2D charts

All 2D traces share the same conventions: with color=None, palette slots are assigned in a fixed, color-vision-safe order; name puts the series in the legend; non-finite values (NaN) break lines into separate runs instead of drawing through the gap. Every add_* returns a trace handle for plot.extend(handle, xs, ys) and plot.set_visible(handle, ...) — see Streaming data. Coordinates accept lists or numpy float32/float64 arrays (arrays are read in one bulk copy).

Line and scatter

plot.add_line(xs, ys, color=None, width=2.0, name=None, axis="y")
plot.add_scatter(xs, ys, color=None, size=2.5, name=None, axis="y")

Bars

plot.add_bar(xs, heights, color=None, name=None, axis="y")

Bars rise (or fall) from the zero baseline. Bar width is derived from the smallest gap between x positions, so adjacent bars keep a visible gap — which also means extend-ing a bar whose x narrows that gap re-flows the width of every bar in the trace.

Stacked bars are a painter's trick: draw cumulative totals first, then the shorter component bars on top:

totals = [s + w + h for s, w, h in zip(solar, wind, hydro)]
plot.add_bar(months, totals, name="hydro")                        # tallest first
plot.add_bar(months, [s + w for s, w in zip(solar, wind)], name="wind")
plot.add_bar(months, solar, name="solar")

Secondary y-axes

axis="y2" or "y3" binds a series to an independent right-hand axis with its own autoscale and tick column, labels tinted to the series color. The grid always belongs to the primary axis.

plot.add_line(xs, mw, name="production")
plot.add_line(xs, price, axis="y2", name="price")

Hover crosshair

2D plots support a hover crosshair: a vertical guide snapped to the nearest sample x, a marker on every series sampled there, and a value readout box. The Textual widget wires it to the mouse automatically; in a raw loop, set it yourself in framebuffer pixels:

changed = plot.set_hover2d(px)   # None clears; returns True when state changed

Series match by exact sample x — series on a shared grid all get a row, while series on their own grids only show where they truly have a sample. Nothing is interpolated silently.