Groupby for hv.Dataset().to.distribution errors

[ahuang11]

But works with hvplot

import numpy as np
import hvplot.pandas
import holoviews as hv
hv.extension('bokeh')

from bokeh.sampledata.autompg import autompg as df

title = "MPG by Cylinders and Data Source, Colored by Cylinders"
df.hvplot.kde('mpg', 'cyl')
hv.Dataset(df).to.distribution('mpg', ['cyl'])

~/anaconda3/envs/bokeh-dev/lib/python3.6/site-packages/param/__init__.py in _check_bounds(self, val)
   1171             if min_length is not None and max_length is not None:
   1172                 if not (min_length <= l <= max_length):
-> 1173                     raise ValueError("%s: list length must be between %s and %s (inclusive)"%(self._attrib_name,min_length,max_length))
   1174             elif min_length is not None:
   1175                 if not min_length <= l:

ValueError: kdims: list length must be between 1 and 1 (inclusive)

[marcbernot]

It is not an error, you need to conform to the syntax of the .to() method.
You operate a groupby on 'cyl' with this line :
df.hvplot.kde('mpg','cyl')
To do the same in holoviews you can use this syntax :
hv.Dataset(df).to(hv.Distribution,'mpg', None,'cyl').overlay()
or
hv.Dataset(df).to(hv.Distribution,kdims='mpg',groupby='cyl').overlay()
Check gridded datasets groupby section to learn more about the .to() method.

[ahuang11]

Ah okay; I guess I'm used to hv.Dataset(df).to.curve('mpg', 'cyl', []) and hv.Dataset(df).to.bars('mpg', 'cyl', []) instead of hv.Dataset(df).to(hv.Curve, 'mpg', 'cyl', []) so I expected hv.Distribution would work similarly. Thanks!

[marcbernot]

Oh sorry, i think you should reopen this issue and that there is indeed an error.
By looking at the signature of the method you want to use :
.to.distribution(dim=None,groupby=[],**kwargs)
it looks like you should be able to use the syntax:
hv.Dataset(df).to.distribution('mpg', ['cyl'])
or even
hv.Dataset(df).to.distribution('mpg', 'cyl')
At first I thought you could do the groupby only within the .to().

Error 1 : this latter yields a "Dimension 'c' not found" due to the absence of brackets ; but it seems to be an error because holoviews is bracket tolerant in a lot of similar situations.

Error 2 : in element/init.py line 44
hv.Dataset(df).to._element.reindex(['cyl'],['mpg']) yields a :Dataset [cyl] (mpg) so after the groupby you are left with no kdim hence the error "ValueError: kdims: list length must be between 1 and 1 (inclusive)".
reindexed = hv.Dataset(df).to._element.reindex(['cyl','mpg']) reindexed.groupby('cyl',hv.HoloMap,hv.Distribution).overlay()
gives the expected result.

[github-actions]

This issue has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.