Mail Archives: djgpp/1997/08/24/01:34:19
tom burgess wrote:
>
(some silly stuff)
Someone was kind enough to point out that my beautiful algorithm was
not going to work. I had hoped to avoid sqrts. To make amends, I offer
the following excellent link and some code that should work.
http://www.cis.ohio-state.edu/hypertext/faq/usenet/graphics/algorithms-faq/faq.html
From the ACM graphics gems code pool (see above FAQ for link).
/*
A Fast 2D Point-On-Line Test
by Alan Paeth
from "Graphics Gems", Academic Press, 1990
*/
#include "GraphicsGems.h"
int PntOnLine(px,py,qx,qy,tx,ty)
int px, py, qx, qy, tx, ty;
{
/*
* given a line through P:(px,py) Q:(qx,qy) and T:(tx,ty)
* return 0 if T is not on the line through <--P--Q-->
* 1 if T is on the open ray ending at P: <--P
* 2 if T is on the closed interior along: P--Q
* 3 if T is on the open ray beginning at Q: Q-->
*
* Example: consider the line P = (3,2), Q = (17,7). A plot
* of the test points T(x,y) (with 0 mapped onto '.') yields:
*
* 8| . . . . . . . . . . . . . . . . . 3 3
* Y 7| . . . . . . . . . . . . . . 2 2 Q 3 3 Q = 2
* 6| . . . . . . . . . . . 2 2 2 2 2 . . .
* a 5| . . . . . . . . 2 2 2 2 2 2 . . . . .
* x 4| . . . . . 2 2 2 2 2 2 . . . . . . . .
* i 3| . . . 2 2 2 2 2 . . . . . . . . . . .
* s 2| 1 1 P 2 2 . . . . . . . . . . . . . . P = 2
* 1| 1 1 . . . . . . . . . . . . . . . . .
* +--------------------------------------
* 1 2 3 4 5 X-axis 10 15 19
*
* Point-Line distance is normalized with the Infinity Norm
* avoiding square-root code and tightening the test vs the
* Manhattan Norm. All math is done on the field of integers.
* The latter replaces the initial ">= MAX(...)" test with
* "> (ABS(qx-px) + ABS(qy-py))" loosening both inequality
* and norm, yielding a broader target line for selection.
* The tightest test is employed here for best discrimination
* in merging collinear (to grid coordinates) vertex chains
* into a larger, spanning vectors within the Lemming editor.
*/
if ( ABS((qy-py)*(tx-px)-(ty-py)*(qx-px)) >=
(MAX(ABS(qx-px), ABS(qy-py)))) return(0);
if (((qx<px)&&(px<tx)) || ((qy<py)&&(py<ty))) return(1);
if (((tx<px)&&(px<qx)) || ((ty<py)&&(py<qy))) return(1);
if (((px<qx)&&(qx<tx)) || ((py<qy)&&(qy<ty))) return(3);
if (((tx<qx)&&(qx<px)) || ((ty<qy)&&(qy<py))) return(3);
return(2);
}
regards, tom (tburgess AT drao DOT nrc DOT ca)
- Raw text -