Closed cbp44 closed 1 year ago
@johanneskoester Can you please approve the workflow so the test can run? It is strange that the tests work on my fork, but not here, and I would like to help fix the GitHub Actions for you.
I have a lot of additional bug fixes and enhancements that I would like to try and get merged here too, but that can't happen without GitHub Actions working first.
Hey @cbp44! Thanks for your contribution here and sorry for not picking this up earlier. Feel free to ping us about such things on the Discord server for snakemake: https://discord.gg/NUdMtmr
Regarding your suggestions here, I integrated the updates to the GitHub Workflow with PR #59, that also updates all wrappers and all conda environments. As this PR comes from your master branch, I would have to modify that to rebase this PR onto the latest master branch here, but didn't want to mess with your stuff.
But as we have passing CI tests again, would you be willing to rebase this fix (and the others you were mentioning) onto the latest master branch?
@dlaehnemann Can do.
I just rebased the master branch onto my fix/count-matrix-groupby branch containing only the count-matrix.py
change that I made.
I opened a new pull request to get that branch merged in (PR #60).
Fix for issue #32.
An analysis can fail if the
count-matrix.py
script inconsistently sorts output columns. The default behavior of thepandas.matrix.groupby
function is to sort the grouping keys, but the sorting is not stable and causes the workflow to fail sometimes in thedeseq2_init
rule of the analysis if the sorting happened to change the order.I ran into this bug a while ago and it prevented analyses where I had dozens of units in the
units.tsv
file that spanned sequencing lanes.