🐛🐍 "Module CerebrospinalFluid_mean has no output called stats"

[shnizzedy]

In 0e7db8c, I'm getting the following error every time I run test_config:

200218-22:29:47,996 nipype.utils INFO:
	 No new version available.
Traceback (most recent call last):
  File "/code/run.py", line 587, in <module>
    test_config = 1 if args.analysis_level == "test_config" else 0
  File "/code/CPAC/pipeline/cpac_runner.py", line 341, in run
    p_name, plugin, plugin_args, test_config)
  File "/code/CPAC/pipeline/cpac_pipeline.py", line 2540, in prep_workflow
    name='nuisance_regressor_{0}_{1}'.format(regressors_selector_i, num_strat)
  File "/code/CPAC/nuisance/nuisance.py", line 1303, in create_regressor_workflow
    build_nuisance_regressors, regressor_arg
  File "/usr/local/miniconda/lib/python3.7/site-packages/nipype/pipeline/engine/workflows.py", line 236, in connect
    raise Exception("\n".join(["Some connections were not found"] + infostr))
Exception: Some connections were not found
Module CerebrospinalFluid_mean has no output called stats

I don't get this error running 3f22187, so I think this problem is in either the Python 2 → 3 / Ubuntu 16 → 18 upgrade.

AFNI took some massaging to get to work in this upgrade, so I wouldn't be surprised if another tool is working not as expected. The 2 → 3 changes are huge as well, so there's also a good chance the problem is there.

What software should be generating CerebrospinalFluid_mean stats?

[anibalsolon]

This is the CerebrospinalFluid_mean:

C-PAC/CPAC/nuisance/nuisance.py

Lines 1190 to 1204 in 0e7db8c

	mean_node = pe.Node(
	afni.ROIStats(quiet=False, args='-1Dformat'),
	name='{}_mean'.format(regressor_type)
	)
	nuisance_wf.connect(
	summary_method_input[0], summary_method_input[1],
	mean_node, 'in_file'
	)
	nuisance_wf.connect(
	union_masks_paths, 'out_file',
	mean_node, 'mask'
	)
	summary_method_input = (mean_node, 'stats')

I guess you have updated the Nipype version, in which the 'stats' output attribute is now called 'out_file':

nipy/nipype@05c056c#diff-851b9f467ed8042d29a89dbd00b754dfL1820-R1897

[shnizzedy]

Thanks! 🙏

That's probably it! I unlocked all the Python package versions to build this new Python3-based image. I'll re-lock everything that's Py3 to what we had before, try again, and then if it all seems to work lock what's left.

[shnizzedy]

If we want to upgrade Nipype, we can decide that, but I think that's beyond the scope of this upgrade

[anibalsolon]

You are right, outside the scope, the port to Py3 can bring several problems already. No need to include a new variable.

[shnizzedy]

Resolved in 8c659d1

[shnizzedy]

😬 maybe this is not well-resolved. Now I'm getting empty CerebrospinalFluid_mean/roi_stats.csv files:

ValueError: Number of time points in /tmp/resting_preproc_sub-0025427_ses-3/nuisance_regressor_0_0/_scan_rest_run-1/_selector_CSF-2mmE-M_aC-CSF+WM-2mm-DPC5_G-M_M-SDB_P-2_BP-B0.01-T0.1/CerebrospinalFluid_mean/roi_stats.csv (0) is inconsistent with length of functional file /tmp/resting_preproc_sub-0025427_ses-3/func_preproc_afni_3dvolreg_0/_scan_rest_run-1/func_normalize/sub-0025427_ses-3_task-rest_run-1_bold_calc_tshift_resample_volreg_calc_maths.nii.gz (300)

ValueError: Number of time points in /tmp/resting_preproc_sub-0025427_ses-3/nuisance_regressor_1_0/_scan_rest_run-1/_selector_CSF-2mmE-M_aC-CSF+WM-2mm-DPC5_M-SDB_P-2_BP-B0.01-T0.1/CerebrospinalFluid_mean/roi_stats.csv (0) is inconsistent with length of functional file /tmp/resting_preproc_sub-0025427_ses-3/func_preproc_afni_3dvolreg_0/_scan_rest_run-1/func_normalize/sub-0025427_ses-3_task-rest_run-1_bold_calc_tshift_resample_volreg_calc_maths.nii.gz (300)

[pn2200]

Is this the same issue as #1140?

[shnizzedy]

Is this the same issue as #1140?

They might be related, but this issue is occurring exclusively on the upgrade/python3 branch while #1140 seems to be occurring in Python 2 as well

[sgiavasis]

I believe this can be removed as a blocker for the Python 3 PR.