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 # --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.
 125 #
 126 # EXAMPLE USAGE
 127 # -------------
 128 # Suppose you want to remove a file (containing confidential information
 129 # or copyright violation) from all commits:
 130 #
 131 #       git-filter-branch --tree-filter 'rm filename' newbranch
 132 #
 133 # A significantly faster version:
 134 #
 135 #       git-filter-branch --index-filter 'git-update-index --remove filename' newbranch
 136 #
 137 # Now, you will get the rewritten history saved in the branch 'newbranch'
 138 # (your current branch is left untouched).
 139 #
 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):
 142 #
 143 #       git-filter-branch --parent-filter sed\ 's/^$/-p graftcommitid/' newbranch
 144 #
 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:
 149 #
 150 #       git-filter-branch --parent-filter 'cat; [ "$GIT_COMMIT" = "COMMIT" ] && echo "-p GRAFTCOMMIT"' newbranch
 151 #
 152 # To remove commits authored by "Darl McBribe" from the history:
 153 #
 154 #       git-filter-branch --commit-filter 'if [ "$GIT_AUTHOR_NAME" = "Darl McBribe" ]; then shift; while [ -n "$1" ]; do shift; echo "$1"; shift; done; else git-commit-tree "$@"; fi' newbranch
 155 #
 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.
 161 #
 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.
 166 #
 167 # Consider this history:
 168 #
 169 #            D--E--F--G--H
 170 #           /     /
 171 #       A--B-----C
 172 #
 173 # To rewrite commits D,E,F,G,H, use:
 174 #
 175 #       git-filter-branch ... new-H C..H
 176 #
 177 # To rewrite commits E,F,G,H, use one of these:
 178 #
 179 #       git-filter-branch ... new-H C..H --not D
 180 #       git-filter-branch ... new-H D..H --not C
 181
 182 # Testsuite: TODO
 183
 184 set -e
 185
 186 USAGE="git-filter-branch [-d TEMPDIR] [FILTERS] DESTBRANCH [REV-RANGE]"
 187 . git-sh-setup
 188
 189 map()
 190 {
 191         # if it was not rewritten, take the original
 192         test -r "$workdir/../map/$1" || echo "$1"
 193         cat "$workdir/../map/$1"
 194 }
 195
 196 # When piped a commit, output a script to set the ident of either
 197 # "author" or "committer
 198
 199 set_ident () {
 200         lid="$(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 tempdir=.git-rewrite
 230 filter_env=
 231 filter_tree=
 232 filter_index=
 233 filter_parent=
 234 filter_msg=cat
 235 filter_commit='git-commit-tree "$@"'
 236 filter_tag_name=
 237 filter_subdir=
 238 while case "$#" 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         --subdirectory-filter)
 284                 filter_subdir="$OPTARG"
 285                 ;;
 286         *)
 287                 usage
 288                 ;;
 289         esac
 290 done
 291
 292 dstbranch="$1"
 293 shift
 294 test -n "$dstbranch" || die "missing branch name"
 295 git-show-ref "refs/heads/$dstbranch" 2> /dev/null &&
 296         die "branch $dstbranch already exists"
 297
 298 test ! -e "$tempdir" || die "$tempdir already exists, please remove it"
 299 mkdir -p "$tempdir/t"
 300 cd "$tempdir/t"
 301 workdir="$(pwd)"
 302
 303 case "$GIT_DIR" in
 304 /*)
 305         ;;
 306 *)
 307         export GIT_DIR="$(pwd)/../../$GIT_DIR"
 308         ;;
 309 esac
 310
 311 export GIT_INDEX_FILE="$(pwd)/../index"
 312 git-read-tree # seed the index file
 313
 314 ret=0
 315
 316
 317 mkdir ../map # map old->new commit ids for rewriting parents
 318
 319 case "$filter_subdir" in
 320 "")
 321         git-rev-list --reverse --topo-order --default HEAD \
 322                 --parents "$@"
 323         ;;
 324 *)
 325         git-rev-list --reverse --topo-order --default HEAD \
 326                 --parents "$@" -- "$filter_subdir"
 327 esac > ../revs
 328 commits=$(cat ../revs | wc -l | tr -d " ")
 329
 330 test $commits -eq 0 && die "Found nothing to rewrite"
 331
 332 i=0
 333 while read commit parents; do
 334         i=$(($i+1))
 335         printf "$commit ($i/$commits) "
 336
 337         case "$filter_subdir" in
 338         "")
 339                 git-read-tree -i -m $commit
 340                 ;;
 341         *)
 342                 git-read-tree -i -m $commit:"$filter_subdir"
 343         esac
 344
 345         export GIT_COMMIT=$commit
 346         git-cat-file commit "$commit" >../commit
 347
 348         eval "$(set_ident AUTHOR <../commit)"
 349         eval "$(set_ident COMMITTER <../commit)"
 350         eval "$filter_env" < /dev/null
 351
 352         if [ "$filter_tree" ]; then
 353                 git-checkout-index -f -u -a
 354                 # files that $commit removed are now still in the working tree;
 355                 # remove them, else they would be added again
 356                 git-ls-files -z --others | xargs -0 rm -f
 357                 eval "$filter_tree" < /dev/null
 358                 git-diff-index -r $commit | cut -f 2- | tr '\n' '\0' | \
 359                         xargs -0 git-update-index --add --replace --remove
 360                 git-ls-files -z --others | \
 361                         xargs -0 git-update-index --add --replace --remove
 362         fi
 363
 364         eval "$filter_index" < /dev/null
 365
 366         parentstr=
 367         for parent in $parents; do
 368                 for reparent in $(map "$parent"); do
 369                         parentstr="$parentstr -p $reparent"
 370                 done
 371         done
 372         if [ "$filter_parent" ]; then
 373                 parentstr="$(echo "$parentstr" | eval "$filter_parent")"
 374         fi
 375
 376         sed -e '1,/^$/d' <../commit | \
 377                 eval "$filter_msg" | \
 378                 sh -c "$filter_commit" git-commit-tree $(git-write-tree) $parentstr | \
 379                 tee ../map/$commit
 380 done <../revs
 381
 382 src_head=$(tail -n 1 ../revs | sed -e 's/ .*//')
 383 target_head=$(head -n 1 ../map/$src_head)
 384 case "$target_head" in
 385 '')
 386         echo Nothing rewritten
 387         ;;
 388 *)
 389         git-update-ref refs/heads/"$dstbranch" $target_head
 390         if [ $(cat ../map/$src_head | wc -l) -gt 1 ]; then
 391                 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 $dstbranch head but you will need to resolve the situation now (probably by manually merging the other commits). These are all the commits:" >&2
 392                 sed 's/^/       /' ../map/$src_head >&2
 393                 ret=1
 394         fi
 395         ;;
 396 esac
 397
 398 if [ "$filter_tag_name" ]; then
 399         git-for-each-ref --format='%(objectname) %(objecttype) %(refname)' refs/tags |
 400         while read sha1 type ref; do
 401                 ref="${ref#refs/tags/}"
 402                 # XXX: Rewrite tagged trees as well?
 403                 if [ "$type" != "commit" -a "$type" != "tag" ]; then
 404                         continue;
 405                 fi
 406
 407                 if [ "$type" = "tag" ]; then
 408                         # Dereference to a commit
 409                         sha1t="$sha1"
 410                         sha1="$(git-rev-parse "$sha1"^{commit} 2>/dev/null)" || continue
 411                 fi
 412
 413                 [ -f "../map/$sha1" ] || continue
 414                 new_sha1="$(cat "../map/$sha1")"
 415                 export GIT_COMMIT="$sha1"
 416                 new_ref="$(echo "$ref" | eval "$filter_tag_name")"
 417
 418                 echo "$ref -> $new_ref ($sha1 -> $new_sha1)"
 419
 420                 if [ "$type" = "tag" ]; then
 421                         # Warn that we are not rewriting the tag object itself.
 422                         warn "unreferencing tag object $sha1t"
 423                 fi
 424
 425                 git-update-ref "refs/tags/$new_ref" "$new_sha1"
 426         done
 427 fi
 428
 429 cd ../..
 430 rm -rf "$tempdir"
 431 echo "Rewritten history saved to the $dstbranch branch"
 432
 433 exit $ret