Process/protocol editing failure with missing unit

[mikkonie]

Something I noticed in our production server timeline. A user has successfully edited a sample sheet, but them clicking on "finish sheet editing" to return to browsing mode and saving a sheet version results in the following error:

Expected 'Unit' not found in node 'p688-s0-a0-Protein Quantification-22-1' after/for attribute 'Comment(name='Protein quantification value', value='')'

The new sample sheet version is not saved. No previous edit actions have logged errors. As can be seen from the error log, this happens in a process.

The sample sheet does remain in a working condition in browsing mode. Alas, when I open it in edit mode I see this for the assay in question:

This is very strange. Needless to say, each process should only have one protocol. In the browsing mode, they show up correctly.

This also only happens in one of the two assays in the study: the other one appears fine.

I'll have to try and reproduce this in dev. We should see if there's a bug related to units here and/or something wrong with the sample sheet itself.

[mikkonie]

This seems to be related to #1598, thanks to @sellth for bringing it to my attention.

My guess is that something uncaught goes wrong when doing the actual cell edit, resulting in a failure of both exporting and reading the data for edit mode. It looks fine in browsing mode, but I guess that's due to cached data being re-read. I'll look into it next week.

[mikkonie]

Postponing this to v1.0, as it's a rare case and the person who'll I have to debug it is currently on vacation ;) May be re-prioritized if this is raised by users.

[mikkonie]

I can reproduce this bug in dev by adding a row to the 1st assay and only filling up required fields.

However, I also noticed the sheet is broken. There is an unnamed Extract Name in the assay. Not sure if that is allowed in ISA-Tab or not, but I have never seen this in a sample sheet before and I'm pretty sure SODAR does not support that. File names can be blank, but not material names.

I have stored a debug log of this edit failing, which should help further in looking into the issue. I can't be posting it here of course, as it contains IDs not meant for the public.

However, before I proceed any further in debugging this, I need to be in touch with the people in charge of this project and sample sheets. We need to ensure the sheets are in fact valid to begin with.