Dependency issues with Scikit-bio

[florian-huber]

I started working on new workflow to check if the memo can be build/installed as pip package.

I ran into issues with scikit-bio which only worked for when installing it via conda.
When installing it with the other dependencies via pip I get an error that looks like this one: scikit-bio/scikit-bio#1671.

Another problem with scikit-bio is that it supports Python up to 3.7, but memo requires Python >= 3.8.
(so far that didn't seem to create bigger issues though).

I started working on it in #4 .

[florian-huber]

In #4 I can deal with the issue if I first pip install numpy before installing memo (with scikit-bio as dependency).
I guess that is OK for now. We would later have to check if this causes trouble when making memo a pypi package (see #8 ).

[oolonek]

Since its required only for the plotting of memo outputs we could:

• get rid of the plotting part completely
• separate it from the memo package

To be discussed with @ArnaudGaudry

[ArnaudGaudry]

I am open to discussion about it, but I still think memo would be more attractive if people can have a quick look at their data without having to install any other package or define a plotting function. It doesn't cover all the possibilities of downstream analyses of course, but gives easily a first idea of what's possible.
Otherwise if this causes too many installation issues, we can still cut it and people have a look at the publication scripts to see how the plots were generated for the article.

[florian-huber]

I agree, it would be nicer to have all in one package. So we should try that first.
But it can still be a good backup-option!

Even in such a case, the visualization part does not have to be removed completely. I would make it optional which with pip means that all code will be installed, but the visualisation functions will only work if people have scikit-bio installed as well.