What do you do if you need to write a Python package and use it, unchanged and unmaintained, for the next ten years or so? Let's assume the user can do a bit of work on their side (e.g. use pip on virtualevs, maybe install Anaconda or something like that) but not install anything from sources or use Docker, etc. -- think lab & workshop folk with some generic IT support. I think I can crudely proof my packages against changes in their dependencies by specifying versions in my "requirements" files (although there is an infinity of install schemes these days and this may be out of date already). But how do you protect yourself against changes in the package installation scheme and the language itself? Any links or tutorials? I cannot afford to maintain a host of packages and need some way to freeze them and have them work in the distant future. Can you do this with Python at all these days? If this is not practically possible in Python, which other scripting language would you use that has something similar to Numpy (Linux platforms only)? Continue reading...
