/* useful to create randomised samples for statistical tests */
void do_split_sample ( char *input, char *output, int in_types, double percentage, char *map,
int all, int processing_mode, int quiet) {
CELL *cellbuf;
DCELL *dcellbuf;
GT_Row_cache_t *cache;
int fd;
int i,j,k,l;
int no_sites;
int sites_tried = 0;
struct Cell_head region;
int error;
char *mapset, errmsg [200];
unsigned int *taken; /* this is an array of 0/1 which signals, if
a certain site has already been 'drawn' */
long row_idx, col_idx;
struct Map_info in_vect_map;
struct Map_info out_vect_map;
struct line_pnts *vect_points;
struct line_cats *vect_cats;
double x,y,z;
int n_points = 1;
int cur_type;
cellbuf = NULL;
dcellbuf = NULL;
cache = NULL;
/* get current region */
G_get_window (®ion);
/* attempt to create new file for output */
Vect_set_open_level (2);
if (0 > Vect_open_new (&out_vect_map, output, 0) ) {
G_fatal_error ("Could not open output vector map.\n");
}
/* open input vector map */
if ((mapset = G_find_vector2 (input, "")) == NULL) {
sprintf (errmsg, "Could not find input %s\n", input);
G_fatal_error ("%s",errmsg);
}
if (1 > Vect_open_old (&in_vect_map, input, "")) {
sprintf (errmsg, "Could not open input map %s.\n", input);
G_fatal_error ("%s",errmsg);
}
vect_points = Vect_new_line_struct ();
vect_cats = Vect_new_cats_struct ();
/* set constraints specified */
if (in_types != 0) {
Vect_set_constraint_type (&in_vect_map, in_types);
}
if (all != 1) {
Vect_set_constraint_region (&in_vect_map, region.north, region.south,
region.east, region.west, 0.0, 0.0);
}
/* get total number of objects with constraints */
i = 0;
while ((cur_type = Vect_read_next_line (&in_vect_map, vect_points, vect_cats) > 0)) {
i ++;
}
k = ( ((float) i/100)) * percentage; /* k now has the number of objects wanted */
if ( quiet != 1 ) {
fprintf (stderr,"Creating randomised sample of size n = %i.\n",k);
}
/* now, we need to acquire exactly 'k' random objects that fall in NON-NULL */
/* coverage raster cells. */
taken = G_calloc (i, sizeof (unsigned int));
for ( l = 0; l < k; l ++ ) {
taken[l] = 0;
}
no_sites = i; /* store this for later use */
/* does user want to filter objects through a raster map? */
if ( map != NULL) {
/* open raster map */
fd = G_open_cell_old (map, G_find_cell (map, ""));
if (fd < 0)
{
G_fatal_error ("Could not open raster map for reading!\n");
}
/* allocate cache and buffer, according to type of coverage */
if ( processing_mode == CELL_TYPE) {
/* INT coverage */
cache = (GT_Row_cache_t *) G_malloc (sizeof (GT_Row_cache_t));
/* TODO: check error value */
error = GT_RC_open (cache, cachesize, fd, CELL_TYPE);
cellbuf = G_allocate_raster_buf (CELL_TYPE);
}
if ( (processing_mode == FCELL_TYPE) || (processing_mode == DCELL_TYPE) ) {
/* FP coverage */
cache = (GT_Row_cache_t *) G_malloc (sizeof (GT_Row_cache_t));
/* TODO: check error value */
error = GT_RC_open (cache, cachesize, fd, DCELL_TYPE);
dcellbuf = G_allocate_raster_buf (DCELL_TYPE);
}
}
srand ( ((unsigned int) time (NULL)) + getpid()); /* set seed for random number generator from system time and process ID*/
i = 0;
/* MAIN LOOP */
while ( i < k ) {
/* get a random index, but one that was not taken already */
l = 0;
while ( l == 0 ) {
j = rand () % ( no_sites - 1 + 1) + 1; /* j now has the random position to try */
if ( taken[j-1] == 0 ) {
l = 1; /* exit loop */
}
}
taken [j-1] = 1; /* mark this index as 'taken' */
sites_tried ++; /* keep track of this so we do not enter an infinite loop */
if ( sites_tried > no_sites ) {
/* could not create a large enough sample */
G_fatal_error ("Could not find enough objects for split sampling.\nDecrease split sample size.\n");
}
/* get next vector object */
cur_type = Vect_read_line (&in_vect_map, vect_points, vect_cats, j);
if (cur_type < 0 ) {
G_fatal_error ("Error reading vector map: premature EOF.\n");
}
/* now, check if coverage under site is NON-NULL and within region */
/* convert site northing to row! */
/* for this check, we use only the first pair of coordinates! */
Vect_copy_pnts_to_xyz (vect_points, &x, &y, &z, &n_points);
row_idx =
(long) G_northing_to_row (y,
®ion);
col_idx =
(long) G_easting_to_col (x,
®ion);
/* do region check, first... OBSOLETE */
/* read row from cache and check for NULL */
/* if required */
if ( map != NULL ) {
if ( processing_mode == CELL_TYPE ) {
cellbuf = GT_RC_get (cache, row_idx);
if (!G_is_c_null_value(&cellbuf[col_idx])) {
i ++;
Vect_write_line (&out_vect_map, cur_type,
vect_points, vect_cats );
fflush (stdout);
}
}
if ( (processing_mode == FCELL_TYPE) || (processing_mode == DCELL_TYPE) ) {
dcellbuf = GT_RC_get (cache, row_idx);
if (!G_is_d_null_value(&dcellbuf[col_idx])) {
i ++;
Vect_write_line (&out_vect_map, cur_type,
vect_points, vect_cats );
fflush (stdout);
}
}
} else {
i ++;
Vect_write_line (&out_vect_map, GV_POINT,
vect_points, vect_cats );
fflush (stdout);
}
/* disregard region setting and map, if -a flag is given */
if ( all == 1 ) {
i ++;
Vect_write_line (&out_vect_map, cur_type,
vect_points, vect_cats );
fflush (stdout);
}
if ( quiet != 1 ) {
G_percent(i,k,1);
}
}
/* END OF MAIN LOOP */
Vect_copy_head_data (&in_vect_map, &out_vect_map);
fprintf (stdout, "Building topology information for output map.\n");
Vect_build (&out_vect_map);
Vect_close (&in_vect_map);
Vect_close (&out_vect_map);
if ( map != NULL ) {
/* close cache, free buffers! */
GT_RC_close (cache);
if ( processing_mode == CELL_TYPE ) {
G_free (cellbuf);
}
if ( (processing_mode == FCELL_TYPE) || (processing_mode == DCELL_TYPE) ) {
G_free (dcellbuf);
}
G_free (cache);
}
}
/* create the actual report */
void do_report_CELL ( char *map, char *mapset, char *sites,
int precision, int null_flag, int uncat_flag,
int all_flag, int quiet_flag, int skip_flag,
char *logfile, int background, int gain, int show_progress) {
CELL *cellbuf;
struct Cell_head region;
GT_Row_cache_t *cache;
unsigned long row_idx, col_idx;
int fd;
unsigned long i,j,k;
unsigned long no_sites;
FILE *lp;
unsigned long nrows, ncols;
unsigned long *share_smp = NULL; /* array that keeps percentage of sites */
double total = 0;
double map_total = 0;
double kvamme_gain;
long null_count = 0; /* keeps count of sites on NULL cells */
long nocat_count = 0;
/* category counts and descriptions */
int cats;
char **cats_description; /* category labels */
long *cat_count; /* category counts */
long null_count_map; /* number of NULL cells in input map */
long nocat_count_map; /* number of cells that do not fall into the category range [0 .. n] */
int debug_mode = 0; /* 1 to enable writing additional output to logfile */
time_t systime;
char errmsg [200];
struct Map_info in_vect_map;
struct line_pnts *vect_points;
double x,y,z;
int n_points = 1;
int cur_type;
/* get current region */
G_get_window (®ion);
nrows = G_window_rows ();
ncols = G_window_cols ();
/* check logfile */
if (logfile != NULL) {
debug_mode = 1;
if ( !G_legal_filename (logfile) ) {
delete_tmpfile (map);
G_fatal_error ("Please specify a legal filename for the logfile.\n");
}
/* attempt to write to logfile */
if ( (lp = fopen ( logfile, "w+" ) ) == NULL ) {
delete_tmpfile (map);
G_fatal_error ("Could not create logfile.\n");
}
/* we want unbuffered output for the logfile */
setvbuf (lp,NULL,_IONBF,0);
fprintf (lp,"This is %s, version %.2f\n",PROGNAME, PROGVERSION);
systime = time (NULL);
fprintf (lp,"Started on %s",ctime(&systime));
fprintf (lp,"\tlocation = %s\n",G_location());
fprintf (lp,"\tmapset = %s\n",G_mapset());
fprintf (lp,"\tinput map = %s\n",map);
fprintf (lp,"\tsample file = %s\n",sites);
} else {
/* log output to stderr by default */
lp = stderr;
}
if (1 > Vect_open_old (&in_vect_map, sites, "")) {
delete_tmpfile (map);
sprintf (errmsg, "Could not open input map %s.\n", sites);
G_fatal_error (errmsg);
}
vect_points = Vect_new_line_struct ();
if (all_flag != 1) {
Vect_set_constraint_region (&in_vect_map, region.north, region.south,
region.east, region.west, 0.0, 0.0);
}
/* get total number of sampling points */
i = 0;
while ((cur_type = Vect_read_next_line (&in_vect_map, vect_points, NULL) > 0)) {
i ++;
}
no_sites = i; /* store this for later use */
/* open raster map */
fd = G_open_cell_old (map, G_find_cell (map, ""));
if (fd < 0)
{
delete_tmpfile (map);
G_fatal_error ("Could not open raster map for reading!\n");
}
/* allocate a cache and a raster buffer */
cache = (GT_Row_cache_t *) G_malloc (sizeof (GT_Row_cache_t));
GT_RC_open (cache, CACHESIZE, fd, CELL_TYPE);
cellbuf = G_allocate_raster_buf (CELL_TYPE);
cats = GT_get_stats (map,mapset,&null_count_map, &nocat_count_map, show_progress);
if ( cats < 2 ) {
delete_tmpfile (map);
G_fatal_error ("Input map must have at least two categories.");
}
/* get category labels and counts */
cats_description = GT_get_labels (map,mapset);
if (cats_description == NULL) {
delete_tmpfile (map);
G_fatal_error ("Could not read category labels from input map.");
}
cat_count = GT_get_c_counts (map,mapset, show_progress);
if (cat_count == NULL) {
delete_tmpfile (map);
G_fatal_error ("Could not count categories in input map.");
}
/* allocate a double array to hold statistics */
share_smp = (unsigned long *) G_malloc ((signed)(cats * sizeof (unsigned long)));
for (i = 0; i < cats; i++)
{
share_smp[i] = 0;
}
/* count raster values under sampling points */
i = 0;
k = 0; /* progress counter for status display */
Vect_rewind (&in_vect_map);
if ( !quiet_flag ) {
fprintf (stdout, "Counting sample: \n");
fflush (stdout);
}
/* we MUST not set constraints so that no raster values outside the current region are
accessed, which would give an "illegal cache request" error */
Vect_set_constraint_region (&in_vect_map, region.north, region.south,
region.east, region.west, 0.0, 0.0);
while ((cur_type = Vect_read_next_line (&in_vect_map, vect_points, NULL) > 0)) {
Vect_copy_pnts_to_xyz (vect_points, &x, &y, &z, &n_points);
k ++;
if ( !quiet_flag ) {
G_percent ((signed) k, (signed) no_sites, 1);
}
/* get raster row with same northing as sample and perform
quantification */
row_idx = (long) G_northing_to_row (y, ®ion);
col_idx = (long) G_easting_to_col (x, ®ion);
cellbuf = GT_RC_get (cache, (signed) row_idx);
/* now read the raster value under the current site */
if (G_is_c_null_value (&cellbuf[col_idx]) == 0) {
/* site on cell within category range [0..cats] ? */
if ( (cellbuf[col_idx] > -1) && (cellbuf[col_idx] <= cats) ) {
share_smp [cellbuf[col_idx] ] ++;
/* i keeps track of samples on non-null coverage only */
/* inside the current region */
i ++;
} else {
if ( uncat_flag ) {
/* also keep count of sites on uncategorised cells? */
i ++;
nocat_count++;
}
}
}
if (G_is_c_null_value (&cellbuf[col_idx]) == 1) {
/* got a NULL value under this site */
if (null_flag) { /* only count this, if null flag is set */
null_count ++;
i ++;
}
}
}
Vect_close (&in_vect_map);
fprintf (lp,"\n");
if ( background ) {
fprintf (lp,"Distribution of categories under %lu points (%lu in region) and in input map:\n",i,no_sites);
} else {
fprintf (lp,"Distribution of categories under %lu points (%lu in region):\n",i,no_sites);
}
/* determine starting value for total of sites analysed */
total = 0;
for ( j=0; j < cats; j ++) {
total = total + share_smp[j];
map_total = map_total + cat_count[j];
}
if (null_flag) { /* add NULL values to total */
total = total + null_count;
map_total = map_total + null_count_map;
}
if (uncat_flag) { /* add uncategorised cells to total */
total = total + nocat_count;
map_total = map_total + nocat_count_map;
}
/* Now display those values which the user has chosen */
if ( (background) && (gain) ) {
fprintf (lp,"Cat.\tPts.\t(%%)\tMap\t(%%)\tGain\tDescription\n");
}
if ( (background) && (!gain) ) {
fprintf (lp,"Cat.\tPts.\t(%%)\tMap\t(%%)\tDescription\n");
}
if ( (!background) && (gain) ) {
fprintf (lp,"Cat.\tPts.\t(%%)\tGain\tDescription\n");
}
if ( (!background) && (!gain) ) {
fprintf (lp,"Cat.\tPts.\t(%%)\tDescription\n");
}
for ( j = 0; j < cats; j ++) {
/* if skip_flag is not set: only show categories that have count > 0 */
if ((skip_flag == 1) || ((skip_flag == 0) && (share_smp[j] > 0))) {
if ( (background) && (gain) ) {
/* Kvamme's Gain = 1 - (%area/%sites) */
kvamme_gain = gstats_gain_K(((double) share_smp[j]*(100/total)),
((double) cat_count[j]*(100/map_total)));
fprintf (lp, "%lu\t%6lu\t%6.2f\t%8lu %6.2f\t%6.2f\t%s\n", j, share_smp[j], (float) share_smp[j]*(100/total),
cat_count[j], (float) cat_count[j]*(100/map_total),
kvamme_gain, cats_description[j]);
}
if ( (background) && (!gain) ) {
fprintf (lp, "%lu\t%6lu\t%6.2f\t%8lu %6.2f\t%s\n", j, share_smp[j], (float) share_smp[j]*(100/total),
cat_count[j], (float) cat_count[j]*(100/map_total),
cats_description[j]);
}
if ( (!background) && (gain) ) {
kvamme_gain = 1-( (float) cat_count[j]*(100/map_total) / (float) share_smp[j]*(100/total) );
fprintf (lp, "%lu\t%6lu\t%6.2f\t%6.2f\t%s\n", j, share_smp[j], (float) share_smp[j]*(100/total),
kvamme_gain, cats_description[j]);
}
if ( (!background) && (!gain) ) {
fprintf (lp, "%lu\t%6lu\t%6.2f\t%s\n", j, share_smp[j], (float) share_smp[j]*(100/total),
cats_description[j]);
}
}
}
if (null_flag) {
if ( background ) {
fprintf (lp,"NULL\t%6lu\t%6.2f\t%8lu %6.2f\n",null_count, (float) null_count * 100 / total
,null_count_map, (float) null_count_map * 100 / map_total);
} else {
fprintf (lp,"NULL\t%6lu\t%6.2f\n",null_count, (float) null_count * 100 / total);
}
}
if (uncat_flag) {
if ( background ) {
fprintf (lp,"NOCAT\t%6lu\t%6.2f\t%8lu %6.2f\n",nocat_count, (float) nocat_count * 100 / total
,nocat_count_map, (float) nocat_count_map * 100 / map_total);
} else {
fprintf (lp,"NOCAT\t%6lu\t%6.2f\n",nocat_count, (float) nocat_count * 100 / total);
}
}
if ( background) {
fprintf (lp,"TOTAL\t%6lu\t%6.2f\t%8lu %6.2f\n",(long) total, (float) 100, (long) map_total, (float) 100);
} else {
fprintf (lp,"TOTAL\t%6lu\t%6.2f\n",(long) total, (float) 100);
}
/* close cache and sites file; free buffers. */
GT_RC_close (cache);
G_free (cellbuf);
G_free (cache);
}
