git-mv is not able to handle big directories
[gitweb.git] / sha1_name.c
index df3ba2d43f14d9dbcc805a639c3f413f60c2e26a..faac158b16ca978914696ed5d2801770a122cce2 100644 (file)
@@ -1,33 +1,54 @@
 #include "cache.h"
+#include "tag.h"
 #include "commit.h"
+#include "tree.h"
+#include "blob.h"
 
 static int find_short_object_filename(int len, const char *name, unsigned char *sha1)
 {
-       static char dirname[PATH_MAX];
+       struct alternate_object_database *alt;
        char hex[40];
-       DIR *dir;
-       int found;
+       int found = 0;
+       static struct alternate_object_database *fakeent;
+
+       if (!fakeent) {
+               const char *objdir = get_object_directory();
+               int objdir_len = strlen(objdir);
+               int entlen = objdir_len + 43;
+               fakeent = xmalloc(sizeof(*fakeent) + entlen);
+               memcpy(fakeent->base, objdir, objdir_len);
+               fakeent->name = fakeent->base + objdir_len + 1;
+               fakeent->name[-1] = '/';
+       }
+       fakeent->next = alt_odb_list;
 
-       snprintf(dirname, sizeof(dirname), "%s/%.2s", get_object_directory(), name);
-       dir = opendir(dirname);
        sprintf(hex, "%.2s", name);
-       found = 0;
-       if (dir) {
+       for (alt = fakeent; alt && found < 2; alt = alt->next) {
                struct dirent *de;
+               DIR *dir;
+               sprintf(alt->name, "%.2s/", name);
+               dir = opendir(alt->base);
+               if (!dir)
+                       continue;
                while ((de = readdir(dir)) != NULL) {
                        if (strlen(de->d_name) != 38)
                                continue;
-                       if (memcmp(de->d_name, name + 2, len-2))
+                       if (memcmp(de->d_name, name + 2, len - 2))
                                continue;
-                       memcpy(hex + 2, de->d_name, 38);
-                       if (++found > 1)
+                       if (!found) {
+                               memcpy(hex + 2, de->d_name, 38);
+                               found++;
+                       }
+                       else if (memcmp(hex + 2, de->d_name, 38)) {
+                               found = 2;
                                break;
+                       }
                }
                closedir(dir);
        }
        if (found == 1)
                return get_sha1_hex(hex, sha1) == 0;
-       return 0;
+       return found;
 }
 
 static int match_sha(unsigned len, const unsigned char *a, const unsigned char *b)
@@ -48,9 +69,11 @@ static int match_sha(unsigned len, const unsigned char *a, const unsigned char *
 static int find_short_packed_object(int len, const unsigned char *match, unsigned char *sha1)
 {
        struct packed_git *p;
+       unsigned char found_sha1[20];
+       int found = 0;
 
        prepare_packed_git();
-       for (p = packed_git; p; p = p->next) {
+       for (p = packed_git; p && found < 2; p = p->next) {
                unsigned num = num_packed_objects(p);
                unsigned first = 0, last = num;
                while (first < last) {
@@ -74,29 +97,71 @@ static int find_short_packed_object(int len, const unsigned char *match, unsigne
                        unsigned char now[20], next[20];
                        nth_packed_object_sha1(p, first, now);
                        if (match_sha(len, match, now)) {
-                               if (nth_packed_object_sha1(p, first+1, next) || !match_sha(len, match, next)) {
-                                       memcpy(sha1, now, 20);
-                                       return 1;
+                               if (nth_packed_object_sha1(p, first+1, next) ||
+                                   !match_sha(len, match, next)) {
+                                       /* unique within this pack */
+                                       if (!found) {
+                                               memcpy(found_sha1, now, 20);
+                                               found++;
+                                       }
+                                       else if (memcmp(found_sha1, now, 20)) {
+                                               found = 2;
+                                               break;
+                                       }
+                               }
+                               else {
+                                       /* not even unique within this pack */
+                                       found = 2;
+                                       break;
                                }
                        }
                }
        }
+       if (found == 1)
+               memcpy(sha1, found_sha1, 20);
+       return found;
+}
+
+#define SHORT_NAME_NOT_FOUND (-1)
+#define SHORT_NAME_AMBIGUOUS (-2)
+
+static int find_unique_short_object(int len, char *canonical,
+                                   unsigned char *res, unsigned char *sha1)
+{
+       int has_unpacked, has_packed;
+       unsigned char unpacked_sha1[20], packed_sha1[20];
+
+       has_unpacked = find_short_object_filename(len, canonical, unpacked_sha1);
+       has_packed = find_short_packed_object(len, res, packed_sha1);
+       if (!has_unpacked && !has_packed)
+               return SHORT_NAME_NOT_FOUND;
+       if (1 < has_unpacked || 1 < has_packed)
+               return SHORT_NAME_AMBIGUOUS;
+       if (has_unpacked != has_packed) {
+               memcpy(sha1, (has_packed ? packed_sha1 : unpacked_sha1), 20);
+               return 0;
+       }
+       /* Both have unique ones -- do they match? */
+       if (memcmp(packed_sha1, unpacked_sha1, 20))
+               return -2;
+       memcpy(sha1, packed_sha1, 20);
        return 0;
 }
 
-static int get_short_sha1(const char *name, unsigned char *sha1)
+static int get_short_sha1(const char *name, int len, unsigned char *sha1,
+                         int quietly)
 {
-       int i;
+       int i, status;
        char canonical[40];
        unsigned char res[20];
 
+       if (len < 4)
+               return -1;
        memset(res, 0, 20);
        memset(canonical, 'x', 40);
-       for (i = 0;;i++) {
+       for (i = 0; i < len ;i++) {
                unsigned char c = name[i];
                unsigned char val;
-               if (!c || i > 40)
-                       break;
                if (c >= '0' && c <= '9')
                        val = c - '0';
                else if (c >= 'a' && c <= 'f')
@@ -112,28 +177,53 @@ static int get_short_sha1(const char *name, unsigned char *sha1)
                        val <<= 4;
                res[i >> 1] |= val;
        }
-       if (i < 4)
-               return -1;
-       if (find_short_object_filename(i, canonical, sha1))
-               return 0;
-       if (find_short_packed_object(i, res, sha1))
-               return 0;
-       return -1;
+
+       status = find_unique_short_object(i, canonical, res, sha1);
+       if (!quietly && (status == SHORT_NAME_AMBIGUOUS))
+               return error("short SHA1 %.*s is ambiguous.", len, canonical);
+       return status;
 }
 
-static int get_sha1_file(const char *path, unsigned char *result)
+const char *find_unique_abbrev(const unsigned char *sha1, int len)
 {
-       char buffer[60];
-       int fd = open(path, O_RDONLY);
-       int len;
+       int status;
+       static char hex[41];
+       memcpy(hex, sha1_to_hex(sha1), 40);
+       while (len < 40) {
+               unsigned char sha1_ret[20];
+               status = get_short_sha1(hex, len, sha1_ret, 1);
+               if (!status) {
+                       hex[len] = 0;
+                       return hex;
+               }
+               if (status != SHORT_NAME_AMBIGUOUS)
+                       return NULL;
+               len++;
+       }
+       return NULL;
+}
 
-       if (fd < 0)
-               return -1;
-       len = read(fd, buffer, sizeof(buffer));
-       close(fd);
-       if (len < 40)
-               return -1;
-       return get_sha1_hex(buffer, result);
+static int ambiguous_path(const char *path)
+{
+       int slash = 1;
+
+       for (;;) {
+               switch (*path++) {
+               case '\0':
+                       break;
+               case '/':
+                       if (slash)
+                               break;
+                       slash = 1;
+                       continue;
+               case '.':
+                       continue;
+               default:
+                       slash = 0;
+                       continue;
+               }
+               return slash;
+       }
 }
 
 static int get_sha1_basic(const char *str, int len, unsigned char *sha1)
@@ -143,20 +233,34 @@ static int get_sha1_basic(const char *str, int len, unsigned char *sha1)
                "refs",
                "refs/tags",
                "refs/heads",
-               "refs/snap",
                NULL
        };
        const char **p;
+       int found = 0;
 
-       if (!get_sha1_hex(str, sha1))
+       if (len == 40 && !get_sha1_hex(str, sha1))
                return 0;
 
+       /* Accept only unambiguous ref paths. */
+       if (ambiguous_path(str))
+               return -1;
+
        for (p = prefix; *p; p++) {
                char *pathname = git_path("%s/%.*s", *p, len, str);
-               if (!get_sha1_file(pathname, sha1))
-                       return 0;
+               if (!read_ref(pathname, sha1)) {
+                       /* Must be unique; i.e. when heads/foo and
+                        * tags/foo are both present, reject "foo".
+                        * Note that read_ref() eventually calls
+                        * get_sha1_hex() which can smudge initial
+                        * part of the buffer even if what is read
+                        * is found to be invalid halfway.
+                        */
+                       if (1 < found++)
+                               return -1;
+               }
        }
-
+       if (found == 1)
+               return 0;
        return -1;
 }
 
@@ -192,9 +296,107 @@ static int get_parent(const char *name, int len,
        return -1;
 }
 
+static int get_nth_ancestor(const char *name, int len,
+                           unsigned char *result, int generation)
+{
+       unsigned char sha1[20];
+       int ret = get_sha1_1(name, len, sha1);
+       if (ret)
+               return ret;
+
+       while (generation--) {
+               struct commit *commit = lookup_commit_reference(sha1);
+
+               if (!commit || parse_commit(commit) || !commit->parents)
+                       return -1;
+               memcpy(sha1, commit->parents->item->object.sha1, 20);
+       }
+       memcpy(result, sha1, 20);
+       return 0;
+}
+
+static int peel_onion(const char *name, int len, unsigned char *sha1)
+{
+       unsigned char outer[20];
+       const char *sp;
+       const char *type_string = NULL;
+       struct object *o;
+
+       /*
+        * "ref^{type}" dereferences ref repeatedly until you cannot
+        * dereference anymore, or you get an object of given type,
+        * whichever comes first.  "ref^{}" means just dereference
+        * tags until you get a non-tag.  "ref^0" is a shorthand for
+        * "ref^{commit}".  "commit^{tree}" could be used to find the
+        * top-level tree of the given commit.
+        */
+       if (len < 4 || name[len-1] != '}')
+               return -1;
+
+       for (sp = name + len - 1; name <= sp; sp--) {
+               int ch = *sp;
+               if (ch == '{' && name < sp && sp[-1] == '^')
+                       break;
+       }
+       if (sp <= name)
+               return -1;
+
+       sp++; /* beginning of type name, or closing brace for empty */
+       if (!strncmp(commit_type, sp, 6) && sp[6] == '}')
+               type_string = commit_type;
+       else if (!strncmp(tree_type, sp, 4) && sp[4] == '}')
+               type_string = tree_type;
+       else if (!strncmp(blob_type, sp, 4) && sp[4] == '}')
+               type_string = blob_type;
+       else if (sp[0] == '}')
+               type_string = NULL;
+       else
+               return -1;
+
+       if (get_sha1_1(name, sp - name - 2, outer))
+               return -1;
+
+       o = parse_object(outer);
+       if (!o)
+               return -1;
+       if (!type_string) {
+               o = deref_tag(o, name, sp - name - 2);
+               if (!o || (!o->parsed && !parse_object(o->sha1)))
+                       return -1;
+               memcpy(sha1, o->sha1, 20);
+       }
+       else {
+               /* At this point, the syntax look correct, so
+                * if we do not get the needed object, we should
+                * barf.
+                */
+
+               while (1) {
+                       if (!o || (!o->parsed && !parse_object(o->sha1)))
+                               return -1;
+                       if (o->type == type_string) {
+                               memcpy(sha1, o->sha1, 20);
+                               return 0;
+                       }
+                       if (o->type == tag_type)
+                               o = ((struct tag*) o)->tagged;
+                       else if (o->type == commit_type)
+                               o = &(((struct commit *) o)->tree->object);
+                       else
+                               return error("%.*s: expected %s type, but the object dereferences to %s type",
+                                            len, name, type_string,
+                                            o->type);
+                       if (!o->parsed)
+                               parse_object(o->sha1);
+               }
+       }
+       return 0;
+}
+
 static int get_sha1_1(const char *name, int len, unsigned char *sha1)
 {
        int parent, ret;
+       const char *cp;
 
        /* foo^[0-9] or foo^ (== foo^1); we do not do more than 9 parents. */
        if (len > 2 && name[len-2] == '^' &&
@@ -202,20 +404,44 @@ static int get_sha1_1(const char *name, int len, unsigned char *sha1)
                parent = name[len-1] - '0';
                len -= 2;
        }
-       else if (len > 1 && name[len-1] == '^')
+       else if (len > 1 && name[len-1] == '^') {
                parent = 1;
-       else
+               len--;
+       } else
                parent = -1;
 
-       if (0 <= parent) {
-               ret = get_parent(name, len-1, sha1, parent);
-               if (!ret)
-                       return 0;
+       if (parent >= 0)
+               return get_parent(name, len, sha1, parent);
+
+       /* "name~3" is "name^^^",
+        * "name~12" is "name^^^^^^^^^^^^", and
+        * "name~" and "name~0" are name -- not "name^0"!
+        */
+       parent = 0;
+       for (cp = name + len - 1; name <= cp; cp--) {
+               int ch = *cp;
+               if ('0' <= ch && ch <= '9')
+                       continue;
+               if (ch != '~')
+                       parent = -1;
+               break;
        }
+       if (!parent && *cp == '~') {
+               int len1 = cp - name;
+               cp++;
+               while (cp < name + len)
+                       parent = parent * 10 + *cp++ - '0';
+               return get_nth_ancestor(name, len1, sha1, parent);
+       }
+
+       ret = peel_onion(name, len, sha1);
+       if (!ret)
+               return 0;
+
        ret = get_sha1_basic(name, len, sha1);
        if (!ret)
                return 0;
-       return get_short_sha1(name, sha1);
+       return get_short_sha1(name, len, sha1, 0);
 }
 
 /*
@@ -224,5 +450,6 @@ static int get_sha1_1(const char *name, int len, unsigned char *sha1)
  */
 int get_sha1(const char *name, unsigned char *sha1)
 {
+       prepare_alt_odb();
        return get_sha1_1(name, strlen(name), sha1);
 }