rsconnect-jupyter 1.2.2.4

Jupyter Notebook integration with RStudio Connect

rsconnect-jupyter

rsconnect-jupyter is a plugin for Jupyter Notebook that enables publishing notebooks to RStudio Connect.

Requirements

• Python 2.7.9 or Python 3.5.0 and higher
• Jupyter Notebook 5.x
• pip
• wheel
• RStudio Connect v1.7.0 or higher, configured with Python support.

If using conda, pip and wheel should already be installed.

Installation

If you are installing rsconnect-jupyter for use in Jupyterhub, please see the Jupyterhub section below.

We recommend working within a virtualenv. If you are unfamiliar, these commands create and activate a virtualenv at /my/path:

pip install virtualenv
virtualenv /my/path
source /my/path/bin/activate

Install Jupyter inside the virtualenv:

pip install jupyter

Note: be sure to run Jupyter from the virtual environment, not from a global installation.

Install the rsconnect-jupyter package with the following command:

pip install rsconnect_jupyter

Enable the rsconnect-jupyter extension with the following commands:

# Install `rsconnect-jupyter` as a jupyter extension
jupyter-nbextension install --sys-prefix --py rsconnect_jupyter

# Enable JavaScript extension
jupyter-nbextension enable --sys-prefix --py rsconnect_jupyter

# Enable Python extension
jupyter-serverextension enable --sys-prefix --py rsconnect_jupyter

Note: The above commands only need to be run once when installing rsconnect_jupyter.

Note: If you run into an issue during installation please let us know by filing a bug here.

Uninstalling

First disable and remove the rsconnect-jupyter notebook extension:

# Disable Python extensions found in `rsconnect-jupyter`
jupyter-serverextension disable --sys-prefix --py rsconnect_jupyter

# Remove JavaScript extension
jupyter-nbextension uninstall --sys-prefix --py rsconnect_jupyter

Finally, uninstall the rsconnect-jupyter python package:

pip uninstall rsconnect_jupyter

Upgrading

To upgrade rsconnect-jupyter, first uninstall the extension and then re-install it.

Usage

Open a notebook and click the blue "Publish to RStudio Connect" icon to publish the current notebook to RStudio Connect.

Entering server information

If this is your first time publishing a notebook, you will be prompted to enter the location and a nickname for the RStudio Connect server.

You will also be prompted to enter your API Key. See the RStudio Connect User Guide for instructions on generating API Keys for your user.

When you click the Add Server button, rsconnect-jupyter will send a request to the RStudio Connect server to verify that it can be reached via the requested URL and that the API key is valid.

If your RStudio Connect server was configured with a self-signed certificate (or other certificate that computer hosting your Jupyter notebook server does not trust), the attempt to contact RStudio Connect will fail with a TLS-related error. In this case, please contact your system administrator for instructions on how to configure your computer to trust the certificate they have provided. If trusting the certificate is not an option, you can check the Disable TLS Certificate Verification box, which will bypass the certificate check. Your connection to the server will still be encrypted, though the server identity will not be verified.

Publishing options

There are two different publication modes. Selecting "Publish finished document only" will publish an HTML snapshot of the notebook to RStudio Connect. HTML snapshots are static and cannot be scheduled or re-run on the RStudio Connect server.

If you select "Publish document with source code", the notebook file and a list of the Python packages installed in your environment will be sent to RStudio Connect. This enables RStudio Connect to recreate the environment and re-run the notebook at a later time.

Environment detection with pip

The list of packages sent along with the notebook comes from the python environment where the notebook kernel is running. In order for environment inspection to work, the rsconnect-jupyter package must be installed in the kernel environment; that is, the environment where the ipykernel package is installed. In most cases that will be the same as the notebook server environment where jupyter is installed.

The command pip freeze will be used to inspect the environment. The output of pip freeze lists all packages currently installed, as well as their versions, which enables RStudio Connect to recreate the same environment.

Handling conflicts

If content that matches your notebook's title is found on RStudio Connect, you may choose to overwrite the existing content or create new content.

Choosing "New location" will create a new document in RStudio Connect. You can choose either publication mode - an HTML snapshot or a document with source code.

Updating an existing document will not change its publication mode.

Upon successful publishing of the document a notification will be shown in toolbar. Clicking the notification will open the published document in the RStudio Connect server you selected in the previous dialog.

Collaboration

To collaborate with others add them as collaborators in RStudio Connect. During publishing they should provide their API key and will be able to choose a content location to publish to if the notebook title is the same.

You may share notebooks if appropriate.

Installation in JupyterHub

In JupyterHub, follow the directions above to install the rsconnect-jupyter package into the Python environment where the Jupyter notebook server and kernel are installed. Typically those will be the same environment. If you've configured separate kernel environments, install the rsconnect-jupyter package in the notebook server environment as well as each kernel environment.

The exact install location depends on your Jupyterhub configuration.

JupyterHub Example Configuration

This section presents a simple working example of a Jupyterhub configuration with rsconnect-jupyter installed.

This example uses Docker, but you can install the rsconnect-jupyter package in any Jupyterhub installation. Docker is not required.

Example Dockerfile:

FROM jupyterhub/jupyterhub:0.9.4

# Install Jupyter notebook into the existing base conda environment
RUN conda install notebook

# Download and install rsconnect-jupyter in the same environment
# Update this to specify the desired version of the rsconnect-jupyter package,
# or pass `--build-arg VERSION=...` to docker build.
ARG VERSION=RSCONNECT_VERSION
ARG REPOSITORY=https://s3.amazonaws.com/rstudio-rsconnect-jupyter

RUN wget ${REPOSITORY}/rsconnect_jupyter-${VERSION}-py2.py3-none-any.whl
RUN pip install rsconnect_jupyter-${VERSION}-py2.py3-none-any.whl && \
	jupyter-nbextension install --sys-prefix --py rsconnect_jupyter && \
	jupyter-nbextension enable --sys-prefix --py rsconnect_jupyter && \
	jupyter-serverextension enable --sys-prefix --py rsconnect_jupyter

# create test users
RUN useradd -m -s /bin/bash user1 && \
	useradd -m -s /bin/bash user2 && \
	useradd -m -s /bin/bash user3 && \
	bash -c 'echo -en "password\npassword" | passwd user1' && \
	bash -c 'echo -en "password\npassword" | passwd user2' && \
	bash -c 'echo -en "password\npassword" | passwd user3'

CMD ["jupyterhub"]

Run these commands to build and start the container:

docker build -t jupyterhub:rsconnect-jupyter .
docker run --rm -p 8000:8000 --name jupyterhub jupyterhub:rsconnect-jupyter

Connect to Jupyterhub on http://localhost:8000 and log in as one of the test users. From there, you can create a notebook and publish it to RStudio Connect. Note that the current Jupyterhub docker image uses Python 3.6.5, so you will need a compatible Python version installed on your RStudio Connect server.