#include "gfxdevice.h"
#include "gfxtools.h"
#include "gfxpoly.h"
+#include "mem.h"
+#ifdef INTERNAL_LIBART
#include "art/libart.h"
+#include "art/art_svp_intersect.h"
+#include "art/art_svp_ops.h"
+#else
+#include <libart_lgpl/libart.h>
+#include <libart_lgpl/art_svp_intersect.h>
+#include <libart_lgpl/art_svp_ops.h>
+#endif
+#include "log.h"
#include <assert.h>
#include <memory.h>
#include <math.h>
+#define PERTURBATE
+//#define SHEAR
+//#define DEBUG
+
+//----------------------------------------- svp renderer ----------------------------------------
+
+typedef struct {
+ int xmin;
+ int ymin;
+ int xmax;
+ int ymax;
+ int width;
+ int height;
+} intbbox_t;
+
+typedef struct _renderpoint
+{
+ double x;
+ signed char direction;
+} renderpoint_t;
+
+typedef struct _renderline
+{
+ renderpoint_t*points;
+ int size;
+ int num;
+} renderline_t;
+
+typedef struct _renderbuf
+{
+ intbbox_t bbox;
+ int width;
+ int height;
+ double zoom;
+ renderline_t*lines;
+} renderbuf_t;
+
+static inline void add_pixel(renderbuf_t*buf, double x, int y, signed char direction)
+{
+ renderpoint_t p;
+ p.x = x;
+ p.direction = direction;
+
+ if(x >= buf->bbox.xmax || y >= buf->bbox.ymax || y < buf->bbox.ymin)
+ return;
+ renderline_t*l = &buf->lines[y-buf->bbox.ymin];
+
+ if(l->num == l->size) {
+ l->size += 32;
+ l->points = (renderpoint_t*)rfx_realloc(l->points, l->size * sizeof(renderpoint_t));
+ }
+ l->points[l->num] = p;
+ l->num++;
+}
+#define CUT 0.5
+#define INT(x) ((int)((x)+16)-16)
+static void add_line(renderbuf_t*buf, double x1, double y1, double x2, double y2, signed char direction)
+{
+ x1 *= buf->zoom;
+ y1 *= buf->zoom;
+ x2 *= buf->zoom;
+ y2 *= buf->zoom;
+ double diffx, diffy;
+ double ny1, ny2, stepx;
+ if(y2 < y1) {
+ double x,y;
+ x = x1;x1 = x2;x2=x;
+ y = y1;y1 = y2;y2=y;
+ }
+ diffx = x2 - x1;
+ diffy = y2 - y1;
+ ny1 = INT(y1)+CUT;
+ ny2 = INT(y2)+CUT;
+
+ if(ny1 < y1) {
+ ny1 = INT(y1) + 1.0 + CUT;
+ }
+ if(ny2 >= y2) {
+ ny2 = INT(y2) - 1.0 + CUT;
+ }
+ if(ny1 > ny2)
+ return;
+
+ stepx = diffx/diffy;
+ x1 = x1 + (ny1-y1)*stepx;
+ x2 = x2 + (ny2-y2)*stepx;
+
+ int posy=INT(ny1);
+ int endy=INT(ny2);
+ double posx=0;
+ double startx = x1;
+
+ while(posy<=endy) {
+ double xx = startx + posx;
+ add_pixel(buf, xx, posy, direction);
+ posx+=stepx;
+ posy++;
+ }
+}
+
+static int compare_renderpoints(const void * _a, const void * _b)
+{
+ renderpoint_t*a = (renderpoint_t*)_a;
+ renderpoint_t*b = (renderpoint_t*)_b;
+ if(a->x < b->x) return -1;
+ if(a->x > b->x) return 1;
+ return 0;
+}
+
+static void fill_bitwise(unsigned char*line, int x1, int x2)
+{
+ int p1 = x1>>3;
+ int p2 = x2>>3;
+ int b1 = 0xff >> (x1&7);
+ int b2 = 0xff << (1+7-(x2&7));
+ if(p1==p2) {
+ line[p1] |= b1&b2;
+ } else {
+ line[p1] |= b1;
+ memset(&line[p1+1], 255, p2-(p1+1));
+ line[p2] = b2;
+ }
+}
+
+unsigned char* render_svp(ArtSVP*svp, intbbox_t*bbox, double zoom, ArtWindRule rule)
+{
+ renderbuf_t _buf, *buf=&_buf;
+ buf->width = (bbox->xmax - bbox->xmin);
+ buf->height = (bbox->ymax - bbox->ymin);
+ buf->bbox = *bbox;
+ buf->zoom = zoom;
+ int width8 = (buf->width+7) >> 3;
+ unsigned char* image = (unsigned char*)malloc(width8*buf->height);
+ memset(image, 0, width8*buf->height);
+
+ buf->lines = (renderline_t*)rfx_alloc(buf->height*sizeof(renderline_t));
+ int y;
+ for(y=0;y<buf->height;y++) {
+ memset(&buf->lines[y], 0, sizeof(renderline_t));
+ buf->lines[y].points = 0;
+ buf->lines[y].num = 0;
+ }
+
+ int t;
+ for(t=0;t<svp->n_segs;t++) {
+ ArtSVPSeg* seg = &svp->segs[t];
+ int s;
+ for(s=0;s<seg->n_points-1;s++) {
+ int dir = seg->dir? 1 : -1;
+ add_line(buf, seg->points[s].x, seg->points[s].y, seg->points[s+1].x, seg->points[s+1].y, dir);
+ }
+ }
+ for(y=0;y<buf->height;y++) {
+ renderpoint_t*points = buf->lines[y].points;
+ unsigned char*line = &image[width8*y];
+ int n;
+ int num = buf->lines[y].num;
+ int wind = 0;
+ qsort(points, num, sizeof(renderpoint_t), compare_renderpoints);
+ int lastx = 0;
+ int fill = 0;
+
+ for(n=0;n<num;n++) {
+ renderpoint_t*p = &points[n];
+ int x = (int)(p->x - bbox->xmin);
+
+ if(x < lastx)
+ x = lastx;
+ if(x > buf->width) {
+ break;
+ }
+ if(fill && x!=lastx) {
+ fill_bitwise(line, lastx, x);
+ }
+ wind += p->direction;
+ if(rule == ART_WIND_RULE_INTERSECT) {
+ fill = wind>=2;
+ } else if (rule == ART_WIND_RULE_NONZERO) {
+ fill = wind!=0;
+ } else if (rule == ART_WIND_RULE_ODDEVEN) {
+ fill = wind&1;
+ } else { // rule == ART_WIND_RULE_POSITIVE
+ fill = wind>=1;
+ }
+ lastx = x;
+ }
+ if(fill && lastx!=buf->width)
+ fill_bitwise(line, lastx, buf->width);
+ }
+
+ for(y=0;y<buf->height;y++) {
+ if(buf->lines[y].points) {
+ free(buf->lines[y].points);
+ }
+ memset(&buf->lines[y], 0, sizeof(renderline_t));
+ }
+ free(buf->lines);buf->lines=0;
+ return image;
+}
+
+#define MAX_WIDTH 8192
+#define MAX_HEIGHT 4096
+
+intbbox_t get_svp_bbox(ArtSVP*svp, double zoom)
+{
+ int t;
+ intbbox_t b = {0,0,0,0};
+ if(svp->n_segs && svp->segs[0].n_points) {
+ b.xmin = svp->segs[0].points[0].x;
+ b.ymin = svp->segs[0].points[0].y;
+ b.xmax = svp->segs[0].points[0].x;
+ b.ymax = svp->segs[0].points[0].y;
+ }
+ for(t=0;t<svp->n_segs;t++) {
+ ArtSVPSeg* seg = &svp->segs[t];
+ int s;
+ for(s=0;s<seg->n_points;s++) {
+ double x = seg->points[s].x*zoom;
+ double y = seg->points[s].y*zoom;
+ int x1 = floor(x);
+ int x2 = ceil(x);
+ int y1 = floor(y);
+ int y2 = ceil(y);
+ if(x1 < b.xmin) b.xmin = x1;
+ if(y1 < b.ymin) b.ymin = y1;
+ if(x2 > b.xmax) b.xmax = x2;
+ if(y2 > b.xmax) b.ymax = y2;
+ }
+ }
+ if(b.xmax > (int)(MAX_WIDTH*zoom))
+ b.xmax = (int)(MAX_WIDTH*zoom);
+ if(b.ymax > (int)(MAX_HEIGHT*zoom))
+ b.ymax = (int)(MAX_HEIGHT*zoom);
+ if(b.xmin < -(int)(MAX_WIDTH*zoom))
+ b.xmin = -(int)(MAX_WIDTH*zoom);
+ if(b.ymin < -(int)(MAX_HEIGHT*zoom))
+ b.ymin = -(int)(MAX_HEIGHT*zoom);
+
+ if(b.xmin > b.xmax)
+ b.xmin = b.xmax;
+ if(b.ymin > b.ymax)
+ b.ymin = b.ymax;
+
+ b.width = b.xmax - b.xmin;
+ b.height = b.ymax - b.ymin;
+ return b;
+}
+
+#define B11100000 0xe0
+#define B01110000 0x70
+#define B00111000 0x38
+#define B00011100 0x1c
+#define B00001110 0x0e
+#define B00000111 0x07
+#define B10000000 0x80
+#define B01000000 0x40
+#define B00100000 0x20
+#define B00010000 0x10
+#define B00001000 0x08
+#define B00000100 0x04
+#define B00000010 0x02
+#define B00000001 0x01
+
+int compare_bitmaps(intbbox_t*bbox, unsigned char*data1, unsigned char*data2)
+{
+ int similar = 0;
+ int x,y;
+ int height = bbox->height;
+ int width = bbox->width;
+ int width8 = (width+7) >> 3;
+ unsigned char*l1 = &data1[width8];
+ unsigned char*l2 = &data2[width8];
+ int fail = 0;
+ for(y=1;y<height-1;y++) {
+ for(x=0;x<width8;x++) {
+ unsigned a = l1[x-width8] & l1[x] & l1[x+width8];
+ unsigned b = l2[x-width8] & l2[x] & l2[x+width8];
+
+ if((a&B11100000) && !(l2[x]&B01000000))
+ fail == 1;
+ if((a&B01110000) && !(l2[x]&B00100000))
+ fail == 1;
+ if((a&B00111000) && !(l2[x]&B00010000))
+ fail == 1;
+ if((a&B00011100) && !(l2[x]&B00001000))
+ fail == 1;
+ if((a&B00001110) && !(l2[x]&B00000100))
+ fail == 1;
+ if((a&B00000111) && !(l2[x]&B00000010))
+ fail == 1;
+
+ if((b&B11100000) && !(l1[x]&B01000000))
+ fail == 1;
+ if((b&B01110000) && !(l1[x]&B00100000))
+ fail == 1;
+ if((b&B00111000) && !(l1[x]&B00010000))
+ fail == 1;
+ if((b&B00011100) && !(l1[x]&B00001000))
+ fail == 1;
+ if((b&B00001110) && !(l1[x]&B00000100))
+ fail == 1;
+ if((b&B00000111) && !(l1[x]&B00000010))
+ fail == 1;
+ }
+ l1 += width8;
+ l2 += width8;
+ }
+ return !fail;
+}
+
+
+//-----------------------------------------------------------------------------------------------
+
static ArtVpath* gfxline_to_ArtVpath(gfxline_t*line, char fill)
{
ArtVpath *vec = NULL;
while(l2) {
if(l2->type == gfx_moveTo) {
pos ++;
- } if(l2->type == gfx_lineTo) {
+ } else if(l2->type == gfx_lineTo) {
pos ++;
- } if(l2->type == gfx_splineTo) {
+ } else if(l2->type == gfx_splineTo) {
int parts = (int)(sqrt(fabs(l2->x-2*l2->sx+x) + fabs(l2->y-2*l2->sy+y))*subfraction);
if(!parts) parts = 1;
pos += parts + 1;
l2 = line;
while(l2) {
if(l2->type == gfx_moveTo) {
- vec[pos].code = ART_MOVETO;
+ vec[pos].code = ART_MOVETO_OPEN;
vec[pos].x = l2->x;
vec[pos].y = l2->y;
pos++;
} else if(l2->type == gfx_splineTo) {
int i;
int parts = (int)(sqrt(fabs(l2->x-2*l2->sx+x) + fabs(l2->y-2*l2->sy+y))*subfraction);
- double stepsize = parts?1.0/parts:0;
+ if(!parts) parts = 1;
+ double stepsize = 1.0/parts;
for(i=0;i<=parts;i++) {
double t = (double)i*stepsize;
vec[pos].code = ART_LINETO;
y = l2->y;
l2 = l2->next;
}
- vec[pos].code = ART_END;
-
+ vec[pos++].code = ART_END;
+ assert(pos == len);
+
if(!fill) {
/* Fix "dotted" lines. Those are lines where singular points are created
by a moveto x,y lineto x,y combination. We "fix" these by shifting the
*/
int t;
for(t=0;vec[t].code!=ART_END;t++) {
- if(t>0 && vec[t-1].code==ART_MOVETO && vec[t].code==ART_LINETO
- && vec[t+1].code!=ART_LINETO
- && vec[t-1].x == vec[t].x
- && vec[t-1].y == vec[t].y) {
+ if(t>0 && (vec[t-1].code==ART_MOVETO_OPEN || vec[t-1].code==ART_MOVETO)
+ && vec[t].code==ART_LINETO && vec[t+1].code!=ART_LINETO &&
+ vec[t-1].x == vec[t].x &&
+ vec[t-1].y == vec[t].y) {
vec[t].x += 0.01;
}
- x = vec[t].x;
- y = vec[t].y;
}
}
/* Find adjacent identical points. If an ajdacent pair of identical
- points is found, the moveto is removed (unless both are movetos).
- So moveto x,y lineto x,y becomes lineto x,y
+ points is found, the second is removed.
+ So moveto x,y lineto x,y becomes moveto x,y
lineto x,y lineto x,y becomes lineto x,y
lineto x,y moveto x,y becomes lineto x,y
moveto x,y moveto x,y becomes moveto x,y
+ lineto x,y lineto x2,y2 becomes lineto x2,y2 (if dir(x,y) ~= dir(x2,y2))
*/
- int t;
- while(t < pos)
+ pos = 0;
+ int outpos = 0;
+ while(1)
{
- int t = 1;
- if ((vec[t-1].x == vec[t].x) && (vec[t-1].y == vec[t].y)) {
- // adjacent identical points; remove one
- int type = ART_MOVETO;
- if(vec[t-1].code==ART_LINETO || vec[t].code==ART_LINETO)
- type = ART_LINETO;
- memcpy(&vec[t], &vec[t+1], sizeof(vec[0]) * (pos - t));
- vec[t].code = type;
- pos--;
- } else {
- t++;
- }
+ if(vec[pos].code == ART_END) {
+ vec[outpos++] = vec[pos++];
+ break;
+ }
+
+ char samedir = 0, samepoint = 0;
+ if(outpos) {
+ double dx = vec[pos].x-vec[pos-1].x;
+ double dy = vec[pos].y-vec[pos-1].y;
+ /*if(pos<len-1 && vec[pos].code == ART_LINETO && vec[pos+1].code == ART_LINETO) {
+ double dx2 = vec[pos+1].x-vec[pos].x;
+ double dy2 = vec[pos+1].y-vec[pos].y;
+ if(fabs(dx*dy2 - dy*dx2) < 0.0001 && dx*dx2 + dy*dy2 >= 0) {
+ samedir=1;
+ }
+ }*/
+ if(fabs(dx) + fabs(dy) < 0.0001) {
+ samepoint=1;
+ }
+ }
+ if(!samepoint && !samedir) {
+ vec[outpos++] = vec[pos++];
+ } else {
+ pos++; // skip
+ }
}
- /* adjacency remover disabled for now, pending code inspection */
return vec;
+}
- // Check for further non-adjacent identical points. We don't want any
- // points other than the first and last points to exactly match.
- //
- // If we do find matching points, move the second point slightly. This
- // currently moves the duplicate 2% towards the midpoint of its neighbours
- // (easier to calculate than 2% down the perpendicular to the line joining
- // the neighbours) but limiting the change to 0.1 twips to avoid visual
- // problems when the shapes are large. Note that there is no minimum
- // change: if the neighbouring points are colinear and equally spaced,
- // e.g. they were generated as part of a straight spline, then the
- // duplicate point may not actually move.
- //
- // The scan for duplicates algorithm is quadratic in the number of points:
- // there's probably a better method but the numbers of points is generally
- // small so this will do for now.
- {
- int i = 1, j;
- for(; i < (pos - 1); ++i)
- {
- for (j = 0; j < i; ++j)
- {
- if ((vec[i].x == vec[j].x)
- && (vec[i].y == vec[j].y))
- {
- // points match; shuffle point
- double dx = (vec[i-1].x + vec[i+1].x - (vec[i].x * 2.0)) / 100.0;
- double dy = (vec[i-1].y + vec[i+1].y - (vec[i].y * 2.0)) / 100.0;
- double dxxyy = (dx*dx) + (dy*dy);
- if (dxxyy > 0.01)
- {
- // This is more than 0.1 twip's distance; scale down
- double dscale = sqrt(dxxyy) * 10.0;
- dx /= dscale;
- dy /= dscale;
- };
- vec[i].x += dx;
- vec[i].y += dy;
- break;
- }
- }
- }
+static void shear_svp(ArtSVP*svp, double v)
+{
+ /* do a "shearing" on the polygon. We do this to eliminate all
+ horizontal lines (which libart can't handle properly, even though
+ it tries). */
+
+ int t;
+ for(t=0;t<svp->n_segs;t++) {
+ ArtSVPSeg* seg = &svp->segs[t];
+ int s;
+ for(s=0;s<seg->n_points;s++) {
+ ArtPoint* point = &seg->points[s];
+ point->y += point->x*v;
+ }
}
+}
- return vec;
+static double find_shear_value(ArtSVP*svp)
+{
+ /* We try random values until we find one
+ where there's no slope below a given value, or if that fails,
+ at least no slope of 0 */
+
+ double v = 0;
+ int tries = 0;
+ while(1) {
+ char fail = 0;
+ int t;
+ for(t=0;t<svp->n_segs;t++) {
+ ArtSVPSeg* seg = &svp->segs[t];
+ int s;
+ for(s=0;s<seg->n_points-1;s++) {
+ ArtPoint* p1 = &seg->points[s];
+ ArtPoint* p2 = &seg->points[s+1];
+ double dx = p2->x - p1->x;
+ double dy = p2->y - p1->y;
+ dy += dx*v;
+ if(dy==0) {
+ fail = 1;
+ break;
+ }
+ if(tries<100 && dx && fabs(dy / dx) < 1e-5) {
+ fail = 1;
+ break;
+ }
+ }
+ if(fail)
+ break;
+ }
+ if(!fail)
+ break;
+#ifdef HAVE_LRAND48
+ v = lrand48() / 2000000000.0;;
+#elif HAVE_RAND
+ v = rand() / 2000000000.0;
+#else
+#error "no lrand48()/rand() implementation found"
+#endif
+ tries++;
+ }
+ return v;
}
void show_path(ArtSVP*path)
printf("\n");
}
-ArtSVP* gfxfillToSVP(gfxline_t*line, int perturb)
+
+typedef struct svp_segment_part
+{
+ double y1;
+ double y2;
+ char active;
+} svp_segment_part_t;
+
+int compare_double(const void*_y1, const void*_y2)
+{
+ const double*y1 = _y1;
+ const double*y2 = _y2;
+ if(*y1<*y2) return -1;
+ if(*y1>*y2) return 1;
+ else return 0;
+}
+
+int compare_seg_start(const void*_s1, const void*_s2)
+{
+ svp_segment_part_t*s1 = *(svp_segment_part_t**)_s1;
+ svp_segment_part_t*s2 = *(svp_segment_part_t**)_s2;
+ if(s1->y1 < s2->y1) return -1;
+ if(s1->y1 > s2->y1) return 1;
+ else return 0;
+}
+
+int compare_seg_end(const void*_s1, const void*_s2)
+{
+ svp_segment_part_t*s1 = *(svp_segment_part_t**)_s1;
+ svp_segment_part_t*s2 = *(svp_segment_part_t**)_s2;
+ if(s1->y2 < s2->y2) return -1;
+ if(s1->y2 > s2->y2) return 1;
+ else return 0;
+}
+
+void clean_svp(ArtSVP*svp)
+{
+ int t;
+ int oldpoints = 0;
+ int newpoints = 0;
+ int oldsegs = 0;
+ int newsegs = 0;
+ for(t=0;t<svp->n_segs;t++) {
+ ArtSVPSeg* seg = &svp->segs[t];
+ int p;
+ int pos=0;
+ double lasty = 0;
+ oldpoints += seg->n_points;
+ for(p=0;p<seg->n_points;p++) {
+ ArtPoint* p1 = &seg->points[p];
+ if(!p || lasty!=p1->y) {
+ seg->points[pos] = seg->points[p];
+ pos++;
+ lasty = p1->y;
+ }
+ }
+ seg->n_points = pos;
+ newpoints += seg->n_points;
+ }
+ int pos = 0;
+ oldsegs = svp->n_segs;
+ for(t=0;t<svp->n_segs;t++) {
+ if(svp->segs[t].n_points > 1) {
+ svp->segs[pos++] = svp->segs[t];
+ }
+ }
+ svp->n_segs = pos;
+ newsegs = svp->n_segs;
+ if(newsegs < oldsegs || newpoints < oldpoints) {
+ msg("<verbose> Simplified polygon from %d points to %d points, %d to %d segs",
+ oldpoints, newpoints, oldsegs, newsegs);
+ }
+}
+
+int check_svp(ArtSVP*svp)
+{
+ /* count number of y coordinates and segs */
+ int t;
+ int num_points = 0;
+ int num_segs = 0;
+ for(t=0;t<svp->n_segs;t++) {
+ if(!svp->segs[t].n_points) {
+ msg("<error> svp contains segment with zero points\n");
+ return 0;
+ }
+ num_points += svp->segs[t].n_points;
+ num_segs += svp->segs[t].n_points - 1;
+ }
+
+ /* create segs and ys */
+ double*y = malloc(sizeof(double)*num_points);
+ svp_segment_part_t*segs = malloc(sizeof(svp_segment_part_t)*num_segs);
+ svp_segment_part_t**seg_start = malloc(sizeof(svp_segment_part_t*)*num_segs);
+ svp_segment_part_t**seg_end = malloc(sizeof(svp_segment_part_t*)*num_segs);
+
+ int c1=0;
+ num_segs = 0;
+ for(t=0;t<svp->n_segs;t++) {
+ ArtSVPSeg* seg = &svp->segs[t];
+ int p;
+ for(p=0;p<seg->n_points;p++) {
+ y[c1++] = seg->points[p].y;
+ assert(c1 <= num_points);
+ }
+ for(p=0;p<seg->n_points-1;p++) {
+ ArtPoint* p1 = &seg->points[p];
+ ArtPoint* p2 = &seg->points[p+1];
+
+ if(p1->y >= p2->y) {
+ if(p1->y > p2->y) {
+ msg("<error> bad svp, y in seg is non-increasing %.16f -> %.16f\n", p1->y, p2->y);
+ }
+ } else {
+ segs[num_segs].y1 = p1->y;
+ segs[num_segs].y2 = p2->y;
+ segs[num_segs].active = 0;
+ seg_start[num_segs] = &segs[num_segs];
+ seg_end[num_segs] = &segs[num_segs];
+ num_segs++;
+ }
+ }
+ }
+
+ qsort(y, num_points, sizeof(double), compare_double);
+ qsort(seg_start, num_segs, sizeof(svp_segment_part_t*), compare_seg_start);
+ qsort(seg_end, num_segs, sizeof(svp_segment_part_t*), compare_seg_end);
+
+ double lasty = num_points?y[0]+1:0;
+ int num_active = 0;
+ int bleed = 0;
+ double bleedy1=0,bleedy2 = 0;
+ for(t=0;t<num_points;t++) {
+ if(lasty == y[t])
+ continue;
+ int s;
+ for(s=0;s<num_segs;s++) {
+ if(segs[s].y1==y[t]) {
+ /* segment is starting */
+ segs[s].active = 1;
+ num_active++;
+ } else if(segs[s].y2==y[t]) {
+ segs[s].active = 0;
+ num_active--;
+ }
+ }
+ if(num_active&1) {
+ bleed = num_active;
+ bleedy1 = y[t];
+ } else {
+ if(bleed) {
+ bleedy2 = y[t];
+ break;
+ }
+ }
+ lasty = y[t];
+ }
+ if(bleed) {
+ msg("<verbose> svp bleeds from y=%.16f to y=%.16f (%d/%d active segments)\n",
+ bleedy1, bleedy2,
+ bleed, num_segs);
+ free(y);free(segs);free(seg_start);free(seg_end);
+ return 0;
+ }
+
+ free(y);
+ free(segs);
+ free(seg_start);
+ free(seg_end);
+
+ return 1;
+}
+
+void write_svp_postscript(const char*filename, ArtSVP*svp)
+{
+ if(!svp)
+ return;
+ FILE*fi = fopen(filename, "wb");
+ int i, j;
+ double xmin=0,ymin=0,xmax=0,ymax=0;
+ fprintf(fi, "%% begin\n");
+ for (i = 0; i < svp->n_segs; i++) {
+ for (j = 0; j < svp->segs[i].n_points; j++) {
+ double x = svp->segs[i].points[j].x;
+ double y = svp->segs[i].points[j].y;
+ if(i==0 && j==0) {
+ xmin = xmax = x;
+ ymin = ymax = y;
+ } else {
+ if(x < xmin) xmin = x;
+ if(x > xmax) xmax = x;
+ if(y < ymin) ymin = y;
+ if(y > ymax) ymax = y;
+ }
+ }
+ }
+ if(xmax == xmin) xmax=xmin+1;
+ if(ymax == ymin) ymax=ymin+1;
+
+ for (i = 0; i < svp->n_segs; i++)
+ {
+ fprintf(fi, "%g setgray\n", svp->segs[i].dir ? 0.7 : 0);
+ for (j = 0; j < svp->segs[i].n_points; j++)
+ {
+ //fprintf(fi, "%g %g %s\n",
+ // 20 + 550*(svp->segs[i].points[j].x-xmin)/(xmax-xmin),
+ // 820 - 800*(svp->segs[i].points[j].y-ymin)/(ymax-ymin),
+ // j ? "lineto" : "moveto");
+ fprintf(fi, "%.32f %.32f %s\n",
+ svp->segs[i].points[j].x,
+ svp->segs[i].points[j].y,
+ j ? "lineto" : "moveto");
+ }
+ fprintf(fi, "stroke\n");
+ }
+
+ fprintf(fi, "showpage\n");
+ fclose(fi);
+}
+
+void write_vpath_postscript(const char*filename, ArtVpath*path)
+{
+ FILE*fi = fopen(filename, "wb");
+ int i, j;
+ double xmin=0,ymin=0,xmax=0,ymax=0;
+ fprintf(fi, "%% begin\n");
+ ArtVpath*p = path;
+ char first = 1;
+ while(p->code != ART_END) {
+ if(p->code == ART_MOVETO || p->code == ART_MOVETO_OPEN) {
+ if(!first)
+ fprintf(fi, "stroke\n");
+ first = 0;
+ fprintf(fi, "0.0 setgray\n");
+ fprintf(fi, "%.32f %.32f moveto\n", p->x, p->y);
+ } else {
+ fprintf(fi, "%.32f %.32f lineto\n", p->x, p->y);
+ }
+ p++;
+ }
+ if(!first)
+ fprintf(fi, "stroke\n");
+ fprintf(fi, "showpage\n");
+ fclose(fi);
+}
+
+void write_gfxline_postscript(const char*filename, gfxline_t*line)
+{
+ FILE*fi = fopen(filename, "wb");
+ int i, j;
+ fprintf(fi, "%% begin\n");
+ char first = 1;
+ while(line) {
+ if(line->type == gfx_moveTo) {
+ if(!first)
+ fprintf(fi, "stroke\n");
+ first = 0;
+ fprintf(fi, "0.0 setgray\n");
+ fprintf(fi, "%.32f %.32f moveto\n", line->x, line->y);
+ } else {
+ fprintf(fi, "%.32f %.32f lineto\n", line->x, line->y);
+ }
+ line = line->next;
+ }
+ if(!first)
+ fprintf(fi, "stroke\n");
+ fprintf(fi, "showpage\n");
+ fclose(fi);
+}
+
+static int vpath_len(ArtVpath*svp)
+{
+ int len = 0;
+ while(svp->code != ART_END) {
+ svp ++;
+ len ++;
+ }
+ return len;
+}
+
+int gfxline_len(gfxline_t*line)
+{
+ gfxline_t*i = line;
+ int len = 0;
+ while(i) {
+ len ++;
+ i = i->next;
+ }
+ return len;
+}
+
+static ArtVpath*pvpath= 0;
+static int cmppos(const void*_p1, const void*_p2)
+{
+ int*p1 = (int*)_p1;
+ int*p2 = (int*)_p2;
+ ArtVpath*v1 = &pvpath[*p1];
+ ArtVpath*v2 = &pvpath[*p2];
+ if(v1->y < v2->y)
+ return -1;
+ else if(v1->y > v2->y)
+ return 1;
+ else if(v1->x < v2->x)
+ return -2;
+ else if(v1->x > v2->x)
+ return 2;
+ else
+ return 0;
+}
+
+#define PERTURBATION 2e-3
+static void my_perturb(ArtVpath*vpath)
+{
+ int t;
+ int len = vpath_len(vpath);
+ int*pos = (int*)malloc(len*sizeof(int));
+ for(t=0;t<len;t++)
+ pos[t] = t;
+ pvpath = vpath;
+ qsort(pos, len, sizeof(int), cmppos);
+ t=0;
+ while(t<len) {
+ int t2=t+1;
+ while(t2<len && !cmppos(&pos[t],&pos[t2])) {
+ t2++;
+ }
+
+ double dx = (PERTURBATION * rand ()) / RAND_MAX - PERTURBATION * 0.5;
+ double dy = (PERTURBATION * rand ()) / RAND_MAX - PERTURBATION * 0.5;
+ int s;
+ for(s=t;s<t2;s++) {
+ vpath[pos[s]].x += dx;
+ vpath[pos[s]].y += dy;
+ }
+ t = t2;
+ }
+ free(pos);
+ pvpath = 0;
+}
+
+static ArtSVP* gfxfillToSVP(gfxline_t*line, int perturb)
{
ArtVpath* vec = gfxline_to_ArtVpath(line, 1);
+ msg("<verbose> Casting gfxline of %d segments (%d line segments) to a gfxpoly", gfxline_len(line), vpath_len(vec));
+
if(perturb) {
- ArtVpath* vec2 = art_vpath_perturb(vec);
- free(vec);
- vec = vec2;
+ //ArtVpath* vec2 = art_vpath_perturb(vec);
+ //free(vec);
+ //vec = vec2;
+ my_perturb(vec);
}
ArtSVP *svp = art_svp_from_vpath(vec);
free(vec);
- // We need to make sure that the SVP we now have bounds an area (i.e. the
- // source line wound anticlockwise) rather than excludes an area (i.e. the
- // line wound clockwise). It seems that PDF (or xpdf) is less strict about
- // this for bitmaps than it is for fill areas.
- //
- // To check this, we'll sum the cross products of all pairs of adjacent
- // lines. If the result is positive, the direction is correct; if not, we
- // need to reverse the sense of the SVP generated. The easiest way to do
- // this is to flip the up/down flags of all the segments.
- //
- // This is approximate; since the gfxline_t structure can contain any
- // combination of moveTo, lineTo and splineTo in any order, not all pairs
- // of lines in the shape that share a point need be described next to each
- // other in the sequence. For ease, we'll consider only pairs of lines
- // stored as lineTos and splineTos without intervening moveTos.
- //
- // TODO is this a valid algorithm? My vector maths is rusty.
- //
- // It may be more correct to instead reverse the line before we feed it
- // into gfxfilltoSVP. However, this seems equivalent and is easier to
- // implement!
- double total_cross_product = 0.0;
- gfxline_t* cursor = line;
- if (cursor != NULL)
- {
- double x_last = cursor->x;
- double y_last = cursor->y;
- cursor = cursor->next;
-
- while((cursor != NULL) && (cursor->next != NULL))
- {
- if (((cursor->type == gfx_lineTo) || (cursor->type == gfx_splineTo))
- && ((cursor->next->type == gfx_lineTo) || (cursor->next->type == gfx_splineTo)))
- {
- // Process these lines
- //
- // In this space (x right, y down) the cross-product is
- // (x1 * y0) - (x0 * y1)
- double x0 = cursor->x - x_last;
- double y0 = cursor->y - y_last;
- double x1 = cursor->next->x - cursor->x;
- double y1 = cursor->next->y - cursor->y;
- total_cross_product += (x1 * y0) - (x0 * y1);
- }
+ return svp;
+}
- x_last = cursor->x;
- y_last = cursor->y;
- cursor = cursor->next;
- }
+//#ifdef SHEAR
+// double shear = find_shear_value(svp);
+// gfxline_t*line = gfxpoly_to_gfxline((gfxpoly_t*)svp);
+// gfxline_t*l = line;
+// while(l) {
+// l->y += l->x*shear;
+// l->sy += l->sx*shear;
+// l= l->next;
+// }
+// svp = (ArtSVP*)gfxpoly_fillToPoly(line);
+// printf("shearing svp by %.16f\n", shear);
+//#endif
+// ....
+//#ifdef SHEAR
+// art_svp_free(svp);
+// shear_svp(result, -shear);
+//#endif
+
+
+ArtSVP* run_intersector(ArtSVP*svp, ArtWindRule rule)
+{
+ ArtSvpWriter * swr = art_svp_writer_rewind_new(rule);
+
+ double zoom = 1.0;
+ intbbox_t bbox = get_svp_bbox(svp, zoom);
+
+ art_svp_intersector(svp, swr);
+ ArtSVP*result = art_svp_writer_rewind_reap(swr);
+ clean_svp(result);
+ if(!check_svp(result)) {
+ current_svp = result;
+ art_report_error(); // might set art_error_in_intersector
+ } else {
+ msg("<verbose> Comparing polygon renderings of size %dx%d and %dx%d", bbox.width, bbox.height, bbox.width, bbox.height);
+ unsigned char*data1 = render_svp(svp, &bbox, zoom, rule);
+ unsigned char*data2 = render_svp(result, &bbox, zoom, ART_WIND_RULE_ODDEVEN);
+ if(!compare_bitmaps(&bbox, data1, data2)) {
+ msg("<verbose> Bad SVP rewinding result- polygons don't match");
+ current_svp = result;
+ art_report_error(); // might set art_error_in_intersector
+ }
+ free(data1);
+ free(data2);
}
- if (total_cross_product < 0.0)
- {
- int i = 0;
- for(; i < svp->n_segs; ++i)
- {
- if (svp->segs[i].dir != 0)
- svp->segs[i].dir = 0;
- else
- svp->segs[i].dir = 1;
- }
+
+ if(art_error_in_intersector) {
+ msg("<verbose> Error in polygon processing");
+ art_svp_free(result);
+ art_error_in_intersector=0;
+ return 0;
}
- return svp;
+ return result;
}
gfxline_t* gfxpoly_to_gfxline(gfxpoly_t*poly)
int size = 0;
int t;
int pos = 0;
+
+ msg("<verbose> Casting polygon of %d segments back to gfxline", svp->n_segs);
+
for(t=0;t<svp->n_segs;t++) {
- size += svp->segs[t].n_points + 1;
+ size += svp->segs[t].n_points;
}
+ size = size + 1;
gfxline_t* lines = (gfxline_t*)rfx_alloc(sizeof(gfxline_t)*size);
for(t=0;t<svp->n_segs;t++) {
gfxpoly_t* gfxpoly_fillToPoly(gfxline_t*line)
{
+ /* I'm not sure whether doing perturbation here is actually
+ a good idea- if that line has been run through the machinery
+ several times already, it might be safer to leave it alone,
+ since it should already be in a format libart can handle */
+#ifdef PERTURBATE
ArtSVP* svp = gfxfillToSVP(line, 1);
- if (svp->n_segs > 500)
- {
- int lineParts = 0;
- gfxline_t* lineCursor = line;
- while(lineCursor != NULL)
- {
- if(lineCursor->type != gfx_moveTo) ++lineParts;
- lineCursor = lineCursor->next;
- }
- fprintf(stderr, "arts_fill abandonning shape with %d segments (%d line parts)\n", svp->n_segs, lineParts);
- art_svp_free(svp);
- return (gfxpoly_t*)gfxpoly_strokeToPoly(0, 0, gfx_capButt, gfx_joinMiter, 0);
+#else
+ ArtSVP* svp = gfxfillToSVP(line, 0);
+#endif
+
+#ifdef DEBUG
+ char filename[80];
+ static int counter = 0;
+ sprintf(filename, "svp%d.ps", counter);
+ write_svp_postscript(filename, svp);
+ sprintf(filename, "gfxline%d.ps", counter);
+ write_gfxline_postscript(filename, line);
+#endif
+
+ /* we do xor-filling by default, so dir is always 1
+ (actually for oddeven rewinding it makes no difference, but
+ it's "cleaner")
+ */
+ int t;
+ for(t=0; t<svp->n_segs; t++) {
+ svp->segs[t].dir = 1;
}
- ArtSVP* svp2 = art_svp_rewind_uncrossed(art_svp_uncross(svp),ART_WIND_RULE_ODDEVEN);
- art_svp_free(svp);svp=svp2;
+
+ /* for some reason, we need to rewind / self-intersect the polygons that gfxfillToSVP
+ returns- art probably uses a different fill mode (circular?) for vpaths */
+ ArtSVP*svp_uncrossed=0;
+
+#ifdef DEBUG
+ sprintf(filename, "svp%d_in.ps", counter);
+ write_svp_postscript(filename, svp);
+ counter++;
+#endif
+
+ svp_uncrossed = run_intersector(svp, ART_WIND_RULE_ODDEVEN);
+
+ art_svp_free(svp);
+ svp=svp_uncrossed;
+
return (gfxpoly_t*)svp;
}
gfxpoly_t* gfxpoly_intersect(gfxpoly_t*poly1, gfxpoly_t*poly2)
{
+ ArtSvpWriter *swr;
+
+ static int counter = 0;
+
ArtSVP* svp1 = (ArtSVP*)poly1;
ArtSVP* svp2 = (ArtSVP*)poly2;
+ msg("<verbose> Intersecting two polygons of %d and %d segments", svp1->n_segs, svp2->n_segs);
+
+#ifdef DEBUG
+ char filename[80];
+ sprintf(filename, "isvp%d_src1.ps", counter);
+ write_svp_postscript(filename, svp1);
+ sprintf(filename, "isvp%d_src2.ps", counter);
+ write_svp_postscript(filename, svp2);
+#endif
+
+ ArtSVP* svp3 = art_svp_merge (svp1, svp2);
+
+#ifdef DEBUG
+ sprintf(filename, "isvp%d_src.ps", counter);
+ write_svp_postscript(filename, svp3);
+#endif
+
+ //write_svp_postscript("svp.ps", svp3);
+ ArtSVP*svp_new = run_intersector(svp3, ART_WIND_RULE_INTERSECT);
+
+ art_free (svp3); /* shallow free because svp3 contains shared segments */
+
+#ifdef DEBUG
+ sprintf(filename, "isvp%d.ps", counter);
+ write_svp_postscript(filename, svp_new);
+#endif
+
+ counter++;
+
+ //write_svp_postscript("svp_new.ps", svp_new);
- ArtSVP* svp = art_svp_intersect(svp1, svp2);
- return (gfxpoly_t*)svp;
+ return (gfxpoly_t*)svp_new;
}
gfxpoly_t* gfxpoly_union(gfxpoly_t*poly1, gfxpoly_t*poly2)
{
ArtSVP* svp1 = (ArtSVP*)poly1;
ArtSVP* svp2 = (ArtSVP*)poly2;
+ msg("<verbose> Unifying two polygons of %d and %d segments", svp1->n_segs, svp2->n_segs);
ArtSVP* svp = art_svp_union(svp1, svp2);
return (gfxpoly_t*)svp;
gfxpoly_t* gfxpoly_strokeToPoly(gfxline_t*line, gfxcoord_t width, gfx_capType cap_style, gfx_joinType joint_style, double miterLimit)
{
ArtVpath* vec = gfxline_to_ArtVpath(line, 0);
+ msg("<verbose> Casting gfxline of %d segments to a stroke-polygon", gfxline_len(line));
+
+ ArtVpath* vec2 = art_vpath_perturb(vec);
+ free(vec);
+ vec = vec2;
ArtSVP *svp = art_svp_vpath_stroke (vec,
(joint_style==gfx_joinMiter)?ART_PATH_STROKE_JOIN_MITER:
gfxline_t* gfxline_circularToEvenOdd(gfxline_t*line)
{
+ msg("<verbose> Converting circular-filled gfxline of %d segments to even-odd filled gfxline", gfxline_len(line));
ArtSVP* svp = gfxfillToSVP(line, 1);
- ArtSVP* svp2 = art_svp_rewind_uncrossed(art_svp_uncross(svp),ART_WIND_RULE_POSITIVE);
- gfxline_t* result = gfxpoly_to_gfxline((gfxpoly_t*)svp2);
+
+ /* TODO: ART_WIND_RULE_POSITIVE means that a shape is visible if
+ positive and negative line segments add up to something positive.
+ I *think* that clockwise fill in PDF is defined in a way, however,
+ that the *last* shape's direction will determine whether something
+ is filled */
+ ArtSVP* svp_rewinded;
+
+ svp_rewinded = run_intersector(svp, ART_WIND_RULE_POSITIVE);
+ if(!svp_rewinded) {
+ art_svp_free(svp);
+ return 0;
+ }
+
+ gfxline_t* result = gfxpoly_to_gfxline((gfxpoly_t*)svp_rewinded);
art_svp_free(svp);
- art_svp_free(svp2);
+ art_svp_free(svp_rewinded);
return result;
}