When you install Git on your computer, it can be run from the terminal with "git". Similarly, Python can be run with just "python" or "py".
If I make a CLI with argparse and Python, is there a way to have users simply install it and use it in the terminal just by typing
$ myCLI -option
Rather than
$ python path\to\myCLI.py -option
Like all the CLI's I use/interact with do?
Your script just needs to be somewhere in the user's $PATH (and to have no extension if you don't want them to type myCLI.py). Additionally it needs to start with a magic comment #!/usr/bin/env python3 (If you are running Linux) and be executable (on linux chmod +x myCLI prior to distributing it). Alternatively you create a script in the native shell language of the desired OS (so sh for linux) which calls python3 /path/to/script or python -m module or the like.
If you you name your main script without a .py it prevents you importing it from anything else, so the usual approach is to have an 'entrypoint' script which consists of something like:
#!/usr/bin/env python3
from myApp import main
main()
(see the PyPa docs for a discussion)
Of course, you need to deploy your entrypoint script somewhere in the end-user's $PATH. This can be more intersting, but luckily the whole thing is handled by setuptools, and you don't even need to write a script: you just need an entry point function:
Use console_script entry points to register your script interfaces. You can then let the toolchain handle the work of turning these interfaces into actual scripts
you just put in your setup.py:
setup(
entry_points = {
'console_scripts': [
'PKG_NAME = package.module:func',
],
}
)
and anyone who types PKG_NAME will effectively run package.module:func() (so func is your entrypoint main() function above).
See this question for a discussion of the merits of this approach over the old approach of writing separate scripts for each platform.
Note that if you're not building your app as a python package, but are using some other packaging standard (archlinux PKGBUILD or .deb or whatever) you can still take the same approach---just write the entry point script as above and have the package place it somewhere in the user's $PATH.
