Duplicate rows created on ISA-Tab export when splitting with unnamed data files

[mikkonie]

While fixing #1904, I noticed a problem with rendering tables for pooled processes. When pooling and then splitting with empty material names, duplicate rows are created on ISA-Tab export.

How to Reproduce

• Create sheets from e.g. the cytof template
• Add assay names to a column of processes
• Add new sample to study
• Insert new row with new sample, set process name to an existing assay name
• Leave data file names empty (not sure if this matters but just in case)
• Click Finish Editing and exit edit mode

What Should Happen

• Exporting should produce a valid ISA-Tab which also re-imports ok

What Actually Happens

• We end up with four rows containing the pooled/split data, with two rows repeated identically as the data file names are empty
• When exporting the ISA-Tab, the duplicate rows are present
• Re-importing this ISA-Tab fails in AltamISA validation because of the duplicated rows

Workaround

• Fill the data file names with something in the ISA-Tab prior to re-importing

Notes

• Data in Assay.arcs is fine, this happens on export
• If the data files are named in editing, export and import work fine
• Technically this is correct: files do not explicitly have to be named
• One weird thing I noticed is that in row import, we set `Derived Data File" in the unique name of the files
  • On import, AltamISA sets Empty Derived Data File
  • This only seems to affect unique_name, nothing else
  • If we supply the correct keyword, does that do something in AltamISA?
  • See TOKEN_EMPTY in AltamISA
  • Update: The token doesn't actually do anything in AltamISA, it just gets set to the unique_name.

[mikkonie]

I guess we should decide how to handle this. The options I can see off the top of my head:

• Prevent inserting rows with empty data file names in case of splitting
  • Not really correct from ISA-Tab standpoint
  • This would require properly implementing the renderting of split rows during editing (see #1927)
• Omit duplicate rows with empty data file names from ISA-Tab export
• Decline to export until data file names have been filled
• Accept this as a feature and instruct users to fill out data file names if they want to re-import
• TBD: something else?

To be discussed. Since it took me so long to discover this problem and no one has reported it yet, I'm inclined to think it's not super urgent to fix :)

Removing the milestone for now but leaving this open.