Skip to content

Orange with Dask

Marko Toplak edited this page Oct 27, 2023 · 8 revisions

Orange has an experimental branch, dask, which includes a subset of widgets that support on-disk data. That means that data can be really big, even bigger than your RAM.

This is very experimental.

Binary packages

We built installer packages for Windows and MacOS. The packages include dask branches of orange3 and add-ons orange-spectroscopy and orange3-survival-analysis.

Download here: https://download.biolab.si/download/files/dask

Note that these packages are not signed. To open the application on MacOS you'll need, after moving the app to appropriate location, Control+click the app and then click "Open" in its context menu (more info). Windows will likely complain that the package is unsigned; please click "More info" and then you'll be able to continue.

Windows users will also have to install Microsoft Visual C++ Redistributable).

Installation from source

Set up the system

The above command assumes a properly set development environment with accessible compilers and library headers (the requirements are the same as for compiling Orange from the master branch). Initializing the environment is OS and distribution-specific. Browse this GitHub repository for more info.

Windows users will surely need to install Microsoft Visual C++ Redistributable).

Create a virtual environment

First, create a new Python environment. With conda, you can use:

conda create python=3.10 --yes --name orange3-dask
conda activate orange3-dask

Other ways of making Python environments, such as pyenv and virtualenv, work too.

Install from the dask branch

Then, in the activated environment, install Orange from the Dask branch:

pip install https://github.com/biolab/orange3/archive/refs/heads/dask.zip

The above installation command installs all dependencies but PyQt. To install PyQt5, use (in the activated conda environment):

Install PyQt

pip install PyQt5>=5.12,!=5.15.1 PyQtWebEngine>=5.12

Alternatively, for PyQt6 (required on ARM Macs without Rosetta):

pip install PyQt6

Add-ons (optional)

In the activated environment:

  • For Spectroscopy, run pip install https://github.com/Quasars/orange-spectroscopy/archive/refs/heads/dask.zip
  • For Survival Analysis, run pip install git+https://github.com/biolab/orange3-survival-analysis.git@dask#egg=orange3-survival-analysis

Running Orange

To run the installed experimental Orange version, try:

conda activate orange3-dask  # activate the environment
python -m Orange.canvas