http://qs321.pair.com?node_id=132469


in reply to Math::Vector - request for comments

A few points on implementation:

  1. The sub orthogonal is not very stable. Floating point numbers should be compared in terms of a small tolerance. A proper choice for two vectors could be pseudocoded as magnitude(v1-v2) < $EPS * sqrt(magnitude(v1)**2 + magnitude(v2)**2); $EPS is epsilon, the builtin precision of floats.
  2. Sub angle loses precision for angles near integer multiples of pi. An implementation in terms of atan2 would be better, extracting sin of the angle from the cross product.
  3. You may prefer to simply return the zero vector from normalize, instead of carping out.
  4. You'll find some help in the Math::Trig module. For high performance, take a look at Math::GSL and Math::Pari. If you decide to follow Masem's suggestion to generalize, PDL is another high performance library specializing in arrays of values.
I implemented some of the sexier parts of this stuff in the snippet Cartesian 3-Vectors.

Update: Added #4. U2: The pseudocode in #1 is for equality ( == operator). Other comparisons are similarly made.

After Compline,
Zaxo

Replies are listed 'Best First'.
Re: Re: Math::Vector - request for comments
by count0 (Friar) on Dec 17, 2001 at 09:24 UTC
    Floating point numbers should be compared in terms of a small tolerance.
    That is a wonderful point! Many thanks. I have been so used to Perl taking care of the nitty gritty numeric stuff, that I didn't even think of that.

    An implementation in terms of atan2 would be better...
    I had originally used atan2(), simply for the lack of an inverse-cosine function.. until I was pointed to Math::Trig and told to use it ;) So i just haphazardly switched, never considering accuracy issues.

    Thanks a lot for the pointers. They're greatly appreciated =)
[reply]

In Section Code Catacombs