Documentation / technical / index-format.txton commit Merge branch 'rr/minimal-stat' (070c57d)
   1GIT index format
   2================
   3
   4== The git index file has the following format
   5
   6  All binary numbers are in network byte order. Version 2 is described
   7  here unless stated otherwise.
   8
   9   - A 12-byte header consisting of
  10
  11     4-byte signature:
  12       The signature is { 'D', 'I', 'R', 'C' } (stands for "dircache")
  13
  14     4-byte version number:
  15       The current supported versions are 2 and 3.
  16
  17     32-bit number of index entries.
  18
  19   - A number of sorted index entries (see below).
  20
  21   - Extensions
  22
  23     Extensions are identified by signature. Optional extensions can
  24     be ignored if GIT does not understand them.
  25
  26     GIT currently supports cached tree and resolve undo extensions.
  27
  28     4-byte extension signature. If the first byte is 'A'..'Z' the
  29     extension is optional and can be ignored.
  30
  31     32-bit size of the extension
  32
  33     Extension data
  34
  35   - 160-bit SHA-1 over the content of the index file before this
  36     checksum.
  37
  38== Index entry
  39
  40  Index entries are sorted in ascending order on the name field,
  41  interpreted as a string of unsigned bytes (i.e. memcmp() order, no
  42  localization, no special casing of directory separator '/'). Entries
  43  with the same name are sorted by their stage field.
  44
  45  32-bit ctime seconds, the last time a file's metadata changed
  46    this is stat(2) data
  47
  48  32-bit ctime nanosecond fractions
  49    this is stat(2) data
  50
  51  32-bit mtime seconds, the last time a file's data changed
  52    this is stat(2) data
  53
  54  32-bit mtime nanosecond fractions
  55    this is stat(2) data
  56
  57  32-bit dev
  58    this is stat(2) data
  59
  60  32-bit ino
  61    this is stat(2) data
  62
  63  32-bit mode, split into (high to low bits)
  64
  65    4-bit object type
  66      valid values in binary are 1000 (regular file), 1010 (symbolic link)
  67      and 1110 (gitlink)
  68
  69    3-bit unused
  70
  71    9-bit unix permission. Only 0755 and 0644 are valid for regular files.
  72    Symbolic links and gitlinks have value 0 in this field.
  73
  74  32-bit uid
  75    this is stat(2) data
  76
  77  32-bit gid
  78    this is stat(2) data
  79
  80  32-bit file size
  81    This is the on-disk size from stat(2), truncated to 32-bit.
  82
  83  160-bit SHA-1 for the represented object
  84
  85  A 16-bit 'flags' field split into (high to low bits)
  86
  87    1-bit assume-valid flag
  88
  89    1-bit extended flag (must be zero in version 2)
  90
  91    2-bit stage (during merge)
  92
  93    12-bit name length if the length is less than 0xFFF; otherwise 0xFFF
  94    is stored in this field.
  95
  96  (Version 3) A 16-bit field, only applicable if the "extended flag"
  97  above is 1, split into (high to low bits).
  98
  99    1-bit reserved for future
 100
 101    1-bit skip-worktree flag (used by sparse checkout)
 102
 103    1-bit intent-to-add flag (used by "git add -N")
 104
 105    13-bit unused, must be zero
 106
 107  Entry path name (variable length) relative to top level directory
 108    (without leading slash). '/' is used as path separator. The special
 109    path components ".", ".." and ".git" (without quotes) are disallowed.
 110    Trailing slash is also disallowed.
 111
 112    The exact encoding is undefined, but the '.' and '/' characters
 113    are encoded in 7-bit ASCII and the encoding cannot contain a NUL
 114    byte (iow, this is a UNIX pathname).
 115
 116  (Version 4) In version 4, the entry path name is prefix-compressed
 117    relative to the path name for the previous entry (the very first
 118    entry is encoded as if the path name for the previous entry is an
 119    empty string).  At the beginning of an entry, an integer N in the
 120    variable width encoding (the same encoding as the offset is encoded
 121    for OFS_DELTA pack entries; see pack-format.txt) is stored, followed
 122    by a NUL-terminated string S.  Removing N bytes from the end of the
 123    path name for the previous entry, and replacing it with the string S
 124    yields the path name for this entry.
 125
 126  1-8 nul bytes as necessary to pad the entry to a multiple of eight bytes
 127  while keeping the name NUL-terminated.
 128
 129  (Version 4) In version 4, the padding after the pathname does not
 130  exist.
 131
 132== Extensions
 133
 134=== Cached tree
 135
 136  Cached tree extension contains pre-computed hashes for trees that can
 137  be derived from the index. It helps speed up tree object generation
 138  from index for a new commit.
 139
 140  When a path is updated in index, the path must be invalidated and
 141  removed from tree cache.
 142
 143  The signature for this extension is { 'T', 'R', 'E', 'E' }.
 144
 145  A series of entries fill the entire extension; each of which
 146  consists of:
 147
 148  - NUL-terminated path component (relative to its parent directory);
 149
 150  - ASCII decimal number of entries in the index that is covered by the
 151    tree this entry represents (entry_count);
 152
 153  - A space (ASCII 32);
 154
 155  - ASCII decimal number that represents the number of subtrees this
 156    tree has;
 157
 158  - A newline (ASCII 10); and
 159
 160  - 160-bit object name for the object that would result from writing
 161    this span of index as a tree.
 162
 163  An entry can be in an invalidated state and is represented by having
 164  a negative number in the entry_count field. In this case, there is no
 165  object name and the next entry starts immediately after the newline.
 166  When writing an invalid entry, -1 should always be used as entry_count.
 167
 168  The entries are written out in the top-down, depth-first order.  The
 169  first entry represents the root level of the repository, followed by the
 170  first subtree---let's call this A---of the root level (with its name
 171  relative to the root level), followed by the first subtree of A (with
 172  its name relative to A), ...
 173
 174=== Resolve undo
 175
 176  A conflict is represented in the index as a set of higher stage entries.
 177  When a conflict is resolved (e.g. with "git add path"), these higher
 178  stage entries will be removed and a stage-0 entry with proper resoluton
 179  is added.
 180
 181  When these higher stage entries are removed, they are saved in the
 182  resolve undo extension, so that conflicts can be recreated (e.g. with
 183  "git checkout -m"), in case users want to redo a conflict resolution
 184  from scratch.
 185
 186  The signature for this extension is { 'R', 'E', 'U', 'C' }.
 187
 188  A series of entries fill the entire extension; each of which
 189  consists of:
 190
 191  - NUL-terminated pathname the entry describes (relative to the root of
 192    the repository, i.e. full pathname);
 193
 194  - Three NUL-terminated ASCII octal numbers, entry mode of entries in
 195    stage 1 to 3 (a missing stage is represented by "0" in this field);
 196    and
 197
 198  - At most three 160-bit object names of the entry in stages from 1 to 3
 199    (nothing is written for a missing stage).
 200