Half-extent (as a fraction of its viewport axis) below which a projected
rect's thin dimension counts as DEGENERATE for projectBoxAreaFraction:
effectively lower-dimensional content (an axis-aligned straight polyline, a
planar dataset viewed edge-on, 1D/2D bounds on a mapped axis) whose
area-product would read ~0 no matter how much of the screen it spans —
permanently pinning it to the coarsest level. 1e-3 ≈ one pixel on a
~1080p viewport: anything rendering thinner than a pixel genuinely reads as
a line, and for a line the faithful "portion of the screen occupied" is its
linear span, not the vanishing area. The fallback is a continuous RAMP over
[0, this] (see the function doc), not a cliff.
Exported so tests pin the ramp against the real constant (and so the
@link references above resolve in TypeDoc).
Half-extent (as a fraction of its viewport axis) below which a projected rect's thin dimension counts as DEGENERATE for projectBoxAreaFraction: effectively lower-dimensional content (an axis-aligned straight polyline, a planar dataset viewed edge-on, 1D/2D bounds on a mapped axis) whose area-product would read ~0 no matter how much of the screen it spans — permanently pinning it to the coarsest level.
1e-3≈ one pixel on a ~1080p viewport: anything rendering thinner than a pixel genuinely reads as a line, and for a line the faithful "portion of the screen occupied" is its linear span, not the vanishing area. The fallback is a continuous RAMP over[0, this](see the function doc), not a cliff.Exported so tests pin the ramp against the real constant (and so the
@linkreferences above resolve in TypeDoc).