Interactive data displays with Jupyter notebooks - Example Notebooks

Ioanna Ioannou and Maoyuan Liu

Forecast Systems Team

Research and Development

Bureau of Meteorology

Installation

Requirements

Python

• pandas - loading data
• matplotlib, bokeh, plotly - plotting libraries
• ipywidgets - provides UI widgets
• jupyter, notebook - provides the notebook environment
• jupyter-dashboards, jupyter-cms - provides dashboards layout extensions
• jupyter-kernel-gateway - provides kernel gateway for dashboards

System

• node, npm - provides server and package manager for node.js applications

npm

• jupyter-dashboards-server, debug - dashboards server front-end

Notebook installation

Prefer installing packages using Miniconda. To download, go to http://conda.pydata.org/miniconda.html

# work in a new conda environment
conda create -n dashboard_example python=3
source activate dashboard_example

# install packages
conda install pandas matplotlib bokeh jupyter
# conda provides the big packages, use pip for the smaller ones that
# are not yet in conda
pip install --upgrade ipywidgets
pip install jupyter-dashboards jupyter-kernel-gateway plotly

# register notebook extensions
jupyter dashboards quick-setup --py --sys-prefix
jupyter cms quick-setup --py --sys-prefix
jupyter nbextension install --py widgetsnbextension --sys-prefix
jupyter nbextension enable --py widgetsnbextension --sys-prefix
jupyter nbextension enable --py jupyter_cms --sys-prefix
jupyter nbextension enable --py jupyter_dashboards --sys-prefix

Dashboard front-end installation

If you don't have the executable npm and node in your PATH, you may need to install it using your system package manager. The system package that provides node and npm is known under different names for different linux distributions. Check the documentation for your OS.

# install jupyter-dashboards-server
npm install jupyter-dashboards-server
npm install debug
# add `node_modules/.bin` to your PATH

Running the notebook and dashboards server

# go to the directory which contains the notebooks

# run the kernel gateway in the background using default settings
jupyter-kernelgateway &
# listening on port 8888

# run the notebook server
juputer notebook &
# listening on port 8889

# run the dashboard server
jupyter-dashboards-server --NOTEBOOKS_DIR=`pwd` --KERNEL_GATEWAY_URL=http://127.0.0.1:8888 &
# listening on port 3000

Notebooks: http://localhost:8889

Dashboards: http://localhost:3000