Add git-filter-branch
[git/git.git] / git-filter-branch.sh
CommitLineData
6f6826c5
JS
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# -r STARTREV:: The commit id to start the rewrite at
46# Normally, the command will rewrite the entire history. If you
47# pass this argument, though, this will be the first commit it
48# will rewrite and keep the previous commits intact.
49#
50# -k KEEPREV:: A commit id until which _not_ to rewrite history
51# If you pass this argument, this commit and all of its
52# predecessors are kept intact.
53#
54# Filters
55# ~~~~~~~
56# The filters are applied in the order as listed below. The COMMAND
57# argument is always evaluated in shell using the 'eval' command.
58# The $GIT_COMMIT environment variable is permanently set to contain
59# the id of the commit being rewritten. The author/committer environment
60# variables are set before the first filter is run.
61#
62# A 'map' function is available that takes an "original sha1 id" argument
63# and outputs a "rewritten sha1 id" if the commit has been already
64# rewritten, fails otherwise; the 'map' function can return several
65# ids on separate lines if your commit filter emitted multiple commits
66# (see below).
67#
68# --env-filter COMMAND:: The filter for modifying environment
69# This is the filter for modifying the environment in which
70# the commit will be performed. Specifically, you might want
71# to rewrite the author/committer name/email/time environment
72# variables (see `git-commit` for details). Do not forget to
73# re-export the variables.
74#
75# --tree-filter COMMAND:: The filter for rewriting tree (and its contents)
76# This is the filter for rewriting the tree and its contents.
77# The COMMAND argument is evaluated in shell with the working
78# directory set to the root of the checked out tree. The new tree
79# is then used as-is (new files are auto-added, disappeared files
80# are auto-removed - .gitignore files nor any other ignore rules
81# HAVE NO EFFECT!).
82#
83# --index-filter COMMAND:: The filter for rewriting index
84# This is the filter for rewriting the Git's directory index.
85# It is similar to the tree filter but does not check out the
86# tree, which makes it much faster. However, you must use the
87# lowlevel Git index manipulation commands to do your work.
88#
89# --parent-filter COMMAND:: The filter for rewriting parents
90# This is the filter for rewriting the commit's parent list.
91# It will receive the parent string on stdin and shall output
92# the new parent string on stdout. The parent string is in
93# format accepted by `git-commit-tree`: empty for initial
94# commit, "-p parent" for a normal commit and "-p parent1
95# -p parent2 -p parent3 ..." for a merge commit.
96#
97# --msg-filter COMMAND:: The filter for rewriting commit message
98# This is the filter for rewriting the commit messages.
99# The COMMAND argument is evaluated in shell with the original
100# commit message on standard input; its standard output is
101# is used as the new commit message.
102#
103# --commit-filter COMMAND:: The filter for performing the commit
104# If this filter is passed, it will be called instead of the
105# `git-commit-tree` command, with those arguments:
106#
107# TREE_ID [-p PARENT_COMMIT_ID]...
108#
109# and the log message on stdin. The commit id is expected on
110# stdout. As a special extension, the commit filter may emit
111# multiple commit ids; in that case, all of them will be used
112# as parents instead of the original commit in further commits.
113#
114# --tag-name-filter COMMAND:: The filter for rewriting tag names.
115# If this filter is passed, it will be called for every tag ref
116# that points to a rewritten object (or to a tag object which
117# points to a rewritten object). The original tag name is passed
118# via standard input, and the new tag name is expected on standard
119# output.
120#
121# The original tags are not deleted, but can be overwritten;
122# use "--tag-name-filter=cat" to simply update the tags. In this
123# case, be very careful and make sure you have the old tags
124# backed up in case the conversion has run afoul.
125#
126# Note that there is currently no support for proper rewriting of
127# tag objects; in layman terms, if the tag has a message or signature
128# attached, the rewritten tag won't have it. Sorry. (It is by
129# definition impossible to preserve signatures at any rate, though.)
130#
131# EXAMPLE USAGE
132# -------------
133# Suppose you want to remove a file (containing confidential information
134# or copyright violation) from all commits:
135#
136# git-filter-branch --tree-filter 'rm filename' newbranch
137#
138# A significantly faster version:
139#
140# git-filter-branch --index-filter 'git-update-index --remove filename' newbranch
141#
142# Now, you will get the rewritten history saved in the branch 'newbranch'
143# (your current branch is left untouched).
144#
145# To "etch-graft" a commit to the revision history (set a commit to be
146# the parent of the current initial commit and propagate that):
147#
148# git-filter-branch --parent-filter sed\ 's/^$/-p graftcommitid/' newbranch
149#
150# (if the parent string is empty - therefore we are dealing with the
151# initial commit - add graftcommit as a parent). Note that this assumes
152# history with a single root (that is, no git-merge without common ancestors
153# happened). If this is not the case, use:
154#
155# git-filter-branch --parent-filter 'cat; [ "$GIT_COMMIT" = "COMMIT" ] && echo "-p GRAFTCOMMIT"' newbranch
156#
157# To remove commits authored by "Darl McBribe" from the history:
158#
159# git-filter-branch --commit-filter 'if [ "$GIT_AUTHOR_NAME" = "Darl McBribe" ]; then shift; while [ -n "$1" ]; do shift; echo "$1"; shift; done; else git-commit-tree "$@"; fi' newbranch
160#
161# (the shift magic first throws away the tree id and then the -p
162# parameters). Note that this handles merges properly! In case Darl
163# committed a merge between P1 and P2, it will be propagated properly
164# and all children of the merge will become merge commits with P1,P2
165# as their parents instead of the merge commit.
166#
167# To restrict rewriting to only part of the history, use -r or -k or both.
168# Consider this history:
169#
170# D--E--F--G--H
171# / /
172# A--B-----C
173#
174# To rewrite only commits F,G,H, use:
175#
176# git-filter-branch -r F ...
177#
178# To rewrite commits E,F,G,H, use one of these:
179#
180# git-filter-branch -r E -k C ...
181# git-filter-branch -k D -k C ...
182
183# Testsuite: TODO
184
185set -e
186
187USAGE="git-filter-branch [-d TEMPDIR] [-r STARTREV]... [-k KEEPREV]... [-s SRCBRANCH] [FILTERS] DESTBRANCH"
188. git-sh-setup
189
190map()
191{
192 [ -r "$workdir/../map/$1" ] || return 1
193 cat "$workdir/../map/$1"
194}
195
196# When piped a commit, output a script to set the ident of either
197# "author" or "committer
198
199set_ident () {
200 lid="$(echo "$1" | tr "A-Z" "a-z")"
201 uid="$(echo "$1" | tr "a-z" "A-Z")"
202 pick_id_script='
203 /^'$lid' /{
204 s/'\''/'\''\\'\'\''/g
205 h
206 s/^'$lid' \([^<]*\) <[^>]*> .*$/\1/
207 s/'\''/'\''\'\'\''/g
208 s/.*/export GIT_'$uid'_NAME='\''&'\''/p
209
210 g
211 s/^'$lid' [^<]* <\([^>]*\)> .*$/\1/
212 s/'\''/'\''\'\'\''/g
213 s/.*/export GIT_'$uid'_EMAIL='\''&'\''/p
214
215 g
216 s/^'$lid' [^<]* <[^>]*> \(.*\)$/\1/
217 s/'\''/'\''\'\'\''/g
218 s/.*/export GIT_'$uid'_DATE='\''&'\''/p
219
220 q
221 }
222 '
223
224 LANG=C LC_ALL=C sed -ne "$pick_id_script"
225 # Ensure non-empty id name.
226 echo "[ -n \"\$GIT_${uid}_NAME\" ] || export GIT_${uid}_NAME=\"\${GIT_${uid}_EMAIL%%@*}\""
227}
228
229# list all parent's object names for a given commit
230get_parents () {
231 git-rev-list -1 --parents "$1" | sed "s/^[0-9a-f]*//"
232}
233
234tempdir=.git-rewrite
235unchanged=" "
236filter_env=
237filter_tree=
238filter_index=
239filter_parent=
240filter_msg=cat
241filter_commit='git-commit-tree "$@"'
242filter_tag_name=
243srcbranch=HEAD
244while case "$#" in 0) usage;; esac
245do
246 case "$1" in
247 --)
248 shift
249 break
250 ;;
251 -*)
252 ;;
253 *)
254 break;
255 esac
256
257 # all switches take one argument
258 ARG="$1"
259 case "$#" in 1) usage ;; esac
260 shift
261 OPTARG="$1"
262 shift
263
264 case "$ARG" in
265 -d)
266 tempdir="$OPTARG"
267 ;;
268 -r)
269 unchanged="$(get_parents "$OPTARG") $unchanged"
270 ;;
271 -k)
272 unchanged="$(git-rev-parse "$OPTARG"^{commit}) $unchanged"
273 ;;
274 --env-filter)
275 filter_env="$OPTARG"
276 ;;
277 --tree-filter)
278 filter_tree="$OPTARG"
279 ;;
280 --index-filter)
281 filter_index="$OPTARG"
282 ;;
283 --parent-filter)
284 filter_parent="$OPTARG"
285 ;;
286 --msg-filter)
287 filter_msg="$OPTARG"
288 ;;
289 --commit-filter)
290 filter_commit="$OPTARG"
291 ;;
292 --tag-name-filter)
293 filter_tag_name="$OPTARG"
294 ;;
295 -s)
296 srcbranch="$OPTARG"
297 ;;
298 *)
299 usage
300 ;;
301 esac
302done
303
304dstbranch="$1"
305test -n "$dstbranch" || die "missing branch name"
306git-show-ref "refs/heads/$dstbranch" 2> /dev/null &&
307 die "branch $dstbranch already exists"
308
309test ! -e "$tempdir" || die "$tempdir already exists, please remove it"
310mkdir -p "$tempdir/t"
311cd "$tempdir/t"
312workdir="$(pwd)"
313
314case "$GIT_DIR" in
315/*)
316 ;;
317*)
318 export GIT_DIR="$(pwd)/../../$GIT_DIR"
319 ;;
320esac
321
322export GIT_INDEX_FILE="$(pwd)/../index"
323git-read-tree # seed the index file
324
325ret=0
326
327
328mkdir ../map # map old->new commit ids for rewriting parents
329
330# seed with identity mappings for the parents where we start off
331for commit in $unchanged; do
332 echo $commit > ../map/$commit
333done
334
335git-rev-list --reverse --topo-order $srcbranch --not $unchanged >../revs
336commits=$(cat ../revs | wc -l | tr -d " ")
337
338test $commits -eq 0 && die "Found nothing to rewrite"
339
340i=0
341while read commit; do
342 i=$((i+1))
343 printf "$commit ($i/$commits) "
344
345 git-read-tree -i -m $commit
346
347 export GIT_COMMIT=$commit
348 git-cat-file commit "$commit" >../commit
349
350 eval "$(set_ident AUTHOR <../commit)"
351 eval "$(set_ident COMMITTER <../commit)"
352 eval "$filter_env"
353
354 if [ "$filter_tree" ]; then
355 git-checkout-index -f -u -a
356 # files that $commit removed are now still in the working tree;
357 # remove them, else they would be added again
358 git-ls-files -z --others | xargs -0 rm -f
359 eval "$filter_tree"
360 git-diff-index -r $commit | cut -f 2- | tr '\n' '\0' | \
361 xargs -0 git-update-index --add --replace --remove
362 git-ls-files -z --others | \
363 xargs -0 git-update-index --add --replace --remove
364 fi
365
366 eval "$filter_index"
367
368 parentstr=
369 for parent in $(get_parents $commit); do
370 if [ -r "../map/$parent" ]; then
371 for reparent in $(cat "../map/$parent"); do
372 parentstr="$parentstr -p $reparent"
373 done
374 else
375 die "assertion failed: parent $parent for commit $commit not found in rewritten ones"
376 fi
377 done
378 if [ "$filter_parent" ]; then
379 parentstr="$(echo "$parentstr" | eval "$filter_parent")"
380 fi
381
382 sed -e '1,/^$/d' <../commit | \
383 eval "$filter_msg" | \
384 sh -c "$filter_commit" git-commit-tree $(git-write-tree) $parentstr | \
385 tee ../map/$commit
386done <../revs
387
388git-update-ref refs/heads/"$dstbranch" $(head -n 1 ../map/$(tail -n 1 ../revs))
389if [ "$(cat ../map/$(tail -n 1 ../revs) | wc -l)" -gt 1 ]; then
390 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 $dstbranch head but you will need to resolve the situation now (probably by manually merging the other commits). These are all the commits:" >&2
391 sed 's/^/ /' ../map/$(tail -n 1 ../revs) >&2
392 ret=1
393fi
394
395if [ "$filter_tag_name" ]; then
396 git-for-each-ref --format='%(objectname) %(objecttype) %(refname)' refs/tags |
397 while read sha1 type ref; do
398 ref="${ref#refs/tags/}"
399 # XXX: Rewrite tagged trees as well?
400 if [ "$type" != "commit" -a "$type" != "tag" ]; then
401 continue;
402 fi
403
404 if [ "$type" = "tag" ]; then
405 # Dereference to a commit
406 sha1t="$sha1"
407 sha1="$(git-rev-parse "$sha1"^{commit} 2>/dev/null)" || continue
408 fi
409
410 [ -f "../map/$sha1" ] || continue
411 new_sha1="$(cat "../map/$sha1")"
412 export GIT_COMMIT="$sha1"
413 new_ref="$(echo "$ref" | eval "$filter_tag_name")"
414
415 echo "$ref -> $new_ref ($sha1 -> $new_sha1)"
416
417 if [ "$type" = "tag" ]; then
418 # Warn that we are not rewriting the tag object itself.
419 warn "unreferencing tag object $sha1t"
420 fi
421
422 git-update-ref "refs/tags/$new_ref" "$new_sha1"
423 done
424fi
425
426cd ../..
427rm -rf "$tempdir"
428echo "Rewritten history saved to the $dstbranch branch"
429
430exit $ret