jeffreysarnoff / anglebetweenvectors.jl Goto Github PK

View Code? Open in Web Editor NEW

28.0 28.0 4.0 123 KB

An accurate and stable calculation of the angle separating two vectors.

License: MIT License

Julia 100.00%

accurate angle floating-point julia n-dimensional vectors

anglebetweenvectors.jl's People

Contributors

Stargazers

Watchers

Forkers

emerali standardgalactic jefffessler mkitti

anglebetweenvectors.jl's Issues

Strange behaviour for points near the origin

Hi, I noticed that angle does not actually do as claimed in the docstrings:

help?> angle
...
If one of the points is at the origin, the result is zero.
...

Specifically:

julia> angle((0.0, 0.0), (1.0, 1.0))
NaN

Also, there is some inconsistent behaviour near the origin:

julia> angle((0.0, 0.001), (1.0, 1.0))
0.7853981633974483

julia> angle((0.001, 0.001), (1.0, 1.0))
1.5700924586837754e-16

julia> angle((0.0001, 0.001), (1.0, 1.0))
0.6857295109062864

julia> angle((0.0001, 0.0001), (1.0, 1.0))
1.5700924586837754e-16

I would expect all of the above to be approximately 0.7853981633974483?

I am using [ec570357] AngleBetweenVectors v0.3.0.

replace unitize by scaling to same magnitude

Maybe I am totally wrong but could this code be even better without divisions:

     
    n1 = norm(point1)
    n2 = norm(point2)

    p1 = n2 * point1
    p2 = n1 * point2

    # p1 and p2 have now same magnitudes

    y = p1 .- p2
    x = p1 .+ p2

    a = 2 * atan(norm(y) , norm(x))  # use atan2(Y,X) result already in [0,pi]