restore the original git history

[milahu]

continue garycourt/uri-js#95 (comment)

PR restoring git history would be welcome

this is impossible via github

---

PR from milahu/uri-js fails because kibertoad/toad-uri-js is not a fork of garycourt/uri-js

I've done forking of other libraries in the past, it gives a very bad developer experience, as then all PRs default to targetting parent repositories, ending up in creating accidental PRs

main...milahu:uri-js:master

There isn’t anything to compare.
We couldn’t figure out how to compare these references, do they point to valid commits?

---

PR from milahu/toad-uri-js fails because the root commit is different, aka "not related git histories"

main...milahu:toad-uri-js:main

There isn’t anything to compare.
kibertoad:main and milahu:main are entirely different commit histories.

---

so as i said...

please help yourself

here is an updated script, to also preserve the committer data

restore-git-history.sh

#!/usr/bin/env bash

set -e
set -x

git clone https://github.com/garycourt/uri-js
mv uri-js toad-uri-js
cd toad-uri-js
git branch -M master main || true
git remote add fork https://github.com/kibertoad/toad-uri-js
git fetch fork main
git worktree add fork-main fork/main

# this is the first commit in the fork
# git log --format=fuller
cat >/dev/null <<'EOF'
commit ab2a6cab3b60120db334118640f0417d0c7a7f52
Author:     Igor Savin <[email protected]>
AuthorDate: Fri Feb 9 20:40:20 2024 +0200
Commit:     GitHub <[email protected]>
CommitDate: Fri Feb 9 20:40:20 2024 +0200

    Migrate to vitest (#1)
EOF

git_commit() {
  local an="$1"
  local ae="$2"
  local ad="$3"
  local m="$4"
  GIT_AUTHOR_NAME="$an" GIT_AUTHOR_EMAIL="$ae" GIT_AUTHOR_DATE="$ad" \
  GIT_COMMITTER_NAME="$an" GIT_COMMITTER_EMAIL="$ae" GIT_COMMITTER_DATE="$ad" \
  git commit -m "$m"
}

an="Igor Savin"
ae="[email protected]"

git rm -rf dist/
git_commit "$an" "$ae" "Fri Feb 9 20:20:00 2024 +0200" "rm -rf dist/"

git -C fork-main checkout fork/main~12
cp fork-main/.gitignore .gitignore
git add .gitignore
git_commit "$an" "$ae" "Fri Feb 9 20:30:00 2024 +0200" "up .gitignore"

# now there should be no difference
# -> ready for cherry-pick
git_diff=$(git diff main~0 fork/main~12)
if [[ "$git_diff" != "" ]]; then
  echo "error: diff is not empty:"
  echo "$git_diff"
  exit 1
fi



# git-cherry-pick-preserve-committer.sh

# bash script to modify committer after git cherry-pick

# either to preserve the original committer
# or to use author as committer
# similar to git rebase --committer-date-is-author-date

#     #echo 'format = "%an\n%ae\n%at\n%ad"' # committer is author
#     echo 'format = "%cn\n%ce\n%ct\n%cd"' # preserve committer

set -e

# preserve committer in "git cherry-pick"

function git_cherry_pick_preserve_committer() {

  # filter commits by range of the "git cherry-pick"
  local head_before_cherry_pick=$(git rev-parse HEAD)
  echo "head_before_cherry_pick: $head_before_cherry_pick"

  # -x: append "(cherry picked from commit ...)" to the original commit message
  local args=(git cherry-pick -x "$@")
  echo -n "running:"; printf " %q" "${args[@]}"; echo
  "${args[@]}"

  # run "git filter-repo" to postprocess the "git cherry-pick"

  # Python code body for processing commit objects
  commit_callback=$(
    #echo 'print("locals", locals())'

    echo 'message_lines = commit.message.split(b"\n")'
    #echo 'print("message_lines", repr(message_lines))'
    echo 'try:'
    echo '  picked_from = message_lines[-2].decode("ascii")'
    echo 'except IndexError:'
    echo '  return'
    echo 'if not picked_from.startswith("(cherry picked from commit "):'
    echo '  return'

    # print newline after "Parsed 123 commits"
    #echo 'print()'

    #echo 'print("message_lines", repr(message_lines))'
    echo 'assert len(picked_from) == 68, f"unexpected picked_from: {repr(picked_from)}"'
    echo 'picked_from = picked_from[27:-1]'

    #echo 'for a in ["author_date", "author_email", "author_name", "message", "branch", "id", "old_id", "original_id", "parents", "type"]:'
    #echo '  print(f"commit.{a}", repr(getattr(commit, a)))'

    #echo 'print("picked_from", repr(picked_from))'

    # _do_not_use_this_var is passed via callback_metadata
    # lets use it! :P
    # we only need repo_filter._repo_working_dir
    # 'commit_rename_func': <bound method RepoFilter._translate_commit_hash of <git_filter_repo.RepoFilter object at ...>>
    echo 'repo_filter = _do_not_use_this_var["commit_rename_func"].__self__'
    #echo 'print("repo_filter", repo_filter)'

    if false; then
    # filter commits by range of the "git cherry-pick"
    echo "head_before_cherry_pick = '$head_before_cherry_pick'"
    echo 'args = ["git", "merge-base", "--is-ancestor", head_before_cherry_pick, commit.original_id]'
    #echo 'print("args", repr(args))'
    echo 'proc = subproc.run(args, cwd=repo_filter._repo_working_dir)'
    echo 'if proc.returncode == 1:'
    echo '  return'
    fi

    #echo 'format = "%an\n%ae\n%at\n%ad"' # committer is author
    echo 'format = "%cn\n%ce\n%ct\n%cd"' # preserve committer

    echo 'args = ["git", "show", "-s", f"--format=format:{format}", picked_from]'
    #echo 'print("args", repr(args))'
    echo 'proc = subproc.Popen(args, stdout=subprocess.PIPE, cwd=repo_filter._repo_working_dir)'
    echo 'lines = proc.stdout.read().splitlines()'
    echo 'if len(lines) != 4:'
    echo '  print("lines", repr(lines))'
    echo '  raise Exception(f"not found picked_from commit {picked_from}")'
    echo 'commit.committer_name = lines[0]'
    echo 'commit.committer_email = lines[1]'
    echo 'commit.committer_date = lines[2] + lines[3][-6:]' # timestamp + timezone

    echo 'message_lines = message_lines[:-2]' # remove "cherry picked from" line
    echo 'commit.message = b"\n".join(message_lines)'
    #echo 'print("commit.message", repr(commit.message))'
  )

  # preserve-commit-hashes
  # fix: fatal: bad object xxx
  # git-filter-repo patches hashes in commit.message

  git filter-repo \
    --force \
    --preserve-commit-hashes \
    --commit-callback "$commit_callback" \
    --refs "$head_before_cherry_pick..HEAD"

}

#git cherry-pick fork/main~12..fork/main
git_cherry_pick_preserve_committer fork/main~12..fork/main



# use author date as committer date
# github shows only the committer date
#git rebase a1acf730b4bba3f1097c9f52e7d9d3aba8cdcaae --committer-date-is-author-date

# undo the cherry-pick
#git reset --hard main~12

git tag 5.0.0 main~4

git tag 5.0.1 main~0

echo "done. verify:"
echo "git -C toad-uri-js/ log --format=fuller"

... then git push --force

[kibertoad]

thanks, I appreciate the instructions. was too swamped to do it yet, but I will get back to it ASAP

[milahu]

push

come on, this is low hanging fruit, this takes 5 minutes

[kibertoad]

thank you for the reminder, I will try to do this over the weekend