G1Joshi Jupyter
Jupyter is the de facto standard for interactive data science. v7 (2025) of the Notebook is built on JupyterLab components, offering a modern, extensible experience.
When to Use
- Exploratory Data Analysis (EDA): Plotting data inline (
matplotlib). - Education: Teaching code with markdown explanations.
- Prototyping: Testing snippets before moving to a script.
Core Concepts
Kernels
The computation engine (IPython, IJulia).
Cells
Code cells (executed) vs Markdown cells (documentation).
Magic Commands
%timeit, !pip install.
Best Practices (2025)
Do:
- Use JupyterLab: The richer, multi-tab interface is standard.
- Use
nbdev: If you want to build libraries from notebooks. - Use Version Control: Use
jupytextto pair notebooks with.pyfiles for git diffs.
Don't:
- Don't store secrets: Clear output before committing.