JupyterOnRails

Integrate Rails and Jupyter

Motivation

Although it was already possible to run jupyter + iruby and require app_full_path to load Rails application context, it is a bit tiring to each time copy all the require statements.

Moreover, since Rails Way works in keeping (generally) everything under the project directory, managing the jupyter configuration installed in user global area (the iruby kernel register thing) is again a bit awkward.

With this gem, these awkwardness is to solved by following instruments:

• rake jupyter:notebook Railtie command which invokes jupyter at your project root, and
• This rake task automatically creates two iruby kernels to load your application in Jupyter:
  • one for normal operation
  • one for sandbox mode in which everything you do is wrapped in a database transaction that is rolled back when the kernel exits (equivalent to rails console --sandbox)

Prerequisites

• iruby's prerequisites must be met.

  • Refer: https://github.com/SciRuby/iruby
    • Either cztop or ffi-rzmq gem must be installable.

• jupyter command must be somehow available.

  • Either
    • jupyter command (pip global install, anaconda, etc), or
    • pipenv run jupyter command (managed by Pipfile at project root)

Installation

Add these lines to your application's Gemfile:

gem 'jupyter_on_rails'

# For sessions pick either:
gem 'ffi-rzmq'
# Or
gem 'cztop'

And then execute:

$ bundle install

Usage

Starting the jupyter server is available as a rake command.

Just execute:

rake jupyter:notebook

Eventually, you'll have jupyter opened, and the kernel being available.

ApplicationRecord.to_df/#to_df

User.to_df # => Daru::DataFrame is returned
user.to_df # => Daru::DataFrame for single record

Daru::DataFrame#write_model

For each row, call to_h and pass it to model.new. And then, imports them using model.insert_all on Rails >= 6, or model.import (requires adding activerecord-import to your Gemfile).

df.write_model(User)

Development

This is a railtie gem, so you'd probably want to do something like:

gem 'jupyter_on_rails', git: 'GIT_URL_OF_YOUR_REPO',
                        branch: 'the-work-branch'

or

gem `jupyter_on_rails`, path: 'jupyter_on_rails_as_sub_project'

Contributing

Bug reports and pull requests are welcome on GitHub at https://github.com/Yuki-Inoue/jupyter_on_rails.