Default behavior is the same

Executors:

• Default: run() w/ a pass-through to show real time stderr of progress

• subprocess.run pass-through

• SubprocessCommand generates functions in dataclasses you can invoke later

  Rather than immediately running a SubprocessCommand, you can introspect, pickle it, adjust it, etc.

  Git(executor=SubprocessCommand)

• Bring your own

  If you're not satisfied and don't think a PR will do, create your own executor class.

  If you do, you may want to either 1.) override the class's #run() or 2.) pass a run_fn to override it.