]>
Commit | Line | Data |
---|---|---|
c401b33c JS |
1 | git-filter-branch(1) |
2 | ==================== | |
3 | ||
4 | NAME | |
5 | ---- | |
6 | git-filter-branch - Rewrite branches | |
7 | ||
8 | SYNOPSIS | |
9 | -------- | |
10 | [verse] | |
b1889c36 | 11 | 'git filter-branch' [--env-filter <command>] [--tree-filter <command>] |
c401b33c JS |
12 | [--index-filter <command>] [--parent-filter <command>] |
13 | [--msg-filter <command>] [--commit-filter <command>] | |
14 | [--tag-name-filter <command>] [--subdirectory-filter <directory>] | |
5433235d | 15 | [--original <namespace>] [-d <directory>] [-f | --force] |
8afa4210 | 16 | [--] [<rev-list options>...] |
c401b33c JS |
17 | |
18 | DESCRIPTION | |
19 | ----------- | |
08203668 JS |
20 | Lets you rewrite git revision history by rewriting the branches mentioned |
21 | in the <rev-list options>, applying custom filters on each revision. | |
c401b33c JS |
22 | Those filters can modify each tree (e.g. removing a file or running |
23 | a perl rewrite on all files) or information about each commit. | |
24 | Otherwise, all information (including original commit times or merge | |
25 | information) will be preserved. | |
26 | ||
08203668 | 27 | The command will only rewrite the _positive_ refs mentioned in the |
bf7c9021 | 28 | command line (e.g. if you pass 'a..b', only 'b' will be rewritten). |
08203668 JS |
29 | If you specify no filters, the commits will be recommitted without any |
30 | changes, which would normally have no effect. Nevertheless, this may be | |
31 | useful in the future for compensating for some git bugs or such, | |
32 | therefore such a usage is permitted. | |
c401b33c | 33 | |
c6d8f763 DCS |
34 | *NOTE*: This command honors `.git/info/grafts`. If you have any grafts |
35 | defined, running this command will make them permanent. | |
36 | ||
73616fd3 | 37 | *WARNING*! The rewritten history will have different object names for all |
c401b33c JS |
38 | the objects and will not converge with the original branch. You will not |
39 | be able to easily push and distribute the rewritten branch on top of the | |
40 | original branch. Please do not use this command if you do not know the | |
41 | full implications, and avoid using it anyway, if a simple single commit | |
97c33c65 TR |
42 | would suffice to fix your problem. (See the "RECOVERING FROM UPSTREAM |
43 | REBASE" section in linkgit:git-rebase[1] for further information about | |
44 | rewriting published history.) | |
c401b33c | 45 | |
dfd05e38 JS |
46 | Always verify that the rewritten version is correct: The original refs, |
47 | if different from the rewritten ones, will be stored in the namespace | |
48 | 'refs/original/'. | |
c401b33c | 49 | |
bf7c9021 | 50 | Note that since this operation is very I/O expensive, it might |
08203668 JS |
51 | be a good idea to redirect the temporary directory off-disk with the |
52 | '-d' option, e.g. on tmpfs. Reportedly the speedup is very noticeable. | |
c401b33c JS |
53 | |
54 | ||
55 | Filters | |
56 | ~~~~~~~ | |
57 | ||
58 | The filters are applied in the order as listed below. The <command> | |
bf7c9021 RW |
59 | argument is always evaluated in the shell context using the 'eval' command |
60 | (with the notable exception of the commit filter, for technical reasons). | |
c401b33c JS |
61 | Prior to that, the $GIT_COMMIT environment variable will be set to contain |
62 | the id of the commit being rewritten. Also, GIT_AUTHOR_NAME, | |
63 | GIT_AUTHOR_EMAIL, GIT_AUTHOR_DATE, GIT_COMMITTER_NAME, GIT_COMMITTER_EMAIL, | |
bf7c9021 RW |
64 | and GIT_COMMITTER_DATE are set according to the current commit. The values |
65 | of these variables after the filters have run, are used for the new commit. | |
66 | If any evaluation of <command> returns a non-zero exit status, the whole | |
67 | operation will be aborted. | |
c401b33c JS |
68 | |
69 | A 'map' function is available that takes an "original sha1 id" argument | |
70 | and outputs a "rewritten sha1 id" if the commit has been already | |
32c37c12 JS |
71 | rewritten, and "original sha1 id" otherwise; the 'map' function can |
72 | return several ids on separate lines if your commit filter emitted | |
73 | multiple commits. | |
c401b33c JS |
74 | |
75 | ||
76 | OPTIONS | |
77 | ------- | |
78 | ||
79 | --env-filter <command>:: | |
bf7c9021 RW |
80 | This filter may be used if you only need to modify the environment |
81 | in which the commit will be performed. Specifically, you might | |
82 | want to rewrite the author/committer name/email/time environment | |
5162e697 | 83 | variables (see linkgit:git-commit[1] for details). Do not forget |
c401b33c JS |
84 | to re-export the variables. |
85 | ||
86 | --tree-filter <command>:: | |
87 | This is the filter for rewriting the tree and its contents. | |
88 | The argument is evaluated in shell with the working | |
89 | directory set to the root of the checked out tree. The new tree | |
90 | is then used as-is (new files are auto-added, disappeared files | |
91 | are auto-removed - neither .gitignore files nor any other ignore | |
73616fd3 | 92 | rules *HAVE ANY EFFECT*!). |
c401b33c JS |
93 | |
94 | --index-filter <command>:: | |
95 | This is the filter for rewriting the index. It is similar to the | |
96 | tree filter but does not check out the tree, which makes it much | |
3bc427e0 TR |
97 | faster. Frequently used with `git rm \--cached |
98 | \--ignore-unmatch ...`, see EXAMPLES below. For hairy | |
99 | cases, see linkgit:git-update-index[1]. | |
c401b33c JS |
100 | |
101 | --parent-filter <command>:: | |
102 | This is the filter for rewriting the commit's parent list. | |
103 | It will receive the parent string on stdin and shall output | |
104 | the new parent string on stdout. The parent string is in | |
483bc4f0 | 105 | the format described in linkgit:git-commit-tree[1]: empty for |
c401b33c JS |
106 | the initial commit, "-p parent" for a normal commit and |
107 | "-p parent1 -p parent2 -p parent3 ..." for a merge commit. | |
108 | ||
109 | --msg-filter <command>:: | |
110 | This is the filter for rewriting the commit messages. | |
111 | The argument is evaluated in the shell with the original | |
112 | commit message on standard input; its standard output is | |
113 | used as the new commit message. | |
114 | ||
115 | --commit-filter <command>:: | |
116 | This is the filter for performing the commit. | |
117 | If this filter is specified, it will be called instead of the | |
ba020ef5 | 118 | 'git-commit-tree' command, with arguments of the form |
c401b33c JS |
119 | "<TREE_ID> [-p <PARENT_COMMIT_ID>]..." and the log message on |
120 | stdin. The commit id is expected on stdout. | |
121 | + | |
122 | As a special extension, the commit filter may emit multiple | |
c5833f6e | 123 | commit ids; in that case, the rewritten children of the original commit will |
c401b33c | 124 | have all of them as parents. |
f95eef15 JS |
125 | + |
126 | You can use the 'map' convenience function in this filter, and other | |
127 | convenience functions, too. For example, calling 'skip_commit "$@"' | |
128 | will leave out the current commit (but not its changes! If you want | |
ba020ef5 | 129 | that, use 'git-rebase' instead). |
d3240d93 PH |
130 | + |
131 | You can also use the 'git_commit_non_empty_tree "$@"' instead of | |
132 | 'git commit-tree "$@"' if you don't wish to keep commits with a single parent | |
133 | and that makes no change to the tree. | |
c401b33c JS |
134 | |
135 | --tag-name-filter <command>:: | |
136 | This is the filter for rewriting tag names. When passed, | |
137 | it will be called for every tag ref that points to a rewritten | |
138 | object (or to a tag object which points to a rewritten object). | |
139 | The original tag name is passed via standard input, and the new | |
140 | tag name is expected on standard output. | |
141 | + | |
142 | The original tags are not deleted, but can be overwritten; | |
5876b8ee | 143 | use "--tag-name-filter cat" to simply update the tags. In this |
c401b33c JS |
144 | case, be very careful and make sure you have the old tags |
145 | backed up in case the conversion has run afoul. | |
146 | + | |
1bf6551e BC |
147 | Nearly proper rewriting of tag objects is supported. If the tag has |
148 | a message attached, a new tag object will be created with the same message, | |
149 | author, and timestamp. If the tag has a signature attached, the | |
150 | signature will be stripped. It is by definition impossible to preserve | |
151 | signatures. The reason this is "nearly" proper, is because ideally if | |
152 | the tag did not change (points to the same object, has the same name, etc.) | |
153 | it should retain any signature. That is not the case, signatures will always | |
154 | be removed, buyer beware. There is also no support for changing the | |
155 | author or timestamp (or the tag message for that matter). Tags which point | |
156 | to other tags will be rewritten to point to the underlying commit. | |
c401b33c JS |
157 | |
158 | --subdirectory-filter <directory>:: | |
73616fd3 JS |
159 | Only look at the history which touches the given subdirectory. |
160 | The result will contain that directory (and only that) as its | |
161 | project root. | |
c401b33c | 162 | |
d3240d93 PH |
163 | --prune-empty:: |
164 | Some kind of filters will generate empty commits, that left the tree | |
165 | untouched. This switch allow git-filter-branch to ignore such | |
166 | commits. Though, this switch only applies for commits that have one | |
167 | and only one parent, it will hence keep merges points. Also, this | |
168 | option is not compatible with the use of '--commit-filter'. Though you | |
169 | just need to use the function 'git_commit_non_empty_tree "$@"' instead | |
170 | of the 'git commit-tree "$@"' idiom in your commit filter to make that | |
171 | happen. | |
172 | ||
5433235d GB |
173 | --original <namespace>:: |
174 | Use this option to set the namespace where the original commits | |
175 | will be stored. The default value is 'refs/original'. | |
176 | ||
c401b33c JS |
177 | -d <directory>:: |
178 | Use this option to set the path to the temporary directory used for | |
179 | rewriting. When applying a tree filter, the command needs to | |
bf7c9021 | 180 | temporarily check out the tree to some directory, which may consume |
c401b33c JS |
181 | considerable space in case of large projects. By default it |
182 | does this in the '.git-rewrite/' directory but you can override | |
183 | that choice by this parameter. | |
184 | ||
3240240f SB |
185 | -f:: |
186 | --force:: | |
ba020ef5 | 187 | 'git-filter-branch' refuses to start with an existing temporary |
dfd05e38 JS |
188 | directory or when there are already refs starting with |
189 | 'refs/original/', unless forced. | |
190 | ||
f448e24e | 191 | <rev-list options>...:: |
8afa4210 TR |
192 | Arguments for 'git-rev-list'. All positive refs included by |
193 | these options are rewritten. You may also specify options | |
194 | such as '--all', but you must use '--' to separate them from | |
195 | the 'git-filter-branch' options. | |
c401b33c JS |
196 | |
197 | ||
198 | Examples | |
199 | -------- | |
200 | ||
201 | Suppose you want to remove a file (containing confidential information | |
202 | or copyright violation) from all commits: | |
203 | ||
204 | ------------------------------------------------------- | |
dfd05e38 | 205 | git filter-branch --tree-filter 'rm filename' HEAD |
c401b33c JS |
206 | ------------------------------------------------------- |
207 | ||
e4d594c6 JL |
208 | However, if the file is absent from the tree of some commit, |
209 | a simple `rm filename` will fail for that tree and commit. | |
210 | Thus you may instead want to use `rm -f filename` as the script. | |
211 | ||
3bc427e0 TR |
212 | Using `\--index-filter` with 'git-rm' yields a significantly faster |
213 | version. Like with using `rm filename`, `git rm --cached filename` | |
214 | will fail if the file is absent from the tree of a commit. If you | |
215 | want to "completely forget" a file, it does not matter when it entered | |
216 | history, so we also add `\--ignore-unmatch`: | |
c401b33c | 217 | |
dfd05e38 | 218 | -------------------------------------------------------------------------- |
3bc427e0 | 219 | git filter-branch --index-filter 'git rm --cached --ignore-unmatch filename' HEAD |
dfd05e38 | 220 | -------------------------------------------------------------------------- |
c401b33c | 221 | |
8ef44519 | 222 | Now, you will get the rewritten history saved in HEAD. |
c401b33c | 223 | |
8afa4210 TR |
224 | To rewrite the repository to look as if `foodir/` had been its project |
225 | root, and discard all other history: | |
226 | ||
227 | ------------------------------------------------------- | |
228 | git filter-branch --subdirectory-filter foodir -- --all | |
229 | ------------------------------------------------------- | |
230 | ||
231 | Thus you can, e.g., turn a library subdirectory into a repository of | |
232 | its own. Note the `\--` that separates 'filter-branch' options from | |
233 | revision options, and the `\--all` to rewrite all branches and tags. | |
234 | ||
32c37c12 JS |
235 | To set a commit (which typically is at the tip of another |
236 | history) to be the parent of the current initial commit, in | |
237 | order to paste the other history behind the current history: | |
c401b33c | 238 | |
dfd05e38 JS |
239 | ------------------------------------------------------------------- |
240 | git filter-branch --parent-filter 'sed "s/^\$/-p <graft-id>/"' HEAD | |
241 | ------------------------------------------------------------------- | |
c401b33c | 242 | |
08203668 JS |
243 | (if the parent string is empty - which happens when we are dealing with |
244 | the initial commit - add graftcommit as a parent). Note that this assumes | |
c401b33c JS |
245 | history with a single root (that is, no merge without common ancestors |
246 | happened). If this is not the case, use: | |
247 | ||
dfd05e38 | 248 | -------------------------------------------------------------------------- |
c401b33c | 249 | git filter-branch --parent-filter \ |
41e86a37 | 250 | 'test $GIT_COMMIT = <commit-id> && echo "-p <graft-id>" || cat' HEAD |
dfd05e38 | 251 | -------------------------------------------------------------------------- |
c401b33c | 252 | |
32c37c12 JS |
253 | or even simpler: |
254 | ||
255 | ----------------------------------------------- | |
256 | echo "$commit-id $graft-id" >> .git/info/grafts | |
dfd05e38 | 257 | git filter-branch $graft-id..HEAD |
32c37c12 JS |
258 | ----------------------------------------------- |
259 | ||
c401b33c JS |
260 | To remove commits authored by "Darl McBribe" from the history: |
261 | ||
262 | ------------------------------------------------------------------------------ | |
263 | git filter-branch --commit-filter ' | |
264 | if [ "$GIT_AUTHOR_NAME" = "Darl McBribe" ]; | |
265 | then | |
f95eef15 | 266 | skip_commit "$@"; |
c401b33c JS |
267 | else |
268 | git commit-tree "$@"; | |
dfd05e38 | 269 | fi' HEAD |
c401b33c JS |
270 | ------------------------------------------------------------------------------ |
271 | ||
8451c565 | 272 | The function 'skip_commit' is defined as follows: |
f95eef15 JS |
273 | |
274 | -------------------------- | |
275 | skip_commit() | |
276 | { | |
277 | shift; | |
278 | while [ -n "$1" ]; | |
279 | do | |
280 | shift; | |
281 | map "$1"; | |
282 | shift; | |
283 | done; | |
284 | } | |
285 | -------------------------- | |
286 | ||
c401b33c JS |
287 | The shift magic first throws away the tree id and then the -p |
288 | parameters. Note that this handles merges properly! In case Darl | |
289 | committed a merge between P1 and P2, it will be propagated properly | |
290 | and all children of the merge will become merge commits with P1,P2 | |
291 | as their parents instead of the merge commit. | |
292 | ||
a1748890 | 293 | You can rewrite the commit log messages using `--msg-filter`. For |
ba020ef5 | 294 | example, 'git-svn-id' strings in a repository created by 'git-svn' can |
ed10d9aa MV |
295 | be removed this way: |
296 | ||
297 | ------------------------------------------------------- | |
a1748890 | 298 | git filter-branch --msg-filter ' |
ed10d9aa MV |
299 | sed -e "/^git-svn-id:/d" |
300 | ' | |
301 | ------------------------------------------------------- | |
f95eef15 | 302 | |
c401b33c JS |
303 | To restrict rewriting to only part of the history, specify a revision |
304 | range in addition to the new branch name. The new branch name will | |
ba020ef5 | 305 | point to the top-most revision that a 'git-rev-list' of this range |
c401b33c JS |
306 | will print. |
307 | ||
b8f42332 JS |
308 | If you need to add 'Acked-by' lines to, say, the last 10 commits (none |
309 | of which is a merge), use this command: | |
310 | ||
311 | -------------------------------------------------------- | |
312 | git filter-branch --msg-filter ' | |
313 | cat && | |
314 | echo "Acked-by: Bugs Bunny <bunny@bugzilla.org>" | |
315 | ' HEAD~10..HEAD | |
316 | -------------------------------------------------------- | |
317 | ||
08203668 JS |
318 | *NOTE* the changes introduced by the commits, and which are not reverted |
319 | by subsequent commits, will still be in the rewritten branch. If you want | |
c401b33c | 320 | to throw out _changes_ together with the commits, you should use the |
ba020ef5 | 321 | interactive mode of 'git-rebase'. |
c401b33c | 322 | |
08203668 | 323 | |
c401b33c JS |
324 | Consider this history: |
325 | ||
326 | ------------------ | |
327 | D--E--F--G--H | |
328 | / / | |
329 | A--B-----C | |
330 | ------------------ | |
331 | ||
332 | To rewrite only commits D,E,F,G,H, but leave A, B and C alone, use: | |
333 | ||
334 | -------------------------------- | |
dfd05e38 | 335 | git filter-branch ... C..H |
c401b33c JS |
336 | -------------------------------- |
337 | ||
338 | To rewrite commits E,F,G,H, use one of these: | |
339 | ||
340 | ---------------------------------------- | |
dfd05e38 JS |
341 | git filter-branch ... C..H --not D |
342 | git filter-branch ... D..H --not C | |
c401b33c JS |
343 | ---------------------------------------- |
344 | ||
345 | To move the whole tree into a subdirectory, or remove it from there: | |
346 | ||
347 | --------------------------------------------------------------- | |
348 | git filter-branch --index-filter \ | |
349 | 'git ls-files -s | sed "s-\t-&newsubdir/-" | | |
350 | GIT_INDEX_FILE=$GIT_INDEX_FILE.new \ | |
351 | git update-index --index-info && | |
dfd05e38 | 352 | mv $GIT_INDEX_FILE.new $GIT_INDEX_FILE' HEAD |
c401b33c JS |
353 | --------------------------------------------------------------- |
354 | ||
355 | ||
d0268de6 TR |
356 | |
357 | Checklist for Shrinking a Repository | |
358 | ------------------------------------ | |
359 | ||
360 | git-filter-branch is often used to get rid of a subset of files, | |
361 | usually with some combination of `\--index-filter` and | |
362 | `\--subdirectory-filter`. People expect the resulting repository to | |
363 | be smaller than the original, but you need a few more steps to | |
364 | actually make it smaller, because git tries hard not to lose your | |
365 | objects until you tell it to. First make sure that: | |
366 | ||
367 | * You really removed all variants of a filename, if a blob was moved | |
368 | over its lifetime. `git log \--name-only \--follow \--all \-- | |
369 | filename` can help you find renames. | |
370 | ||
371 | * You really filtered all refs: use `\--tag-name-filter cat \-- | |
372 | \--all` when calling git-filter-branch. | |
373 | ||
374 | Then there are two ways to get a smaller repository. A safer way is | |
375 | to clone, that keeps your original intact. | |
376 | ||
377 | * Clone it with `git clone +++file:///path/to/repo+++`. The clone | |
378 | will not have the removed objects. See linkgit:git-clone[1]. (Note | |
379 | that cloning with a plain path just hardlinks everything!) | |
380 | ||
381 | If you really don't want to clone it, for whatever reasons, check the | |
382 | following points instead (in this order). This is a very destructive | |
383 | approach, so *make a backup* or go back to cloning it. You have been | |
384 | warned. | |
385 | ||
386 | * Remove the original refs backed up by git-filter-branch: say `git | |
387 | for-each-ref \--format="%(refname)" refs/original/ | xargs -n 1 git | |
388 | update-ref -d`. | |
389 | ||
390 | * Expire all reflogs with `git reflog expire \--expire=now \--all`. | |
391 | ||
392 | * Garbage collect all unreferenced objects with `git gc \--prune=now` | |
393 | (or if your git-gc is not new enough to support arguments to | |
394 | `\--prune`, use `git repack -ad; git prune` instead). | |
395 | ||
396 | ||
c401b33c JS |
397 | Author |
398 | ------ | |
399 | Written by Petr "Pasky" Baudis <pasky@suse.cz>, | |
400 | and the git list <git@vger.kernel.org> | |
401 | ||
402 | Documentation | |
403 | -------------- | |
404 | Documentation by Petr Baudis and the git list. | |
405 | ||
406 | GIT | |
407 | --- | |
9e1f0a85 | 408 | Part of the linkgit:git[1] suite |