[Gambas-user] gb.gsl: pseudo-complex Vectors make different cross product

[Benoît Minisini]

Le 11/02/2014 21:15, Tobias Boege a écrit :
> Hi Benoit,
>
> a strange thing this time. Using gb.gsl, we noticed that there is a way to
> compute the cross product of two Vector objects in "complex" mode - but
> without any complex numbers in them - which yields a different result than
> other methods to compute it. Attached is a project that shows what I mean.
>
> Note that everything is fine if we don't force the Vector to be complex and
> also everything is fine if we don't insist on computing the components in
> separate Float variables, as said in the comments.
>
> Oh, and for the record, the Y component is wrong. It should be -18 every
> time but using that special method above, we get +18. The vectors are
> immediately taken from an application, seemingly nothing special about them.
>
> I also noticed that wrapping each product of components into a CFloat() call
> yields the correct result again... I tried that because even the debugger
> didn't seem to get the type of hA[iIndex] figured out...
>
> What may be going on there?
>
> Regards,
> Tobi
>

Arf... This is because Complex to Float conversion takes the modulus 
value, so -18 as a complex becomes +18.

Mmm... that was a weird choice. We can even call that a bug.

I must leave now, but I think I will replace that by taking the real 
part of the complex number if the imaginary part is zero, and raising a 
conversion error instead.

Regards,

-- 
Benoît Minisini