9d61b7fff6670bcf40e4135749cec9bdb3fd10ae
[git/git.git] / git-filter-branch.sh
1 #!/bin/sh
2 #
3 # Rewrite revision history
4 # Copyright (c) Petr Baudis, 2006
5 # Minimal changes to "port" it to core-git (c) Johannes Schindelin, 2007
6 #
7 # Lets you rewrite GIT revision history by creating a new branch from
8 # your current branch by applying custom filters on each revision.
9 # Those filters can modify each tree (e.g. removing a file or running
10 # a perl rewrite on all files) or information about each commit.
11 # Otherwise, all information (including original commit times or merge
12 # information) will be preserved.
13 #
14 # The command takes the new branch name as a mandatory argument and
15 # the filters as optional arguments. If you specify no filters, the
16 # commits will be recommitted without any changes, which would normally
17 # have no effect and result with the new branch pointing to the same
18 # branch as your current branch. (Nevertheless, this may be useful in
19 # the future for compensating for some Git bugs or such, therefore
20 # such a usage is permitted.)
21 #
22 # WARNING! The rewritten history will have different ids for all the
23 # objects and will not converge with the original branch. You will not
24 # be able to easily push and distribute the rewritten branch. Please do
25 # not use this command if you do not know the full implications, and
26 # avoid using it anyway - do not do what a simple single commit on top
27 # of the current version would fix.
28 #
29 # Always verify that the rewritten version is correct before disposing
30 # the original branch.
31 #
32 # Note that since this operation is extensively I/O expensive, it might
33 # be a good idea to do it off-disk, e.g. on tmpfs. Reportedly the speedup
34 # is very noticeable.
35 #
36 # OPTIONS
37 # -------
38 # -d TEMPDIR:: The path to the temporary tree used for rewriting
39 # When applying a tree filter, the command needs to temporary
40 # checkout the tree to some directory, which may consume
41 # considerable space in case of large projects. By default it
42 # does this in the '.git-rewrite/' directory but you can override
43 # that choice by this parameter.
44 #
45 # Filters
46 # ~~~~~~~
47 # The filters are applied in the order as listed below. The COMMAND
48 # argument is always evaluated in shell using the 'eval' command.
49 # The $GIT_COMMIT environment variable is permanently set to contain
50 # the id of the commit being rewritten. The author/committer environment
51 # variables are set before the first filter is run.
52 #
53 # A 'map' function is available that takes an "original sha1 id" argument
54 # and outputs a "rewritten sha1 id" if the commit has been already
55 # rewritten, fails otherwise; the 'map' function can return several
56 # ids on separate lines if your commit filter emitted multiple commits
57 # (see below).
58 #
59 # --env-filter COMMAND:: The filter for modifying environment
60 # This is the filter for modifying the environment in which
61 # the commit will be performed. Specifically, you might want
62 # to rewrite the author/committer name/email/time environment
63 # variables (see `git-commit` for details). Do not forget to
64 # re-export the variables.
65 #
66 # --tree-filter COMMAND:: The filter for rewriting tree (and its contents)
67 # This is the filter for rewriting the tree and its contents.
68 # The COMMAND argument is evaluated in shell with the working
69 # directory set to the root of the checked out tree. The new tree
70 # is then used as-is (new files are auto-added, disappeared files
71 # are auto-removed - .gitignore files nor any other ignore rules
72 # HAVE NO EFFECT!).
73 #
74 # --index-filter COMMAND:: The filter for rewriting index
75 # This is the filter for rewriting the Git's directory index.
76 # It is similar to the tree filter but does not check out the
77 # tree, which makes it much faster. However, you must use the
78 # lowlevel Git index manipulation commands to do your work.
79 #
80 # --parent-filter COMMAND:: The filter for rewriting parents
81 # This is the filter for rewriting the commit's parent list.
82 # It will receive the parent string on stdin and shall output
83 # the new parent string on stdout. The parent string is in
84 # format accepted by `git-commit-tree`: empty for initial
85 # commit, "-p parent" for a normal commit and "-p parent1
86 # -p parent2 -p parent3 ..." for a merge commit.
87 #
88 # --msg-filter COMMAND:: The filter for rewriting commit message
89 # This is the filter for rewriting the commit messages.
90 # The COMMAND argument is evaluated in shell with the original
91 # commit message on standard input; its standard output is
92 # is used as the new commit message.
93 #
94 # --commit-filter COMMAND:: The filter for performing the commit
95 # If this filter is passed, it will be called instead of the
96 # `git-commit-tree` command, with those arguments:
97 #
98 # TREE_ID [-p PARENT_COMMIT_ID]...
99 #
100 # and the log message on stdin. The commit id is expected on
101 # stdout. As a special extension, the commit filter may emit
102 # multiple commit ids; in that case, all of them will be used
103 # as parents instead of the original commit in further commits.
104 #
105 # --tag-name-filter COMMAND:: The filter for rewriting tag names.
106 # If this filter is passed, it will be called for every tag ref
107 # that points to a rewritten object (or to a tag object which
108 # points to a rewritten object). The original tag name is passed
109 # via standard input, and the new tag name is expected on standard
110 # output.
111 #
112 # The original tags are not deleted, but can be overwritten;
113 # use "--tag-name-filter=cat" to simply update the tags. In this
114 # case, be very careful and make sure you have the old tags
115 # backed up in case the conversion has run afoul.
116 #
117 # Note that there is currently no support for proper rewriting of
118 # tag objects; in layman terms, if the tag has a message or signature
119 # attached, the rewritten tag won't have it. Sorry. (It is by
120 # definition impossible to preserve signatures at any rate, though.)
121 #
122 # EXAMPLE USAGE
123 # -------------
124 # Suppose you want to remove a file (containing confidential information
125 # or copyright violation) from all commits:
126 #
127 # git-filter-branch --tree-filter 'rm filename' newbranch
128 #
129 # A significantly faster version:
130 #
131 # git-filter-branch --index-filter 'git-update-index --remove filename' newbranch
132 #
133 # Now, you will get the rewritten history saved in the branch 'newbranch'
134 # (your current branch is left untouched).
135 #
136 # To "etch-graft" a commit to the revision history (set a commit to be
137 # the parent of the current initial commit and propagate that):
138 #
139 # git-filter-branch --parent-filter sed\ 's/^$/-p graftcommitid/' newbranch
140 #
141 # (if the parent string is empty - therefore we are dealing with the
142 # initial commit - add graftcommit as a parent). Note that this assumes
143 # history with a single root (that is, no git-merge without common ancestors
144 # happened). If this is not the case, use:
145 #
146 # git-filter-branch --parent-filter 'cat; [ "$GIT_COMMIT" = "COMMIT" ] && echo "-p GRAFTCOMMIT"' newbranch
147 #
148 # To remove commits authored by "Darl McBribe" from the history:
149 #
150 # git-filter-branch --commit-filter 'if [ "$GIT_AUTHOR_NAME" = "Darl McBribe" ]; then shift; while [ -n "$1" ]; do shift; echo "$1"; shift; done; else git-commit-tree "$@"; fi' newbranch
151 #
152 # (the shift magic first throws away the tree id and then the -p
153 # parameters). Note that this handles merges properly! In case Darl
154 # committed a merge between P1 and P2, it will be propagated properly
155 # and all children of the merge will become merge commits with P1,P2
156 # as their parents instead of the merge commit.
157 #
158 # To restrict rewriting to only part of the history, specify a revision
159 # range in addition to the new branch name. The new branch name will
160 # point to the top-most revision that a 'git rev-list' of this range
161 # will print.
162 #
163 # Consider this history:
164 #
165 # D--E--F--G--H
166 # / /
167 # A--B-----C
168 #
169 # To rewrite commits D,E,F,G,H, use:
170 #
171 # git-filter-branch ... new-H C..H
172 #
173 # To rewrite commits E,F,G,H, use one of these:
174 #
175 # git-filter-branch ... new-H C..H --not D
176 # git-filter-branch ... new-H D..H --not C
177
178 # Testsuite: TODO
179
180 set -e
181
182 USAGE="git-filter-branch [-d TEMPDIR] [FILTERS] DESTBRANCH [REV-RANGE]"
183 . git-sh-setup
184
185 map()
186 {
187 # if it was not rewritten, take the original
188 test -r "$workdir/../map/$1" || echo "$1"
189 cat "$workdir/../map/$1"
190 }
191
192 # When piped a commit, output a script to set the ident of either
193 # "author" or "committer
194
195 set_ident () {
196 lid="$(echo "$1" | tr "A-Z" "a-z")"
197 uid="$(echo "$1" | tr "a-z" "A-Z")"
198 pick_id_script='
199 /^'$lid' /{
200 s/'\''/'\''\\'\'\''/g
201 h
202 s/^'$lid' \([^<]*\) <[^>]*> .*$/\1/
203 s/'\''/'\''\'\'\''/g
204 s/.*/export GIT_'$uid'_NAME='\''&'\''/p
205
206 g
207 s/^'$lid' [^<]* <\([^>]*\)> .*$/\1/
208 s/'\''/'\''\'\'\''/g
209 s/.*/export GIT_'$uid'_EMAIL='\''&'\''/p
210
211 g
212 s/^'$lid' [^<]* <[^>]*> \(.*\)$/\1/
213 s/'\''/'\''\'\'\''/g
214 s/.*/export GIT_'$uid'_DATE='\''&'\''/p
215
216 q
217 }
218 '
219
220 LANG=C LC_ALL=C sed -ne "$pick_id_script"
221 # Ensure non-empty id name.
222 echo "[ -n \"\$GIT_${uid}_NAME\" ] || export GIT_${uid}_NAME=\"\${GIT_${uid}_EMAIL%%@*}\""
223 }
224
225 # list all parent's object names for a given commit
226 get_parents () {
227 git-rev-list -1 --parents "$1" | sed "s/^[0-9a-f]*//"
228 }
229
230 tempdir=.git-rewrite
231 filter_env=
232 filter_tree=
233 filter_index=
234 filter_parent=
235 filter_msg=cat
236 filter_commit='git-commit-tree "$@"'
237 filter_tag_name=
238 while case "$#" in 0) usage;; esac
239 do
240 case "$1" in
241 --)
242 shift
243 break
244 ;;
245 -*)
246 ;;
247 *)
248 break;
249 esac
250
251 # all switches take one argument
252 ARG="$1"
253 case "$#" in 1) usage ;; esac
254 shift
255 OPTARG="$1"
256 shift
257
258 case "$ARG" in
259 -d)
260 tempdir="$OPTARG"
261 ;;
262 --env-filter)
263 filter_env="$OPTARG"
264 ;;
265 --tree-filter)
266 filter_tree="$OPTARG"
267 ;;
268 --index-filter)
269 filter_index="$OPTARG"
270 ;;
271 --parent-filter)
272 filter_parent="$OPTARG"
273 ;;
274 --msg-filter)
275 filter_msg="$OPTARG"
276 ;;
277 --commit-filter)
278 filter_commit="$OPTARG"
279 ;;
280 --tag-name-filter)
281 filter_tag_name="$OPTARG"
282 ;;
283 *)
284 usage
285 ;;
286 esac
287 done
288
289 dstbranch="$1"
290 shift
291 test -n "$dstbranch" || die "missing branch name"
292 git-show-ref "refs/heads/$dstbranch" 2> /dev/null &&
293 die "branch $dstbranch already exists"
294
295 test ! -e "$tempdir" || die "$tempdir already exists, please remove it"
296 mkdir -p "$tempdir/t"
297 cd "$tempdir/t"
298 workdir="$(pwd)"
299
300 case "$GIT_DIR" in
301 /*)
302 ;;
303 *)
304 export GIT_DIR="$(pwd)/../../$GIT_DIR"
305 ;;
306 esac
307
308 export GIT_INDEX_FILE="$(pwd)/../index"
309 git-read-tree # seed the index file
310
311 ret=0
312
313
314 mkdir ../map # map old->new commit ids for rewriting parents
315
316 git-rev-list --reverse --topo-order --default HEAD "$@" >../revs
317 commits=$(cat ../revs | wc -l | tr -d " ")
318
319 test $commits -eq 0 && die "Found nothing to rewrite"
320
321 i=0
322 while read commit; do
323 i=$(($i+1))
324 printf "$commit ($i/$commits) "
325
326 git-read-tree -i -m $commit
327
328 export GIT_COMMIT=$commit
329 git-cat-file commit "$commit" >../commit
330
331 eval "$(set_ident AUTHOR <../commit)"
332 eval "$(set_ident COMMITTER <../commit)"
333 eval "$filter_env" < /dev/null
334
335 if [ "$filter_tree" ]; then
336 git-checkout-index -f -u -a
337 # files that $commit removed are now still in the working tree;
338 # remove them, else they would be added again
339 git-ls-files -z --others | xargs -0 rm -f
340 eval "$filter_tree" < /dev/null
341 git-diff-index -r $commit | cut -f 2- | tr '\n' '\0' | \
342 xargs -0 git-update-index --add --replace --remove
343 git-ls-files -z --others | \
344 xargs -0 git-update-index --add --replace --remove
345 fi
346
347 eval "$filter_index" < /dev/null
348
349 parentstr=
350 for parent in $(get_parents $commit); do
351 for reparent in $(map "$parent"); do
352 parentstr="$parentstr -p $reparent"
353 done
354 done
355 if [ "$filter_parent" ]; then
356 parentstr="$(echo "$parentstr" | eval "$filter_parent")"
357 fi
358
359 sed -e '1,/^$/d' <../commit | \
360 eval "$filter_msg" | \
361 sh -c "$filter_commit" git-commit-tree $(git-write-tree) $parentstr | \
362 tee ../map/$commit
363 done <../revs
364
365 src_head=$(tail -n 1 ../revs)
366 target_head=$(head -n 1 ../map/$src_head)
367 case "$target_head" in
368 '')
369 echo Nothing rewritten
370 ;;
371 *)
372 git-update-ref refs/heads/"$dstbranch" $target_head
373 if [ $(cat ../map/$src_head | wc -l) -gt 1 ]; then
374 echo "WARNING: Your commit filter caused the head commit to expand to several rewritten commits. Only the first such commit was recorded as the current $dstbranch head but you will need to resolve the situation now (probably by manually merging the other commits). These are all the commits:" >&2
375 sed 's/^/ /' ../map/$src_head >&2
376 ret=1
377 fi
378 ;;
379 esac
380
381 if [ "$filter_tag_name" ]; then
382 git-for-each-ref --format='%(objectname) %(objecttype) %(refname)' refs/tags |
383 while read sha1 type ref; do
384 ref="${ref#refs/tags/}"
385 # XXX: Rewrite tagged trees as well?
386 if [ "$type" != "commit" -a "$type" != "tag" ]; then
387 continue;
388 fi
389
390 if [ "$type" = "tag" ]; then
391 # Dereference to a commit
392 sha1t="$sha1"
393 sha1="$(git-rev-parse "$sha1"^{commit} 2>/dev/null)" || continue
394 fi
395
396 [ -f "../map/$sha1" ] || continue
397 new_sha1="$(cat "../map/$sha1")"
398 export GIT_COMMIT="$sha1"
399 new_ref="$(echo "$ref" | eval "$filter_tag_name")"
400
401 echo "$ref -> $new_ref ($sha1 -> $new_sha1)"
402
403 if [ "$type" = "tag" ]; then
404 # Warn that we are not rewriting the tag object itself.
405 warn "unreferencing tag object $sha1t"
406 fi
407
408 git-update-ref "refs/tags/$new_ref" "$new_sha1"
409 done
410 fi
411
412 cd ../..
413 rm -rf "$tempdir"
414 echo "Rewritten history saved to the $dstbranch branch"
415
416 exit $ret