Mismatch in -ctab option between json and tsv files

[cjl2007]

Hi - In previous versions of tedana, the input for --ctab was a tab delimitated text file (formatted just like the then output file, "comp_table_ica.txt"). If a user wanted to manually accept/reject components, they could edit the classification fields, and then rerun tedana, specifying the edited .txt file as the input for "--ctab"

In this latest version of tedana, however, the tab delimitated file output I am used to seeing has been replaced by a .json file called "ica_decomposition.json".

I assumed that I could edit the classification fields in this .json file, just as I had done with the .txt files, and rerun tedana using the edited .json file as the input for the --ctab option. This was not working.

I looked at the code briefly and it looked like the same bit of code that was used to load the classification table in earlier version of tedana, is still being used now ...

line 513 in tedana.py
comptable = pd.read_csv(ctab, sep='\t', index_col='component')

Which prompted me to try using a tab delimitated file instead, and this seems to have done the trick.

My question is whether this is expected behavior? It is not clear to me if the code just had not yet been updated to handle the new .json based component classification files, or if the user is expected to understand that a tab delimitated text file in a very specific format is expected. If the latter, then I think the documentation should be updated a bit to provide more information for users, because this wasn't obvious to me at first. Right now it just says,

--ctab = "File containing a component table from which to extract pre-computed classifications."

Thanks!
Chuck

[handwerkerd]

This does look like a bug to me. @tsalo worked to convert the comptable to a BIDS compatible JSON file, which will have several longer-term advantages. As of now, it looks like the the comptable is saved in JSON format with the io.save_comptable function, but the load_comptable function isn't ever used to load the comptable. The acute problem might be addressed by swapping pd.read_csv with io.load_comptable on line 513 of tedana.py, but I'll defer to @tsalo.

[tsalo]

#508 handles this.

[cjl2007]

Great! Thanks for point that this is raised already in #508, sorry for the duplicate.

[tsalo]

No, thank you for opening the issue. I should have opened one to log the bug before fixing it in the linked PR.