Accumulated documentation updates.
[gitweb.git] / pack-objects.c
index 322c4bc3e9da508900801a316c6ac4eeac4ab468..3d622787cc554eb2936995441e4591ec1986bded 100644 (file)
@@ -1,78 +1,31 @@
+#include <ctype.h>
 #include "cache.h"
 #include "object.h"
 #include "delta.h"
+#include "pack.h"
+#include "csum-file.h"
 
-static const char pack_usage[] = "git-pack-objects [--window=N] base-name < object-list";
-
-enum object_type {
-       OBJ_NONE,
-       OBJ_COMMIT,
-       OBJ_TREE,
-       OBJ_BLOB,
-       OBJ_DELTA       // NOTE! This is _not_ the same as a "delta" object in the filesystem
-};
+static const char pack_usage[] = "git-pack-objects [--incremental] [--window=N] [--depth=N] {--stdout | base-name} < object-list";
 
 struct object_entry {
        unsigned char sha1[20];
        unsigned long size;
        unsigned long offset;
        unsigned int depth;
-       unsigned int flags;
+       unsigned int hash;
        enum object_type type;
        unsigned long delta_size;
        struct object_entry *delta;
 };
 
+static unsigned char object_list_sha1[20];
+static int non_empty = 0;
+static int incremental = 0;
 static struct object_entry **sorted_by_sha, **sorted_by_type;
 static struct object_entry *objects = NULL;
 static int nr_objects = 0, nr_alloc = 0;
-static int delta_window = 10;
 static const char *base_name;
-
-struct myfile {
-       int fd;
-       unsigned long chars;
-       unsigned char buffer[8192];
-};
-
-static FILE *create_file(const char *suffix)
-{
-       static char filename[PATH_MAX];
-       unsigned len;
-
-       len = snprintf(filename, PATH_MAX, "%s.%s", base_name, suffix);
-       if (len >= PATH_MAX)
-               die("you wascally wabbit, you");
-       return fopen(filename, "w");
-}
-
-static unsigned long fwrite_compressed(void *in, unsigned long size, FILE *f)
-{
-       z_stream stream;
-       unsigned long maxsize;
-       void *out;
-
-       memset(&stream, 0, sizeof(stream));
-       deflateInit(&stream, Z_DEFAULT_COMPRESSION);
-       maxsize = deflateBound(&stream, size);
-       out = xmalloc(maxsize);
-
-       /* Compress it */
-       stream.next_in = in;
-       stream.avail_in = size;
-
-       stream.next_out = out;
-       stream.avail_out = maxsize;
-
-       while (deflate(&stream, Z_FINISH) == Z_OK)
-               /* nothing */;
-       deflateEnd(&stream);
-
-       size = stream.total_out;
-       fwrite(out, size, 1, f);
-       free(out);
-       return size;
-}
+static unsigned char pack_file_sha1[20];
 
 static void *delta_against(void *buf, unsigned long size, struct object_entry *entry)
 {
@@ -83,21 +36,50 @@ static void *delta_against(void *buf, unsigned long size, struct object_entry *e
 
        if (!otherbuf)
                die("unable to read %s", sha1_to_hex(entry->delta->sha1));
-        delta_buf = diff_delta(buf, size, otherbuf, othersize, &delta_size);
-        if (delta_size != entry->delta_size)
+        delta_buf = diff_delta(otherbuf, othersize,
+                              buf, size, &delta_size, 0);
+        if (!delta_buf || delta_size != entry->delta_size)
                die("delta size changed");
         free(buf);
         free(otherbuf);
        return delta_buf;
 }
 
-static unsigned long write_object(FILE *f, struct object_entry *entry)
+/*
+ * The per-object header is a pretty dense thing, which is
+ *  - first byte: low four bits are "size", then three bits of "type",
+ *    and the high bit is "size continues".
+ *  - each byte afterwards: low seven bits are size continuation,
+ *    with the high bit being "size continues"
+ */
+static int encode_header(enum object_type type, unsigned long size, unsigned char *hdr)
+{
+       int n = 1;
+       unsigned char c;
+
+       if (type < OBJ_COMMIT || type > OBJ_DELTA)
+               die("bad type %d", type);
+
+       c = (type << 4) | (size & 15);
+       size >>= 4;
+       while (size) {
+               *hdr++ = c | 0x80;
+               c = size & 0x7f;
+               size >>= 7;
+               n++;
+       }
+       *hdr = c;
+       return n;
+}
+
+static unsigned long write_object(struct sha1file *f, struct object_entry *entry)
 {
        unsigned long size;
        char type[10];
        void *buf = read_sha1_file(entry->sha1, type, &size);
-       char header[21];
+       unsigned char header[10];
        unsigned hdrlen, datalen;
+       enum object_type obj_type;
 
        if (!buf)
                die("unable to read %s", sha1_to_hex(entry->sha1));
@@ -105,39 +87,65 @@ static unsigned long write_object(FILE *f, struct object_entry *entry)
                die("object %s size inconsistency (%lu vs %lu)", sha1_to_hex(entry->sha1), size, entry->size);
 
        /*
-        * The object header is a byte of 'type' followed by four bytes of
-        * length, except for deltas that has the 20 bytes of delta sha
-        * instead.
+        * The object header is a byte of 'type' followed by zero or
+        * more bytes of length.  For deltas, the 20 bytes of delta sha1
+        * follows that.
         */
-       header[0] = ".CTB"[entry->type];
-       datalen = htonl(size);
-       memcpy(header+1, &datalen, 4);
-       hdrlen = 5;
+       obj_type = entry->type;
        if (entry->delta) {
-               memcpy(header+1, entry->delta, 20);
                buf = delta_against(buf, size, entry);
                size = entry->delta_size;
-               hdrlen = 21;
+               obj_type = OBJ_DELTA;
        }
-       fwrite(header, hdrlen, 1, f);
-       datalen = fwrite_compressed(buf, size, f);
+       hdrlen = encode_header(obj_type, size, header);
+       sha1write(f, header, hdrlen);
+       if (entry->delta) {
+               sha1write(f, entry->delta, 20);
+               hdrlen += 20;
+       }
+       datalen = sha1write_compressed(f, buf, size);
        free(buf);
        return hdrlen + datalen;
 }
 
+static unsigned long write_one(struct sha1file *f,
+                              struct object_entry *e,
+                              unsigned long offset)
+{
+       if (e->offset)
+               /* offset starts from header size and cannot be zero
+                * if it is written already.
+                */
+               return offset;
+       e->offset = offset;
+       offset += write_object(f, e);
+       /* if we are delitified, write out its base object. */
+       if (e->delta)
+               offset = write_one(f, e->delta, offset);
+       return offset;
+}
+
 static void write_pack_file(void)
 {
        int i;
-       FILE *f = create_file("pack");
-       unsigned long offset = 0;
+       struct sha1file *f;
+       unsigned long offset;
        unsigned long mb;
+       struct pack_header hdr;
 
-       for (i = 0; i < nr_objects; i++) {
-               struct object_entry *entry = objects + i;
-               entry->offset = offset;
-               offset += write_object(f, entry);
-       }
-       fclose(f);
+       if (!base_name)
+               f = sha1fd(1, "<stdout>");
+       else
+               f = sha1create("%s-%s.%s", base_name, sha1_to_hex(object_list_sha1), "pack");
+       hdr.hdr_signature = htonl(PACK_SIGNATURE);
+       hdr.hdr_version = htonl(PACK_VERSION);
+       hdr.hdr_entries = htonl(nr_objects);
+       sha1write(f, &hdr, sizeof(hdr));
+       offset = sizeof(hdr);
+       for (i = 0; i < nr_objects; i++)
+               offset = write_one(f, objects + i, offset);
+
+       sha1close(f, pack_file_sha1, 1);
        mb = offset >> 20;
        offset &= 0xfffff;
 }
@@ -145,7 +153,7 @@ static void write_pack_file(void)
 static void write_index_file(void)
 {
        int i;
-       FILE *f = create_file("idx");
+       struct sha1file *f = sha1create("%s-%s.%s", base_name, sha1_to_hex(object_list_sha1), "idx");
        struct object_entry **list = sorted_by_sha;
        struct object_entry **last = list + nr_objects;
        unsigned int array[256];
@@ -153,7 +161,7 @@ static void write_index_file(void)
        /*
         * Write the first-level table (the list is sorted,
         * but we use a 256-entry lookup to be able to avoid
-        * having to do eight extra binary search iterations)
+        * having to do eight extra binary search iterations).
         */
        for (i = 0; i < 256; i++) {
                struct object_entry **next = list;
@@ -166,26 +174,30 @@ static void write_index_file(void)
                array[i] = htonl(next - sorted_by_sha);
                list = next;
        }
-       fwrite(array, 256, sizeof(int), f);
+       sha1write(f, array, 256 * sizeof(int));
 
        /*
         * Write the actual SHA1 entries..
         */
        list = sorted_by_sha;
-       for (i < 0; i < nr_objects; i++) {
+       for (i = 0; i < nr_objects; i++) {
                struct object_entry *entry = *list++;
                unsigned int offset = htonl(entry->offset);
-               fwrite(&offset, 4, 1, f);
-               fwrite(entry->sha1, 20, 1, f);
+               sha1write(f, &offset, 4);
+               sha1write(f, entry->sha1, 20);
        }
-       fclose(f);
+       sha1write(f, pack_file_sha1, 20);
+       sha1close(f, NULL, 1);
 }
 
-static void add_object_entry(unsigned char *sha1)
+static int add_object_entry(unsigned char *sha1, unsigned int hash)
 {
        unsigned int idx = nr_objects;
        struct object_entry *entry;
 
+       if (incremental && has_sha1_pack(sha1))
+               return 0;
+
        if (idx >= nr_alloc) {
                unsigned int needed = (idx + 1024) * 3 / 2;
                objects = xrealloc(objects, needed * sizeof(*entry));
@@ -194,33 +206,31 @@ static void add_object_entry(unsigned char *sha1)
        entry = objects + idx;
        memset(entry, 0, sizeof(*entry));
        memcpy(entry->sha1, sha1, 20);
+       entry->hash = hash;
        nr_objects = idx+1;
+       return 1;
 }
 
 static void check_object(struct object_entry *entry)
 {
-       char buffer[128];
-       char type[10];
-       unsigned long mapsize;
-       z_stream stream;
-       void *map;
-
-       map = map_sha1_file(entry->sha1, &mapsize);
-       if (!map)
-               die("unable to map %s", sha1_to_hex(entry->sha1));
-       if (unpack_sha1_header(&stream, map, mapsize, buffer, sizeof(buffer)) < 0)
-               die("unable to unpack %s header", sha1_to_hex(entry->sha1));
-       munmap(map, mapsize);
-       if (parse_sha1_header(buffer, type, &entry->size) < 0)
-               die("unable to parse %s header", sha1_to_hex(entry->sha1));
-       if (!strcmp(type, "commit")) {
-               entry->type = OBJ_COMMIT;
-       } else if (!strcmp(type, "tree")) {
-               entry->type = OBJ_TREE;
-       } else if (!strcmp(type, "blob")) {
-               entry->type = OBJ_BLOB;
-       } else
-               die("unable to pack object %s of type %s", sha1_to_hex(entry->sha1), type);
+       char type[20];
+
+       if (!sha1_object_info(entry->sha1, type, &entry->size)) {
+               if (!strcmp(type, "commit")) {
+                       entry->type = OBJ_COMMIT;
+               } else if (!strcmp(type, "tree")) {
+                       entry->type = OBJ_TREE;
+               } else if (!strcmp(type, "blob")) {
+                       entry->type = OBJ_BLOB;
+               } else if (!strcmp(type, "tag")) {
+                       entry->type = OBJ_TAG;
+               } else
+                       die("unable to pack object %s of type %s",
+                           sha1_to_hex(entry->sha1), type);
+       }
+       else
+               die("unable to get type of object %s",
+                   sha1_to_hex(entry->sha1));
 }
 
 static void get_object_details(void)
@@ -266,6 +276,10 @@ static int type_size_sort(const struct object_entry *a, const struct object_entr
                return -1;
        if (a->type > b->type)
                return 1;
+       if (a->hash < b->hash)
+               return -1;
+       if (a->hash > b->hash)
+               return 1;
        if (a->size < b->size)
                return -1;
        if (a->size > b->size)
@@ -279,29 +293,34 @@ struct unpacked {
 };
 
 /*
- * We search for deltas in a list sorted by type and by size, and
- * walk the "old" chain backwards in the list, so if the type
- * has changed or the size difference is too big, there's no point
- * in even continuing the walk, since the other old objects are
- * going to be even smaller or of a different type. So return -1
- * once we determine that there's no point even trying.
+ * We search for deltas _backwards_ in a list sorted by type and
+ * by size, so that we see progressively smaller and smaller files.
+ * That's because we prefer deltas to be from the bigger file
+ * to the smaller - deletes are potentially cheaper, but perhaps
+ * more importantly, the bigger file is likely the more recent
+ * one.
  */
-static int try_delta(struct unpacked *cur, struct unpacked *old)
+static int try_delta(struct unpacked *cur, struct unpacked *old, unsigned max_depth)
 {
        struct object_entry *cur_entry = cur->entry;
        struct object_entry *old_entry = old->entry;
-       unsigned long size, oldsize, delta_size;
+       unsigned long size, oldsize, delta_size, sizediff;
+       long max_size;
        void *delta_buf;
 
        /* Don't bother doing diffs between different types */
        if (cur_entry->type != old_entry->type)
                return -1;
 
-       /* Size is guaranteed to be larger than or equal to oldsize */
        size = cur_entry->size;
+       if (size < 50)
+               return -1;
        oldsize = old_entry->size;
-       if (size - oldsize > oldsize / 4)
+       sizediff = oldsize > size ? oldsize - size : size - oldsize;
+       if (sizediff > size / 8)
                return -1;
+       if (old_entry->depth >= max_depth)
+               return 0;
 
        /*
         * NOTE!
@@ -310,27 +329,32 @@ static int try_delta(struct unpacked *cur, struct unpacked *old)
         * more space-efficient (deletes don't have to say _what_ they
         * delete).
         */
-       delta_buf = diff_delta(cur->data, size, old->data, oldsize, &delta_size);
+       max_size = size / 2 - 20;
+       if (cur_entry->delta)
+               max_size = cur_entry->delta_size-1;
+       if (sizediff >= max_size)
+               return -1;
+       delta_buf = diff_delta(old->data, oldsize,
+                              cur->data, size, &delta_size, max_size);
        if (!delta_buf)
-               die("unable to create delta");
-       if (delta_size + 20 < size / 2) {
-               if (!cur_entry->delta || cur_entry->delta_size > delta_size) {
-                       cur_entry->delta = old_entry;
-                       cur_entry->delta_size = delta_size;
-               }
-       }
+               return 0;
+       cur_entry->delta = old_entry;
+       cur_entry->delta_size = delta_size;
+       cur_entry->depth = old_entry->depth + 1;
        free(delta_buf);
+       return 0;
 }
 
-static void find_deltas(struct object_entry **list, int window)
+static void find_deltas(struct object_entry **list, int window, int depth)
 {
-       unsigned int i;
+       int i, idx;
        unsigned int array_size = window * sizeof(struct unpacked);
        struct unpacked *array = xmalloc(array_size);
 
        memset(array, 0, array_size);
-       for (i = 0; i < nr_objects; i++) {
-               unsigned int idx = i % window;
+       i = nr_objects;
+       idx = 0;
+       while (--i >= 0) {
                struct object_entry *entry = list[i];
                struct unpacked *n = array + idx;
                unsigned long size;
@@ -342,36 +366,65 @@ static void find_deltas(struct object_entry **list, int window)
                n->data = read_sha1_file(entry->sha1, type, &size);
                if (size != entry->size)
                        die("object %s inconsistent object length (%lu vs %lu)", sha1_to_hex(entry->sha1), size, entry->size);
-               for (j = 0; j < window; j++) {
-                       unsigned int other_idx = idx - 1 - j;
+               j = window;
+               while (--j > 0) {
+                       unsigned int other_idx = idx + j;
                        struct unpacked *m;
-                       if (other_idx < 0)
-                               other_idx += window;
+                       if (other_idx >= window)
+                               other_idx -= window;
                        m = array + other_idx;
                        if (!m->entry)
                                break;
-                       if (try_delta(n, m) < 0)
+                       if (try_delta(n, m, depth) < 0)
                                break;
                }
+               idx++;
+               if (idx >= window)
+                       idx = 0;
        }
+
+       for (i = 0; i < window; ++i)
+               free(array[i].data);
+       free(array);
 }
 
 int main(int argc, char **argv)
 {
-       char line[128];
+       SHA_CTX ctx;
+       char line[PATH_MAX + 20];
+       int window = 10, depth = 10, pack_to_stdout = 0;
        int i;
 
        for (i = 1; i < argc; i++) {
                const char *arg = argv[i];
 
                if (*arg == '-') {
+                       if (!strcmp("--non-empty", arg)) {
+                               non_empty = 1;
+                               continue;
+                       }
+                       if (!strcmp("--incremental", arg)) {
+                               incremental = 1;
+                               continue;
+                       }
                        if (!strncmp("--window=", arg, 9)) {
                                char *end;
-                               delta_window = strtoul(arg+9, &end, 0);
+                               window = strtoul(arg+9, &end, 0);
                                if (!arg[9] || *end)
                                        usage(pack_usage);
                                continue;
                        }
+                       if (!strncmp("--depth=", arg, 8)) {
+                               char *end;
+                               depth = strtoul(arg+8, &end, 0);
+                               if (!arg[8] || *end)
+                                       usage(pack_usage);
+                               continue;
+                       }
+                       if (!strcmp("--stdout", arg)) {
+                               pack_to_stdout = 1;
+                               continue;
+                       }
                        usage(pack_usage);
                }
                if (base_name)
@@ -379,25 +432,44 @@ int main(int argc, char **argv)
                base_name = arg;
        }
 
-       if (!base_name)
+       if (pack_to_stdout != !base_name)
                usage(pack_usage);
 
+       SHA1_Init(&ctx);
        while (fgets(line, sizeof(line), stdin) != NULL) {
+               unsigned int hash;
+               char *p;
                unsigned char sha1[20];
+
                if (get_sha1_hex(line, sha1))
                        die("expected sha1, got garbage");
-               add_object_entry(sha1);
+               hash = 0;
+               p = line+40;
+               while (*p) {
+                       unsigned char c = *p++;
+                       if (isspace(c))
+                               continue;
+                       hash = hash * 11 + c;
+               }
+               if (add_object_entry(sha1, hash))
+                       SHA1_Update(&ctx, sha1, 20);
        }
+       SHA1_Final(object_list_sha1, &ctx);
+       if (non_empty && !nr_objects)
+               return 0;
        get_object_details();
 
-       printf("Packing %d objects\n", nr_objects);
+       fprintf(stderr, "Packing %d objects\n", nr_objects);
 
        sorted_by_sha = create_sorted_list(sha1_sort);
        sorted_by_type = create_sorted_list(type_size_sort);
-       if (delta_window)
-               find_deltas(sorted_by_type, delta_window);
+       if (window && depth)
+               find_deltas(sorted_by_type, window+1, depth);
 
        write_pack_file();
-       write_index_file();
+       if (!pack_to_stdout) {
+               write_index_file();
+               puts(sha1_to_hex(object_list_sha1));
+       }
        return 0;
 }