Python Jupyter中跨子包相对导入超出顶级包报错求助

First, let's break down why this error pops up: When you run from A.foo import Foo directly in your notebook.ipynb, Python treats the A directory as a top-level package instead of recognizing src as the actual root of your project. That's why the relative import from ..B.bar import Bar in foo.py fails—it's trying to jump outside what Python thinks is the top-level package.

Here are three clean, maintainable solutions to fix this without messy sys.path hacks:

1. Launch Jupyter from the parent directory of src #

This is the quickest fix for small projects:

• Navigate to the folder that contains your src directory (let's call this project_root).
• Run jupyter notebook from this project_root folder.
• In your notebook, update the import line to:

  from src.A.foo import Foo
  

Now Python recognizes src as the top-level package, so the relative import from ..B.bar import Bar in foo.py will work perfectly—.. correctly points back to src, allowing access to the B subpackage.

2. Set the Python path in your notebook (cleaner than manual sys.path edits) #

If you don't want to change where you launch Jupyter, use IPython magic commands to adjust the module search path directly in your notebook:

• At the very top of your notebook, add:

  # Change to the parent directory of src (replace with your actual path)
  %cd /path/to/your/project_root
  # Or set PYTHONPATH to include project_root
  %env PYTHONPATH=/path/to/your/project_root
  

Then you can keep using from A.foo import Foo in the notebook, and the relative import in foo.py will resolve correctly because Python now looks for packages starting from project_root.

3. Install your project as an editable package (best for long-term projects) #

This is the most professional approach, especially if your project will grow:

1. In your project_root directory, create a pyproject.toml file with this minimal content:

   [build-system]
   requires = ["setuptools>=61.0"]
   build-backend = "setuptools.build_meta"
   
   [project]
   name = "your-project-name"
   version = "0.1.0"
   packages = ["src"]
   

2. From project_root, run:

   pip install -e .
   

This installs your src package in "editable" mode, meaning any changes you make to your code are immediately reflected without re-installing.
3. Now, from anywhere (including your notebook), you can import with:

from src.A.foo import Foo

The relative import in foo.py will work flawlessly because src is now a properly recognized top-level package in Python's module system.

Pick the solution that fits your workflow: use option 1 for quick tests, option 2 for temporary debugging, and option 3 for projects you plan to maintain long-term.

内容的提问来源于stack exchange，提问作者user9569574