dot draws one dot per row. Reach for it where the position of a point is the
thing worth seeing: where a line turns provisional, an outlier worth naming, a
handful of readings too sparse to join up. Where the number is what matters,
autoPositionedLabel writes it
at the point instead.
radius sizes the dot in screen pixels, so it keeps its size as the scene around
it is scaled. Dots sitting on a line of their own color disappear into it, which
is what ringColor is for: set it to the surface the chart sits on and each dot
is punched out of whatever it lands on.
Dots over part of a line say "this stretch is different" without a legend entry to explain it. Here the last few readings are marked as provisional, the dot placed where the settled data ends:
decorative wraps any mark so it draws and contributes to the scales — the axis
still covers what it shows — but stops owning chart points. That takes it out of
the focused group, and so out of the tooltip, off the keyboard cursor's path, and
out of what onSelect and onFocusChange report.
Reach for it whenever a mark restates something already on the plot: a marker on
a line, a threshold, a target. Without it the annotation arrives in the tooltip
as a series of its own, named after whatever id the mark was given — a row
reading dot-2 beside the real one.
It is not only for dots. Any mark can be decorative, and a mark that genuinely carries its own reading should not be: the tooltip is where a reader looks to find out what a shape means, so taking a real series out of it hides the number rather than tidying the chart.