pub fn point_to_f64_from_lam(lam: &Lam) -> Option<[f64; 3]>Expand description
f64 coordinates from an ALREADY-CACHED Lam (the interner’s per-Vid I512
lambda), skipping the from-scratch I1024 recompute point_to_f64 does. The
cached lambda is stored in one homogeneous COARSE convention (fine-scale points
fold their 2^20 scale ratio into d), so real = λ/(d·COARSE) for either grid
and the result is bit-identical to point_to_f64. None only when the cached
lambda was absent (overflow) — the caller then falls back to point_to_f64.