How to generate an installable package from adapted source?

I set up a (virtual) dev-machine in order to have a clean environment for some bokeh adaptations.

The code is good enough for my urgent need and I would like to transfer the modified bokeh to my usual working machine with Ubuntu 18.04 and many Python packages installed via Ubuntu’s pip3.
I don’t want to wait for the next release and hope for a merge of my code.
I also don’t like the idea to set-up a dev-environment with conda on a machine where pip is managing all other packages and the OS depends on the Python install for OS-management.

The simplest way seems something along the lines of …

• pip3 uninstall bokeh
• pip3 install mybokeh

Is there a foreseen “good practice” to deploy interim solutions for work?
I couldn’t find a hint on the developers webpage.

Thanks in advance.

We don’t have any opinions on how downstream packages should handle their packaging, and this is essentially the same thing. TLDR; use whatever works best for you.

To whom it may concern,

what worked for me was to generate a pypi-wheel-package on the dev system by:

bokeh> python3 setup.py sdist bdist_wheel

this generates (among others) a bokeh_some_stuff.whl file in the dist/. subdirectory
(here ist was dist/bokeh-0+unknown-py3-none-any.whl).
Transfer this file to the data analysis machine and install via pip by

~/>pip uninstall bokeh
to remove the original release and

~/>pip install /path/where/tranfered/to/bokeh_some_stuff.whl
to install my modified version.

On Ubuntu systems with python 2.7 and python3 replace pip by pip3.

Now I can enjoy my patches during work, while waiting for the next official release of bokeh.