This project shows how to package data files within a Python distribution, and has some example code for reading the data files. To build this distribution, create a venv with build installed and then execute

python -m build

The distributions (an sdist .tar.gz and a bdist .whl) will be written to ./dist/ subdirectory. To test it out, install the distribution and run the console script resources-example.

Here's a compatibility summary of the five approaches demonstrated:

Module	Description	In stdlib?	Works on Py2?	Works on Py3?	Works in zipfile?	Run as script? [*]
example1.py	os.path.join	yes	yes	yes	no	yes
example2.py	pkgutil	yes	yes	yes	yes	no
example3.py	pkg_resources	no	yes	deprecated	yes	yes
example4.py	importlib.resources. read_binary/read_text	deprecated	no	yes (3.7+)	yes	yes
example5.py	importlib.resources. files	yes (3.9+)	yes [†]	yes	yes	yes

If you are interested in creating an executable zip from source, you can use stdlib zipapp utility (Python 3.5+):

python3 -m zipapp --compress /path/to/resources-example --main="myapp:main" --output=myapp.zip

If this command is slow or the .zip is surprisingly large, make sure don't have any stray subdirs in the source path beforehand (e.g. .venv, .git, .idea).

Now you can run the zip directly with the interpreter (any Python version):

python myapp.zip

[*]	"Run as script" means executing the submodule directly, e.g. python myapp/example2.py. Note that Guido considers this an anti-pattern.

[†]	The same APIs are available in 2.7 by using an importlib_resources backport.