repos

`repos` is a CLI tool that helps to manage multiple git repositories.

Installation

git clone https://github.com/amadev/repos.git && cd repos && sudo ./install.sh

Basic operations

Add some repositories with `repos add` command.

repos add https://review.opendev.org/openstack/nova
repos add https://review.opendev.org/openstack/cinder

Repository names must be unique. By default the last segment of repository path used as a name, so conflict could happen. To overcome, use the `–name` option.

The following basic operations are available.

repos list
repos pull

Repositories could be organized hierarchically into groups and operations can be issued on a particular group only. The most useful way of using groups is a `search` operation that helps to search on multiples repositories not depending on a path they were cloned.

repos search --group qemu needle -- --ignore tests

The search backend could be changed with a ‘search_command` config option.

`repos sync`

A `repos sync` command pulls and pushes repositories to make sure everything is synchronized. As a pure man dropbox alternative, you can mark your repo with `auto-commit` tag like `repos add … –tags auto-commit` and `repos sync` will auto-commit changes in the repository, pull changes from upstream and push resulting set to upstream. Surely this may be applied to a specific group only.

Group hierarchy

With `–group` and `–parent` options, repositories can be easily organized into a hierarchy.

repos add-group group-1
repos add repo-1 --group group-1
repos add-group group-1.1 --parent group-1
repos add repo-1.1 --group group-1.1

Group names must be unique. Be careful with circular dependencies.

Config options

Config by default resided in $HOME/.repos/conf.yaml.

• db: path where `repos` db yaml stored;
• groups: path to group hierarchy symlinks;
• search_command - a template for command, `grep -HRin` by default, but `ag -f –search-files` recommended;
• repos_default_directory - all repositories without `–path` specification will be cloned here;
• log - path to a log file.

Repository tags

• auto-commit - for `repos sync` command it will stage all changes and commit them;
• shallow - quick clone for big repositories pushing is not planned from.

Advanced usage

Besides CLI commands, there is a way to edit DB directly. Especially as DB is a simple YAML file.

Just after installation, a DB file can be found in $HOME/.repos/db.yaml. The actual path can be modified via config DB property (See Config options).

Basically, it has two entities:

Group:

• name
• parent

Repo:

• name - the name of a repository
• path - absolute path in FS where a repository resided
• tags - special properties of a repository (see Repository tags)
• group - group name, that must be defined in Group entity
• remotes - a list of git remotes (not available via CLI)
• branches - a list of branches (not available via CLI)
• config - a list of git config keys and values (not available via CLI)

An example of DB can be seen in file:tests/fixtures/db.yaml

Gerrit repositories download

`repos gerrit <url>` generates a database file can be used later with `repos` commands to download all gerrit projects hierarchy.

First, let’s create a new config file and export its path as an env variable.

cat << 'EOF' > ~/.repos/mcp-config.yaml
db: $HOME/.repos/mcp-db.yaml
groups: $HOME/.repos/mcp-groups/
search_command: ag -f --search-files
repos_default_directory: $HOME/src/mcp/
EOF

export REPOS_CONFIG=~/.repos/mcp-config.yaml

Then download gerrit project and convert them into a `repos` DB format. As Gerrit may require authentication an additional http headers could be provided.

repos gerrit \
  https://gerrit.mcp.mirantis.com/ \
  ssh://avolkov@gerrit.mcp.mirantis.com:29418/ --headers "$HEADERS" \
  > $HOME/.repos/mcp-db.yaml

And last step is to make pull.

repos pull