Reading freesurfer stats files correctly

[simkarwin]

No description provided.

[effigies]

A couple thoughts:

1. Why a class with only static methods? I think flat functions would make more sense and be consistent with this module.

2. Is it possible to do what you want with plain numpy? Pandas is a very heavy dependency that we are unlikely to make an installation requirement.

[simkarwin]

The code has been revised regard to your comments

[codecov]

Codecov Report

Attention: Patch coverage is 35.71429% with 9 lines in your changes missing coverage. Please review.

Project coverage is 95.33%. Comparing base (276f786) to head (02334f4).

Files with missing lines	Patch %	Lines
nibabel/freesurfer/io.py	35.71%	9 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1152      +/-   ##
==========================================
- Coverage   95.36%   95.33%   -0.03%     
==========================================
  Files         207      207              
  Lines       29680    29694      +14     
  Branches     4992     4996       +4     
==========================================
+ Hits        28305    28310       +5     
- Misses        935      944       +9     
  Partials      440      440              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[effigies]

Thanks for this. I've made a couple comments.

A larger question is whether this is intended to be for any stats files in FreeSurfer, or just the ones generated by mris_anatomical_stats? We might want to do some detecting of the kind of stats in order to correctly build a dtype. I imagine a column header -> dtype lookup table would not be too difficult to construct.

What do you think?

[effigies]

If we're going to have example code, it needs to run. We could drop an example file in nibabel/tests/data/ and access it with

from nibabel.testing import test_data
stats_a2009 = read_stats_file(test_data(fname='lh.aparc.a2009s.stats'))

[effigies]

You're no longer returning column names. I think you can use numpy.lib.recfunctions.join_by to do what you want.

I would also use the following pattern for doctests:

    .. testsetup::

        >>> import os
        >>> from nibabel.testing import get_test_data
        >>> cwd = os.getcwd()
        >>> os.chdir(get_test_data())

    >>> stats_a2009 = read_stats_file_both_hemispheres(r'lh.aparc.a2009s.stats')

    .. testcleanup::

        >>> os.chdir(cwd)

[effigies]

@simkarwin Sorry about the long radio silence. Any interest in finishing this up?