Entrypoint Consolidation: All Functionality accessible via mlflow-export-import

[juftin]

This PR addresses Issue #29

Changes

1) All Click Commands have been added to a new master Click Group

2) All entrypoints have been replaced with the new Click Group:

❯ mlflow-export-import --help
Usage: mlflow-export-import [OPTIONS] COMMAND [ARGS]...

  MLflow Export / Import CLI: Command Line Interface

Options:
  --version  Show the version and exit.
  --help     Show this message and exit.

Commands:
  export-all          Export the entire tracking server.
  export-experiment   Exports an experiment to a directory.
  export-experiments  Exports experiments to a directory.
  export-model        Export a registered model.
  export-models       Exports models and their versions' backing.
  export-run          Exports a run to a directory.
  find-artifacts      Find artifacts that match a filename.
  http-client         Interact with the MLflow / Databricks HTTP Client.
  import-all          Imports models and their experiments and runs.
  import-experiment   Imports an experiment from a directory.
  import-experiments  Import a list of experiments from a directory.
  import-model        Import a registered model.
  import-run          Imports a run from a directory.
  list-models         Lists all registered models.

3) All packaging configuration moved to setup.cfg - setup.py inherits from this automatically

• This also includes a pyproject.toml file which instructs Python to use setuptools as the build tool. Again nothing actually changes here - we're just using the setup.cfg file to be a little more declarative and having no arbitrary code execution at install time via setup.py

4) Version management moved to the _version.py file - this allows us to do things like from mlflow_export_import import __version__

5) Standard Python .gitignore added (with mlruns/ exclusion)

Notes

• Missing Entrypoints
  • import-models - this was a duplicate of import-all
  • export-model-list - this code is missing

[juftin]

Here's where all the magic happens to consolidate each click command. You can distribute a click CLI over different files in a few ways, but this way results in the least refactor.

[juftin]

Here's how the official MLflow CLI is organized using a similar method: https://github.com/mlflow/mlflow/blob/fd1f3aa246ca55a6dc0121ecf582c48b1c0287d0/mlflow/cli.py#L578-L583

[juftin]

Here's where I tell the main function that it's click command is "list-models". Click will handle this for us automatically if the function being decorated with click.command() was named list_models.

I also place the docstring on the function - click will automatically pass this to the CLI