Talk:Xiaolin Wu's line algorithm

Any guy can give me Xiaolin Wu's line algorithms code in C#. Than you.

Bug in the C code?

<lang c> } else {

   if ( y2 < y1 ) {
     swap_(x1, x2);
     swap_(y1, y2);
   }
   double gradient = dx / dy;
   double yend = round_(y1);
   double xend = x1 + gradient*(yend - y1);
   double ygap = rfpart_(y1 + 0.5);
   int ypxl1 = yend;
   int xpxl1 = ipart_(xend);
   plot_(xpxl1, ypxl1, rfpart_(xend)*ygap);
   plot_(xpxl1, ypxl1+1, fpart_(xend)*ygap);
   double interx = xend + gradient;

   yend = round_(y2);
   xend = x2 + gradient*(yend - y2);
   ygap = fpart_(y2+0.5);
   int ypxl2 = yend;
   int xpxl2 = ipart_(xend);
   plot_(xpxl2, ypxl2, rfpart_(xend) * ygap);
   plot_(xpxl2, ypxl2 + 1, fpart_(xend) * ygap);

   int y;
   for(y=ypxl1+1; y <= (ypxl2-1); y++) {
     plot_(ipart_(interx), y, rfpart_(interx));
     plot_(ipart_(interx) + 1, y, fpart_(interx));
     interx += gradient;
   }
 }

} </lang>

The start and end pixels here in the else where fabs(dx) <= fabs(dy), should have a secondary plot point +1 in the x-direction, not y-direction. It's correct for the main loop and the first if, but not the two end points, for any drawn line with a slope greater or equal to 1.

So where it says: <lang c>

   plot_(xpxl2, ypxl2, rfpart_(xend) * ygap);
   plot_(xpxl2, ypxl2 + 1, fpart_(xend) * ygap);

</lang> it should say: <lang c>

   plot_(xpxl2, ypxl2, rfpart_(xend) * ygap);
   plot_(xpxl2 + 1, ypxl2, fpart_(xend) * ygap);
</lang>

76.175.220.55 11:20, 3 July 2011 (UTC)

REXX comment

The new REXX entry currently says "Apparently, there may be an error in the definition of the algorithm (which only manifests itself with negative numbers): use of the IPART function should probably be FLOOR"

But, in the sort of context where this line algorithm would be useful that I have seen, screen coordinates are always have been non-negative integers, so in practice this kind of ambiguity should not matter. --Rdm 15:52, 4 June 2012 (UTC)