OpenEuphoria

Re: Fw: Distance between 2 points

• Posted by Ralf Nieuwenhuijsen <nieuwen at XS4ALL.NL> Sep 30, 1998
• 567 views

>try this instead. 1 function call vs 3, should be worlds

Yes, I could inline it also. I just loven another thread on the sum ()
function, it will add atoms it can find.
I could just have used the '+' operator, since Patrat was prolly refering to
a 2D field.
However, it is cleaner to use recursive functions. And to be dimension free,
who knows how many dimensions we are gonna explore in the future ? Dont want
no Y2K typ of problem with my software when we reach 4K  Just kidding, I
dunno why I do this things... impulses maybe ?

>faster for atoms and sequences both, i think...

For both atoms and sequences ? Actually I think with atoms, it will should
return the difference. pos (a - b)
That is also the problem with your function here, you forget to substract.

>   function distance(object a, object b)
>      return sqrt( (a*a)+(b*b) )
>   end function

I think you meant this:

-- Assuming 2D world:

function distance (sequence a, sequence b)
    return sqrt ( ((a[1] - b[1]) * (a[1] - b[1])) + ((a[2] - b[2]) * (a[2] -
b[2])) )
end function

-- Assuming 3D world:

function distance (sequence a, sequence b)
    return sqrt ( ((a[1] - b[1]) * (a[1] - b[1])) + ((a[2] - b[2]) * (a[2] -
b[2])) + ((a[3] - b[3]) * (a[3] - b[3])) )
end function

-- Assuming 1D world

function distance (atom a, atom b)
    return sqrt ( (a - b) * (a - b))    -- in other words: the positive
difference between 'a' and 'b'
end function

>and something that should be benchmarked against the
>above 2 distance() functions as well should be:
>   function distance(object a, object b)
>      return power( (a*a)+(b*b),0.5 )
>   end function

It might be less precize, because the result will always be a float because
you use .5
I wonder when Robert is gonna make calculations with floats convert back to
machine integers when possible.

>to see if power is faster than sqrt.


Something I dont expect. First of all, sqrt takes less arguments, and thus
needs less conversion. (if the argument passed is an float.. use the
sqrt_float routine, other wise the sqrt_sequence or the sqrt_integer
routine) .. for power I suspect there are internally 9 different routines.
(Robert, if i'm wrong please say so .. )

Off course a simple benchmark would tell wouldn't it ?

And now here's my fastest function I could come up with (allowing a number
of dimension ranging from 0 to eternity)

function distance (object a, object b)
    if atom (a) and atom(b) then
        if a > b then
            return a-b
        else
            return b-a
        end if
    end if
    a = power(a - b,2)
    b = a[1]
    for index = 2 to length(a) do
        b = b + a[index]
    end for
    return sqrt (b)
end function

Now, how does this benchmark ?

Ralf