AFNI Qwarp output paths are incorrect

[salma1601]

Summary

output paths are not correct: current directory is always used as parent directory and +tlrc is appended even to nifti outputs

How to replicate the behavior

from nipype.interfaces import afni
a = afni.Qwarp()
a.inputs.in_file = '/tmp/in_file.nii.gz'
a.inputs.base_file = '/tmp/base_file.nii.gz'
a.inputs.out_file = '/tmp/warped.nii.gz'
res = a.run()
print(res.outputs.warped_source)

prints a wrong path

In [20]:  /home/salma/CODE/nipype/warped14+tlrc.nii.gz

[oesteban]

Please note that the input spec for out_file defines both name_source and genfile. That needs be fixed. Attached comment on your PR. Thanks for spotting this.

[salma1601]

I am not sure I understand correctly but I removed the genfile from here https://github.com/nipy/nipype/blob/master/nipype/interfaces/afni/preprocess.py#L2883 and still have the same issue

[oesteban]

Hi @salma1601, name_source and genfile are two metadata you can use on input File traits to automatically generate names of outputs.

• genfile is a legacy method. You set it True in the File metadata (meaning name = File(genfile=True)) and then, your interface must define a _gen_filename() method (the name may not be exact, I don't recall on top of my head).

• name_source is the currently advised method. It requires you to define up to 3 metadata bits like so: some_input = File(name_source=['other_input'], name_template='%s_suffix', output_name='some_output'). This bit of code is saying if some_input is not set, then create a new name reading the base name of the file in other_input format it through name_template, and use that new name to set the output some_output. Default values for name_template and output_name are %s_generated and the input name (some_input in this case).

Therefore, genfile and name_source are mutually exclusive and should not be used together. I'll have a look to see why this is not working in your PR.

[oesteban]

Hi again, sorry, I can't see where you are trying removing genfile. Could you point me to this?

[salma1601]

I just removed the line genfile=True

out_file = File(argstr='-prefix %s',
                    name_template='%s_QW',
                    name_source=['in_file'],
                    genfile=True,
                    desc='out_file ppp'
'Sets the prefix for the output datasets.'

but this doesn't solve the issue.

[oesteban]

what do you get when you do interface_object.cmdline in your use-case and why that seems to be wrong?

[salma1601]

In [2]: a.cmdline
Out[2]: u'3dQwarp -base /tmp/base_file.nii.gz -source /tmp/in_file.nii.gz -prefix /tmp/warped.nii.gz'

In [3]: res.outputs
Out[3]: 

base_warp = <undefined>
source_warp = /home/salma/CODE/nipype/warped_WARP+tlrc.nii.gz
warped_base = <undefined>
warped_source = /home/salma/CODE/nipype/warped+tlrc.nii.gz
weights = <undefined>

what is wrong is that

1. it is appending +tlrc to the filename
2. it is using the current directory as a parent directory instead of the directory given in the command (\tmp)
   so all the paths in the outputs do not exist

[oesteban]

Some thoughts:

1. As I mentioned before, this code should not be necessary:

https://github.com/salma1601/nipype/blob/ca3e8a3f28e37faff2bac5de272a486a41823206/nipype/interfaces/afni/preprocess.py#L3513-L3515

2. Second, the name_source automatically updates the name of the outputs unless a workaround is set in place. That is exactly what _list_outputs() is doing here:

https://github.com/salma1601/nipype/blob/ca3e8a3f28e37faff2bac5de272a486a41823206/nipype/interfaces/afni/preprocess.py#L3476-L3489

So that logic should also go away if you want the name_source to perform its magic.