Closed effigies closed 4 years ago
@bpinsard There's a question of whether this should be handled in 20.2.1 or 20.3.0. It will change the derivatives, but I think the current derivatives are in error. Therefore I think it would be acceptable to fix in the LTS series.
An uglier but more conservative option would be to could keep the .dtseries.json
files with their current data, and add .json
files with valid BIDS keys.
I agree - I think this is a 20.2.1 issue. I'm inclined to go the conservative route, since any breaking derivatives change should be avoided in the LTS. @bpinsard WDYT?
The question is whether the current outputs are in error. We permit changes to derivatives that remove or modify outputs that are in error.
If we do want to use the conservative route, I think the strategy needs to be to declare a frozen set of keys that will continue to go into .dtseries.json
, and then all others go into .json
.
@mgxd Let's go with the conservative route for now. If it's acceptable to remove the "invalid" JSON files now, it's acceptable to remove them later.
fMRIPrep produces outputs named
sub-19_task-MerlinMovie_space-fsLR_den-91k_bold.dtseries.json
to matchsub-19_task-MerlinMovie_space-fsLR_den-91k_bold.dtseries.nii
.In the spec, the example given is:
PyBIDS also looks for sidecars with the extension
.json
, not.dtseries.json
.A first pass at this was taken in https://github.com/nipreps/niworkflows/pull/558 but dropped.
There was concern about not having sufficient disambiguation, so if we still can see that, then we need to make an issue on the spec. For my part, I would prefer not to change the spec, but to ensure that we always provide enough entities to ensure that no ambiguity is possible.
What version of fMRIPrep are you using?
20.2.0