OK, with no feedback on this, I have removed the copied comments from
geo_ops.c, reorganized a little, and removed the GPL copyright notice.

Patch attached as unified diff so you can see the line changes easier.
Should this be backpatched to 8.2.X?

---------------------------------------------------------------------------

Bruce Momjian wrote:
> 
> I did some research on this item.  I downloaded the source code to WN from:
> 
>       http://hopf.math.northwestern.edu/index.html
> 
> I could only find the most recent version. wn-2.4.7.  I then looked at
> its image.c file:
> 
>       http://momjian.us/expire/image.c
> 
> I looked at the last two functions in the file and compared it to what
> we have in CVS, particularly the version of the code when it was first
> added to CVS:
> 
>       
> http://developer.postgresql.org/cvsweb.cgi/pgsql/src/backend/utils/adt/geo_ops.c?rev=1.13;content-type=text%2Fplain
> 
> Again, look at the last two functions in the file.
> 
> You will see similarities and differences.  Many of the variable names
> are the same, and there is an identical comment block.  The algorithm
> used is very similar as well, but the style and formatting is different.
> 
> I have updated the code comment in CVS to mention the web site, GPL
> license, and article describing the algorithm.
> 
> Not sure where to go from here.
> 
> ---------------------------------------------------------------------------
> 
> bruce wrote:
> > Tom Lane wrote:
> > > [ redirecting to -hackers, as I see no need for this to be a core issue ]
> > > 
> > > Charles Comiskey <[EMAIL PROTECTED]> writes:
> > > > Hello,
> > > > I've recently looked through the PostgreSQL code and a couple of 
> > > > questions 
> > > > surfaced.  I was hoping someone here may be able to answer them.  Two 
> > > > have 
> > > > links to possible GPL sources and the third is just a contribution 
> > > > question. 
> > > 
> > > > item #1: Does the geo_ops.c file contain GPL code?
> > > > Embedded within the geo_ops.c file is a John Franks copyright statement 
> > > > referring to wn/image.c file from WN Server 1.15.1.  WN Server appears 
> > > > to 
> > > > have been under the GPL license since 0.94 and continues to be offered 
> > > > under the GPL license today.  John's letter to Linux Journal seems to 
> > > > only 
> > > > point the user to his WN Server distribution vs granting any specific 
> > > > license.
> > 
> > The comment is:
> > 
> >     /* poly_contain_pt()
> >      * Test to see if the point is inside the polygon.
> >      * Code adapted from integer-based routines in
> >      *  Wn: A Server for the HTTP
> >      *  File: wn/image.c
> >      *  Version 1.15.1
> >      *  Copyright (C) 1995  <by John Franks>
> >      * (code offered for use by J. Franks in Linux Journal letter.)
> >      */
> > 
> > That term "adapted from" isn't something Thomas would idly type, I
> > think.  I bet it means he looked at John Franks' code and used it as a
> > base for our code.  I am not concerned.
> > 
> > > > Questions:
> > > > 1) Is any John Franks code really in this file?
> > > > 2) Did John provide a separate license for PostgreSQL to license it 
> > > > under 
> > > > the BSD license?
> > > 
> > > This code seems to have been inserted by Tom Lockhart on 1997-07-29
> > > (geo_ops.c rev 1.13).  Tom, any info on the copyright status?
> > > 
> > > > References:
> > > > - 1994 e-mail with GPL reference to WN Server v0.94: 
> > > > http://1997.webhistory.org/www.lists/www-talk.1994q4/1080.html
> > > > - 1995 e-mail from John with GPL license text reference: 
> > > > http://1997.webhistory.org/www.lists/www-talk.1995q1/0482.html
> > > > - WN Server url today: http://hopf.math.northwestern.edu/
> > > > - Link to Linux Journal article: 
> > > > http://www.linuxjournal.com/article/2197
> > > 
> > > 
> 
> -- 
>   Bruce Momjian   [EMAIL PROTECTED]
>   EnterpriseDB    http://www.enterprisedb.com
> 
>   + If your life is a hard drive, Christ can be your backup. +
> 
> ---------------------------(end of broadcast)---------------------------
> TIP 6: explain analyze is your friend

-- 
  Bruce Momjian  <[EMAIL PROTECTED]>          http://momjian.us
  EnterpriseDB                               http://www.enterprisedb.com

  + If your life is a hard drive, Christ can be your backup. +
Index: geo_ops.c
===================================================================
RCS file: /cvsroot/pgsql/src/backend/utils/adt/geo_ops.c,v
retrieving revision 1.95
retrieving revision 1.96
diff -u -r1.95 -r1.96
--- geo_ops.c	27 Feb 2007 23:48:08 -0000	1.95
+++ geo_ops.c	5 Mar 2007 23:29:14 -0000	1.96
@@ -8,7 +8,7 @@
  *
  *
  * IDENTIFICATION
- *	  $PostgreSQL: pgsql/src/backend/utils/adt/geo_ops.c,v 1.95 2007/02/27 23:48:08 tgl Exp $
+ *	  $PostgreSQL: pgsql/src/backend/utils/adt/geo_ops.c,v 1.96 2007/03/05 23:29:14 momjian Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -5063,128 +5063,126 @@
  ***********************************************************************/
 
 /*
- *	Test to see if the point is inside the polygon.
- *	Code adapted from integer-based routines in WN: A Server for the HTTP
+ *	Test to see if the point is inside the polygon, returns 1/0, or 2 if
+ *	the point is on the polygon.
+ *	Code adapted but not copied from integer-based routines in WN: A
+ *	Server for the HTTP
  *	version 1.15.1, file wn/image.c
- *	GPL Copyright (C) 1995 by John Franks
  *	http://hopf.math.northwestern.edu/index.html
  *	Description of algorithm:  http://www.linuxjournal.com/article/2197
+ *							   http://www.linuxjournal.com/article/2029
  */
 
-#define HIT_IT INT_MAX
+#define POINT_ON_POLYGON INT_MAX
 
 static int
 point_inside(Point *p, int npts, Point *plist)
 {
 	double		x0,
 				y0;
-	double		px,
-				py;
-	int			i;
+	double		prev_x,
+				prev_y;
+	int			i = 0;
 	double		x,
 				y;
-	int			cross,
-				crossnum;
-
-/*
- * We calculate crossnum, which is twice the crossing number of a
- * ray from the origin parallel to the positive X axis.
- * A coordinate change is made to move the test point to the origin.
- * Then the function lseg_crossing() is called to calculate the crossnum of
- * one segment of the translated polygon with the ray which is the
- * positive X-axis.
- */
+	int			cross, total_cross = 0;
 
-	crossnum = 0;
-	i = 0;
 	if (npts <= 0)
 		return 0;
 
+	/* compute first polygon point relative to single point */
 	x0 = plist[0].x - p->x;
 	y0 = plist[0].y - p->y;
 
-	px = x0;
-	py = y0;
+	prev_x = x0;
+	prev_y = y0;
+	/* loop over polygon points and aggregate total_cross */
 	for (i = 1; i < npts; i++)
 	{
+		/* compute next polygon point relative to single point */
 		x = plist[i].x - p->x;
 		y = plist[i].y - p->y;
 
-		if ((cross = lseg_crossing(x, y, px, py)) == HIT_IT)
+		/* compute previous to current point crossing */
+		if ((cross = lseg_crossing(x, y, prev_x, prev_y)) == POINT_ON_POLYGON)
 			return 2;
-		crossnum += cross;
-
-		px = x;
-		py = y;
+		total_cross += cross;
+		
+		prev_x = x;
+		prev_y = y;
 	}
-	if ((cross = lseg_crossing(x0, y0, px, py)) == HIT_IT)
+
+	/* now do the first point */
+	if ((cross = lseg_crossing(x0, y0, prev_x, prev_y)) == POINT_ON_POLYGON)
 		return 2;
-	crossnum += cross;
-	if (crossnum != 0)
+	total_cross += cross;
+
+	if (total_cross != 0)
 		return 1;
 	return 0;
 }
 
 
 /* lseg_crossing()
- * The function lseg_crossing() returns +2, or -2 if the segment from (x,y)
- * to previous (x,y) crosses the positive X-axis positively or negatively.
- * It returns +1 or -1 if one endpoint is on this ray, or 0 if both are.
- * It returns 0 if the ray and the segment don't intersect.
- * It returns HIT_IT if the segment contains (0,0)
+ * Returns +/-2 if line segment crosses the positive X-axis in a +/- direction.
+ * Returns +/-1 if one point is on the positive X-axis.
+ * Returns 0 if both points are on the positive X-axis, or there is no crossing.
+ * Returns POINT_ON_POLYGON if the segment contains (0,0).
+ * Wow, that is one confusing API, but it is used above, and when summed,
+ * can tell is if a point is in a polygon.
  */
 
 static int
-lseg_crossing(double x, double y, double px, double py)
+lseg_crossing(double x, double y, double prev_x, double prev_y)
 {
 	double		z;
-	int			sgn;
-
-	/* If (px,py) = (0,0) and not first call we have already sent HIT_IT */
+	int			y_sign;
 
 	if (FPzero(y))
-	{
-		if (FPzero(x))
-		{
-			return HIT_IT;
-
-		}
+	{	/* y == 0, on X axis */
+		if (FPzero(x))	/* (x,y) is (0,0)? */
+			return POINT_ON_POLYGON;
 		else if (FPgt(x, 0))
-		{
-			if (FPzero(py))
-				return FPgt(px, 0) ? 0 : HIT_IT;
-			return FPlt(py, 0) ? 1 : -1;
-
+		{	/* x > 0 */
+			if (FPzero(prev_y))	/* y and prev_y are zero */
+				/* prev_x > 0? */
+				return FPgt(prev_x, 0) ? 0 : POINT_ON_POLYGON;
+			return FPlt(prev_y, 0) ? 1 : -1;
 		}
 		else
-		{						/* x < 0 */
-			if (FPzero(py))
-				return FPlt(px, 0) ? 0 : HIT_IT;
+		{	/* x < 0, x not on positive X axis */
+			if (FPzero(prev_y))
+				/* prev_x < 0? */
+				return FPlt(prev_x, 0) ? 0 : POINT_ON_POLYGON;
 			return 0;
 		}
 	}
-
-	/* Now we know y != 0;	set sgn to sign of y */
-	sgn = (FPgt(y, 0) ? 1 : -1);
-	if (FPzero(py))
-		return FPlt(px, 0) ? 0 : sgn;
-
-	if (FPgt((sgn * py), 0))
-	{							/* y and py have same sign */
-		return 0;
-
-	}
 	else
-	{							/* y and py have opposite signs */
-		if (FPge(x, 0) && FPgt(px, 0))
-			return 2 * sgn;
-		if (FPlt(x, 0) && FPle(px, 0))
-			return 0;
-
-		z = (x - px) * y - (y - py) * x;
-		if (FPzero(z))
-			return HIT_IT;
-		return FPgt((sgn * z), 0) ? 0 : 2 * sgn;
+	{	/* y != 0 */
+		/* compute y crossing direction from previous point */
+		y_sign = FPgt(y, 0) ? 1 : -1;
+
+		if (FPzero(prev_y))
+			/* previous point was on X axis, so new point is either off or on */
+			return FPlt(prev_x, 0) ? 0 : y_sign;
+		else if (FPgt(y_sign * prev_y, 0))
+			/* both above or below X axis */
+			return 0;	/* same sign */
+		else
+		{	/* y and prev_y cross X-axis */
+			if (FPge(x, 0) && FPgt(prev_x, 0))
+				/* both non-negative so cross positive X-axis */
+				return 2 * y_sign;
+			if (FPlt(x, 0) && FPle(prev_x, 0))
+				/* both non-positive so do not cross positive X-axis */
+				return 0;
+
+			/* x and y cross axises, see URL above point_inside() */
+			z = (x - prev_x) * y - (y - prev_y) * x;
+			if (FPzero(z))
+				return POINT_ON_POLYGON;
+			return FPgt((y_sign * z), 0) ? 0 : 2 * y_sign;
+		}
 	}
 }
 
---------------------------(end of broadcast)---------------------------
TIP 9: In versions below 8.0, the planner will ignore your desire to
       choose an index scan if your joining column's datatypes do not
       match

Reply via email to