Skip to content

SETUP.md

Latest commit

 

History

History
115 lines (84 loc) · 3.07 KB

SETUP.md

File metadata and controls

115 lines (84 loc) · 3.07 KB
Raw

Set up the training environment with Pixi

This repository uses Pixi to install Python, JupyterLab, and the libraries used by the training material. The commands below are for a Linux terminal.

Install Pixi

The installer recommended by the Pixi project is:

curl -fsSL https://pixi.sh/install.sh | sh

This command downloads and immediately executes the installer. If you prefer to inspect a remote script before running it, use:

curl -fsSL https://pixi.sh/install.sh -o /tmp/pixi-install.sh
less /tmp/pixi-install.sh
sh /tmp/pixi-install.sh

The installer puts pixi in ~/.pixi/bin and updates the shell startup file so that the command is on PATH. Start a new terminal, then verify the installation:

pixi --version

See the official Pixi installation instructions for installer options and alternative installation methods.

Create an environment

Run Pixi commands from the top-level directory of this repository, where pixi.toml is located:

cd /path/to/Python-for-data-science

To create the default environment containing the complete current training stack, run:

pixi install

The initial solve and download can take some time, particularly because the default environment includes the geospatial libraries. Later invocations reuse the environment. The first install also creates or updates pixi.lock, which records the resolved package versions.

Create a smaller topic environment

If you only need one part of the training, install its environment explicitly. For example:

pixi install --environment pandas
pixi install --environment gis

The available topic environments are:

  • altair
  • db-access
  • duckdb
  • gis
  • holoviews
  • networkx
  • pandas
  • polars
  • regexes
  • seaborn
  • web-scraping
  • xarray

There is also a separate pandas-bokeh environment for the historical notebook under source-code/pandas/bokeh. It deliberately uses an older Python, NumPy, Pandas, and Bokeh stack and should not be used for the other notebooks.

You can create every environment with pixi install --all, but this duplicates many packages across separate environments and consumes substantially more disk space. Usually the default environment or one topic environment is sufficient.

Run a notebook

Start JupyterLab with the complete default environment:

pixi run lab

JupyterLab opens in source-code. Select a topic directory and open the desired .ipynb file. Run notebook cells with Shift+Enter.

To use a topic-specific environment, select it with --environment (or -e):

pixi run --environment pandas lab
pixi run -e gis lab

For the isolated legacy Bokeh example, run:

pixi run --environment pandas-bokeh lab

Stop JupyterLab by returning to the terminal and pressing Ctrl+C. Pixi does not require manual environment activation for these commands; pixi run runs the task inside the selected environment and creates that environment first if it has not yet been installed.