New version 1.2.0 generates str error with mrTrix denoise, biascorrect

[lucindasisk]

Hi! I have a DWI workflow that has run smoothly in Nipype version 1.1.8, utilizing mrTrix nodes including DWIDenoise. I upgraded to version 1.2.0 so I could incorporate MRDeGibbs (Gibbs ringing removal), but the denoising node generated the following error after running: "TypeError: expected str, bytes or os.PathLike object, not _Undefined". DWIBiasCorrect also generated same error. I downgraded to version 1.1.8 again, removed the Gibbs node, and everything ran smoothly. Thanks for any help!

Script/Workflow details

Node

#Denoise DWI data using local PCA correction - mrTrix3
denoise = Node(mtx.DWIDenoise(),
              name='denoise')

Workflow

(drop2, denoise, [('roi_file', 'in_file')]),
                      (denoise, datasink, [('out_file', '1_Check_Unwarped.@par.@par.@par')])

Platform details:

mac OS Mojave 10.14.5

[effigies]

Could you provide the full traceback and workflow? And can you verify that upgrading to 1.2.0 while not changing the script causes the problem?

[lucindasisk]

Definitely, thanks for your help! I just tested it again, and while it ran fine in version 1.1.8, it crashed again when I upgraded to 1.2.0. It also runs in version 1.1.9. Script is here. Traceback below:

Traceback (most recent call last):
  File "1_DTI_Preprocessing-SHAPES.py", line 247, in <module>
    preproc = preproc_flow.run('MultiProc', plugin_args={'n_procs': 4})
  File "/Users/lucindasisk/.local/share/virtualenvs/lucindasisk-jFZfvjp6/lib/python3.6/site-packages/nipype/pipeline/engine/workflows.py", line 599, in run
    runner.run(execgraph, updatehash=updatehash, config=self.config)
  File "/Users/lucindasisk/.local/share/virtualenvs/lucindasisk-jFZfvjp6/lib/python3.6/site-packages/nipype/pipeline/plugins/base.py", line 161, in run
    self._clean_queue(jobid, graph, result=result))
  File "/Users/lucindasisk/.local/share/virtualenvs/lucindasisk-jFZfvjp6/lib/python3.6/site-packages/nipype/pipeline/plugins/base.py", line 224, in _clean_queue
    raise RuntimeError("".join(result['traceback']))
RuntimeError: Traceback (most recent call last):
  File "/Users/lucindasisk/.local/share/virtualenvs/lucindasisk-jFZfvjp6/lib/python3.6/site-packages/nipype/pipeline/plugins/multiproc.py", line 69, in run_node
    result['result'] = node.run(updatehash=updatehash)
  File "/Users/lucindasisk/.local/share/virtualenvs/lucindasisk-jFZfvjp6/lib/python3.6/site-packages/nipype/pipeline/engine/nodes.py", line 472, in run
    result = self._run_interface(execute=True)
  File "/Users/lucindasisk/.local/share/virtualenvs/lucindasisk-jFZfvjp6/lib/python3.6/site-packages/nipype/pipeline/engine/nodes.py", line 563, in _run_interface
    return self._run_command(execute)
  File "/Users/lucindasisk/.local/share/virtualenvs/lucindasisk-jFZfvjp6/lib/python3.6/site-packages/nipype/pipeline/engine/nodes.py", line 643, in _run_command
    result = self._interface.run(cwd=outdir)
  File "/Users/lucindasisk/.local/share/virtualenvs/lucindasisk-jFZfvjp6/lib/python3.6/site-packages/nipype/interfaces/base/core.py", line 377, in run
    outputs = self.aggregate_outputs(runtime)
  File "/Users/lucindasisk/.local/share/virtualenvs/lucindasisk-jFZfvjp6/lib/python3.6/site-packages/nipype/interfaces/base/core.py", line 455, in aggregate_outputs
    predicted_outputs = self._list_outputs()
  File "/Users/lucindasisk/.local/share/virtualenvs/lucindasisk-jFZfvjp6/lib/python3.6/site-packages/nipype/interfaces/mrtrix3/preprocess.py", line 81, in _list_outputs
    outputs['out_file'] = op.abspath(self.inputs.out_file)
  File "/Users/lucindasisk/.local/share/virtualenvs/lucindasisk-jFZfvjp6/lib/python3.6/posixpath.py", line 369, in abspath
    path = os.fspath(path)
TypeError: expected str, bytes or os.PathLike object, not _Undefined

[effigies]

Okay. It looks like this was introduce in #2904. @josephmje Could you clarify the reasons for the changes to DWIDenoise outputs? Looking at it, I don't immediately understand why out_file was changed.

[lucindasisk]

Thank you @effigies!

[josephmje]

Sorry @effigies @lucindasisk . I wanted to work with the output noise file in the rest of my workflow and modified the _list_outputs() method as suggested in the PR. But it looks like the change to the DWIDenoise class causes things to break unless out_file is specifically defined.

I've been working through this by modifying the noise trait in DWIDenoiseInputSpec to:

    noise = File(
        argstr="-noise %s",
        name_template="%s_noise",
        name_source=["in_file"],
        keep_extension=True,
        desc="the output noise map",
    )

and removing the _list_outputs() method. But I'm not sure if this is the correct solution.

[oesteban]

Is the -noise argument optional?

[josephmje]

Yes it is.

[oesteban]

Then the name_template logic will include it automatically, so your suggestion has made it mandatory in practice. Do you know if there is any change in the methodology if you use -noise?

If it is just enabling the output, I don't think it would hurt to make it present always (via name_template).

[josephmje]

There is no change in the methodology by including -noise. The function creates it anyways but adding the option saves it to the filesystem.