How to recreate git commit history but with applying pre-commit hook?

Question

I'm trying to rewrite git repository history and apply new pre-commit hook:

1. Take every commit
2. Apply pre-commit hook
3. Keep the original metadata (author, date, message)
4. Resolve conflicts manually, if any (the hook can alter the commit)
5. Commit to a new repo

The end state is a new repo with a different commit history.

What I already found:

• cherry-pick doesn't run pre-commit hook.
• I can do

git cherry-pick --no-commit
git commit --no-edit

But it doesn't preserve the commit date. Also, not sure how to do that for each commit in history (unless I write a e.g. Python script for that).

Any ideas on how to do that efficiently?

Answer 1

Use the --exec flag to git rebase, possibly with a custom GIT_SEQUENCE_EDITOR to skip the interactive prompt with the pick list. So something like:

GIT_SEQUENCE_EDIT=cat git rebase --root --exec .git/hooks/pre-commit

This will add exec .git/hooks/pre-commit after every pick <commit> in the pick list. If the pre-commit hooks fails, that will interrupt the rebase:

Executing: .git/hooks/pre-commit
warning: execution failed: .git/hooks/pre-commit
You can fix the problem, and then run

  git rebase --continue

You can manually resolve the issues, and then git rebase --continue.

Answer 2

Using @larsks's idea, I achieved the goal using the following commands. I'm leaving it here for reference.

1. Create an empty branch [1]:

git switch --orphan new_history

2. Rebase the main branch onto it, executing pre-commit hook on every commit:

git rebase --root main --exec .git/hooks/pre-commit -X theirs

Note that I used theirs merge strategy, so that in case of conflicts, the version from the main branch would be used ([2], [3]).

3. After that, the commit date was set to today in all commits. To fix that, I ran ([4]):

git filter-branch --env-filter 'export GIT_COMMITTER_DATE="$GIT_AUTHOR_DATE"' 

Here is my pre-commit script:

#!/bin/sh

set -x 

IFS=$'\n'

files=$(git diff-tree --no-commit-id  --name-status -r HEAD | grep -v ^D | cut -c3-)

if [ "$files" != "" ]
then

  for f in $files
  do
      if [[ "$f" =~ [.]md$ ]]
      then

          # Add a linebreak to the file if it doesn't have one
          if [ "$(tail -c1 $f)" != '\n' ]
          then
            echo >> $f
          fi

          # Remove trailing whitespace if it exists
          if grep -q "[[:blank:]]$" $f
          then
            sed -i "" -e $'s/[ \t]*$//g' $f
          fi

          # Remove blank lines
          python3 ~/bin/remove_empty_lines_md.py $f $f 
      fi
  done

fi

unset IFS

git add -A
git commit --amend --no-edit --no-verify

Note that --no-verify in the end is important (it skips hooks), otherwise there is an infinite loop.

For completeness, here is the remove_empty_lines_md.py script:

def remove_empty_lines(text):
    r"""
    >>> text = "\n\n- test\n  - test2\n\n- test3\n\n\n  cointinue\n\n- test 4\n\n- test 5\n  - test 6\n    continue\n    continue\n\n\n\n    test7\n\n\n    tset8\n\n\n- test 9\n\n\n- final\n\n\n"
    >>> remove_empty_lines(text)
    '- test\n  - test2\n- test3\n\n\n  cointinue\n- test 4\n- test 5\n  - test 6\n    continue\n    continue\n\n\n\n    test7\n\n\n    tset8\n- test 9\n- final\n'
    """    
    new_lines = []
    buffer = []
    for i, line in enumerate(text.split('\n')):
        if not line.strip():
            buffer.append(line)
        else:
            if buffer:
                if line.lstrip() and line.lstrip()[0] != '-':
                    new_lines.extend(buffer)
                buffer = []
            new_lines.append(line)
    return '\n'.join(new_lines) + '\n'


if __name__ == '__main__':
    import doctest
    import sys
    doctest.testmod()
    filename_in, filename_out = sys.argv[1], sys.argv[2]
    with open(filename_in) as fin:
        text = fin.read()
    with open(filename_out, 'w') as fout:
        fout.write(remove_empty_lines(text))