Re: Re: gdk_draw_polygon +stipple fill has problem

[Melvin Hadasht <melvin hadasht free fr>]

Hi,

> The width and height calculations are wrong. Obvioously, one cannot
> calculate them before knowing what the actual minimum x and y will
> be. I.e. two separate loops are needed:
> 
>   for (i = 0; i < npoints; i++)
>     {
>       bounds.x = MIN (bounds.x, points[i].x);
>       bounds.y = MIN (bounds.y, points[i].y);
>     }
> 
>   for (i = 0; i < npoints; i++)
>     {
>       bounds.width = MAX (bounds.width, points[i].x - bounds.x);
>       bounds.height = MAX (bounds.height, points[i].y - bounds.y);
>     }

Two separate loops are not needed if we introduce two temp variables
to hold the max coordinates:

/* Initial valid values for the bounds taken from the first point*/
bounds.x = tmpmax.x = points[0].x;
bounds.y = tmpmax.y = points[0].y;

/* Find min and max by iterating through the rest of the points*/
for (i = 1; i < npoints; i++) /* Note: start at 1 not 0 */
  {
    bounds.x = MIN (bounds.x, points[i].x);
    bounds.y = MIN (bounds.y, points[i].y);
    tmpmax.x = MAX(tmpmax.x, points[i].x);
    tmpmax.y = MAX(tmpmax.y, points[i].y);
  }

/* Compute extents */
bounds.width = tmpmax.x - bounds.x
bounds.height = tmpmax.y - bounds.y

(In fact, we can use bounds.width/height as temp variables, but then
the code is less readable because bounds.width/height will temporarily
contains the max coordinates instead of the width/height)

This gets rid of the second loop that in addition contained
two substractions.

Cheers

-- 
Melvin Hadasht