Unconditional summary stats results on numerical data: explicit 1D case

[alextanski]

In conjunction with #156 , is it generally possible to get statistics like cube_mean, cube_stddev, etc. in form of a CrunchCube for numeric variables if I am interested in the unconditional result? From what I understand, the measure payload for a mean is not considered as a dimension on its own, so I need to provide at least one variable as a dimension, which will only show the results per subgroup of that variable. Exposing a margin-like result for such cube results would not be enough though, since that one could be different to the real unconditional result for the numeric variable in question.

[alextanski]

With the main issue already solved within the discussion in #156, a problem with the "empty" cube for numerical variables came up in that the CrunchCube.margin() will now throw an IndexError when trying to return the margin (now in the actual sense of the N of the aggregation).

Example dataset: 81242f64e6304afb87486e467dde51fa

numeric    = 'open_realrange'
mean = {
        "function": 'cube_mean',
        "args": [
            {
                "function": "cast",
                "args": [
                    {
                        "variable": "datasets/{}/variables/{}".format(
                            ds.id,
                            ds[numeric].id
                        )
                    },
                    {"class": "numeric"}
                ]
            }
        ]
    }

cube = fetch_cube(ds.resource, [], mean=mean)
cube = CrunchCube(cube)
cube.margin()

IndexErrorTraceback (most recent call last)
<ipython-input-11-ff300642f1d1> in <module>()
      2 cube = fetch_cube(ds.resource, [], mean=mean)
      3 cube = CrunchCube(cube)
----> 4 cube.margin()
      5 raise

C:\Users\alt\AppData\Local\Continuum\anaconda2\lib\site-packages\cr\cube\crunch_cube.pyc in margin(self, axis, weighted, include_missing, include_transforms_for_dims, prune)
    412 
    413         den = self._drop_mr_cat_dims(den, fix_valids=include_missing)
--> 414         if den.shape[0] == 1 and len(den.shape) > 1 and self.ndim < 3:
    415             den = den.reshape(den.shape[1:])
    416         return den

IndexError: tuple index out of range

[alextanski]

bump!