FigureWidget update issues after single bin histogram.

Running into an (admittedly quite edgy) issue updating a plotly graph_object Histogram in a FigureWidget.

The issue only seems to arise when using barmode='overlay' and seems to result from updating .data[0].x to a degenerative
(single bin) array (i.e. all elements equal).

Running in Jupyterlab the issue can be reproduced by:

import plotly.graph_objects as go
import numpy as np
hist_wgt = go.FigureWidget([go.Histogram(x=np.random.random(200), histnorm='probability', opacity=0.7)],
                          layout=dict(barmode='overlay'))
display(hist_wgt)

which produces a nicely binned histogram e.g.

and

print("xrange:", hist_wgt.layout.xaxis.range)
print("nbinsx:", hist_wgt.data[0].nbinsx)
print("yrange:",hist_wgt.layout.yaxis.range)
print("start:", hist_wgt.data[0].xbins.start)
print("end:", hist_wgt.data[0].xbins.end)
print("size:", hist_wgt.data[0].xbins.size)

xrange: (6.938893903907228e-17, 0.9999999999999999)
nbinsx: 0
yrange: (0, 0.1736842105263158)
start: 0
end: 1
size: 0.1

However, if we update hist_wgt.data[0].x with a (terrible) uniform array:

hist_wgt.data[0].x = np.ones(200)

the histogram becomes a single bin (as expected):

and the layout and bin info:

xrange: (0.5, 1.5)
nbinsx: 0
yrange: (0, 1.0526315789473684)
start: 0.5
end: 1.5
size: 1

The issue now happens when we try to update hist_wgt.data[0].x, but with with a more sensible array:

hist_wgt.data[0].x = np.random.random(200) * 1e6

it seems the widget layout is not re-evaluated or updated for this new x:

although nbinsx does get set to None (unsure if this is significant):

xrange: (0.5, 1.5)
nbinsx: None
yrange: (0, 1.0526315789473684)
start: 0.5
end: 1.5
size: 1

My guess is that the ambiguous nature of bin-width where the data range is 0 means that something is being set for the widget layout for that special case -- and then this setting overrides any later auto-range operations?

I am happy to admit that trying to plot a single bin histogram is slightly pointless but I want the histogram to be updatable for user selected variables, some of which could could collapse to a single value. I have tried to find an option to force the layout update through but haven't stumble upon the right combination of calls.

Start with the FigureWidget and go.Histogram reproduction in JupyterLab, first confirming the update sequence from varied data to a single-bin array and back. Trace how changing FigureWidget.data[0].x updates histogram bin metadata and axis ranges; done means the layout and bin information are recalculated for the new data rather than retaining the single-bin values.