postgis/postgis_estimate.c

/**********************************************************************
 * $Id$
 *
 * PostGIS - Spatial Types for PostgreSQL
 * http://postgis.refractions.net
 * Copyright 2001-2003 Refractions Research Inc.
 *
 * This is free software; you can redistribute and/or modify it under
 * the terms of hte GNU General Public Licence. See the COPYING file.
 * 
 **********************************************************************
 * $Log$
 * Revision 1.19  2004/03/09 00:21:02  strk
 * Removed useless code blocks in histogram builder
 *
 * Revision 1.18  2004/03/09 00:09:56  strk
 * estimator applies a gain of AOI/cell_area on each cell it intersects (reverted to previous behaviour)
 *
 * Revision 1.17  2004/03/04 13:50:45  strk
 * postgis_gist_sel(): added warnings if search_box goes outside of histogram grid
 *
 * Revision 1.16  2004/03/04 09:44:57  strk
 * The selectivity estimator does add the full value of each cell it overlaps,
 * regardless of the actual overlapping area. Final gain is not applied
 * (formerly 1 / minimun between average feature cells occupation and
 *  search_box cells occupation)
 *
 * Revision 1.15  2004/03/03 21:59:48  strk
 * added check to keep selectivity value in the range of validity (suggested by m.cave)
 *
 * Revision 1.14  2004/03/01 16:02:41  strk
 * histogram's boxesPerSide computed as a function of the column's statistic target
 *
 * Revision 1.13  2004/02/29 21:53:42  strk
 * bug fix in postgis_gist_sel (for PG75): SysCache is not released if not acquired
 *
 * Revision 1.12  2004/02/26 16:42:59  strk
 * Fixed bugs reported by Mark Cave-Ayland <m.cave-ayland@webbased.co.uk>.
 * Re-introduced previously removed estimate value incrementation by
 * the fractional part of each of the cells' value computed as the fraction
 * of overlapping area.
 *
 * Revision 1.11  2004/02/25 12:00:32  strk
 * Added handling for point features in histogram creation (add 1 instead of AOI/cell_area when AOI is 0).
 * Fixed a wrong cast of BOX3D to BOX (called the convertion func).
 * Added some comments and an implementation on how to change evaluation
 * based on the average feature and search box cells occupation.
 *
 * Revision 1.10  2004/02/25 00:46:26  strk
 * initial version of && selectivity estimation for PG75
 *
 * Revision 1.9  2004/02/23 21:59:16  strk
 * geometry analyzer builds the histogram
 *
 * Revision 1.8  2004/02/23 12:18:55  strk
 * added skeleton functions for pg75 stats integration
 *
 * Revision 1.7  2003/11/11 10:14:57  strk
 * Added support for PG74
 *
 * Revision 1.6  2003/07/25 17:08:37  pramsey
 * Moved Cygwin endian define out of source files into postgis.h common
 * header file.
 *
 * Revision 1.5  2003/07/01 18:30:55  pramsey
 * Added CVS revision headers.
 *
 *
 **********************************************************************/

 // If you're modifiying this file you should read the postgis mail list as it has
 // detailed descriptions of whats happening here and why.

#include "postgres.h"

#include <math.h>
#include <float.h>
#include <string.h>
#include <stdio.h>
#include <errno.h>

#include "access/gist.h"
#include "access/itup.h"
#include "access/rtree.h"

#include "fmgr.h"

#include "postgis.h"
#include "utils/elog.h"

#define SHOW_DIGS_DOUBLE 15
#define MAX_DIGS_DOUBLE (SHOW_DIGS_DOUBLE + 6 + 1 + 3 +1)

//--------------------------------------------------------------------------

#include "access/heapam.h"
#include "catalog/catname.h"
#include "catalog/pg_operator.h"
#include "catalog/pg_proc.h"
#include "catalog/pg_statistic.h"
#include "catalog/pg_type.h"
#include "mb/pg_wchar.h"
#include "nodes/makefuncs.h"
#include "optimizer/clauses.h"
#include "optimizer/cost.h"
#include "optimizer/pathnode.h"
#include "optimizer/plancat.h"
#include "optimizer/prep.h"
#include "parser/parse_func.h"
#include "parser/parse_oper.h"
#include "parser/parsetree.h"
#include "utils/builtins.h"
#include "utils/date.h"
#include "utils/int8.h"
#include "utils/lsyscache.h"
#include "utils/selfuncs.h"
#include "utils/syscache.h"



#include "executor/spi.h"

#if USE_VERSION >= 75

#include "commands/vacuum.h"

/*
 * 	Assign a number to the postgis statistics kind
 *
 * 	tgl suggested:
 *
 * 	1-100:	reserved for assignment by the core Postgres project
 * 	100-199: reserved for assignment by PostGIS
 * 	200-9999: reserved for other globally-known stats kinds
 * 	10000-32767: reserved for private site-local use
 *
 */
#define STATISTIC_KIND_GEOMETRY 100

#define DEBUG_GEOMETRY_STATS 1

/*
 * Default geometry selectivity factor
 */
#define DEFAULT_GEOMETRY_SEL 0.000005 

typedef struct GEOM_STATS_T
{
	//boxesPerSide * boxesPerSide = total boxes in grid
	float4 boxesPerSide;
	
	// average bounding box area of not-null features 
	float4 avgFeatureArea;

	// average number of histogram cells
	// covered by the sample not-null features
	float4 avgFeatureCells;

	// BOX of area
	float4 xmin,ymin, xmax, ymax;
	// variable length # of floats for histogram
	float4 value[1];
} GEOM_STATS;

#endif


//estimate_histogram2d(histogram2d, box)
// returns a % estimate of the # of features that will be returned by that box query
//
//For each grid cell that intersects the query box
//	  Calculate area of intersection (AOI)
//    IF AOI < avgFeatureArea THEN set AOI = avgFeatureArea
//    SUM AOI/area-of-cell*value-of-cell
//
// change : instead of avgFeatureArea, use avgFeatureArea or 10% of a grid cell (whichever is smaller)

PG_FUNCTION_INFO_V1(estimate_histogram2d);
Datum estimate_histogram2d(PG_FUNCTION_ARGS)
{
	HISTOGRAM2D   *histo = (HISTOGRAM2D *)  PG_DETOAST_DATUM(PG_GETARG_DATUM(0));
	BOX			  *box = (BOX *) PG_GETARG_POINTER(1);
	double box_area;
	int			x_idx_min, x_idx_max, y_idx_min, y_idx_max;
	double intersect_x, intersect_y, AOI;
	int x,y;
	double xmin,ymin,xmax,ymax;
	int32 result_sum;
	double cell_area;
	int total,t;
	double 	avg_feature_size;



	result_sum = 0;
    xmin = histo->xmin;
    ymin = histo->ymin;
    xmax = histo->xmax;
    ymax = histo->ymax;

			cell_area = ( (xmax-xmin)*(ymax-ymin)/(histo->boxesPerSide*histo->boxesPerSide) );

			avg_feature_size = histo->avgFeatureArea;
			if ( avg_feature_size > cell_area*0.1)
			{
				avg_feature_size = cell_area*0.1;
			}


//elog(NOTICE,"start estimate_histogram2d: ");
//elog(NOTICE,"box is : (%.15g,%.15g to %.15g,%.15g)",box->low.x,box->low.y, box->high.x, box->high.y);

					box_area = (box->high.x-box->low.x)*(box->high.y-box->low.y);

					if (box_area<0)
					    box_area =0;  // for precision!

					//check to see which boxes this intersects
					x_idx_min = (box->low.x-xmin)/(xmax-xmin)*histo->boxesPerSide;
					if (x_idx_min <0)
						x_idx_min = 0;
					if (x_idx_min >= histo->boxesPerSide)
						x_idx_min = histo->boxesPerSide-1;
					y_idx_min = (box->low.y-ymin)/(ymax-ymin)*histo->boxesPerSide;
					if (y_idx_min <0)
						y_idx_min = 0;
					if (y_idx_min >= histo->boxesPerSide)
						y_idx_min = histo->boxesPerSide-1;

					x_idx_max = (box->high.x-xmin)/(xmax-xmin)*histo->boxesPerSide;
					if (x_idx_max <0)
						x_idx_max = 0;
					if (x_idx_max >= histo->boxesPerSide)
						x_idx_max = histo->boxesPerSide-1;
					y_idx_max = (box->high.y-ymin)/(ymax-ymin)*histo->boxesPerSide ;
					if (y_idx_max <0)
						y_idx_max = 0;
					if (y_idx_max >= histo->boxesPerSide)
						y_idx_max = histo->boxesPerSide-1;

					//the {x,y}_idx_{min,max} define the grid squares that the box intersects


//elog(NOTICE,"        search is in x: %i to %i   y: %i to %i",x_idx_min, x_idx_max, y_idx_min,y_idx_max);
					for (y= y_idx_min; y<=y_idx_max;y++)
					{
						for (x=x_idx_min;x<= x_idx_max;x++)
						{
								intersect_x = min(box->high.x, xmin+ (x+1) * (xmax-xmin)/histo->boxesPerSide ) -
											max(box->low.x, xmin+ x*(xmax-xmin)/histo->boxesPerSide ) ;
								intersect_y = min(box->high.y, ymin+ (y+1) * (ymax-ymin)/histo->boxesPerSide ) -
											max(box->low.y, ymin+ y*(ymax-ymin)/histo->boxesPerSide ) ;

									// for a point, intersect_x=0, intersect_y=0, box_area =0
//elog(NOTICE,"x=%i,y=%i, intersect_x= %.15g, intersect_y = %.15g",x,y,intersect_x,intersect_y);
								if ( (intersect_x>=0) && (intersect_y>=0) )
								{
									AOI = intersect_x*intersect_y;
									if (AOI< avg_feature_size)
											AOI = avg_feature_size;
									result_sum += AOI/cell_area* histo->value[x+y*histo->boxesPerSide];
								}
						}
					}
			total = 0;
			for(t=0;t<histo->boxesPerSide*histo->boxesPerSide;t++)
			{
				total+=histo->value[t];
			}

			if ( (histo->avgFeatureArea <=0) && (box_area <=0) )
				PG_RETURN_FLOAT8(1.0/((double)(total)));
			else
				PG_RETURN_FLOAT8(result_sum/((double)total));

}

//explode_histogram2d(histogram2d, tablename::text)
// executes CREATE TABLE tablename (the_geom geometry, id int, hits int, percent float)
// then populates it
//  DOES NOT UPDATE GEOMETRY_COLUMNS
PG_FUNCTION_INFO_V1(explode_histogram2d);
Datum explode_histogram2d(PG_FUNCTION_ARGS)
{
	HISTOGRAM2D   *histo = (HISTOGRAM2D *)  PG_DETOAST_DATUM(PG_GETARG_DATUM(0));
	char	*tablename;
	char	sql[1000];
	char	geom[1000];
	int t;
    int total;
	double cellx,celly;
	int x,y;
	int SPIcode;


	cellx = (histo->xmax-histo->xmin)/histo->boxesPerSide;
	celly = (histo->ymax-histo->ymin)/histo->boxesPerSide;

	tablename  = DatumGetCString(DirectFunctionCall1(textout,
	                PointerGetDatum(PG_GETARG_DATUM(1))));

	total = 0;
	for(t=0;t<histo->boxesPerSide*histo->boxesPerSide;t++)
	{
		total+=histo->value[t];
	}
	if (total==0)
		total=1;

		SPIcode = SPI_connect();
		if (SPIcode  != SPI_OK_CONNECT)
		{
			elog(ERROR,"build_histogram2d: couldnt open a connection to SPI");
			PG_RETURN_NULL() ;

		}

		sprintf(sql,"CREATE TABLE %s (the_geom geometry, id int, hits int, percent float)",tablename);

		SPIcode = SPI_exec(sql, 2147483640 ); // max signed int32

			if (SPIcode  != SPI_OK_UTILITY )
			{
					elog(ERROR,"explode_histogram2d: couldnt create table");
					PG_RETURN_NULL() ;
			}
			t=0;
			for(y=0;y<histo->boxesPerSide;y++)
			{
					for(x=0;x<histo->boxesPerSide;x++)
					{

						sprintf(geom,"POLYGON((%.15g %.15g, %.15g %.15g, %.15g %.15g, %.15g %.15g, %.15g %.15g ))",
							histo->xmin + x*cellx,     histo->ymin+y*celly,
							histo->xmin + (x)*cellx,   histo->ymin+ (y+1)*celly,
							histo->xmin + (x+1)*cellx, histo->ymin+ (y+1)*celly,
							histo->xmin + (x+1)*cellx, histo->ymin+y*celly,
							histo->xmin + x*cellx,     histo->ymin+y*celly
							);
						sprintf(sql,"INSERT INTO %s VALUES('%s'::geometry,%i,%i,%.15g)",tablename,geom,t,histo->value[t],histo->value[t]/((double)total)*100.0);
//elog(NOTICE,"SQL:%s",sql);
						t++;
						SPIcode = SPI_exec(sql, 2147483640 ); // max signed int32
						if (SPIcode  != SPI_OK_INSERT )
						{
								elog(ERROR,"explode_histogram2d: couldnt insert into");
								PG_RETURN_NULL() ;
						}
					}
			}

	    SPIcode =SPI_finish();
	    if (SPIcode  != SPI_OK_FINISH )
		{
					elog(ERROR,"build_histogram2d: couldnt disconnect from SPI");
					PG_RETURN_NULL() ;
		}

	PG_RETURN_POINTER(histo) ;
}


//build_histogram2d (HISTOGRAM2D, tablename, columnname)
// executes the SPI 'SELECT box3d(columnname) FROM tablename'
// and sticks all the results in the histogram
PG_FUNCTION_INFO_V1(build_histogram2d);
Datum build_histogram2d(PG_FUNCTION_ARGS)
{
	HISTOGRAM2D   *histo = (HISTOGRAM2D *)  PG_DETOAST_DATUM(PG_GETARG_DATUM(0));

	char	*tablename, *columnname;
	HISTOGRAM2D   *result;
	int SPIcode;
	char	sql[1000];
	SPITupleTable *tuptable;
	TupleDesc tupdesc ;
	int ntuples,t;
	Datum	datum;
	bool	isnull;
	HeapTuple tuple ;
	BOX	*box;
    double box_area, area_intersect, cell_area;
    int x_idx_min, x_idx_max;
    int y_idx_min, y_idx_max;
    double xmin,ymin, xmax,ymax;
    double intersect_x, intersect_y;
    int x,y;
    int total;
    double sum_area;
    int    sum_area_numb;

  	double sum_area_new      = 0;
    int    sum_area_numb_new =0;
	int bump=0;

	int tuplimit = 500000;	// No. of tuples returned on each cursor fetch
	bool moredata;
	void *SPIplan;
	void *SPIportal;


    xmin = histo->xmin;
    ymin = histo->ymin;
    xmax = histo->xmax;
    ymax = histo->ymax;


	result = (HISTOGRAM2D *) malloc(histo->size);
	memcpy(result,histo,histo->size);


		total = 0;
		for(t=0;t<histo->boxesPerSide*histo->boxesPerSide;t++)
		{
			total+=histo->value[t];
		}



		sum_area = histo->avgFeatureArea * total;
		sum_area_numb = total;



	    tablename  = DatumGetCString(DirectFunctionCall1(textout,
	                                                PointerGetDatum(PG_GETARG_DATUM(1))));

		columnname = DatumGetCString(DirectFunctionCall1(textout,
	                                                PointerGetDatum(PG_GETARG_DATUM(2))));

	//elog(NOTICE,"Start build_histogram2d with %i items already existing", sum_area_numb);
	//elog(NOTICE,"table=\"%s\", column = \"%s\"", tablename, columnname);


		SPIcode = SPI_connect();

		if (SPIcode  != SPI_OK_CONNECT)
		{
			elog(ERROR,"build_histogram2d: couldnt open a connection to SPI");
			PG_RETURN_NULL() ;

		}


			sprintf(sql,"SELECT box(\"%s\") FROM \"%s\"",columnname,tablename);
			//elog(NOTICE,"executing %s",sql);

			SPIplan = SPI_prepare(sql, 0, NULL);
			if (SPIplan  == NULL)
			{
					elog(ERROR,"build_histogram2d: couldnt create query plan via SPI");
					PG_RETURN_NULL() ;
			}

			SPIportal = SPI_cursor_open(NULL, SPIplan, NULL, NULL);
			if (SPIportal == NULL)
			{
					elog(ERROR,"build_histogram2d: couldn't create cursor via SPI");
					PG_RETURN_NULL() ;
			}

			
			moredata = TRUE;
			while (moredata==TRUE) {

				//elog(NOTICE,"about to fetch...");
				SPI_cursor_fetch(SPIportal, TRUE, tuplimit);

				ntuples = SPI_processed;
				//elog(NOTICE,"processing %d records", ntuples);

				if (ntuples > 0) {

					tuptable = SPI_tuptable;
					tupdesc = SPI_tuptable->tupdesc;

					cell_area = ( (xmax-xmin)*(ymax-ymin)/(histo->boxesPerSide*histo->boxesPerSide) );

					for (t=0;t<ntuples;t++)
					{
						tuple = tuptable->vals[t];
						datum = SPI_getbinval(tuple, tupdesc, 1, &isnull);
						if (!(isnull))
						{
							box = (BOX *)DatumGetPointer(datum);
							box_area = (box->high.x-box->low.x)*(box->high.y-box->low.y);

							sum_area_new += box_area;
							sum_area_numb_new ++;

							if (box_area > cell_area )
								box_area = cell_area;
							if (box_area<0)
								box_area =0;  // for precision!

							//check to see which boxes this intersects
							x_idx_min = (box->low.x-xmin)/(xmax-xmin)*histo->boxesPerSide;
							if (x_idx_min <0)
								x_idx_min = 0;
							if (x_idx_min >= histo->boxesPerSide)
								x_idx_min = histo->boxesPerSide-1;
							y_idx_min = (box->low.y-ymin)/(ymax-ymin)*histo->boxesPerSide;
							if (y_idx_min <0)
								y_idx_min = 0;
							if (y_idx_min >= histo->boxesPerSide)
								y_idx_min = histo->boxesPerSide-1;

							x_idx_max = (box->high.x-xmin)/(xmax-xmin)*histo->boxesPerSide;
							if (x_idx_max <0)
								x_idx_max = 0;
							if (x_idx_max >= histo->boxesPerSide)
								x_idx_max = histo->boxesPerSide-1;
							y_idx_max = (box->high.y-ymin)/(ymax-ymin)*histo->boxesPerSide ;
							if (y_idx_max <0)
								y_idx_max = 0;
							if (y_idx_max >= histo->boxesPerSide)
								y_idx_max = histo->boxesPerSide-1;

							//the {x,y}_idx_{min,max} define the grid squares that the box intersects
							// if the area of the intersect between the box and the grid square > 5% of

			//elog(NOTICE,"box is : (%.15g,%.15g to %.15g,%.15g)",box->low.x,box->low.y, box->high.x, box->high.y);
		//elog(NOTICE,"        search is in x: %i to %i   y: %i to %i",x_idx_min, x_idx_max, y_idx_min,y_idx_max);
							for (y= y_idx_min; y<=y_idx_max;y++)
							{
								for (x=x_idx_min;x<= x_idx_max;x++)
								{
										intersect_x = min(box->high.x, xmin+ (x+1) * (xmax-xmin)/histo->boxesPerSide ) -
													max(box->low.x, xmin+ x*(xmax-xmin)/histo->boxesPerSide ) ;
										intersect_y = min(box->high.y, ymin+ (y+1) * (ymax-ymin)/histo->boxesPerSide ) -
													max(box->low.y, ymin+ y*(ymax-ymin)/histo->boxesPerSide ) ;

											// for a point, intersect_x=0, intersect_y=0, box_area =0
		//elog(NOTICE,"x=%i,y=%i, intersect_x= %.15g, intersect_y = %.15g",x,y,intersect_x,intersect_y);
										if ( (intersect_x>=0) && (intersect_y>=0) )
										{
											area_intersect = intersect_x*intersect_y;
											if (area_intersect >= box_area*0.05)
											{
		//elog(NOTICE,"bump");
												bump++;
												result->value[x+y*histo->boxesPerSide]++;
											}
										}
								}
							} // End of y

						} // End isnull

					} // End of for loop

					// Free all the results after each fetch, otherwise all tuples stay
					// in memory until the end of the table...
					SPI_freetuptable(tuptable);

				} else {
						moredata = FALSE;
				} // End of if ntuples > 0

			} // End of while loop


		// Close the cursor
		SPI_cursor_close(SPIportal);

	    SPIcode =SPI_finish();
	    if (SPIcode  != SPI_OK_FINISH )
		{
					elog(ERROR,"build_histogram2d: couldnt disconnect from SPI");
					PG_RETURN_NULL() ;
		}

		//elog(NOTICE,"finishing up build_histogram2d ");

	//pfree(tablename);
	//pfree(columnname);

	total = 0;
	for(t=0;t<histo->boxesPerSide*histo->boxesPerSide;t++)
	{
		total+=result->value[t];
	}
	//elog(NOTICE ,"histogram finishes with %i items in it - acutally added %i rows and %i bumps\n",total,sum_area_numb_new,bump);
	//elog(NOTICE,"done build_histogram2d ");


	//re-calculate statistics on avg bbox size
	if (sum_area_numb_new >0)
		result->avgFeatureArea = (sum_area_new+sum_area)/((double)(sum_area_numb_new+sum_area_numb));

	PG_RETURN_POINTER(result) ;
}





#if USE_VERSION < 75
/*
 * get_restriction_var
 *		Examine the args of a restriction clause to see if it's of the
 *		form (var op something) or (something op var).	If so, extract
 *		and return the var and the other argument.
 *
 * Inputs:
 *	args: clause argument list
 *	varRelid: see specs for restriction selectivity functions
 *
 * Outputs: (these are set only if TRUE is returned)
 *	*var: gets Var node
 *	*other: gets other clause argument
 *	*varonleft: set TRUE if var is on the left, FALSE if on the right
 *
 * Returns TRUE if a Var is identified, otherwise FALSE.
 */
static bool
get_restriction_var(List *args,
					int varRelid,
					Var **var,
					Node **other,
					bool *varonleft)
{
	Node	   *left,
			   *right;

	if (length(args) != 2)
		return false;

	left = (Node *) lfirst(args);
	right = (Node *) lsecond(args);

	/* Ignore any binary-compatible relabeling */

	if (IsA(left, RelabelType))
		left = (Node *)((RelabelType *) left)->arg;
	if (IsA(right, RelabelType))
		right = (Node *)((RelabelType *) right)->arg;

	/* Look for the var */

	if (IsA(left, Var) &&
		(varRelid == 0 || varRelid == ((Var *) left)->varno))
	{
		*var = (Var *) left;
		*other = right;
		*varonleft = true;
	}
	else if (IsA(right, Var) &&
			 (varRelid == 0 || varRelid == ((Var *) right)->varno))
	{
		*var = (Var *) right;
		*other = left;
		*varonleft = false;
	}
	else
	{
		/* Duh, it's too complicated for me... */
		return false;
	}

	return true;
}

//restriction in the GiST && operator
PG_FUNCTION_INFO_V1(postgis_gist_sel);
Datum postgis_gist_sel(PG_FUNCTION_ARGS)
{
		Query	   *root = (Query *) PG_GETARG_POINTER(0);
	//	Oid			operator = PG_GETARG_OID(1);
		List	   *args = (List *) PG_GETARG_POINTER(2);
		int			varRelid = PG_GETARG_INT32(3);
		GEOMETRY	*in;
		BOX			*search_box;
		char		sql[1000];

		SPITupleTable *tuptable;
		TupleDesc tupdesc ;
		HeapTuple tuple ;

		Datum datum;
		bool isnull;


		Var	*var;
		Node	*other;
		bool varonleft;
		Oid relid;
		int SPIcode;

		double myest;

#ifndef USE_STATS
	PG_RETURN_FLOAT8(0.000005);
#endif

		  //PG_RETURN_FLOAT8(0.000005);

		//elog(NOTICE,"postgis_gist_sel was called");

		if (!get_restriction_var(args, varRelid,
							 &var, &other, &varonleft))
		{
			//elog(NOTICE,"get_restriction_var FAILED -returning early");
			PG_RETURN_FLOAT8(0.000005);
		}

		relid = getrelid(var->varno, root->rtable);
		if (relid == InvalidOid)
		{
			//elog(NOTICE,"getrelid FAILED (invalid oid) -returning early");
			PG_RETURN_FLOAT8(0.000005);
		}

		//elog(NOTICE,"operator's oid = %i (this should be GEOMETRY && GEOMETRY)",operator);
		//elog(NOTICE,"relations' oid = %i (this should be the relation that the && is working on) ",relid);
		//elog(NOTICE,"varatt oid = %i (basically relations column #) ",var->varattno);


		if (IsA(other, Const) &&((Const *) other)->constisnull)
		{
			//elog(NOTICE,"other operand of && is NULL - returning early");
			PG_RETURN_FLOAT8(0.000005);
		}

		if (IsA(other, Const))
		{
			//elog(NOTICE,"The other side of the && is a constant with type (oid) = %i and length %i.  This should be GEOMETRY with length -1 (variable length)",((Const*)other)->consttype,((Const*)other)->constlen);

		}
		else
		{
			//elog(NOTICE,"the other side of && isnt a constant - returning early");
			PG_RETURN_FLOAT8(0.000005);
		}

		//get the BOX thats being searched in
		in = (GEOMETRY*)PG_DETOAST_DATUM( ((Const*)other)->constvalue );
		search_box = convert_box3d_to_box(&in->bvol);

		//elog(NOTICE,"requested search box is : (%.15g %.15g, %.15g %.15g)",search_box->low.x,search_box->low.y,search_box->high.x,search_box->high.y);


			SPIcode = SPI_connect();
			if (SPIcode  != SPI_OK_CONNECT)
			{
				elog(NOTICE,"postgis_gist_sel: couldnt open a connection to SPI:%i",SPIcode);
				PG_RETURN_FLOAT8(0.000005) ;
			}

			sprintf(sql,"SELECT stats FROM GEOMETRY_COLUMNS WHERE attrelid=%u AND varattnum=%i",relid,var->varattno);
			//elog(NOTICE,"sql:%s",sql);
			SPIcode = SPI_exec(sql, 1 );
			if (SPIcode  != SPI_OK_SELECT )
			{
				SPI_finish();
				elog(NOTICE,"postgis_gist_sel: couldnt execute sql via SPI");
				PG_RETURN_FLOAT8(0.000005) ;
			}

			if (SPI_processed !=1)
			{
				SPI_finish();
				//elog(NOTICE,"postgis_gist_sel: geometry_columns didnt return a unique value");
				PG_RETURN_FLOAT8(0.000005) ;
			}

			tuptable = SPI_tuptable;
			tupdesc = SPI_tuptable->tupdesc;
			tuple = tuptable->vals[0];
			datum = SPI_getbinval(tuple, tupdesc, 1, &isnull);
			if (isnull)
			{
				SPI_finish();
				//elog(NOTICE,"postgis_gist_sel: geometry_columns returned a null histogram");
				PG_RETURN_FLOAT8(0.000005) ;
			}
//elog(NOTICE,"postgis_gist_sel: checking against estimate_histogram2d");
			// now we have the histogram, and our search box - use the estimate_histogram2d(histo,box) to get the result!
			myest =
					DatumGetFloat8( DirectFunctionCall2( estimate_histogram2d, datum, PointerGetDatum(search_box) ) );

			if ( (myest<0) || (myest!=myest) ) // <0?  or NaN?
			{
				//elog(NOTICE,"postgis_gist_sel: got something crazy back from estimate_histogram2d");
				PG_RETURN_FLOAT8(0.000005) ;
			}




			SPIcode =SPI_finish();
			if (SPIcode  != SPI_OK_FINISH )
			{
				//elog(NOTICE,"postgis_gist_sel: couldnt disconnect from SPI");
				PG_RETURN_FLOAT8(0.000005) ;
			}
//elog(NOTICE,"postgis_gist_sel: finished, returning with %lf",myest);
        PG_RETURN_FLOAT8(myest);
}

static void
genericcostestimate2(Query *root, RelOptInfo *rel,
					IndexOptInfo *index, List *indexQuals,
					Cost *indexStartupCost,
					Cost *indexTotalCost,
					Selectivity *indexSelectivity,
					double *indexCorrelation)
{
	double		numIndexTuples;
	double		numIndexPages;
	List		*selectivityQuals = indexQuals;
#if USE_VERSION >= 74
	QualCost	index_qual_cost;
#endif


//elog(NOTICE,"in genericcostestimate");


	/*
	 * If the index is partial, AND the index predicate with the
	 * explicitly given indexquals to produce a more accurate idea of the
	 * index restriction.  This may produce redundant clauses, which we
	 * hope that cnfify and clauselist_selectivity will deal with
	 * intelligently.
	 *
	 * Note that index->indpred and indexQuals are both in implicit-AND form
	 * to start with, which we have to make explicit to hand to
	 * canonicalize_qual, and then we get back implicit-AND form again.
	 */
	if (index->indpred != NIL)
	{
		Expr	   *andedQuals;

		andedQuals = make_ands_explicit(nconc(listCopy(index->indpred),
											  indexQuals));
		selectivityQuals = canonicalize_qual(andedQuals, true);
	}



	/* Estimate the fraction of main-table tuples that will be visited */
	*indexSelectivity = clauselist_selectivity(root, selectivityQuals,
#if USE_VERSION < 74
		lfirsti(rel->relids));
#else
		rel->relid, JOIN_INNER);
#endif

	/*
	 * Estimate the number of tuples that will be visited.	We do it in
	 * this rather peculiar-looking way in order to get the right answer
	 * for partial indexes.  We can bound the number of tuples by the
	 * index size, in any case.
	 */
	numIndexTuples = *indexSelectivity * rel->tuples;

	//elog(NOTICE,"relation has %li pages",rel->pages);
	//elog(NOTICE,"indexselectivity is %lf, ntuples = %lf, numindexTuples = %lf, index->tuples = %lf",*indexSelectivity, rel->tuples, numIndexTuples,index->tuples);


	if (numIndexTuples > index->tuples)
		numIndexTuples = index->tuples;

	/*
	 * Always estimate at least one tuple is touched, even when
	 * indexSelectivity estimate is tiny.
	 */
	if (numIndexTuples < 1.0)
		numIndexTuples = 1.0;

	/*
	 * Estimate the number of index pages that will be retrieved.
	 *
	 * For all currently-supported index types, the first page of the index
	 * is a metadata page, and we should figure on fetching that plus a
	 * pro-rated fraction of the remaining pages.
	 */


	 //elog(NOTICE,"index->pages =  %li ",index->pages);

	if (index->pages > 1 && index->tuples > 0)
	{
		numIndexPages = (numIndexTuples / index->tuples) * (index->pages - 1);
		numIndexPages += 1;		/* count the metapage too */
		numIndexPages = ceil(numIndexPages);
	}
	else
		numIndexPages = 1.0;


//elog(NOTICE,"numIndexPages =  %lf ",numIndexPages);
	/*
	 * Compute the index access cost.
	 *
	 * Our generic assumption is that the index pages will be read
	 * sequentially, so they have cost 1.0 each, not random_page_cost.
	 * Also, we charge for evaluation of the indexquals at each index
	 * tuple. All the costs are assumed to be paid incrementally during
	 * the scan.
	 */

#if USE_VERSION < 74
	*indexStartupCost = 0;
	*indexTotalCost = numIndexPages +
		(cpu_index_tuple_cost + cost_qual_eval(indexQuals)) *
		numIndexTuples;
#else
	cost_qual_eval(&index_qual_cost, indexQuals);
	*indexStartupCost = index_qual_cost.startup;
	*indexTotalCost = numIndexPages +
		(cpu_index_tuple_cost + index_qual_cost.per_tuple) *
		numIndexTuples;
#endif


	//elog(NOTICE,"cpu_index_tuple_cost = %lf, cost_qual_eval(indexQuals)) = %lf",
	//	cpu_index_tuple_cost,cost_qual_eval(indexQuals));

	//elog(NOTICE,"indexTotalCost =  %lf ",*indexTotalCost);

	/*
	 * Generic assumption about index correlation: there isn't any.
	 */
	*indexCorrelation = 0.97;
	//elog(NOTICE,"indexcorrelation =  %lf ",*indexCorrelation);
}


PG_FUNCTION_INFO_V1(postgisgistcostestimate);
Datum
postgisgistcostestimate(PG_FUNCTION_ARGS)
{
    Query      *root = (Query *) PG_GETARG_POINTER(0);
    RelOptInfo *rel = (RelOptInfo *) PG_GETARG_POINTER(1);
    IndexOptInfo *index = (IndexOptInfo *) PG_GETARG_POINTER(2);
    List       *indexQuals = (List *) PG_GETARG_POINTER(3);
    Cost       *indexStartupCost = (Cost *) PG_GETARG_POINTER(4);
    Cost       *indexTotalCost = (Cost *) PG_GETARG_POINTER(5);
    Selectivity *indexSelectivity = (Selectivity *) PG_GETARG_POINTER(6);
    double     *indexCorrelation = (double *) PG_GETARG_POINTER(7);

//elog(NOTICE,"postgisgistcostestimate was called");

    genericcostestimate2(root, rel, index, indexQuals,
                        indexStartupCost, indexTotalCost,
                        indexSelectivity, indexCorrelation);
//elog(NOTICE,"postgisgistcostestimate is going to return void");

    PG_RETURN_VOID();
}

#else // USE_VERSION >= 75

/*
 * This function should return an estimation of the number of
 * rows returned by a query involving an overlap check 
 * ( it's the restrict function for the && operator )
 *
 * It can make use (if available) of the statistics collected
 * by the geometry analyzer function.
 *
 * This is the one used for PG version >= 7.5
 *
 */
PG_FUNCTION_INFO_V1(postgis_gist_sel);
Datum postgis_gist_sel(PG_FUNCTION_ARGS)
{
	Query *root = (Query *) PG_GETARG_POINTER(0);
	//Oid operator = PG_GETARG_OID(1);
	List *args = (List *) PG_GETARG_POINTER(2);
	int varRelid = PG_GETARG_INT32(3);
	Oid relid;
	HeapTuple stats_tuple;
	GEOM_STATS *geomstats;
	int geomstats_nvalues=0;
	Const *other;
	Var *self;
	GEOMETRY *in;
	BOX *search_box;
	float8 selectivity=0;

#if DEBUG_GEOMETRY_STATS
	elog(NOTICE, "postgis_gist_sel called");
#endif

	/*
	 * Find the constant part
	 */

	other = (Const *) lfirst(args);
	if ( ! IsA(other, Const) ) 
	{
		self = (Var *)other;
		other = (Const *) lsecond(args);
	}
	else
	{
		self = (Var *) lsecond(args);
	}

	if ( ! IsA(other, Const) ) 
	{
#if DEBUG_GEOMETRY_STATS
		elog(NOTICE, " no constant arguments - returning default selectivity");
#endif
		PG_RETURN_FLOAT8(DEFAULT_GEOMETRY_SEL);
	}

	/*
	 * We are working on two constants..
	 * TODO: check if expression is true,
	 *       returned set would be either 
	 *       the whole or none.
	 */
	if ( ! IsA(self, Var) ) 
	{
#if DEBUG_GEOMETRY_STATS
		elog(NOTICE, " no variable argument ? - returning default selectivity");
#endif
		PG_RETURN_FLOAT8(DEFAULT_GEOMETRY_SEL);
	}

	/*
	 * Convert the constant to a BOX
	 */

	in = (GEOMETRY*)PG_DETOAST_DATUM( ((Const*)other)->constvalue );
	search_box = convert_box3d_to_box(&in->bvol);
#if DEBUG_GEOMETRY_STATS > 1
	elog(NOTICE," requested search box is : %.15g %.15g, %.15g %.15g",search_box->low.x,search_box->low.y,search_box->high.x,search_box->high.y);
#endif

	/*
	 * Get pg_statistic row
	 */

	relid = getrelid(varRelid, root->rtable);
	stats_tuple = SearchSysCache(STATRELATT, ObjectIdGetDatum(relid), Int16GetDatum(self->varattno), 0, 0);
	if ( ! stats_tuple )
	{
#if DEBUG_GEOMETRY_STATS
		elog(NOTICE, " SearchSysCache returned NULL - ret.def.");
#endif
		PG_RETURN_FLOAT8(DEFAULT_GEOMETRY_SEL);
	}


	if ( ! get_attstatsslot(stats_tuple, 0, 0,
		STATISTIC_KIND_GEOMETRY, InvalidOid, NULL, NULL,
		(float4 **)&geomstats, &geomstats_nvalues) )
	{
#if DEBUG_GEOMETRY_STATS
		elog(NOTICE, " STATISTIC_KIND_GEOMETRY stats not found - returning default geometry selectivity");
#endif
		ReleaseSysCache(stats_tuple);
		PG_RETURN_FLOAT8(DEFAULT_GEOMETRY_SEL);

	}

#if DEBUG_GEOMETRY_STATS > 1
		elog(NOTICE, " %d read from stats", geomstats_nvalues);
#endif

#if DEBUG_GEOMETRY_STATS > 1
	elog(NOTICE, " histo: xmin,ymin: %f,%f",
		geomstats->xmin, geomstats->ymin);
	elog(NOTICE, " histo: xmax,ymax: %f,%f",
		geomstats->xmax, geomstats->ymax);
	elog(NOTICE, " histo: boxesPerSide: %f", geomstats->boxesPerSide);
	elog(NOTICE, " histo: avgFeatureArea: %f", geomstats->avgFeatureArea);
	elog(NOTICE, " histo: avgFeatureCells: %f", geomstats->avgFeatureCells);
#endif

	/*
	 * Do the estimation
	 */
	{
		int x, y;
		int x_idx_min, x_idx_max, y_idx_min, y_idx_max;
		double intersect_x, intersect_y, AOI;
		double cell_area, box_area;
		double geow, geoh; // width and height of histogram
		int bps; // boxesPerSide 
		BOX *box;
		double value;
		float overlapping_cells;
		float avg_feat_cells;
		double gain;

		box = search_box;
		geow = geomstats->xmax-geomstats->xmin;
		geoh = geomstats->ymax-geomstats->ymin;
		bps = geomstats->boxesPerSide;
		cell_area = (geow*geoh) / (bps*bps);
		box_area = (box->high.x-box->low.x)*(box->high.y-box->low.y);
		value = 0;

		/* Find first overlapping column */
		x_idx_min = (box->low.x-geomstats->xmin) / geow * bps;
		if (x_idx_min < 0) {
#if DEBUG_GEOMETRY_STATS
			elog(NOTICE, " search_box overlaps %d columns on the left of histogram grid", -x_idx_min);
#endif
			// should increment the value somehow
			x_idx_min = 0;
		}
		if (x_idx_min >= bps)
		{
#if DEBUG_GEOMETRY_STATS
			elog(NOTICE, " search_box overlaps %d columns on the right of histogram grid", x_idx_min-bps+1);
#endif
			// should increment the value somehow
			x_idx_min = bps-1;
		}

		/* Find first overlapping row */
		y_idx_min = (box->low.y-geomstats->ymin) / geoh * bps;
		if (y_idx_min <0)
		{
#if DEBUG_GEOMETRY_STATS
			elog(NOTICE, " search_box overlaps %d columns on the bottom of histogram grid", -y_idx_min);
#endif
			// should increment the value somehow
			y_idx_min = 0;
		}
		if (y_idx_min >= bps)
		{
#if DEBUG_GEOMETRY_STATS
			elog(NOTICE, " search_box overlaps %d columns on the top of histogram grid", y_idx_min-bps+1);
#endif
			// should increment the value somehow
			y_idx_min = bps-1;
		}

		/* Find last overlapping column */
		x_idx_max = (box->high.x-geomstats->xmin) / geow * bps;
		if (x_idx_max <0)
		{
			// should increment the value somehow
			x_idx_max = 0;
		}
		if (x_idx_max >= bps )
		{
			// should increment the value somehow
			x_idx_max = bps-1;
		}

		/* Find last overlapping row */
		y_idx_max = (box->high.y-geomstats->ymin) / geoh * bps;
		if (y_idx_max <0)
		{
			// should increment the value somehow
			y_idx_max = 0;
		}
		if (y_idx_max >= bps)
		{
			// should increment the value somehow
			y_idx_max = bps-1;
		}

		/*
		 * the {x,y}_idx_{min,max}
		 * define the grid squares that the box intersects
		 */
		for (y=y_idx_min; y<=y_idx_max; y++)
		{
			for (x=x_idx_min; x<=x_idx_max; x++)
			{
				double val;
				double gain;

				val = geomstats->value[x+y*bps];

				/*
				 * Of the cell value we get
				 * only the overlap fraction.
				 */

				intersect_x = min(box->high.x, geomstats->xmin + (x+1) * geow / bps) - max(box->low.x, geomstats->xmin + x * geow / bps );
				intersect_y = min(box->high.y, geomstats->ymin + (y+1) * geoh / bps) - max(box->low.y, geomstats->ymin+ y * geoh / bps) ;
				
				AOI = intersect_x*intersect_y;
				gain = AOI/cell_area;

#if DEBUG_GEOMETRY_STATS > 1
				elog(NOTICE, " [%d,%d] cell val %.15f",
						x, y, val);
				elog(NOTICE, " [%d,%d] gain %.15f",
						x, y, gain);
#endif

				val *= gain;

#if DEBUG_GEOMETRY_STATS > 1
				elog(NOTICE, " [%d,%d] adding %.15f to value",
					x, y, val);
#endif
				value += val;
			}
		}


		/*
		 * If the search_box is a point, it will
		 * overlap a single cell and thus get
		 * it's value, which is the fraction of
		 * samples (we can presume of row set also)
		 * which bumped to that cell.
		 *
		 * If the table features are points, each
		 * of them will overlap a single histogram cell.
		 * Our search_box value would then be correctly
		 * computed as the sum of the bumped cells values.
		 *
		 * If both our search_box AND the sample features
		 * overlap more then a single histogram cell we
		 * need to consider the fact that our sum computation
		 * will have many duplicated included. E.g. each
		 * single sample feature would have contributed to
		 * raise the search_box value by as many times as
		 * many cells in the histogram are commonly overlapped
		 * by both searc_box and feature. We should then
		 * divide our value by the number of cells in the virtual
		 * 'intersection' between average feature cell occupation
		 * and occupation of the search_box. This is as 
		 * fuzzy as you understand it :)
		 *
		 * Consistency check: whenever the number of cells is
		 * one of whichever part (search_box_occupation,
		 * avg_feature_occupation) the 'intersection' must be 1.
		 * If sounds that our 'intersaction' is actually the
		 * minimun number between search_box_occupation and
		 * avg_feat_occupation.
		 *
		 */
		overlapping_cells = (x_idx_max-x_idx_min+1) *
			(y_idx_max-y_idx_min+1);
		avg_feat_cells = geomstats->avgFeatureCells;

#if DEBUG_GEOMETRY_STATS
	elog(NOTICE, " search_box overlaps %f cells", overlapping_cells);
	elog(NOTICE, " avg feat overlaps %f cells", avg_feat_cells);
#endif

		gain = 1/min(overlapping_cells, avg_feat_cells);
		selectivity = value*gain;

#if DEBUG_GEOMETRY_STATS
		elog(NOTICE, " SUM(ov_histo_cells)=%f", value);
		elog(NOTICE, " gain=%f", gain);
		elog(NOTICE, " selectivity=%f", selectivity);
#endif

		/* prevent rounding overflows */
		if (selectivity > 1.0) selectivity = 1.0;
		else if (selectivity < 0) selectivity = 0.0;
	}
	


#if DEBUG_GEOMETRY_STATS
	elog(NOTICE, " returning computed value: %f", selectivity);
#endif

	free_attstatsslot(0, NULL, 0, (float *)geomstats, geomstats_nvalues);
	ReleaseSysCache(stats_tuple);
	PG_RETURN_FLOAT8(selectivity);

}

/*
 * This function is called by the analyze function iff
 * the geometry_analyze() function give it its pointer
 * (this is always the case so far).
 * The geometry_analyze() function is also responsible
 * of deciding the number of "sample" rows we will receive
 * here. It is able to give use other 'custom' data, but we
 * won't use them so far.
 *
 * Our job is to build some statistics on the sample data
 * for use by operator estimators.
 *
 * Currently we only need statistics to estimate the number of rows
 * overlapping a given extent (estimation function bound
 * to the && operator).
 *
 */
static void
compute_geometry_stats(VacAttrStats *stats, AnalyzeAttrFetchFunc fetchfunc,
		int samplerows, double totalrows)
{
	MemoryContext old_context;
	int i;
	int geom_stats_size;
	BOX **sampleboxes;
	GEOM_STATS *geomstats;
	bool isnull;
	int null_cnt=0, notnull_cnt=0;
	BOX3D *sample_extent=NULL;
	double total_width=0;
	double total_boxes_area=0;
	int total_boxes_cells=0;
	double cell_area;
	double cell_width;
	double cell_height;
	double geow, geoh; // width and height of histogram
	int bps; // boxesPerSide  (alias)
	int boxesPerSide;
	/*
	 * This is where geometry_analyze
	 * should put its' custom parameters.
	 */
	//void *mystats = stats->extra_data;
	
	/*
	 * We'll build an histogram having from 40 to 400 boxesPerSide
	 */
	boxesPerSide = sqrt(160*stats->attr->attstattarget);



#if DEBUG_GEOMETRY_STATS
	elog(NOTICE, "compute_geometry_stats called");
	elog(NOTICE, " samplerows: %d", samplerows);
	elog(NOTICE, " boxesPerSide: %d", boxesPerSide);
#endif

	sampleboxes = palloc(sizeof(BOX *)*samplerows);

	/*
	 * First scan:
	 *  o find extent of the sample rows
	 *  o count null/not-null values
	 *  o compute total_width
	 *  o compute total features's box area (for avgFeatureArea)
	 */
	for (i=0; i<samplerows; i++)
	{
		Datum datum;
		GEOMETRY *geom;
		BOX *box;

		datum = fetchfunc(stats, i, &isnull);

		if ( isnull ) {
			null_cnt++;
			sampleboxes[i] = NULL;
			continue;
		}
		notnull_cnt++;

		geom = (GEOMETRY *) PG_DETOAST_DATUM(datum);
		box = convert_box3d_to_box(&(geom->bvol));
		sampleboxes[i] = box;

		sample_extent = union_box3d(&(geom->bvol), sample_extent);
		
		// TODO: ask if we need geom or bvol size for stawidth
		total_width += geom->size;
		total_boxes_area += (box->high.x-box->low.x)*(box->high.y-box->low.y);

	}

	if ( ! notnull_cnt ) {
		elog(NOTICE, " no notnull values, invalid stats");
		stats->stats_valid = false;
		return;
	}

	/*
	 * Create the histogram (GEOM_STATS)
	 */
	old_context = MemoryContextSwitchTo(stats->anl_context);
	geom_stats_size=sizeof(GEOM_STATS)+
		(boxesPerSide*boxesPerSide-1)*sizeof(float4);
	geomstats = palloc(geom_stats_size);
	MemoryContextSwitchTo(old_context);

	geomstats->xmin = sample_extent->LLB.x;
	geomstats->ymin = sample_extent->LLB.y;
	geomstats->xmax = sample_extent->URT.x;
	geomstats->ymax = sample_extent->URT.y;
	geomstats->boxesPerSide = boxesPerSide;
	geomstats->avgFeatureArea = total_boxes_area/notnull_cnt;
	// Initialize all values to 0
	for (i=0;i<boxesPerSide*boxesPerSide; i++) geomstats->value[i] = 0;


	geow = geomstats->xmax-geomstats->xmin;
	geoh = geomstats->ymax-geomstats->ymin;
	bps = geomstats->boxesPerSide;
	cell_width = geow/bps;
	cell_height = geoh/bps;
	cell_area = cell_width*cell_height;

#if DEBUG_GEOMETRY_STATS > 2
	elog(NOTICE, "cell_width: %f", cell_width);
	elog(NOTICE, "cell_height: %f", cell_height);
#endif
	

	/*
	 * Second scan:
	 *  o fill histogram values with the number of
	 *    features' bbox overlaps: a feature's bvol
	 *    can fully overlap (1) or partially overlap
	 *    (fraction of 1) an histogram cell.
	 *
	 *  o compute total cells occupation
	 */
	for (i=0; i<samplerows; i++)
	{
		BOX *box;
		int x_idx_min, x_idx_max, x;
		int y_idx_min, y_idx_max, y;
		int numcells=0;

		if ( sampleboxes[i] == NULL ) continue;
		box = (BOX *)sampleboxes[i];

#if DEBUG_GEOMETRY_STATS > 2
		elog(NOTICE, " feat %d box is %f %f, %f %f",
				i, box->high.x, box->high.y,
				box->low.x, box->low.y);
#endif
							
		/* Find first overlapping column */
		x_idx_min = (box->low.x-geomstats->xmin) / geow * bps;
		if (x_idx_min <0) x_idx_min = 0;
		if (x_idx_min >= bps) x_idx_min = bps-1;

		/* Find first overlapping row */
		y_idx_min = (box->low.y-geomstats->ymin) / geoh * bps;
		if (y_idx_min <0) y_idx_min = 0;
		if (y_idx_min >= bps) y_idx_min = bps-1;

		/* Find last overlapping column */
		x_idx_max = (box->high.x-geomstats->xmin) / geow * bps;
		if (x_idx_max <0) x_idx_max = 0;
		if (x_idx_max >= bps ) x_idx_max = bps-1;

		/* Find last overlapping row */
		y_idx_max = (box->high.y-geomstats->ymin) / geoh * bps;
		if (y_idx_max <0) y_idx_max = 0;
		if (y_idx_max >= bps) y_idx_max = bps-1;
#if DEBUG_GEOMETRY_STATS > 2
		elog(NOTICE, " feat %d overlaps columns %d-%d, rows %d-%d",
				i, x_idx_min, x_idx_max, y_idx_min, y_idx_max);
#endif

		/*
		 * the {x,y}_idx_{min,max}
		 * define the grid squares that the box intersects
		 */
		for (y=y_idx_min; y<=y_idx_max; y++)
		{
			for (x=x_idx_min; x<=x_idx_max; x++)
			{
				geomstats->value[x+y*bps] += 1;
				numcells++;
			}
		}

		// before adding to the total cells
		// we could decide if we really 
		// want this feature to count
		total_boxes_cells += numcells;
	}

	// what about null features (TODO) ?
	geomstats->avgFeatureCells = (float4)total_boxes_cells/notnull_cnt;

#if DEBUG_GEOMETRY_STATS
	elog(NOTICE, " histo: total_boxes_cells: %d", total_boxes_cells);
	elog(NOTICE, " histo: xmin,ymin: %f,%f",
		geomstats->xmin, geomstats->ymin);
	elog(NOTICE, " histo: xmax,ymax: %f,%f",
		geomstats->xmax, geomstats->ymax);
	elog(NOTICE, " histo: boxesPerSide: %f", geomstats->boxesPerSide);
	elog(NOTICE, " histo: avgFeatureArea: %f", geomstats->avgFeatureArea);
	elog(NOTICE, " histo: avgFeatureCells: %f", geomstats->avgFeatureCells);
#endif


	/*
	 * Normalize histogram
	 *
	 * We divide each histogram cell value
	 * by the number of samples examined.
	 *
	 */
	for (i=0; i<boxesPerSide*boxesPerSide; i++)
		geomstats->value[i] /= samplerows;

#if DEBUG_GEOMETRY_STATS > 1
	{
		int x, y;
		for (x=0; x<bps; x++)
		{
			for (y=0; y<bps; y++)
			{
				elog(NOTICE, " histo[%d,%d] = %.15f", x, y, geomstats->value[x+y*bps]);
			}
		}
	}
#endif

	/*
	 * Write the statistics data 
	 */
	stats->stakind[0] = STATISTIC_KIND_GEOMETRY;
	stats->staop[0] = InvalidOid;
	stats->stanumbers[0] = (float4 *)geomstats;
	stats->numnumbers[0] = geom_stats_size/sizeof(float4);

	stats->stanullfrac = null_cnt/samplerows;
	stats->stawidth = total_width/notnull_cnt;
	stats->stadistinct = -1.0;

#if DEBUG_GEOMETRY_STATS
	elog(NOTICE, " out: slot 0: kind %d (STATISTIC_KIND_GEOMETRY)",
		stats->stakind[0]);
	elog(NOTICE, " out: slot 0: op %d (InvalidOid)", stats->staop[0]);
	elog(NOTICE, " out: slot 0: numnumbers %d", stats->numnumbers[0]);
	elog(NOTICE, " out: null fraction: %d/%d", null_cnt, samplerows);
	elog(NOTICE, " out: average width: %d bytes", stats->stawidth);
	elog(NOTICE, " out: distinct values: all (no check done)");
#endif

	stats->stats_valid = true;
}

/*
 * This function will be called when the ANALYZE command is run
 * on a column of the "geometry" type.
 * 
 * It will need to return a stats builder function reference
 * and a "minimum" sample rows to feed it.
 * If we want analisys to be completely skipped we can return
 * FALSE and leave output vals untouched.
 *
 * What we know from this call is:
 *
 * 	o The pg_attribute row referring to the specific column.
 * 	  Could be used to get reltuples from pg_class (which 
 * 	  might quite inexact though...) and use them to set an
 * 	  appropriate minimum number of sample rows to feed to
 * 	  the stats builder. The stats builder will also receive
 * 	  a more accurate "estimation" of the number or rows.
 *
 * 	o The pg_type row for the specific column.
 * 	  Could be used to set stat builder / sample rows
 * 	  based on domain type (when postgis will be implemented
 * 	  that way).
 *
 * Being this experimental we'll stick to a static stat_builder/sample_rows
 * value for now.
 *
 */
PG_FUNCTION_INFO_V1(geometry_analyze);
Datum geometry_analyze(PG_FUNCTION_ARGS)
{
	VacAttrStats *stats = (VacAttrStats *)PG_GETARG_POINTER(0);
	Form_pg_attribute attr = stats->attr;

#if DEBUG_GEOMETRY_STATS
	elog(NOTICE, "geometry_analyze called");
#endif

	/* If the attstattarget column is negative, use the default value */
	/* NB: it is okay to scribble on stats->attr since it's a copy */
        if (attr->attstattarget < 0)
                attr->attstattarget = default_statistics_target;

#if DEBUG_GEOMETRY_STATS
	elog(NOTICE, " attribute stat target: %d", attr->attstattarget);
#endif

	/*
	 * There might be a reason not to analyze this column
	 * (can we detect the absence of an index?)
	 */
	//elog(NOTICE, "compute_geometry_stats not implemented yet");
	//PG_RETURN_BOOL(false);

	/* Setup the minimum rows and the algorithm function */
	stats->minrows = 300 * stats->attr->attstattarget;
	stats->compute_stats = compute_geometry_stats;

#if DEBUG_GEOMETRY_STATS
	elog(NOTICE, " minrows: %d", stats->minrows);
#endif

	/* Indicate we are done successfully */
	PG_RETURN_BOOL(true);
}

#endif