1GIT-PACK-OBJECTS(1) Git Manual GIT-PACK-OBJECTS(1)
2
6 git-pack-objects - Create a packed archive of objects
7
9 git pack-objects [-q | --progress | --all-progress] [--all-progress-implied]
10 [--no-reuse-delta] [--delta-base-offset] [--non-empty]
11 [--local] [--incremental] [--window=<n>] [--depth=<n>]
12 [--revs [--unpacked | --all]] [--keep-pack=<pack-name>]
13 [--stdout [--filter=<filter-spec>] | base-name]
14 [--shallow] [--keep-true-parents] [--sparse] < object-list
15
17 Reads list of objects from the standard input, and writes either one or
18 more packed archives with the specified base-name to disk, or a packed
19 archive to the standard output.
20 A packed archive is an efficient way to transfer a set of objects
22 between two repositories as well as an access efficient archival
23 format. In a packed archive, an object is either stored as a compressed
24 whole or as a difference from some other object. The latter is often
25 called a delta.
26 The packed archive format (.pack) is designed to be self-contained so
28 that it can be unpacked without any further information. Therefore,
29 each object that a delta depends upon must be present within the pack.
30 A pack index file (.idx) is generated for fast, random access to the
32 objects in the pack. Placing both the index file (.idx) and the packed
33 archive (.pack) in the pack/ subdirectory of $GIT_OBJECT_DIRECTORY (or
34 any of the directories on $GIT_ALTERNATE_OBJECT_DIRECTORIES) enables
35 Git to read from the pack archive.
36 The git unpack-objects command can read the packed archive and expand
38 the objects contained in the pack into "one-file one-object" format;
39 this is typically done by the smart-pull commands when a pack is
40 created on-the-fly for efficient network transport by their peers.
41
43 base-name
44 Write into pairs of files (.pack and .idx), using <base-name> to
45 determine the name of the created file. When this option is used,
46 the two files in a pair are written in
47 <base-name>-<SHA-1>.{pack,idx} files. <SHA-1> is a hash based on
48 the pack content and is written to the standard output of the
49 command.
50 --stdout
52 Write the pack contents (what would have been written to .pack
53 file) out to the standard output.
54 --revs
56 Read the revision arguments from the standard input, instead of
57 individual object names. The revision arguments are processed the
58 same way as git rev-list with the --objects flag uses its commit
59 arguments to build the list of objects it outputs. The objects on
60 the resulting list are packed. Besides revisions, --not or
61 --shallow <SHA-1> lines are also accepted.
62 --unpacked
64 This implies --revs. When processing the list of revision arguments
65 read from the standard input, limit the objects packed to those
66 that are not already packed.
67 --all
69 This implies --revs. In addition to the list of revision arguments
70 read from the standard input, pretend as if all refs under refs/
71 are specified to be included.
72 --include-tag
74 Include unasked-for annotated tags if the object they reference was
75 included in the resulting packfile. This can be useful to send new
76 tags to native Git clients.
77 --window=<n>, --depth=<n>
79 These two options affect how the objects contained in the pack are
80 stored using delta compression. The objects are first internally
81 sorted by type, size and optionally names and compared against the
82 other objects within --window to see if using delta compression
83 saves space. --depth limits the maximum delta depth; making it too
84 deep affects the performance on the unpacker side, because delta
85 data needs to be applied that many times to get to the necessary
86 object.
87 The default value for --window is 10 and --depth is 50. The maximum
89 depth is 4095.
90 --window-memory=<n>
92 This option provides an additional limit on top of --window; the
93 window size will dynamically scale down so as to not take up more
94 than <n> bytes in memory. This is useful in repositories with a mix
95 of large and small objects to not run out of memory with a large
96 window, but still be able to take advantage of the large window for
97 the smaller objects. The size can be suffixed with "k", "m", or
98 "g". --window-memory=0 makes memory usage unlimited. The default
99 is taken from the pack.windowMemory configuration variable.
100 --max-pack-size=<n>
102 In unusual scenarios, you may not be able to create files larger
103 than a certain size on your filesystem, and this option can be used
104 to tell the command to split the output packfile into multiple
105 independent packfiles, each not larger than the given size. The
106 size can be suffixed with "k", "m", or "g". The minimum size
107 allowed is limited to 1 MiB. This option prevents the creation of a
108 bitmap index. The default is unlimited, unless the config variable
109 pack.packSizeLimit is set.
110 --honor-pack-keep
112 This flag causes an object already in a local pack that has a .keep
113 file to be ignored, even if it would have otherwise been packed.
114 --keep-pack=<pack-name>
116 This flag causes an object already in the given pack to be ignored,
117 even if it would have otherwise been packed. <pack-name> is the
118 pack file name without leading directory (e.g. pack-123.pack). The
119 option could be specified multiple times to keep multiple packs.
120 --incremental
122 This flag causes an object already in a pack to be ignored even if
123 it would have otherwise been packed.
124 --local
126 This flag causes an object that is borrowed from an alternate
127 object store to be ignored even if it would have otherwise been
128 packed.
129 --non-empty
131 Only create a packed archive if it would contain at least one
132 object.
133 --progress
135 Progress status is reported on the standard error stream by default
136 when it is attached to a terminal, unless -q is specified. This
137 flag forces progress status even if the standard error stream is
138 not directed to a terminal.
139 --all-progress
141 When --stdout is specified then progress report is displayed during
142 the object count and compression phases but inhibited during the
143 write-out phase. The reason is that in some cases the output stream
144 is directly linked to another command which may wish to display
145 progress status of its own as it processes incoming pack data. This
146 flag is like --progress except that it forces progress report for
147 the write-out phase as well even if --stdout is used.
148 --all-progress-implied
150 This is used to imply --all-progress whenever progress display is
151 activated. Unlike --all-progress this flag doesn’t actually force
152 any progress display by itself.
153 -q
155 This flag makes the command not to report its progress on the
156 standard error stream.
157 --no-reuse-delta
159 When creating a packed archive in a repository that has existing
160 packs, the command reuses existing deltas. This sometimes results
161 in a slightly suboptimal pack. This flag tells the command not to
162 reuse existing deltas but compute them from scratch.
163 --no-reuse-object
165 This flag tells the command not to reuse existing object data at
166 all, including non deltified object, forcing recompression of
167 everything. This implies --no-reuse-delta. Useful only in the
168 obscure case where wholesale enforcement of a different compression
169 level on the packed data is desired.
170 --compression=<n>
172 Specifies compression level for newly-compressed data in the
173 generated pack. If not specified, pack compression level is
174 determined first by pack.compression, then by core.compression, and
175 defaults to -1, the zlib default, if neither is set. Add
176 --no-reuse-object if you want to force a uniform compression level
177 on all data no matter the source.
178 --sparse
180 Use the "sparse" algorithm to determine which objects to include in
181 the pack, when combined with the "--revs" option. This algorithm
182 only walks trees that appear in paths that introduce new objects.
183 This can have significant performance benefits when computing a
184 pack to send a small change. However, it is possible that extra
185 objects are added to the pack-file if the included commits contain
186 certain types of direct renames.
187 --thin
189 Create a "thin" pack by omitting the common objects between a
190 sender and a receiver in order to reduce network transfer. This
191 option only makes sense in conjunction with --stdout.
192 Note: A thin pack violates the packed archive format by omitting
194 required objects and is thus unusable by Git without making it
195 self-contained. Use git index-pack --fix-thin (see git-index-
196 pack(1)) to restore the self-contained property.
197 --shallow
199 Optimize a pack that will be provided to a client with a shallow
200 repository. This option, combined with --thin, can result in a
201 smaller pack at the cost of speed.
202 --delta-base-offset
204 A packed archive can express the base object of a delta as either a
205 20-byte object name or as an offset in the stream, but ancient
206 versions of Git don’t understand the latter. By default, git
207 pack-objects only uses the former format for better compatibility.
208 This option allows the command to use the latter format for
209 compactness. Depending on the average delta chain length, this
210 option typically shrinks the resulting packfile by 3-5 per-cent.
211 Note: Porcelain commands such as git gc (see git-gc(1)), git repack
213 (see git-repack(1)) pass this option by default in modern Git when
214 they put objects in your repository into pack files. So does git
215 bundle (see git-bundle(1)) when it creates a bundle.
216 --threads=<n>
218 Specifies the number of threads to spawn when searching for best
219 delta matches. This requires that pack-objects be compiled with
220 pthreads otherwise this option is ignored with a warning. This is
221 meant to reduce packing time on multiprocessor machines. The
222 required amount of memory for the delta search window is however
223 multiplied by the number of threads. Specifying 0 will cause Git to
224 auto-detect the number of CPU’s and set the number of threads
225 accordingly.
226 --index-version=<version>[,<offset>]
228 This is intended to be used by the test suite only. It allows to
229 force the version for the generated pack index, and to force 64-bit
230 index entries on objects located above the given offset.
231 --keep-true-parents
233 With this option, parents that are hidden by grafts are packed
234 nevertheless.
235 --filter=<filter-spec>
237 Requires --stdout. Omits certain objects (usually blobs) from the
238 resulting packfile. See git-rev-list(1) for valid <filter-spec>
239 forms.
240 --no-filter
242 Turns off any previous --filter= argument.
243 --missing=<missing-action>
245 A debug option to help with future "partial clone" development.
246 This option specifies how missing objects are handled.
247 The form --missing=error requests that pack-objects stop with an
249 error if a missing object is encountered. This is the default
250 action.
251 The form --missing=allow-any will allow object traversal to
253 continue if a missing object is encountered. Missing objects will
254 silently be omitted from the results.
255 The form --missing=allow-promisor is like allow-any, but will only
257 allow object traversal to continue for EXPECTED promisor missing
258 objects. Unexpected missing object will raise an error.
259 --exclude-promisor-objects
261 Omit objects that are known to be in the promisor remote. (This
262 option has the purpose of operating only on locally created
263 objects, so that when we repack, we still maintain a distinction
264 between locally created objects [without .promisor] and objects
265 from the promisor remote [with .promisor].) This is used with
266 partial clone.
267 --keep-unreachable
269 Objects unreachable from the refs in packs named with --unpacked=
270 option are added to the resulting pack, in addition to the
271 reachable objects that are not in packs marked with *.keep files.
272 This implies --revs.
273 --pack-loose-unreachable
275 Pack unreachable loose objects (and their loose counterparts
276 removed). This implies --revs.
277 --unpack-unreachable
279 Keep unreachable objects in loose form. This implies --revs.
280 --delta-islands
282 Restrict delta matches based on "islands". See DELTA ISLANDS below.
283
285 When possible, pack-objects tries to reuse existing on-disk deltas to
286 avoid having to search for new ones on the fly. This is an important
287 optimization for serving fetches, because it means the server can avoid
288 inflating most objects at all and just send the bytes directly from
289 disk. This optimization can’t work when an object is stored as a delta
290 against a base which the receiver does not have (and which we are not
291 already sending). In that case the server "breaks" the delta and has to
292 find a new one, which has a high CPU cost. Therefore it’s important for
293 performance that the set of objects in on-disk delta relationships
294 match what a client would fetch.
295 In a normal repository, this tends to work automatically. The objects
297 are mostly reachable from the branches and tags, and that’s what
298 clients fetch. Any deltas we find on the server are likely to be
299 between objects the client has or will have.
300 But in some repository setups, you may have several related but
302 separate groups of ref tips, with clients tending to fetch those groups
303 independently. For example, imagine that you are hosting several
304 "forks" of a repository in a single shared object store, and letting
305 clients view them as separate repositories through GIT_NAMESPACE or
306 separate repos using the alternates mechanism. A naive repack may find
307 that the optimal delta for an object is against a base that is only
308 found in another fork. But when a client fetches, they will not have
309 the base object, and we’ll have to find a new delta on the fly.
310 A similar situation may exist if you have many refs outside of
312 refs/heads/ and refs/tags/ that point to related objects (e.g.,
313 refs/pull or refs/changes used by some hosting providers). By default,
314 clients fetch only heads and tags, and deltas against objects found
315 only in those other groups cannot be sent as-is.
316 Delta islands solve this problem by allowing you to group your refs
318 into distinct "islands". Pack-objects computes which objects are
319 reachable from which islands, and refuses to make a delta from an
320 object A against a base which is not present in all of A's islands.
321 This results in slightly larger packs (because we miss some delta
322 opportunities), but guarantees that a fetch of one island will not have
323 to recompute deltas on the fly due to crossing island boundaries.
324 When repacking with delta islands the delta window tends to get clogged
326 with candidates that are forbidden by the config. Repacking with a big
327 --window helps (and doesn’t take as long as it otherwise might because
328 we can reject some object pairs based on islands before doing any
329 computation on the content).
330 Islands are configured via the pack.island option, which can be
332 specified multiple times. Each value is a left-anchored regular
333 expressions matching refnames. For example:
334 [pack]
336 island = refs/heads/
337 island = refs/tags/
338 puts heads and tags into an island (whose name is the empty string; see
340 below for more on naming). Any refs which do not match those regular
341 expressions (e.g., refs/pull/123) is not in any island. Any object
342 which is reachable only from refs/pull/ (but not heads or tags) is
343 therefore not a candidate to be used as a base for refs/heads/.
344 Refs are grouped into islands based on their "names", and two regexes
346 that produce the same name are considered to be in the same island. The
347 names are computed from the regexes by concatenating any capture groups
348 from the regex, with a - dash in between. (And if there are no capture
349 groups, then the name is the empty string, as in the above example.)
350 This allows you to create arbitrary numbers of islands. Only up to 14
351 such capture groups are supported though.
352 For example, imagine you store the refs for each fork in
354 refs/virtual/ID, where ID is a numeric identifier. You might then
355 configure:
356 [pack]
358 island = refs/virtual/([0-9]+)/heads/
359 island = refs/virtual/([0-9]+)/tags/
360 island = refs/virtual/([0-9]+)/(pull)/
361 That puts the heads and tags for each fork in their own island (named
363 "1234" or similar), and the pull refs for each go into their own
364 "1234-pull".
365 Note that we pick a single island for each regex to go into, using
367 "last one wins" ordering (which allows repo-specific config to take
368 precedence over user-wide config, and so forth).
369
371 git-rev-list(1) git-repack(1) git-prune-packed(1)
372
374 Part of the git(1) suite
375Git 2.26.2 2020-04-20 GIT-PACK-OBJECTS(1)