rhngla / minpypack Goto Github PK

I have codes that are under development while also being used across multiple local projects that use custom environments (e.g. specified with conda).

• I don't want to make copies of the code
• appending paths through sys.path.append() is cumbersome, and will have to be cleaned up eventually if I'm sharing my code

Using pip to install codes as a package locally is a clean solution. It also promotes better practices for eventually releasing codes. This repository is a minimal package showing how to structure code, write the setup file and complete the installation with pip in any local environment:

The parent minpypack directory contains a subfolder with the same name that contains all codes to use in multiple environments. The setup.py file contains installation instructions used by pip. For reference, this is the flat-layout packaging style

#minpypack directory structure:
└── minpypack
    ├── README.md
    ├── minpypack
    │   ├── A
    │   │   ├── A_file.py
    │   │   └── __init__.py
    │   ├── B
    │   │   ├── B_file.py
    │   │   └── __init__.py
    │   ├── C.py
    │   └── __init__.py
    └── setup.py

Some terminology:

• A and B are (sub) packages (because they contain the __init__.py file).
• A_file.py, B_file.py, C.py are modules - i.e. .py files containing functions and classes.

⚠️ Only modules or the functions, classes inside of them can be imported (explanation).

#Valid ways of importing:
import SomeModule
from SomeModule import SomeFuncName, SomeClassName, etc.
from SomeModule import SomeFuncName as SomeOtherName
from SomeModule import *

Accordingly for our project:

🚫 This fails:

#Fails because B is not a module, but a package:
import minpypack.B as bb
bb.B_file.main() 

✅ These work:

#Works because B_file is a module with function main:
import minpypack.B.B_file as bb
bpack.main() 

#Works because C is a module with function BfromC
import minpypack.C as cc
cc.BfromC()

Navigate to where the setup.py file is located. Install minpypack into current environment:

pip install -e .

Installation performed in this manner includes symbolic links to the files.

• 🛠 Changes to any minpypack source files are immediately available in all environments it is installed.
• ⚠️ If minpypack folder is moved after installation, you will have to reinstall it.

pip also provides an easy way to install remote repositories, including private github repositories you have access to.

Below is an example to create a new conda environment and install minpypack as a package without explicitly cloning with git:

#Create a new conda environment with python 3.8
conda create -n demoenv
conda activate demoenv
conda install python==3.8

#Install minpypack into this environment directly from github:
pip install git+https://github.com/rhngla/minpypack