) %s\n",
393 pathlen, path, it->entry_count, it->subtree_nr,
394 sha1_to_hex(it->sha1));
395 else
396 fprintf(stderr, "cache-tree <%.*s> (%d subtree) invalid\n",
397 pathlen, path, it->subtree_nr);
398 #endif
400 if (0 <= it->entry_count) {
401 strbuf_add(buffer, it->sha1, 20);
403 for (i = 0; i < it->subtree_nr; i++) {
404 struct cache_tree_sub *down = it->down[i];
405 if (i) {
406 struct cache_tree_sub *prev = it->down[i-1];
407 if (subtree_name_cmp(down->name, down->namelen,
408 prev->name, prev->namelen) <= 0)
409 die("fatal - unsorted cache subtree");
411 write_one(buffer, down->cache_tree, down->name, down->namelen);
415 void cache_tree_write(struct strbuf *sb, struct cache_tree *root)
417 write_one(sb, root, "", 0);
420 static struct cache_tree *read_one(const char **buffer, unsigned long *size_p)
422 const char *buf = *buffer;
423 unsigned long size = *size_p;
424 const char *cp;
425 char *ep;
426 struct cache_tree *it;
427 int i, subtree_nr;
429 it = NULL;
430 /* skip name, but make sure name exists */
431 while (size && *buf) {
432 size--;
433 buf++;
435 if (!size)
436 goto free_return;
437 buf++; size--;
438 it = cache_tree();
440 cp = buf;
441 it->entry_count = strtol(cp, &ep, 10);
442 if (cp == ep)
443 goto free_return;
444 cp = ep;
445 subtree_nr = strtol(cp, &ep, 10);
446 if (cp == ep)
447 goto free_return;
448 while (size && *buf && *buf != '\n') {
449 size--;
450 buf++;
452 if (!size)
453 goto free_return;
454 buf++; size--;
455 if (0 <= it->entry_count) {
456 if (size < 20)
457 goto free_return;
458 hashcpy(it->sha1, (const unsigned char*)buf);
459 buf += 20;
460 size -= 20;
463 #if DEBUG
464 if (0 <= it->entry_count)
465 fprintf(stderr, "cache-tree <%s> (%d ent, %d subtree) %s\n",
466 *buffer, it->entry_count, subtree_nr,
467 sha1_to_hex(it->sha1));
468 else
469 fprintf(stderr, "cache-tree <%s> (%d subtrees) invalid\n",
470 *buffer, subtree_nr);
471 #endif
474 * Just a heuristic -- we do not add directories that often but
475 * we do not want to have to extend it immediately when we do,
476 * hence +2.
478 it->subtree_alloc = subtree_nr + 2;
479 it->down = xcalloc(it->subtree_alloc, sizeof(struct cache_tree_sub *));
480 for (i = 0; i < subtree_nr; i++) {
481 /* read each subtree */
482 struct cache_tree *sub;
483 struct cache_tree_sub *subtree;
484 const char *name = buf;
486 sub = read_one(&buf, &size);
487 if (!sub)
488 goto free_return;
489 subtree = cache_tree_sub(it, name);
490 subtree->cache_tree = sub;
492 if (subtree_nr != it->subtree_nr)
493 die("cache-tree: internal error");
494 *buffer = buf;
495 *size_p = size;
496 return it;
498 free_return:
499 cache_tree_free(&it);
500 return NULL;
503 struct cache_tree *cache_tree_read(const char *buffer, unsigned long size)
505 if (buffer[0])
506 return NULL; /* not the whole tree */
507 return read_one(&buffer, &size);
510 static struct cache_tree *cache_tree_find(struct cache_tree *it, const char *path)
512 while (*path) {
513 const char *slash;
514 struct cache_tree_sub *sub;
516 slash = strchr(path, '/');
517 if (!slash)
518 slash = path + strlen(path);
519 /* between path and slash is the name of the
520 * subtree to look for.
522 sub = find_subtree(it, path, slash - path, 0);
523 if (!sub)
524 return NULL;
525 it = sub->cache_tree;
526 if (slash)
527 while (*slash && *slash == '/')
528 slash++;
529 if (!slash || !*slash)
530 return it; /* prefix ended with slashes */
531 path = slash;
533 return it;
536 int write_cache_as_tree(unsigned char *sha1, int missing_ok, const char *prefix)
538 int entries, was_valid, newfd;
541 * We can't free this memory, it becomes part of a linked list
542 * parsed atexit()
544 struct lock_file *lock_file = xcalloc(1, sizeof(struct lock_file));
546 newfd = hold_locked_index(lock_file, 1);
548 entries = read_cache();
549 if (entries < 0)
550 return WRITE_TREE_UNREADABLE_INDEX;
552 if (!active_cache_tree)
553 active_cache_tree = cache_tree();
555 was_valid = cache_tree_fully_valid(active_cache_tree);
557 if (!was_valid) {
558 if (cache_tree_update(active_cache_tree,
559 active_cache, active_nr,
560 missing_ok, 0) < 0)
561 return WRITE_TREE_UNMERGED_INDEX;
562 if (0 <= newfd) {
563 if (!write_cache(newfd, active_cache, active_nr) &&
564 !commit_lock_file(lock_file))
565 newfd = -1;
567 /* Not being able to write is fine -- we are only interested
568 * in updating the cache-tree part, and if the next caller
569 * ends up using the old index with unupdated cache-tree part
570 * it misses the work we did here, but that is just a
571 * performance penalty and not a big deal.
575 if (prefix) {
576 struct cache_tree *subtree =
577 cache_tree_find(active_cache_tree, prefix);
578 if (!subtree)
579 return WRITE_TREE_PREFIX_ERROR;
580 hashcpy(sha1, subtree->sha1);
582 else
583 hashcpy(sha1, active_cache_tree->sha1);
585 if (0 <= newfd)
586 rollback_lock_file(lock_file);
588 return 0;