Create subplots over groupby rather then by

[jsignell]

I am not sure how the ambiguity of this should be handled, but I have a plot with both a groupby and a by. I don't want to have a selector widget so I'd like to be able to get subplots using subplots=True. But this splits into subplots over the by field rather than the groupby

To Reproduce:

import pandas as pd
import hvplot.pandas

df = pd.read_csv('benchmarking.txt')
df['total_cores'] = df['njobs'] * df['ncores']

df.hvplot('total_cores', 'seconds', by='nworkers', groupby='operation', subplots=True,
                width=400, height=400, ylim=(0,1))

benchmarking.txt

[jbednar]

I admit I've been confused between by and groupby...

[jsignell]

In general by creates overlays and groupby creates selectors. The work-around that Philipp suggests is to use call .layout() on the output:

[philippjfr]

@tonyfast suggests that we could alias by with hue which is used by seaborn and is a lot clearer.

[philippjfr]

@holoviz/triaging-team any thoughts?

[jbednar]

No objection; just need to document that hue is the same as by...

[maximlt]

I've also been confused by by and groupby. @philippjfr @jbednar still in favor of aliasing by with hue? It sounds like a good idea.

[maximlt]

xarray plotting also uses hue.

[droumis]

Our discussion on discord is leaning toward the use of colorby instead of hue

[philippjfr]

Tbh, I have some issues with both colorby and hue because we already support color=<dimension≥. The current by operation effectively maps onto an NdOverlay in HoloViews. Most users won't care about that but it's still conceptually different from simply coloring by a dimensions, it just so happens that we define a color cycle for layers in an overlay. Sadly I don't have any good suggestions to communicate the subtleties.