Python Installation, Setting, Project Packaging

Installation and PATH Variable Setting

After installing python you need to set PATH variable to let your system know where your python executable resides. Usually you don’t need to do this manually if you checked Add Python to PATH checkbox when you installed python.

However, sometimes you might want to maintain multiple versions of pythons or simply forget about the checkbox so need manual setting. What should you do then?

Install Python

I installed Python 3.10.

After installation you can check it was successful by running python --version command in your command prompt. If you checked Add Python to PATH during installaton this must work in most cases, as it makes your system know the what it should do when you type python in the command prompt.

Check PATH Variables

Now let’s check how my PATH variables look like. In my case I was able to access [Control Panel]-[System Properties] to check this.

Now click [Environment Variables…] button, and from bottom panel of [System Variables] activate Path and click Edit button.

You can see above kinds of paths are registered as Path variable.

First of all, /Python310 directory contains python.exe executable file. So when you type python in the command prompt, what system actually do is finding this exe file and execute it. If you force to rename python.exe file, you can make it fail. Try for fun!

On the other hands, in /Python310/Scripts you can find pip.exe. pip is a package management system of Python and pip.exe is executable file for this package. So after installing python you can use command like pip install to install packages you need with assistance of pip.

You can check version of your installed pip also in the same manner of checking python version.

setuptools for Project Packaging

If you have worked on your own python packages and got some results, your next natural eager is packaging it and distribute for many different users or inside your team.

Example Usage

Assume I have myproject counstructed as follows. I want other people to be able to install my ambitious coffemaster, by their running pip install coffeemaster or so, rather than sending my zipped source code via e-mail and asking them to unzip it and locate at proper location in there PC (as it is dumb). Furthermore, I am also tired of being asked from my team about what dependency libraries they have to install in their local to avoid stupid ModuleNotFoundError or so.

You can use setuptools package in this case. It helps you to pack your codes into single binary file installable via pip and to specify all the dependency packages so that they can be automatically installed when your custom package is installed.

myproject
├─ README.md
├─ setup.py
└─ coffeemaster
   ├─ __init__.py
   ├─ __main__.py
   ├─ drinktypes
   │  ├─ __init__.py
   │  ├─ americano.py
   │  └─ latte.py
   └─ coffeebeans
      ├─ __init__.py
      ├─ arabica.py
      └─ robusta.py

setup.py is the most important file. See below code.

from setuptools import setup, find_packages

setup(
    name='coffeemaster',
    version='0.0.1',
    packages=find_packages(),
    install_requires = [
	    'torch',
	    'pandas>=1.4.2'
    ],
    package_data={'examplecoffee': ['tallSizeAmericano.txt']}
)

In this file, I speficied my package name (name) and version (version). Also I specified all the dependency packages under install_requires item. If I want to distribute files other than python codes (.py files), specify them as package_data item.

Then I can move to the directory where setup.py resides and rum python setup.py sdist to build distributable binary file for my coffeemaster package. coffeemaster.0.0.1.tar.gz file results under new directory myproject/dist.

I can attach this .tar.gz file when I release new version from my repository or just send this file to teammates. Anyone downloading/receiving this file can run pip install <path of tar.gz file> in their environment to install coffeemaster package.

If I run python setup.py bdist_wheel, .whl file results. It seems that they say .whl format is better than tarball(.tar.gz) format.

But in my case .whl format didn’t work better than tarball. What I expected was to install coffeemaster only in the target environment, but when using .whl formatted binary file three different packages of coffeemaster, drinktypes, and coffeebeans were installed altogether. Currently I am only using sdist command therefore. This phenomenon seems not to happen but depending on the environment. I am not sure up until now why such case occur.

Updates from setuptools 58.3.0.

It is said that from setuptools version 58.3.0. use of python setup.py sdist and python setup.py bdist_wheel was deprecated.

Above two build commands are replaced by single command below:

python -m build

To run this, you may need to install build to your local environment first:

pip install build

When you run above command, .tar and whl files both output as results. If you need only one of them, you can add additional argument --sdist or --wheel.

Providing Standalone Executable - pyinstaller

Oftentimes you may need to distribute your python projects as a stand-alone executable. Such a case actually happen for example when you elaborated your solution but the potential users are not SW-acquainted engineers. pyinstaller is a decent option in such cases.

First of all, we need to install pyinstaller

pip install pyinstaller

The simplist way to build exe file using pyinstaller is to move to the source directory where the .py file you want to convert into exe and to execute below command,

pyinstaller example.py

given the file I want to convert into exe is example.py.

You can use more detailed specifications by writing down your custom .spec file. An example .spec file is formatted as follows.

a = Analysis(['minimal.py'],
         pathex=['/Developer/PItests/minimal'],
         binaries=None,
         datas=None,
         hiddenimports=[],
         hookspath=None,
         runtime_hooks=None,
         excludes=None)
pyz = PYZ(a.pure)
exe = EXE(pyz,... )
coll = COLLECT(...)

Dependency to Visual Studio Pyinstaller-generated executables need Visual Studio 2015 or later to be run! If you get complaints from your program’s users that your executable is not working on their PC, make sure they have proper VS intalled on their PC. This might settle the issue in most cases.