Enhance resolve_dependent_value if value is list or tuples

[hoxbro]

Recursive calling resolve_dependent_value if value is a list or tuple. If this implementation crash and burn in the CI an alternative is the changes described here holoviz/hvplot#697 (comment).

This PR should make it able to run:

import holoviews as hv
import panel as pn
from bokeh.sampledata.autompg import autompg_clean as df

hv.extension("bokeh")

xaxis = pn.widgets.RadioButtonGroup(name="X axis", options=["mpg", "yr"])
ds = hv.Dataset(df)
transform = hv.util.transform.df_dim("*")
transform.groupby(["origin", xaxis]).apply(ds, keep_index=True, compute=False).mean()

Will add a unit test later this week.

[philippjfr]

Nice, presumably we also need to ensure that it actually watches the widget/parameter though. Don't recall how offhand so if you don't see a way let me know and I'll dig into it.

[hoxbro]

I don't know if anything is needed in holoviews, but in hvplot the function hvplot.interactive._find_widgets should be updated with something similar to this.

[hoxbro]

This should now watch the widget/parameter. See:

example.mp4

The code:

import holoviews as hv
import hvplot.pandas  # noqa
import panel as pn
from bokeh.sampledata.autompg import autompg_clean as df

hv.extension("bokeh")

idf = df.interactive()

cylinders = pn.widgets.IntSlider(start=4, end=8, step=2, name="Cylinders")
mfr = pn.widgets.ToggleGroup(
    options=["ford", "chevrolet", "plymouth", "honda", "sbaru", "volkswagen"],
    value=["ford", "chevrolet", "plymouth", "honda", "sbaru", "volkswagen"],
)
xaxis = pn.widgets.RadioButtonGroup(name="X axis", options=["mpg", "yr"])
yaxis = pn.widgets.RadioButtonGroup(name="Y axis", options=["hp", "weight"])

ipipeline = (
    idf[(idf.mfr.isin(mfr)) & (idf.cyl == cylinders)]
    .groupby(["origin", xaxis])[yaxis]
    .mean()
)

pn.Column(ipipeline.hvplot(by="origin"), sizing_mode="stretch_both").servable()

I have noted the linecolor change when changing the yaxis, I don't know if this is a feature or a bug.

[hoxbro]

I have added an unittest. When trying to run the test suite on my computer it stops at holoviews/tests/plotting/bokeh/test_server.py and when I ctrl+c it returns TypeError: cannot create weak reference to 'NoneType' object. This seems to be related to changes made here holoviz/panel#2597.

[jbednar]

This is really cool, @hoxbro ! Does anyone know of other Pandas or Xarray API argument types that remain unsupported? So far it seems like scalars, lists, tuples, and dictionaries are handled. I don't expect there are any arguments that take sets? Any other containers worth adding?

[hoxbro]

This needs to be updated again with dictionary but will wait with an update until a final decision have been made.

[jbednar]

@philippjfr , I'm eager to get this in place; is there anything else needed before merging, other than the docs fix above?

[philippjfr]

Nice, all looks good. Will merge.

[github-actions]

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