1GITFORMAT-INDEX(5) Git Manual GITFORMAT-INDEX(5)
2
3
4
6 gitformat-index - Git index format
7
9 $GIT_DIR/index
10
12 Git index format
13
15 All binary numbers are in network byte order.
16 In a repository using the traditional SHA-1, checksums and object IDs
17 (object names) mentioned below are all computed using SHA-1. Similarly,
18 in SHA-256 repositories, these values are computed using SHA-256.
19 Version 2 is described here unless stated otherwise.
20
21 • A 12-byte header consisting of
22
23 4-byte signature:
24 The signature is { 'D', 'I', 'R', 'C' } (stands for "dircache")
25
26 4-byte version number:
27 The current supported versions are 2, 3 and 4.
28
29 32-bit number of index entries.
30
31 • A number of sorted index entries (see below).
32
33 • Extensions
34
35 Extensions are identified by signature. Optional extensions can
36 be ignored if Git does not understand them.
37
38 4-byte extension signature. If the first byte is 'A'..'Z' the
39 extension is optional and can be ignored.
40
41 32-bit size of the extension
42
43 Extension data
44
45 • Hash checksum over the content of the index file before this
46 checksum.
47
49 Index entries are sorted in ascending order on the name field,
50 interpreted as a string of unsigned bytes (i.e. memcmp() order, no
51 localization, no special casing of directory separator '/'). Entries
52 with the same name are sorted by their stage field.
53
54 An index entry typically represents a file. However, if sparse-checkout
55 is enabled in cone mode (`core.sparseCheckoutCone` is enabled) and the
56 `extensions.sparseIndex` extension is enabled, then the index may
57 contain entries for directories outside of the sparse-checkout definition.
58 These entries have mode `040000`, include the `SKIP_WORKTREE` bit, and
59 the path ends in a directory separator.
60
61 32-bit ctime seconds, the last time a file's metadata changed
62 this is stat(2) data
63
64 32-bit ctime nanosecond fractions
65 this is stat(2) data
66
67 32-bit mtime seconds, the last time a file's data changed
68 this is stat(2) data
69
70 32-bit mtime nanosecond fractions
71 this is stat(2) data
72
73 32-bit dev
74 this is stat(2) data
75
76 32-bit ino
77 this is stat(2) data
78
79 32-bit mode, split into (high to low bits)
80
81 16-bit unused, must be zero
82
83 4-bit object type
84 valid values in binary are 1000 (regular file), 1010 (symbolic link)
85 and 1110 (gitlink)
86
87 3-bit unused, must be zero
88
89 9-bit unix permission. Only 0755 and 0644 are valid for regular files.
90 Symbolic links and gitlinks have value 0 in this field.
91
92 32-bit uid
93 this is stat(2) data
94
95 32-bit gid
96 this is stat(2) data
97
98 32-bit file size
99 This is the on-disk size from stat(2), truncated to 32-bit.
100
101 Object name for the represented object
102
103 A 16-bit 'flags' field split into (high to low bits)
104
105 1-bit assume-valid flag
106
107 1-bit extended flag (must be zero in version 2)
108
109 2-bit stage (during merge)
110
111 12-bit name length if the length is less than 0xFFF; otherwise 0xFFF
112 is stored in this field.
113
114 (Version 3 or later) A 16-bit field, only applicable if the
115 "extended flag" above is 1, split into (high to low bits).
116
117 1-bit reserved for future
118
119 1-bit skip-worktree flag (used by sparse checkout)
120
121 1-bit intent-to-add flag (used by "git add -N")
122
123 13-bit unused, must be zero
124
125 Entry path name (variable length) relative to top level directory
126 (without leading slash). '/' is used as path separator. The special
127 path components ".", ".." and ".git" (without quotes) are disallowed.
128 Trailing slash is also disallowed.
129
130 The exact encoding is undefined, but the '.' and '/' characters
131 are encoded in 7-bit ASCII and the encoding cannot contain a NUL
132 byte (iow, this is a UNIX pathname).
133
134 (Version 4) In version 4, the entry path name is prefix-compressed
135 relative to the path name for the previous entry (the very first
136 entry is encoded as if the path name for the previous entry is an
137 empty string). At the beginning of an entry, an integer N in the
138 variable width encoding (the same encoding as the offset is encoded
139 for OFS_DELTA pack entries; see linkgit:gitformat-pack[5]) is stored, followed
140 by a NUL-terminated string S. Removing N bytes from the end of the
141 path name for the previous entry, and replacing it with the string S
142 yields the path name for this entry.
143
144 1-8 nul bytes as necessary to pad the entry to a multiple of eight bytes
145 while keeping the name NUL-terminated.
146
147 (Version 4) In version 4, the padding after the pathname does not
148 exist.
149
150 Interpretation of index entries in split index mode is completely
151 different. See below for details.
152
154 Cache tree
155 Since the index does not record entries for directories, the cache
156 entries cannot describe tree objects that already exist in the object
157 database for regions of the index that are unchanged from an existing
158 commit. The cache tree extension stores a recursive tree structure that
159 describes the trees that already exist and completely match sections of
160 the cache entries. This speeds up tree object generation from the index
161 for a new commit by only computing the trees that are "new" to that
162 commit. It also assists when comparing the index to another tree, such
163 as `HEAD^{tree}`, since sections of the index can be skipped when a tree
164 comparison demonstrates equality.
165
166 The recursive tree structure uses nodes that store a number of cache
167 entries, a list of subnodes, and an object ID (OID). The OID references
168 the existing tree for that node, if it is known to exist. The subnodes
169 correspond to subdirectories that themselves have cache tree nodes. The
170 number of cache entries corresponds to the number of cache entries in
171 the index that describe paths within that tree's directory.
172
173 The extension tracks the full directory structure in the cache tree
174 extension, but this is generally smaller than the full cache entry list.
175
176 When a path is updated in index, Git invalidates all nodes of the
177 recursive cache tree corresponding to the parent directories of that
178 path. We store these tree nodes as being "invalid" by using "-1" as the
179 number of cache entries. Invalid nodes still store a span of index
180 entries, allowing Git to focus its efforts when reconstructing a full
181 cache tree.
182
183 The signature for this extension is { 'T', 'R', 'E', 'E' }.
184
185 A series of entries fill the entire extension; each of which
186 consists of:
187
188 • NUL-terminated path component (relative to its parent directory);
189
190 • ASCII decimal number of entries in the index that is covered by the
191 tree this entry represents (entry_count);
192
193 • A space (ASCII 32);
194
195 • ASCII decimal number that represents the number of subtrees this
196 tree has;
197
198 • A newline (ASCII 10); and
199
200 • Object name for the object that would result from writing this span
201 of index as a tree.
202
203 An entry can be in an invalidated state and is represented by having
204 a negative number in the entry_count field. In this case, there is no
205 object name and the next entry starts immediately after the newline.
206 When writing an invalid entry, -1 should always be used as entry_count.
207
208 The entries are written out in the top-down, depth-first order. The
209 first entry represents the root level of the repository, followed by the
210 first subtree--let's call this A--of the root level (with its name
211 relative to the root level), followed by the first subtree of A (with
212 its name relative to A), and so on. The specified number of subtrees
213 indicates when the current level of the recursive stack is complete.
214
215 Resolve undo
216 A conflict is represented in the index as a set of higher stage entries.
217 When a conflict is resolved (e.g. with "git add path"), these higher
218 stage entries will be removed and a stage-0 entry with proper resolution
219 is added.
220
221 When these higher stage entries are removed, they are saved in the
222 resolve undo extension, so that conflicts can be recreated (e.g. with
223 "git checkout -m"), in case users want to redo a conflict resolution
224 from scratch.
225
226 The signature for this extension is { 'R', 'E', 'U', 'C' }.
227
228 A series of entries fill the entire extension; each of which
229 consists of:
230
231 • NUL-terminated pathname the entry describes (relative to the root
232 of the repository, i.e. full pathname);
233
234 • Three NUL-terminated ASCII octal numbers, entry mode of entries in
235 stage 1 to 3 (a missing stage is represented by "0" in this field);
236 and
237
238 • At most three object names of the entry in stages from 1 to 3
239 (nothing is written for a missing stage).
240
241 Split index
242 In split index mode, the majority of index entries could be stored
243 in a separate file. This extension records the changes to be made on
244 top of that to produce the final index.
245
246 The signature for this extension is { 'l', 'i', 'n', 'k' }.
247
248 The extension consists of:
249
250 • Hash of the shared index file. The shared index file path is
251 $GIT_DIR/sharedindex.<hash>. If all bits are zero, the index does
252 not require a shared index file.
253
254 • An ewah-encoded delete bitmap, each bit represents an entry in the
255 shared index. If a bit is set, its corresponding entry in the
256 shared index will be removed from the final index. Note, because a
257 delete operation changes index entry positions, but we do need
258 original positions in replace phase, it’s best to just mark entries
259 for removal, then do a mass deletion after replacement.
260
261 • An ewah-encoded replace bitmap, each bit represents an entry in the
262 shared index. If a bit is set, its corresponding entry in the
263 shared index will be replaced with an entry in this index file. All
264 replaced entries are stored in sorted order in this index. The
265 first "1" bit in the replace bitmap corresponds to the first index
266 entry, the second "1" bit to the second entry and so on. Replaced
267 entries may have empty path names to save space.
268
269 The remaining index entries after replaced ones will be added to the
270 final index. These added entries are also sorted by entry name then
271 stage.
272
274 Untracked cache saves the untracked file list and necessary data to
275 verify the cache. The signature for this extension is { 'U', 'N',
276 'T', 'R' }.
277
278 The extension starts with
279
280 • A sequence of NUL-terminated strings, preceded by the size of the
281 sequence in variable width encoding. Each string describes the
282 environment where the cache can be used.
283
284 • Stat data of $GIT_DIR/info/exclude. See "Index entry" section from
285 ctime field until "file size".
286
287 • Stat data of core.excludesFile
288
289 • 32-bit dir_flags (see struct dir_struct)
290
291 • Hash of $GIT_DIR/info/exclude. A null hash means the file does not
292 exist.
293
294 • Hash of core.excludesFile. A null hash means the file does not
295 exist.
296
297 • NUL-terminated string of per-dir exclude file name. This usually is
298 ".gitignore".
299
300 • The number of following directory blocks, variable width encoding.
301 If this number is zero, the extension ends here with a following
302 NUL.
303
304 • A number of directory blocks in depth-first-search order, each
305 consists of
306
307 • The number of untracked entries, variable width encoding.
308
309 • The number of sub-directory blocks, variable width encoding.
310
311 • The directory name terminated by NUL.
312
313 • A number of untracked file/dir names terminated by NUL.
314
315 The remaining data of each directory block is grouped by type:
316
317 • An ewah bitmap, the n-th bit marks whether the n-th directory has
318 valid untracked cache entries.
319
320 • An ewah bitmap, the n-th bit records "check-only" bit of
321 read_directory_recursive() for the n-th directory.
322
323 • An ewah bitmap, the n-th bit indicates whether hash and stat data
324 is valid for the n-th directory and exists in the next data.
325
326 • An array of stat data. The n-th data corresponds with the n-th
327 "one" bit in the previous ewah bitmap.
328
329 • An array of hashes. The n-th hash corresponds with the n-th "one"
330 bit in the previous ewah bitmap.
331
332 • One NUL.
333
335 The file system monitor cache tracks files for which the core.fsmonitor
336 hook has told us about changes. The signature for this extension is
337 { 'F', 'S', 'M', 'N' }.
338
339 The extension starts with
340
341 • 32-bit version number: the current supported versions are 1 and 2.
342
343 • (Version 1) 64-bit time: the extension data reflects all changes
344 through the given time which is stored as the nanoseconds elapsed
345 since midnight, January 1, 1970.
346
347 • (Version 2) A null terminated string: an opaque token defined by
348 the file system monitor application. The extension data reflects
349 all changes relative to that token.
350
351 • 32-bit bitmap size: the size of the CE_FSMONITOR_VALID bitmap.
352
353 • An ewah bitmap, the n-th bit indicates whether the n-th index entry
354 is not CE_FSMONITOR_VALID.
355
357 The End of Index Entry (EOIE) is used to locate the end of the variable
358 length index entries and the beginning of the extensions. Code can take
359 advantage of this to quickly locate the index extensions without having
360 to parse through all of the index entries.
361
362 Because it must be able to be loaded before the variable length cache
363 entries and other index extensions, this extension must be written last.
364 The signature for this extension is { 'E', 'O', 'I', 'E' }.
365
366 The extension consists of:
367
368 • 32-bit offset to the end of the index entries
369
370 • Hash over the extension types and their sizes (but not their
371 contents). E.g. if we have "TREE" extension that is N-bytes long,
372 "REUC" extension that is M-bytes long, followed by "EOIE", then the
373 hash would be:
374
375 Hash("TREE" + <binary representation of N> +
376 "REUC" + <binary representation of M>)
377
379 The Index Entry Offset Table (IEOT) is used to help address the CPU
380 cost of loading the index by enabling multi-threading the process of
381 converting cache entries from the on-disk format to the in-memory format.
382 The signature for this extension is { 'I', 'E', 'O', 'T' }.
383
384 The extension consists of:
385
386 • 32-bit version (currently 1)
387
388 • A number of index offset entries each consisting of:
389
390 • 32-bit offset from the beginning of the file to the first cache
391 entry in this block of entries.
392
393 • 32-bit count of cache entries in this block
394
396 When using sparse-checkout in cone mode, some entire directories within
397 the index can be summarized by pointing to a tree object instead of the
398 entire expanded list of paths within that tree. An index containing such
399 entries is a "sparse index". Index format versions 4 and less were not
400 implemented with such entries in mind. Thus, for these versions, an
401 index containing sparse directory entries will include this extension
402 with signature { 's', 'd', 'i', 'r' }. Like the split-index extension,
403 tools should avoid interacting with a sparse index unless they understand
404 this extension.
405
407 Part of the git(1) suite
408
409
410
411Git 2.43.0 11/20/2023 GITFORMAT-INDEX(5)