timdiels / varbio

Internal library shared by CoExpNetViz and MORPH

varbio (various bioinformatics) is an internal library with common code of CoExpNetViz and MORPH.

Links

• Conda
• GitHub
• Old PyPI, for older versions of the library.

Usage

Add varbio as a dependency to your conda recipe.

Development guide

Guide for developing varbio itself. It's analog to pytil's dev guide.

You do not need to release pytil just to try it out in varbio, conda install pytil's dependencies and finally pip install -e . in your local pytil checkout. If for some reason you want to try out the pytil conda pkg instead, you could build the pytil pkg locall first and then conda install --use-local pytil.

varbio's Issues

Find an optimised vectorised pearson

In GitLab by @timdiels on Apr 3, 2019, 17:55

Find an optimised vectorised pearson in a different project. If switching to YAML and no longer needing our own vectorised pearson, we may be able to get rid of this library, which simplifies things.

Input musn't be finnicky

In GitLab by @timdiels on Apr 3, 2019, 16:32

Column names with " in the name, a row with fewer columns than the rest, strange line endings, trailing tabs, consecutive tabs, UTF16-LE BOM encoding, dos line endings. All these weird cases should either be supported or yield an error which tells the user exactly what's wrong and how to fix it. Plus it shouldn't look like a traceback, users don't read those because that's not how you show errors to users, that's just for devs.

So:

• Handle various encodings
• Handle strange line endings
• Whitespace is insignificant, i.e. it can't break things
• User friendly error messages for common input errors

Remove mutual information

In GitLab by @timdiels on Apr 3, 2019, 17:51

I removed mutual inf from CoExpNetViz which erroneously used it as a correlation statistic; I do not have time to change CoExpNetViz to use mutual inf properly. This means none of our projects use it, so remove it and simplify by removing genericness too (generic df).