Toggle navigation

.rst .pdf

Specifying dependencies and publishing code

Missing dependencies, incompatible versions, or conflicting version requirements are an effective cause of “works on my computer but not on yours”.

At the first level, the idea is to document dependencies, enable a quick installation of those using pip install -r. The code itself is simply downloaded from GitHub, shared by email, … without being packaged further.

At the second level, we use Poetry which is a turn-key solution to manage dependencies, isolate virtual environments, publish package on PyPI.

At the third level, we’ll want to provide additional guarantees about the code robustness. I wonder how many academic projects will reach that stage.

Try out these tools in a sample project before publishing existing code!

(But please use TestPyPI to not pollute the Python pacakge namespace).