ENH: Allow input weight for afni's volreg.

[williford]

This PR should be discussed before accepting.

I've added the following to VolregInputSpec:

    in_weight_file = File(
        desc='File for input weighting volume',
        argstr="-weight '%s[0]'",
        exists=True)

Apparently, 3dvolreg requires the volume number for the weight. I don't know of a good way to do this with NiPype, so I make it just take the first volume (I think typically someone would only pass a single volume image). Is this sufficient?

Also there doesn't seem to be a standard way for naming the input weights. I went with a more verbose version, but feedback is welcome.

I tried to run make check-before-commit and received the error "ImportError: No module named packaging.version". I'm running the code in virtualenv with Python3, using pip to install.

[codecov-io]

Codecov Report

Merging #2155 into master will increase coverage by <.01%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master    #2155      +/-   ##
==========================================
+ Coverage   72.25%   72.25%   +<.01%     
==========================================
  Files        1168     1168              
  Lines       58405    58406       +1     
  Branches     8400     8400              
==========================================
+ Hits        42201    42204       +3     
+ Misses      14870    14868       -2     
  Partials     1334     1334

Flag	Coverage Δ
#smoketests	72.25% <100%> (ø)	⬆️
#unittests	69.91% <100%> (ø)	⬆️

Impacted Files	Coverage Δ
nipype/interfaces/afni/preprocess.py	81.4% <100%> (+0.02%)	⬆️
nipype/interfaces/base.py	83.48% <0%> (+0.18%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 32e724c...ab14a0e. Read the comment docs.

[codecov-io]

Codecov Report

Merging #2155 into master will increase coverage by <.01%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master    #2155      +/-   ##
==========================================
+ Coverage   72.25%   72.25%   +<.01%     
==========================================
  Files        1168     1168              
  Lines       58405    58406       +1     
  Branches     8400     8400              
==========================================
+ Hits        42201    42204       +3     
+ Misses      14870    14868       -2     
  Partials     1334     1334

Flag	Coverage Δ
#smoketests	72.25% <100%> (ø)	⬆️
#unittests	69.91% <100%> (ø)	⬆️

Impacted Files	Coverage Δ
nipype/interfaces/afni/preprocess.py	81.4% <100%> (+0.02%)	⬆️
nipype/interfaces/base.py	83.48% <0%> (+0.18%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 32e724c...ab14a0e. Read the comment docs.

[codecov-io]

Codecov Report

Merging #2155 into master will increase coverage by <.01%.
The diff coverage is 100%.

@@            Coverage Diff             @@
##           master    #2155      +/-   ##
==========================================
+ Coverage   72.25%   72.25%   +<.01%     
==========================================
  Files        1168     1168              
  Lines       58405    58406       +1     
  Branches     8400     8400              
==========================================
+ Hits        42203    42204       +1     
  Misses      14868    14868              
  Partials     1334     1334

Flag	Coverage Δ
#smoketests	72.25% <100%> (ø)	⬆️
#unittests	69.91% <100%> (ø)	⬆️

Impacted Files	Coverage Δ
nipype/interfaces/afni/preprocess.py	81.4% <100%> (+0.02%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 32e724c...6e1a392. Read the comment docs.

[effigies]

One option is to use a tuple trait, which would require the user to specify the volume.

in_weight_volume = traits.Tuple(File(exists=True), traits.Int, argstr="-weight '%s[%d]'", ...)

However, using the 0th volume seems like a reasonable default, so what about the following:

in_weight_volume = traits.Either(traits.Tuple(File(exists=True), traits.Int),
                                 File(exists=True), argstr="-weight '%s[%d]'", ...)

Then, in the interface class, override _format_arg:

def _format_arg(self, name, trait_spec, value):
    if name == 'in_weight_volume' and not isinstance(value, Tuple):
        value = (value, 0)
    return super(Volreg, self)._format_arg(name, trait_spec, value)

Thus, if you pass volreg.inputs.in_weight_volume = 'functional.nii', you get -weight 'functional.nii[0]', but if you pass volreg.inputs.in_weight_volume = ('functional.nii', 10), you get -weight 'functional.nii[10]'. (I'd add these to the docstring to make sure that it works as I expect.)

[satra]

@williford - could you please merge with current master and update this PR w.r.t @effigies comments?

[effigies]

Hi @williford. Do you have a few minutes to finish this? It would be nice to have it in for 1.0.

If you don't have the time, I can push my suggested changes to your branch.

[williford]

@effigies Sorry for my delay! I just implemented your suggestions, updated the minimal testing, and enabled maintainers to modify my PR. Feel free to make the changes you want.

I did get one error when running tests, but it was unrelated to volreg. The volreg tests passed.

[williford]

I failed to fetch the current changes from nipype master branch...

[williford]

I'm closing this PR, since I created a clean version that is up-to-date with Master here:
#2396