ENH: Modify Directory and File traits to get along with pathlib

[oesteban]

Ref #2944

Summary

Fixes # .

List of changes proposed in this PR (pull-request)

Acknowledgment

• (Mandatory) I acknowledge that this contribution will be available under the Apache 2 license.

[satra]

looks like a reasonable PR. the File trait should take hash_files into account.

the Directory trait is a weird one since hash_files doesn't apply to it as we don't hash the contents of a directory.

[codecov-io]

Codecov Report

Merging #2959 into master will decrease coverage by 0.03%.
The diff coverage is 78.22%.

@@            Coverage Diff            @@
##           master   #2959      +/-   ##
=========================================
- Coverage   64.24%   64.2%   -0.04%     
=========================================
  Files         342     342              
  Lines       43692   43727      +35     
  Branches     5453    5465      +12     
=========================================
+ Hits        28068   28074       +6     
- Misses      14545   14562      +17     
- Partials     1079    1091      +12

Flag	Coverage Δ
#unittests	64.2% <78.22%> (-0.04%)	⬇️

Impacted Files	Coverage Δ
nipype/info.py	88.05% <ø> (ø)	⬆️
nipype/interfaces/afni/base.py	66.66% <ø> (-0.27%)	⬇️
nipype/interfaces/base/support.py	79.65% <ø> (ø)	⬆️
nipype/interfaces/io.py	49.52% <0%> (-0.12%)	⬇️
nipype/interfaces/base/specs.py	87.24% <100%> (+0.54%)	⬆️
nipype/interfaces/dipy/base.py	77.96% <100%> (+0.77%)	⬆️
nipype/interfaces/spm/base.py	57.94% <100%> (-0.14%)	⬇️
nipype/interfaces/afni/preprocess.py	82.1% <100%> (+0.08%)	⬆️
nipype/interfaces/base/traits_extension.py	84.35% <80.68%> (-10.22%)	⬇️
nipype/pipeline/plugins/legacymultiproc.py	63% <0%> (-4.5%)	⬇️
... and 4 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 e722d61...8069604. Read the comment docs.

[satra]

@oesteban - do we really need to enable usedefault=True for all File traits?

[oesteban]

I'm thinking it should be enabled only if the default value is given.

[effigies]

@oesteban That's counter to how all of the other traits work, which require manual enabling of usedefault=True. While it's a little annoying to have to add when you set a default, this change breaks consistency.

[oesteban]

The current implementation of interfaces setting a default value on Files behave as if they had set usedefault=True.

[effigies]

Oh, fun.

[satra]

The current implementation of interfaces setting a default value on Files behave as if they had set usedefault=True.

i don't think this is true.

In [1]: from nipype.interfaces.base import TraitedSpec, File

In [2]: class A(TraitedSpec):
   ...:     foo = File()

In [3]: a = A()

In [4]: a
Out[4]: 

foo = <undefined>

In [5]: class A(TraitedSpec):
   ...:     foo = File('/foo')

In [6]: a = A()

In [7]: a
Out[7]: 

foo = <undefined>

In [12]: class A(TraitedSpec):
    ...:     foo = File(value='/foo')

In [13]: a = A()

In [14]: a
Out[14]: 

foo = <undefined>

and when usedefault=True then:

In [15]: class A(TraitedSpec):
    ...:     foo = File(value='/foo', usedefault=True)

In [16]: a = A()

In [17]: a
Out[17]: 

foo = /foo

[oesteban]

Okay, yes, I double-checked and realized I was misinterpreting one test. Reverting.

[effigies]

It's pretty unclear what are the necessary changes and which ones were done while you were at it. If the new trait is a drop-in replacement, then all the SPM/AFNI changes are superfluous. If not, then are we breaking downstream code?

[oesteban]

Closed in favor of #2962