Google drive for jupyter notebooks
JavaScript TypeScript Python Makefile
Latest commit 5458133 Mar 23, 2016 @Carreau Carreau Merge pull request #136 from rjw57/fix-auth-timeout
Fix Google API timeout in long-lived notebooks

README.md

Google Drive support for Jupyter Notebook

This repository contains custom Contents classes that allows IPython to use Google Drive for file management. The code is organized as a python package that contains functions to install a Jupyter Notebook JavaScript extension, and activate/deactivate different IPython profiles to be used with Google drive.

To install this package, run

git clone git://github.com/jupyter/jupyter-drive.git
pip install -e jupyter-drive

This will install the package in development mode with pip, which means that any change you make to the repository will be reflected into the importable version immediately.

To install the notebook extension and activate your configuration with Google Drive, run

python -m jupyterdrive

To deactivate, run

python -m jupyterdrive --deactivate

Note on Jupyter/IPython

We try to support both IPython 3.x and above version, though many changes in configuration between IPython 3.x and later versions may cause the exact configuration path to vary from system to system.

Symlink files

By default all the Javascript files will be symlinked, so any update to the package python or javascript will be reflected on your python server. Use the --no-symlink option to actually copy the file on their final destination.

Mixed contents

You can also install the MixedContentsManager, to do so install as before with the --mixed options. For example, on the default profile:

python -m jupyterdrive --mixed

The mixed content manager will show contents from both the local hard drive and the remote google drive as two directories in your dashboard.

mixed content configuration.

All of the following files should be created automatically the first time that you run the content manager. You might need to modify these files manually to get the contents manager to work.

To modify the configuration of the mixed contents manager you need to update the following files:

Under Jupyter/IPython 4.x+ <config-dir>/jupyter_notebook_config.json which by default should have the following structure :

{
  "nbformat": 1,
  "NotebookApp": {
    "contents_manager_class": "jupyterdrive.mixednbmanager.MixedContentsManager",
    "tornado_settings": {
      "contents_js_source": "nbextensions/gdrive/mixed-contents"
    }
  },
  "MixedContentsManager":{
    "filesystem_scheme": [
        {
          "root":"local",
          "contents":"IPython.html.services.contents.filemanager.FileContentsManager"
        },
        {
          "root": "gdrive",
          "contents": "jupyterdrive.clientsidenbmanager.ClientSideContentsManager"
        }
      ]
  }

}

Under IPython 3.x This file would be <profile>/ipython_notebook_config.json.

The root field of filesystem_scheme represents the name that would be used as the virtual mount points for the contents manager in the dashbord and should be consistent with the name given in nbconfig/common.json described below.

The contents field contains the fully qualified name of a Contents manager to mount on the mountpoint.

The second config file that deals with configuring the frontend should be:

  {
    "schema": [
      {
        "stripjs": false,
        "contents": "services/contents",
        "root": "local"
      },
      {
        "stripjs": true,
        "contents": "./drive-contents",
        "root": "gdrive"
      }
    ]
  }

To modify it, in a notebook access the config object, and extract default value: >>> data = IPython.notebook.contents.config.data . Change one of the values, for example : >>> data.mixed_contents.schema[0].root = 'new_root'. Update the config to the new value : >>> IPython.notebook.contents.config.update(data).

As stated previously the root value should match python side config file, contents represents the client-side content manager that need to be used. stripjs is a boolean value that indicates whether the name of the mount point should be stripped from the various paths on the javascript side before passing it to the different subcontent managers.

Other options

If IPython has been installed system wide, in a virtual environment or with some custom setup you might need to pass --user, -prefix <prefix> or --path <path> option to install the extension in a user-owned profile location.

See python -m jupyterdrive --help for more information.

deactivate a profile

It is not yet possible to deactivate the drive integration automatically. But you can get rid of ~/.ipython/profile_defaut/ipython_notebook_config.json config file to deactivate Google Drive as well as other config options.

The configuration of IPython/Jupyter is done through the .json file in the profile and will take precedence on configuration done in .py files, though you can still use non conflicting options.

First launch

Launch IPython with the profile in which you activated jupyter-drive:

ipython notebook --profile <profilename>

On first launch, the application will ask you for the authorization to access your files on Google Drive. It only asks for permission to create new files or to access files it has created or that you manually open with this application. It also requires permission to access file/directory metadata, in order to display the list of files/directories in the tree view.

The request pop-up looks like the following:

auth screenshot

Clicking ok will open a Google OAuth pop-up. You will see that the Jupyter Drive application wants access to some information about your files. Keep that in mind if you want to revoke access at a later point.

Once you click Accept you should be able to start creating new notebooks on Google Drive, and open existing ones created by this application, and view files/directories in the tree view.

Advanced configuration

The contents manager can access the common section of nbconfig, thus you can set config values in <profile_dir>/nbconfig/common.json. The default values that are used are the following:

{"gdrive":
    {
    "METADATA_SCOPE": true,
    "FILE_SCOPE": true,
    "CLIENT_ID": "763546234320-uvcktfp0udklafjqv00qjgivpjh0t33p.apps.googleusercontent.com"
    }
}

The APP_ID section is not yet configurable, but should be configurable in the same way at some point in the future.

Developing

The client-side code for jupyter-drive is written in TypeScript. The TypeScript transpiler can be installed using npm:

npm install

The JavaScript may then be re-compiled using npm run:

npm run gulp js