ENH: Drop testing notebooks; only run them

[jhlegarreta]

Drop testing notebooks; only run them. pytest and the nbval plugin
were being used to actually collect notebooks for testing, and to test
cells that output some value. However, nbval's timeout flag to
explicitly set the cell execution time is not working properly, and most
preprocessing notebook cells were not being run.

Thus, this commit favors running all cells over testing the few cells
that print some output value.

[netlify]

✔️ Deploy Preview for carpentries-dmri ready!

🔨 Explore the source changes: b7dbaa9

🔍 Inspect the deploy log: https://app.netlify.com/sites/carpentries-dmri/deploys/623397609d724f0008acf624

😎 Browse the preview: https://deploy-preview-205--carpentries-dmri.netlify.app

[jhlegarreta]

@kaitj @josephmje The CIs are doing their job, and the preprocessing notebook seems to crash:
https://github.com/carpentries-incubator/SDC-BIDS-dMRI/runs/5591600735?check_suite_focus=true#step:13:190

Any chance to have a look and fix that, please? Thanks.

A separate PR would be best, as I'd need to slightly clean the additions in here once we get all notebooks running.

[kaitj]

Took a quick look but I suspect it has something to do with the way the command is called.

\nmkdir -p ../data/ds000221/derivatives/uncorrected_topup_eddy/sub-010006/ses-01/dwi/work\n\n# Create an index file mapping the 67 volumes in 4D dwi volume to the pedir.txt file\nindx=""\nfor i in `seq 1 67`; do\n indx="$indx 1"\ndone\necho $indx > ../data/ds000221/derivatives/uncorrected_topup_eddy/sub-010006/ses-01/dwi/work/index.txt\n\neddy_openmp \\\n --imain=../data/ds000221/sub-010006/ses-01/dwi/sub-010006_ses-01_dwi.nii.gz \\\n --mask=../data/ds000221/derivatives/uncorrected/sub-010006/ses-01/dwi/sub-010006_ses-01_brainmask.nii.gz \\\n --acqp=../data/ds000221/derivatives/uncorrected_topup/sub-010006/ses-01/dwi/work/pedir.txt \\\n --index=../data/ds000221/derivatives/uncorrected_topup_eddy/sub-010006/ses-01/dwi/work/index.txt \\\n --bvecs=../data/ds000221/sub-010006/ses-01/dwi/sub-010006_ses-01_dwi.bvec \\\n --bvals=../data/ds000221/sub-010006/ses-01/dwi/sub-010006_ses-01_dwi.bval \\\n --topup=../data/ds000221/derivatives/uncorrected_topup/sub-010006/ses-01/dwi/work/topup \\\n --out=../data/ds000221/derivatives/uncorrected_topup_eddy/sub-010006/ses-01/dwi/dwi \\\n --repol\n'

It looks like its adding the newline characters with how the notebook cells are run in the CI now. As far as I know, nothing has changed since we ran those tests to time these cells (and when we had tried it with nbval) so it should still run locally. I suspect moving everything to a single line would fix the failure.

[jhlegarreta]

Thanks for the investigation Jason. Maybe defining the paths/filenames as variables will likely help to avoid having an endless, unreadable line, and hopefully make it digestible to the CI (?)

Splitting the mkdir and any command previous to the loop would also help. In general, making our cells smaller might help when teaching the lesson, and would facilitate the type-as-you-teach Carpentries way of teaching I think.

[jhlegarreta]

I've just come across this:
https://pypi.org/project/nbmake/

Might be worthwhile trying: looks to accept/successfully apply the timeout value, avoids the need to convert the notebooks, and if we are lucky we might get back the cell-wise pass/fail reports. I'll have a look as time permits.

[jhlegarreta]

@kaitj The same eddy_openmp step is failing in PRs #207 and #208 (I tried these two options to investigate which one provides the best report, allows to drop additional plugins, etc.)