spread_plot traces carry no hovertemplate — turning on spread loses all hover detail

[jepegit]

Found while wiring group-averaged summary plots in cellpy-simple-gui (against cellpy 2.1.2).

What happens

On the collected summary path, hover is informative — until spread=True, at which point every trace has hovertemplate = None.

Same collection, same columns, three renderings:

mode	hovertemplate on the first trace
per-cell	cellpy<br>group=1<br>sub_group=1<br>variable=coulombic_efficiency<br>Cycle (n.)=%{x}<br>value=%{y}
group_it=True	group=1<br>variable=coulombic_efficiency<br>Cycle (n.)=%{x}<br>mean=%{y}
group_it=True, spread=True	None — on all 18 traces, mean lines included

Why

summary_plotter goes through plotly.express, which attaches hover from the frame. spread_plot builds traces directly with go.Scatter (make_subplots + add_trace) and never sets hovertemplate or hoverinfo, so the figure ships with Plotly's bare default.

The docstring already flags the function as experimental — this looks like one of the gaps that implies, rather than a deliberate choice.

Why it matters for an app

Spread is the mode where hover is most useful: the band hides the individual cells, so the tooltip is the only way to read a value. Losing group, variable, cycle and value at exactly that moment is a visible regression to a user toggling one checkbox.

There is also a smaller issue underneath: the Upper Bound / Lower Bound traces are hoverable. They are construction artefacts (mean ± std), so a tooltip on them reads like a measurement that was never taken.

Suggestions

1. Set hovertemplate on the mean traces in spread_plot, matching the group_it path's fields (group, variable, x, y) — ideally with the spread itself, since mean and std are both in hand at that point.
2. Set hoverinfo="skip" on the band edges so only the mean is hoverable.
3. Longer term, the docstring says plotly.express is to be replaced by this path for all summary plots — worth making hover parity part of that, since it is currently the main functional difference.

Workaround (app)

cellpy-simple-gui reconstructs hover after the fact: it groups traces by subplot, takes the variable from the y-axis title and the group from the mean trace's name, derives std from the upper-bound trace, and skips hover on the bounds. It works, but it depends on trace naming ("Upper Bound <group>") and on emission order (mean, upper, lower) — both internal details that would be better not to rely on.

Environment

cellpy 2.1.2, Python 3.13, Windows.