基于打包（setup.py）的另一个文件的函数导入导致“找不到 .egg-info 目录”

Question

My goal is to access a function called train_test_split(...) from the file utils.py within dnn.py. My structure is the following:

├───MachineLearning
│   └───src
│       ├───deep_learning
|           └─── dnn.py
|           └─── __init__.py
│       └───utils
|           └─── preprocessing.py
|           └─── __init__.py
│   └───readme.md
│   └───setup.py

I don't want to use or change my windows path because it should run on other computers too that clone from my github. The solution I found is using a setup.py file. I spent a really long time trying to understand how it works and ended up with the structure above. My setup.py file contains this:

from setuptools import setup, find_packages

setup(
    name='example',
    version='0.1.0',
    packages=find_packages()
)

Now when I run pip install -e . from the the root folder (MachineLearning) I get the following error:

(tf38_clone) F:\Code\MachineLearning>pip install -e .          
Obtaining file:///F:/Code/MachineLearning
  Preparing metadata (setup.py) ... done
ERROR: No .egg-info directory found in C:\Users\Nick\AppData\Local\Temp\pip-pip-egg-info-pvq_am4y

Can anybody point me into the right direction? The only thing I want is to import my function into another .py file as described above.

Answer 1

If your

goal is to access a function called train_test_split(...) from the file utils.py within dnn.py

which you also indicated by writing

The only thing I want is to import my function into another .py file as described above.

then you should try from utils import train_test_split at the beginning your dnn.py to import the function from the other module. There is no need for creating a package for this use case. More information about modules in python can be found here .

However, if you want to install an editable package, you need to build one first (ie, python -m build ), which will create *.egg-info in your stacktrace. You can read more about packaging here .