Closed michellewang closed 2 weeks ago
All modified and coverable lines are covered by tests :white_check_mark:
Project coverage is 98.64%. Comparing base (
5f107e0
) to head (51051bb
).
:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.
@nikhil153 what should we populate output_spaces
with?
@nikhil153 I'm thinking of renaming the descriptors:
dcm2bids-3.1.0-convert
-> dcm2bids-3.1.0.json
dcm2bids-3.1.0-prepare
-> dcm2bids_helper-3.1.0.json
Mostly because I'm looking at BIDSCOIN now and it has the bidsmapper
, bidseditor
and bidscoiner
CLIs, and I'm wondering if it would make more sense to just name the descriptors based on the actual command name.
However, I still think the invocations should probably still have the step name in them, so it might be confusing to have e.g. the descriptor called dcm2bids_helper-3.1.0.json
and the invocation called dcm2bids-3.1.0-prepare.json
?
Do you have any thoughts/preference?
@nikhil153 what should we populate
output_spaces
with?
The default one, which i think is this: MNI152NLin2009cAsym
. The reason we should explicitly add this into invocation is better visibility. Often people forget what space / resolution they use or how to change it before launching large number of jobs. This would help, hopefully!
@nikhil153 I'm thinking of renaming the descriptors:
dcm2bids-3.1.0-convert
->dcm2bids-3.1.0.json
dcm2bids-3.1.0-prepare
->dcm2bids_helper-3.1.0.json
Mostly because I'm looking at BIDSCOIN now and it has the
bidsmapper
,bidseditor
andbidscoiner
CLIs, and I'm wondering if it would make more sense to just name the descriptors based on the actual command name.However, I still think the invocations should probably still have the step name in them, so it might be confusing to have e.g. the descriptor called
dcm2bids_helper-3.1.0.json
and the invocation calleddcm2bids-3.1.0-prepare.json
?Do you have any thoughts/preference?
I think mapping descriptors 1-1 on the tools CLI makes sense. It also detaches them from nipoppy terms which is consistent with what we discussed last time.
As for the invocations, I think it's fine to name them differently since now we are in nipoppy scope where we try to abstract out / standardize run calls. From user perspective neither the passengers
nor crew members
will ever see descriptors, so it should be fine! :-)
The default one, which i think is this: MNI152NLin2009cAsym.
Okay, I'll add that. In the tracker I have *
instead of the specific space
entities -- I assume that is still fine? This way even if users change the output spaces then the tracker would still work out-of-the-box.
Summary:
fmriprep_tracker.py
andfs_tracker.py
files