Jupyter Notebook Cheat Sheet
Jupyter Notebook reference covering magic commands, shell integration, cell types, and essential keyboard shortcuts for efficient interactive computing.
1 PageBeginnerMar 25, 2026
Common Magic Commands
Line and cell magics for the IPython kernel.
python
%matplotlib inline # render plots inline%timeit sum(range(1000)) # time a single line%%time # time a whole cell (must be first line)%who # list variables in the current namespace%load_ext autoreload%autoreload 2 # auto-reload modules on change!pip install requests # run a shell command
Cell Magics & Shell
Write files and run shell scripts from a cell.
python
%%writefile script.pyprint("hello from a generated file")%%bashecho "runs as a bash script"ls -la%run script.py # execute an external .py file%pwd # print working directory
Keyboard Shortcuts
Command mode shortcuts (press Esc first).
- Esc, A- insert a new cell above
- Esc, B- insert a new cell below
- Esc, D D- delete the selected cell
- Shift+Enter- run cell and select the next one
- Ctrl+Enter- run the cell in place
- Esc, M- convert cell to Markdown
- Esc, Y- convert cell to code
Cell & Kernel Concepts
Building blocks of a notebook document.
- Code cell- executes in the kernel and shows output below
- Markdown cell- renders formatted text, LaTeX, and images
- Raw cell- passed through unmodified during nbconvert (no execution)
- Kernel- the language runtime (ipykernel for Python) executing your code
- Checkpoint- autosaved notebook snapshot used for recovery
Interactive Widgets & Rich Display
Build interactive UIs with ipywidgets and customize how objects render via the display protocol.
python
import ipywidgets as widgetsfrom IPython.display import display, HTML@widgets.interact(x=(0, 10), label=["a", "b", "c"])def show(x=5, label="a"): print(x, label)# Manual widget wiringslider = widgets.FloatSlider(min=0, max=1, step=0.05, description="threshold")output = widgets.Output()def on_change(change): with output: output.clear_output() print(f"threshold = {change['new']}")slider.observe(on_change, names="value")display(slider, output)# Custom rich repr for your own classesclass Report: def _repr_html_(self): return "<b>Custom HTML rendering</b>"
Headless Execution: nbconvert & Papermill
Run notebooks as parameterized scripts from the CLI or a pipeline, without opening the UI.
bash
# Execute a notebook end-to-end and overwrite it with outputsjupyter nbconvert --to notebook --execute --inplace analysis.ipynb# Export to a script or HTML reportjupyter nbconvert --to script analysis.ipynbjupyter nbconvert --to html --no-input analysis.ipynb # hide code cells# Papermill: inject parameters into a tagged cell and produce a new output notebookpip install papermillpapermill template.ipynb output.ipynb -p threshold 0.8 -p dataset "prod"
Debugging & Profiling Magics
Drop into the debugger on exception and profile slow cells without leaving the notebook.
python
%pdb on # auto-launch pdb whenever a cell raises%debug # post-mortem debug the last exception%prun some_function(df) # cProfile a single call, ranked by cumulative time%%prun -s cumulativeresult = expensive_pipeline(df)%load_ext memory_profiler%memit train(model, X, y) # peak memory used by a call%env MY_VAR=value # inspect/set environment variables for the kernel
Kernel Config & Notebook Metadata
Register alternate kernels and inspect the JSON structure behind a .ipynb file.
bash
# Register a virtualenv/conda env as a selectable kernelpython -m ipykernel install --user --name myenv --display-name "Python (myenv)"jupyter kernelspec listjupyter kernelspec remove myenv# A .ipynb is JSON: cells[], each with cell_type, source, outputs, metadata# Cell tags (Edit > Cell Tags in UI) drive tools like papermill and nbconvert# templates, e.g. tagging a cell "parameters" or "remove-cell"
Ecosystem Tools & Extensions
Common tooling that extends notebooks beyond the base kernel loop.
- JupyterLab- the successor UI to classic Notebook, with a file browser, terminals, and extension manager built in
- nbdime- structural diff/merge for notebooks so git diffs show cell-level changes instead of raw JSON noise
- jupytext- pairs a .ipynb with a plain .py/.md file for cleaner version control and code review
- voila- turns a notebook into a standalone, code-hidden interactive web app
- nbval / pytest-notebook- runs notebooks as regression tests, asserting cell outputs match expectations
- JupyterHub- multi-user server that spawns a notebook instance per authenticated user
- ipyparallel- distributes cell execution across a cluster of IPython engines
Pro Tip
Restart the kernel and run all cells top-to-bottom (Kernel > Restart & Run All) before trusting a notebook's results — out-of-order execution during interactive editing easily leaves you with state that won't reproduce from a fresh run.
Was this cheat sheet helpful?
Explore Topics
#JupyterNotebook#JupyterNotebookCheatSheet#DataScience#Beginner#CommonMagicCommands#CellMagicsShell#KeyboardShortcuts#CellKernelConcepts#MachineLearning#CommandLine#CheatSheet#SkillVeris
