Allow using lists of kwargs in morph_stats features

[eleftherioszisis]

Summary:

• Added new config support for adding the same feature multiple times with different arguments (see below)
• If kwargs are populated in feature they are appended to its name (f'{mode}_{feature_name}__{suffix}')
• The suffix includes each feature's kwarg present in the config, apart from neurite_type, e.g. variant:branch-order__method:uylings
• Config sanitization has become the point where all different versions are converted to the single valid one for the rest of the program.
• Made protected all the functions apart from extract_stats, extract_dataframe, and main.
• Sanitization has moved from main to extract_stats to guarantee that regardless of the entry point in the module, i.e. using any of the three functions in the previous point, the config is sanitized the same way.

A list of features looks like this:

morphology:
  soma_radius:
    modes:
      - mean
  max_radial_distance:
    modes:
      - mean
neurite:
  section_lengths:
    modes:
      - max
      - sum
  section_volumes:
    modes:
      - sum
  section_branch_orders:
    modes:
      - max
      - raw
  segment_midpoints:
    modes:
      - max
  max_radial_distance:
    modes:
      - mean
  principal_direction_extents:
    kwargs:
      - direction: 1
      - direction: 0
    modes:
      - min
      - sum
  partition_asymmetry:
    kwargs:
        - variant: branch-order
          method: petilla
        - variant: length
          method: uylings
    modes:
        - min
        - max
neurite_type:
- AXON
- APICAL_DENDRITE
- BASAL_DENDRITE
- ALL

Using that extract_stats results to:

{'all': {'max_partition_asymmetry__variant:branch-order__method:petilla': 0.9,
         'max_partition_asymmetry__variant:length__method:uylings': 0.8795344229855895,
         'max_section_branch_orders': 10,
         'max_section_lengths': 11.758282,
         'max_segment_midpoints_0': 64.40167236328125,
         'max_segment_midpoints_1': 48.48197937011719,
         'max_segment_midpoints_2': 53.750946044921875,
         'mean_max_radial_distance': 99.58946,
         'min_partition_asymmetry__variant:branch-order__method:petilla': 0.0,
         'min_partition_asymmetry__variant:length__method:uylings': 0.00030289197373727377,
         'min_principal_direction_extents__direction:0': 80.01260723801528,
         'min_principal_direction_extents__direction:1': 5.274489134421081,
         'raw_section_branch_orders': [0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5, 6, 6, 7,
                                       7, 8, 8, 9, 9, 10, 10, 0, 1, 1, 2, 2, 3,
                                       3, 4, 4, 5, 5, 6, 6, 7, 7, 8, 8, 9, 9,
                                       10, 10, 0, 1, 1, 2, 2, 3, 3, 4, 4, 5, 5,
                                       6, 6, 7, 7, 8, 8, 9, 9, 10, 10, 0, 1, 1,
                                       2, 2, 3, 3, 4, 4, 5, 5, 6, 6, 7, 7, 8, 8,
                                       9, 9, 10, 10],
         'sum_principal_direction_extents__direction:0': 356.3872399244001,
         'sum_principal_direction_extents__direction:1': 25.26438016226154,
         'sum_section_lengths': 840.68524,
         'sum_section_volumes': 1104.9077745403258},
...
 'morphology': {'mean_max_radial_distance': 99.58946,
                'mean_soma_radius': 0.13065629977308288}}

Notice how the multiple feature entries with different kwargs result into different name:

 min_principal_direction_extents__direction:1

Closes #1015

[codecov-commenter]

Codecov Report

Merging #1021 (25106ce) into master (f4cb039) will not change coverage.
The diff coverage is 100.00%.

@@            Coverage Diff            @@
##            master     #1021   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files           36        36           
  Lines         2401      2403    +2     
=========================================
+ Hits          2401      2403    +2     

[adrien-berchet]

Nice, thanks for this!

[adrien-berchet]

Do we want to keep both formats? Shouldn't we raise a deprecation warning for the dict format?

[eleftherioszisis]

Not sure where we stand wrt to deprecating the old formats. @mgeplf , should I add warnings for all previous config layouts?

[mgeplf]

Hmm, the way I was thinking about adding this was to make, the kwargs (which was poorly named), a list:

neurite:
    partition_asymmetry:
        kwargs:
           - {}
           - { variant: 'length',  method: 'petilla' }

and then isinstance kwargs to see if it was a dict (and then use the old behavior), or a list(and then use the new behavior. I think the config could be more minimal this way; however, the modes would be the same for both kwargs variants.

[eleftherioszisis]

I like this. I will change it accordingly!
@lidakanari , @adrien-berchet , that would work for you right?

[eleftherioszisis]

I have implemented @mgeplf suggestion for multiple kwargs. @lidakanari , @adrien-berchet could you please try it and let me know if it works for you?Thanks.

[mgeplf]

LGTM; but let's wait for the feedback on usage.

[lidakanari]

Yep, that works well for me, thanks!

[adrien-berchet]

Looks perfect, thank you very much!