pub const SMOOTH_CORNER_SHARE: f32 = 0.25;Expand description
The largest share of a Maurer walk’s vertices that may be corners before
maurer_rose_pieces declines to fit it — is this a curve at all?
The two families a Maurer walk holds are not near each other on this
number, which is the whole reason a threshold can exist. At the shipped
chord-web steps (d = 29 to 71) more than 85 % of the walk’s vertices
turn past biarc::CORNER_TURN; at d = 2, the smooth rose, under 15 % do —
and those few are the genuine cusps where the radius crosses zero and the
figure passes through the origin.
A per-corner rule alone would not do, and that is a measurement rather
than a worry: a d = 29 web is ~90 % corners, so the fit would turn its
remaining tenth into arcs and redraw a figure whose chords are the figure.
The decision has to be about the walk as a whole.