The Datashader project is focused on building better ways to visualize very large datasets, using intelligent server-side downsampling, automatic computations performed on the data as it progresses through the visualization pipeline, and other related techniques. The project is under active development, and all the code and documentation is subject to frequent changes.

# Create a new conda environment, if desired
conda create -n datashader python=2.7
source activate datashader

# Install required packages, including latest fixes required
conda install numpy pandas xarray toolz numba datashape odo dask pillow
pip install --upgrade --no-deps git+https://github.com/Blaze/odo
pip install --upgrade --no-deps git+https://github.com/Blaze/datashape

# Install Bokeh for running examples
conda install -c https://conda.anaconda.org/bokeh/channel/dev bokeh

# Install the datashader library
git clone https://github.com/bokeh/datashader.git
cd datashader
python setup.py develop

Running the examples

cd examples

Download the sample data. This may take 20 minutes on a good connection, and more otherwise:

python download_sample_data.py

Dashboard example:

python dashboard/dashboard.py --config dashboard/nyc_taxi.yml 

(which should launch a browser tab pointing to the appropriate URL, which is localhost:5000 by default.)

Jupyter notebook examples:

jupyter notebook

(and then select nyc_taxi.ipynb or nyc_taxi-nongeo.ipynb from within the Jupyter tab in your browser, and select Cell/Run all to create interactive plots.)