]> git.ipfire.org Git - thirdparty/git.git/blob - Documentation/git-filter-branch.txt
t4014: make output-directory tests self-contained
[thirdparty/git.git] / Documentation / git-filter-branch.txt
1 git-filter-branch(1)
2 ====================
3
4 NAME
5 ----
6 git-filter-branch - Rewrite branches
7
8 SYNOPSIS
9 --------
10 [verse]
11 'git filter-branch' [--setup <command>] [--subdirectory-filter <directory>]
12 [--env-filter <command>] [--tree-filter <command>]
13 [--index-filter <command>] [--parent-filter <command>]
14 [--msg-filter <command>] [--commit-filter <command>]
15 [--tag-name-filter <command>] [--prune-empty]
16 [--original <namespace>] [-d <directory>] [-f | --force]
17 [--state-branch <branch>] [--] [<rev-list options>...]
18
19 DESCRIPTION
20 -----------
21 Lets you rewrite Git revision history by rewriting the branches mentioned
22 in the <rev-list options>, applying custom filters on each revision.
23 Those filters can modify each tree (e.g. removing a file or running
24 a perl rewrite on all files) or information about each commit.
25 Otherwise, all information (including original commit times or merge
26 information) will be preserved.
27
28 The command will only rewrite the _positive_ refs mentioned in the
29 command line (e.g. if you pass 'a..b', only 'b' will be rewritten).
30 If you specify no filters, the commits will be recommitted without any
31 changes, which would normally have no effect. Nevertheless, this may be
32 useful in the future for compensating for some Git bugs or such,
33 therefore such a usage is permitted.
34
35 *NOTE*: This command honors `.git/info/grafts` file and refs in
36 the `refs/replace/` namespace.
37 If you have any grafts or replacement refs defined, running this command
38 will make them permanent.
39
40 *WARNING*! The rewritten history will have different object names for all
41 the objects and will not converge with the original branch. You will not
42 be able to easily push and distribute the rewritten branch on top of the
43 original branch. Please do not use this command if you do not know the
44 full implications, and avoid using it anyway, if a simple single commit
45 would suffice to fix your problem. (See the "RECOVERING FROM UPSTREAM
46 REBASE" section in linkgit:git-rebase[1] for further information about
47 rewriting published history.)
48
49 Always verify that the rewritten version is correct: The original refs,
50 if different from the rewritten ones, will be stored in the namespace
51 'refs/original/'.
52
53 Note that since this operation is very I/O expensive, it might
54 be a good idea to redirect the temporary directory off-disk with the
55 `-d` option, e.g. on tmpfs. Reportedly the speedup is very noticeable.
56
57
58 Filters
59 ~~~~~~~
60
61 The filters are applied in the order as listed below. The <command>
62 argument is always evaluated in the shell context using the 'eval' command
63 (with the notable exception of the commit filter, for technical reasons).
64 Prior to that, the `$GIT_COMMIT` environment variable will be set to contain
65 the id of the commit being rewritten. Also, GIT_AUTHOR_NAME,
66 GIT_AUTHOR_EMAIL, GIT_AUTHOR_DATE, GIT_COMMITTER_NAME, GIT_COMMITTER_EMAIL,
67 and GIT_COMMITTER_DATE are taken from the current commit and exported to
68 the environment, in order to affect the author and committer identities of
69 the replacement commit created by linkgit:git-commit-tree[1] after the
70 filters have run.
71
72 If any evaluation of <command> returns a non-zero exit status, the whole
73 operation will be aborted.
74
75 A 'map' function is available that takes an "original sha1 id" argument
76 and outputs a "rewritten sha1 id" if the commit has been already
77 rewritten, and "original sha1 id" otherwise; the 'map' function can
78 return several ids on separate lines if your commit filter emitted
79 multiple commits.
80
81
82 OPTIONS
83 -------
84
85 --setup <command>::
86 This is not a real filter executed for each commit but a one
87 time setup just before the loop. Therefore no commit-specific
88 variables are defined yet. Functions or variables defined here
89 can be used or modified in the following filter steps except
90 the commit filter, for technical reasons.
91
92 --subdirectory-filter <directory>::
93 Only look at the history which touches the given subdirectory.
94 The result will contain that directory (and only that) as its
95 project root. Implies <<Remap_to_ancestor>>.
96
97 --env-filter <command>::
98 This filter may be used if you only need to modify the environment
99 in which the commit will be performed. Specifically, you might
100 want to rewrite the author/committer name/email/time environment
101 variables (see linkgit:git-commit-tree[1] for details).
102
103 --tree-filter <command>::
104 This is the filter for rewriting the tree and its contents.
105 The argument is evaluated in shell with the working
106 directory set to the root of the checked out tree. The new tree
107 is then used as-is (new files are auto-added, disappeared files
108 are auto-removed - neither .gitignore files nor any other ignore
109 rules *HAVE ANY EFFECT*!).
110
111 --index-filter <command>::
112 This is the filter for rewriting the index. It is similar to the
113 tree filter but does not check out the tree, which makes it much
114 faster. Frequently used with `git rm --cached
115 --ignore-unmatch ...`, see EXAMPLES below. For hairy
116 cases, see linkgit:git-update-index[1].
117
118 --parent-filter <command>::
119 This is the filter for rewriting the commit's parent list.
120 It will receive the parent string on stdin and shall output
121 the new parent string on stdout. The parent string is in
122 the format described in linkgit:git-commit-tree[1]: empty for
123 the initial commit, "-p parent" for a normal commit and
124 "-p parent1 -p parent2 -p parent3 ..." for a merge commit.
125
126 --msg-filter <command>::
127 This is the filter for rewriting the commit messages.
128 The argument is evaluated in the shell with the original
129 commit message on standard input; its standard output is
130 used as the new commit message.
131
132 --commit-filter <command>::
133 This is the filter for performing the commit.
134 If this filter is specified, it will be called instead of the
135 'git commit-tree' command, with arguments of the form
136 "<TREE_ID> [(-p <PARENT_COMMIT_ID>)...]" and the log message on
137 stdin. The commit id is expected on stdout.
138 +
139 As a special extension, the commit filter may emit multiple
140 commit ids; in that case, the rewritten children of the original commit will
141 have all of them as parents.
142 +
143 You can use the 'map' convenience function in this filter, and other
144 convenience functions, too. For example, calling 'skip_commit "$@"'
145 will leave out the current commit (but not its changes! If you want
146 that, use 'git rebase' instead).
147 +
148 You can also use the `git_commit_non_empty_tree "$@"` instead of
149 `git commit-tree "$@"` if you don't wish to keep commits with a single parent
150 and that makes no change to the tree.
151
152 --tag-name-filter <command>::
153 This is the filter for rewriting tag names. When passed,
154 it will be called for every tag ref that points to a rewritten
155 object (or to a tag object which points to a rewritten object).
156 The original tag name is passed via standard input, and the new
157 tag name is expected on standard output.
158 +
159 The original tags are not deleted, but can be overwritten;
160 use "--tag-name-filter cat" to simply update the tags. In this
161 case, be very careful and make sure you have the old tags
162 backed up in case the conversion has run afoul.
163 +
164 Nearly proper rewriting of tag objects is supported. If the tag has
165 a message attached, a new tag object will be created with the same message,
166 author, and timestamp. If the tag has a signature attached, the
167 signature will be stripped. It is by definition impossible to preserve
168 signatures. The reason this is "nearly" proper, is because ideally if
169 the tag did not change (points to the same object, has the same name, etc.)
170 it should retain any signature. That is not the case, signatures will always
171 be removed, buyer beware. There is also no support for changing the
172 author or timestamp (or the tag message for that matter). Tags which point
173 to other tags will be rewritten to point to the underlying commit.
174
175 --prune-empty::
176 Some filters will generate empty commits that leave the tree untouched.
177 This option instructs git-filter-branch to remove such commits if they
178 have exactly one or zero non-pruned parents; merge commits will
179 therefore remain intact. This option cannot be used together with
180 `--commit-filter`, though the same effect can be achieved by using the
181 provided `git_commit_non_empty_tree` function in a commit filter.
182
183 --original <namespace>::
184 Use this option to set the namespace where the original commits
185 will be stored. The default value is 'refs/original'.
186
187 -d <directory>::
188 Use this option to set the path to the temporary directory used for
189 rewriting. When applying a tree filter, the command needs to
190 temporarily check out the tree to some directory, which may consume
191 considerable space in case of large projects. By default it
192 does this in the `.git-rewrite/` directory but you can override
193 that choice by this parameter.
194
195 -f::
196 --force::
197 'git filter-branch' refuses to start with an existing temporary
198 directory or when there are already refs starting with
199 'refs/original/', unless forced.
200
201 --state-branch <branch>::
202 This option will cause the mapping from old to new objects to
203 be loaded from named branch upon startup and saved as a new
204 commit to that branch upon exit, enabling incremental of large
205 trees. If '<branch>' does not exist it will be created.
206
207 <rev-list options>...::
208 Arguments for 'git rev-list'. All positive refs included by
209 these options are rewritten. You may also specify options
210 such as `--all`, but you must use `--` to separate them from
211 the 'git filter-branch' options. Implies <<Remap_to_ancestor>>.
212
213
214 [[Remap_to_ancestor]]
215 Remap to ancestor
216 ~~~~~~~~~~~~~~~~~
217
218 By using linkgit:git-rev-list[1] arguments, e.g., path limiters, you can limit the
219 set of revisions which get rewritten. However, positive refs on the command
220 line are distinguished: we don't let them be excluded by such limiters. For
221 this purpose, they are instead rewritten to point at the nearest ancestor that
222 was not excluded.
223
224
225 EXIT STATUS
226 -----------
227
228 On success, the exit status is `0`. If the filter can't find any commits to
229 rewrite, the exit status is `2`. On any other error, the exit status may be
230 any other non-zero value.
231
232
233 EXAMPLES
234 --------
235
236 Suppose you want to remove a file (containing confidential information
237 or copyright violation) from all commits:
238
239 -------------------------------------------------------
240 git filter-branch --tree-filter 'rm filename' HEAD
241 -------------------------------------------------------
242
243 However, if the file is absent from the tree of some commit,
244 a simple `rm filename` will fail for that tree and commit.
245 Thus you may instead want to use `rm -f filename` as the script.
246
247 Using `--index-filter` with 'git rm' yields a significantly faster
248 version. Like with using `rm filename`, `git rm --cached filename`
249 will fail if the file is absent from the tree of a commit. If you
250 want to "completely forget" a file, it does not matter when it entered
251 history, so we also add `--ignore-unmatch`:
252
253 --------------------------------------------------------------------------
254 git filter-branch --index-filter 'git rm --cached --ignore-unmatch filename' HEAD
255 --------------------------------------------------------------------------
256
257 Now, you will get the rewritten history saved in HEAD.
258
259 To rewrite the repository to look as if `foodir/` had been its project
260 root, and discard all other history:
261
262 -------------------------------------------------------
263 git filter-branch --subdirectory-filter foodir -- --all
264 -------------------------------------------------------
265
266 Thus you can, e.g., turn a library subdirectory into a repository of
267 its own. Note the `--` that separates 'filter-branch' options from
268 revision options, and the `--all` to rewrite all branches and tags.
269
270 To set a commit (which typically is at the tip of another
271 history) to be the parent of the current initial commit, in
272 order to paste the other history behind the current history:
273
274 -------------------------------------------------------------------
275 git filter-branch --parent-filter 'sed "s/^\$/-p <graft-id>/"' HEAD
276 -------------------------------------------------------------------
277
278 (if the parent string is empty - which happens when we are dealing with
279 the initial commit - add graftcommit as a parent). Note that this assumes
280 history with a single root (that is, no merge without common ancestors
281 happened). If this is not the case, use:
282
283 --------------------------------------------------------------------------
284 git filter-branch --parent-filter \
285 'test $GIT_COMMIT = <commit-id> && echo "-p <graft-id>" || cat' HEAD
286 --------------------------------------------------------------------------
287
288 or even simpler:
289
290 -----------------------------------------------
291 git replace --graft $commit-id $graft-id
292 git filter-branch $graft-id..HEAD
293 -----------------------------------------------
294
295 To remove commits authored by "Darl McBribe" from the history:
296
297 ------------------------------------------------------------------------------
298 git filter-branch --commit-filter '
299 if [ "$GIT_AUTHOR_NAME" = "Darl McBribe" ];
300 then
301 skip_commit "$@";
302 else
303 git commit-tree "$@";
304 fi' HEAD
305 ------------------------------------------------------------------------------
306
307 The function 'skip_commit' is defined as follows:
308
309 --------------------------
310 skip_commit()
311 {
312 shift;
313 while [ -n "$1" ];
314 do
315 shift;
316 map "$1";
317 shift;
318 done;
319 }
320 --------------------------
321
322 The shift magic first throws away the tree id and then the -p
323 parameters. Note that this handles merges properly! In case Darl
324 committed a merge between P1 and P2, it will be propagated properly
325 and all children of the merge will become merge commits with P1,P2
326 as their parents instead of the merge commit.
327
328 *NOTE* the changes introduced by the commits, and which are not reverted
329 by subsequent commits, will still be in the rewritten branch. If you want
330 to throw out _changes_ together with the commits, you should use the
331 interactive mode of 'git rebase'.
332
333 You can rewrite the commit log messages using `--msg-filter`. For
334 example, 'git svn-id' strings in a repository created by 'git svn' can
335 be removed this way:
336
337 -------------------------------------------------------
338 git filter-branch --msg-filter '
339 sed -e "/^git-svn-id:/d"
340 '
341 -------------------------------------------------------
342
343 If you need to add 'Acked-by' lines to, say, the last 10 commits (none
344 of which is a merge), use this command:
345
346 --------------------------------------------------------
347 git filter-branch --msg-filter '
348 cat &&
349 echo "Acked-by: Bugs Bunny <bunny@bugzilla.org>"
350 ' HEAD~10..HEAD
351 --------------------------------------------------------
352
353 The `--env-filter` option can be used to modify committer and/or author
354 identity. For example, if you found out that your commits have the wrong
355 identity due to a misconfigured user.email, you can make a correction,
356 before publishing the project, like this:
357
358 --------------------------------------------------------
359 git filter-branch --env-filter '
360 if test "$GIT_AUTHOR_EMAIL" = "root@localhost"
361 then
362 GIT_AUTHOR_EMAIL=john@example.com
363 fi
364 if test "$GIT_COMMITTER_EMAIL" = "root@localhost"
365 then
366 GIT_COMMITTER_EMAIL=john@example.com
367 fi
368 ' -- --all
369 --------------------------------------------------------
370
371 To restrict rewriting to only part of the history, specify a revision
372 range in addition to the new branch name. The new branch name will
373 point to the top-most revision that a 'git rev-list' of this range
374 will print.
375
376 Consider this history:
377
378 ------------------
379 D--E--F--G--H
380 / /
381 A--B-----C
382 ------------------
383
384 To rewrite only commits D,E,F,G,H, but leave A, B and C alone, use:
385
386 --------------------------------
387 git filter-branch ... C..H
388 --------------------------------
389
390 To rewrite commits E,F,G,H, use one of these:
391
392 ----------------------------------------
393 git filter-branch ... C..H --not D
394 git filter-branch ... D..H --not C
395 ----------------------------------------
396
397 To move the whole tree into a subdirectory, or remove it from there:
398
399 ---------------------------------------------------------------
400 git filter-branch --index-filter \
401 'git ls-files -s | sed "s-\t\"*-&newsubdir/-" |
402 GIT_INDEX_FILE=$GIT_INDEX_FILE.new \
403 git update-index --index-info &&
404 mv "$GIT_INDEX_FILE.new" "$GIT_INDEX_FILE"' HEAD
405 ---------------------------------------------------------------
406
407
408
409 CHECKLIST FOR SHRINKING A REPOSITORY
410 ------------------------------------
411
412 git-filter-branch can be used to get rid of a subset of files,
413 usually with some combination of `--index-filter` and
414 `--subdirectory-filter`. People expect the resulting repository to
415 be smaller than the original, but you need a few more steps to
416 actually make it smaller, because Git tries hard not to lose your
417 objects until you tell it to. First make sure that:
418
419 * You really removed all variants of a filename, if a blob was moved
420 over its lifetime. `git log --name-only --follow --all -- filename`
421 can help you find renames.
422
423 * You really filtered all refs: use `--tag-name-filter cat -- --all`
424 when calling git-filter-branch.
425
426 Then there are two ways to get a smaller repository. A safer way is
427 to clone, that keeps your original intact.
428
429 * Clone it with `git clone file:///path/to/repo`. The clone
430 will not have the removed objects. See linkgit:git-clone[1]. (Note
431 that cloning with a plain path just hardlinks everything!)
432
433 If you really don't want to clone it, for whatever reasons, check the
434 following points instead (in this order). This is a very destructive
435 approach, so *make a backup* or go back to cloning it. You have been
436 warned.
437
438 * Remove the original refs backed up by git-filter-branch: say `git
439 for-each-ref --format="%(refname)" refs/original/ | xargs -n 1 git
440 update-ref -d`.
441
442 * Expire all reflogs with `git reflog expire --expire=now --all`.
443
444 * Garbage collect all unreferenced objects with `git gc --prune=now`
445 (or if your git-gc is not new enough to support arguments to
446 `--prune`, use `git repack -ad; git prune` instead).
447
448 NOTES
449 -----
450
451 git-filter-branch allows you to make complex shell-scripted rewrites
452 of your Git history, but you probably don't need this flexibility if
453 you're simply _removing unwanted data_ like large files or passwords.
454 For those operations you may want to consider
455 http://rtyley.github.io/bfg-repo-cleaner/[The BFG Repo-Cleaner],
456 a JVM-based alternative to git-filter-branch, typically at least
457 10-50x faster for those use-cases, and with quite different
458 characteristics:
459
460 * Any particular version of a file is cleaned exactly _once_. The BFG,
461 unlike git-filter-branch, does not give you the opportunity to
462 handle a file differently based on where or when it was committed
463 within your history. This constraint gives the core performance
464 benefit of The BFG, and is well-suited to the task of cleansing bad
465 data - you don't care _where_ the bad data is, you just want it
466 _gone_.
467
468 * By default The BFG takes full advantage of multi-core machines,
469 cleansing commit file-trees in parallel. git-filter-branch cleans
470 commits sequentially (i.e. in a single-threaded manner), though it
471 _is_ possible to write filters that include their own parallelism,
472 in the scripts executed against each commit.
473
474 * The http://rtyley.github.io/bfg-repo-cleaner/#examples[command options]
475 are much more restrictive than git-filter branch, and dedicated just
476 to the tasks of removing unwanted data- e.g:
477 `--strip-blobs-bigger-than 1M`.
478
479 GIT
480 ---
481 Part of the linkgit:git[1] suite