init-monorepo.sh

README.md

init-monorepo

TODO

• we can keep the history under the subdirectory via --to-subdirectory-filter

git filter-repo --source "$tmp_dir/$package" --target "$tmp_dir/$package" --to-subdirectory-filter "packages/$package" --commit-callback "commit.message += b'\n\nPreMonorepoCommit: https://github.com/$repository/commit/' + commit.original_id + b'\n'"

summary

• git merge --allow-unrelated-histories to merge multiple repositories without common parent commit
• git filter-repo to inject links to the original commit on Github
• git tag to rename existing tags

example

• original repositories
  • https://github.com/hi-ogawa/demo-init-monorepo--node-csv-stringify
  • https://github.com/hi-ogawa/demo-init-monorepo--node-csv-generate
  • https://github.com/hi-ogawa/demo-init-monorepo--node-csv-parse
• result
  • https://github.com/hi-ogawa/demo-init-monorepo

shellcheck -x init-monorepo.sh

# initialize monorepo locally
bash init-monorepo.sh |& tee -a init-monorepo.log

# push to remote if the result looks okay
cd demo-init-monorepo
git remote add origin [email protected]:hi-ogawa/demo-init-monorepo.git
git push origin init-monorepo # --force
git push origin --tags # --force

misc

• reset tags (when testing repeatedly in the same repository)

git fetch
git tag -l | xargs -n 1 git push --delete origin
git tag -l | xargs -n 1 git tag -d

• port old diff to monorepo (e.g. to migrate old PRs)

# Create diff in an old repository (e.g. diff from master to HEAD in "some-module" repository)
git diff master > some-module.diff

# Convert diff
python prefix-diff.py packages/some-module < some-module.diff > some-module.monorepo.diff

# Apply diff in monorepo
git apply some-module.monorepo.diff

references

• https://www.adaltas.com/en/2021/05/21/js-monorepos-merging-git-repositories/
• https://github.com/newren/git-filter-repo

init-monorepo-config.sh

#!/bin/bash
set -eoux pipefail

monorepo_root="$PWD/demo-init-monorepo"
monorepo_branch=init-monorepo
repositories=(
  hi-ogawa/demo-init-monorepo--node-csv-stringify
  hi-ogawa/demo-init-monorepo--node-csv-generate
  hi-ogawa/demo-init-monorepo--node-csv-parse
)
packages_dir=packages
tmp_dir="$PWD/tmp"

init-monorepo.sh

#!/bin/bash
set -eoux pipefail

source ./init-monorepo-config.sh

#
# Validate
#

if ! type git-filter-repo >/dev/null 2>&1; then
  echo "error: git-filter-repo not found"
  exit 1
fi

if [ -e "$monorepo_root" ]; then
  echo "error: not empty '$monorepo_root'"
  exit 1
fi

#
# Setup
#

rm -rf "$tmp_dir"
mkdir -p "$tmp_dir" "$monorepo_root/$packages_dir"
cd "$monorepo_root"
git init
git checkout -b "$monorepo_branch"

#
# Merge each repository
#

for repository in "${repositories[@]}"; do
  package="${repository##*/}"
  mkdir -p "$tmp_dir/$package" "$packages_dir/$package"
  git clone "[email protected]:$repository.git" "$tmp_dir/$package"

  #
  # Manipulate history within a temporary repository
  #

  # inject old commit links (convenient to find relevant PRs via git blame)
  git filter-repo --source "$tmp_dir/$package" --target "$tmp_dir/$package" --commit-callback "commit.message += b'\n\nPreMonorepoCommit: https://github.com/$repository/commit/' + commit.original_id + b'\n'"

  # rename tags (prepend a package name)
  cd "$tmp_dir/$package"
  for tag in $(git tag -l); do
    git tag "$package-$tag" "$tag"
    git tag -d "$tag"
  done
  cd -

  #
  # Add to monorepo as remote branch and merge
  #

  git remote add -f "$package" "$tmp_dir/$package"
  default_branch=$(git -C "$tmp_dir/$package" rev-parse --abbrev-ref HEAD)
  git merge --allow-unrelated-histories "$package/$default_branch" -m "chore(monorepo): merge '$package'"

  #
  # Move files and commit
  #

  find . -maxdepth 1 -not \( -name '.' -o -name '.git' -o -name "$packages_dir" \) -exec mv '{}' "$packages_dir/$package" \;
  git add .
  git commit -m "chore(monorepo): create '$packages_dir/$package'"
done

prefix-diff.py

# cf. https://stackoverflow.com/a/2530012

import re
import argparse
import sys


def main(prefix: str) -> None:
    for line in sys.stdin:
        #
        # diff --git a/src/app/hello.ts b/src/app/foo.ts
        #   ⇓
        # diff --git a/PREFIX/src/app/hello.ts b/PREFIX/src/app/foo.ts
        #
        line = re.sub(
            r"^diff --git a/(.*) b/(.*)$",
            f"diff --git a/{prefix}/\\1 b/{prefix}/\\2",
            line,
        )

        #
        # --- a/src/app/hello.ts
        #   ⇓
        # --- a/PREFIX/src/app/hello.ts
        #
        line = re.sub(r"^--- a/(.*)$", f"--- a/{prefix}/\\1", line)

        #
        # +++ b/src/app/foo.ts
        #
        # +++ b/PREFIX/src/app/foo.ts
        #
        line = re.sub(r"^\+\+\+ b/(.*)$", f"+++ b/{prefix}/\\1", line)

        print(line, end="")


def main_cli() -> None:
    parser = argparse.ArgumentParser()
    parser.add_argument("prefix")
    main(**parser.parse_args().__dict__)


if __name__ == "__main__":
    main_cli()