[RTM] FIX: Ensure un-nested lists are not over-flattened #602
Conversation
|
@bkraft4257 Is there any chance you could give this a shot and verify that it solves your issue? git clone git@github.com:effigies/fmriprep.git
cd fmriprep
git checkout fix/flatten
fmriprep-docker -f fmriprep [...]Where |
| @@ -447,7 +447,8 @@ def init_prepare_epi_wf(ants_nthreads, name="prepare_epi_wf"): | |||
| workflow = pe.Workflow(name=name) | |||
|
|
|||
| def _flatten(l): | |||
| return [item for sublist in l for item in sublist] | |||
| from niworkflows.nipype.utils.filemanip import filename_to_list | |||
| return [item for sublist in l for item in filename_to_list(sublist)] | |||
There was a problem hiding this comment.
Guess I'll just request a review. So the situation is that the input to merge from split is being interpreted as a list of single-character strings.
Given that the output is passed to _flatten before being passed to merge, I believe the correct interpretation is that we get a list with a single element ['/path/to/file']. Expanding that out:
[item for sublist in ['/path/to/file'] for item in sublist]
[item for item in '/path/to/file']
['/', 'p', 'a', 't', 'h', '/', 't', 'o', '/', 'f', 'i', 'l', 'e']Thus, by using filename_to_list in the inner loop, we get:
[item for sublist in ['/path/to/file'] for item in filename_to_list(sublist)]
[item for item in filename_to_list('/path/to/file')]
[item for item in ['/path/to/file']]
['/path/to/file']I believe that because this is the output of a MapNode, the original input must always be a list, so there's no concern about l itself being a string. However, we could be more cautious for basically no cost and apply filename_to_list at both the inner and outer loops.
effigies
changed the title
Do we have a good dataset to test this? It showed up with blip-up/blip-down, presumably single volume field maps.
Closes #598.