git-snapshot-sh - a script to save junk code

git-snapshot.sh

#!/bin/sh
#
# git-snapshot.sh - save junk code to a snapshots branch
#
# USAGE: git-snapshot.sh [-m message] [files...]
#
# Run `git add files` and then create a new commit on branch 'snapshot'
# with two parents, the previous snapshot and HEAD.
#
# If -m is given, the 'message' is used as the commit message; otherwise
# "snapshot" is used.  This option must be given as the first argument.
#
# The 'files' argument can be any argument for git add; it defaults to "-a".

# Configuration:
BRANCH=refs/heads/snapshot  # full refname of branch to update
MESSAGE=snapshot            # default commit message
EXT=.snapshot               # extension for temporary index file


# Parse the -m argument.
if [ "$1" = "-m" ]; then
    if [ $# -lt 2 ]; then
        echo "ERROR: -m takes an argument" >&2
        exit 1
    fi
    MESSAGE="$2"
    shift 2
fi

ORIGINAL_INDEX=`git rev-parse --git-dir`/index
GIT_INDEX_FILE=$ORIGINAL_INDEX$EXT
export GIT_INDEX_FILE

# If the last command failed, exit with its error code after restoring the
# index.
e() {
    if [ $? -ne 0 ]; then
        RC=$?
        rm $GIT_INDEX_FILE
        exit $RC
    fi
}

# Use a temporary index file.
cp $ORIGINAL_INDEX $GIT_INDEX_FILE || exit $?

# Save the commit ID of where we currently are.
HEAD_ID=`git rev-parse HEAD` ; e

# Save the commit ID of the old head of the snapshots branch.
SNAPSHOT_ID=`git rev-parse $BRANCH 2>/dev/null`
if [ $? -ne 0 ]; then
    # The branch doesn't exist.  Create it with an initial empty commit.
    # For a description of these commands, see below.
    rm $GIT_INDEX_FILE
    TREE_ID=`git write-tree` ; e
    SNAPSHOT_ID=`echo initial snapshot | git commit-tree $TREE_ID` ; e
    cp $ORIGINAL_INDEX $GIT_INDEX_FILE
fi

# Add all of the files listed on the command line, or use '-a' if no arguments
# given.
if [ $# -eq 0 ]; then
    git add --all ; e
else
    git add "$@" ; e
fi

# Record the index as a tree in the object database, and save its ID.
TREE_ID=`git write-tree` ; e

# Create a new commit with the tree we just created, the message given,
# and two parents: the previous snapshot and the current HEAD.
# Save the ID of this new commit.
COMMIT_ID=`echo "$MESSAGE" | \
    git commit-tree $TREE_ID -p $SNAPSHOT_ID -p $HEAD_ID` ; e

# Update our branch to point to this new commit.
git update-ref $BRANCH $COMMIT_ID ; e

# Restore the original index.
rm $GIT_INDEX_FILE || exit $?