git-send-email: Add --thread option
[git/aroben.git] / git-filter-branch.sh
blob8fa5ce6467b12119868a4c3c74c3e3740f6d6b8e
1 #!/bin/sh
3 # Rewrite revision history
4 # Copyright (c) Petr Baudis, 2006
5 # Minimal changes to "port" it to core-git (c) Johannes Schindelin, 2007
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.
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.)
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.
29 # Always verify that the rewritten version is correct before disposing
30 # the original branch.
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.
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.
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.
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).
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.
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!).
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.
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.
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.
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:
98 # TREE_ID [-p PARENT_COMMIT_ID]...
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.
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.
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.
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.)
122 # --subdirectory-filter DIRECTORY:: Only regard the history, as seen by
123 # the given subdirectory. The result will contain that directory as
124 # its project root.
126 # EXAMPLE USAGE
127 # -------------
128 # Suppose you want to remove a file (containing confidential information
129 # or copyright violation) from all commits:
131 # git-filter-branch --tree-filter 'rm filename' newbranch
133 # A significantly faster version:
135 # git-filter-branch --index-filter 'git-update-index --remove filename' newbranch
137 # Now, you will get the rewritten history saved in the branch 'newbranch'
138 # (your current branch is left untouched).
140 # To "etch-graft" a commit to the revision history (set a commit to be
141 # the parent of the current initial commit and propagate that):
143 # git-filter-branch --parent-filter sed\ 's/^$/-p graftcommitid/' newbranch
145 # (if the parent string is empty - therefore we are dealing with the
146 # initial commit - add graftcommit as a parent). Note that this assumes
147 # history with a single root (that is, no git-merge without common ancestors
148 # happened). If this is not the case, use:
150 # git-filter-branch --parent-filter 'cat; [ "$GIT_COMMIT" = "COMMIT" ] && echo "-p GRAFTCOMMIT"' newbranch
152 # To remove commits authored by "Darl McBribe" from the history:
154 # git-filter-branch --commit-filter 'if [ "$GIT_AUTHOR_NAME" = "Darl McBribe" ]; then shift; while [ -n "$1" ]; do shift; echo "$1"; shift; done; else git-commit-tree "$@"; fi' newbranch
156 # (the shift magic first throws away the tree id and then the -p
157 # parameters). Note that this handles merges properly! In case Darl
158 # committed a merge between P1 and P2, it will be propagated properly
159 # and all children of the merge will become merge commits with P1,P2
160 # as their parents instead of the merge commit.
162 # To restrict rewriting to only part of the history, specify a revision
163 # range in addition to the new branch name. The new branch name will
164 # point to the top-most revision that a 'git rev-list' of this range
165 # will print.
167 # Consider this history:
169 # D--E--F--G--H
170 # / /
171 # A--B-----C
173 # To rewrite commits D,E,F,G,H, use:
175 # git-filter-branch ... new-H C..H
177 # To rewrite commits E,F,G,H, use one of these:
179 # git-filter-branch ... new-H C..H --not D
180 # git-filter-branch ... new-H D..H --not C
182 # To move the whole tree into a subdirectory, or remove it from there:
184 # git-filter-branch --index-filter \
185 # 'git-ls-files -s | sed "s-\t-&newsubdir/-" |
186 # GIT_INDEX_FILE=$GIT_INDEX_FILE.new \
187 # git-update-index --index-info &&
188 # mv $GIT_INDEX_FILE.new $GIT_INDEX_FILE' directorymoved
190 # Testsuite: TODO
192 set -e
194 USAGE="git-filter-branch [-d TEMPDIR] [FILTERS] DESTBRANCH [REV-RANGE]"
195 . git-sh-setup
197 map()
199 # if it was not rewritten, take the original
200 test -r "$workdir/../map/$1" || echo "$1"
201 cat "$workdir/../map/$1"
204 # When piped a commit, output a script to set the ident of either
205 # "author" or "committer
207 set_ident () {
208 lid="$(echo "$1" | tr "A-Z" "a-z")"
209 uid="$(echo "$1" | tr "a-z" "A-Z")"
210 pick_id_script='
211 /^'$lid' /{
212 s/'\''/'\''\\'\'\''/g
214 s/^'$lid' \([^<]*\) <[^>]*> .*$/\1/
215 s/'\''/'\''\'\'\''/g
216 s/.*/export GIT_'$uid'_NAME='\''&'\''/p
219 s/^'$lid' [^<]* <\([^>]*\)> .*$/\1/
220 s/'\''/'\''\'\'\''/g
221 s/.*/export GIT_'$uid'_EMAIL='\''&'\''/p
224 s/^'$lid' [^<]* <[^>]*> \(.*\)$/\1/
225 s/'\''/'\''\'\'\''/g
226 s/.*/export GIT_'$uid'_DATE='\''&'\''/p
232 LANG=C LC_ALL=C sed -ne "$pick_id_script"
233 # Ensure non-empty id name.
234 echo "[ -n \"\$GIT_${uid}_NAME\" ] || export GIT_${uid}_NAME=\"\${GIT_${uid}_EMAIL%%@*}\""
237 tempdir=.git-rewrite
238 filter_env=
239 filter_tree=
240 filter_index=
241 filter_parent=
242 filter_msg=cat
243 filter_commit='git-commit-tree "$@"'
244 filter_tag_name=
245 filter_subdir=
246 while case "$#" in 0) usage;; esac
248 case "$1" in
250 shift
251 break
256 break;
257 esac
259 # all switches take one argument
260 ARG="$1"
261 case "$#" in 1) usage ;; esac
262 shift
263 OPTARG="$1"
264 shift
266 case "$ARG" in
268 tempdir="$OPTARG"
270 --env-filter)
271 filter_env="$OPTARG"
273 --tree-filter)
274 filter_tree="$OPTARG"
276 --index-filter)
277 filter_index="$OPTARG"
279 --parent-filter)
280 filter_parent="$OPTARG"
282 --msg-filter)
283 filter_msg="$OPTARG"
285 --commit-filter)
286 filter_commit="$OPTARG"
288 --tag-name-filter)
289 filter_tag_name="$OPTARG"
291 --subdirectory-filter)
292 filter_subdir="$OPTARG"
295 usage
297 esac
298 done
300 dstbranch="$1"
301 shift
302 test -n "$dstbranch" || die "missing branch name"
303 git-show-ref "refs/heads/$dstbranch" 2> /dev/null &&
304 die "branch $dstbranch already exists"
306 test ! -e "$tempdir" || die "$tempdir already exists, please remove it"
307 mkdir -p "$tempdir/t"
308 cd "$tempdir/t"
309 workdir="$(pwd)"
311 case "$GIT_DIR" in
315 export GIT_DIR="$(pwd)/../../$GIT_DIR"
317 esac
319 export GIT_INDEX_FILE="$(pwd)/../index"
320 git-read-tree # seed the index file
322 ret=0
325 mkdir ../map # map old->new commit ids for rewriting parents
327 case "$filter_subdir" in
329 git-rev-list --reverse --topo-order --default HEAD \
330 --parents "$@"
333 git-rev-list --reverse --topo-order --default HEAD \
334 --parents --full-history "$@" -- "$filter_subdir"
335 esac > ../revs
336 commits=$(cat ../revs | wc -l | tr -d " ")
338 test $commits -eq 0 && die "Found nothing to rewrite"
341 while read commit parents; do
342 i=$(($i+1))
343 printf "$commit ($i/$commits) "
345 case "$filter_subdir" in
347 git-read-tree -i -m $commit
350 git-read-tree -i -m $commit:"$filter_subdir"
351 esac
353 export GIT_COMMIT=$commit
354 git-cat-file commit "$commit" >../commit
356 eval "$(set_ident AUTHOR <../commit)"
357 eval "$(set_ident COMMITTER <../commit)"
358 eval "$filter_env" < /dev/null
360 if [ "$filter_tree" ]; then
361 git-checkout-index -f -u -a
362 # files that $commit removed are now still in the working tree;
363 # remove them, else they would be added again
364 git-ls-files -z --others | xargs -0 rm -f
365 eval "$filter_tree" < /dev/null
366 git-diff-index -r $commit | cut -f 2- | tr '\n' '\0' | \
367 xargs -0 git-update-index --add --replace --remove
368 git-ls-files -z --others | \
369 xargs -0 git-update-index --add --replace --remove
372 eval "$filter_index" < /dev/null
374 parentstr=
375 for parent in $parents; do
376 for reparent in $(map "$parent"); do
377 parentstr="$parentstr -p $reparent"
378 done
379 done
380 if [ "$filter_parent" ]; then
381 parentstr="$(echo "$parentstr" | eval "$filter_parent")"
384 sed -e '1,/^$/d' <../commit | \
385 eval "$filter_msg" | \
386 sh -c "$filter_commit" git-commit-tree $(git-write-tree) $parentstr | \
387 tee ../map/$commit
388 done <../revs
390 src_head=$(tail -n 1 ../revs | sed -e 's/ .*//')
391 target_head=$(head -n 1 ../map/$src_head)
392 case "$target_head" in
394 echo Nothing rewritten
397 git-update-ref refs/heads/"$dstbranch" $target_head
398 if [ $(cat ../map/$src_head | wc -l) -gt 1 ]; then
399 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 $dstbranch head but you will need to resolve the situation now (probably by manually merging the other commits). These are all the commits:" >&2
400 sed 's/^/ /' ../map/$src_head >&2
401 ret=1
404 esac
406 if [ "$filter_tag_name" ]; then
407 git-for-each-ref --format='%(objectname) %(objecttype) %(refname)' refs/tags |
408 while read sha1 type ref; do
409 ref="${ref#refs/tags/}"
410 # XXX: Rewrite tagged trees as well?
411 if [ "$type" != "commit" -a "$type" != "tag" ]; then
412 continue;
415 if [ "$type" = "tag" ]; then
416 # Dereference to a commit
417 sha1t="$sha1"
418 sha1="$(git-rev-parse "$sha1"^{commit} 2>/dev/null)" || continue
421 [ -f "../map/$sha1" ] || continue
422 new_sha1="$(cat "../map/$sha1")"
423 export GIT_COMMIT="$sha1"
424 new_ref="$(echo "$ref" | eval "$filter_tag_name")"
426 echo "$ref -> $new_ref ($sha1 -> $new_sha1)"
428 if [ "$type" = "tag" ]; then
429 # Warn that we are not rewriting the tag object itself.
430 warn "unreferencing tag object $sha1t"
433 git-update-ref "refs/tags/$new_ref" "$new_sha1"
434 done
437 cd ../..
438 rm -rf "$tempdir"
439 echo "Rewritten history saved to the $dstbranch branch"
441 exit $ret