[Bug] Issue with transcripts patches

[lguerard]

Hi again and sorry to add more 😅

Description

During the patches creation for Baysor, some CSV files have issues and are incorrectly parsed resulting in errors later in the pipeline.
When opening one of the patches CSV, a couple of lines show

I tried seeing how to access the raw data to understand the issue, but there isn't enough information to find it back.

Reproducing the issue

Unfortunately, this is happening only on some patches and no reproducibly. Running twice the patches creation ended up with the same issue but on different patches :/

Expected behavior

All transcripts should have all the columns in all patches.

System

• OS: Windows
• Python version 3.10

I got this issue with both the sopa2 branch and the main one.

[quentinblampey]

Hi @lguerard, it seems to be the same issue as #123, can you confirm it?
I'm working on this already, but I couldn't reproduce it yet...

[lguerard]

Ah yes, that seems to be simiar, something weird happening when creating the transcripts.csv for the patches..

[quentinblampey]

Ok, then I'm closing this issue, and I'll continue on the other issue.

This is definitely a highly important issue, so I'll try to fix it quickly!
I never experienced this issue though, and I wasn't able to reproduce it. I suspect it may be a Windows-specific issue.

[lguerard]

Ah maybe, I haven't tried on Linux at all. But even on Windows, it's quite difficult as it seems to be fully randomized unfortunately :/