Updating jupyter install with python3

[josueneo]

While working for Singularity 101 course, I came across these examples and found out the versions are very old, this is an update.

I do not understand why using conda if a pip install is enough to install it, maybe I am missing something. However this is an urgent update since the previous jupyter install was using ubuntu 16 and python 2.7.

[dtrudg]

Please try and keep to a conda install if you can.

The reason to use conda here is that the jupyter example is just a starting point. Users will then want to add other python packages they need to perform their analysis inside jupyter.

It is not easy, or sometimes even possible, to install up to date ML, bioinformatics, data science python packages with a distribution's pip. A lot of ML / bioinformatics / data science stuff depends on some lower level C libraries or similar. Conda supplies these non-python dependencies easily, and is probably the most common tool used by researchers and data scientists for this purpose.

[dtrudg]

Also, as a general rule, when creating containers using python.... either...

• Use the distributions own python packages, installing them with the package manage. E.g. apt install jupyter.

-or-

• Use conda

-or-

• Use pip inside a virtual environment, e.g. via venv

When you use pip as root, you can end up installing packages in system locations that can conflict or update over distribution python- packages. Thos distribution python packages may be dependencies of the distros own system admin tools etc. This can cause confusing issues for users who try to take the example as a starting point.