AutoMovie API
    Preparing search index...

    Namespace AutoMovieSpline

    The interpolating spline under tension on caller-owned arrays: one shared tension chosen per profile by a sampled overshoot policy, the coefficients solved in place and evaluation without allocation.

    Segment k of width h between knots x_k and x_k+1 is s = h^2 (M_k sinh(tau (1 - t)) + M_k+1 sinh(tau t)) / (tau^2 sinh tau) + (y_k - M_k h^2 / tau^2)(1 - t) + (y_k+1 - M_k+1 h^2 / tau^2) t at t = (x - x_k) / h, the spline under tension of Schweikert (1966, J. Math. Phys. 45:312-317) as Cline wrote it (1974, Comm. ACM 17(4):218-220) with tension sigma = tau / h; M is its second derivative at the knots, so the curve is C2 for every tension, a cubic spline at tension 0 and the piecewise-linear interpolant in the limit. An open profile has natural ends, second derivative 0, and holds its end values outside its knots; a periodic one closes across its period.

    fit first tries tension 0. It checks fifteen interior locations per segment against the knot range expanded by a tenth of the largest value step within two segments. If that scan fails, it doubles tension from 1/2, then bisects the rejected/accepted bracket ten times, up to 64. The scan does not bound extrema between probes, and reaching 64 returns that tension even if the scan still fails. Positive values need not stay positive between knots. Unlike a monotone cubic, the spline preserves curvature continuity at interior knots; its constant open-end extension need not preserve the endpoint's first derivative.

    BOUND
    evaluate
    fit