ENH: Interfaces with automated population of outputs

[oesteban]

trafficstars

Proof of concept for #3149

This is a very rough draft of how such an interface would look like. It only tests automatic templated names filling (glob patterns should be fairly easy to add, and regexes too).

Some wrinkles around the corners need to be ironed out. E.g., implementing a keep_ext metadata for outputs, and raising an error if two inputs are set in the template and extensions are different. Similarly, raise error if not all the arguments were replaced.

But in general, it seems doable.

BTW the pattern matching and replacing part of this PR could also be useful to accept the name_template metadata in BaseInterfaces (currently, it is only allowed in command line interfaces).

WDYT @satra, @mgxd, @djarecka, @mgxd ?

EDIT: The new AutoOutputInterface is basically a copy of the current BaseInterface, but removing the _list_outputs and aggregate_outputs, and adding this: https://github.com/nipy/nipype/pull/3150/files#diff-1353462138ab420e9b6ce7f8a19ab284R407-R456

[codecov[bot]]

Codecov Report

Merging #3150 into master will decrease coverage by 3.76%. The diff coverage is 45.16%.

@@            Coverage Diff             @@
##           master    #3150      +/-   ##
==========================================
- Coverage   67.62%   63.86%   -3.77%     
==========================================
  Files         299      298       -1     
  Lines       39487    39623     +136     
  Branches     5220     5268      +48     
==========================================
- Hits        26703    25304    -1399     
- Misses      12073    13252    +1179     
- Partials      711     1067     +356

Flag	Coverage Δ
#smoketests	?
#unittests	63.86% <45.16%> (-1.01%)	:arrow_down:

Impacted Files	Coverage Δ
nipype/interfaces/base/experimental.py	45.16% <45.16%> (ø)
nipype/interfaces/nilearn.py	38.98% <0%> (-57.63%)	:arrow_down:
nipype/workflows/__init__.py	47.05% <0%> (-52.95%)	:arrow_down:
nipype/utils/spm_docs.py	20% <0%> (-48%)	:arrow_down:
nipype/interfaces/freesurfer/base.py	45.68% <0%> (-32.76%)	:arrow_down:
nipype/utils/logger.py	58.46% <0%> (-30.77%)	:arrow_down:
nipype/algorithms/rapidart.py	34.42% <0%> (-29.68%)	:arrow_down:
nipype/interfaces/spm/base.py	57.66% <0%> (-29.34%)	:arrow_down:
nipype/utils/provenance.py	55.16% <0%> (-28.71%)	:arrow_down:
nipype/interfaces/fsl/model.py	55.18% <0%> (-25.4%)	:arrow_down:
... and 37 more

---

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 5d2fe1d...75d12f4. Read the comment docs.

[effigies]

This seems reasonable on first glance. I think problems are likely to be found while trying to work through a few examples.

I assume that we'll still support the old style, though? Otherwise a lot of interfaces will need to be rewritten mostly from scratch.

[satra]

@oesteban - this looks reasonable.