Cube.aggregated_by() broken

[bblay]

Cube.aggregated_by() appears to be broken.
The aggregated data dimension is not being collapsed.

https://github.com/bblay/iris_issue_files/blob/master/test_agg.py

$  python test_agg.py 
------------------
original_cube                       (w: 3; z: 3; y: 3; x: 3)
     Dimension coordinates:
          w                           x     -     -     -
          z                           -     x     -     -
          y                           -     -     x     -
          x                           -     -     -     x
shape (3, 3, 3, 3)
------------------
aggregated_cube                     (*ANONYMOUS*: 3; z: 3; y: 3; x: 3)
     Dimension coordinates:
          z                                     -     x     -     -
          y                                     -     -     x     -
          x                                     -     -     -     x
     Auxiliary coordinates:
          w                                     x     -     -     -
     Attributes:
          history: Mean of original_cube aggregated over w
     Cell methods:
          mean: w
shape (3, 3, 3, 3)

[bblay]

Thanks to esc24 for pointing out this is not a bug, I was using aggregated_by when I should have been using collapsed.

[esc24]

I believe there is some confusion here between Cube.aggregated_by() and Cube.collapsed(). You've specified the 'w' coordinate as the "group coordinate" but it has no repeating values so it takes the means of each group containing a single value. I suspect you want to use Cube.collapsed() passing it an aggregator.

It is worth noting that this seemingly null process does however convert the DimCoord to an AuxCoord. In my opinion it should not do this.

[pelson]

It is worth noting that this seemingly null process does however convert the DimCoord to an AuxCoord. In my opinion it should not do this.

@esc24: Issue for it?

[bblay]

It's also worth noting that the name can cause confusion. Aggregation often means collapsed(), as in the NIMROD user story text (which is how I jumped on the wrong routine). I suppose partially_collapsed() is too long...compressed() is ok...unique_coord_points() is accurate but ugly...reduced() is ok...