Skip to content

IPython magics.md

Latest commit

 

History

History
122 lines (80 loc) · 4.07 KB

IPython magics.md

File metadata and controls

122 lines (80 loc) · 4.07 KB

IPython magics: %reload_ext autoreload and %autoreload 2

%reload_ext autoreload and %autoreload 2 — usage, explanation, use cases, and related magics

Usage (copy into the top cell of a Jupyter notebook)

%reload_ext autoreload
%autoreload 2

What these commands do (line by line)

  • %reload_ext autoreload

    • Loads or reloads the IPython extension named autoreload into the current IPython/Jupyter session.
    • This makes the autoreload functionality available for use in that session.

  • %autoreload 2

    • Configures the autoreload extension to mode 2.

    • Mode meanings:

      • 0: disable autoreload (default).
      • 1: reload modules that are explicitly imported using %aimport.
      • 2: reload all modules (except certain builtins/C-extension modules) automatically before executing each cell.

Why use them together

  • %reload_ext autoreload ensures the extension is loaded into the kernel.
  • %autoreload 2 activates automatic reloading in the broadest mode (reload all modules).
  • Together they let you edit Python source files on disk (for example, modules inside your project src/) and immediately run notebook cells that import those modules without restarting the kernel or manually re-importing/reloading modules.

Typical use cases

  • Active development of local modules while iterating in a notebook (e.g., src/data.py, src/models.py).
  • Quick experimentation where you change functions/classes in source files and want the notebook to see updates immediately.
  • Teaching or demos where you modify code then re-run cells to show changes.

Example (typical placement)

# Top cell of notebook
%reload_ext autoreload
%autoreload 2

import pandas as pd
from src.data import load_raw, split_and_save  # edit src/data.py and changes will be picked up

Important caveats and gotchas

  • autoreload tries to update module objects in-place; this can fail or behave unexpectedly for:

    • objects created before a reload (e.g., existing class instances keep their old method bindings).
    • C-extension modules or compiled libraries (they are not reloadable).
    • complex import patterns (reloading does not rewrite references held by other modules).

  • For class definition changes, objects created before the edit may still behave according to the old class; you may need to recreate instances.

  • For deterministic behavior in production code, prefer restarting the kernel or using explicit import/reload workflows.

  • %autoreload 2 reloads a lot; if you want more control, use %autoreload 1 + %aimport modulename for selected modules.

  • Not a substitute for proper unit tests or CI — it is a developer convenience.

Alternatives and related commands (useful magics)

  • %autoreload 1

    • Reload only modules explicitly registered with %aimport.

    • Example:

      %reload_ext autoreload
%autoreload 1
%aimport src.data

  • %aimport modulename

    • Registers (or removes) modules for autoreload in mode 1.

  • importlib.reload(module) (regular Python)

    • Explicitly reload a specific module from Python standard library.

    • Example:

      import importlib
import src.data
importlib.reload(src.data)

  • %load_ext <extension>

    • Load an IPython extension (alias of %reload_ext for first load).
    • Example: %load_ext autoreload

  • %matplotlib inline / %matplotlib widget

    • Configure matplotlib display in notebooks (inline creates static images; widget enables interactive backends).

  • %timeit

    • Quick micro-benchmarking of small snippets.
    • Example: %timeit my_function(x)

  • %run script.py

    • Run an external Python script in the current kernel (useful to execute scripts and keep variables in the session).

  • %who / %whos

    • Inspect variables currently defined in the session.

  • %debug

    • Enter post-mortem debugger after an exception.

Minimal printable summary (one-liner)

Put these two lines at the top of a Jupyter notebook to auto-reload edited Python modules while you iterate:

%reload_ext autoreload
%autoreload 2