Extract old and new versions of specific files throughout git history

extract-feature-files.sh

#!/usr/bin/env bash
#
# Run this script from the root of a git repo to create directories containing
# the old and new version of a .feature file.
#
# A directory will be created for each commit that modifies a .feature file,
# and inside this directory there will be a directory with the name of the file.
# Finally, inside that directory there will be a `new` and `old` file showing
# the contents before and after the commit.
#
# The commit directories are prefixed with the unix timestamp of the commit.
# This makes it possible to sort them in chronological order during further processing.
#
# Example:
#
# tmp/
# ├── 1445335992-b22c74de8d6c2667b715ad47e1f190a9b23ce58d
# │   └── features
# │       └── publish_features.feature
# │           ├── new
# │           └── old
# ├── 1453378524-b6525f8ef499889d8dcceb9eda56de1ad0e75315
# │   └── features
# │       ├── create_projects.feature
# │       │   ├── new
# │       │   └── old
# │       └── publish_features.feature
# │           ├── new
# │           └── old
# └── 1459348969-cf95d79349a24eeb68f47528a6e25189b180fc6b
#     └── features
#         └── publish_features.feature
#             ├── new
#             └── old
#
#
#

out_dir="./tmp"

# Loop over all commit SHAs, starting with the oldest one
git log --reverse --reflog --pretty=format:"%H" | \
while read -r commit; do
  # Loop over all feature files that were modified in this commit
  git diff-tree -r ${commit} | \
  grep "\.feature$" | \
  grep --invert-match "0000000000000000000000000000000000000000" | \
  while read -r blob_line; do
    # Create a directory for the commit SHA followed by the filename
    path=$(echo "${blob_line}" | cut -d$'\t' -f 2)
    timestamp=$(git show -s --format=%ct ${commit})
    dir="${out_dir}/${timestamp}-${commit}/${path}"
    mkdir -p "${dir}"

    # Write the old and new file
    old_oid=$(echo "${blob_line}" | cut -d' ' -f3)
    new_oid=$(echo "${blob_line}" | cut -d' ' -f4)
    git show "${old_oid}" > "${dir}/old"
    git show "${new_oid}" > "${dir}/new"

    # Print some status
    echo "${commit} ${path}"
  done
done