7390 lines
174 KiB
C
7390 lines
174 KiB
C
/*
|
|
* see COPYRIGHT
|
|
*/
|
|
|
|
#include <stdio.h>
|
|
#include <stdlib.h>
|
|
#include <string.h>
|
|
#include <sys/types.h>
|
|
#include <sys/stat.h>
|
|
#include <fcntl.h>
|
|
#include <time.h>
|
|
#include <ctype.h>
|
|
#include <math.h>
|
|
|
|
#ifndef WINDOWS
|
|
# include <netinet/in.h>
|
|
# include <unistd.h>
|
|
#else
|
|
# include "windows.h"
|
|
#endif
|
|
|
|
#include "ttf.h"
|
|
#include "pt1.h"
|
|
#include "global.h"
|
|
|
|
/* big and small values for comparisons */
|
|
#define FBIGVAL (1e20)
|
|
#define FEPS (100000./FBIGVAL)
|
|
|
|
/* names of the axes */
|
|
#define X 0
|
|
#define Y 1
|
|
|
|
/* the GENTRY extension structure used in fforceconcise() */
|
|
|
|
struct gex_con {
|
|
double d[2 /*X, Y*/]; /* sizes of curve */
|
|
double sin2; /* squared sinus of the angle to the next gentry */
|
|
double len2; /* squared distance between the endpoints */
|
|
|
|
/* number of reference dots taken from each curve */
|
|
#define NREFDOTS 3
|
|
|
|
double dots[NREFDOTS][2]; /* reference dots */
|
|
|
|
int flags; /* flags for gentry and tits joint to the next gentry */
|
|
/* a vertical or horizontal line may be in 2 quadrants at once */
|
|
#define GEXF_QUL 0x00000001 /* in up-left quadrant */
|
|
#define GEXF_QUR 0x00000002 /* in up-right quadrant */
|
|
#define GEXF_QDR 0x00000004 /* in down-right quadrant */
|
|
#define GEXF_QDL 0x00000008 /* in down-left quadrant */
|
|
#define GEXF_QMASK 0x0000000F /* quadrant mask */
|
|
|
|
/* if a line is nearly vertical or horizontal, we remember that idealized quartant too */
|
|
#define GEXF_QTO_IDEAL(f) (((f)&0xF)<<4)
|
|
#define GEXF_QFROM_IDEAL(f) (((f)&0xF0)>>4)
|
|
#define GEXF_IDQ_L 0x00000090 /* left */
|
|
#define GEXF_IDQ_R 0x00000060 /* right */
|
|
#define GEXF_IDQ_U 0x00000030 /* up */
|
|
#define GEXF_IDQ_D 0x000000C0 /* down */
|
|
|
|
/* possibly can be joined with conditions:
|
|
* (in order of increasing preference, the numeric order is important)
|
|
*/
|
|
#define GEXF_JLINE 0x00000100 /* into one line */
|
|
#define GEXF_JIGN 0x00000200 /* if one entry's tangent angle is ignored */
|
|
#define GEXF_JID 0x00000400 /* if one entry is idealized to hor/vert */
|
|
#define GEXF_JFLAT 0x00000800 /* if one entry is flattened */
|
|
#define GEXF_JGOOD 0x00001000 /* perfectly, no additional maodifications */
|
|
|
|
#define GEXF_JMASK 0x00001F00 /* the mask of all above */
|
|
#define GEXF_JCVMASK 0x00001E00 /* the mask of all above except JLINE */
|
|
|
|
/* which entry needs to be modified for conditional joining */
|
|
#define GEXF_JIGN1 0x00002000
|
|
#define GEXF_JIGN2 0x00004000
|
|
#define GEXF_JIGNDIR(dir) (GEXF_JIGN1<<(dir))
|
|
#define GEXF_JID1 0x00008000
|
|
#define GEXF_JID2 0x00010000
|
|
#define GEXF_JIDDIR(dir) (GEXF_JID1<<(dir))
|
|
#define GEXF_JFLAT1 0x00020000
|
|
#define GEXF_JFLAT2 0x00040000
|
|
#define GEXF_JFLATDIR(dir) (GEXF_JFLAT1<<(dir))
|
|
|
|
#define GEXF_VERT 0x00100000 /* is nearly vertical */
|
|
#define GEXF_HOR 0x00200000 /* is nearly horizontal */
|
|
#define GEXF_FLAT 0x00400000 /* is nearly flat */
|
|
|
|
#define GEXF_VDOTS 0x01000000 /* the dots are valid */
|
|
|
|
signed char isd[2 /*X,Y*/]; /* signs of the sizes */
|
|
};
|
|
typedef struct gex_con GEX_CON;
|
|
|
|
/* convenience macros */
|
|
#define X_CON(ge) ((GEX_CON *)((ge)->ext))
|
|
#define X_CON_D(ge) (X_CON(ge)->d)
|
|
#define X_CON_DX(ge) (X_CON(ge)->d[0])
|
|
#define X_CON_DY(ge) (X_CON(ge)->d[1])
|
|
#define X_CON_ISD(ge) (X_CON(ge)->isd)
|
|
#define X_CON_ISDX(ge) (X_CON(ge)->isd[0])
|
|
#define X_CON_ISDY(ge) (X_CON(ge)->isd[1])
|
|
#define X_CON_SIN2(ge) (X_CON(ge)->sin2)
|
|
#define X_CON_LEN2(ge) (X_CON(ge)->len2)
|
|
#define X_CON_F(ge) (X_CON(ge)->flags)
|
|
|
|
/* performance statistics about guessing the concise curves */
|
|
static int ggoodcv=0, ggoodcvdots=0, gbadcv=0, gbadcvdots=0;
|
|
|
|
int stdhw, stdvw; /* dominant stems widths */
|
|
int stemsnaph[12], stemsnapv[12]; /* most typical stem width */
|
|
|
|
int bluevalues[14];
|
|
int nblues;
|
|
int otherblues[10];
|
|
int notherb;
|
|
int bbox[4]; /* the FontBBox array */
|
|
double italic_angle;
|
|
|
|
GLYPH *glyph_list;
|
|
int encoding[ENCTABSZ]; /* inverse of glyph[].char_no */
|
|
int kerning_pairs = 0;
|
|
|
|
/* prototypes */
|
|
static void fixcvdir( GENTRY * ge, int dir);
|
|
static void fixcvends( GENTRY * ge);
|
|
static int fgetcvdir( GENTRY * ge);
|
|
static int igetcvdir( GENTRY * ge);
|
|
static int fiszigzag( GENTRY *ge);
|
|
static int iiszigzag( GENTRY *ge);
|
|
static GENTRY * freethisge( GENTRY *ge);
|
|
static void addgeafter( GENTRY *oge, GENTRY *nge );
|
|
static GENTRY * newgentry( int flags);
|
|
static void debugstems( char *name, STEM * hstems, int nhs, STEM * vstems, int nvs);
|
|
static int addbluestems( STEM *s, int n);
|
|
static void sortstems( STEM * s, int n);
|
|
static int stemoverlap( STEM * s1, STEM * s2);
|
|
static int steminblue( STEM *s);
|
|
static void markbluestems( STEM *s, int nold);
|
|
static int joinmainstems( STEM * s, int nold, int useblues);
|
|
static void joinsubstems( STEM * s, short *pairs, int nold, int useblues);
|
|
static void fixendpath( GENTRY *ge);
|
|
static void fdelsmall( GLYPH *g, double minlen);
|
|
static void alloc_gex_con( GENTRY *ge);
|
|
static double fjointsin2( GENTRY *ge1, GENTRY *ge2);
|
|
static double fcvarea( GENTRY *ge);
|
|
static double fcvval( GENTRY *ge, int axis, double t);
|
|
static void fsampledots( GENTRY *ge, double dots[][2], int ndots);
|
|
static void fnormalizege( GENTRY *ge);
|
|
static void fanalyzege( GENTRY *ge);
|
|
static void fanalyzejoint( GENTRY *ge);
|
|
static void fconcisecontour( GLYPH *g, GENTRY *ge);
|
|
static double fclosegap( GENTRY *from, GENTRY *to, int axis,
|
|
double gap, double *ret);
|
|
|
|
int
|
|
isign(
|
|
int x
|
|
)
|
|
{
|
|
if (x > 0)
|
|
return 1;
|
|
else if (x < 0)
|
|
return -1;
|
|
else
|
|
return 0;
|
|
}
|
|
|
|
int
|
|
fsign(
|
|
double x
|
|
)
|
|
{
|
|
if (x > 0.0)
|
|
return 1;
|
|
else if (x < 0.0)
|
|
return -1;
|
|
else
|
|
return 0;
|
|
}
|
|
|
|
static GENTRY *
|
|
newgentry(
|
|
int flags
|
|
)
|
|
{
|
|
GENTRY *ge;
|
|
|
|
ge = calloc(1, sizeof(GENTRY));
|
|
|
|
if (ge == 0) {
|
|
fprintf(stderr, "***** Memory allocation error *****\n");
|
|
exit(255);
|
|
}
|
|
ge->stemid = -1;
|
|
ge->flags = flags;
|
|
/* the rest is set to 0 by calloc() */
|
|
return ge;
|
|
}
|
|
|
|
/*
|
|
* Routines to print out Postscript functions with optimization
|
|
*/
|
|
|
|
void
|
|
rmoveto(
|
|
int dx,
|
|
int dy
|
|
)
|
|
{
|
|
if (optimize && dx == 0)
|
|
fprintf(pfa_file, "%d vmoveto\n", dy);
|
|
else if (optimize && dy == 0)
|
|
fprintf(pfa_file, "%d hmoveto\n", dx);
|
|
else
|
|
fprintf(pfa_file, "%d %d rmoveto\n", dx, dy);
|
|
}
|
|
|
|
void
|
|
rlineto(
|
|
int dx,
|
|
int dy
|
|
)
|
|
{
|
|
if (optimize && dx == 0 && dy == 0) /* for special pathologic
|
|
* case */
|
|
return;
|
|
else if (optimize && dx == 0)
|
|
fprintf(pfa_file, "%d vlineto\n", dy);
|
|
else if (optimize && dy == 0)
|
|
fprintf(pfa_file, "%d hlineto\n", dx);
|
|
else
|
|
fprintf(pfa_file, "%d %d rlineto\n", dx, dy);
|
|
}
|
|
|
|
void
|
|
rrcurveto(
|
|
int dx1,
|
|
int dy1,
|
|
int dx2,
|
|
int dy2,
|
|
int dx3,
|
|
int dy3
|
|
)
|
|
{
|
|
/* first two ifs are for crazy cases that occur surprisingly often */
|
|
if (optimize && dx1 == 0 && dx2 == 0 && dx3 == 0)
|
|
rlineto(0, dy1 + dy2 + dy3);
|
|
else if (optimize && dy1 == 0 && dy2 == 0 && dy3 == 0)
|
|
rlineto(dx1 + dx2 + dx3, 0);
|
|
else if (optimize && dy1 == 0 && dx3 == 0)
|
|
fprintf(pfa_file, "%d %d %d %d hvcurveto\n",
|
|
dx1, dx2, dy2, dy3);
|
|
else if (optimize && dx1 == 0 && dy3 == 0)
|
|
fprintf(pfa_file, "%d %d %d %d vhcurveto\n",
|
|
dy1, dx2, dy2, dx3);
|
|
else
|
|
fprintf(pfa_file, "%d %d %d %d %d %d rrcurveto\n",
|
|
dx1, dy1, dx2, dy2, dx3, dy3);
|
|
}
|
|
|
|
void
|
|
closepath(void)
|
|
{
|
|
fprintf(pfa_file, "closepath\n");
|
|
}
|
|
|
|
/*
|
|
* Many of the path processing routines exist (or will exist) in
|
|
* both floating-point and integer version. Fimally most of the
|
|
* processing will go in floating point and the integer processing
|
|
* will become legacy.
|
|
* The names of floating routines start with f, names of integer
|
|
* routines start with i, and those old routines existing in one
|
|
* version only have no such prefix at all.
|
|
*/
|
|
|
|
/*
|
|
** Routine that checks integrity of the path, for debugging
|
|
*/
|
|
|
|
void
|
|
assertpath(
|
|
GENTRY * from,
|
|
char *file,
|
|
int line,
|
|
char *name
|
|
)
|
|
{
|
|
GENTRY *first, *pe, *ge;
|
|
int isfloat;
|
|
|
|
if(from==0)
|
|
return;
|
|
isfloat = (from->flags & GEF_FLOAT);
|
|
pe = from->prev;
|
|
for (ge = from; ge != 0; pe = ge, ge = ge->next) {
|
|
if( (ge->flags & GEF_FLOAT) ^ isfloat ) {
|
|
fprintf(stderr, "**! assertpath: called from %s line %d (%s) ****\n", file, line, name);
|
|
fprintf(stderr, "float flag changes from %s to %s at 0x%p (type %c, prev type %c)\n",
|
|
(isfloat ? "TRUE" : "FALSE"), (isfloat ? "FALSE" : "TRUE"), ge, ge->type, pe->type);
|
|
abort();
|
|
}
|
|
if (pe != ge->prev) {
|
|
fprintf(stderr, "**! assertpath: called from %s line %d (%s) ****\n", file, line, name);
|
|
fprintf(stderr, "unidirectional chain 0x%x -next-> 0x%x -prev-> 0x%x \n",
|
|
pe, ge, ge->prev);
|
|
abort();
|
|
}
|
|
|
|
switch(ge->type) {
|
|
case GE_MOVE:
|
|
break;
|
|
case GE_PATH:
|
|
if (ge->prev == 0) {
|
|
fprintf(stderr, "**! assertpath: called from %s line %d (%s) ****\n", file, line, name);
|
|
fprintf(stderr, "empty path at 0x%x \n", ge);
|
|
abort();
|
|
}
|
|
break;
|
|
case GE_LINE:
|
|
case GE_CURVE:
|
|
if(ge->frwd->bkwd != ge) {
|
|
fprintf(stderr, "**! assertpath: called from %s line %d (%s) ****\n", file, line, name);
|
|
fprintf(stderr, "unidirectional chain 0x%x -frwd-> 0x%x -bkwd-> 0x%x \n",
|
|
ge, ge->frwd, ge->frwd->bkwd);
|
|
abort();
|
|
}
|
|
if(ge->prev->type == GE_MOVE) {
|
|
first = ge;
|
|
if(ge->bkwd->next->type != GE_PATH) {
|
|
fprintf(stderr, "**! assertpath: called from %s line %d (%s) ****\n", file, line, name);
|
|
fprintf(stderr, "broken first backlink 0x%x -bkwd-> 0x%x -next-> 0x%x \n",
|
|
ge, ge->bkwd, ge->bkwd->next);
|
|
abort();
|
|
}
|
|
}
|
|
if(ge->next->type == GE_PATH) {
|
|
if(ge->frwd != first) {
|
|
fprintf(stderr, "**! assertpath: called from %s line %d (%s) ****\n", file, line, name);
|
|
fprintf(stderr, "broken loop 0x%x -...-> 0x%x -frwd-> 0x%x \n",
|
|
first, ge, ge->frwd);
|
|
abort();
|
|
}
|
|
}
|
|
break;
|
|
}
|
|
|
|
}
|
|
}
|
|
|
|
void
|
|
assertisfloat(
|
|
GLYPH *g,
|
|
char *msg
|
|
)
|
|
{
|
|
if( !(g->flags & GF_FLOAT) ) {
|
|
fprintf(stderr, "**! Glyph %s is not float: %s\n", g->name, msg);
|
|
abort();
|
|
}
|
|
if(g->lastentry) {
|
|
if( !(g->lastentry->flags & GEF_FLOAT) ) {
|
|
fprintf(stderr, "**! Glyphs %s last entry is int: %s\n", g->name, msg);
|
|
abort();
|
|
}
|
|
}
|
|
}
|
|
|
|
void
|
|
assertisint(
|
|
GLYPH *g,
|
|
char *msg
|
|
)
|
|
{
|
|
if( (g->flags & GF_FLOAT) ) {
|
|
fprintf(stderr, "**! Glyph %s is not int: %s\n", g->name, msg);
|
|
abort();
|
|
}
|
|
if(g->lastentry) {
|
|
if( (g->lastentry->flags & GEF_FLOAT) ) {
|
|
fprintf(stderr, "**! Glyphs %s last entry is float: %s\n", g->name, msg);
|
|
abort();
|
|
}
|
|
}
|
|
}
|
|
|
|
|
|
/*
|
|
* Routines to save the generated data about glyph
|
|
*/
|
|
|
|
void
|
|
fg_rmoveto(
|
|
GLYPH * g,
|
|
double x,
|
|
double y)
|
|
{
|
|
GENTRY *oge;
|
|
|
|
if (ISDBG(BUILDG))
|
|
fprintf(stderr, "%s: f rmoveto(%g, %g)\n", g->name, x, y);
|
|
|
|
assertisfloat(g, "adding float MOVE");
|
|
|
|
if ((oge = g->lastentry) != 0) {
|
|
if (oge->type == GE_MOVE) { /* just eat up the first move */
|
|
oge->fx3 = x;
|
|
oge->fy3 = y;
|
|
} else if (oge->type == GE_LINE || oge->type == GE_CURVE) {
|
|
fprintf(stderr, "Glyph %s: MOVE in middle of path\n", g->name);
|
|
} else {
|
|
GENTRY *nge;
|
|
|
|
nge = newgentry(GEF_FLOAT);
|
|
nge->type = GE_MOVE;
|
|
nge->fx3 = x;
|
|
nge->fy3 = y;
|
|
|
|
oge->next = nge;
|
|
nge->prev = oge;
|
|
g->lastentry = nge;
|
|
}
|
|
} else {
|
|
GENTRY *nge;
|
|
|
|
nge = newgentry(GEF_FLOAT);
|
|
nge->type = GE_MOVE;
|
|
nge->fx3 = x;
|
|
nge->fy3 = y;
|
|
nge->bkwd = (GENTRY*)&g->entries;
|
|
g->entries = g->lastentry = nge;
|
|
}
|
|
|
|
if (0 && ISDBG(BUILDG))
|
|
dumppaths(g, NULL, NULL);
|
|
}
|
|
|
|
void
|
|
ig_rmoveto(
|
|
GLYPH * g,
|
|
int x,
|
|
int y)
|
|
{
|
|
GENTRY *oge;
|
|
|
|
if (ISDBG(BUILDG))
|
|
fprintf(stderr, "%s: i rmoveto(%d, %d)\n", g->name, x, y);
|
|
|
|
assertisint(g, "adding int MOVE");
|
|
|
|
if ((oge = g->lastentry) != 0) {
|
|
if (oge->type == GE_MOVE) { /* just eat up the first move */
|
|
oge->ix3 = x;
|
|
oge->iy3 = y;
|
|
} else if (oge->type == GE_LINE || oge->type == GE_CURVE) {
|
|
fprintf(stderr, "Glyph %s: MOVE in middle of path, ignored\n", g->name);
|
|
} else {
|
|
GENTRY *nge;
|
|
|
|
nge = newgentry(0);
|
|
nge->type = GE_MOVE;
|
|
nge->ix3 = x;
|
|
nge->iy3 = y;
|
|
|
|
oge->next = nge;
|
|
nge->prev = oge;
|
|
g->lastentry = nge;
|
|
}
|
|
} else {
|
|
GENTRY *nge;
|
|
|
|
nge = newgentry(0);
|
|
nge->type = GE_MOVE;
|
|
nge->ix3 = x;
|
|
nge->iy3 = y;
|
|
nge->bkwd = (GENTRY*)&g->entries;
|
|
g->entries = g->lastentry = nge;
|
|
}
|
|
|
|
}
|
|
|
|
void
|
|
fg_rlineto(
|
|
GLYPH * g,
|
|
double x,
|
|
double y)
|
|
{
|
|
GENTRY *oge, *nge;
|
|
|
|
if (ISDBG(BUILDG))
|
|
fprintf(stderr, "%s: f rlineto(%g, %g)\n", g->name, x, y);
|
|
|
|
assertisfloat(g, "adding float LINE");
|
|
|
|
nge = newgentry(GEF_FLOAT);
|
|
nge->type = GE_LINE;
|
|
nge->fx3 = x;
|
|
nge->fy3 = y;
|
|
|
|
if ((oge = g->lastentry) != 0) {
|
|
if (x == oge->fx3 && y == oge->fy3) { /* empty line */
|
|
/* ignore it or we will get in troubles later */
|
|
free(nge);
|
|
return;
|
|
}
|
|
if (g->path == 0) {
|
|
g->path = nge;
|
|
nge->bkwd = nge->frwd = nge;
|
|
} else {
|
|
oge->frwd = nge;
|
|
nge->bkwd = oge;
|
|
g->path->bkwd = nge;
|
|
nge->frwd = g->path;
|
|
}
|
|
|
|
oge->next = nge;
|
|
nge->prev = oge;
|
|
g->lastentry = nge;
|
|
} else {
|
|
WARNING_1 fprintf(stderr, "Glyph %s: LINE outside of path\n", g->name);
|
|
free(nge);
|
|
}
|
|
|
|
if (0 && ISDBG(BUILDG))
|
|
dumppaths(g, NULL, NULL);
|
|
}
|
|
|
|
void
|
|
ig_rlineto(
|
|
GLYPH * g,
|
|
int x,
|
|
int y)
|
|
{
|
|
GENTRY *oge, *nge;
|
|
|
|
if (ISDBG(BUILDG))
|
|
fprintf(stderr, "%s: i rlineto(%d, %d)\n", g->name, x, y);
|
|
|
|
assertisint(g, "adding int LINE");
|
|
|
|
nge = newgentry(0);
|
|
nge->type = GE_LINE;
|
|
nge->ix3 = x;
|
|
nge->iy3 = y;
|
|
|
|
if ((oge = g->lastentry) != 0) {
|
|
if (x == oge->ix3 && y == oge->iy3) { /* empty line */
|
|
/* ignore it or we will get in troubles later */
|
|
free(nge);
|
|
return;
|
|
}
|
|
if (g->path == 0) {
|
|
g->path = nge;
|
|
nge->bkwd = nge->frwd = nge;
|
|
} else {
|
|
oge->frwd = nge;
|
|
nge->bkwd = oge;
|
|
g->path->bkwd = nge;
|
|
nge->frwd = g->path;
|
|
}
|
|
|
|
oge->next = nge;
|
|
nge->prev = oge;
|
|
g->lastentry = nge;
|
|
} else {
|
|
WARNING_1 fprintf(stderr, "Glyph %s: LINE outside of path\n", g->name);
|
|
free(nge);
|
|
}
|
|
|
|
}
|
|
|
|
void
|
|
fg_rrcurveto(
|
|
GLYPH * g,
|
|
double x1,
|
|
double y1,
|
|
double x2,
|
|
double y2,
|
|
double x3,
|
|
double y3)
|
|
{
|
|
GENTRY *oge, *nge;
|
|
|
|
oge = g->lastentry;
|
|
|
|
if (ISDBG(BUILDG))
|
|
fprintf(stderr, "%s: f rrcurveto(%g, %g, %g, %g, %g, %g)\n"
|
|
,g->name, x1, y1, x2, y2, x3, y3);
|
|
|
|
assertisfloat(g, "adding float CURVE");
|
|
|
|
if (oge && oge->fx3 == x1 && x1 == x2 && x2 == x3) /* check if it's
|
|
* actually a line */
|
|
fg_rlineto(g, x1, y3);
|
|
else if (oge && oge->fy3 == y1 && y1 == y2 && y2 == y3)
|
|
fg_rlineto(g, x3, y1);
|
|
else {
|
|
nge = newgentry(GEF_FLOAT);
|
|
nge->type = GE_CURVE;
|
|
nge->fx1 = x1;
|
|
nge->fy1 = y1;
|
|
nge->fx2 = x2;
|
|
nge->fy2 = y2;
|
|
nge->fx3 = x3;
|
|
nge->fy3 = y3;
|
|
|
|
if (oge != 0) {
|
|
if (x3 == oge->fx3 && y3 == oge->fy3) {
|
|
free(nge); /* consider this curve empty */
|
|
/* ignore it or we will get in troubles later */
|
|
return;
|
|
}
|
|
if (g->path == 0) {
|
|
g->path = nge;
|
|
nge->bkwd = nge->frwd = nge;
|
|
} else {
|
|
oge->frwd = nge;
|
|
nge->bkwd = oge;
|
|
g->path->bkwd = nge;
|
|
nge->frwd = g->path;
|
|
}
|
|
|
|
oge->next = nge;
|
|
nge->prev = oge;
|
|
g->lastentry = nge;
|
|
} else {
|
|
WARNING_1 fprintf(stderr, "Glyph %s: CURVE outside of path\n", g->name);
|
|
free(nge);
|
|
}
|
|
}
|
|
|
|
if (0 && ISDBG(BUILDG))
|
|
dumppaths(g, NULL, NULL);
|
|
}
|
|
|
|
void
|
|
ig_rrcurveto(
|
|
GLYPH * g,
|
|
int x1,
|
|
int y1,
|
|
int x2,
|
|
int y2,
|
|
int x3,
|
|
int y3)
|
|
{
|
|
GENTRY *oge, *nge;
|
|
|
|
oge = g->lastentry;
|
|
|
|
if (ISDBG(BUILDG))
|
|
fprintf(stderr, "%s: i rrcurveto(%d, %d, %d, %d, %d, %d)\n"
|
|
,g->name, x1, y1, x2, y2, x3, y3);
|
|
|
|
assertisint(g, "adding int CURVE");
|
|
|
|
if (oge && oge->ix3 == x1 && x1 == x2 && x2 == x3) /* check if it's
|
|
* actually a line */
|
|
ig_rlineto(g, x1, y3);
|
|
else if (oge && oge->iy3 == y1 && y1 == y2 && y2 == y3)
|
|
ig_rlineto(g, x3, y1);
|
|
else {
|
|
nge = newgentry(0);
|
|
nge->type = GE_CURVE;
|
|
nge->ix1 = x1;
|
|
nge->iy1 = y1;
|
|
nge->ix2 = x2;
|
|
nge->iy2 = y2;
|
|
nge->ix3 = x3;
|
|
nge->iy3 = y3;
|
|
|
|
if (oge != 0) {
|
|
if (x3 == oge->ix3 && y3 == oge->iy3) {
|
|
free(nge); /* consider this curve empty */
|
|
/* ignore it or we will get in troubles later */
|
|
return;
|
|
}
|
|
if (g->path == 0) {
|
|
g->path = nge;
|
|
nge->bkwd = nge->frwd = nge;
|
|
} else {
|
|
oge->frwd = nge;
|
|
nge->bkwd = oge;
|
|
g->path->bkwd = nge;
|
|
nge->frwd = g->path;
|
|
}
|
|
|
|
oge->next = nge;
|
|
nge->prev = oge;
|
|
g->lastentry = nge;
|
|
} else {
|
|
WARNING_1 fprintf(stderr, "Glyph %s: CURVE outside of path\n", g->name);
|
|
free(nge);
|
|
}
|
|
}
|
|
}
|
|
|
|
void
|
|
g_closepath(
|
|
GLYPH * g
|
|
)
|
|
{
|
|
GENTRY *oge, *nge;
|
|
|
|
if (ISDBG(BUILDG))
|
|
fprintf(stderr, "%s: closepath\n", g->name);
|
|
|
|
oge = g->lastentry;
|
|
|
|
if (g->path == 0) {
|
|
WARNING_1 fprintf(stderr, "Warning: **** closepath on empty path in glyph \"%s\" ****\n",
|
|
g->name);
|
|
if (oge == 0) {
|
|
WARNING_1 fprintf(stderr, "No previois entry\n");
|
|
} else {
|
|
WARNING_1 fprintf(stderr, "Previous entry type: %c\n", oge->type);
|
|
if (oge->type == GE_MOVE) {
|
|
g->lastentry = oge->prev;
|
|
if (oge->prev == 0)
|
|
g->entries = 0;
|
|
else
|
|
g->lastentry->next = 0;
|
|
free(oge);
|
|
}
|
|
}
|
|
return;
|
|
}
|
|
|
|
nge = newgentry(oge->flags & GEF_FLOAT); /* keep the same type */
|
|
nge->type = GE_PATH;
|
|
|
|
g->path = 0;
|
|
|
|
oge->next = nge;
|
|
nge->prev = oge;
|
|
g->lastentry = nge;
|
|
|
|
if (0 && ISDBG(BUILDG))
|
|
dumppaths(g, NULL, NULL);
|
|
}
|
|
|
|
/*
|
|
* * SB * Routines to smooth and fix the glyphs
|
|
*/
|
|
|
|
/*
|
|
** we don't want to see the curves with coinciding middle and
|
|
** outer points
|
|
*/
|
|
|
|
static void
|
|
fixcvends(
|
|
GENTRY * ge
|
|
)
|
|
{
|
|
int dx, dy;
|
|
int x0, y0, x1, y1, x2, y2, x3, y3;
|
|
|
|
if (ge->type != GE_CURVE)
|
|
return;
|
|
|
|
if(ge->flags & GEF_FLOAT) {
|
|
fprintf(stderr, "**! fixcvends(0x%x) on floating entry, ABORT\n", ge);
|
|
abort(); /* dump core */
|
|
}
|
|
|
|
x0 = ge->prev->ix3;
|
|
y0 = ge->prev->iy3;
|
|
x1 = ge->ix1;
|
|
y1 = ge->iy1;
|
|
x2 = ge->ix2;
|
|
y2 = ge->iy2;
|
|
x3 = ge->ix3;
|
|
y3 = ge->iy3;
|
|
|
|
|
|
/* look at the start of the curve */
|
|
if (x1 == x0 && y1 == y0) {
|
|
dx = x2 - x1;
|
|
dy = y2 - y1;
|
|
|
|
if (dx == 0 && dy == 0
|
|
|| x2 == x3 && y2 == y3) {
|
|
/* Oops, we actually have a straight line */
|
|
/*
|
|
* if it's small, we hope that it will get optimized
|
|
* later
|
|
*/
|
|
if (abs(x3 - x0) <= 2 || abs(y3 - y0) <= 2) {
|
|
ge->ix1 = x3;
|
|
ge->iy1 = y3;
|
|
ge->ix2 = x0;
|
|
ge->iy2 = y0;
|
|
} else {/* just make it a line */
|
|
ge->type = GE_LINE;
|
|
}
|
|
} else {
|
|
if (abs(dx) < 4 && abs(dy) < 4) { /* consider it very
|
|
* small */
|
|
ge->ix1 = x2;
|
|
ge->iy1 = y2;
|
|
} else if (abs(dx) < 8 && abs(dy) < 8) { /* consider it small */
|
|
ge->ix1 += dx / 2;
|
|
ge->iy1 += dy / 2;
|
|
} else {
|
|
ge->ix1 += dx / 4;
|
|
ge->iy1 += dy / 4;
|
|
}
|
|
/* make sure that it's still on the same side */
|
|
if (abs(x3 - x0) * abs(dy) < abs(y3 - y0) * abs(dx)) {
|
|
if (abs(x3 - x0) * abs(ge->iy1 - y0) > abs(y3 - y0) * abs(ge->ix1 - x0))
|
|
ge->ix1 += isign(dx);
|
|
} else {
|
|
if (abs(x3 - x0) * abs(ge->iy1 - y0) < abs(y3 - y0) * abs(ge->ix1 - x0))
|
|
ge->iy1 += isign(dy);
|
|
}
|
|
|
|
ge->ix2 += (x3 - x2) / 8;
|
|
ge->iy2 += (y3 - y2) / 8;
|
|
/* make sure that it's still on the same side */
|
|
if (abs(x3 - x0) * abs(y3 - y2) < abs(y3 - y0) * abs(x3 - x2)) {
|
|
if (abs(x3 - x0) * abs(y3 - ge->iy2) > abs(y3 - y0) * abs(x3 - ge->ix2))
|
|
ge->iy1 -= isign(y3 - y2);
|
|
} else {
|
|
if (abs(x3 - x0) * abs(y3 - ge->iy2) < abs(y3 - y0) * abs(x3 - ge->ix2))
|
|
ge->ix1 -= isign(x3 - x2);
|
|
}
|
|
|
|
}
|
|
} else if (x2 == x3 && y2 == y3) {
|
|
dx = x1 - x2;
|
|
dy = y1 - y2;
|
|
|
|
if (dx == 0 && dy == 0) {
|
|
/* Oops, we actually have a straight line */
|
|
/*
|
|
* if it's small, we hope that it will get optimized
|
|
* later
|
|
*/
|
|
if (abs(x3 - x0) <= 2 || abs(y3 - y0) <= 2) {
|
|
ge->ix1 = x3;
|
|
ge->iy1 = y3;
|
|
ge->ix2 = x0;
|
|
ge->iy2 = y0;
|
|
} else {/* just make it a line */
|
|
ge->type = GE_LINE;
|
|
}
|
|
} else {
|
|
if (abs(dx) < 4 && abs(dy) < 4) { /* consider it very
|
|
* small */
|
|
ge->ix2 = x1;
|
|
ge->iy2 = y1;
|
|
} else if (abs(dx) < 8 && abs(dy) < 8) { /* consider it small */
|
|
ge->ix2 += dx / 2;
|
|
ge->iy2 += dy / 2;
|
|
} else {
|
|
ge->ix2 += dx / 4;
|
|
ge->iy2 += dy / 4;
|
|
}
|
|
/* make sure that it's still on the same side */
|
|
if (abs(x3 - x0) * abs(dy) < abs(y3 - y0) * abs(dx)) {
|
|
if (abs(x3 - x0) * abs(ge->iy2 - y3) > abs(y3 - y0) * abs(ge->ix2 - x3))
|
|
ge->ix2 += isign(dx);
|
|
} else {
|
|
if (abs(x3 - x0) * abs(ge->iy2 - y3) < abs(y3 - y0) * abs(ge->ix2 - x3))
|
|
ge->iy2 += isign(dy);
|
|
}
|
|
|
|
ge->ix1 += (x0 - x1) / 8;
|
|
ge->iy1 += (y0 - y1) / 8;
|
|
/* make sure that it's still on the same side */
|
|
if (abs(x3 - x0) * abs(y0 - y1) < abs(y3 - y0) * abs(x0 - x1)) {
|
|
if (abs(x3 - x0) * abs(y0 - ge->iy1) > abs(y3 - y0) * abs(x0 - ge->ix1))
|
|
ge->iy1 -= isign(y0 - y1);
|
|
} else {
|
|
if (abs(x3 - x0) * abs(y0 - ge->iy1) < abs(y3 - y0) * abs(x0 - ge->ix1))
|
|
ge->ix1 -= isign(x0 - x1);
|
|
}
|
|
|
|
}
|
|
}
|
|
}
|
|
|
|
/*
|
|
** After transformations we want to make sure that the resulting
|
|
** curve is going in the same quadrant as the original one,
|
|
** because rounding errors introduced during transformations
|
|
** may make the result completeley wrong.
|
|
**
|
|
** `dir' argument describes the direction of the original curve,
|
|
** it is the superposition of two values for the front and
|
|
** rear ends of curve:
|
|
**
|
|
** >EQUAL - goes over the line connecting the ends
|
|
** =EQUAL - coincides with the line connecting the ends
|
|
** <EQUAL - goes under the line connecting the ends
|
|
**
|
|
** See CVDIR_* for exact definitions.
|
|
*/
|
|
|
|
static void
|
|
fixcvdir(
|
|
GENTRY * ge,
|
|
int dir
|
|
)
|
|
{
|
|
int a, b, c, d;
|
|
double kk, kk1, kk2;
|
|
int changed;
|
|
int fdir, rdir;
|
|
|
|
if(ge->flags & GEF_FLOAT) {
|
|
fprintf(stderr, "**! fixcvdir(0x%x) on floating entry, ABORT\n", ge);
|
|
abort(); /* dump core */
|
|
}
|
|
|
|
fdir = (dir & CVDIR_FRONT) - CVDIR_FEQUAL;
|
|
if ((dir & CVDIR_REAR) == CVDIR_RSAME)
|
|
rdir = fdir; /* we need only isign, exact value doesn't matter */
|
|
else
|
|
rdir = (dir & CVDIR_REAR) - CVDIR_REQUAL;
|
|
|
|
fixcvends(ge);
|
|
|
|
c = isign(ge->ix3 - ge->prev->ix3); /* note the direction of
|
|
* curve */
|
|
d = isign(ge->iy3 - ge->prev->iy3);
|
|
|
|
a = ge->iy3 - ge->prev->iy3;
|
|
b = ge->ix3 - ge->prev->ix3;
|
|
kk = fabs(a == 0 ? (b == 0 ? 1. : 100000.) : ((double) b / (double) a));
|
|
a = ge->iy1 - ge->prev->iy3;
|
|
b = ge->ix1 - ge->prev->ix3;
|
|
kk1 = fabs(a == 0 ? (b == 0 ? 1. : 100000.) : ((double) b / (double) a));
|
|
a = ge->iy3 - ge->iy2;
|
|
b = ge->ix3 - ge->ix2;
|
|
kk2 = fabs(a == 0 ? (b == 0 ? 1. : 100000.) : ((double) b / (double) a));
|
|
|
|
changed = 1;
|
|
while (changed) {
|
|
if (ISDBG(FIXCVDIR)) {
|
|
/* for debugging */
|
|
fprintf(stderr, "fixcvdir %d %d (%d %d %d %d %d %d) %f %f %f\n",
|
|
fdir, rdir,
|
|
ge->ix1 - ge->prev->ix3,
|
|
ge->iy1 - ge->prev->iy3,
|
|
ge->ix2 - ge->ix1,
|
|
ge->iy2 - ge->iy1,
|
|
ge->ix3 - ge->ix2,
|
|
ge->iy3 - ge->iy2,
|
|
kk1, kk, kk2);
|
|
}
|
|
changed = 0;
|
|
|
|
if (fdir > 0) {
|
|
if (kk1 > kk) { /* the front end has problems */
|
|
if (c * (ge->ix1 - ge->prev->ix3) > 0) {
|
|
ge->ix1 -= c;
|
|
changed = 1;
|
|
} if (d * (ge->iy2 - ge->iy1) > 0) {
|
|
ge->iy1 += d;
|
|
changed = 1;
|
|
}
|
|
/* recalculate the coefficients */
|
|
a = ge->iy3 - ge->prev->iy3;
|
|
b = ge->ix3 - ge->prev->ix3;
|
|
kk = fabs(a == 0 ? (b == 0 ? 1. : 100000.) : ((double) b / (double) a));
|
|
a = ge->iy1 - ge->prev->iy3;
|
|
b = ge->ix1 - ge->prev->ix3;
|
|
kk1 = fabs(a == 0 ? (b == 0 ? 1. : 100000.) : ((double) b / (double) a));
|
|
}
|
|
} else if (fdir < 0) {
|
|
if (kk1 < kk) { /* the front end has problems */
|
|
if (c * (ge->ix2 - ge->ix1) > 0) {
|
|
ge->ix1 += c;
|
|
changed = 1;
|
|
} if (d * (ge->iy1 - ge->prev->iy3) > 0) {
|
|
ge->iy1 -= d;
|
|
changed = 1;
|
|
}
|
|
/* recalculate the coefficients */
|
|
a = ge->iy1 - ge->prev->iy3;
|
|
b = ge->ix1 - ge->prev->ix3;
|
|
kk1 = fabs(a == 0 ? (b == 0 ? 1. : 100000.) : ((double) b / (double) a));
|
|
a = ge->iy3 - ge->prev->iy3;
|
|
b = ge->ix3 - ge->prev->ix3;
|
|
kk = fabs(a == 0 ? (b == 0 ? 1. : 100000.) : ((double) b / (double) a));
|
|
}
|
|
}
|
|
if (rdir > 0) {
|
|
if (kk2 < kk) { /* the rear end has problems */
|
|
if (c * (ge->ix2 - ge->ix1) > 0) {
|
|
ge->ix2 -= c;
|
|
changed = 1;
|
|
} if (d * (ge->iy3 - ge->iy2) > 0) {
|
|
ge->iy2 += d;
|
|
changed = 1;
|
|
}
|
|
/* recalculate the coefficients */
|
|
a = ge->iy3 - ge->prev->iy3;
|
|
b = ge->ix3 - ge->prev->ix3;
|
|
kk = fabs(a == 0 ? (b == 0 ? 1. : 100000.) : ((double) b / (double) a));
|
|
a = ge->iy3 - ge->iy2;
|
|
b = ge->ix3 - ge->ix2;
|
|
kk2 = fabs(a == 0 ? (b == 0 ? 1. : 100000.) : ((double) b / (double) a));
|
|
}
|
|
} else if (rdir < 0) {
|
|
if (kk2 > kk) { /* the rear end has problems */
|
|
if (c * (ge->ix3 - ge->ix2) > 0) {
|
|
ge->ix2 += c;
|
|
changed = 1;
|
|
} if (d * (ge->iy2 - ge->iy1) > 0) {
|
|
ge->iy2 -= d;
|
|
changed = 1;
|
|
}
|
|
/* recalculate the coefficients */
|
|
a = ge->iy3 - ge->prev->iy3;
|
|
b = ge->ix3 - ge->prev->ix3;
|
|
kk = fabs(a == 0 ? (b == 0 ? 1. : 100000.) : ((double) b / (double) a));
|
|
a = ge->iy3 - ge->iy2;
|
|
b = ge->ix3 - ge->ix2;
|
|
kk2 = fabs(a == 0 ? (b == 0 ? 1. : 100000.) : ((double) b / (double) a));
|
|
}
|
|
}
|
|
}
|
|
fixcvends(ge);
|
|
}
|
|
|
|
/* Get the directions of ends of curve for further usage */
|
|
|
|
/* expects that the previous element is also float */
|
|
|
|
static int
|
|
fgetcvdir(
|
|
GENTRY * ge
|
|
)
|
|
{
|
|
double a, b;
|
|
double k, k1, k2;
|
|
int dir = 0;
|
|
|
|
if( !(ge->flags & GEF_FLOAT) ) {
|
|
fprintf(stderr, "**! fgetcvdir(0x%x) on int entry, ABORT\n", ge);
|
|
abort(); /* dump core */
|
|
}
|
|
|
|
a = fabs(ge->fy3 - ge->prev->fy3);
|
|
b = fabs(ge->fx3 - ge->prev->fx3);
|
|
k = a < FEPS ? (b < FEPS ? 1. : 100000.) : ( b / a);
|
|
|
|
a = fabs(ge->fy1 - ge->prev->fy3);
|
|
b = fabs(ge->fx1 - ge->prev->fx3);
|
|
if(a < FEPS) {
|
|
if(b < FEPS) {
|
|
a = fabs(ge->fy2 - ge->prev->fy3);
|
|
b = fabs(ge->fx2 - ge->prev->fx3);
|
|
k1 = a < FEPS ? (b < FEPS ? k : 100000.) : ( b / a);
|
|
} else
|
|
k1 = FBIGVAL;
|
|
} else
|
|
k1 = b / a;
|
|
|
|
a = fabs(ge->fy3 - ge->fy2);
|
|
b = fabs(ge->fx3 - ge->fx2);
|
|
if(a < FEPS) {
|
|
if(b < FEPS) {
|
|
a = fabs(ge->fy3 - ge->fy1);
|
|
b = fabs(ge->fx3 - ge->fx1);
|
|
k2 = a < FEPS ? (b < FEPS ? k : 100000.) : ( b / a);
|
|
} else
|
|
k2 = FBIGVAL;
|
|
} else
|
|
k2 = b / a;
|
|
|
|
if(fabs(k1-k) < 0.0001)
|
|
dir |= CVDIR_FEQUAL;
|
|
else if (k1 < k)
|
|
dir |= CVDIR_FUP;
|
|
else
|
|
dir |= CVDIR_FDOWN;
|
|
|
|
if(fabs(k2-k) < 0.0001)
|
|
dir |= CVDIR_REQUAL;
|
|
else if (k2 > k)
|
|
dir |= CVDIR_RUP;
|
|
else
|
|
dir |= CVDIR_RDOWN;
|
|
|
|
return dir;
|
|
}
|
|
|
|
|
|
/* expects that the previous element is also int */
|
|
|
|
static int
|
|
igetcvdir(
|
|
GENTRY * ge
|
|
)
|
|
{
|
|
int a, b;
|
|
double k, k1, k2;
|
|
int dir = 0;
|
|
|
|
if(ge->flags & GEF_FLOAT) {
|
|
fprintf(stderr, "**! igetcvdir(0x%x) on floating entry, ABORT\n", ge);
|
|
abort(); /* dump core */
|
|
}
|
|
|
|
a = ge->iy3 - ge->prev->iy3;
|
|
b = ge->ix3 - ge->prev->ix3;
|
|
k = (a == 0) ? (b == 0 ? 1. : 100000.) : fabs((double) b / (double) a);
|
|
|
|
a = ge->iy1 - ge->prev->iy3;
|
|
b = ge->ix1 - ge->prev->ix3;
|
|
if(a == 0) {
|
|
if(b == 0) {
|
|
a = ge->iy2 - ge->prev->iy3;
|
|
b = ge->ix2 - ge->prev->ix3;
|
|
k1 = (a == 0) ? (b == 0 ? k : 100000.) : fabs((double) b / (double) a);
|
|
} else
|
|
k1 = FBIGVAL;
|
|
} else
|
|
k1 = fabs((double) b / (double) a);
|
|
|
|
a = ge->iy3 - ge->iy2;
|
|
b = ge->ix3 - ge->ix2;
|
|
if(a == 0) {
|
|
if(b == 0) {
|
|
a = ge->iy3 - ge->iy1;
|
|
b = ge->ix3 - ge->ix1;
|
|
k2 = (a == 0) ? (b == 0 ? k : 100000.) : fabs((double) b / (double) a);
|
|
} else
|
|
k2 = FBIGVAL;
|
|
} else
|
|
k2 = fabs((double) b / (double) a);
|
|
|
|
if(fabs(k1-k) < 0.0001)
|
|
dir |= CVDIR_FEQUAL;
|
|
else if (k1 < k)
|
|
dir |= CVDIR_FUP;
|
|
else
|
|
dir |= CVDIR_FDOWN;
|
|
|
|
if(fabs(k2-k) < 0.0001)
|
|
dir |= CVDIR_REQUAL;
|
|
else if (k2 > k)
|
|
dir |= CVDIR_RUP;
|
|
else
|
|
dir |= CVDIR_RDOWN;
|
|
|
|
return dir;
|
|
}
|
|
|
|
#if 0
|
|
/* a function just to test the work of fixcvdir() */
|
|
static void
|
|
testfixcvdir(
|
|
GLYPH * g
|
|
)
|
|
{
|
|
GENTRY *ge;
|
|
int dir;
|
|
|
|
for (ge = g->entries; ge != 0; ge = ge->next) {
|
|
if (ge->type == GE_CURVE) {
|
|
dir = igetcvdir(ge);
|
|
fixcvdir(ge, dir);
|
|
}
|
|
}
|
|
}
|
|
#endif
|
|
|
|
static int
|
|
iround(
|
|
double val
|
|
)
|
|
{
|
|
return (int) (val > 0 ? val + 0.5 : val - 0.5);
|
|
}
|
|
|
|
/* for debugging - dump the glyph
|
|
* mark with a star the entries from start to end inclusive
|
|
* (start == NULL means don't mark any, end == NULL means to the last)
|
|
*/
|
|
|
|
void
|
|
dumppaths(
|
|
GLYPH *g,
|
|
GENTRY *start,
|
|
GENTRY *end
|
|
)
|
|
{
|
|
GENTRY *ge;
|
|
int i;
|
|
char mark=' ';
|
|
|
|
fprintf(stderr, "Glyph %s:\n", g->name);
|
|
|
|
/* now do the conversion */
|
|
for(ge = g->entries; ge != 0; ge = ge->next) {
|
|
if(ge == start)
|
|
mark = '*';
|
|
fprintf(stderr, " %c %8x", mark, ge);
|
|
switch(ge->type) {
|
|
case GE_MOVE:
|
|
case GE_LINE:
|
|
if(ge->flags & GEF_FLOAT)
|
|
fprintf(stderr," %c float (%g, %g)\n", ge->type, ge->fx3, ge->fy3);
|
|
else
|
|
fprintf(stderr," %c int (%d, %d)\n", ge->type, ge->ix3, ge->iy3);
|
|
break;
|
|
case GE_CURVE:
|
|
if(ge->flags & GEF_FLOAT) {
|
|
fprintf(stderr," C float ");
|
|
for(i=0; i<3; i++)
|
|
fprintf(stderr,"(%g, %g) ", ge->fxn[i], ge->fyn[i]);
|
|
fprintf(stderr,"\n");
|
|
} else {
|
|
fprintf(stderr," C int ");
|
|
for(i=0; i<3; i++)
|
|
fprintf(stderr,"(%d, %d) ", ge->ixn[i], ge->iyn[i]);
|
|
fprintf(stderr,"\n");
|
|
}
|
|
break;
|
|
default:
|
|
fprintf(stderr, " %c\n", ge->type);
|
|
break;
|
|
}
|
|
if(ge == end)
|
|
mark = ' ';
|
|
}
|
|
}
|
|
|
|
/*
|
|
* Routine that converts all entries in the path from float to int
|
|
*/
|
|
|
|
void
|
|
pathtoint(
|
|
GLYPH *g
|
|
)
|
|
{
|
|
GENTRY *ge;
|
|
int x[3], y[3];
|
|
int i;
|
|
|
|
|
|
if(ISDBG(TOINT))
|
|
fprintf(stderr, "TOINT: glyph %s\n", g->name);
|
|
assertisfloat(g, "converting path to int\n");
|
|
|
|
fdelsmall(g, 1.0); /* get rid of sub-pixel contours */
|
|
assertpath(g->entries, __FILE__, __LINE__, g->name);
|
|
|
|
/* 1st pass, collect the directions of the curves: have
|
|
* to do that in advance, while everyting is float
|
|
*/
|
|
for(ge = g->entries; ge != 0; ge = ge->next) {
|
|
if( !(ge->flags & GEF_FLOAT) ) {
|
|
fprintf(stderr, "**! glyphs %s has int entry, found in conversion to int\n",
|
|
g->name);
|
|
exit(1);
|
|
}
|
|
if(ge->type == GE_CURVE) {
|
|
ge->dir = fgetcvdir(ge);
|
|
}
|
|
}
|
|
|
|
/* now do the conversion */
|
|
for(ge = g->entries; ge != 0; ge = ge->next) {
|
|
switch(ge->type) {
|
|
case GE_MOVE:
|
|
case GE_LINE:
|
|
if(ISDBG(TOINT))
|
|
fprintf(stderr," %c float x=%g y=%g\n", ge->type, ge->fx3, ge->fy3);
|
|
x[0] = iround(ge->fx3);
|
|
y[0] = iround(ge->fy3);
|
|
for(i=0; i<3; i++) { /* put some valid values everywhere, for convenience */
|
|
ge->ixn[i] = x[0];
|
|
ge->iyn[i] = y[0];
|
|
}
|
|
if(ISDBG(TOINT))
|
|
fprintf(stderr," int x=%d y=%d\n", ge->ix3, ge->iy3);
|
|
break;
|
|
case GE_CURVE:
|
|
if(ISDBG(TOINT))
|
|
fprintf(stderr," %c float ", ge->type);
|
|
|
|
for(i=0; i<3; i++) {
|
|
if(ISDBG(TOINT))
|
|
fprintf(stderr,"(%g, %g) ", ge->fxn[i], ge->fyn[i]);
|
|
x[i] = iround(ge->fxn[i]);
|
|
y[i] = iround(ge->fyn[i]);
|
|
}
|
|
|
|
if(ISDBG(TOINT))
|
|
fprintf(stderr,"\n int ");
|
|
|
|
for(i=0; i<3; i++) {
|
|
ge->ixn[i] = x[i];
|
|
ge->iyn[i] = y[i];
|
|
if(ISDBG(TOINT))
|
|
fprintf(stderr,"(%d, %d) ", ge->ixn[i], ge->iyn[i]);
|
|
}
|
|
ge->flags &= ~GEF_FLOAT; /* for fixcvdir */
|
|
fixcvdir(ge, ge->dir);
|
|
|
|
if(ISDBG(TOINT)) {
|
|
fprintf(stderr,"\n fixed ");
|
|
for(i=0; i<3; i++)
|
|
fprintf(stderr,"(%d, %d) ", ge->ixn[i], ge->iyn[i]);
|
|
fprintf(stderr,"\n");
|
|
}
|
|
|
|
break;
|
|
}
|
|
ge->flags &= ~GEF_FLOAT;
|
|
}
|
|
g->flags &= ~GF_FLOAT;
|
|
}
|
|
|
|
|
|
/* check whether we can fix up the curve to change its size by (dx,dy) */
|
|
/* 0 means NO, 1 means YES */
|
|
|
|
/* for float: if scaling would be under 10% */
|
|
|
|
int
|
|
fcheckcv(
|
|
GENTRY * ge,
|
|
double dx,
|
|
double dy
|
|
)
|
|
{
|
|
if( !(ge->flags & GEF_FLOAT) ) {
|
|
fprintf(stderr, "**! fcheckcv(0x%x) on int entry, ABORT\n", ge);
|
|
abort(); /* dump core */
|
|
}
|
|
|
|
if (ge->type != GE_CURVE)
|
|
return 0;
|
|
|
|
if( fabs(ge->fx3 - ge->prev->fx3) < fabs(dx) * 10 )
|
|
return 0;
|
|
|
|
if( fabs(ge->fy3 - ge->prev->fy3) < fabs(dy) * 10 )
|
|
return 0;
|
|
|
|
return 1;
|
|
}
|
|
|
|
/* for int: if won't create new zigzags at the ends */
|
|
|
|
int
|
|
icheckcv(
|
|
GENTRY * ge,
|
|
int dx,
|
|
int dy
|
|
)
|
|
{
|
|
int xdep, ydep;
|
|
|
|
if(ge->flags & GEF_FLOAT) {
|
|
fprintf(stderr, "**! icheckcv(0x%x) on floating entry, ABORT\n", ge);
|
|
abort(); /* dump core */
|
|
}
|
|
|
|
if (ge->type != GE_CURVE)
|
|
return 0;
|
|
|
|
xdep = ge->ix3 - ge->prev->ix3;
|
|
ydep = ge->iy3 - ge->prev->iy3;
|
|
|
|
if (ge->type == GE_CURVE
|
|
&& (xdep * (xdep + dx)) > 0
|
|
&& (ydep * (ydep + dy)) > 0) {
|
|
return 1;
|
|
} else
|
|
return 0;
|
|
}
|
|
|
|
/* float connect the ends of open contours */
|
|
|
|
void
|
|
fclosepaths(
|
|
GLYPH * g
|
|
)
|
|
{
|
|
GENTRY *ge, *fge, *xge, *nge;
|
|
int i;
|
|
|
|
assertisfloat(g, "fclosepaths float\n");
|
|
|
|
for (xge = g->entries; xge != 0; xge = xge->next) {
|
|
if( xge->type != GE_PATH )
|
|
continue;
|
|
|
|
ge = xge->prev;
|
|
if(ge == 0 || ge->type != GE_LINE && ge->type!= GE_CURVE) {
|
|
fprintf(stderr, "**! Glyph %s got empty path\n",
|
|
g->name);
|
|
exit(1);
|
|
}
|
|
|
|
fge = ge->frwd;
|
|
if (fge->prev == 0 || fge->prev->type != GE_MOVE) {
|
|
fprintf(stderr, "**! Glyph %s got strange beginning of path\n",
|
|
g->name);
|
|
exit(1);
|
|
}
|
|
fge = fge->prev;
|
|
if (fge->fx3 != ge->fx3 || fge->fy3 != ge->fy3) {
|
|
/* we have to fix this open path */
|
|
|
|
WARNING_4 fprintf(stderr, "Glyph %s got path open by dx=%g dy=%g\n",
|
|
g->name, fge->fx3 - ge->fx3, fge->fy3 - ge->fy3);
|
|
|
|
|
|
/* add a new line */
|
|
nge = newgentry(GEF_FLOAT);
|
|
(*nge) = (*ge);
|
|
nge->fx3 = fge->fx3;
|
|
nge->fy3 = fge->fy3;
|
|
nge->type = GE_LINE;
|
|
|
|
addgeafter(ge, nge);
|
|
|
|
if (fabs(ge->fx3 - fge->fx3) <= 2 && fabs(ge->fy3 - fge->fy3) <= 2) {
|
|
/*
|
|
* small change, try to get rid of the new entry
|
|
*/
|
|
|
|
double df[2];
|
|
|
|
for(i=0; i<2; i++) {
|
|
df[i] = ge->fpoints[i][2] - fge->fpoints[i][2];
|
|
df[i] = fclosegap(nge, nge, i, df[i], NULL);
|
|
}
|
|
|
|
if(df[0] == 0. && df[1] == 0.) {
|
|
/* closed gap successfully, remove the added entry */
|
|
freethisge(nge);
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
void
|
|
smoothjoints(
|
|
GLYPH * g
|
|
)
|
|
{
|
|
GENTRY *ge, *ne;
|
|
int dx1, dy1, dx2, dy2, k;
|
|
int dir;
|
|
|
|
return; /* this stuff seems to create problems */
|
|
|
|
assertisint(g, "smoothjoints int");
|
|
|
|
if (g->entries == 0) /* nothing to do */
|
|
return;
|
|
|
|
for (ge = g->entries->next; ge != 0; ge = ge->next) {
|
|
ne = ge->frwd;
|
|
|
|
/*
|
|
* although there should be no one-line path * and any path
|
|
* must end with CLOSEPATH, * nobody can say for sure
|
|
*/
|
|
|
|
if (ge == ne || ne == 0)
|
|
continue;
|
|
|
|
/* now handle various joints */
|
|
|
|
if (ge->type == GE_LINE && ne->type == GE_LINE) {
|
|
dx1 = ge->ix3 - ge->prev->ix3;
|
|
dy1 = ge->iy3 - ge->prev->iy3;
|
|
dx2 = ne->ix3 - ge->ix3;
|
|
dy2 = ne->iy3 - ge->iy3;
|
|
|
|
/* check whether they have the same direction */
|
|
/* and the same slope */
|
|
/* then we can join them into one line */
|
|
|
|
if (dx1 * dx2 >= 0 && dy1 * dy2 >= 0 && dx1 * dy2 == dy1 * dx2) {
|
|
/* extend the previous line */
|
|
ge->ix3 = ne->ix3;
|
|
ge->iy3 = ne->iy3;
|
|
|
|
/* and get rid of the next line */
|
|
freethisge(ne);
|
|
}
|
|
} else if (ge->type == GE_LINE && ne->type == GE_CURVE) {
|
|
fixcvends(ne);
|
|
|
|
dx1 = ge->ix3 - ge->prev->ix3;
|
|
dy1 = ge->iy3 - ge->prev->iy3;
|
|
dx2 = ne->ix1 - ge->ix3;
|
|
dy2 = ne->iy1 - ge->iy3;
|
|
|
|
/* if the line is nearly horizontal and we can fix it */
|
|
if (dx1 != 0 && 5 * abs(dy1) / abs(dx1) == 0
|
|
&& icheckcv(ne, 0, -dy1)
|
|
&& abs(dy1) <= 4) {
|
|
dir = igetcvdir(ne);
|
|
ge->iy3 -= dy1;
|
|
ne->iy1 -= dy1;
|
|
fixcvdir(ne, dir);
|
|
if (ge->next != ne)
|
|
ne->prev->iy3 -= dy1;
|
|
dy1 = 0;
|
|
} else if (dy1 != 0 && 5 * abs(dx1) / abs(dy1) == 0
|
|
&& icheckcv(ne, -dx1, 0)
|
|
&& abs(dx1) <= 4) {
|
|
/* the same but vertical */
|
|
dir = igetcvdir(ne);
|
|
ge->ix3 -= dx1;
|
|
ne->ix1 -= dx1;
|
|
fixcvdir(ne, dir);
|
|
if (ge->next != ne)
|
|
ne->prev->ix3 -= dx1;
|
|
dx1 = 0;
|
|
}
|
|
/*
|
|
* if line is horizontal and curve begins nearly
|
|
* horizontally
|
|
*/
|
|
if (dy1 == 0 && dx2 != 0 && 5 * abs(dy2) / abs(dx2) == 0) {
|
|
dir = igetcvdir(ne);
|
|
ne->iy1 -= dy2;
|
|
fixcvdir(ne, dir);
|
|
dy2 = 0;
|
|
} else if (dx1 == 0 && dy2 != 0 && 5 * abs(dx2) / abs(dy2) == 0) {
|
|
/* the same but vertical */
|
|
dir = igetcvdir(ne);
|
|
ne->ix1 -= dx2;
|
|
fixcvdir(ne, dir);
|
|
dx2 = 0;
|
|
}
|
|
} else if (ge->type == GE_CURVE && ne->type == GE_LINE) {
|
|
fixcvends(ge);
|
|
|
|
dx1 = ge->ix3 - ge->ix2;
|
|
dy1 = ge->iy3 - ge->iy2;
|
|
dx2 = ne->ix3 - ge->ix3;
|
|
dy2 = ne->iy3 - ge->iy3;
|
|
|
|
/* if the line is nearly horizontal and we can fix it */
|
|
if (dx2 != 0 && 5 * abs(dy2) / abs(dx2) == 0
|
|
&& icheckcv(ge, 0, dy2)
|
|
&& abs(dy2) <= 4) {
|
|
dir = igetcvdir(ge);
|
|
ge->iy3 += dy2;
|
|
ge->iy2 += dy2;
|
|
fixcvdir(ge, dir);
|
|
if (ge->next != ne)
|
|
ne->prev->iy3 += dy2;
|
|
dy2 = 0;
|
|
} else if (dy2 != 0 && 5 * abs(dx2) / abs(dy2) == 0
|
|
&& icheckcv(ge, dx2, 0)
|
|
&& abs(dx2) <= 4) {
|
|
/* the same but vertical */
|
|
dir = igetcvdir(ge);
|
|
ge->ix3 += dx2;
|
|
ge->ix2 += dx2;
|
|
fixcvdir(ge, dir);
|
|
if (ge->next != ne)
|
|
ne->prev->ix3 += dx2;
|
|
dx2 = 0;
|
|
}
|
|
/*
|
|
* if line is horizontal and curve ends nearly
|
|
* horizontally
|
|
*/
|
|
if (dy2 == 0 && dx1 != 0 && 5 * abs(dy1) / abs(dx1) == 0) {
|
|
dir = igetcvdir(ge);
|
|
ge->iy2 += dy1;
|
|
fixcvdir(ge, dir);
|
|
dy1 = 0;
|
|
} else if (dx2 == 0 && dy1 != 0 && 5 * abs(dx1) / abs(dy1) == 0) {
|
|
/* the same but vertical */
|
|
dir = igetcvdir(ge);
|
|
ge->ix2 += dx1;
|
|
fixcvdir(ge, dir);
|
|
dx1 = 0;
|
|
}
|
|
} else if (ge->type == GE_CURVE && ne->type == GE_CURVE) {
|
|
fixcvends(ge);
|
|
fixcvends(ne);
|
|
|
|
dx1 = ge->ix3 - ge->ix2;
|
|
dy1 = ge->iy3 - ge->iy2;
|
|
dx2 = ne->ix1 - ge->ix3;
|
|
dy2 = ne->iy1 - ge->iy3;
|
|
|
|
/*
|
|
* check if we have a rather smooth joint at extremal
|
|
* point
|
|
*/
|
|
/* left or right extremal point */
|
|
if (abs(dx1) <= 4 && abs(dx2) <= 4
|
|
&& dy1 != 0 && 5 * abs(dx1) / abs(dy1) == 0
|
|
&& dy2 != 0 && 5 * abs(dx2) / abs(dy2) == 0
|
|
&& (ge->iy3 < ge->prev->iy3 && ne->iy3 < ge->iy3
|
|
|| ge->iy3 > ge->prev->iy3 && ne->iy3 > ge->iy3)
|
|
&& (ge->ix3 - ge->prev->ix3) * (ne->ix3 - ge->ix3) < 0
|
|
) {
|
|
dir = igetcvdir(ge);
|
|
ge->ix2 += dx1;
|
|
dx1 = 0;
|
|
fixcvdir(ge, dir);
|
|
dir = igetcvdir(ne);
|
|
ne->ix1 -= dx2;
|
|
dx2 = 0;
|
|
fixcvdir(ne, dir);
|
|
}
|
|
/* top or down extremal point */
|
|
else if (abs(dy1) <= 4 && abs(dy2) <= 4
|
|
&& dx1 != 0 && 5 * abs(dy1) / abs(dx1) == 0
|
|
&& dx2 != 0 && 5 * abs(dy2) / abs(dx2) == 0
|
|
&& (ge->ix3 < ge->prev->ix3 && ne->ix3 < ge->ix3
|
|
|| ge->ix3 > ge->prev->ix3 && ne->ix3 > ge->ix3)
|
|
&& (ge->iy3 - ge->prev->iy3) * (ne->iy3 - ge->iy3) < 0
|
|
) {
|
|
dir = igetcvdir(ge);
|
|
ge->iy2 += dy1;
|
|
dy1 = 0;
|
|
fixcvdir(ge, dir);
|
|
dir = igetcvdir(ne);
|
|
ne->iy1 -= dy2;
|
|
dy2 = 0;
|
|
fixcvdir(ne, dir);
|
|
}
|
|
/* or may be we just have a smooth junction */
|
|
else if (dx1 * dx2 >= 0 && dy1 * dy2 >= 0
|
|
&& 10 * abs(k = abs(dx1 * dy2) - abs(dy1 * dx2)) < (abs(dx1 * dy2) + abs(dy1 * dx2))) {
|
|
int tries[6][4];
|
|
int results[6];
|
|
int i, b;
|
|
|
|
/* build array of changes we are going to try */
|
|
/* uninitalized entries are 0 */
|
|
if (k > 0) {
|
|
static int t1[6][4] = {
|
|
{0, 0, 0, 0},
|
|
{-1, 0, 1, 0},
|
|
{-1, 0, 0, 1},
|
|
{0, -1, 1, 0},
|
|
{0, -1, 0, 1},
|
|
{-1, -1, 1, 1}};
|
|
memcpy(tries, t1, sizeof tries);
|
|
} else {
|
|
static int t1[6][4] = {
|
|
{0, 0, 0, 0},
|
|
{1, 0, -1, 0},
|
|
{1, 0, 0, -1},
|
|
{0, 1, -1, 0},
|
|
{0, 1, 0, -1},
|
|
{1, 1, -1, -1}};
|
|
memcpy(tries, t1, sizeof tries);
|
|
}
|
|
|
|
/* now try the changes */
|
|
results[0] = abs(k);
|
|
for (i = 1; i < 6; i++) {
|
|
results[i] = abs((abs(dx1) + tries[i][0]) * (abs(dy2) + tries[i][1]) -
|
|
(abs(dy1) + tries[i][2]) * (abs(dx2) + tries[i][3]));
|
|
}
|
|
|
|
/* and find the best try */
|
|
k = abs(k);
|
|
b = 0;
|
|
for (i = 1; i < 6; i++)
|
|
if (results[i] < k) {
|
|
k = results[i];
|
|
b = i;
|
|
}
|
|
/* and finally apply it */
|
|
if (dx1 < 0)
|
|
tries[b][0] = -tries[b][0];
|
|
if (dy2 < 0)
|
|
tries[b][1] = -tries[b][1];
|
|
if (dy1 < 0)
|
|
tries[b][2] = -tries[b][2];
|
|
if (dx2 < 0)
|
|
tries[b][3] = -tries[b][3];
|
|
|
|
dir = igetcvdir(ge);
|
|
ge->ix2 -= tries[b][0];
|
|
ge->iy2 -= tries[b][2];
|
|
fixcvdir(ge, dir);
|
|
dir = igetcvdir(ne);
|
|
ne->ix1 += tries[b][3];
|
|
ne->iy1 += tries[b][1];
|
|
fixcvdir(ne, dir);
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
/* debugging: print out stems of a glyph */
|
|
static void
|
|
debugstems(
|
|
char *name,
|
|
STEM * hstems,
|
|
int nhs,
|
|
STEM * vstems,
|
|
int nvs
|
|
)
|
|
{
|
|
int i;
|
|
|
|
fprintf(pfa_file, "%% %s\n", name);
|
|
fprintf(pfa_file, "%% %d horizontal stems:\n", nhs);
|
|
for (i = 0; i < nhs; i++)
|
|
fprintf(pfa_file, "%% %3d %d (%d...%d) %c %c%c%c%c\n", i, hstems[i].value,
|
|
hstems[i].from, hstems[i].to,
|
|
((hstems[i].flags & ST_UP) ? 'U' : 'D'),
|
|
((hstems[i].flags & ST_END) ? 'E' : '-'),
|
|
((hstems[i].flags & ST_FLAT) ? 'F' : '-'),
|
|
((hstems[i].flags & ST_ZONE) ? 'Z' : ' '),
|
|
((hstems[i].flags & ST_TOPZONE) ? 'T' : ' '));
|
|
fprintf(pfa_file, "%% %d vertical stems:\n", nvs);
|
|
for (i = 0; i < nvs; i++)
|
|
fprintf(pfa_file, "%% %3d %d (%d...%d) %c %c%c\n", i, vstems[i].value,
|
|
vstems[i].from, vstems[i].to,
|
|
((vstems[i].flags & ST_UP) ? 'U' : 'D'),
|
|
((vstems[i].flags & ST_END) ? 'E' : '-'),
|
|
((vstems[i].flags & ST_FLAT) ? 'F' : '-'));
|
|
}
|
|
|
|
/* add pseudo-stems for the limits of the Blue zones to the stem array */
|
|
static int
|
|
addbluestems(
|
|
STEM *s,
|
|
int n
|
|
)
|
|
{
|
|
int i;
|
|
|
|
for(i=0; i<nblues && i<2; i+=2) { /* baseline */
|
|
s[n].value=bluevalues[i];
|
|
s[n].flags=ST_UP|ST_ZONE;
|
|
/* don't overlap with anything */
|
|
s[n].origin=s[n].from=s[n].to= -10000+i;
|
|
n++;
|
|
s[n].value=bluevalues[i+1];
|
|
s[n].flags=ST_ZONE;
|
|
/* don't overlap with anything */
|
|
s[n].origin=s[n].from=s[n].to= -10000+i+1;
|
|
n++;
|
|
}
|
|
for(i=2; i<nblues; i+=2) { /* top zones */
|
|
s[n].value=bluevalues[i];
|
|
s[n].flags=ST_UP|ST_ZONE|ST_TOPZONE;
|
|
/* don't overlap with anything */
|
|
s[n].origin=s[n].from=s[n].to= -10000+i;
|
|
n++;
|
|
s[n].value=bluevalues[i+1];
|
|
s[n].flags=ST_ZONE|ST_TOPZONE;
|
|
/* don't overlap with anything */
|
|
s[n].origin=s[n].from=s[n].to= -10000+i+1;
|
|
n++;
|
|
}
|
|
for(i=0; i<notherb; i+=2) { /* bottom zones */
|
|
s[n].value=otherblues[i];
|
|
s[n].flags=ST_UP|ST_ZONE;
|
|
/* don't overlap with anything */
|
|
s[n].origin=s[n].from=s[n].to= -10000+i+nblues;
|
|
n++;
|
|
s[n].value=otherblues[i+1];
|
|
s[n].flags=ST_ZONE;
|
|
/* don't overlap with anything */
|
|
s[n].origin=s[n].from=s[n].to= -10000+i+1+nblues;
|
|
n++;
|
|
}
|
|
return n;
|
|
}
|
|
|
|
/* sort stems in array */
|
|
static void
|
|
sortstems(
|
|
STEM * s,
|
|
int n
|
|
)
|
|
{
|
|
int i, j;
|
|
STEM x;
|
|
|
|
|
|
/* a simple sorting */
|
|
/* hm, the ordering criteria are not quite simple :-)
|
|
* if the values are tied
|
|
* ST_UP always goes under not ST_UP
|
|
* ST_ZONE goes on the most outer side
|
|
* ST_END goes towards inner side after ST_ZONE
|
|
* ST_FLAT goes on the inner side
|
|
*/
|
|
|
|
for (i = 0; i < n; i++)
|
|
for (j = i + 1; j < n; j++) {
|
|
if(s[i].value < s[j].value)
|
|
continue;
|
|
if(s[i].value == s[j].value) {
|
|
if( (s[i].flags & ST_UP) < (s[j].flags & ST_UP) )
|
|
continue;
|
|
if( (s[i].flags & ST_UP) == (s[j].flags & ST_UP) ) {
|
|
if( s[i].flags & ST_UP ) {
|
|
if(
|
|
(s[i].flags & (ST_ZONE|ST_FLAT|ST_END) ^ ST_FLAT)
|
|
>
|
|
(s[j].flags & (ST_ZONE|ST_FLAT|ST_END) ^ ST_FLAT)
|
|
)
|
|
continue;
|
|
} else {
|
|
if(
|
|
(s[i].flags & (ST_ZONE|ST_FLAT|ST_END) ^ ST_FLAT)
|
|
<
|
|
(s[j].flags & (ST_ZONE|ST_FLAT|ST_END) ^ ST_FLAT)
|
|
)
|
|
continue;
|
|
}
|
|
}
|
|
}
|
|
x = s[j];
|
|
s[j] = s[i];
|
|
s[i] = x;
|
|
}
|
|
}
|
|
|
|
/* check whether two stem borders overlap */
|
|
|
|
static int
|
|
stemoverlap(
|
|
STEM * s1,
|
|
STEM * s2
|
|
)
|
|
{
|
|
int result;
|
|
|
|
if (s1->from <= s2->from && s1->to >= s2->from
|
|
|| s2->from <= s1->from && s2->to >= s1->from)
|
|
result = 1;
|
|
else
|
|
result = 0;
|
|
|
|
if (ISDBG(STEMOVERLAP))
|
|
fprintf(pfa_file, "%% overlap %d(%d..%d)x%d(%d..%d)=%d\n",
|
|
s1->value, s1->from, s1->to, s2->value, s2->from, s2->to, result);
|
|
return result;
|
|
}
|
|
|
|
/*
|
|
* check if the stem [border] is in an appropriate blue zone
|
|
* (currently not used)
|
|
*/
|
|
|
|
static int
|
|
steminblue(
|
|
STEM *s
|
|
)
|
|
{
|
|
int i, val;
|
|
|
|
val=s->value;
|
|
if(s->flags & ST_UP) {
|
|
/* painted size up, look at lower zones */
|
|
if(nblues>=2 && val>=bluevalues[0] && val<=bluevalues[1] )
|
|
return 1;
|
|
for(i=0; i<notherb; i++) {
|
|
if( val>=otherblues[i] && val<=otherblues[i+1] )
|
|
return 1;
|
|
}
|
|
} else {
|
|
/* painted side down, look at upper zones */
|
|
for(i=2; i<nblues; i++) {
|
|
if( val>=bluevalues[i] && val<=bluevalues[i+1] )
|
|
return 1;
|
|
}
|
|
}
|
|
|
|
return 0;
|
|
}
|
|
|
|
/* mark the outermost stem [borders] in the blue zones */
|
|
|
|
static void
|
|
markbluestems(
|
|
STEM *s,
|
|
int nold
|
|
)
|
|
{
|
|
int i, j, a, b, c;
|
|
/*
|
|
* traverse the list of Blue Values, mark the lowest upper
|
|
* stem in each bottom zone and the topmost lower stem in
|
|
* each top zone with ST_BLUE
|
|
*/
|
|
|
|
/* top zones */
|
|
for(i=2; i<nblues; i+=2) {
|
|
a=bluevalues[i]; b=bluevalues[i+1];
|
|
if(ISDBG(BLUESTEMS))
|
|
fprintf(pfa_file, "%% looking at blue zone %d...%d\n", a, b);
|
|
for(j=nold-1; j>=0; j--) {
|
|
if( s[j].flags & (ST_ZONE|ST_UP|ST_END) )
|
|
continue;
|
|
c=s[j].value;
|
|
if(c<a) /* too low */
|
|
break;
|
|
if(c<=b) { /* found the topmost stem border */
|
|
/* mark all the stems with the same value */
|
|
if(ISDBG(BLUESTEMS))
|
|
fprintf(pfa_file, "%% found D BLUE at %d\n", s[j].value);
|
|
/* include ST_END values */
|
|
while( s[j+1].value==c && (s[j+1].flags & ST_ZONE)==0 )
|
|
j++;
|
|
s[j].flags |= ST_BLUE;
|
|
for(j--; j>=0 && s[j].value==c
|
|
&& (s[j].flags & (ST_UP|ST_ZONE))==0 ; j--)
|
|
s[j].flags |= ST_BLUE;
|
|
break;
|
|
}
|
|
}
|
|
}
|
|
/* baseline */
|
|
if(nblues>=2) {
|
|
a=bluevalues[0]; b=bluevalues[1];
|
|
for(j=0; j<nold; j++) {
|
|
if( (s[j].flags & (ST_ZONE|ST_UP|ST_END))!=ST_UP )
|
|
continue;
|
|
c=s[j].value;
|
|
if(c>b) /* too high */
|
|
break;
|
|
if(c>=a) { /* found the lowest stem border */
|
|
/* mark all the stems with the same value */
|
|
if(ISDBG(BLUESTEMS))
|
|
fprintf(pfa_file, "%% found U BLUE at %d\n", s[j].value);
|
|
/* include ST_END values */
|
|
while( s[j-1].value==c && (s[j-1].flags & ST_ZONE)==0 )
|
|
j--;
|
|
s[j].flags |= ST_BLUE;
|
|
for(j++; j<nold && s[j].value==c
|
|
&& (s[j].flags & (ST_UP|ST_ZONE))==ST_UP ; j++)
|
|
s[j].flags |= ST_BLUE;
|
|
break;
|
|
}
|
|
}
|
|
}
|
|
/* bottom zones: the logic is the same as for baseline */
|
|
for(i=0; i<notherb; i+=2) {
|
|
a=otherblues[i]; b=otherblues[i+1];
|
|
for(j=0; j<nold; j++) {
|
|
if( (s[j].flags & (ST_UP|ST_ZONE|ST_END))!=ST_UP )
|
|
continue;
|
|
c=s[j].value;
|
|
if(c>b) /* too high */
|
|
break;
|
|
if(c>=a) { /* found the lowest stem border */
|
|
/* mark all the stems with the same value */
|
|
if(ISDBG(BLUESTEMS))
|
|
fprintf(pfa_file, "%% found U BLUE at %d\n", s[j].value);
|
|
/* include ST_END values */
|
|
while( s[j-1].value==c && (s[j-1].flags & ST_ZONE)==0 )
|
|
j--;
|
|
s[j].flags |= ST_BLUE;
|
|
for(j++; j<nold && s[j].value==c
|
|
&& (s[j].flags & (ST_UP|ST_ZONE))==ST_UP ; j++)
|
|
s[j].flags |= ST_BLUE;
|
|
break;
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
/* Eliminate invalid stems, join equivalent lines and remove nested stems
|
|
* to build the main (non-substituted) set of stems.
|
|
* XXX add consideration of the italic angle
|
|
*/
|
|
static int
|
|
joinmainstems(
|
|
STEM * s,
|
|
int nold,
|
|
int useblues /* do we use the blue values ? */
|
|
)
|
|
{
|
|
#define MAX_STACK 1000
|
|
STEM stack[MAX_STACK];
|
|
int nstack = 0;
|
|
int sbottom = 0;
|
|
int nnew;
|
|
int i, j, k;
|
|
int a, b, c, w1, w2, w3;
|
|
int fw, fd;
|
|
/*
|
|
* priority of the last found stem:
|
|
* 0 - nothing found yet
|
|
* 1 - has ST_END in it (one or more)
|
|
* 2 - has no ST_END and no ST_FLAT, can override only one stem
|
|
* with priority 1
|
|
* 3 - has no ST_END and at least one ST_FLAT, can override one
|
|
* stem with priority 2 or any number of stems with priority 1
|
|
* 4 (handled separately) - has ST_BLUE, can override anything
|
|
*/
|
|
int readystem = 0;
|
|
int pri;
|
|
int nlps = 0; /* number of non-committed
|
|
* lowest-priority stems */
|
|
|
|
|
|
for (i = 0, nnew = 0; i < nold; i++) {
|
|
if (s[i].flags & (ST_UP|ST_ZONE)) {
|
|
if(s[i].flags & ST_BLUE) {
|
|
/* we just HAVE to use this value */
|
|
if (readystem)
|
|
nnew += 2;
|
|
readystem=0;
|
|
|
|
/* remember the list of Blue zone stems with the same value */
|
|
for(a=i, i++; i<nold && s[a].value==s[i].value
|
|
&& (s[i].flags & ST_BLUE); i++)
|
|
{}
|
|
b=i; /* our range is a <= i < b */
|
|
c= -1; /* index of our best guess up to now */
|
|
pri=0;
|
|
/* try to find a match, don't cross blue zones */
|
|
for(; i<nold && (s[i].flags & ST_BLUE)==0; i++) {
|
|
if(s[i].flags & ST_UP) {
|
|
if(s[i].flags & ST_TOPZONE)
|
|
break;
|
|
else
|
|
continue;
|
|
}
|
|
for(j=a; j<b; j++) {
|
|
if(!stemoverlap(&s[j], &s[i]) )
|
|
continue;
|
|
/* consider priorities */
|
|
if( ( (s[j].flags|s[i].flags) & (ST_FLAT|ST_END) )==ST_FLAT ) {
|
|
c=i;
|
|
goto bluematch;
|
|
}
|
|
if( ((s[j].flags|s[i].flags) & ST_END)==0 ) {
|
|
if(pri < 2) {
|
|
c=i; pri=2;
|
|
}
|
|
} else {
|
|
if(pri == 0) {
|
|
c=i; pri=1;
|
|
}
|
|
}
|
|
}
|
|
}
|
|
bluematch:
|
|
/* clean up the stack */
|
|
nstack=sbottom=0;
|
|
readystem=0;
|
|
/* add this stem */
|
|
s[nnew++]=s[a];
|
|
if(c<0) { /* make one-dot-wide stem */
|
|
if(nnew>=b) { /* have no free space */
|
|
for(j=nold; j>=b; j--) /* make free space */
|
|
s[j]=s[j-1];
|
|
b++;
|
|
nold++;
|
|
}
|
|
s[nnew]=s[a];
|
|
s[nnew].flags &= ~(ST_UP|ST_BLUE);
|
|
nnew++;
|
|
i=b-1;
|
|
} else {
|
|
s[nnew++]=s[c];
|
|
i=c; /* skip up to this point */
|
|
}
|
|
if (ISDBG(MAINSTEMS))
|
|
fprintf(pfa_file, "%% +stem %d...%d U BLUE\n",
|
|
s[nnew-2].value, s[nnew-1].value);
|
|
} else {
|
|
if (nstack >= MAX_STACK) {
|
|
WARNING_1 fprintf(stderr, "Warning: **** converter's stem stack overflow ****\n");
|
|
nstack = 0;
|
|
}
|
|
stack[nstack++] = s[i];
|
|
}
|
|
} else if(s[i].flags & ST_BLUE) {
|
|
/* again, we just HAVE to use this value */
|
|
if (readystem)
|
|
nnew += 2;
|
|
readystem=0;
|
|
|
|
/* remember the list of Blue zone stems with the same value */
|
|
for(a=i, i++; i<nold && s[a].value==s[i].value
|
|
&& (s[i].flags & ST_BLUE); i++)
|
|
{}
|
|
b=i; /* our range is a <= i < b */
|
|
c= -1; /* index of our best guess up to now */
|
|
pri=0;
|
|
/* try to find a match */
|
|
for (i = nstack - 1; i >= 0; i--) {
|
|
if( (stack[i].flags & ST_UP)==0 ) {
|
|
if( (stack[i].flags & (ST_ZONE|ST_TOPZONE))==ST_ZONE )
|
|
break;
|
|
else
|
|
continue;
|
|
}
|
|
for(j=a; j<b; j++) {
|
|
if(!stemoverlap(&s[j], &stack[i]) )
|
|
continue;
|
|
/* consider priorities */
|
|
if( ( (s[j].flags|stack[i].flags) & (ST_FLAT|ST_END) )==ST_FLAT ) {
|
|
c=i;
|
|
goto bluedownmatch;
|
|
}
|
|
if( ((s[j].flags|stack[i].flags) & ST_END)==0 ) {
|
|
if(pri < 2) {
|
|
c=i; pri=2;
|
|
}
|
|
} else {
|
|
if(pri == 0) {
|
|
c=i; pri=1;
|
|
}
|
|
}
|
|
}
|
|
}
|
|
bluedownmatch:
|
|
/* if found no match make a one-dot-wide stem */
|
|
if(c<0) {
|
|
c=0;
|
|
stack[0]=s[b-1];
|
|
stack[0].flags |= ST_UP;
|
|
stack[0].flags &= ~ST_BLUE;
|
|
}
|
|
/* remove all the stems conflicting with this one */
|
|
readystem=0;
|
|
for(j=nnew-2; j>=0; j-=2) {
|
|
if (ISDBG(MAINSTEMS))
|
|
fprintf(pfa_file, "%% ?stem %d...%d -- %d\n",
|
|
s[j].value, s[j+1].value, stack[c].value);
|
|
if(s[j+1].value < stack[c].value) /* no conflict */
|
|
break;
|
|
if(s[j].flags & ST_BLUE) {
|
|
/* oops, we don't want to spoil other blue zones */
|
|
stack[c].value=s[j+1].value+1;
|
|
break;
|
|
}
|
|
if( (s[j].flags|s[j+1].flags) & ST_END ) {
|
|
if (ISDBG(MAINSTEMS))
|
|
fprintf(pfa_file, "%% -stem %d...%d p=1\n",
|
|
s[j].value, s[j+1].value);
|
|
continue; /* pri==1, silently discard it */
|
|
}
|
|
/* we want to discard no nore than 2 stems of pri>=2 */
|
|
if( ++readystem > 2 ) {
|
|
/* change our stem to not conflict */
|
|
stack[c].value=s[j+1].value+1;
|
|
break;
|
|
} else {
|
|
if (ISDBG(MAINSTEMS))
|
|
fprintf(pfa_file, "%% -stem %d...%d p>=2\n",
|
|
s[j].value, s[j+1].value);
|
|
continue;
|
|
}
|
|
}
|
|
nnew=j+2;
|
|
/* add this stem */
|
|
if(nnew>=b-1) { /* have no free space */
|
|
for(j=nold; j>=b-1; j--) /* make free space */
|
|
s[j]=s[j-1];
|
|
b++;
|
|
nold++;
|
|
}
|
|
s[nnew++]=stack[c];
|
|
s[nnew++]=s[b-1];
|
|
/* clean up the stack */
|
|
nstack=sbottom=0;
|
|
readystem=0;
|
|
/* set the next position to search */
|
|
i=b-1;
|
|
if (ISDBG(MAINSTEMS))
|
|
fprintf(pfa_file, "%% +stem %d...%d D BLUE\n",
|
|
s[nnew-2].value, s[nnew-1].value);
|
|
} else if (nstack > 0) {
|
|
|
|
/*
|
|
* check whether our stem overlaps with anything in
|
|
* stack
|
|
*/
|
|
for (j = nstack - 1; j >= sbottom; j--) {
|
|
if (s[i].value <= stack[j].value)
|
|
break;
|
|
if (stack[j].flags & ST_ZONE)
|
|
continue;
|
|
|
|
if ((s[i].flags & ST_END)
|
|
|| (stack[j].flags & ST_END))
|
|
pri = 1;
|
|
else if ((s[i].flags & ST_FLAT)
|
|
|| (stack[j].flags & ST_FLAT))
|
|
pri = 3;
|
|
else
|
|
pri = 2;
|
|
|
|
if (pri < readystem && s[nnew + 1].value >= stack[j].value
|
|
|| !stemoverlap(&stack[j], &s[i]))
|
|
continue;
|
|
|
|
if (readystem > 1 && s[nnew + 1].value < stack[j].value) {
|
|
nnew += 2;
|
|
readystem = 0;
|
|
nlps = 0;
|
|
}
|
|
/*
|
|
* width of the previous stem (if it's
|
|
* present)
|
|
*/
|
|
w1 = s[nnew + 1].value - s[nnew].value;
|
|
|
|
/* width of this stem */
|
|
w2 = s[i].value - stack[j].value;
|
|
|
|
if (readystem == 0) {
|
|
/* nothing yet, just add a new stem */
|
|
s[nnew] = stack[j];
|
|
s[nnew + 1] = s[i];
|
|
readystem = pri;
|
|
if (pri == 1)
|
|
nlps = 1;
|
|
else if (pri == 2)
|
|
sbottom = j;
|
|
else {
|
|
sbottom = j + 1;
|
|
while (sbottom < nstack
|
|
&& stack[sbottom].value <= stack[j].value)
|
|
sbottom++;
|
|
}
|
|
if (ISDBG(MAINSTEMS))
|
|
fprintf(pfa_file, "%% +stem %d...%d p=%d n=%d\n",
|
|
stack[j].value, s[i].value, pri, nlps);
|
|
} else if (pri == 1) {
|
|
if (stack[j].value > s[nnew + 1].value) {
|
|
/*
|
|
* doesn't overlap with the
|
|
* previous one
|
|
*/
|
|
nnew += 2;
|
|
nlps++;
|
|
s[nnew] = stack[j];
|
|
s[nnew + 1] = s[i];
|
|
if (ISDBG(MAINSTEMS))
|
|
fprintf(pfa_file, "%% +stem %d...%d p=%d n=%d\n",
|
|
stack[j].value, s[i].value, pri, nlps);
|
|
} else if (w2 < w1) {
|
|
/* is narrower */
|
|
s[nnew] = stack[j];
|
|
s[nnew + 1] = s[i];
|
|
if (ISDBG(MAINSTEMS))
|
|
fprintf(pfa_file, "%% /stem %d...%d p=%d n=%d %d->%d\n",
|
|
stack[j].value, s[i].value, pri, nlps, w1, w2);
|
|
}
|
|
} else if (pri == 2) {
|
|
if (readystem == 2) {
|
|
/* choose the narrower stem */
|
|
if (w1 > w2) {
|
|
s[nnew] = stack[j];
|
|
s[nnew + 1] = s[i];
|
|
sbottom = j;
|
|
if (ISDBG(MAINSTEMS))
|
|
fprintf(pfa_file, "%% /stem %d...%d p=%d n=%d\n",
|
|
stack[j].value, s[i].value, pri, nlps);
|
|
}
|
|
/* else readystem==1 */
|
|
} else if (stack[j].value > s[nnew + 1].value) {
|
|
/*
|
|
* value doesn't overlap with
|
|
* the previous one
|
|
*/
|
|
nnew += 2;
|
|
nlps = 0;
|
|
s[nnew] = stack[j];
|
|
s[nnew + 1] = s[i];
|
|
sbottom = j;
|
|
readystem = pri;
|
|
if (ISDBG(MAINSTEMS))
|
|
fprintf(pfa_file, "%% +stem %d...%d p=%d n=%d\n",
|
|
stack[j].value, s[i].value, pri, nlps);
|
|
} else if (nlps == 1
|
|
|| stack[j].value > s[nnew - 1].value) {
|
|
/*
|
|
* we can replace the top
|
|
* stem
|
|
*/
|
|
nlps = 0;
|
|
s[nnew] = stack[j];
|
|
s[nnew + 1] = s[i];
|
|
readystem = pri;
|
|
sbottom = j;
|
|
if (ISDBG(MAINSTEMS))
|
|
fprintf(pfa_file, "%% /stem %d...%d p=%d n=%d\n",
|
|
stack[j].value, s[i].value, pri, nlps);
|
|
}
|
|
} else if (readystem == 3) { /* that means also
|
|
* pri==3 */
|
|
/* choose the narrower stem */
|
|
if (w1 > w2) {
|
|
s[nnew] = stack[j];
|
|
s[nnew + 1] = s[i];
|
|
sbottom = j + 1;
|
|
while (sbottom < nstack
|
|
&& stack[sbottom].value <= stack[j].value)
|
|
sbottom++;
|
|
if (ISDBG(MAINSTEMS))
|
|
fprintf(pfa_file, "%% /stem %d...%d p=%d n=%d\n",
|
|
stack[j].value, s[i].value, pri, nlps);
|
|
}
|
|
} else if (pri == 3) {
|
|
/*
|
|
* we can replace as many stems as
|
|
* neccessary
|
|
*/
|
|
nnew += 2;
|
|
while (nnew > 0 && s[nnew - 1].value >= stack[j].value) {
|
|
nnew -= 2;
|
|
if (ISDBG(MAINSTEMS))
|
|
fprintf(pfa_file, "%% -stem %d..%d\n",
|
|
s[nnew].value, s[nnew + 1].value);
|
|
}
|
|
nlps = 0;
|
|
s[nnew] = stack[j];
|
|
s[nnew + 1] = s[i];
|
|
readystem = pri;
|
|
sbottom = j + 1;
|
|
while (sbottom < nstack
|
|
&& stack[sbottom].value <= stack[j].value)
|
|
sbottom++;
|
|
if (ISDBG(MAINSTEMS))
|
|
fprintf(pfa_file, "%% +stem %d...%d p=%d n=%d\n",
|
|
stack[j].value, s[i].value, pri, nlps);
|
|
}
|
|
}
|
|
}
|
|
}
|
|
if (readystem)
|
|
nnew += 2;
|
|
|
|
/* change the 1-pixel-wide stems to 20-pixel-wide stems if possible
|
|
* the constant 20 is recommended in the Type1 manual
|
|
*/
|
|
if(useblues) {
|
|
for(i=0; i<nnew; i+=2) {
|
|
if(s[i].value != s[i+1].value)
|
|
continue;
|
|
if( ((s[i].flags ^ s[i+1].flags) & ST_BLUE)==0 )
|
|
continue;
|
|
if( s[i].flags & ST_BLUE ) {
|
|
if(nnew>i+2 && s[i+2].value<s[i].value+22)
|
|
s[i+1].value=s[i+2].value-2; /* compensate for fuzziness */
|
|
else
|
|
s[i+1].value+=20;
|
|
} else {
|
|
if(i>0 && s[i-1].value>s[i].value-22)
|
|
s[i].value=s[i-1].value+2; /* compensate for fuzziness */
|
|
else
|
|
s[i].value-=20;
|
|
}
|
|
}
|
|
}
|
|
/* make sure that no stem it stretched between
|
|
* a top zone and a bottom zone
|
|
*/
|
|
if(useblues) {
|
|
for(i=0; i<nnew; i+=2) {
|
|
a=10000; /* lowest border of top zone crosing the stem */
|
|
b= -10000; /* highest border of bottom zone crossing the stem */
|
|
|
|
for(j=2; j<nblues; j++) {
|
|
c=bluevalues[j];
|
|
if( c>=s[i].value && c<=s[i+1].value && c<a )
|
|
a=c;
|
|
}
|
|
if(nblues>=2) {
|
|
c=bluevalues[1];
|
|
if( c>=s[i].value && c<=s[i+1].value && c>b )
|
|
b=c;
|
|
}
|
|
for(j=1; j<notherb; j++) {
|
|
c=otherblues[j];
|
|
if( c>=s[i].value && c<=s[i+1].value && c>b )
|
|
b=c;
|
|
}
|
|
if( a!=10000 && b!= -10000 ) { /* it is stretched */
|
|
/* split the stem into 2 ghost stems */
|
|
for(j=nnew+1; j>i+1; j--) /* make free space */
|
|
s[j]=s[j-2];
|
|
nnew+=2;
|
|
|
|
if(s[i].value+22 >= a)
|
|
s[i+1].value=a-2; /* leave space for fuzziness */
|
|
else
|
|
s[i+1].value=s[i].value+20;
|
|
|
|
if(s[i+3].value-22 <= b)
|
|
s[i+2].value=b+2; /* leave space for fuzziness */
|
|
else
|
|
s[i+2].value=s[i+3].value-20;
|
|
|
|
i+=2;
|
|
}
|
|
}
|
|
}
|
|
/* look for triple stems */
|
|
for (i = 0; i < nnew; i += 2) {
|
|
if (nnew - i >= 6) {
|
|
a = s[i].value + s[i + 1].value;
|
|
b = s[i + 2].value + s[i + 3].value;
|
|
c = s[i + 4].value + s[i + 5].value;
|
|
|
|
w1 = s[i + 1].value - s[i].value;
|
|
w2 = s[i + 3].value - s[i + 2].value;
|
|
w3 = s[i + 5].value - s[i + 4].value;
|
|
|
|
fw = w3 - w1; /* fuzz in width */
|
|
fd = ((c - b) - (b - a)); /* fuzz in distance
|
|
* (doubled) */
|
|
|
|
/* we are able to handle some fuzz */
|
|
/*
|
|
* it doesn't hurt if the declared stem is a bit
|
|
* narrower than actual unless it's an edge in
|
|
* a blue zone
|
|
*/
|
|
if (abs(abs(fd) - abs(fw)) * 5 < w2
|
|
&& abs(fw) * 20 < (w1 + w3)) { /* width dirrerence <10% */
|
|
|
|
if(useblues) { /* check that we don't disturb any blue stems */
|
|
j=c; k=a;
|
|
if (fw > 0) {
|
|
if (fd > 0) {
|
|
if( s[i+5].flags & ST_BLUE )
|
|
continue;
|
|
j -= fw;
|
|
} else {
|
|
if( s[i+4].flags & ST_BLUE )
|
|
continue;
|
|
j += fw;
|
|
}
|
|
} else if(fw < 0) {
|
|
if (fd > 0) {
|
|
if( s[i+1].flags & ST_BLUE )
|
|
continue;
|
|
k -= fw;
|
|
} else {
|
|
if( s[i].flags & ST_BLUE )
|
|
continue;
|
|
k += fw;
|
|
}
|
|
}
|
|
pri = ((j - b) - (b - k));
|
|
|
|
if (pri > 0) {
|
|
if( s[i+2].flags & ST_BLUE )
|
|
continue;
|
|
} else if(pri < 0) {
|
|
if( s[i+3].flags & ST_BLUE )
|
|
continue;
|
|
}
|
|
}
|
|
|
|
/*
|
|
* first fix up the width of 1st and 3rd
|
|
* stems
|
|
*/
|
|
if (fw > 0) {
|
|
if (fd > 0) {
|
|
s[i + 5].value -= fw;
|
|
c -= fw;
|
|
} else {
|
|
s[i + 4].value += fw;
|
|
c += fw;
|
|
}
|
|
} else {
|
|
if (fd > 0) {
|
|
s[i + 1].value -= fw;
|
|
a -= fw;
|
|
} else {
|
|
s[i].value += fw;
|
|
a += fw;
|
|
}
|
|
}
|
|
fd = ((c - b) - (b - a));
|
|
|
|
if (fd > 0) {
|
|
s[i + 2].value += abs(fd) / 2;
|
|
} else {
|
|
s[i + 3].value -= abs(fd) / 2;
|
|
}
|
|
|
|
s[i].flags |= ST_3;
|
|
i += 4;
|
|
}
|
|
}
|
|
}
|
|
|
|
return (nnew & ~1); /* number of lines must be always even */
|
|
}
|
|
|
|
/*
|
|
* these macros and function allow to set the base stem,
|
|
* check that it's not empty and subtract another stem
|
|
* from the base stem (possibly dividing it into multiple parts)
|
|
*/
|
|
|
|
/* pairs for pieces of the base stem */
|
|
static short xbstem[MAX_STEMS*2];
|
|
/* index of the last point */
|
|
static int xblast= -1;
|
|
|
|
#define setbasestem(from, to) \
|
|
(xbstem[0]=from, xbstem[1]=to, xblast=1)
|
|
#define isbaseempty() (xblast<=0)
|
|
|
|
/* returns 1 if was overlapping, 0 otherwise */
|
|
static int
|
|
subfrombase(
|
|
int from,
|
|
int to
|
|
)
|
|
{
|
|
int a, b;
|
|
int i, j;
|
|
|
|
if(isbaseempty())
|
|
return 0;
|
|
|
|
/* handle the simple case simply */
|
|
if(from > xbstem[xblast] || to < xbstem[0])
|
|
return 0;
|
|
|
|
/* the binary search may be more efficient */
|
|
/* but for now the linear search is OK */
|
|
for(b=1; from > xbstem[b]; b+=2) {} /* result: from <= xbstem[b] */
|
|
for(a=xblast-1; to < xbstem[a]; a-=2) {} /* result: to >= xbstem[a] */
|
|
|
|
/* now the interesting examples are:
|
|
* (it was hard for me to understand, so I looked at the examples)
|
|
* 1
|
|
* a|-----| |-----|b |-----| |-----|
|
|
* f|-----|t
|
|
* 2
|
|
* a|-----|b |-----| |-----| |-----|
|
|
* f|--|t
|
|
* 3
|
|
* a|-----|b |-----| |-----| |-----|
|
|
* f|-----|t
|
|
* 4
|
|
* |-----|b a|-----| |-----| |-----|
|
|
* f|------------|t
|
|
* 5
|
|
* |-----| |-----|b |-----| a|-----|
|
|
* f|-----------------------------|t
|
|
* 6
|
|
* |-----|b |-----| |-----| a|-----|
|
|
* f|--------------------------------------------------|t
|
|
* 7
|
|
* |-----|b |-----| a|-----| |-----|
|
|
* f|--------------------------|t
|
|
*/
|
|
|
|
if(a < b-1) /* hits a gap - example 1 */
|
|
return 0;
|
|
|
|
/* now the subtraction itself */
|
|
|
|
if(a==b-1 && from > xbstem[a] && to < xbstem[b]) {
|
|
/* overlaps with only one subrange and splits it - example 2 */
|
|
j=xblast; i=(xblast+=2);
|
|
while(j>=b)
|
|
xbstem[i--]=xbstem[j--];
|
|
xbstem[b]=from-1;
|
|
xbstem[b+1]=to+1;
|
|
return 1;
|
|
/* becomes
|
|
* 2a
|
|
* a|b || |-----| |-----| |-----|
|
|
* f|--|t
|
|
*/
|
|
}
|
|
|
|
if(xbstem[b-1] < from) {
|
|
/* cuts the back of this subrange - examples 3, 4, 7 */
|
|
xbstem[b] = from-1;
|
|
b+=2;
|
|
/* becomes
|
|
* 3a
|
|
* a|----| |-----|b |-----| |-----|
|
|
* f|-----|t
|
|
* 4a
|
|
* |---| a|-----|b |-----| |-----|
|
|
* f|------------|t
|
|
* 7a
|
|
* |---| |-----|b a|-----| |-----|
|
|
* f|--------------------------|t
|
|
*/
|
|
}
|
|
|
|
if(xbstem[a+1] > to) {
|
|
/* cuts the front of this subrange - examples 4a, 5, 7a */
|
|
xbstem[a] = to+1;
|
|
a-=2;
|
|
/* becomes
|
|
* 4b
|
|
* a|---| |---|b |-----| |-----|
|
|
* f|------------|t
|
|
* 5b
|
|
* |-----| |-----|b a|-----| ||
|
|
* f|-----------------------------|t
|
|
* 7b
|
|
* |---| a|-----|b || |-----|
|
|
* f|--------------------------|t
|
|
*/
|
|
}
|
|
|
|
if(a < b-1) /* now after modification it hits a gap - examples 3a, 4b */
|
|
return 1; /* because we have removed something */
|
|
|
|
/* now remove the subranges completely covered by the new stem */
|
|
/* examples 5b, 6, 7b */
|
|
i=b-1; j=a+2;
|
|
/* positioned as:
|
|
* 5b i j
|
|
* |-----| |-----|b a|-----| ||
|
|
* f|-----------------------------|t
|
|
* 6 i xblast j
|
|
* |-----|b |-----| |-----| a|-----|
|
|
* f|--------------------------------------------------|t
|
|
* 7b i j
|
|
* |---| a|-----|b || |-----|
|
|
* f|--------------------------|t
|
|
*/
|
|
while(j <= xblast)
|
|
xbstem[i++]=xbstem[j++];
|
|
xblast=i-1;
|
|
return 1;
|
|
}
|
|
|
|
/* for debugging */
|
|
static void
|
|
printbasestem(void)
|
|
{
|
|
int i;
|
|
|
|
printf("( ");
|
|
for(i=0; i<xblast; i+=2)
|
|
printf("%d-%d ", xbstem[i], xbstem[i+1]);
|
|
printf(") %d\n", xblast);
|
|
}
|
|
|
|
/*
|
|
* Join the stem borders to build the sets of substituted stems
|
|
* XXX add consideration of the italic angle
|
|
*/
|
|
static void
|
|
joinsubstems(
|
|
STEM * s,
|
|
short *pairs,
|
|
int nold,
|
|
int useblues /* do we use the blue values ? */
|
|
)
|
|
{
|
|
int i, j, x;
|
|
static unsigned char mx[MAX_STEMS][MAX_STEMS];
|
|
|
|
/* we do the substituted groups of stems first
|
|
* and it looks like it's going to be REALLY SLOW
|
|
* AND PAINFUL but let's bother about it later
|
|
*/
|
|
|
|
/* for the substituted stems we don't bother about [hv]stem3 -
|
|
* anyway the X11R6 rasterizer does not bother about hstem3
|
|
* at all and is able to handle only one global vstem3
|
|
* per glyph
|
|
*/
|
|
|
|
/* clean the used part of matrix */
|
|
for(i=0; i<nold; i++)
|
|
for(j=0; j<nold; j++)
|
|
mx[i][j]=0;
|
|
|
|
/* build the matrix of stem pairs */
|
|
for(i=0; i<nold; i++) {
|
|
if( s[i].flags & ST_ZONE )
|
|
continue;
|
|
if(s[i].flags & ST_BLUE)
|
|
mx[i][i]=1; /* allow to pair with itself if no better pair */
|
|
if(s[i].flags & ST_UP) { /* the down-stems are already matched */
|
|
setbasestem(s[i].from, s[i].to);
|
|
for(j=i+1; j<nold; j++) {
|
|
if(s[i].value==s[j].value
|
|
|| s[j].flags & ST_ZONE ) {
|
|
continue;
|
|
}
|
|
x=subfrombase(s[j].from, s[j].to);
|
|
|
|
if(s[j].flags & ST_UP) /* match only up+down pairs */
|
|
continue;
|
|
|
|
mx[i][j]=mx[j][i]=x;
|
|
|
|
if(isbaseempty()) /* nothing else to do */
|
|
break;
|
|
}
|
|
}
|
|
}
|
|
|
|
if(ISDBG(SUBSTEMS)) {
|
|
fprintf(pfa_file, "%% ");
|
|
for(j=0; j<nold; j++)
|
|
putc( j%10==0 ? '0'+(j/10)%10 : ' ', pfa_file);
|
|
fprintf(pfa_file, "\n%% ");
|
|
for(j=0; j<nold; j++)
|
|
putc('0'+j%10, pfa_file);
|
|
putc('\n', pfa_file);
|
|
for(i=0; i<nold; i++) {
|
|
fprintf(pfa_file, "%% %3d ",i);
|
|
for(j=0; j<nold; j++)
|
|
putc( mx[i][j] ? 'X' : '.', pfa_file);
|
|
putc('\n', pfa_file);
|
|
}
|
|
}
|
|
|
|
/* now use the matrix to find the best pair for each stem */
|
|
for(i=0; i<nold; i++) {
|
|
int pri, lastpri, v, f;
|
|
|
|
x= -1; /* best pair: none */
|
|
lastpri=0;
|
|
|
|
v=s[i].value;
|
|
f=s[i].flags;
|
|
|
|
if(f & ST_ZONE) {
|
|
pairs[i]= -1;
|
|
continue;
|
|
}
|
|
|
|
if(f & ST_UP) {
|
|
for(j=i+1; j<nold; j++) {
|
|
if(mx[i][j]==0)
|
|
continue;
|
|
|
|
if( (f | s[j].flags) & ST_END )
|
|
pri=1;
|
|
else if( (f | s[j].flags) & ST_FLAT )
|
|
pri=3;
|
|
else
|
|
pri=2;
|
|
|
|
if(lastpri==0
|
|
|| pri > lastpri
|
|
&& ( lastpri==1 || s[j].value-v<20 || (s[x].value-v)*2 >= s[j].value-v ) ) {
|
|
lastpri=pri;
|
|
x=j;
|
|
}
|
|
}
|
|
} else {
|
|
for(j=i-1; j>=0; j--) {
|
|
if(mx[i][j]==0)
|
|
continue;
|
|
|
|
if( (f | s[j].flags) & ST_END )
|
|
pri=1;
|
|
else if( (f | s[j].flags) & ST_FLAT )
|
|
pri=3;
|
|
else
|
|
pri=2;
|
|
|
|
if(lastpri==0
|
|
|| pri > lastpri
|
|
&& ( lastpri==1 || v-s[j].value<20 || (v-s[x].value)*2 >= v-s[j].value ) ) {
|
|
lastpri=pri;
|
|
x=j;
|
|
}
|
|
}
|
|
}
|
|
if(x== -1 && mx[i][i])
|
|
pairs[i]=i; /* a special case */
|
|
else
|
|
pairs[i]=x;
|
|
}
|
|
|
|
if(ISDBG(SUBSTEMS)) {
|
|
for(i=0; i<nold; i++) {
|
|
j=pairs[i];
|
|
if(j>0)
|
|
fprintf(pfa_file, "%% %d...%d (%d x %d)\n", s[i].value, s[j].value, i, j);
|
|
}
|
|
}
|
|
}
|
|
|
|
/*
|
|
* Make all the stems originating at the same value get the
|
|
* same width. Without this the rasterizer may move the dots
|
|
* randomly up or down by one pixel, and that looks bad.
|
|
* The prioritisation is the same as in findstemat().
|
|
*/
|
|
static void
|
|
uniformstems(
|
|
STEM * s,
|
|
short *pairs,
|
|
int ns
|
|
)
|
|
{
|
|
int i, j, from, to, val, dir;
|
|
int pri, prevpri[2], wd, prevwd[2], prevbest[2];
|
|
|
|
for(from=0; from<ns; from=to) {
|
|
prevpri[0] = prevpri[1] = 0;
|
|
prevwd[0] = prevwd[1] = 0;
|
|
prevbest[0] = prevbest[1] = -1;
|
|
val = s[from].value;
|
|
|
|
for(to = from; to<ns && s[to].value == val; to++) {
|
|
dir = ((s[to].flags & ST_UP)!=0);
|
|
|
|
i=pairs[to]; /* the other side of this stem */
|
|
if(i<0 || i==to)
|
|
continue; /* oops, no other side */
|
|
wd=abs(s[i].value-val);
|
|
if(wd == 0)
|
|
continue;
|
|
pri=1;
|
|
if( (s[to].flags | s[i].flags) & ST_END )
|
|
pri=0;
|
|
if( prevbest[dir] == -1 || pri > prevpri[dir] || wd<prevwd[dir] ) {
|
|
prevbest[dir]=i;
|
|
prevpri[dir]=pri;
|
|
prevwd[dir]=wd;
|
|
}
|
|
}
|
|
|
|
for(i=from; i<to; i++) {
|
|
dir = ((s[i].flags & ST_UP)!=0);
|
|
if(prevbest[dir] >= 0) {
|
|
if(ISDBG(SUBSTEMS)) {
|
|
fprintf(stderr, "at %d (%s %d) pair %d->%d(%d)\n", i,
|
|
(dir ? "UP":"DOWN"), s[i].value, pairs[i], prevbest[dir],
|
|
s[prevbest[dir]].value);
|
|
}
|
|
pairs[i] = prevbest[dir];
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
/*
|
|
* Find the best stem in the array at the specified (value, origin),
|
|
* related to the entry ge.
|
|
* Returns its index in the array sp, -1 means "none".
|
|
* prevbest is the result for the other end of the line, we must
|
|
* find something better than it or leave it as it is.
|
|
*/
|
|
static int
|
|
findstemat(
|
|
int value,
|
|
int origin,
|
|
GENTRY *ge,
|
|
STEM *sp,
|
|
short *pairs,
|
|
int ns,
|
|
int prevbest /* -1 means "none" */
|
|
)
|
|
{
|
|
int i, min, max;
|
|
int v, si;
|
|
int pri, prevpri; /* priority, 0 = has ST_END, 1 = no ST_END */
|
|
int wd, prevwd; /* stem width */
|
|
|
|
si= -1; /* nothing yet */
|
|
|
|
/* stems are ordered by value, binary search */
|
|
min=0; max=ns; /* min <= i < max */
|
|
while( min < max ) {
|
|
i=(min+max)/2;
|
|
v=sp[i].value;
|
|
if(v<value)
|
|
min=i+1;
|
|
else if(v>value)
|
|
max=i;
|
|
else {
|
|
si=i; /* temporary value */
|
|
break;
|
|
}
|
|
}
|
|
|
|
if( si < 0 ) /* found nothing this time */
|
|
return prevbest;
|
|
|
|
/* find the priority of the prevbest */
|
|
/* we expect that prevbest has a pair */
|
|
if(prevbest>=0) {
|
|
i=pairs[prevbest];
|
|
prevpri=1;
|
|
if( (sp[prevbest].flags | sp[i].flags) & ST_END )
|
|
prevpri=0;
|
|
prevwd=abs(sp[i].value-value);
|
|
}
|
|
|
|
/* stems are not ordered by origin, so now do the linear search */
|
|
|
|
while( si>0 && sp[si-1].value==value ) /* find the first one */
|
|
si--;
|
|
|
|
for(; si<ns && sp[si].value==value; si++) {
|
|
if(sp[si].origin != origin)
|
|
continue;
|
|
if(sp[si].ge != ge) {
|
|
if(ISDBG(SUBSTEMS)) {
|
|
fprintf(stderr,
|
|
"dbg: possible self-intersection at v=%d o=%d exp_ge=0x%x ge=0x%x\n",
|
|
value, origin, ge, sp[si].ge);
|
|
}
|
|
continue;
|
|
}
|
|
i=pairs[si]; /* the other side of this stem */
|
|
if(i<0)
|
|
continue; /* oops, no other side */
|
|
pri=1;
|
|
if( (sp[si].flags | sp[i].flags) & ST_END )
|
|
pri=0;
|
|
wd=abs(sp[i].value-value);
|
|
if( prevbest == -1 || pri >prevpri
|
|
|| pri==prevpri && prevwd==0 || wd!=0 && wd<prevwd ) {
|
|
prevbest=si;
|
|
prevpri=pri;
|
|
prevwd=wd;
|
|
continue;
|
|
}
|
|
}
|
|
|
|
return prevbest;
|
|
}
|
|
|
|
/* add the substems for one glyph entry
|
|
* (called from groupsubstems())
|
|
* returns 0 if all OK, 1 if too many groups
|
|
*/
|
|
|
|
static int gssentry_lastgrp=0; /* reset to 0 for each new glyph */
|
|
|
|
static int
|
|
gssentry( /* crazy number of parameters */
|
|
GENTRY *ge,
|
|
STEM *hs, /* horizontal stems, sorted by value */
|
|
short *hpairs,
|
|
int nhs,
|
|
STEM *vs, /* vertical stems, sorted by value */
|
|
short *vpairs,
|
|
int nvs,
|
|
STEMBOUNDS *s,
|
|
short *egp,
|
|
int *nextvsi,
|
|
int *nexthsi /* -2 means "check by yourself" */
|
|
) {
|
|
enum {
|
|
SI_VP, /* vertical primary */
|
|
SI_HP, /* horizontal primary */
|
|
SI_SIZE /* size of the array */
|
|
};
|
|
int si[SI_SIZE]; /* indexes of relevant stems */
|
|
|
|
/* the bounds of the existing relevant stems */
|
|
STEMBOUNDS r[ sizeof(si) / sizeof(si[0]) * 2 ];
|
|
char rexpand; /* by how much we need to expand the group */
|
|
int nr; /* and the number of them */
|
|
|
|
/* yet more temporary storage */
|
|
short lb, hb, isvert;
|
|
int conflict, grp;
|
|
int i, j, x, y;
|
|
|
|
|
|
/* for each line or curve we try to find a horizontal and
|
|
* a vertical stem corresponding to its first point
|
|
* (corresponding to the last point of the previous
|
|
* glyph entry), because the directions of the lines
|
|
* will be eventually reversed and it will then become the last
|
|
* point. And the T1 rasterizer applies the hints to
|
|
* the last point.
|
|
*
|
|
*/
|
|
|
|
/* start with the common part, the first point */
|
|
x=ge->prev->ix3;
|
|
y=ge->prev->iy3;
|
|
|
|
if(*nextvsi == -2)
|
|
si[SI_VP]=findstemat(x, y, ge, vs, vpairs, nvs, -1);
|
|
else {
|
|
si[SI_VP]= *nextvsi; *nextvsi= -2;
|
|
}
|
|
if(*nexthsi == -2)
|
|
si[SI_HP]=findstemat(y, x, ge, hs, hpairs, nhs, -1);
|
|
else {
|
|
si[SI_HP]= *nexthsi; *nexthsi= -2;
|
|
}
|
|
|
|
/*
|
|
* For the horizontal lines we make sure that both
|
|
* ends of the line have the same horizontal stem,
|
|
* and the same thing for vertical lines and stems.
|
|
* In both cases we enforce the stem for the next entry.
|
|
* Otherwise unpleasant effects may arise.
|
|
*/
|
|
|
|
if(ge->type==GE_LINE) {
|
|
if(ge->ix3==x) { /* vertical line */
|
|
*nextvsi=si[SI_VP]=findstemat(x, ge->iy3, ge->frwd, vs, vpairs, nvs, si[SI_VP]);
|
|
} else if(ge->iy3==y) { /* horizontal line */
|
|
*nexthsi=si[SI_HP]=findstemat(y, ge->ix3, ge->frwd, hs, hpairs, nhs, si[SI_HP]);
|
|
}
|
|
}
|
|
|
|
if(si[SI_VP]+si[SI_HP] == -2) /* no stems, leave it alone */
|
|
return 0;
|
|
|
|
/* build the array of relevant bounds */
|
|
nr=0;
|
|
for(i=0; i< sizeof(si) / sizeof(si[0]); i++) {
|
|
STEM *sp;
|
|
short *pairs;
|
|
int step;
|
|
int f;
|
|
int nzones, firstzone, binzone, einzone;
|
|
int btype, etype;
|
|
|
|
if(si[i] < 0)
|
|
continue;
|
|
|
|
if(i<SI_HP) {
|
|
r[nr].isvert=1; sp=vs; pairs=vpairs;
|
|
} else {
|
|
r[nr].isvert=0; sp=hs; pairs=hpairs;
|
|
}
|
|
|
|
r[nr].low=sp[ si[i] ].value;
|
|
r[nr].high=sp[ pairs[ si[i] ] ].value;
|
|
|
|
if(r[nr].low > r[nr].high) {
|
|
j=r[nr].low; r[nr].low=r[nr].high; r[nr].high=j;
|
|
step= -1;
|
|
} else {
|
|
step=1;
|
|
}
|
|
|
|
/* handle the interaction with Blue Zones */
|
|
|
|
if(i>=SI_HP) { /* only for horizontal stems */
|
|
if(si[i]==pairs[si[i]]) {
|
|
/* special case, the outermost stem in the
|
|
* Blue Zone without a pair, simulate it to 20-pixel
|
|
*/
|
|
if(sp[ si[i] ].flags & ST_UP) {
|
|
r[nr].high+=20;
|
|
for(j=si[i]+1; j<nhs; j++)
|
|
if( (sp[j].flags & (ST_ZONE|ST_TOPZONE))
|
|
== (ST_ZONE|ST_TOPZONE) ) {
|
|
if(r[nr].high > sp[j].value-2)
|
|
r[nr].high=sp[j].value-2;
|
|
break;
|
|
}
|
|
} else {
|
|
r[nr].low-=20;
|
|
for(j=si[i]-1; j>=0; j--)
|
|
if( (sp[j].flags & (ST_ZONE|ST_TOPZONE))
|
|
== (ST_ZONE) ) {
|
|
if(r[nr].low < sp[j].value+2)
|
|
r[nr].low=sp[j].value+2;
|
|
break;
|
|
}
|
|
}
|
|
}
|
|
|
|
/* check that the stem borders don't end up in
|
|
* different Blue Zones */
|
|
f=sp[ si[i] ].flags;
|
|
nzones=0; einzone=binzone=0;
|
|
for(j=si[i]; j!=pairs[ si[i] ]; j+=step) {
|
|
if( (sp[j].flags & ST_ZONE)==0 )
|
|
continue;
|
|
/* if see a zone border going in the same direction */
|
|
if( ((f ^ sp[j].flags) & ST_UP)==0 ) {
|
|
if( ++nzones == 1 ) {
|
|
firstzone=sp[j].value; /* remember the first one */
|
|
etype=sp[j].flags & ST_TOPZONE;
|
|
}
|
|
einzone=1;
|
|
|
|
} else { /* the opposite direction */
|
|
if(nzones==0) { /* beginning is in a blue zone */
|
|
binzone=1;
|
|
btype=sp[j].flags & ST_TOPZONE;
|
|
}
|
|
einzone=0;
|
|
}
|
|
}
|
|
|
|
/* beginning and end are in Blue Zones of different types */
|
|
if( binzone && einzone && (btype ^ etype)!=0 ) {
|
|
if( sp[si[i]].flags & ST_UP ) {
|
|
if(firstzone > r[nr].low+22)
|
|
r[nr].high=r[nr].low+20;
|
|
else
|
|
r[nr].high=firstzone-2;
|
|
} else {
|
|
if(firstzone < r[nr].high-22)
|
|
r[nr].low=r[nr].high-20;
|
|
else
|
|
r[nr].low=firstzone+2;
|
|
}
|
|
}
|
|
}
|
|
|
|
if(ISDBG(SUBSTEMS))
|
|
fprintf(pfa_file, "%% at(%d,%d)[%d,%d] %d..%d %c (%d x %d)\n", x, y, i, nr,
|
|
r[nr].low, r[nr].high, r[nr].isvert ? 'v' : 'h',
|
|
si[i], pairs[si[i]]);
|
|
|
|
nr++;
|
|
}
|
|
|
|
/* now try to find a group */
|
|
conflict=0; /* no conflicts found yet */
|
|
for(j=0; j<nr; j++)
|
|
r[j].already=0;
|
|
|
|
/* check if it fits into the last group */
|
|
grp = gssentry_lastgrp;
|
|
i = (grp==0)? 0 : egp[grp-1];
|
|
for(; i<egp[grp]; i++) {
|
|
lb=s[i].low; hb=s[i].high; isvert=s[i].isvert;
|
|
for(j=0; j<nr; j++)
|
|
if( r[j].isvert==isvert /* intersects */
|
|
&& r[j].low <= hb && r[j].high >= lb ) {
|
|
if( r[j].low == lb && r[j].high == hb ) /* coincides */
|
|
r[j].already=1;
|
|
else
|
|
conflict=1;
|
|
}
|
|
|
|
if(conflict)
|
|
break;
|
|
}
|
|
|
|
if(conflict) { /* nope, check all the groups */
|
|
for(j=0; j<nr; j++)
|
|
r[j].already=0;
|
|
|
|
for(i=0, grp=0; i<egp[NSTEMGRP-1]; i++) {
|
|
if(i == egp[grp]) { /* checked all stems in a group */
|
|
if(conflict) {
|
|
grp++; conflict=0; /* check the next group */
|
|
for(j=0; j<nr; j++)
|
|
r[j].already=0;
|
|
} else
|
|
break; /* insert into this group */
|
|
}
|
|
|
|
lb=s[i].low; hb=s[i].high; isvert=s[i].isvert;
|
|
for(j=0; j<nr; j++)
|
|
if( r[j].isvert==isvert /* intersects */
|
|
&& r[j].low <= hb && r[j].high >= lb ) {
|
|
if( r[j].low == lb && r[j].high == hb ) /* coincides */
|
|
r[j].already=1;
|
|
else
|
|
conflict=1;
|
|
}
|
|
|
|
if(conflict)
|
|
i=egp[grp]-1; /* fast forward to the next group */
|
|
}
|
|
}
|
|
|
|
/* do we have any empty group ? */
|
|
if(conflict && grp < NSTEMGRP-1) {
|
|
grp++; conflict=0;
|
|
for(j=0; j<nr; j++)
|
|
r[j].already=0;
|
|
}
|
|
|
|
if(conflict) { /* oops, can't find any group to fit */
|
|
return 1;
|
|
}
|
|
|
|
/* OK, add stems to this group */
|
|
|
|
rexpand = nr;
|
|
for(j=0; j<nr; j++)
|
|
rexpand -= r[j].already;
|
|
|
|
if(rexpand > 0) {
|
|
for(i=egp[NSTEMGRP-1]-1; i>=egp[grp]; i--)
|
|
s[i+rexpand]=s[i];
|
|
for(i=0; i<nr; i++)
|
|
if(!r[i].already)
|
|
s[egp[grp]++]=r[i];
|
|
for(i=grp+1; i<NSTEMGRP; i++)
|
|
egp[i]+=rexpand;
|
|
}
|
|
|
|
ge->stemid = gssentry_lastgrp = grp;
|
|
return 0;
|
|
}
|
|
|
|
/*
|
|
* Create the groups of substituted stems from the list.
|
|
* Each group will be represented by a subroutine in the Subs
|
|
* array.
|
|
*/
|
|
|
|
static void
|
|
groupsubstems(
|
|
GLYPH *g,
|
|
STEM *hs, /* horizontal stems, sorted by value */
|
|
short *hpairs,
|
|
int nhs,
|
|
STEM *vs, /* vertical stems, sorted by value */
|
|
short *vpairs,
|
|
int nvs
|
|
)
|
|
{
|
|
GENTRY *ge;
|
|
int i, j;
|
|
|
|
/* temporary storage */
|
|
STEMBOUNDS s[MAX_STEMS*2];
|
|
/* indexes in there, pointing past the end each stem group */
|
|
short egp[NSTEMGRP];
|
|
|
|
int nextvsi, nexthsi; /* -2 means "check by yourself" */
|
|
|
|
for(i=0; i<NSTEMGRP; i++)
|
|
egp[i]=0;
|
|
|
|
nextvsi=nexthsi= -2; /* processed no horiz/vert line */
|
|
|
|
gssentry_lastgrp = 0; /* reset the last group for new glyph */
|
|
|
|
for (ge = g->entries; ge != 0; ge = ge->next) {
|
|
if(ge->type!=GE_LINE && ge->type!=GE_CURVE) {
|
|
nextvsi=nexthsi= -2; /* next path is independent */
|
|
continue;
|
|
}
|
|
|
|
if( gssentry(ge, hs, hpairs, nhs, vs, vpairs, nvs, s, egp, &nextvsi, &nexthsi) ) {
|
|
WARNING_2 fprintf(stderr, "*** glyph %s requires over %d hint subroutines, ignored them\n",
|
|
g->name, NSTEMGRP);
|
|
/* it's better to have no substituted hints at all than have only part */
|
|
for (ge = g->entries; ge != 0; ge = ge->next)
|
|
ge->stemid= -1;
|
|
g->nsg=0; /* just to be safe, already is 0 by initialization */
|
|
return;
|
|
}
|
|
|
|
/*
|
|
* handle the last vert/horiz line of the path specially,
|
|
* correct the hint for the first entry of the path
|
|
*/
|
|
if(ge->frwd != ge->next && (nextvsi != -2 || nexthsi != -2) ) {
|
|
if( gssentry(ge->frwd, hs, hpairs, nhs, vs, vpairs, nvs, s, egp, &nextvsi, &nexthsi) ) {
|
|
WARNING_2 fprintf(stderr, "*** glyph %s requires over %d hint subroutines, ignored them\n",
|
|
g->name, NSTEMGRP);
|
|
/* it's better to have no substituted hints at all than have only part */
|
|
for (ge = g->entries; ge != 0; ge = ge->next)
|
|
ge->stemid= -1;
|
|
g->nsg=0; /* just to be safe, already is 0 by initialization */
|
|
return;
|
|
}
|
|
}
|
|
|
|
}
|
|
|
|
/* find the index of the first empty group - same as the number of groups */
|
|
if(egp[0]>0) {
|
|
for(i=1; i<NSTEMGRP && egp[i]!=egp[i-1]; i++)
|
|
{}
|
|
g->nsg=i;
|
|
} else
|
|
g->nsg=0;
|
|
|
|
if(ISDBG(SUBSTEMS)) {
|
|
fprintf(pfa_file, "%% %d substem groups (%d %d %d)\n", g->nsg,
|
|
g->nsg>1 ? egp[g->nsg-2] : -1,
|
|
g->nsg>0 ? egp[g->nsg-1] : -1,
|
|
g->nsg<NSTEMGRP ? egp[g->nsg] : -1 );
|
|
j=0;
|
|
for(i=0; i<g->nsg; i++) {
|
|
fprintf(pfa_file, "%% grp %3d: ", i);
|
|
for(; j<egp[i]; j++) {
|
|
fprintf(pfa_file, " %4d...%-4d %c ", s[j].low, s[j].high,
|
|
s[j].isvert ? 'v' : 'h');
|
|
}
|
|
fprintf(pfa_file, "\n");
|
|
}
|
|
}
|
|
|
|
if(g->nsg==1) { /* it would be the same as the main stems */
|
|
/* so erase it */
|
|
for (ge = g->entries; ge != 0; ge = ge->next)
|
|
ge->stemid= -1;
|
|
g->nsg=0;
|
|
}
|
|
|
|
if(g->nsg>0) {
|
|
if( (g->nsbs=malloc(g->nsg * sizeof (egp[0]))) == 0 ) {
|
|
fprintf(stderr, "**** not enough memory for substituted hints ****\n");
|
|
exit(255);
|
|
}
|
|
memmove(g->nsbs, egp, g->nsg * sizeof(short));
|
|
if( (g->sbstems=malloc(egp[g->nsg-1] * sizeof (s[0]))) == 0 ) {
|
|
fprintf(stderr, "**** not enough memory for substituted hints ****\n");
|
|
exit(255);
|
|
}
|
|
memmove(g->sbstems, s, egp[g->nsg-1] * sizeof(s[0]));
|
|
}
|
|
}
|
|
|
|
void
|
|
buildstems(
|
|
GLYPH * g
|
|
)
|
|
{
|
|
STEM hs[MAX_STEMS], vs[MAX_STEMS]; /* temporary working
|
|
* storage */
|
|
short hs_pairs[MAX_STEMS], vs_pairs[MAX_STEMS]; /* best pairs for these stems */
|
|
STEM *sp;
|
|
GENTRY *ge, *nge, *pge;
|
|
int nx, ny;
|
|
int ovalue;
|
|
int totals, grp, lastgrp;
|
|
|
|
assertisint(g, "buildstems int");
|
|
|
|
g->nhs = g->nvs = 0;
|
|
memset(hs, 0, sizeof hs);
|
|
memset(vs, 0, sizeof vs);
|
|
|
|
/* first search the whole character for possible stem points */
|
|
|
|
for (ge = g->entries; ge != 0; ge = ge->next) {
|
|
if (ge->type == GE_CURVE) {
|
|
|
|
/*
|
|
* SURPRISE!
|
|
* We consider the stems bound by the
|
|
* H/V ends of the curves as flat ones.
|
|
*
|
|
* But we don't include the point on the
|
|
* other end into the range.
|
|
*/
|
|
|
|
/* first check the beginning of curve */
|
|
/* if it is horizontal, add a hstem */
|
|
if (ge->iy1 == ge->prev->iy3) {
|
|
hs[g->nhs].value = ge->iy1;
|
|
|
|
if (ge->ix1 < ge->prev->ix3)
|
|
hs[g->nhs].flags = ST_FLAT | ST_UP;
|
|
else
|
|
hs[g->nhs].flags = ST_FLAT;
|
|
|
|
hs[g->nhs].origin = ge->prev->ix3;
|
|
hs[g->nhs].ge = ge;
|
|
|
|
if (ge->ix1 < ge->prev->ix3) {
|
|
hs[g->nhs].from = ge->ix1+1;
|
|
hs[g->nhs].to = ge->prev->ix3;
|
|
if(hs[g->nhs].from > hs[g->nhs].to)
|
|
hs[g->nhs].from--;
|
|
} else {
|
|
hs[g->nhs].from = ge->prev->ix3;
|
|
hs[g->nhs].to = ge->ix1-1;
|
|
if(hs[g->nhs].from > hs[g->nhs].to)
|
|
hs[g->nhs].to++;
|
|
}
|
|
if (ge->ix1 != ge->prev->ix3)
|
|
g->nhs++;
|
|
}
|
|
/* if it is vertical, add a vstem */
|
|
else if (ge->ix1 == ge->prev->ix3) {
|
|
vs[g->nvs].value = ge->ix1;
|
|
|
|
if (ge->iy1 > ge->prev->iy3)
|
|
vs[g->nvs].flags = ST_FLAT | ST_UP;
|
|
else
|
|
vs[g->nvs].flags = ST_FLAT;
|
|
|
|
vs[g->nvs].origin = ge->prev->iy3;
|
|
vs[g->nvs].ge = ge;
|
|
|
|
if (ge->iy1 < ge->prev->iy3) {
|
|
vs[g->nvs].from = ge->iy1+1;
|
|
vs[g->nvs].to = ge->prev->iy3;
|
|
if(vs[g->nvs].from > vs[g->nvs].to)
|
|
vs[g->nvs].from--;
|
|
} else {
|
|
vs[g->nvs].from = ge->prev->iy3;
|
|
vs[g->nvs].to = ge->iy1-1;
|
|
if(vs[g->nvs].from > vs[g->nvs].to)
|
|
vs[g->nvs].to++;
|
|
}
|
|
|
|
if (ge->iy1 != ge->prev->iy3)
|
|
g->nvs++;
|
|
}
|
|
/* then check the end of curve */
|
|
/* if it is horizontal, add a hstem */
|
|
if (ge->iy3 == ge->iy2) {
|
|
hs[g->nhs].value = ge->iy3;
|
|
|
|
if (ge->ix3 < ge->ix2)
|
|
hs[g->nhs].flags = ST_FLAT | ST_UP;
|
|
else
|
|
hs[g->nhs].flags = ST_FLAT;
|
|
|
|
hs[g->nhs].origin = ge->ix3;
|
|
hs[g->nhs].ge = ge->frwd;
|
|
|
|
if (ge->ix3 < ge->ix2) {
|
|
hs[g->nhs].from = ge->ix3;
|
|
hs[g->nhs].to = ge->ix2-1;
|
|
if( hs[g->nhs].from > hs[g->nhs].to )
|
|
hs[g->nhs].to++;
|
|
} else {
|
|
hs[g->nhs].from = ge->ix2+1;
|
|
hs[g->nhs].to = ge->ix3;
|
|
if( hs[g->nhs].from > hs[g->nhs].to )
|
|
hs[g->nhs].from--;
|
|
}
|
|
|
|
if (ge->ix3 != ge->ix2)
|
|
g->nhs++;
|
|
}
|
|
/* if it is vertical, add a vstem */
|
|
else if (ge->ix3 == ge->ix2) {
|
|
vs[g->nvs].value = ge->ix3;
|
|
|
|
if (ge->iy3 > ge->iy2)
|
|
vs[g->nvs].flags = ST_FLAT | ST_UP;
|
|
else
|
|
vs[g->nvs].flags = ST_FLAT;
|
|
|
|
vs[g->nvs].origin = ge->iy3;
|
|
vs[g->nvs].ge = ge->frwd;
|
|
|
|
if (ge->iy3 < ge->iy2) {
|
|
vs[g->nvs].from = ge->iy3;
|
|
vs[g->nvs].to = ge->iy2-1;
|
|
if( vs[g->nvs].from > vs[g->nvs].to )
|
|
vs[g->nvs].to++;
|
|
} else {
|
|
vs[g->nvs].from = ge->iy2+1;
|
|
vs[g->nvs].to = ge->iy3;
|
|
if( vs[g->nvs].from > vs[g->nvs].to )
|
|
vs[g->nvs].from--;
|
|
}
|
|
|
|
if (ge->iy3 != ge->iy2)
|
|
g->nvs++;
|
|
} else {
|
|
|
|
/*
|
|
* check the end of curve for a not smooth
|
|
* local extremum
|
|
*/
|
|
nge = ge->frwd;
|
|
|
|
if (nge == 0)
|
|
continue;
|
|
else if (nge->type == GE_LINE) {
|
|
nx = nge->ix3;
|
|
ny = nge->iy3;
|
|
} else if (nge->type == GE_CURVE) {
|
|
nx = nge->ix1;
|
|
ny = nge->iy1;
|
|
} else
|
|
continue;
|
|
|
|
/* check for vertical extremums */
|
|
if (ge->iy3 > ge->iy2 && ge->iy3 > ny
|
|
|| ge->iy3 < ge->iy2 && ge->iy3 < ny) {
|
|
hs[g->nhs].value = ge->iy3;
|
|
hs[g->nhs].from
|
|
= hs[g->nhs].to
|
|
= hs[g->nhs].origin = ge->ix3;
|
|
hs[g->nhs].ge = ge->frwd;
|
|
|
|
if (ge->ix3 < ge->ix2
|
|
|| nx < ge->ix3)
|
|
hs[g->nhs].flags = ST_UP;
|
|
else
|
|
hs[g->nhs].flags = 0;
|
|
|
|
if (ge->ix3 != ge->ix2 || nx != ge->ix3)
|
|
g->nhs++;
|
|
}
|
|
/*
|
|
* the same point may be both horizontal and
|
|
* vertical extremum
|
|
*/
|
|
/* check for horizontal extremums */
|
|
if (ge->ix3 > ge->ix2 && ge->ix3 > nx
|
|
|| ge->ix3 < ge->ix2 && ge->ix3 < nx) {
|
|
vs[g->nvs].value = ge->ix3;
|
|
vs[g->nvs].from
|
|
= vs[g->nvs].to
|
|
= vs[g->nvs].origin = ge->iy3;
|
|
vs[g->nvs].ge = ge->frwd;
|
|
|
|
if (ge->iy3 > ge->iy2
|
|
|| ny > ge->iy3)
|
|
vs[g->nvs].flags = ST_UP;
|
|
else
|
|
vs[g->nvs].flags = 0;
|
|
|
|
if (ge->iy3 != ge->iy2 || ny != ge->iy3)
|
|
g->nvs++;
|
|
}
|
|
}
|
|
|
|
} else if (ge->type == GE_LINE) {
|
|
nge = ge->frwd;
|
|
|
|
/* if it is horizontal, add a hstem */
|
|
/* and the ends as vstems if they brace the line */
|
|
if (ge->iy3 == ge->prev->iy3
|
|
&& ge->ix3 != ge->prev->ix3) {
|
|
hs[g->nhs].value = ge->iy3;
|
|
if (ge->ix3 < ge->prev->ix3) {
|
|
hs[g->nhs].flags = ST_FLAT | ST_UP;
|
|
hs[g->nhs].from = ge->ix3;
|
|
hs[g->nhs].to = ge->prev->ix3;
|
|
} else {
|
|
hs[g->nhs].flags = ST_FLAT;
|
|
hs[g->nhs].from = ge->prev->ix3;
|
|
hs[g->nhs].to = ge->ix3;
|
|
}
|
|
hs[g->nhs].origin = ge->ix3;
|
|
hs[g->nhs].ge = ge->frwd;
|
|
|
|
pge = ge->bkwd;
|
|
|
|
/* add beginning as vstem */
|
|
vs[g->nvs].value = pge->ix3;
|
|
vs[g->nvs].origin
|
|
= vs[g->nvs].from
|
|
= vs[g->nvs].to = pge->iy3;
|
|
vs[g->nvs].ge = ge;
|
|
|
|
if(pge->type==GE_CURVE)
|
|
ovalue=pge->iy2;
|
|
else
|
|
ovalue=pge->prev->iy3;
|
|
|
|
if (pge->iy3 > ovalue)
|
|
vs[g->nvs].flags = ST_UP | ST_END;
|
|
else if (pge->iy3 < ovalue)
|
|
vs[g->nvs].flags = ST_END;
|
|
else
|
|
vs[g->nvs].flags = 0;
|
|
|
|
if( vs[g->nvs].flags != 0 )
|
|
g->nvs++;
|
|
|
|
/* add end as vstem */
|
|
vs[g->nvs].value = ge->ix3;
|
|
vs[g->nvs].origin
|
|
= vs[g->nvs].from
|
|
= vs[g->nvs].to = ge->iy3;
|
|
vs[g->nvs].ge = ge->frwd;
|
|
|
|
if(nge->type==GE_CURVE)
|
|
ovalue=nge->iy1;
|
|
else
|
|
ovalue=nge->iy3;
|
|
|
|
if (ovalue > ge->iy3)
|
|
vs[g->nvs].flags = ST_UP | ST_END;
|
|
else if (ovalue < ge->iy3)
|
|
vs[g->nvs].flags = ST_END;
|
|
else
|
|
vs[g->nvs].flags = 0;
|
|
|
|
if( vs[g->nvs].flags != 0 )
|
|
g->nvs++;
|
|
|
|
g->nhs++;
|
|
}
|
|
/* if it is vertical, add a vstem */
|
|
/* and the ends as hstems if they brace the line */
|
|
else if (ge->ix3 == ge->prev->ix3
|
|
&& ge->iy3 != ge->prev->iy3) {
|
|
vs[g->nvs].value = ge->ix3;
|
|
if (ge->iy3 > ge->prev->iy3) {
|
|
vs[g->nvs].flags = ST_FLAT | ST_UP;
|
|
vs[g->nvs].from = ge->prev->iy3;
|
|
vs[g->nvs].to = ge->iy3;
|
|
} else {
|
|
vs[g->nvs].flags = ST_FLAT;
|
|
vs[g->nvs].from = ge->iy3;
|
|
vs[g->nvs].to = ge->prev->iy3;
|
|
}
|
|
vs[g->nvs].origin = ge->iy3;
|
|
vs[g->nvs].ge = ge->frwd;
|
|
|
|
pge = ge->bkwd;
|
|
|
|
/* add beginning as hstem */
|
|
hs[g->nhs].value = pge->iy3;
|
|
hs[g->nhs].origin
|
|
= hs[g->nhs].from
|
|
= hs[g->nhs].to = pge->ix3;
|
|
hs[g->nhs].ge = ge;
|
|
|
|
if(pge->type==GE_CURVE)
|
|
ovalue=pge->ix2;
|
|
else
|
|
ovalue=pge->prev->ix3;
|
|
|
|
if (pge->ix3 < ovalue)
|
|
hs[g->nhs].flags = ST_UP | ST_END;
|
|
else if (pge->ix3 > ovalue)
|
|
hs[g->nhs].flags = ST_END;
|
|
else
|
|
hs[g->nhs].flags = 0;
|
|
|
|
if( hs[g->nhs].flags != 0 )
|
|
g->nhs++;
|
|
|
|
/* add end as hstem */
|
|
hs[g->nhs].value = ge->iy3;
|
|
hs[g->nhs].origin
|
|
= hs[g->nhs].from
|
|
= hs[g->nhs].to = ge->ix3;
|
|
hs[g->nhs].ge = ge->frwd;
|
|
|
|
if(nge->type==GE_CURVE)
|
|
ovalue=nge->ix1;
|
|
else
|
|
ovalue=nge->ix3;
|
|
|
|
if (ovalue < ge->ix3)
|
|
hs[g->nhs].flags = ST_UP | ST_END;
|
|
else if (ovalue > ge->ix3)
|
|
hs[g->nhs].flags = ST_END;
|
|
else
|
|
hs[g->nhs].flags = 0;
|
|
|
|
if( hs[g->nhs].flags != 0 )
|
|
g->nhs++;
|
|
|
|
g->nvs++;
|
|
}
|
|
/*
|
|
* check the end of line for a not smooth local
|
|
* extremum
|
|
*/
|
|
nge = ge->frwd;
|
|
|
|
if (nge == 0)
|
|
continue;
|
|
else if (nge->type == GE_LINE) {
|
|
nx = nge->ix3;
|
|
ny = nge->iy3;
|
|
} else if (nge->type == GE_CURVE) {
|
|
nx = nge->ix1;
|
|
ny = nge->iy1;
|
|
} else
|
|
continue;
|
|
|
|
/* check for vertical extremums */
|
|
if (ge->iy3 > ge->prev->iy3 && ge->iy3 > ny
|
|
|| ge->iy3 < ge->prev->iy3 && ge->iy3 < ny) {
|
|
hs[g->nhs].value = ge->iy3;
|
|
hs[g->nhs].from
|
|
= hs[g->nhs].to
|
|
= hs[g->nhs].origin = ge->ix3;
|
|
hs[g->nhs].ge = ge->frwd;
|
|
|
|
if (ge->ix3 < ge->prev->ix3
|
|
|| nx < ge->ix3)
|
|
hs[g->nhs].flags = ST_UP;
|
|
else
|
|
hs[g->nhs].flags = 0;
|
|
|
|
if (ge->ix3 != ge->prev->ix3 || nx != ge->ix3)
|
|
g->nhs++;
|
|
}
|
|
/*
|
|
* the same point may be both horizontal and vertical
|
|
* extremum
|
|
*/
|
|
/* check for horizontal extremums */
|
|
if (ge->ix3 > ge->prev->ix3 && ge->ix3 > nx
|
|
|| ge->ix3 < ge->prev->ix3 && ge->ix3 < nx) {
|
|
vs[g->nvs].value = ge->ix3;
|
|
vs[g->nvs].from
|
|
= vs[g->nvs].to
|
|
= vs[g->nvs].origin = ge->iy3;
|
|
vs[g->nvs].ge = ge->frwd;
|
|
|
|
if (ge->iy3 > ge->prev->iy3
|
|
|| ny > ge->iy3)
|
|
vs[g->nvs].flags = ST_UP;
|
|
else
|
|
vs[g->nvs].flags = 0;
|
|
|
|
if (ge->iy3 != ge->prev->iy3 || ny != ge->iy3)
|
|
g->nvs++;
|
|
}
|
|
}
|
|
}
|
|
|
|
g->nhs=addbluestems(hs, g->nhs);
|
|
sortstems(hs, g->nhs);
|
|
sortstems(vs, g->nvs);
|
|
|
|
if (ISDBG(STEMS))
|
|
debugstems(g->name, hs, g->nhs, vs, g->nvs);
|
|
|
|
/* find the stems interacting with the Blue Zones */
|
|
markbluestems(hs, g->nhs);
|
|
|
|
if(subhints) {
|
|
if (ISDBG(SUBSTEMS))
|
|
fprintf(pfa_file, "%% %s: joining subst horizontal stems\n", g->name);
|
|
joinsubstems(hs, hs_pairs, g->nhs, 1);
|
|
uniformstems(hs, hs_pairs, g->nhs);
|
|
|
|
if (ISDBG(SUBSTEMS))
|
|
fprintf(pfa_file, "%% %s: joining subst vertical stems\n", g->name);
|
|
joinsubstems(vs, vs_pairs, g->nvs, 0);
|
|
|
|
groupsubstems(g, hs, hs_pairs, g->nhs, vs, vs_pairs, g->nvs);
|
|
}
|
|
|
|
if (ISDBG(MAINSTEMS))
|
|
fprintf(pfa_file, "%% %s: joining main horizontal stems\n", g->name);
|
|
g->nhs = joinmainstems(hs, g->nhs, 1);
|
|
if (ISDBG(MAINSTEMS))
|
|
fprintf(pfa_file, "%% %s: joining main vertical stems\n", g->name);
|
|
g->nvs = joinmainstems(vs, g->nvs, 0);
|
|
|
|
if (ISDBG(MAINSTEMS))
|
|
debugstems(g->name, hs, g->nhs, vs, g->nvs);
|
|
|
|
if(g->nhs > 0) {
|
|
if ((sp = malloc(sizeof(STEM) * g->nhs)) == 0) {
|
|
fprintf(stderr, "**** not enough memory for hints ****\n");
|
|
exit(255);
|
|
}
|
|
g->hstems = sp;
|
|
memcpy(sp, hs, sizeof(STEM) * g->nhs);
|
|
} else
|
|
g->hstems = 0;
|
|
|
|
if(g->nvs > 0) {
|
|
if ((sp = malloc(sizeof(STEM) * g->nvs)) == 0) {
|
|
fprintf(stderr, "**** not enough memory for hints ****\n");
|
|
exit(255);
|
|
}
|
|
g->vstems = sp;
|
|
memcpy(sp, vs, sizeof(STEM) * g->nvs);
|
|
} else
|
|
g->vstems = 0;
|
|
|
|
/* now check that the stems won't overflow the interpreter's stem stack:
|
|
* some interpreters (like X11) push the stems on each change into
|
|
* stack and pop them only after the whole glyphs is completed.
|
|
*/
|
|
|
|
totals = (g->nhs+g->nvs) / 2; /* we count whole stems, not halves */
|
|
lastgrp = -1;
|
|
|
|
for (ge = g->entries; ge != 0; ge = ge->next) {
|
|
grp=ge->stemid;
|
|
if(grp >= 0 && grp != lastgrp) {
|
|
if(grp==0)
|
|
totals += g->nsbs[0];
|
|
else
|
|
totals += g->nsbs[grp] - g->nsbs[grp-1];
|
|
|
|
lastgrp = grp;
|
|
}
|
|
}
|
|
|
|
/* be on the safe side, check for >= , not > */
|
|
if(totals >= max_stemdepth) { /* oops, too deep */
|
|
WARNING_2 {
|
|
fprintf(stderr, "Warning: glyph %s needs hint stack depth %d\n", g->name, totals);
|
|
fprintf(stderr, " (limit %d): removed the substituted hints from it\n", max_stemdepth);
|
|
}
|
|
if(g->nsg > 0) {
|
|
for (ge = g->entries; ge != 0; ge = ge->next)
|
|
ge->stemid = -1;
|
|
free(g->sbstems); g->sbstems = 0;
|
|
free(g->nsbs); g->nsbs = 0;
|
|
g->nsg = 0;
|
|
}
|
|
}
|
|
|
|
/* now check if there are too many main stems */
|
|
totals = (g->nhs+g->nvs) / 2; /* we count whole stems, not halves */
|
|
if(totals >= max_stemdepth) {
|
|
/* even worse, too much of non-substituted stems */
|
|
WARNING_2 {
|
|
fprintf(stderr, "Warning: glyph %s has %d main hints\n", g->name, totals);
|
|
fprintf(stderr, " (limit %d): removed the hints from it\n", max_stemdepth);
|
|
}
|
|
if(g->vstems) {
|
|
free(g->vstems); g->vstems = 0; g->nvs = 0;
|
|
}
|
|
if(g->hstems) {
|
|
free(g->hstems); g->hstems = 0; g->nhs = 0;
|
|
}
|
|
}
|
|
}
|
|
|
|
/* convert weird curves that are close to lines into lines.
|
|
*/
|
|
|
|
void
|
|
fstraighten(
|
|
GLYPH * g
|
|
)
|
|
{
|
|
GENTRY *ge, *pge, *nge, *ige;
|
|
double df;
|
|
int dir;
|
|
double iln, oln;
|
|
int svdir, i, o;
|
|
|
|
for (ige = g->entries; ige != 0; ige = ige->next) {
|
|
if (ige->type != GE_CURVE)
|
|
continue;
|
|
|
|
ge = ige;
|
|
pge = ge->bkwd;
|
|
nge = ge->frwd;
|
|
|
|
df = 0.;
|
|
|
|
/* look for vertical then horizontal */
|
|
for(i=0; i<2; i++) {
|
|
o = !i; /* other axis */
|
|
|
|
iln = fabs(ge->fpoints[i][2] - pge->fpoints[i][2]);
|
|
oln = fabs(ge->fpoints[o][2] - pge->fpoints[o][2]);
|
|
/*
|
|
* if current curve is almost a vertical line, and it
|
|
* doesn't begin or end horizontally (and the prev/next
|
|
* line doesn't join smoothly ?)
|
|
*/
|
|
if( oln < 1.
|
|
|| ge->fpoints[o][2] == ge->fpoints[o][1]
|
|
|| ge->fpoints[o][0] == pge->fpoints[o][2]
|
|
|| iln > 2.
|
|
|| iln > 1. && iln/oln > 0.1 )
|
|
continue;
|
|
|
|
|
|
if(ISDBG(STRAIGHTEN))
|
|
fprintf(stderr,"** straighten almost %s\n", (i? "horizontal":"vertical"));
|
|
|
|
df = ge->fpoints[i][2] - pge->fpoints[i][2];
|
|
dir = fsign(ge->fpoints[o][2] - pge->fpoints[o][2]);
|
|
ge->type = GE_LINE;
|
|
|
|
/*
|
|
* suck in all the sequence of such almost lines
|
|
* going in the same direction but not deviating
|
|
* too far from vertical
|
|
*/
|
|
iln = fabs(nge->fpoints[i][2] - ge->fpoints[i][2]);
|
|
oln = nge->fpoints[o][2] - ge->fpoints[o][2];
|
|
|
|
while (fabs(df) <= 5 && nge->type == GE_CURVE
|
|
&& dir == fsign(oln) /* that also gives oln != 0 */
|
|
&& iln <= 2.
|
|
&& ( iln <= 1. || iln/fabs(oln) <= 0.1 ) ) {
|
|
ge->fx3 = nge->fx3;
|
|
ge->fy3 = nge->fy3;
|
|
|
|
if(ISDBG(STRAIGHTEN))
|
|
fprintf(stderr,"** straighten collapsing %s\n", (i? "horizontal":"vertical"));
|
|
freethisge(nge);
|
|
fixendpath(ge);
|
|
pge = ge->bkwd;
|
|
nge = ge->frwd;
|
|
|
|
df = ge->fpoints[i][2] - pge->fpoints[i][2];
|
|
|
|
iln = fabs(nge->fpoints[i][2] - ge->fpoints[i][2]);
|
|
oln = nge->fpoints[o][2] - ge->fpoints[o][2];
|
|
}
|
|
|
|
/* now check what do we have as previous/next line */
|
|
|
|
if(ge != pge) {
|
|
if( pge->type == GE_LINE && pge->fpoints[i][2] == pge->prev->fpoints[i][2]
|
|
&& fabs(pge->fpoints[o][2] != pge->prev->fpoints[o][2]) ) {
|
|
if(ISDBG(STRAIGHTEN)) fprintf(stderr,"** straighten join with previous 0x%x 0x%x\n", pge, ge);
|
|
/* join the previous line with current */
|
|
pge->fx3 = ge->fx3;
|
|
pge->fy3 = ge->fy3;
|
|
|
|
ige = freethisge(ge)->prev; /* keep the iterator valid */
|
|
ge = pge;
|
|
fixendpath(ge);
|
|
pge = ge->bkwd;
|
|
}
|
|
}
|
|
|
|
if(ge != nge) {
|
|
if (nge->type == GE_LINE && nge->fpoints[i][2] == ge->fpoints[i][2]
|
|
&& fabs(nge->fpoints[o][2] != ge->fpoints[o][2]) ) {
|
|
if(ISDBG(STRAIGHTEN)) fprintf(stderr,"** straighten join with next 0x%x 0x%x\n", ge, nge);
|
|
/* join the next line with current */
|
|
ge->fx3 = nge->fx3;
|
|
ge->fy3 = nge->fy3;
|
|
|
|
freethisge(nge);
|
|
fixendpath(ge);
|
|
pge = ge->bkwd;
|
|
nge = ge->frwd;
|
|
|
|
}
|
|
}
|
|
|
|
if(ge != pge) {
|
|
/* try to align the lines if neccessary */
|
|
if(df != 0.)
|
|
fclosegap(ge, ge, i, df, NULL);
|
|
} else {
|
|
/* contour consists of only one line, get rid of it */
|
|
ige = freethisge(ge); /* keep the iterator valid */
|
|
if(ige == 0) /* this was the last contour */
|
|
return;
|
|
ige = ige->prev;
|
|
}
|
|
|
|
break; /* don't bother looking at the other axis */
|
|
}
|
|
}
|
|
}
|
|
|
|
/* solve a square equation,
|
|
* returns the number of solutions found, the solutions
|
|
* are stored in res which should point to array of two doubles.
|
|
* min and max limit the area for solutions
|
|
*/
|
|
|
|
static int
|
|
fsqequation(
|
|
double a,
|
|
double b,
|
|
double c,
|
|
double *res,
|
|
double min,
|
|
double max
|
|
)
|
|
{
|
|
double D;
|
|
int n;
|
|
|
|
if(ISDBG(SQEQ)) fprintf(stderr, "sqeq(%g,%g,%g) [%g;%g]\n", a, b, c, min, max);
|
|
|
|
if(fabs(a) < 0.000001) { /* if a linear equation */
|
|
n=0;
|
|
if(fabs(b) < 0.000001) /* not an equation at all */
|
|
return 0;
|
|
res[0] = -c/b;
|
|
if(ISDBG(SQEQ)) fprintf(stderr, "sqeq: linear t=%g\n", res[0]);
|
|
if(res[0] >= min && res[0] <= max)
|
|
n++;
|
|
return n;
|
|
}
|
|
|
|
D = b*b - 4.0*a*c;
|
|
if(ISDBG(SQEQ)) fprintf(stderr, "sqeq: D=%g\n", D);
|
|
if(D<0)
|
|
return 0;
|
|
|
|
D = sqrt(D);
|
|
|
|
n=0;
|
|
res[0] = (-b+D) / (2*a);
|
|
if(ISDBG(SQEQ)) fprintf(stderr, "sqeq: t1=%g\n", res[0]);
|
|
if(res[0] >= min && res[0] <= max)
|
|
n++;
|
|
|
|
res[n] = (-b-D) / (2*a);
|
|
if(ISDBG(SQEQ)) fprintf(stderr, "sqeq: t2=%g\n", res[n]);
|
|
if(res[n] >= min && res[n] <= max)
|
|
n++;
|
|
|
|
/* return 2nd solution only if it's different enough */
|
|
if(n==2 && fabs(res[0]-res[1])<0.000001)
|
|
n=1;
|
|
|
|
return n;
|
|
}
|
|
|
|
/* check that the curves don't cross quadrant boundary */
|
|
/* (float) */
|
|
|
|
/*
|
|
Here we make sure that the curve does not continue past
|
|
horizontal or vertical extremums. The horizontal points are
|
|
explained, vertical points are by analogy.
|
|
|
|
The horizontal points are where the derivative
|
|
dy/dx is equal to 0. But the Bezier curves are defined by
|
|
parametric formulas
|
|
x=fx(t)
|
|
y=fy(t)
|
|
so finding this derivative is complicated.
|
|
Also even if we find some point (x,y) splitting at this point
|
|
is far not obvious. Fortunately we can use dy/dt = 0 instead,
|
|
this gets to a rather simple square equation and splitting
|
|
at a known value of t is simple.
|
|
|
|
The formulas are:
|
|
|
|
y = A*(1-t)^3 + 3*B*(1-t)^2*t + 3*C*(1-t)*t^2 + D*t^3
|
|
y = (-A+3*B-3*C+D)*t^3 + (3*A-6*B+3*C)*t^2 + (-3*A+3*B)*t + A
|
|
dy/dt = 3*(-A+3*B-3*C+D)*t^2 + 2*(3*A-6*B+3*C)*t + (-3*A+3*B)
|
|
*/
|
|
|
|
void
|
|
ffixquadrants(
|
|
GLYPH *g
|
|
)
|
|
{
|
|
GENTRY *ge, *nge;
|
|
int i, j, np, oldnp;
|
|
double sp[5]; /* split points, last one empty */
|
|
char dir[5]; /* for debugging, direction by which split happened */
|
|
double a, b, *pts; /* points of a curve */
|
|
|
|
for (ge = g->entries; ge != 0; ge = ge->next) {
|
|
if (ge->type != GE_CURVE)
|
|
continue;
|
|
|
|
doagain:
|
|
np = 0; /* no split points yet */
|
|
if(ISDBG(QUAD)) {
|
|
fprintf(stderr, "%s: trying 0x%x (%g %g) (%g %g) (%g %g) (%g %g)\n ", g->name,
|
|
ge, ge->prev->fx3, ge->prev->fy3, ge->fx1, ge->fy1, ge->fx2, ge->fy2,
|
|
ge->fx3, ge->fy3);
|
|
}
|
|
for(i=0; i<2; i++) { /* first for x then for y */
|
|
/* find the cooridnates of control points */
|
|
a = ge->prev->fpoints[i][2];
|
|
pts = &ge->fpoints[i][0];
|
|
|
|
oldnp = np;
|
|
np += fsqequation(
|
|
3.0*(-a + 3.0*pts[0] - 3.0*pts[1] + pts[2]),
|
|
6.0*(a - 2.0*pts[0] + pts[1]),
|
|
3.0*(-a + pts[0]),
|
|
&sp[np],
|
|
0.0, 1.0); /* XXX range is [0;1] */
|
|
|
|
if(np == oldnp)
|
|
continue;
|
|
|
|
if(ISDBG(QUAD))
|
|
fprintf(stderr, "%s: 0x%x: %d pts(%c): ",
|
|
g->name, ge, np-oldnp, i? 'y':'x');
|
|
|
|
/* remove points that are too close to the ends
|
|
* because hor/vert ends are permitted, also
|
|
* if the split point is VERY close to the ends
|
|
* but not exactly then just flatten it and check again.
|
|
*/
|
|
for(j = oldnp; j<np; j++) {
|
|
dir[j] = i;
|
|
if(ISDBG(QUAD))
|
|
fprintf(stderr, "%g ", sp[j]);
|
|
if(sp[j] < 0.03) { /* front end of curve */
|
|
if(ge->fpoints[i][0] != ge->prev->fpoints[i][2]) {
|
|
ge->fpoints[i][0] = ge->prev->fpoints[i][2];
|
|
if(ISDBG(QUAD)) fprintf(stderr, "flattened at front\n");
|
|
goto doagain;
|
|
}
|
|
if( ge->fpoints[i][1] != ge->fpoints[i][0]
|
|
&& fsign(ge->fpoints[i][2] - ge->fpoints[i][1])
|
|
!= fsign(ge->fpoints[i][1] - ge->fpoints[i][0]) ) {
|
|
ge->fpoints[i][1] = ge->fpoints[i][0];
|
|
if(ISDBG(QUAD)) fprintf(stderr, "flattened zigzag at front\n");
|
|
goto doagain;
|
|
}
|
|
sp[j] = sp[j+1]; np--; j--;
|
|
if(ISDBG(QUAD)) fprintf(stderr, "(front flat) ");
|
|
} else if(sp[j] > 0.97) { /* rear end of curve */
|
|
if(ge->fpoints[i][1] != ge->fpoints[i][2]) {
|
|
ge->fpoints[i][1] = ge->fpoints[i][2];
|
|
if(ISDBG(QUAD)) fprintf(stderr, "flattened at rear\n");
|
|
goto doagain;
|
|
}
|
|
if( ge->fpoints[i][0] != ge->fpoints[i][1]
|
|
&& fsign(ge->prev->fpoints[i][2] - ge->fpoints[i][0])
|
|
!= fsign(ge->fpoints[i][0] - ge->fpoints[i][1]) ) {
|
|
ge->fpoints[i][0] = ge->fpoints[i][1];
|
|
if(ISDBG(QUAD)) fprintf(stderr, "flattened zigzag at rear\n");
|
|
goto doagain;
|
|
}
|
|
sp[j] = sp[j+1]; np--; j--;
|
|
if(ISDBG(QUAD)) fprintf(stderr, "(rear flat) ");
|
|
}
|
|
}
|
|
if(ISDBG(QUAD)) fprintf(stderr, "\n");
|
|
}
|
|
|
|
if(np==0) /* no split points, leave it alone */
|
|
continue;
|
|
|
|
if(ISDBG(QUAD)) {
|
|
fprintf(stderr, "%s: splitting 0x%x (%g %g) (%g %g) (%g %g) (%g %g) at %d points\n ", g->name,
|
|
ge, ge->prev->fx3, ge->prev->fy3, ge->fx1, ge->fy1, ge->fx2, ge->fy2,
|
|
ge->fx3, ge->fy3, np);
|
|
for(i=0; i<np; i++)
|
|
fprintf(stderr, "%g(%c) ", sp[i], dir[i] ? 'y':'x');
|
|
fprintf(stderr, "\n");
|
|
}
|
|
|
|
/* sort the points ascending */
|
|
for(i=0; i<np; i++)
|
|
for(j=i+1; j<np; j++)
|
|
if(sp[i] > sp[j]) {
|
|
a = sp[i]; sp[i] = sp[j]; sp[j] = a;
|
|
}
|
|
|
|
/* now finally do the split on each point */
|
|
for(j=0; j<np; j++) {
|
|
double k1, k2, c;
|
|
|
|
k1 = sp[j];
|
|
k2 = 1 - k1;
|
|
|
|
if(ISDBG(QUAD)) fprintf(stderr, " 0x%x %g/%g\n", ge, k1, k2);
|
|
|
|
nge = newgentry(GEF_FLOAT);
|
|
(*nge) = (*ge);
|
|
|
|
#define SPLIT(pt1, pt2) ( (pt1) + k1*((pt2)-(pt1)) ) /* order is important! */
|
|
for(i=0; i<2; i++) { /* for x and y */
|
|
a = ge->fpoints[i][0]; /* get the middle points */
|
|
b = ge->fpoints[i][1];
|
|
|
|
/* calculate new internal points */
|
|
c = SPLIT(a, b);
|
|
|
|
ge->fpoints[i][0] = SPLIT(ge->prev->fpoints[i][2], a);
|
|
ge->fpoints[i][1] = SPLIT(ge->fpoints[i][0], c);
|
|
|
|
nge->fpoints[i][1] = SPLIT(b, nge->fpoints[i][2]);
|
|
nge->fpoints[i][0] = SPLIT(c, nge->fpoints[i][1]);
|
|
|
|
ge->fpoints[i][2] = SPLIT(ge->fpoints[i][1],
|
|
+ nge->fpoints[i][0]);
|
|
}
|
|
#undef SPLIT
|
|
|
|
addgeafter(ge, nge);
|
|
|
|
/* go to the next part, adjust remaining points */
|
|
ge = nge;
|
|
for(i=j+1; i<np; i++)
|
|
sp[i] = (sp[i]-k1) / k2;
|
|
}
|
|
}
|
|
|
|
}
|
|
|
|
/* check if a curve is a zigzag */
|
|
|
|
static int
|
|
iiszigzag(
|
|
GENTRY *ge
|
|
)
|
|
{
|
|
double k, k1, k2;
|
|
int a, b;
|
|
|
|
if (ge->type != GE_CURVE)
|
|
return 0;
|
|
|
|
a = ge->iy2 - ge->iy1;
|
|
b = ge->ix2 - ge->ix1;
|
|
if(a == 0) {
|
|
if(b == 0) {
|
|
return 0;
|
|
} else
|
|
k = FBIGVAL;
|
|
} else
|
|
k = fabs((double) b / (double) a);
|
|
|
|
a = ge->iy1 - ge->prev->iy3;
|
|
b = ge->ix1 - ge->prev->ix3;
|
|
if(a == 0) {
|
|
if(b == 0) {
|
|
return 0;
|
|
} else
|
|
k1 = FBIGVAL;
|
|
} else
|
|
k1 = fabs((double) b / (double) a);
|
|
|
|
a = ge->iy3 - ge->iy2;
|
|
b = ge->ix3 - ge->ix2;
|
|
if(a == 0) {
|
|
if(b == 0) {
|
|
return 0;
|
|
} else
|
|
k2 = FBIGVAL;
|
|
} else
|
|
k2 = fabs((double) b / (double) a);
|
|
|
|
/* if the curve is not a zigzag */
|
|
if (k1+0.0001 >= k && k2 <= k+0.0001 || k1 <= k+0.0001 && k2+0.0001 >= k)
|
|
return 0;
|
|
else
|
|
return 1;
|
|
}
|
|
|
|
/* check if a curve is a zigzag - floating */
|
|
|
|
static int
|
|
fiszigzag(
|
|
GENTRY *ge
|
|
)
|
|
{
|
|
double k, k1, k2;
|
|
double a, b;
|
|
|
|
if (ge->type != GE_CURVE)
|
|
return 0;
|
|
|
|
a = fabs(ge->fy2 - ge->fy1);
|
|
b = fabs(ge->fx2 - ge->fx1);
|
|
if(a < FEPS) {
|
|
if(b < FEPS) {
|
|
return 0;
|
|
} else
|
|
k = FBIGVAL;
|
|
} else
|
|
k = b / a;
|
|
|
|
a = fabs(ge->fy1 - ge->prev->fy3);
|
|
b = fabs(ge->fx1 - ge->prev->fx3);
|
|
if(a < FEPS) {
|
|
if(b < FEPS) {
|
|
return 0;
|
|
} else
|
|
k1 = FBIGVAL;
|
|
} else
|
|
k1 = b / a;
|
|
|
|
a = fabs(ge->fy3 - ge->fy2);
|
|
b = fabs(ge->fx3 - ge->fx2);
|
|
if(a < FEPS) {
|
|
if(b < FEPS) {
|
|
return 0;
|
|
} else
|
|
k2 = FBIGVAL;
|
|
} else
|
|
k2 = b / a;
|
|
|
|
/* if the curve is not a zigzag */
|
|
if (k1+0.0001 >= k && k2 <= k+0.0001 || k1 <= k+0.0001 && k2+0.0001 >= k)
|
|
return 0;
|
|
else
|
|
return 1;
|
|
}
|
|
|
|
/* split the zigzag-like curves into two parts */
|
|
|
|
void
|
|
fsplitzigzags(
|
|
GLYPH * g
|
|
)
|
|
{
|
|
GENTRY *ge, *nge;
|
|
double a, b, c, d;
|
|
|
|
assertisfloat(g, "splitting zigzags");
|
|
for (ge = g->entries; ge != 0; ge = ge->next) {
|
|
if (ge->type != GE_CURVE)
|
|
continue;
|
|
|
|
/* if the curve is not a zigzag */
|
|
if ( !fiszigzag(ge) ) {
|
|
continue;
|
|
}
|
|
|
|
if(ISDBG(FCONCISE)) {
|
|
double maxsc1, maxsc2;
|
|
fprintf(stderr, "split a zigzag ");
|
|
fnormalizege(ge);
|
|
if( fcrossraysge(ge, ge, &maxsc1, &maxsc2, NULL) ) {
|
|
fprintf(stderr, "sc1=%g sc2=%g\n", maxsc1, maxsc2);
|
|
} else {
|
|
fprintf(stderr, "(rays don't cross)\n");
|
|
}
|
|
}
|
|
/* split the curve by t=0.5 */
|
|
nge = newgentry(GEF_FLOAT);
|
|
(*nge) = (*ge);
|
|
nge->type = GE_CURVE;
|
|
|
|
a = ge->prev->fx3;
|
|
b = ge->fx1;
|
|
c = ge->fx2;
|
|
d = ge->fx3;
|
|
nge->fx3 = d;
|
|
nge->fx2 = (c + d) / 2.;
|
|
nge->fx1 = (b + 2. * c + d) / 4.;
|
|
ge->fx3 = (a + b * 3. + c * 3. + d) / 8.;
|
|
ge->fx2 = (a + 2. * b + c) / 4.;
|
|
ge->fx1 = (a + b) / 2.;
|
|
|
|
a = ge->prev->fy3;
|
|
b = ge->fy1;
|
|
c = ge->fy2;
|
|
d = ge->fy3;
|
|
nge->fy3 = d;
|
|
nge->fy2 = (c + d) / 2.;
|
|
nge->fy1 = (b + 2. * c + d) / 4.;
|
|
ge->fy3 = (a + b * 3. + c * 3. + d) / 8.;
|
|
ge->fy2 = (a + 2. * b + c) / 4.;
|
|
ge->fy1 = (a + b) / 2.;
|
|
|
|
addgeafter(ge, nge);
|
|
|
|
if(ISDBG(FCONCISE)) {
|
|
dumppaths(g, ge, nge);
|
|
}
|
|
}
|
|
}
|
|
|
|
/* free this GENTRY, returns what was ge->next
|
|
* (ge must be of type GE_LINE or GE_CURVE)
|
|
* works on both float and int entries
|
|
*/
|
|
|
|
static GENTRY *
|
|
freethisge(
|
|
GENTRY *ge
|
|
)
|
|
{
|
|
GENTRY *xge;
|
|
|
|
if (ge->bkwd != ge->prev) {
|
|
/* at beginning of the contour */
|
|
|
|
xge = ge->bkwd;
|
|
if(xge == ge) { /* was the only line in contour */
|
|
/* remove the contour completely */
|
|
/* prev is GE_MOVE, next is GE_PATH, remove them all */
|
|
|
|
/* may be the first contour, then ->bkwd points to ge->entries */
|
|
if(ge->prev->prev == 0)
|
|
*(GENTRY **)(ge->prev->bkwd) = ge->next->next;
|
|
else
|
|
ge->prev->prev->next = ge->next->next;
|
|
|
|
if(ge->next->next) {
|
|
ge->next->next->prev = ge->prev->prev;
|
|
ge->next->next->bkwd = ge->prev->bkwd;
|
|
}
|
|
|
|
xge = ge->next->next;
|
|
free(ge->prev); free(ge->next); free(ge);
|
|
return xge;
|
|
}
|
|
|
|
/* move the start point of the contour */
|
|
if(ge->flags & GEF_FLOAT) {
|
|
ge->prev->fx3 = xge->fx3;
|
|
ge->prev->fy3 = xge->fy3;
|
|
} else {
|
|
ge->prev->ix3 = xge->ix3;
|
|
ge->prev->iy3 = xge->iy3;
|
|
}
|
|
} else if(ge->frwd != ge->next) {
|
|
/* at end of the contour */
|
|
|
|
xge = ge->frwd->prev;
|
|
/* move the start point of the contour */
|
|
if(ge->flags & GEF_FLOAT) {
|
|
xge->fx3 = ge->bkwd->fx3;
|
|
xge->fy3 = ge->bkwd->fy3;
|
|
} else {
|
|
xge->ix3 = ge->bkwd->ix3;
|
|
xge->iy3 = ge->bkwd->iy3;
|
|
}
|
|
}
|
|
|
|
ge->prev->next = ge->next;
|
|
ge->next->prev = ge->prev;
|
|
ge->bkwd->frwd = ge->frwd;
|
|
ge->frwd->bkwd = ge->bkwd;
|
|
|
|
xge = ge->next;
|
|
free(ge);
|
|
return xge;
|
|
}
|
|
|
|
/* inserts a new gentry (LINE or CURVE) after another (MOVE
|
|
* or LINE or CURVE)
|
|
* corrects the first GE_MOVE if neccessary
|
|
*/
|
|
|
|
static void
|
|
addgeafter(
|
|
GENTRY *oge, /* after this */
|
|
GENTRY *nge /* insert this */
|
|
)
|
|
{
|
|
if(oge->type == GE_MOVE) {
|
|
/* insert before next */
|
|
if(oge->next->type == GE_PATH) {
|
|
/* first and only GENTRY in path */
|
|
nge->frwd = nge->bkwd = nge;
|
|
} else {
|
|
nge->frwd = oge->next;
|
|
nge->bkwd = oge->next->bkwd;
|
|
oge->next->bkwd->frwd = nge;
|
|
oge->next->bkwd = nge;
|
|
}
|
|
} else {
|
|
nge->frwd = oge->frwd;
|
|
nge->bkwd = oge;
|
|
oge->frwd->bkwd = nge;
|
|
oge->frwd = nge;
|
|
}
|
|
|
|
nge->next = oge->next;
|
|
nge->prev = oge;
|
|
oge->next->prev = nge;
|
|
oge->next = nge;
|
|
|
|
if(nge->frwd->prev->type == GE_MOVE) {
|
|
/* fix up the GE_MOVE entry */
|
|
if(nge->flags & GEF_FLOAT) {
|
|
nge->frwd->prev->fx3 = nge->fx3;
|
|
nge->frwd->prev->fy3 = nge->fy3;
|
|
} else {
|
|
nge->frwd->prev->ix3 = nge->ix3;
|
|
nge->frwd->prev->iy3 = nge->iy3;
|
|
}
|
|
}
|
|
}
|
|
|
|
/*
|
|
* Check if this GENTRY happens to be at the end of path
|
|
* and fix the first MOVETO accordingly
|
|
* handles both int and float
|
|
*/
|
|
|
|
static void
|
|
fixendpath(
|
|
GENTRY *ge
|
|
)
|
|
{
|
|
GENTRY *mge;
|
|
|
|
mge = ge->frwd->prev;
|
|
if(mge->type == GE_MOVE) {
|
|
if(ge->flags & GEF_FLOAT) {
|
|
mge->fx3 = ge->fx3;
|
|
mge->fy3 = ge->fy3;
|
|
} else {
|
|
mge->ix3 = ge->ix3;
|
|
mge->iy3 = ge->iy3;
|
|
}
|
|
}
|
|
}
|
|
|
|
/*
|
|
* This function adjusts the rest of path (the part from...to is NOT changed)
|
|
* to cover the specified gap by the specified axis (0 - X, 1 - Y).
|
|
* Gap is counted in direction (end_of_to - beginning_of_from).
|
|
* Returns by how much the gap was not closed (0.0 if it was fully closed).
|
|
* Ret contains by how much the first and last points of [from...to]
|
|
* were moved to bring them in consistence to the rest of the path.
|
|
* If ret==NULL then this info is not returned.
|
|
*/
|
|
|
|
static double
|
|
fclosegap(
|
|
GENTRY *from,
|
|
GENTRY *to,
|
|
int axis,
|
|
double gap,
|
|
double *ret
|
|
)
|
|
{
|
|
#define TIMESLARGER 10. /* how many times larger must be a curve to not change too much */
|
|
double rm[2];
|
|
double oldpos[2];
|
|
double times, limit, df, dx;
|
|
int j, k;
|
|
GENTRY *xge, *pge, *nge, *bge[2];
|
|
|
|
/* remember the old points to calculate ret */
|
|
oldpos[0] = from->prev->fpoints[axis][2];
|
|
oldpos[1] = to->fpoints[axis][2];
|
|
|
|
rm[0] = rm[1] = gap / 2. ;
|
|
|
|
bge[0] = from; /* this is convenient for iterations */
|
|
bge[1] = to;
|
|
|
|
/* first try to modify large curves but if have none then settle for small */
|
|
for(times = (TIMESLARGER-1); times > 0.1; times /= 2. ) {
|
|
|
|
if(rm[0]+rm[1] == 0.)
|
|
break;
|
|
|
|
/* iterate in both directions, backwards then forwards */
|
|
for(j = 0; j<2; j++) {
|
|
|
|
if(rm[j] == 0.) /* if this direction is exhausted */
|
|
continue;
|
|
|
|
limit = fabs(rm[j]) * (1.+times);
|
|
|
|
for(xge = bge[j]->cntr[j]; xge != bge[!j]; xge = xge->cntr[j]) {
|
|
dx = xge->fpoints[axis][2] - xge->prev->fpoints[axis][2];
|
|
df = fabs(dx) - limit;
|
|
if( df <= FEPS ) /* curve is too small to change */
|
|
continue;
|
|
|
|
if( df >= fabs(rm[j]) )
|
|
df = rm[j];
|
|
else
|
|
df *= fsign(rm[j]); /* we may cover this part of rm */
|
|
|
|
rm[j] -= df;
|
|
limit = fabs(rm[j]) * (1.+times);
|
|
|
|
if(xge->type == GE_CURVE) { /* correct internal points */
|
|
double scale = ((dx+df) / dx) - 1.;
|
|
double base;
|
|
|
|
if(j)
|
|
base = xge->fpoints[axis][2];
|
|
else
|
|
base = xge->prev->fpoints[axis][2];
|
|
|
|
for(k = 0; k<2; k++)
|
|
xge->fpoints[axis][k] += scale *
|
|
(xge->fpoints[axis][k] - base);
|
|
}
|
|
|
|
/* move all the intermediate lines */
|
|
if(j) {
|
|
df = -df; /* absolute direction */
|
|
pge = bge[1]->bkwd;
|
|
nge = xge->bkwd;
|
|
} else {
|
|
xge->fpoints[axis][2] += df;
|
|
pge = bge[0];
|
|
nge = xge->frwd;
|
|
}
|
|
while(nge != pge) {
|
|
if(nge->type == GE_CURVE) {
|
|
nge->fpoints[axis][0] +=df;
|
|
nge->fpoints[axis][1] +=df;
|
|
}
|
|
nge->fpoints[axis][2] += df;
|
|
if(nge->next != nge->frwd) { /* last entry of contour */
|
|
nge->frwd->prev->fpoints[axis][2] += df;
|
|
}
|
|
nge = nge->cntr[!j];
|
|
}
|
|
|
|
if(rm[j] == 0.)
|
|
break;
|
|
}
|
|
}
|
|
}
|
|
|
|
/* find the difference */
|
|
oldpos[0] -= from->prev->fpoints[axis][2];
|
|
oldpos[1] -= to->fpoints[axis][2];
|
|
|
|
if(ret) {
|
|
ret[0] = oldpos[0] - from->prev->fpoints[axis][2];
|
|
ret[1] = oldpos[1] - to->fpoints[axis][2];
|
|
}
|
|
|
|
#if 0
|
|
if( rm[0]+rm[1] != gap - oldpos[1] + oldpos[0]) {
|
|
fprintf(stderr, "** gap=%g rm[0]=%g rm[1]=%g o[0]=%g o[1]=%g rg=%g og=%g\n",
|
|
gap, rm[0], rm[1], oldpos[0], oldpos[1], rm[0]+rm[1],
|
|
gap - oldpos[1] + oldpos[0]);
|
|
}
|
|
#endif
|
|
|
|
return rm[0]+rm[1];
|
|
#undef TIMESLARGER
|
|
}
|
|
|
|
/* remove the lines or curves smaller or equal to the size limit */
|
|
|
|
static void
|
|
fdelsmall(
|
|
GLYPH *g,
|
|
double minlen
|
|
)
|
|
{
|
|
GENTRY *ge, *nge, *pge, *xge, *next;
|
|
int i, k;
|
|
double dx, dy, d2, d2m;
|
|
double minlen2;
|
|
#define TIMESLARGER 10. /* how much larger must be a curve to not change too much */
|
|
|
|
minlen2 = minlen*minlen;
|
|
|
|
for (ge = g->entries; ge != 0; ge = next) {
|
|
next = ge->next;
|
|
|
|
if (ge->type != GE_CURVE && ge->type != GE_LINE)
|
|
continue;
|
|
|
|
d2m = 0;
|
|
for(i= (ge->type==GE_CURVE? 0: 2); i<3; i++) {
|
|
dx = ge->fxn[i] - ge->prev->fx3;
|
|
dy = ge->fyn[i] - ge->prev->fy3;
|
|
d2 = dx*dx + dy*dy;
|
|
if(d2m < d2)
|
|
d2m = d2;
|
|
}
|
|
|
|
if( d2m > minlen2 ) { /* line is not too small */
|
|
/* XXX add more normalization here */
|
|
continue;
|
|
}
|
|
|
|
/* if the line is too small */
|
|
|
|
/* check forwards if we have a whole sequence of them */
|
|
nge = ge;
|
|
for(xge = ge->frwd; xge != ge; xge = xge->frwd) {
|
|
d2m = 0;
|
|
for(i= (xge->type==GE_CURVE? 0: 2); i<3; i++) {
|
|
dx = xge->fxn[i] - xge->prev->fx3;
|
|
dy = xge->fyn[i] - xge->prev->fy3;
|
|
d2 = dx*dx + dy*dy;
|
|
if(d2m < d2)
|
|
d2m = d2;
|
|
}
|
|
if( d2m > minlen2 ) /* line is not too small */
|
|
break;
|
|
nge = xge;
|
|
if(next == nge) /* move the next step past this sequence */
|
|
next = next->next;
|
|
}
|
|
|
|
/* check backwards if we have a whole sequence of them */
|
|
pge = ge;
|
|
for(xge = ge->bkwd; xge != ge; xge = xge->bkwd) {
|
|
d2m = 0;
|
|
for(i= (xge->type==GE_CURVE? 0: 2); i<3; i++) {
|
|
dx = xge->fxn[i] - xge->prev->fx3;
|
|
dy = xge->fyn[i] - xge->prev->fy3;
|
|
d2 = dx*dx + dy*dy;
|
|
if(d2m < d2)
|
|
d2m = d2;
|
|
}
|
|
if( d2m > minlen2 ) /* line is not too small */
|
|
break;
|
|
pge = xge;
|
|
}
|
|
|
|
/* now we have a sequence of small fragments in pge...nge (inclusive) */
|
|
|
|
if(ISDBG(FCONCISE)) {
|
|
fprintf(stderr, "glyph %s has very small fragments(%x..%x..%x)\n",
|
|
g->name, pge, ge, nge);
|
|
dumppaths(g, pge, nge);
|
|
}
|
|
|
|
/* reduce whole sequence to one part and remember the middle point */
|
|
if(pge != nge) {
|
|
while(1) {
|
|
xge = pge->frwd;
|
|
if(xge == nge) {
|
|
pge->fx1 = pge->fx2 = pge->fx3;
|
|
pge->fx3 = nge->fx3;
|
|
pge->fy1 = pge->fy2 = pge->fy3;
|
|
pge->fy3 = nge->fy3;
|
|
pge->type = GE_CURVE;
|
|
freethisge(nge);
|
|
break;
|
|
}
|
|
if(xge == nge->bkwd) {
|
|
pge->fx1 = pge->fx2 = (pge->fx3+xge->fx3)/2.;
|
|
pge->fx3 = nge->fx3;
|
|
pge->fy1 = pge->fy2 = (pge->fy3+xge->fy3)/2.;
|
|
pge->fy3 = nge->fy3;
|
|
pge->type = GE_CURVE;
|
|
freethisge(nge);
|
|
freethisge(xge);
|
|
break;
|
|
}
|
|
freethisge(pge); pge = xge;
|
|
xge = nge->bkwd; freethisge(nge); nge = xge;
|
|
}
|
|
}
|
|
ge = pge;
|
|
|
|
/* check if the whole sequence is small */
|
|
dx = ge->fx3 - ge->prev->fx3;
|
|
dy = ge->fy3 - ge->prev->fy3;
|
|
d2 = dx*dx + dy*dy;
|
|
|
|
if( d2 > minlen2 ) { /* no, it is not */
|
|
double b, d;
|
|
|
|
WARNING_3 fprintf(stderr, "glyph %s had a sequence of fragments < %g points each, reduced to one curve\n",
|
|
g->name, minlen);
|
|
|
|
/* check that we did not create a monstrosity spanning quadrants */
|
|
if(fsign(ge->fx1 - ge->prev->fx1) * fsign(ge->fx3 - ge->fx1) < 0
|
|
|| fsign(ge->fy1 - ge->prev->fy1) * fsign(ge->fy3 - ge->fy1) < 0 ) {
|
|
/* yes, we did; are both parts of this thing big enough ? */
|
|
dx = ge->fx1 - ge->prev->fx3;
|
|
dy = ge->fy1 - ge->prev->fy3;
|
|
d2 = dx*dx + dy*dy;
|
|
|
|
dx = ge->fx3 - ge->fx1;
|
|
dy = ge->fy3 - ge->fy1;
|
|
d2m = dx*dx + dy*dy;
|
|
|
|
if(d2 > minlen2 && d2m > minlen2) { /* make two straights */
|
|
nge = newgentry(GEF_FLOAT);
|
|
*nge = *ge;
|
|
|
|
for(i=0; i<2; i++) {
|
|
ge->fpoints[i][2] = ge->fpoints[i][0];
|
|
b = nge->fpoints[i][0];
|
|
d = nge->fpoints[i][2] - b;
|
|
nge->fpoints[i][0] = b + 0.1*d;
|
|
nge->fpoints[i][1] = b + 0.9*d;
|
|
}
|
|
}
|
|
for(i=0; i<2; i++) { /* make one straight or first of two straights */
|
|
b = ge->prev->fpoints[i][2];
|
|
d = ge->fpoints[i][2] - b;
|
|
ge->fpoints[i][0] = b + 0.1*d;
|
|
ge->fpoints[i][1] = b + 0.9*d;
|
|
}
|
|
}
|
|
continue;
|
|
}
|
|
|
|
if(ge->frwd == ge) { /* points to itself, just remove the path completely */
|
|
WARNING_3 fprintf(stderr, "glyph %s had a path made of fragments < %g points each, removed\n",
|
|
g->name, minlen);
|
|
|
|
next = freethisge(ge);
|
|
continue;
|
|
}
|
|
|
|
/* now close the gap by x and y */
|
|
for(i=0; i<2; i++) {
|
|
double gap;
|
|
|
|
gap = ge->fpoints[i][2] - ge->prev->fpoints[i][2];
|
|
if( fclosegap(ge, ge, i, gap, NULL) != 0.0 ) {
|
|
double scale, base;
|
|
|
|
/* not good, as the last resort just scale the next line */
|
|
gap = ge->fpoints[i][2] - ge->prev->fpoints[i][2];
|
|
|
|
if(ISDBG(FCONCISE))
|
|
fprintf(stderr, " last resort on %c: closing next by %g\n",
|
|
(i==0 ? 'x' : 'y'), gap);
|
|
|
|
nge = ge->frwd;
|
|
base = nge->fpoints[i][2];
|
|
dx = ge->fpoints[i][2] - base;
|
|
if(fabs(dx) < FEPS)
|
|
continue;
|
|
|
|
scale = ((dx-gap) / dx);
|
|
|
|
if(nge->type == GE_CURVE)
|
|
for(k = 0; k<2; k++)
|
|
nge->fpoints[i][k] = base +
|
|
scale * (nge->fpoints[i][k] - base);
|
|
|
|
ge->fpoints[i][2] -= gap;
|
|
}
|
|
}
|
|
|
|
/* OK, the gap is closed - remove this useless GENTRY */
|
|
freethisge(ge);
|
|
}
|
|
#undef TIMESLARGER
|
|
}
|
|
|
|
/* find the point where two rays continuing vectors cross
|
|
* returns 1 if they cross, 0 if they don't
|
|
* If they cross optionally (if the pointers are not NULL) returns
|
|
* the maximal scales for both vectors and also optionally the point
|
|
* where the rays cross (twice).
|
|
* Expects that the curves are normalized.
|
|
*
|
|
* For convenience there are 2 front-end functions taking
|
|
* arguments in different formats
|
|
*/
|
|
|
|
struct ray {
|
|
double x1, y1, x2, y2;
|
|
int isvert;
|
|
double k, b; /* lines are represented as y = k*x + b */
|
|
double *maxp;
|
|
};
|
|
static struct ray ray[3];
|
|
|
|
/* the back-end doing the actual work
|
|
* the rays are defined in the static array ray[]
|
|
*/
|
|
|
|
static int
|
|
fcrossraysxx(
|
|
double crossdot[2][2]
|
|
)
|
|
{
|
|
double x, y, max;
|
|
int i;
|
|
|
|
for(i=0; i<2; i++) {
|
|
if(ray[i].x1 == ray[i].x2)
|
|
ray[i].isvert = 1;
|
|
else {
|
|
ray[i].isvert = 0;
|
|
ray[i].k = (ray[i].y2 - ray[i].y1) / (ray[i].x2 - ray[i].x1);
|
|
ray[i].b = ray[i].y2 - ray[i].k * ray[i].x2;
|
|
}
|
|
}
|
|
|
|
if(ray[0].isvert && ray[1].isvert) {
|
|
if(ISDBG(FCONCISE)) fprintf(stderr, "crossrays: both vertical\n");
|
|
return 0; /* both vertical, don't cross */
|
|
}
|
|
|
|
if(ray[1].isvert) {
|
|
ray[2] = ray[0]; /* exchange them */
|
|
ray[0] = ray[1];
|
|
ray[1] = ray[2];
|
|
}
|
|
|
|
if(ray[0].isvert) {
|
|
x = ray[0].x1;
|
|
} else {
|
|
if( fabs(ray[0].k - ray[1].k) < FEPS) {
|
|
if(ISDBG(FCONCISE)) fprintf(stderr, "crossrays: parallel lines, k = %g, %g\n",
|
|
ray[0].k, ray[1].k);
|
|
return 0; /* parallel lines */
|
|
}
|
|
x = (ray[1].b - ray[0].b) / (ray[0].k - ray[1].k) ;
|
|
}
|
|
y = ray[1].k * x + ray[1].b;
|
|
|
|
for(i=0; i<2; i++) {
|
|
if(ray[i].isvert)
|
|
max = (y - ray[i].y1) / (ray[i].y2 - ray[i].y1);
|
|
else
|
|
max = (x - ray[i].x1) / (ray[i].x2 - ray[i].x1);
|
|
/* check if wrong sides of rays cross */
|
|
if( max < 0 ) {
|
|
if(ISDBG(FCONCISE)) fprintf(stderr, "crossrays: %c scale=%g @(%g,%g) (%g,%g)<-(%g,%g)\n",
|
|
(i?'Y':'X'), max, x, y, ray[i].x2, ray[i].y2, ray[i].x1, ray[i].y1);
|
|
return 0;
|
|
}
|
|
if(ray[i].maxp)
|
|
*ray[i].maxp = max;
|
|
}
|
|
if(crossdot != 0) {
|
|
crossdot[0][0] = crossdot[1][0] = x;
|
|
crossdot[0][1] = crossdot[1][1] = y;
|
|
}
|
|
return 1;
|
|
}
|
|
|
|
/* the front-end getting the arguments from 4 dots defining
|
|
* a curve in the same format as for fapproxcurve():
|
|
* rays are defined as beginning and end of the curve,
|
|
* the crossdot is inserted as the two middle dots of the curve
|
|
*/
|
|
|
|
int
|
|
fcrossrayscv(
|
|
double curve[4][2 /*X,Y*/],
|
|
double *max1,
|
|
double *max2
|
|
)
|
|
{
|
|
ray[0].x1 = curve[0][X];
|
|
ray[0].y1 = curve[0][Y];
|
|
ray[0].x2 = curve[1][X];
|
|
ray[0].y2 = curve[1][Y];
|
|
ray[0].maxp = max1;
|
|
|
|
ray[1].x1 = curve[2][X];
|
|
ray[1].y1 = curve[2][Y];
|
|
ray[1].x2 = curve[3][X];
|
|
ray[1].y2 = curve[3][Y];
|
|
ray[1].maxp = max2;
|
|
|
|
return fcrossraysxx(&curve[1]);
|
|
}
|
|
|
|
/* the front-end getting the arguments from gentries:
|
|
* rays are defined as beginning of curve1 and end of curve 2
|
|
*/
|
|
|
|
int
|
|
fcrossraysge(
|
|
GENTRY *ge1,
|
|
GENTRY *ge2,
|
|
double *max1,
|
|
double *max2,
|
|
double crossdot[2][2]
|
|
)
|
|
{
|
|
ray[0].x1 = ge1->prev->fx3;
|
|
ray[0].y1 = ge1->prev->fy3;
|
|
ray[0].x2 = ge1->fpoints[X][ge1->ftg];
|
|
ray[0].y2 = ge1->fpoints[Y][ge1->ftg];
|
|
ray[0].maxp = max1;
|
|
|
|
ray[1].x1 = ge2->fx3;
|
|
ray[1].y1 = ge2->fy3;
|
|
if(ge2->rtg < 0) {
|
|
ray[1].x2 = ge2->prev->fx3;
|
|
ray[1].y2 = ge2->prev->fy3;
|
|
} else {
|
|
ray[1].x2 = ge2->fpoints[X][ge2->rtg];
|
|
ray[1].y2 = ge2->fpoints[Y][ge2->rtg];
|
|
}
|
|
ray[1].maxp = max2;
|
|
|
|
return fcrossraysxx(crossdot);
|
|
}
|
|
|
|
/* debugging printout functions */
|
|
|
|
#if defined(DEBUG_DOTSEG) || defined(DEBUG_DOTCURVE) || defined(DEBUG_APPROXCV)
|
|
|
|
/* for debugging */
|
|
static
|
|
printdot(
|
|
double dot[2]
|
|
)
|
|
{
|
|
fprintf(stderr, "(%g,%g)", dot[0], dot[1]);
|
|
}
|
|
|
|
static
|
|
printseg(
|
|
double seg[2][2]
|
|
)
|
|
{
|
|
putc('[', stderr);
|
|
printdot(seg[0]);
|
|
putc(' ', stderr);
|
|
printdot(seg[1]);
|
|
putc(']', stderr);
|
|
}
|
|
|
|
#endif /* DEBUG_* */
|
|
|
|
/*
|
|
* Find squared distance from a dot to a line segment
|
|
*/
|
|
|
|
double
|
|
fdotsegdist2(
|
|
double seg[2][2 /*X,Y*/],
|
|
double dot[2 /*X,Y*/]
|
|
)
|
|
{
|
|
#define x1 seg[0][X]
|
|
#define y1 seg[0][Y]
|
|
#define x2 seg[1][X]
|
|
#define y2 seg[1][Y]
|
|
#define xdot dot[X]
|
|
#define ydot dot[Y]
|
|
|
|
double dx, dy; /* segment dimensions */
|
|
double kline, bline; /* segment line formula is y=k*x+b */
|
|
double kperp, bperp; /* perpendicular from the dot to the line */
|
|
double xcross, ycross; /* where the perpendicular crosses the segment */
|
|
|
|
/* handle the situation where the nearest point of the segment is its end */
|
|
#define HANDLE_LIMITS(less12, lesscr1, lesscr2) \
|
|
if( less12 ) { \
|
|
if( lesscr1 ) { \
|
|
xcross = x1; \
|
|
ycross = y1; \
|
|
} else if( !(lesscr2) ) { \
|
|
xcross = x2; \
|
|
ycross = y2; \
|
|
} \
|
|
} else { \
|
|
if( !(lesscr1) ) { \
|
|
xcross = x1; \
|
|
ycross = y1; \
|
|
} else if( lesscr2 ) { \
|
|
xcross = x2; \
|
|
ycross = y2; \
|
|
} \
|
|
} \
|
|
/* end of macro */
|
|
|
|
|
|
dx = x2 - x1;
|
|
dy = y2 - y1;
|
|
|
|
if(fabs(dx) < FEPS) {
|
|
/* special case - vertical line */
|
|
#ifdef DEBUG_DOTSEG
|
|
printf("vertical line!\n");
|
|
#endif
|
|
xcross = x1;
|
|
ycross = ydot;
|
|
HANDLE_LIMITS( y1 < y2, ycross < y1, ycross < y2);
|
|
} else if(fabs(dy) < FEPS) {
|
|
/* special case - horizontal line */
|
|
#ifdef DEBUG_DOTSEG
|
|
printf("horizontal line!\n");
|
|
#endif
|
|
xcross = xdot;
|
|
ycross = y1;
|
|
HANDLE_LIMITS( x1 < x2, xcross < x1, xcross < x2)
|
|
} else {
|
|
kline = dy/dx;
|
|
bline = y1 - x1*kline;
|
|
kperp = -1./kline;
|
|
bperp = ydot - xdot*kperp;
|
|
|
|
xcross = (bline-bperp) / (kperp-kline);
|
|
ycross = kline*xcross + bline;
|
|
|
|
HANDLE_LIMITS( x1 < x2, xcross < x1, xcross < x2)
|
|
}
|
|
#ifdef DEBUG_DOTSEG
|
|
printf("crossover at (%g,%g)\n", xcross, ycross);
|
|
#endif
|
|
|
|
dx = xdot-xcross;
|
|
dy = ydot-ycross;
|
|
return dx*dx+dy*dy;
|
|
#undef x1
|
|
#undef y1
|
|
#undef x2
|
|
#undef y2
|
|
#undef xdot
|
|
#undef ydot
|
|
#undef HANDLE_LIMITS
|
|
}
|
|
|
|
/* find the weighted quadratic average for the distance of a set
|
|
* of dots from the curve; also fills out the individual distances
|
|
* for each dot; if maxp!=NULL then returns the maximal squared
|
|
* distance in there
|
|
*/
|
|
|
|
double
|
|
fdotcurvdist2(
|
|
double curve[4][2 /*X,Y*/ ],
|
|
struct dot_dist *dots,
|
|
int ndots, /* number of entries in dots */
|
|
double *maxp
|
|
)
|
|
{
|
|
/* a curve is approximated by this many straight segments */
|
|
#define NAPSECT 16
|
|
/* a curve is divided into this many sections with equal weight each */
|
|
#define NWSECT 4
|
|
/* table of coefficients for finding the dots on the curve */
|
|
/* tt[0] is left unused */
|
|
static double tt[NAPSECT][4];
|
|
static int havett = 0; /* flag: tt is initialized */
|
|
/* dots on the curve */
|
|
double cvd[NAPSECT+1][2 /*X,Y*/];
|
|
/* sums by sections */
|
|
double sum[NWSECT];
|
|
/* counts by sections */
|
|
double count[NWSECT];
|
|
int d, i, j;
|
|
int id1, id2;
|
|
double dist1, dist2, dist3, dx, dy, x, y;
|
|
double max = 0.;
|
|
|
|
if(!havett) {
|
|
double t, nt, t2, nt2, step;
|
|
|
|
havett++;
|
|
step = 1. / NAPSECT;
|
|
t = 0;
|
|
for(i=1; i<NAPSECT; i++) {
|
|
t += step;
|
|
nt = 1 - t;
|
|
t2 = t*t;
|
|
nt2 = nt*nt;
|
|
tt[i][0] = nt2*nt; /* (1-t)^3 */
|
|
tt[i][1] = 3*nt2*t; /* 3*(1-t)^2*t */
|
|
tt[i][2] = 3*nt*t2; /* 3*(1-t)*t^2 */
|
|
tt[i][3] = t2*t; /* t^3 */
|
|
}
|
|
}
|
|
|
|
for(i=0; i<NWSECT; i++) {
|
|
sum[i] = 0.;
|
|
count[i] = 0;
|
|
}
|
|
|
|
/* split the curve into segments */
|
|
for(d=0; d<2; d++) { /* X and Y */
|
|
cvd[0][d] = curve[0][d]; /* endpoints */
|
|
cvd[NAPSECT][d] = curve[3][d];
|
|
for(i=1; i<NAPSECT; i++) {
|
|
cvd[i][d] = curve[0][d] * tt[i][0]
|
|
+ curve[1][d] * tt[i][1]
|
|
+ curve[2][d] * tt[i][2]
|
|
+ curve[3][d] * tt[i][3];
|
|
}
|
|
}
|
|
|
|
for(d=0; d<ndots; d++) {
|
|
#ifdef DEBUG_DOTCURVE
|
|
printf("dot %d ", d); printdot(dots[d].p); printf(":\n");
|
|
|
|
/* for debugging */
|
|
for(i=0; i< NAPSECT; i++) {
|
|
dist1 = fdotsegdist2(&cvd[i], dots[d].p);
|
|
printf(" seg %d ",i); printseg(&cvd[i]); printf(" dist=%g\n", sqrt(dist1));
|
|
}
|
|
#endif
|
|
|
|
x = dots[d].p[X];
|
|
y = dots[d].p[Y];
|
|
|
|
/* find the nearest dot on the curve
|
|
* there may be up to 2 local minimums - so we start from the
|
|
* ends of curve and go to the center
|
|
*/
|
|
|
|
id1 = 0;
|
|
dx = x - cvd[0][X];
|
|
dy = y - cvd[0][Y];
|
|
dist1 = dx*dx + dy*dy;
|
|
#ifdef DEBUG_DOTCURVE
|
|
printf(" dot 0 "); printdot(cvd[id1]); printf(" dist=%g\n", sqrt(dist1));
|
|
#endif
|
|
for(i = 1; i<=NAPSECT; i++) {
|
|
dx = x - cvd[i][X];
|
|
dy = y - cvd[i][Y];
|
|
dist3 = dx*dx + dy*dy;
|
|
#ifdef DEBUG_DOTCURVE
|
|
printf(" dot %d ",i); printdot(cvd[i]); printf(" dist=%g\n", sqrt(dist3));
|
|
#endif
|
|
if(dist3 < dist1) {
|
|
dist1 = dist3;
|
|
id1 = i;
|
|
} else
|
|
break;
|
|
}
|
|
|
|
if(id1 < NAPSECT-1) {
|
|
id2 = NAPSECT;
|
|
dx = x - cvd[NAPSECT][X];
|
|
dy = y - cvd[NAPSECT][Y];
|
|
dist2 = dx*dx + dy*dy;
|
|
#ifdef DEBUG_DOTCURVE
|
|
printf(" +dot %d ", id2); printdot(cvd[id2]); printf(" dist=%g\n", sqrt(dist2));
|
|
#endif
|
|
for(i = NAPSECT-1; i>id1+1; i--) {
|
|
dx = x - cvd[i][X];
|
|
dy = y - cvd[i][Y];
|
|
dist3 = dx*dx + dy*dy;
|
|
#ifdef DEBUG_DOTCURVE
|
|
printf(" dot %d ",i); printdot(cvd[i]); printf(" dist=%g\n", sqrt(dist3));
|
|
#endif
|
|
if(dist3 < dist2) {
|
|
dist2 = dist3;
|
|
id2 = i;
|
|
} else
|
|
break;
|
|
}
|
|
|
|
/* now find which of the local minimums is smaller */
|
|
if(dist2 < dist1) {
|
|
id1 = id2;
|
|
}
|
|
}
|
|
|
|
/* the nearest segment must include the nearest dot */
|
|
if(id1==0) {
|
|
dots[d].seg = 0;
|
|
dots[d].dist2 = fdotsegdist2(&cvd[0], dots[d].p);
|
|
} else if(id1==NAPSECT) {
|
|
dots[d].seg = NAPSECT-1;
|
|
dots[d].dist2 = fdotsegdist2(&cvd[NAPSECT-1], dots[d].p);
|
|
} else {
|
|
dist1 = fdotsegdist2(&cvd[id1], dots[d].p);
|
|
dist2 = fdotsegdist2(&cvd[id1-1], dots[d].p);
|
|
if(dist2 < dist1) {
|
|
dots[d].seg = id1-1;
|
|
dots[d].dist2 = dist2;
|
|
} else {
|
|
dots[d].seg = id1;
|
|
dots[d].dist2 = dist1;
|
|
}
|
|
}
|
|
|
|
i = dots[d].seg % NWSECT;
|
|
sum[i] += dots[d].dist2;
|
|
if(dots[d].dist2 > max)
|
|
max = dots[d].dist2;
|
|
count[i]++;
|
|
#ifdef DEBUG_DOTCURVE
|
|
printf(" best seg %d sect %d dist=%g\n", dots[d].seg, i, sqrt(dots[d].dist2));
|
|
#endif
|
|
}
|
|
|
|
/* calculate the weighted average */
|
|
id1=0;
|
|
dist1=0.;
|
|
for(i=0; i<NWSECT; i++) {
|
|
if(count[i]==0)
|
|
continue;
|
|
id1++;
|
|
dist1 += sum[i]/count[i];
|
|
}
|
|
if(maxp)
|
|
*maxp = max;
|
|
if(id1==0) /* no dots, strange */
|
|
return 0.;
|
|
else
|
|
return dist1/id1; /* to get the average distance apply sqrt() */
|
|
}
|
|
|
|
/*
|
|
* Approximate a curve matching the giving set of points and with
|
|
* middle reference points going along the given segments (and no farther
|
|
* than these segments).
|
|
*/
|
|
|
|
void
|
|
fapproxcurve(
|
|
double cv[4][2 /*X,Y*/ ], /* points 0-3 are passed in, points 1,2 - out */
|
|
struct dot_dist *dots, /* the dots to approximate - distances returned
|
|
* there may be invalid */
|
|
int ndots
|
|
)
|
|
{
|
|
/* b and c are the middle control points */
|
|
#define B 0
|
|
#define C 1
|
|
/* maximal number of sections on each axis - used for the first step */
|
|
#define MAXSECT 2
|
|
/* number of sections used for the other steps */
|
|
#define NORMSECT 2
|
|
/* when the steps become less than this many points, it's time to stop */
|
|
#define STEPEPS 1.
|
|
double from[2 /*B,C*/], to[2 /*B,C*/];
|
|
double middf[2 /*B,C*/][2 /*X,Y*/], df;
|
|
double coef[2 /*B,C*/][MAXSECT];
|
|
double res[MAXSECT][MAXSECT], thisres, bestres, goodres;
|
|
int ncoef[2 /*B,C*/], best[2 /*B,C*/], good[2 /*B,C*/];
|
|
int i, j, k, keepsym;
|
|
char bc[]="BC";
|
|
char xy[]="XY";
|
|
|
|
#ifdef DEBUG_APPROXCV
|
|
fprintf(stderr, "Curve points:");
|
|
for(i=0; i<4; i++) {
|
|
fprintf(stderr, " ");
|
|
printdot(cv[i]);
|
|
}
|
|
fprintf(stderr, "\nDots:");
|
|
for(i=0; i<ndots; i++) {
|
|
fprintf(stderr, " ");
|
|
printdot(dots[i].p);
|
|
}
|
|
fprintf(stderr, "\n");
|
|
#endif
|
|
|
|
/* load the endpoints and calculate differences */
|
|
for(i=0; i<2; i++) {
|
|
/* i is X, Y */
|
|
middf[B][i] = cv[1][i]-cv[0][i];
|
|
middf[C][i] = cv[2][i]-cv[3][i];
|
|
|
|
/* i is B, C */
|
|
from[i] = 0.;
|
|
to[i] = 1.;
|
|
ncoef[i] = MAXSECT;
|
|
}
|
|
|
|
while(ncoef[B] != 1 || ncoef[C] != 1) {
|
|
/* prepare the values of coefficients */
|
|
for(i=0; i<2; i++) { /*B,C*/
|
|
#ifdef DEBUG_APPROXCV
|
|
fprintf(stderr, "Coefficients by %c(%g,%g):", bc[i], from[i], to[i]);
|
|
#endif
|
|
df = (to[i]-from[i]) / (ncoef[i]*2);
|
|
for(j=0; j<ncoef[i]; j++) {
|
|
coef[i][j] = from[i] + df*(2*j+1);
|
|
#ifdef DEBUG_APPROXCV
|
|
fprintf(stderr, " %g", coef[i][j]);
|
|
#endif
|
|
}
|
|
#ifdef DEBUG_APPROXCV
|
|
fprintf(stderr, "\n");
|
|
#endif
|
|
}
|
|
bestres = FBIGVAL;
|
|
best[B] = best[C] = 0;
|
|
/* i iterates by ncoef[B], j iterates by ncoef[C] */
|
|
for(i=0; i<ncoef[B]; i++) {
|
|
for(j=0; j<ncoef[C]; j++) {
|
|
for(k=0; k<2; k++) { /*X, Y*/
|
|
cv[1][k] = cv[0][k] + middf[B][k]*coef[B][i];
|
|
cv[2][k] = cv[3][k] + middf[C][k]*coef[C][j];
|
|
}
|
|
res[i][j] = thisres = fdotcurvdist2(cv, dots, ndots, NULL);
|
|
if(thisres < bestres) {
|
|
goodres = bestres;
|
|
good[B] = best[B];
|
|
good[C] = best[C];
|
|
bestres = thisres;
|
|
best[B] = i;
|
|
best[C] = j;
|
|
} else if(thisres < goodres) {
|
|
goodres = thisres;
|
|
good[B] = i;
|
|
good[C] = j;
|
|
}
|
|
#ifdef DEBUG_APPROXCV
|
|
fprintf(stderr, " at (%g,%g) dist=%g %s\n", coef[B][i], coef[C][j], sqrt(thisres),
|
|
(best[B]==i && best[C]==j)? "(BEST)":"");
|
|
#endif
|
|
}
|
|
}
|
|
#ifdef DEBUG_APPROXCV
|
|
fprintf(stderr, " best: at (%g, %g) dist=%g\n",
|
|
coef[B][best[B]], coef[C][best[C]], sqrt(bestres));
|
|
fprintf(stderr, " B:%d,%d C:%d,%d -- 2nd best: at (%g, %g) dist=%g\n",
|
|
best[B], good[B], best[C], good[C], coef[B][good[B]], coef[C][good[C]], sqrt(goodres));
|
|
#endif
|
|
|
|
if(bestres < (0.1*0.1)) { /* consider it close enough */
|
|
/* calculate the coordinates to return */
|
|
for(k=0; k<2; k++) { /*X, Y*/
|
|
cv[1][k] = cv[0][k] + middf[B][k]*coef[B][best[B]];
|
|
cv[2][k] = cv[3][k] + middf[C][k]*coef[C][best[C]];
|
|
}
|
|
#ifdef DEBUG_APPROXCV
|
|
fprintf(stderr, "quick approximated middle points "); printdot(cv[1]);
|
|
fprintf(stderr, " "); printdot(cv[2]); fprintf(stderr, "\n");
|
|
#endif
|
|
return;
|
|
}
|
|
keepsym = 0;
|
|
if(best[B] != best[C] && best[B]-best[C] == good[C]-good[B]) {
|
|
keepsym = 1;
|
|
#ifdef DEBUG_APPROXCV
|
|
fprintf(stderr, "keeping symmetry!\n");
|
|
#endif
|
|
}
|
|
for(i=0; i<2; i++) { /*B,C*/
|
|
if(ncoef[i]==1)
|
|
continue;
|
|
if(keepsym) {
|
|
/* try to keep the symmetry */
|
|
if(best[i] < good[i]) {
|
|
from[i] = coef[i][best[i]];
|
|
to[i] = coef[i][good[i]];
|
|
} else {
|
|
from[i] = coef[i][good[i]];
|
|
to[i] = coef[i][best[i]];
|
|
}
|
|
} else {
|
|
df = (to[i]-from[i]) / ncoef[i];
|
|
from[i] += df*best[i];
|
|
to[i] = from[i] + df;
|
|
}
|
|
if( fabs(df*middf[i][0]) < STEPEPS && fabs(df*middf[i][1]) < STEPEPS) {
|
|
/* this side has converged */
|
|
from[i] = to[i] = (from[i]+to[i]) / 2.;
|
|
ncoef[i] = 1;
|
|
} else
|
|
ncoef[i] = NORMSECT;
|
|
}
|
|
|
|
}
|
|
/* calculate the coordinates to return */
|
|
for(k=0; k<2; k++) { /*X, Y*/
|
|
cv[1][k] = cv[0][k] + middf[B][k]*from[B];
|
|
cv[2][k] = cv[3][k] + middf[C][k]*from[C];
|
|
}
|
|
#ifdef DEBUG_APPROXCV
|
|
fprintf(stderr, "approximated middle points "); printdot(cv[1]);
|
|
fprintf(stderr, " "); printdot(cv[2]); fprintf(stderr, "\n");
|
|
#endif
|
|
#undef B
|
|
#undef C
|
|
#undef MAXSECT
|
|
#undef NORMSECT
|
|
#undef STEPEPS
|
|
}
|
|
|
|
/*
|
|
* Find the squared value of the sinus of the angle between the
|
|
* end of ge1 and the beginning of ge2
|
|
* The curve must be normalized.
|
|
*/
|
|
|
|
static double
|
|
fjointsin2(
|
|
GENTRY *ge1,
|
|
GENTRY *ge2
|
|
)
|
|
{
|
|
double d[3][2 /*X,Y*/];
|
|
double scale1, scale2, len1, len2;
|
|
int axis;
|
|
|
|
if(ge1->rtg < 0) {
|
|
d[1][X] = ge1->fx3 - ge1->prev->fx3;
|
|
d[1][Y] = ge1->fy3 - ge1->prev->fy3;
|
|
} else {
|
|
d[1][X] = ge1->fx3 - ge1->fpoints[X][ge1->rtg];
|
|
d[1][Y] = ge1->fy3 - ge1->fpoints[Y][ge1->rtg];
|
|
}
|
|
d[2][X] = ge2->fpoints[X][ge2->ftg] - ge2->prev->fx3;
|
|
d[2][Y] = ge2->fpoints[Y][ge2->ftg] - ge2->prev->fy3;
|
|
|
|
len1 = sqrt( d[1][X]*d[1][X] + d[1][Y]*d[1][Y] );
|
|
len2 = sqrt( d[2][X]*d[2][X] + d[2][Y]*d[2][Y] );
|
|
/* scale the 2nd segment to the length of 1
|
|
* and to make sure that the 1st segment is longer scale it to
|
|
* the length of 2 and extend to the same distance backwards
|
|
*/
|
|
scale1 = 2./len1;
|
|
scale2 = 1./len2;
|
|
|
|
for(axis=0; axis <2; axis++) {
|
|
d[0][axis] = -( d[1][axis] *= scale1 );
|
|
d[2][axis] *= scale2;
|
|
}
|
|
return fdotsegdist2(d, d[2]);
|
|
}
|
|
|
|
#if 0
|
|
/* find the area covered by the curve
|
|
* (limited by the projections to the X axis)
|
|
*/
|
|
|
|
static double
|
|
fcvarea(
|
|
GENTRY *ge
|
|
)
|
|
{
|
|
double Ly, My, Ny, Py, Qx, Rx, Sx;
|
|
double area;
|
|
|
|
/* y = Ly*t^3 + My*t^2 + Ny*t + Py */
|
|
Ly = -ge->prev->fy3 + 3*(ge->fy1 - ge->fy2) + ge->fy3;
|
|
My = 3*ge->prev->fy3 - 6*ge->fy1 + 3*ge->fy2;
|
|
Ny = 3*(-ge->prev->fy3 + ge->fy1);
|
|
Py = ge->prev->fy3;
|
|
|
|
/* dx/dt = Qx*t^2 + Rx*t + Sx */
|
|
Qx = 3*(-ge->prev->fx3 + 3*(ge->fx1 - ge->fx2) + ge->fx3);
|
|
Rx = 6*(ge->prev->fx3 - 2*ge->fx1 + ge->fx2);
|
|
Sx = 3*(-ge->prev->fx3 + ge->fx1);
|
|
|
|
/* area is integral[from 0 to 1]( y(t) * dx(t)/dt *dt) */
|
|
area = 1./6.*(Ly*Qx) + 1./5.*(Ly*Rx + My*Qx)
|
|
+ 1./4.*(Ly*Sx + My*Rx + Ny*Qx) + 1./3.*(My*Sx + Ny*Rx + Py*Qx)
|
|
+ 1./2.*(Ny*Sx + Py*Rx) + Py*Sx;
|
|
|
|
return area;
|
|
}
|
|
#endif
|
|
|
|
/* find the value of point on the curve at the given parameter t,
|
|
* along the given axis (0 - X, 1 - Y).
|
|
*/
|
|
|
|
static double
|
|
fcvval(
|
|
GENTRY *ge,
|
|
int axis,
|
|
double t
|
|
)
|
|
{
|
|
double t2, mt, mt2;
|
|
|
|
/* val = A*(1-t)^3 + 3*B*(1-t)^2*t + 3*C*(1-t)*t^2 + D*t^3 */
|
|
t2 = t*t;
|
|
mt = 1-t;
|
|
mt2 = mt*mt;
|
|
|
|
return ge->prev->fpoints[axis][2]*mt2*mt
|
|
+ 3*(ge->fpoints[axis][0]*mt2*t + ge->fpoints[axis][1]*mt*t2)
|
|
+ ge->fpoints[axis][2]*t*t2;
|
|
}
|
|
|
|
/*
|
|
* Find ndots equally spaced dots on a curve or line and fill
|
|
* their coordinates into the dots array
|
|
*/
|
|
|
|
static void
|
|
fsampledots(
|
|
GENTRY *ge,
|
|
double dots[][2], /* the dots to fill */
|
|
int ndots
|
|
)
|
|
{
|
|
int i, axis;
|
|
double t, nf, dx, d[2];
|
|
|
|
nf = ndots+1;
|
|
if(ge->type == GE_CURVE) {
|
|
for(i=0; i<ndots; i++) {
|
|
t = (i+1)/nf;
|
|
for(axis=0; axis<2; axis++)
|
|
dots[i][axis] = fcvval(ge, axis, t);
|
|
}
|
|
} else { /* line */
|
|
d[0] = ge->fx3 - ge->prev->fx3;
|
|
d[1] = ge->fy3 - ge->prev->fy3;
|
|
for(i=0; i<ndots; i++) {
|
|
t = (i+1)/nf;
|
|
for(axis=0; axis<2; axis++)
|
|
dots[i][axis] = ge->prev->fpoints[axis][2]
|
|
+ t*d[axis];
|
|
}
|
|
}
|
|
}
|
|
|
|
/*
|
|
* Allocate a structure gex_con
|
|
*/
|
|
|
|
static void
|
|
alloc_gex_con(
|
|
GENTRY *ge
|
|
)
|
|
{
|
|
ge->ext = (void*)calloc(1, sizeof(GEX_CON));
|
|
if(ge->ext == 0) {
|
|
fprintf (stderr, "****malloc failed %s line %d\n", __FILE__, __LINE__);
|
|
exit(255);
|
|
}
|
|
}
|
|
|
|
/*
|
|
* Normalize a gentry for fforceconcise() : find the points that
|
|
* can be used to calculate the tangents.
|
|
*/
|
|
|
|
static void
|
|
fnormalizege(
|
|
GENTRY *ge
|
|
)
|
|
{
|
|
int midsame, frontsame, rearsame;
|
|
|
|
if(ge->type == GE_LINE) {
|
|
ge->ftg = 2;
|
|
ge->rtg = -1;
|
|
} else { /* assume it's a curve */
|
|
midsame = (fabs(ge->fx1-ge->fx2)<FEPS && fabs(ge->fy1-ge->fy2)<FEPS);
|
|
frontsame = (fabs(ge->fx1-ge->prev->fx3)<FEPS && fabs(ge->fy1-ge->prev->fy3)<FEPS);
|
|
rearsame = (fabs(ge->fx3-ge->fx2)<FEPS && fabs(ge->fy3-ge->fy2)<FEPS);
|
|
|
|
if(midsame && (frontsame || rearsame) ) {
|
|
/* essentially a line */
|
|
ge->ftg = 2;
|
|
ge->rtg = -1;
|
|
} else {
|
|
if(frontsame) {
|
|
ge->ftg = 1;
|
|
} else {
|
|
ge->ftg = 0;
|
|
}
|
|
if(rearsame) {
|
|
ge->rtg = 0;
|
|
} else {
|
|
ge->rtg = 1;
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
/* various definition for the processing of outlines */
|
|
|
|
/* maximal average quadratic distance from the original curve
|
|
* (in dots) to consider the joined curve good
|
|
*/
|
|
#define CVEPS 1.5
|
|
#define CVEPS2 (CVEPS*CVEPS)
|
|
/* squared sinus of the maximal angle that we consider a smooth joint */
|
|
#define SMOOTHSIN2 0.25 /* 0.25==sin(30 degrees)^2 */
|
|
/* squared line length that we consider small */
|
|
#define SMALL_LINE2 (15.*15.)
|
|
/* how many times a curve must be bigger than a line to join, squared */
|
|
#define TIMES_LINE2 (3.*3.)
|
|
|
|
/*
|
|
* Normalize and analyse a gentry for fforceconcise() and fill out the gex_con
|
|
* structure
|
|
*/
|
|
|
|
static void
|
|
fanalyzege(
|
|
GENTRY *ge
|
|
)
|
|
{
|
|
int i, ix, iy;
|
|
double avsd2, dots[3][2 /*X,Y*/];
|
|
GEX_CON *gex;
|
|
|
|
gex = X_CON(ge);
|
|
memset(gex, 0, sizeof *gex);
|
|
|
|
gex->len2 = 0;
|
|
for(i=0; i<2; i++) {
|
|
avsd2 = gex->d[i] = ge->fpoints[i][2] - ge->prev->fpoints[i][2];
|
|
gex->len2 += avsd2*avsd2;
|
|
}
|
|
gex->sin2 = fjointsin2(ge, ge->frwd);
|
|
if(ge->type == GE_CURVE) {
|
|
ge->dir = fgetcvdir(ge);
|
|
for(i=0; i<2; i++) {
|
|
dots[0][i] = ge->prev->fpoints[i][2];
|
|
dots[1][i] = ge->fpoints[i][2];
|
|
dots[2][i] = fcvval(ge, i, 0.5);
|
|
}
|
|
avsd2 = fdotsegdist2(dots, dots[2]);
|
|
if(avsd2 <= CVEPS2) {
|
|
gex->flags |= GEXF_FLAT;
|
|
}
|
|
} else {
|
|
ge->dir = CVDIR_FEQUAL|CVDIR_REQUAL;
|
|
gex->flags |= GEXF_FLAT;
|
|
}
|
|
if(gex->flags & GEXF_FLAT) {
|
|
if( fabs(gex->d[X]) > FEPS && fabs(gex->d[Y]) < 5.
|
|
&& fabs(gex->d[Y] / gex->d[X]) < 0.2)
|
|
gex->flags |= GEXF_HOR;
|
|
else if( fabs(gex->d[Y]) > FEPS && fabs(gex->d[X]) < 5.
|
|
&& fabs(gex->d[X] / gex->d[Y]) < 0.2)
|
|
gex->flags |= GEXF_VERT;
|
|
}
|
|
ix = gex->isd[X] = fsign(gex->d[X]);
|
|
iy = gex->isd[Y] = fsign(gex->d[Y]);
|
|
if(ix <= 0) {
|
|
if(iy <= 0)
|
|
gex->flags |= GEXF_QDL;
|
|
if(iy >= 0)
|
|
gex->flags |= GEXF_QUL;
|
|
if(gex->flags & GEXF_HOR)
|
|
gex->flags |= GEXF_IDQ_L;
|
|
}
|
|
if(ix >= 0) {
|
|
if(iy <= 0)
|
|
gex->flags |= GEXF_QDR;
|
|
if(iy >= 0)
|
|
gex->flags |= GEXF_QUR;
|
|
if(gex->flags & GEXF_HOR)
|
|
gex->flags |= GEXF_IDQ_R;
|
|
}
|
|
if(gex->flags & GEXF_VERT) {
|
|
if(iy <= 0) {
|
|
gex->flags |= GEXF_IDQ_U;
|
|
} else { /* supposedly there is no 0-sized entry */
|
|
gex->flags |= GEXF_IDQ_D;
|
|
}
|
|
}
|
|
}
|
|
|
|
/*
|
|
* Analyse a joint between this and following gentry for fforceconcise()
|
|
* and fill out the corresponding parts of the gex_con structure
|
|
* Bothe entries must be analyzed first.
|
|
*/
|
|
|
|
static void
|
|
fanalyzejoint(
|
|
GENTRY *ge
|
|
)
|
|
{
|
|
GENTRY *nge = ge->frwd;
|
|
GENTRY tge;
|
|
GEX_CON *gex, *ngex;
|
|
double avsd2, dots[3][2 /*X,Y*/];
|
|
int i;
|
|
|
|
gex = X_CON(ge); ngex = X_CON(nge);
|
|
|
|
/* look if they can be joined honestly */
|
|
|
|
/* if any is flat, they should join smoothly */
|
|
if( (gex->flags & GEXF_FLAT || ngex->flags & GEXF_FLAT)
|
|
&& gex->sin2 > SMOOTHSIN2)
|
|
goto try_flatboth;
|
|
|
|
if(ge->type == GE_LINE) {
|
|
if(nge->type == GE_LINE) {
|
|
if(gex->len2 > SMALL_LINE2 || ngex->len2 > SMALL_LINE2)
|
|
goto try_flatboth;
|
|
} else {
|
|
if(gex->len2*TIMES_LINE2 > ngex->len2)
|
|
goto try_flatboth;
|
|
}
|
|
} else if(nge->type == GE_LINE) {
|
|
if(ngex->len2*TIMES_LINE2 > gex->len2)
|
|
goto try_flatboth;
|
|
}
|
|
|
|
/* if curve changes direction */
|
|
if( gex->isd[X]*ngex->isd[X]<0 || gex->isd[Y]*ngex->isd[Y]<0)
|
|
goto try_idealone;
|
|
|
|
/* if would create a zigzag */
|
|
if( ((ge->dir&CVDIR_FRONT)-CVDIR_FEQUAL) * ((nge->dir&CVDIR_REAR)-CVDIR_REQUAL) < 0 )
|
|
goto try_flatone;
|
|
|
|
if( fcrossraysge(ge, nge, NULL, NULL, NULL) )
|
|
gex->flags |= GEXF_JGOOD;
|
|
|
|
try_flatone:
|
|
/* look if they can be joined by flatting out one of the entries */
|
|
|
|
/* at this point we know that the general direction of the
|
|
* gentries is OK
|
|
*/
|
|
|
|
if( gex->flags & GEXF_FLAT ) {
|
|
tge = *ge;
|
|
tge.fx1 = tge.fx3;
|
|
tge.fy1 = tge.fy3;
|
|
fnormalizege(&tge);
|
|
if( fcrossraysge(&tge, nge, NULL, NULL, NULL) )
|
|
gex->flags |= GEXF_JFLAT|GEXF_JFLAT1;
|
|
}
|
|
if( ngex->flags & GEXF_FLAT ) {
|
|
tge = *nge;
|
|
tge.fx2 = ge->fx3;
|
|
tge.fy2 = ge->fy3;
|
|
fnormalizege(&tge);
|
|
if( fcrossraysge(ge, &tge, NULL, NULL, NULL) )
|
|
gex->flags |= GEXF_JFLAT|GEXF_JFLAT2;
|
|
}
|
|
|
|
try_idealone:
|
|
/* look if one of the entries can be brought to an idealized
|
|
* horizontal or vertical position and then joined
|
|
*/
|
|
if( gex->flags & GEXF_HOR && gex->isd[X]*ngex->isd[X]>=0 ) {
|
|
tge = *ge;
|
|
tge.fx1 = tge.fx3;
|
|
tge.fy1 = ge->prev->fy3; /* force horizontal */
|
|
fnormalizege(&tge);
|
|
if( fcrossraysge(&tge, nge, NULL, NULL, NULL) )
|
|
gex->flags |= GEXF_JID|GEXF_JID1;
|
|
} else if( gex->flags & GEXF_VERT && gex->isd[Y]*ngex->isd[Y]>=0 ) {
|
|
tge = *ge;
|
|
tge.fx1 = ge->prev->fx3; /* force vertical */
|
|
tge.fy1 = tge.fy3;
|
|
fnormalizege(&tge);
|
|
if( fcrossraysge(&tge, nge, NULL, NULL, NULL) )
|
|
gex->flags |= GEXF_JID|GEXF_JID1;
|
|
}
|
|
if( ngex->flags & GEXF_HOR && gex->isd[X]*ngex->isd[X]>=0 ) {
|
|
tge = *nge;
|
|
tge.fx2 = ge->fx3;
|
|
tge.fy2 = nge->fy3; /* force horizontal */
|
|
fnormalizege(&tge);
|
|
if( fcrossraysge(ge, &tge, NULL, NULL, NULL) )
|
|
gex->flags |= GEXF_JID|GEXF_JID2;
|
|
} else if( ngex->flags & GEXF_VERT && gex->isd[Y]*ngex->isd[Y]>=0 ) {
|
|
tge = *nge;
|
|
tge.fx2 = nge->fx3; /* force vertical */
|
|
tge.fy2 = ge->fy3;
|
|
fnormalizege(&tge);
|
|
if( fcrossraysge(ge, &tge, NULL, NULL, NULL) )
|
|
gex->flags |= GEXF_JID|GEXF_JID2;
|
|
}
|
|
|
|
try_flatboth:
|
|
/* look if we can change them to one line */
|
|
if(gex->flags & GEXF_FLAT && ngex->flags & GEXF_FLAT) {
|
|
for(i=0; i<2; i++) {
|
|
dots[0][i] = ge->prev->fpoints[i][2];
|
|
dots[1][i] = nge->fpoints[i][2];
|
|
dots[2][i] = ge->fpoints[i][2];
|
|
}
|
|
if( fdotsegdist2(dots, dots[2]) <= CVEPS2)
|
|
gex->flags |= GEXF_JLINE;
|
|
}
|
|
}
|
|
|
|
/*
|
|
* Force conciseness of one contour in the glyph,
|
|
* the contour is indicated by one entry from it.
|
|
*/
|
|
|
|
static void
|
|
fconcisecontour(
|
|
GLYPH *g,
|
|
GENTRY *startge
|
|
)
|
|
{
|
|
/* initial maximal number of dots to be used as reference */
|
|
#define MAXDOTS ((NREFDOTS+1)*12)
|
|
|
|
GENTRY *ge, *pge, *nge, *ige;
|
|
GEX_CON *gex, *pgex, *ngex, *nngex;
|
|
GENTRY tpge, tnge;
|
|
int quad, qq, i, j, ndots, maxdots;
|
|
int found[2];
|
|
int joinmask, pflag, nflag;
|
|
struct dot_dist *dots;
|
|
double avsd2, maxd2, eps2;
|
|
double apcv[4][2];
|
|
|
|
if(startge == 0) {
|
|
fprintf(stderr, "WARNING: assertion in %s line %d, please report it to the ttf2pt1 project\n",
|
|
__FILE__, __LINE__);
|
|
fprintf(stderr, "Strange contour in glyph %s\n", g->name);
|
|
dumppaths(g, NULL, NULL);
|
|
return;
|
|
}
|
|
|
|
if(startge->type != GE_CURVE && startge->type != GE_LINE)
|
|
return; /* probably a degenerate contour */
|
|
|
|
if(ISDBG(FCONCISE))
|
|
fprintf(stderr, "processing contour 0x%p of glyph %s\n", startge, g->name);
|
|
|
|
maxdots = MAXDOTS;
|
|
dots = (struct dot_dist *)malloc(sizeof(*dots)*maxdots);
|
|
if(dots == NULL) {
|
|
fprintf (stderr, "****malloc failed %s line %d\n", __FILE__, __LINE__);
|
|
exit(255);
|
|
}
|
|
|
|
ge = startge;
|
|
joinmask = GEXF_JGOOD;
|
|
while(1) {
|
|
restart:
|
|
gex = X_CON(ge);
|
|
if((gex->flags & GEXF_JMASK) > ((joinmask<<1)-1)) {
|
|
if(ISDBG(FCONCISE))
|
|
fprintf(stderr, "found higher flag (%x>%x) at 0x%p\n",
|
|
gex->flags & GEXF_JMASK, ((joinmask<<1)-1), ge);
|
|
joinmask <<= 1;
|
|
startge = ge; /* have to redo the pass */
|
|
continue;
|
|
}
|
|
if(( gex->flags & joinmask )==0)
|
|
goto next;
|
|
|
|
/* if we happen to be in the middle of a string of
|
|
* joinable entries, find its beginning
|
|
*/
|
|
if( gex->flags & (GEXF_JCVMASK^GEXF_JID) )
|
|
quad = gex->flags & X_CON_F(ge->frwd) & GEXF_QMASK;
|
|
else if( gex->flags & GEXF_JID2 )
|
|
quad = gex->flags & GEXF_QFROM_IDEAL(X_CON_F(ge->frwd)) & GEXF_QMASK;
|
|
else /* must be GEXF_JID1 */
|
|
quad = GEXF_QFROM_IDEAL(gex->flags) & X_CON_F(ge->frwd) & GEXF_QMASK;
|
|
|
|
pge = ge;
|
|
pgex = X_CON(pge->bkwd);
|
|
|
|
if(ISDBG(FCONCISE))
|
|
fprintf(stderr, "ge %p prev -> 0x%p ", ge, pge);
|
|
|
|
while(pgex->flags & GEXF_JCVMASK) {
|
|
if( !(pgex->flags & ((GEXF_JCVMASK^GEXF_JID)|GEXF_JID2)) )
|
|
qq = GEXF_QFROM_IDEAL(pgex->flags);
|
|
else
|
|
qq = pgex->flags & GEXF_QMASK;
|
|
|
|
if(ISDBG(FCONCISE))
|
|
fprintf(stderr, "(%x?%x)", quad, qq);
|
|
|
|
if( !(quad & qq) ) {
|
|
if( !(X_CON_F(pge) & (GEXF_JCVMASK^GEXF_JID))
|
|
&& pgex->flags & (GEXF_JCVMASK^GEXF_JID) ) {
|
|
/* the previos entry is definitely a better match */
|
|
if(pge == ge) {
|
|
if(ISDBG(FCONCISE))
|
|
fprintf(stderr, "\nprev is a better match at %p\n", pge);
|
|
startge = ge;
|
|
goto next;
|
|
} else
|
|
pge = pge->frwd;
|
|
}
|
|
break;
|
|
}
|
|
|
|
quad &= qq;
|
|
pge = pge->bkwd;
|
|
pgex = X_CON(pge->bkwd);
|
|
if(ISDBG(FCONCISE))
|
|
fprintf(stderr, "0x%p ", pge);
|
|
}
|
|
|
|
/* collect as many entries for joining as possible */
|
|
nge = ge->frwd;
|
|
ngex = X_CON(nge);
|
|
nngex = X_CON(nge->frwd);
|
|
|
|
if(ISDBG(FCONCISE))
|
|
fprintf(stderr, ": 0x%x\nnext -> 0x%p ", pge, nge);
|
|
|
|
while(ngex->flags & GEXF_JCVMASK) {
|
|
if( !(ngex->flags & ((GEXF_JCVMASK^GEXF_JID)|GEXF_JID1)) )
|
|
qq = GEXF_QFROM_IDEAL(nngex->flags);
|
|
else
|
|
qq = nngex->flags & GEXF_QMASK;
|
|
|
|
if(ISDBG(FCONCISE))
|
|
fprintf(stderr, "(%x?%x)", quad, qq);
|
|
if( !(quad & qq) ) {
|
|
if( !(X_CON_F(nge->bkwd) & (GEXF_JCVMASK^GEXF_JID))
|
|
&& ngex->flags & (GEXF_JCVMASK^GEXF_JID) ) {
|
|
/* the next-next entry is definitely a better match */
|
|
if(nge == ge->frwd) {
|
|
if(ISDBG(FCONCISE))
|
|
fprintf(stderr, "\nnext %x is a better match than %x at %p (jmask %x)\n",
|
|
ngex->flags & GEXF_JCVMASK, gex->flags & GEXF_JCVMASK, nge, joinmask);
|
|
goto next;
|
|
} else
|
|
nge = nge->bkwd;
|
|
}
|
|
break;
|
|
}
|
|
|
|
quad &= qq;
|
|
nge = nge->frwd;
|
|
ngex = nngex;
|
|
nngex = X_CON(nge->frwd);
|
|
if(ISDBG(FCONCISE))
|
|
fprintf(stderr, "0x%p ", nge);
|
|
}
|
|
|
|
if(ISDBG(FCONCISE))
|
|
fprintf(stderr, ": 0x%x\n", nge);
|
|
|
|
/* XXX add splitting of last entries if neccessary */
|
|
|
|
/* make sure that all the reference dots are valid */
|
|
for(ige = pge; ige != nge->frwd; ige = ige->frwd) {
|
|
nngex = X_CON(ige);
|
|
if( !(nngex->flags & GEXF_VDOTS) ) {
|
|
fsampledots(ige, nngex->dots, NREFDOTS);
|
|
nngex->flags |= GEXF_VDOTS;
|
|
}
|
|
}
|
|
|
|
/* do the actual joining */
|
|
while(1) {
|
|
pgex = X_CON(pge);
|
|
ngex = X_CON(nge->bkwd);
|
|
/* now the segments to be joined are pge...nge */
|
|
|
|
ndots = 0;
|
|
for(ige = pge; ige != nge->frwd; ige = ige->frwd) {
|
|
if(maxdots < ndots+(NREFDOTS+1)) {
|
|
maxdots += MAXDOTS;
|
|
dots = (struct dot_dist *)realloc((void *)dots, sizeof(*dots)*maxdots);
|
|
if(dots == NULL) {
|
|
fprintf (stderr, "****malloc failed %s line %d\n", __FILE__, __LINE__);
|
|
exit(255);
|
|
}
|
|
}
|
|
nngex = X_CON(ige);
|
|
for(i=0; i<NREFDOTS; i++) {
|
|
for(j=0; j<2; j++)
|
|
dots[ndots].p[j] = nngex->dots[i][j];
|
|
ndots++;
|
|
}
|
|
for(j=0; j<2; j++)
|
|
dots[ndots].p[j] = ige->fpoints[j][2];
|
|
ndots++;
|
|
}
|
|
ndots--; /* the last point is not interesting */
|
|
|
|
tpge = *pge;
|
|
pflag = pgex->flags;
|
|
if(pflag & (GEXF_JGOOD|GEXF_JFLAT2|GEXF_JID2)) {
|
|
/* nothing */
|
|
} else if(pflag & GEXF_JFLAT) {
|
|
tpge.fx1 = tpge.fx3;
|
|
tpge.fy1 = tpge.fy3;
|
|
} else if(pflag & GEXF_JID) {
|
|
if(pflag & GEXF_HOR)
|
|
tpge.fy1 = tpge.bkwd->fy3;
|
|
else
|
|
tpge.fx1 = tpge.bkwd->fx3;
|
|
}
|
|
|
|
tnge = *nge;
|
|
nflag = ngex->flags;
|
|
if(nflag & (GEXF_JGOOD|GEXF_JFLAT1|GEXF_JID)
|
|
&& !(nflag & GEXF_JID2)) {
|
|
/* nothing */
|
|
} else if(nflag & GEXF_JFLAT) {
|
|
tnge.fx2 = tnge.bkwd->fx3;
|
|
tnge.fy2 = tnge.bkwd->fy3;
|
|
} else if(nflag & GEXF_JID) {
|
|
if(X_CON_F(nge) & GEXF_HOR)
|
|
tnge.fy2 = tnge.fy3;
|
|
else
|
|
tnge.fx2 = tnge.fx3;
|
|
}
|
|
|
|
fnormalizege(&tpge);
|
|
fnormalizege(&tnge);
|
|
if( fcrossraysge(&tpge, &tnge, NULL, NULL, &apcv[1]) ) {
|
|
apcv[0][X] = tpge.bkwd->fx3;
|
|
apcv[0][Y] = tpge.bkwd->fy3;
|
|
/* apcv[1] and apcv[2] were filled by fcrossraysge() */
|
|
apcv[3][X] = tnge.fx3;
|
|
apcv[3][Y] = tnge.fy3;
|
|
|
|
/* calculate the precision depending on the smaller dimension of the curve */
|
|
maxd2 = apcv[3][X]-apcv[0][X];
|
|
maxd2 *= maxd2;
|
|
eps2 = apcv[3][Y]-apcv[0][Y];
|
|
eps2 *= eps2;
|
|
if(maxd2 < eps2)
|
|
eps2 = maxd2;
|
|
eps2 *= (CVEPS2*4.) / (400.*400.);
|
|
if(eps2 < CVEPS2)
|
|
eps2 = CVEPS2;
|
|
else if(eps2 > CVEPS2*4.)
|
|
eps2 = CVEPS2*4.;
|
|
|
|
fapproxcurve(apcv, dots, ndots);
|
|
|
|
avsd2 = fdotcurvdist2(apcv, dots, ndots, &maxd2);
|
|
if(ISDBG(FCONCISE))
|
|
fprintf(stderr, "avsd = %g, maxd = %g, ", sqrt(avsd2), sqrt(maxd2));
|
|
if(avsd2 <= eps2 && maxd2 <= eps2*2.) {
|
|
/* we've guessed a curve that is close enough */
|
|
ggoodcv++; ggoodcvdots += ndots;
|
|
|
|
if(ISDBG(FCONCISE)) {
|
|
fprintf(stderr, "in %s joined %p-%p to ", g->name, pge, nge);
|
|
for(i=0; i<4; i++) {
|
|
fprintf(stderr, " (%g, %g)", apcv[i][X], apcv[i][Y]);
|
|
}
|
|
fprintf(stderr, " from\n");
|
|
dumppaths(g, pge, nge);
|
|
}
|
|
for(i=0; i<3; i++) {
|
|
pge->fxn[i] = apcv[i+1][X];
|
|
pge->fyn[i] = apcv[i+1][Y];
|
|
}
|
|
pge->type = GE_CURVE;
|
|
ge = pge;
|
|
for(ige = pge->frwd; ; ige = pge->frwd) {
|
|
if(ige == pge) {
|
|
fprintf(stderr, "WARNING: assertion in %s line %d, please report it to the ttf2pt1 project\n",
|
|
__FILE__, __LINE__);
|
|
free(dots);
|
|
return;
|
|
}
|
|
if(startge == ige)
|
|
startge = pge;
|
|
free(ige->ext);
|
|
freethisge(ige);
|
|
if(ige == nge)
|
|
break;
|
|
}
|
|
fnormalizege(ge);
|
|
if(ISDBG(FCONCISE)) {
|
|
fprintf(stderr, "normalized ");
|
|
for(i=0; i<3; i++) {
|
|
fprintf(stderr, " (%g, %g)", ge->fpoints[X][i], ge->fpoints[Y][i]);
|
|
}
|
|
fprintf(stderr, "\n");
|
|
}
|
|
fanalyzege(ge);
|
|
fanalyzejoint(ge);
|
|
fanalyzege(ge->bkwd);
|
|
fanalyzejoint(ge->bkwd);
|
|
|
|
/* the results of this join will have to be reconsidered */
|
|
startge = ge = ge->frwd;
|
|
goto restart;
|
|
} else {
|
|
gbadcv++; gbadcvdots += ndots;
|
|
}
|
|
}
|
|
|
|
/* if we're down to 2 entries then the join has failed */
|
|
if(pge->frwd == nge) {
|
|
pgex->flags &= ~joinmask;
|
|
if(ISDBG(FCONCISE))
|
|
fprintf(stderr, "no match\n");
|
|
goto next;
|
|
}
|
|
|
|
/* reduce the number of entries by dropping one at some end,
|
|
* should never drop the original ge from the range
|
|
*/
|
|
|
|
if(nge->bkwd == ge
|
|
|| pge != ge && (pgex->flags & GEXF_JCVMASK) <= (ngex->flags & GEXF_JCVMASK) ) {
|
|
pge = pge->frwd;
|
|
} else {
|
|
nge = nge->bkwd;
|
|
}
|
|
if(ISDBG(FCONCISE))
|
|
fprintf(stderr, "next try: %p to %p\n", pge, nge);
|
|
}
|
|
|
|
next:
|
|
ge = ge->frwd;
|
|
if(ge == startge) {
|
|
joinmask = (joinmask >> 1) & GEXF_JCVMASK;
|
|
if(joinmask == 0)
|
|
break;
|
|
}
|
|
}
|
|
|
|
/* join flat segments into lines */
|
|
/* here ge==startge */
|
|
while(1) {
|
|
gex = X_CON(ge);
|
|
if( !(gex->flags & GEXF_JLINE) )
|
|
goto next2;
|
|
|
|
ndots = 0;
|
|
dots[ndots].p[X] = ge->fx3;
|
|
dots[ndots].p[Y] = ge->fy3;
|
|
ndots++;
|
|
|
|
pge = ge->bkwd;
|
|
nge = ge->frwd;
|
|
|
|
if(ISDBG(FCONCISE))
|
|
fprintf(stderr, "joining LINE from %p-%p\n", ge, nge);
|
|
|
|
while(pge!=nge) {
|
|
pgex = X_CON(pge);
|
|
ngex = X_CON(nge);
|
|
if(ISDBG(FCONCISE))
|
|
fprintf(stderr, "(p=%p/%x n=0x%x/%x) ", pge, pgex->flags & GEXF_JLINE,
|
|
nge, ngex->flags & GEXF_JLINE);
|
|
if( !((pgex->flags | ngex->flags) & GEXF_JLINE) ) {
|
|
if(ISDBG(FCONCISE))
|
|
fprintf(stderr, "(end p=%p n=%p) ", pge, nge);
|
|
break;
|
|
}
|
|
|
|
if(maxdots < ndots+2) {
|
|
maxdots += MAXDOTS;
|
|
dots = (struct dot_dist *)realloc((void *)dots, sizeof(*dots)*maxdots);
|
|
if(dots == NULL) {
|
|
fprintf (stderr, "****malloc failed %s line %d\n", __FILE__, __LINE__);
|
|
exit(255);
|
|
}
|
|
}
|
|
if( pgex->flags & GEXF_JLINE ) {
|
|
for(i=0; i<2; i++) {
|
|
apcv[0][i] = pge->bkwd->fpoints[i][2];
|
|
apcv[1][i] = nge->fpoints[i][2];
|
|
dots[ndots].p[i] = pge->fpoints[i][2];
|
|
}
|
|
ndots++;
|
|
for(i=0; i<ndots; i++) {
|
|
avsd2 = fdotsegdist2(apcv, dots[i].p);
|
|
if(avsd2 > CVEPS2)
|
|
break;
|
|
}
|
|
if(i<ndots) { /* failed to join */
|
|
if(ISDBG(FCONCISE))
|
|
fprintf(stderr, "failed to join prev %p ", pge);
|
|
ndots--;
|
|
pgex->flags &= ~GEXF_JLINE;
|
|
} else {
|
|
pge = pge->bkwd;
|
|
if(pge == nge) {
|
|
if(ISDBG(FCONCISE))
|
|
fprintf(stderr, "intersected at prev %p ", pge);
|
|
break; /* oops, tried to self-intersect */
|
|
}
|
|
}
|
|
} else if(ISDBG(FCONCISE))
|
|
fprintf(stderr, "(p=%p) ", pge);
|
|
|
|
if( ngex->flags & GEXF_JLINE ) {
|
|
for(i=0; i<2; i++) {
|
|
apcv[0][i] = pge->fpoints[i][2]; /* pge points before the 1st segment */
|
|
apcv[1][i] = nge->frwd->fpoints[i][2];
|
|
dots[ndots].p[i] = nge->fpoints[i][2];
|
|
}
|
|
ndots++;
|
|
for(i=0; i<ndots; i++) {
|
|
avsd2 = fdotsegdist2(apcv, dots[i].p);
|
|
if(avsd2 > CVEPS2)
|
|
break;
|
|
}
|
|
if(i<ndots) { /* failed to join */
|
|
if(ISDBG(FCONCISE))
|
|
fprintf(stderr, "failed to join next %p ", nge->frwd);
|
|
ndots--;
|
|
ngex->flags &= ~GEXF_JLINE;
|
|
} else {
|
|
nge = nge->frwd;
|
|
}
|
|
} else if(ISDBG(FCONCISE))
|
|
fprintf(stderr, "(n=%p) ", nge);
|
|
}
|
|
|
|
pge = pge->frwd; /* now the limits are pge...nge inclusive */
|
|
if(pge == nge) /* a deeply perversive contour */
|
|
break;
|
|
|
|
if(ISDBG(FCONCISE)) {
|
|
fprintf(stderr, "\nin %s joined LINE %p-%p from\n", g->name, pge, nge);
|
|
dumppaths(g, pge, nge);
|
|
}
|
|
pge->type = GE_LINE;
|
|
for(i=0; i<2; i++) {
|
|
pge->fpoints[i][2] = nge->fpoints[i][2];
|
|
}
|
|
fnormalizege(pge);
|
|
X_CON_F(pge) &= ~GEXF_JLINE;
|
|
|
|
ge = pge;
|
|
for(ige = pge->frwd; ; ige = pge->frwd) {
|
|
if(ige == pge) {
|
|
fprintf(stderr, "WARNING: assertion in %s line %d, please report it to the ttf2pt1 project\n",
|
|
__FILE__, __LINE__);
|
|
free(dots);
|
|
return;
|
|
}
|
|
if(startge == ige)
|
|
startge = pge;
|
|
free(ige->ext);
|
|
freethisge(ige);
|
|
if(ige == nge)
|
|
break;
|
|
}
|
|
next2:
|
|
ge = ge->frwd;
|
|
if(ge == startge)
|
|
break;
|
|
}
|
|
|
|
free(dots);
|
|
}
|
|
|
|
/* force conciseness: substitute 2 or more curves going in the
|
|
** same quadrant with one curve
|
|
** in floating point
|
|
*/
|
|
|
|
void
|
|
fforceconcise(
|
|
GLYPH * g
|
|
)
|
|
{
|
|
|
|
GENTRY *ge, *nge, *endge, *xge;
|
|
|
|
assertisfloat(g, "enforcing conciseness");
|
|
|
|
fdelsmall(g, 0.05);
|
|
assertpath(g->entries, __FILE__, __LINE__, g->name);
|
|
|
|
if(ISDBG(FCONCISE))
|
|
dumppaths(g, NULL, NULL);
|
|
|
|
/* collect more information about each gentry and their joints */
|
|
for (ge = g->entries; ge != 0; ge = ge->next)
|
|
if (ge->type == GE_CURVE || ge->type == GE_LINE)
|
|
fnormalizege(ge);
|
|
|
|
for (ge = g->entries; ge != 0; ge = ge->next)
|
|
if (ge->type == GE_CURVE || ge->type == GE_LINE) {
|
|
alloc_gex_con(ge);
|
|
fanalyzege(ge);
|
|
}
|
|
|
|
/* see what we can do about joining */
|
|
for (ge = g->entries; ge != 0; ge = ge->next)
|
|
if (ge->type == GE_CURVE || ge->type == GE_LINE)
|
|
fanalyzejoint(ge);
|
|
|
|
/* now do the joining */
|
|
for (ge = g->entries; ge != 0; ge = ge->next)
|
|
if(ge->type == GE_MOVE)
|
|
fconcisecontour(g, ge->next);
|
|
|
|
for (ge = g->entries; ge != 0; ge = ge->next)
|
|
if (ge->type == GE_CURVE || ge->type == GE_LINE)
|
|
free(ge->ext);
|
|
}
|
|
|
|
void
|
|
print_glyph(
|
|
int glyphno
|
|
)
|
|
{
|
|
GLYPH *g;
|
|
GENTRY *ge;
|
|
int x = 0, y = 0;
|
|
int i;
|
|
int grp, lastgrp= -1;
|
|
|
|
if(ISDBG(FCONCISE) && glyphno == 0) {
|
|
fprintf(stderr, "Guessed curves: bad %d/%d good %d/%d\n",
|
|
gbadcv, gbadcvdots, ggoodcv, ggoodcvdots);
|
|
}
|
|
|
|
g = &glyph_list[glyphno];
|
|
|
|
fprintf(pfa_file, "/%s { \n", g->name);
|
|
|
|
/* consider widths >MAXLEGALWIDTH as bugs */
|
|
if( g->scaledwidth <= MAXLEGALWIDTH ) {
|
|
fprintf(pfa_file, "0 %d hsbw\n", g->scaledwidth);
|
|
} else {
|
|
fprintf(pfa_file, "0 1000 hsbw\n");
|
|
WARNING_2 fprintf(stderr, "glyph %s: width %d seems to be buggy, set to 1000\n",
|
|
g->name, g->scaledwidth);
|
|
}
|
|
|
|
#if 0
|
|
fprintf(pfa_file, "%% contours: ");
|
|
for (i = 0; i < g->ncontours; i++)
|
|
fprintf(pfa_file, "%s(%d,%d) ", (g->contours[i].direction == DIR_OUTER ? "out" : "in"),
|
|
g->contours[i].xofmin, g->contours[i].ymin);
|
|
fprintf(pfa_file, "\n");
|
|
|
|
if (g->rymin < 5000)
|
|
fprintf(pfa_file, "%d lower%s\n", g->rymin, (g->flatymin ? "flat" : "curve"));
|
|
if (g->rymax > -5000)
|
|
fprintf(pfa_file, "%d upper%s\n", g->rymax, (g->flatymax ? "flat" : "curve"));
|
|
#endif
|
|
|
|
if (g->hstems)
|
|
for (i = 0; i < g->nhs; i += 2) {
|
|
if (g->hstems[i].flags & ST_3) {
|
|
fprintf(pfa_file, "%d %d %d %d %d %d hstem3\n",
|
|
g->hstems[i].value,
|
|
g->hstems[i + 1].value - g->hstems[i].value,
|
|
g->hstems[i + 2].value,
|
|
g->hstems[i + 3].value - g->hstems[i + 2].value,
|
|
g->hstems[i + 4].value,
|
|
g->hstems[i + 5].value - g->hstems[i + 4].value
|
|
);
|
|
i += 4;
|
|
} else {
|
|
fprintf(pfa_file, "%d %d hstem\n", g->hstems[i].value,
|
|
g->hstems[i + 1].value - g->hstems[i].value);
|
|
}
|
|
}
|
|
|
|
if (g->vstems)
|
|
for (i = 0; i < g->nvs; i += 2) {
|
|
if (g->vstems[i].flags & ST_3) {
|
|
fprintf(pfa_file, "%d %d %d %d %d %d vstem3\n",
|
|
g->vstems[i].value,
|
|
g->vstems[i + 1].value - g->vstems[i].value,
|
|
g->vstems[i + 2].value,
|
|
g->vstems[i + 3].value - g->vstems[i + 2].value,
|
|
g->vstems[i + 4].value,
|
|
g->vstems[i + 5].value - g->vstems[i + 4].value
|
|
);
|
|
i += 4;
|
|
} else {
|
|
fprintf(pfa_file, "%d %d vstem\n", g->vstems[i].value,
|
|
g->vstems[i + 1].value - g->vstems[i].value);
|
|
}
|
|
}
|
|
|
|
for (ge = g->entries; ge != 0; ge = ge->next) {
|
|
if(g->nsg>0) {
|
|
grp=ge->stemid;
|
|
if(grp >= 0 && grp != lastgrp) {
|
|
fprintf(pfa_file, "%d 4 callsubr\n", grp+g->firstsubr);
|
|
lastgrp=grp;
|
|
}
|
|
}
|
|
|
|
switch (ge->type) {
|
|
case GE_MOVE:
|
|
if (absolute)
|
|
fprintf(pfa_file, "%d %d amoveto\n", ge->ix3, ge->iy3);
|
|
else
|
|
rmoveto(ge->ix3 - x, ge->iy3 - y);
|
|
if (0)
|
|
fprintf(stderr, "Glyph %s: print moveto(%d, %d)\n",
|
|
g->name, ge->ix3, ge->iy3);
|
|
x = ge->ix3;
|
|
y = ge->iy3;
|
|
break;
|
|
case GE_LINE:
|
|
if (absolute)
|
|
fprintf(pfa_file, "%d %d alineto\n", ge->ix3, ge->iy3);
|
|
else
|
|
rlineto(ge->ix3 - x, ge->iy3 - y);
|
|
x = ge->ix3;
|
|
y = ge->iy3;
|
|
break;
|
|
case GE_CURVE:
|
|
if (absolute)
|
|
fprintf(pfa_file, "%d %d %d %d %d %d arcurveto\n",
|
|
ge->ix1, ge->iy1, ge->ix2, ge->iy2, ge->ix3, ge->iy3);
|
|
else
|
|
rrcurveto(ge->ix1 - x, ge->iy1 - y,
|
|
ge->ix2 - ge->ix1, ge->iy2 - ge->iy1,
|
|
ge->ix3 - ge->ix2, ge->iy3 - ge->iy2);
|
|
x = ge->ix3;
|
|
y = ge->iy3;
|
|
break;
|
|
case GE_PATH:
|
|
closepath();
|
|
break;
|
|
default:
|
|
WARNING_1 fprintf(stderr, "**** Glyph %s: unknown entry type '%c'\n",
|
|
g->name, ge->type);
|
|
break;
|
|
}
|
|
}
|
|
|
|
fprintf(pfa_file, "endchar } ND\n");
|
|
}
|
|
|
|
/* print the subroutines for this glyph, returns the number of them */
|
|
int
|
|
print_glyph_subs(
|
|
int glyphno,
|
|
int startid /* start numbering subroutines from this id */
|
|
)
|
|
{
|
|
GLYPH *g;
|
|
int i, grp;
|
|
|
|
g = &glyph_list[glyphno];
|
|
|
|
if(!hints || !subhints || g->nsg<1)
|
|
return 0;
|
|
|
|
g->firstsubr=startid;
|
|
|
|
#if 0
|
|
fprintf(pfa_file, "%% %s %d\n", g->name, g->nsg);
|
|
#endif
|
|
for(grp=0; grp<g->nsg; grp++) {
|
|
fprintf(pfa_file, "dup %d {\n", startid++);
|
|
for(i= (grp==0)? 0 : g->nsbs[grp-1]; i<g->nsbs[grp]; i++)
|
|
fprintf(pfa_file, "\t%d %d %cstem\n", g->sbstems[i].low,
|
|
g->sbstems[i].high-g->sbstems[i].low,
|
|
g->sbstems[i].isvert ? 'v' : 'h');
|
|
fprintf(pfa_file, "\treturn\n\t} NP\n");
|
|
}
|
|
|
|
return g->nsg;
|
|
}
|
|
|
|
void
|
|
print_glyph_metrics(
|
|
FILE *afm_file,
|
|
int code,
|
|
int glyphno
|
|
)
|
|
{
|
|
GLYPH *g;
|
|
|
|
g = &glyph_list[glyphno];
|
|
|
|
if(transform)
|
|
fprintf(afm_file, "C %d ; WX %d ; N %s ; B %d %d %d %d ;\n",
|
|
code, g->scaledwidth, g->name,
|
|
iscale(g->xMin), iscale(g->yMin), iscale(g->xMax), iscale(g->yMax));
|
|
else
|
|
fprintf(afm_file, "C %d ; WX %d ; N %s ; B %d %d %d %d ;\n",
|
|
code, g->scaledwidth, g->name,
|
|
g->xMin, g->yMin, g->xMax, g->yMax);
|
|
|
|
}
|
|
|
|
void
|
|
print_glyph_metrics_ufm(
|
|
FILE *ufm_file,
|
|
int code,
|
|
int glyphno
|
|
)
|
|
{
|
|
GLYPH *g;
|
|
|
|
g = &glyph_list[glyphno];
|
|
|
|
fprintf(ufm_file, "U %d ; WX %d ; N %s ; G %d ;\n",
|
|
code, g->scaledwidth, g->name, glyphno);
|
|
}
|
|
/*
|
|
SB:
|
|
An important note about the BlueValues.
|
|
|
|
The Adobe documentation says that the maximal width of a Blue zone
|
|
is connected to the value of BlueScale, which is by default 0.039625.
|
|
The BlueScale value defines, at which point size the overshoot
|
|
suppression be disabled.
|
|
|
|
The formula for it that is given in the manual is:
|
|
|
|
BlueScale=point_size/240, for a 300dpi device
|
|
|
|
that makes us wonder what is this 240 standing for. Incidentally
|
|
240=72*1000/300, where 72 is the relation between inches and points,
|
|
1000 is the size of the glyph matrix, and 300dpi is the resolution of
|
|
the output device. Knowing that we can recalculate the formula for
|
|
the font size in pixels rather than points:
|
|
|
|
BlueScale=pixel_size/1000
|
|
|
|
That looks a lot simpler than the original formula, does not it ?
|
|
And the limitation about the maximal width of zone also looks
|
|
a lot simpler after the transformation:
|
|
|
|
max_width < 1000/pixel_size
|
|
|
|
that ensures that even at the maximal pixel size when the overshoot
|
|
suppression is disabled the zone width will be less than one pixel.
|
|
This is important, failure to comply to this limit will result in
|
|
really ugly fonts (been there, done that). But knowing the formula
|
|
for the pixel width, we see that in fact we can use the maximal width
|
|
of 24, not 23 as specified in the manual.
|
|
|
|
*/
|
|
|
|
#define MAXBLUEWIDTH (24)
|
|
|
|
/*
|
|
* Find the indexes of the most frequent values
|
|
* in the hystogram, sort them in ascending order, and save which one
|
|
* was the best one (if asked).
|
|
* Returns the number of values found (may be less than maximal because
|
|
* we ignore the zero values)
|
|
*/
|
|
|
|
#define MAXHYST (2000) /* size of the hystogram */
|
|
#define HYSTBASE 500
|
|
|
|
static int
|
|
besthyst(
|
|
int *hyst, /* the hystogram */
|
|
int base, /* the base point of the hystogram */
|
|
int *best, /* the array for indexes of best values */
|
|
int nbest, /* its allocated size */
|
|
int width, /* minimal difference between indexes */
|
|
int *bestindp /* returned top point */
|
|
)
|
|
{
|
|
unsigned char hused[MAXHYST / 8 + 1];
|
|
int i, max, j, w, last = 0;
|
|
int nf = 0;
|
|
|
|
width--;
|
|
|
|
memset(hused, 0 , sizeof hused);
|
|
|
|
max = 1;
|
|
for (i = 0; i < nbest && max != 0; i++) {
|
|
best[i] = 0;
|
|
max = 0;
|
|
for (j = 1; j < MAXHYST - 1; j++) {
|
|
w = hyst[j];
|
|
|
|
if (w > max && (hused[j>>3] & (1 << (j & 0x07))) == 0) {
|
|
best[i] = j;
|
|
max = w;
|
|
}
|
|
}
|
|
if (max != 0) {
|
|
if (max < last/2) {
|
|
/* do not pick the too low values */
|
|
break;
|
|
}
|
|
for (j = best[i] - width; j <= best[i] + width; j++) {
|
|
if (j >= 0 && j < MAXHYST)
|
|
hused[j >> 3] |= (1 << (j & 0x07));
|
|
}
|
|
last = max;
|
|
best[i] -= base;
|
|
nf = i + 1;
|
|
}
|
|
}
|
|
|
|
if (bestindp)
|
|
*bestindp = best[0];
|
|
|
|
/* sort the indexes in ascending order */
|
|
for (i = 0; i < nf; i++) {
|
|
for (j = i + 1; j < nf; j++)
|
|
if (best[j] < best[i]) {
|
|
w = best[i];
|
|
best[i] = best[j];
|
|
best[j] = w;
|
|
}
|
|
}
|
|
|
|
return nf;
|
|
}
|
|
|
|
/*
|
|
* Find the next best Blue zone in the hystogram.
|
|
* Return the weight of the found zone.
|
|
*/
|
|
|
|
static int
|
|
bestblue(
|
|
short *zhyst, /* the zones hystogram */
|
|
short *physt, /* the points hystogram */
|
|
short *ozhyst, /* the other zones hystogram */
|
|
int *bluetab /* where to put the found zone */
|
|
)
|
|
{
|
|
int i, j, w, max, ind, first, last;
|
|
|
|
/* find the highest point in the zones hystogram */
|
|
/* if we have a plateau, take its center */
|
|
/* if we have multiple peaks, take the first one */
|
|
|
|
max = -1;
|
|
first = last = -10;
|
|
for (i = 0; i <= MAXHYST - MAXBLUEWIDTH; i++) {
|
|
w = zhyst[i];
|
|
if (w > max) {
|
|
first = last = i;
|
|
max = w;
|
|
} else if (w == max) {
|
|
if (last == i - 1)
|
|
last = i;
|
|
}
|
|
}
|
|
ind = (first + last) / 2;
|
|
|
|
if (max == 0) /* no zones left */
|
|
return 0;
|
|
|
|
/* now we reuse `first' and `last' as inclusive borders of the zone */
|
|
first = ind;
|
|
last = ind + (MAXBLUEWIDTH - 1);
|
|
|
|
/* our maximal width is far too big, so we try to make it narrower */
|
|
w = max;
|
|
j = (w & 1); /* a pseudo-random bit */
|
|
while (1) {
|
|
while (physt[first] == 0)
|
|
first++;
|
|
while (physt[last] == 0)
|
|
last--;
|
|
if (last - first < (MAXBLUEWIDTH * 2 / 3) || (max - w) * 10 > max)
|
|
break;
|
|
|
|
if (physt[first] < physt[last]
|
|
|| physt[first] == physt[last] && j) {
|
|
if (physt[first] * 20 > w) /* if weight is >5%,
|
|
* stop */
|
|
break;
|
|
w -= physt[first];
|
|
first++;
|
|
j = 0;
|
|
} else {
|
|
if (physt[last] * 20 > w) /* if weight is >5%,
|
|
* stop */
|
|
break;
|
|
w -= physt[last];
|
|
last--;
|
|
j = 1;
|
|
}
|
|
}
|
|
|
|
/* save our zone */
|
|
bluetab[0] = first - HYSTBASE;
|
|
bluetab[1] = last - HYSTBASE;
|
|
|
|
/* invalidate all the zones overlapping with this one */
|
|
/* the constant of 2 is determined by the default value of BlueFuzz */
|
|
for (i = first - (MAXBLUEWIDTH - 1) - 2; i <= last + 2; i++)
|
|
if (i >= 0 && i < MAXHYST) {
|
|
zhyst[i] = 0;
|
|
ozhyst[i] = 0;
|
|
}
|
|
return w;
|
|
}
|
|
|
|
/*
|
|
* Try to find the Blue Values, bounding box and italic angle
|
|
*/
|
|
|
|
void
|
|
findblues(void)
|
|
{
|
|
/* hystograms for upper and lower zones */
|
|
short hystl[MAXHYST];
|
|
short hystu[MAXHYST];
|
|
short zuhyst[MAXHYST];
|
|
short zlhyst[MAXHYST];
|
|
int nchars;
|
|
int i, j, k, w, max;
|
|
GENTRY *ge;
|
|
GLYPH *g;
|
|
double ang;
|
|
|
|
/* find the lowest and highest points of glyphs */
|
|
/* and by the way build the values for FontBBox */
|
|
/* and build the hystogram for the ItalicAngle */
|
|
|
|
/* re-use hystl for the hystogram of italic angle */
|
|
|
|
bbox[0] = bbox[1] = 5000;
|
|
bbox[2] = bbox[3] = -5000;
|
|
|
|
for (i = 0; i < MAXHYST; i++)
|
|
hystl[i] = 0;
|
|
|
|
nchars = 0;
|
|
|
|
for (i = 0, g = glyph_list; i < numglyphs; i++, g++) {
|
|
if (g->flags & GF_USED) {
|
|
nchars++;
|
|
|
|
g->rymin = 5000;
|
|
g->rymax = -5000;
|
|
for (ge = g->entries; ge != 0; ge = ge->next) {
|
|
if (ge->type == GE_LINE) {
|
|
|
|
j = ge->iy3 - ge->prev->iy3;
|
|
k = ge->ix3 - ge->prev->ix3;
|
|
if (j > 0)
|
|
ang = atan2(-k, j) * 180.0 / M_PI;
|
|
else
|
|
ang = atan2(k, -j) * 180.0 / M_PI;
|
|
|
|
k /= 100;
|
|
j /= 100;
|
|
if (ang > -45.0 && ang < 45.0) {
|
|
/*
|
|
* be careful to not overflow
|
|
* the counter
|
|
*/
|
|
hystl[HYSTBASE + (int) (ang * 10.0)] += (k * k + j * j) / 4;
|
|
}
|
|
if (ge->iy3 == ge->prev->iy3) {
|
|
if (ge->iy3 <= g->rymin) {
|
|
g->rymin = ge->iy3;
|
|
g->flatymin = 1;
|
|
}
|
|
if (ge->iy3 >= g->rymax) {
|
|
g->rymax = ge->iy3;
|
|
g->flatymax = 1;
|
|
}
|
|
} else {
|
|
if (ge->iy3 < g->rymin) {
|
|
g->rymin = ge->iy3;
|
|
g->flatymin = 0;
|
|
}
|
|
if (ge->iy3 > g->rymax) {
|
|
g->rymax = ge->iy3;
|
|
g->flatymax = 0;
|
|
}
|
|
}
|
|
} else if (ge->type == GE_CURVE) {
|
|
if (ge->iy3 < g->rymin) {
|
|
g->rymin = ge->iy3;
|
|
g->flatymin = 0;
|
|
}
|
|
if (ge->iy3 > g->rymax) {
|
|
g->rymax = ge->iy3;
|
|
g->flatymax = 0;
|
|
}
|
|
}
|
|
if (ge->type == GE_LINE || ge->type == GE_CURVE) {
|
|
if (ge->ix3 < bbox[0])
|
|
bbox[0] = ge->ix3;
|
|
if (ge->ix3 > bbox[2])
|
|
bbox[2] = ge->ix3;
|
|
if (ge->iy3 < bbox[1])
|
|
bbox[1] = ge->iy3;
|
|
if (ge->iy3 > bbox[3])
|
|
bbox[3] = ge->iy3;
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
/* get the most popular angle */
|
|
max = 0;
|
|
w = 0;
|
|
for (i = 0; i < MAXHYST; i++) {
|
|
if (hystl[i] > w) {
|
|
w = hystl[i];
|
|
max = i;
|
|
}
|
|
}
|
|
ang = (double) (max - HYSTBASE) / 10.0;
|
|
WARNING_2 fprintf(stderr, "Guessed italic angle: %f\n", ang);
|
|
if (italic_angle == 0.0)
|
|
italic_angle = ang;
|
|
|
|
/* build the hystogram of the lower points */
|
|
for (i = 0; i < MAXHYST; i++)
|
|
hystl[i] = 0;
|
|
|
|
for (i = 0, g = glyph_list; i < numglyphs; i++, g++) {
|
|
if ((g->flags & GF_USED)
|
|
&& g->rymin + HYSTBASE >= 0 && g->rymin < MAXHYST - HYSTBASE) {
|
|
hystl[g->rymin + HYSTBASE]++;
|
|
}
|
|
}
|
|
|
|
/* build the hystogram of the upper points */
|
|
for (i = 0; i < MAXHYST; i++)
|
|
hystu[i] = 0;
|
|
|
|
for (i = 0, g = glyph_list; i < numglyphs; i++, g++) {
|
|
if ((g->flags & GF_USED)
|
|
&& g->rymax + HYSTBASE >= 0 && g->rymax < MAXHYST - HYSTBASE) {
|
|
hystu[g->rymax + HYSTBASE]++;
|
|
}
|
|
}
|
|
|
|
/* build the hystogram of all the possible lower zones with max width */
|
|
for (i = 0; i < MAXHYST; i++)
|
|
zlhyst[i] = 0;
|
|
|
|
for (i = 0; i <= MAXHYST - MAXBLUEWIDTH; i++) {
|
|
for (j = 0; j < MAXBLUEWIDTH; j++)
|
|
zlhyst[i] += hystl[i + j];
|
|
}
|
|
|
|
/* build the hystogram of all the possible upper zones with max width */
|
|
for (i = 0; i < MAXHYST; i++)
|
|
zuhyst[i] = 0;
|
|
|
|
for (i = 0; i <= MAXHYST - MAXBLUEWIDTH; i++) {
|
|
for (j = 0; j < MAXBLUEWIDTH; j++)
|
|
zuhyst[i] += hystu[i + j];
|
|
}
|
|
|
|
/* find the baseline */
|
|
w = bestblue(zlhyst, hystl, zuhyst, &bluevalues[0]);
|
|
if (0)
|
|
fprintf(stderr, "BaselineBlue zone %d%% %d...%d\n", w * 100 / nchars,
|
|
bluevalues[0], bluevalues[1]);
|
|
|
|
if (w == 0) /* no baseline, something weird */
|
|
return;
|
|
|
|
/* find the upper zones */
|
|
for (nblues = 2; nblues < 14; nblues += 2) {
|
|
w = bestblue(zuhyst, hystu, zlhyst, &bluevalues[nblues]);
|
|
|
|
if (0)
|
|
fprintf(stderr, "Blue zone %d%% %d...%d\n", w * 100 / nchars,
|
|
bluevalues[nblues], bluevalues[nblues+1]);
|
|
|
|
if (w * 20 < nchars)
|
|
break; /* don't save this zone */
|
|
}
|
|
|
|
/* find the lower zones */
|
|
for (notherb = 0; notherb < 10; notherb += 2) {
|
|
w = bestblue(zlhyst, hystl, zuhyst, &otherblues[notherb]);
|
|
|
|
if (0)
|
|
fprintf(stderr, "OtherBlue zone %d%% %d...%d\n", w * 100 / nchars,
|
|
otherblues[notherb], otherblues[notherb+1]);
|
|
|
|
|
|
if (w * 20 < nchars)
|
|
break; /* don't save this zone */
|
|
}
|
|
|
|
}
|
|
|
|
/*
|
|
* Find the actual width of the glyph and modify the
|
|
* description to reflect it. Not guaranteed to do
|
|
* any good, may make character spacing too wide.
|
|
*/
|
|
|
|
void
|
|
docorrectwidth(void)
|
|
{
|
|
int i;
|
|
GENTRY *ge;
|
|
GLYPH *g;
|
|
int xmin, xmax;
|
|
int maxwidth, minsp;
|
|
|
|
/* enforce this minimal spacing,
|
|
* we limit the amount of the enforced spacing to avoid
|
|
* spacing the bold wonts too widely
|
|
*/
|
|
minsp = (stdhw>60 || stdhw<10)? 60 : stdhw;
|
|
|
|
for (i = 0, g = glyph_list; i < numglyphs; i++, g++) {
|
|
g->oldwidth=g->scaledwidth; /* save the old width, will need for AFM */
|
|
|
|
if (correctwidth && g->flags & GF_USED) {
|
|
xmin = 5000;
|
|
xmax = -5000;
|
|
for (ge = g->entries; ge != 0; ge = ge->next) {
|
|
if (ge->type != GE_LINE && ge->type != GE_CURVE)
|
|
continue;
|
|
|
|
if (ge->ix3 <= xmin) {
|
|
xmin = ge->ix3;
|
|
}
|
|
if (ge->ix3 >= xmax) {
|
|
xmax = ge->ix3;
|
|
}
|
|
}
|
|
|
|
maxwidth=xmax+minsp;
|
|
if( g->scaledwidth < maxwidth ) {
|
|
g->scaledwidth = maxwidth;
|
|
WARNING_3 fprintf(stderr, "glyph %s: extended from %d to %d\n",
|
|
g->name, g->oldwidth, g->scaledwidth );
|
|
}
|
|
}
|
|
}
|
|
|
|
}
|
|
|
|
/*
|
|
* Try to find the typical stem widths
|
|
*/
|
|
|
|
void
|
|
stemstatistics(void)
|
|
{
|
|
#define MINDIST 10 /* minimal distance between the widths */
|
|
int hyst[MAXHYST+MINDIST*2];
|
|
int best[12];
|
|
int i, j, k, w;
|
|
int nchars;
|
|
int ns;
|
|
STEM *s;
|
|
GLYPH *g;
|
|
|
|
/* start with typical stem width */
|
|
|
|
nchars=0;
|
|
|
|
/* build the hystogram of horizontal stem widths */
|
|
memset(hyst, 0, sizeof hyst);
|
|
|
|
for (i = 0, g = glyph_list; i < numglyphs; i++, g++) {
|
|
if (g->flags & GF_USED) {
|
|
nchars++;
|
|
s = g->hstems;
|
|
for (j = 0; j < g->nhs; j += 2) {
|
|
if ((s[j].flags | s[j + 1].flags) & ST_END)
|
|
continue;
|
|
w = s[j + 1].value - s[j].value+1;
|
|
if(w==20) /* split stems should not be counted */
|
|
continue;
|
|
if (w > 0 && w < MAXHYST - 1) {
|
|
/*
|
|
* handle some fuzz present in
|
|
* converted fonts
|
|
*/
|
|
hyst[w+MINDIST] += MINDIST-1;
|
|
for(k=1; k<MINDIST-1; k++) {
|
|
hyst[w+MINDIST + k] += MINDIST-1-k;
|
|
hyst[w+MINDIST - k] += MINDIST-1-k;
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
/* find 12 most frequent values */
|
|
ns = besthyst(hyst+MINDIST, 0, best, 12, MINDIST, &stdhw);
|
|
|
|
/* store data in stemsnaph */
|
|
for (i = 0; i < ns; i++)
|
|
stemsnaph[i] = best[i];
|
|
if (ns < 12)
|
|
stemsnaph[ns] = 0;
|
|
|
|
/* build the hystogram of vertical stem widths */
|
|
memset(hyst, 0, sizeof hyst);
|
|
|
|
for (i = 0, g = glyph_list; i < numglyphs; i++, g++) {
|
|
if (g->flags & GF_USED) {
|
|
s = g->vstems;
|
|
for (j = 0; j < g->nvs; j += 2) {
|
|
if ((s[j].flags | s[j + 1].flags) & ST_END)
|
|
continue;
|
|
w = s[j + 1].value - s[j].value+1;
|
|
if (w > 0 && w < MAXHYST - 1) {
|
|
/*
|
|
* handle some fuzz present in
|
|
* converted fonts
|
|
*/
|
|
hyst[w+MINDIST] += MINDIST-1;
|
|
for(k=1; k<MINDIST-1; k++) {
|
|
hyst[w+MINDIST + k] += MINDIST-1-k;
|
|
hyst[w+MINDIST - k] += MINDIST-1-k;
|
|
}
|
|
}
|
|
}
|
|
}
|
|
}
|
|
|
|
/* find 12 most frequent values */
|
|
ns = besthyst(hyst+MINDIST, 0, best, 12, MINDIST, &stdvw);
|
|
|
|
/* store data in stemsnaph */
|
|
for (i = 0; i < ns; i++)
|
|
stemsnapv[i] = best[i];
|
|
if (ns < 12)
|
|
stemsnapv[ns] = 0;
|
|
|
|
#undef MINDIST
|
|
}
|
|
|
|
/*
|
|
* SB
|
|
* A funny thing: TTF paths are going in reverse direction compared
|
|
* to Type1. So after all (because the rest of logic uses TTF
|
|
* path directions) we have to reverse the paths.
|
|
*
|
|
* It was a big headache to discover that.
|
|
*/
|
|
|
|
/* works on both int and float paths */
|
|
|
|
void
|
|
reversepathsfromto(
|
|
GENTRY * from,
|
|
GENTRY * to
|
|
)
|
|
{
|
|
GENTRY *ge, *nge, *pge;
|
|
GENTRY *cur, *next;
|
|
int i, n, ilast[2];
|
|
double flast[2], f;
|
|
|
|
for (ge = from; ge != 0 && ge != to; ge = ge->next) {
|
|
if(ge->type == GE_LINE || ge->type == GE_CURVE) {
|
|
if (ISDBG(REVERSAL))
|
|
fprintf(stderr, "reverse path 0x%x <- 0x%x, 0x%x\n", ge, ge->prev, ge->bkwd);
|
|
|
|
/* cut out the path itself */
|
|
pge = ge->prev; /* GE_MOVE */
|
|
if (pge == 0) {
|
|
fprintf(stderr, "**! No MOVE before line !!! Fatal. ****\n");
|
|
exit(1);
|
|
}
|
|
nge = ge->bkwd->next; /* GE_PATH */
|
|
pge->next = nge;
|
|
nge->prev = pge;
|
|
ge->bkwd->next = 0; /* mark end of chain */
|
|
|
|
/* remember the starting point */
|
|
if(ge->flags & GEF_FLOAT) {
|
|
flast[0] = pge->fx3;
|
|
flast[1] = pge->fy3;
|
|
} else {
|
|
ilast[0] = pge->ix3;
|
|
ilast[1] = pge->iy3;
|
|
}
|
|
|
|
/* then reinsert them in backwards order */
|
|
for(cur = ge; cur != 0; cur = next ) {
|
|
next = cur->next; /* or addgeafter() will screw it up */
|
|
if(cur->flags & GEF_FLOAT) {
|
|
for(i=0; i<2; i++) {
|
|
/* reverse the direction of path element */
|
|
f = cur->fpoints[i][0];
|
|
cur->fpoints[i][0] = cur->fpoints[i][1];
|
|
cur->fpoints[i][1] = f;
|
|
f = flast[i];
|
|
flast[i] = cur->fpoints[i][2];
|
|
cur->fpoints[i][2] = f;
|
|
}
|
|
} else {
|
|
for(i=0; i<2; i++) {
|
|
/* reverse the direction of path element */
|
|
n = cur->ipoints[i][0];
|
|
cur->ipoints[i][0] = cur->ipoints[i][1];
|
|
cur->ipoints[i][1] = n;
|
|
n = ilast[i];
|
|
ilast[i] = cur->ipoints[i][2];
|
|
cur->ipoints[i][2] = n;
|
|
}
|
|
}
|
|
addgeafter(pge, cur);
|
|
}
|
|
|
|
/* restore the starting point */
|
|
if(ge->flags & GEF_FLOAT) {
|
|
pge->fx3 = flast[0];
|
|
pge->fy3 = flast[1];
|
|
} else {
|
|
pge->ix3 = ilast[0];
|
|
pge->iy3 = ilast[1];
|
|
}
|
|
|
|
ge = nge;
|
|
}
|
|
|
|
}
|
|
}
|
|
|
|
void
|
|
reversepaths(
|
|
GLYPH * g
|
|
)
|
|
{
|
|
reversepathsfromto(g->entries, NULL);
|
|
}
|
|
|
|
/* add a kerning pair information, scales the value */
|
|
|
|
void
|
|
addkernpair(
|
|
unsigned id1,
|
|
unsigned id2,
|
|
int unscval
|
|
)
|
|
{
|
|
static unsigned char *bits = 0;
|
|
static int lastid;
|
|
GLYPH *g = &glyph_list[id1];
|
|
int i, n;
|
|
struct kern *p;
|
|
|
|
if(unscval == 0 || id1 >= numglyphs || id2 >= numglyphs)
|
|
return;
|
|
|
|
if( (glyph_list[id1].flags & GF_USED)==0
|
|
|| (glyph_list[id2].flags & GF_USED)==0 )
|
|
return;
|
|
|
|
if(bits == 0) {
|
|
bits = calloc( BITMAP_BYTES(numglyphs), 1);
|
|
if (bits == NULL) {
|
|
fprintf (stderr, "****malloc failed %s line %d\n", __FILE__, __LINE__);
|
|
exit(255);
|
|
}
|
|
lastid = id1;
|
|
}
|
|
|
|
if(lastid != id1) {
|
|
/* refill the bitmap cache */
|
|
memset(bits, 0,BITMAP_BYTES(numglyphs));
|
|
p = g->kern;
|
|
for(i=g->kerncount; i>0; i--) {
|
|
n = (p++)->id;
|
|
SET_BITMAP(bits, n);
|
|
}
|
|
lastid = id1;
|
|
}
|
|
|
|
if(IS_BITMAP(bits, id2))
|
|
return; /* duplicate */
|
|
|
|
if(g->kerncount <= g->kernalloc) {
|
|
g->kernalloc += 8;
|
|
p = realloc(g->kern, sizeof(struct kern) * g->kernalloc);
|
|
if(p == 0) {
|
|
fprintf (stderr, "** realloc failed, kerning data will be incomplete\n");
|
|
}
|
|
g->kern = p;
|
|
}
|
|
|
|
SET_BITMAP(bits, id2);
|
|
p = &g->kern[g->kerncount];
|
|
p->id = id2;
|
|
p->val = iscale(unscval) - (g->scaledwidth - g->oldwidth);
|
|
g->kerncount++;
|
|
kerning_pairs++;
|
|
}
|
|
|
|
/* print out the kerning information */
|
|
|
|
void
|
|
print_kerning(
|
|
FILE *afm_file
|
|
)
|
|
{
|
|
int i, j, n;
|
|
GLYPH *g;
|
|
struct kern *p;
|
|
|
|
if( kerning_pairs == 0 )
|
|
return;
|
|
|
|
fprintf(afm_file, "StartKernData\n");
|
|
fprintf(afm_file, "StartKernPairs %hd\n", kerning_pairs);
|
|
|
|
for(i=0; i<numglyphs; i++) {
|
|
g = &glyph_list[i];
|
|
if( (g->flags & GF_USED) ==0)
|
|
continue;
|
|
p = g->kern;
|
|
for(j=g->kerncount; j>0; j--, p++) {
|
|
fprintf(afm_file, "KPX %s %s %d\n", g->name,
|
|
glyph_list[ p->id ].name, p->val );
|
|
}
|
|
}
|
|
|
|
fprintf(afm_file, "EndKernPairs\n");
|
|
fprintf(afm_file, "EndKernData\n");
|
|
}
|
|
|
|
|
|
#if 0
|
|
|
|
/*
|
|
** This function is commented out because the information
|
|
** collected by it is not used anywhere else yet. Now
|
|
** it only collects the directions of contours. And the
|
|
** direction of contours gets fixed already in draw_glyf().
|
|
**
|
|
***********************************************
|
|
**
|
|
** Here we expect that the paths are already closed.
|
|
** We also expect that the contours do not intersect
|
|
** and that curves doesn't cross any border of quadrant.
|
|
**
|
|
** Find which contours go inside which and what is
|
|
** their proper direction. Then fix the direction
|
|
** to make it right.
|
|
**
|
|
*/
|
|
|
|
#define MAXCONT 1000
|
|
|
|
void
|
|
fixcontours(
|
|
GLYPH * g
|
|
)
|
|
{
|
|
CONTOUR cont[MAXCONT];
|
|
short ymax[MAXCONT]; /* the highest point */
|
|
short xofmax[MAXCONT]; /* X-coordinate of any point
|
|
* at ymax */
|
|
short ymin[MAXCONT]; /* the lowest point */
|
|
short xofmin[MAXCONT]; /* X-coordinate of any point
|
|
* at ymin */
|
|
short count[MAXCONT]; /* count of lines */
|
|
char dir[MAXCONT]; /* in which direction they must go */
|
|
GENTRY *start[MAXCONT], *minptr[MAXCONT], *maxptr[MAXCONT];
|
|
int ncont;
|
|
int i;
|
|
int dx1, dy1, dx2, dy2;
|
|
GENTRY *ge, *nge;
|
|
|
|
/* find the contours and their most upper/lower points */
|
|
ncont = 0;
|
|
ymax[0] = -5000;
|
|
ymin[0] = 5000;
|
|
for (ge = g->entries; ge != 0; ge = ge->next) {
|
|
if (ge->type == GE_LINE || ge->type == GE_CURVE) {
|
|
if (ge->iy3 > ymax[ncont]) {
|
|
ymax[ncont] = ge->iy3;
|
|
xofmax[ncont] = ge->ix3;
|
|
maxptr[ncont] = ge;
|
|
}
|
|
if (ge->iy3 < ymin[ncont]) {
|
|
ymin[ncont] = ge->iy3;
|
|
xofmin[ncont] = ge->ix3;
|
|
minptr[ncont] = ge;
|
|
}
|
|
}
|
|
if (ge->frwd != ge->next) {
|
|
start[ncont++] = ge->frwd;
|
|
ymax[ncont] = -5000;
|
|
ymin[ncont] = 5000;
|
|
}
|
|
}
|
|
|
|
/* determine the directions of contours */
|
|
for (i = 0; i < ncont; i++) {
|
|
ge = minptr[i];
|
|
nge = ge->frwd;
|
|
|
|
if (ge->type == GE_CURVE) {
|
|
dx1 = ge->ix3 - ge->ix2;
|
|
dy1 = ge->iy3 - ge->iy2;
|
|
|
|
if (dx1 == 0 && dy1 == 0) { /* a pathological case */
|
|
dx1 = ge->ix3 - ge->ix1;
|
|
dy1 = ge->iy3 - ge->iy1;
|
|
}
|
|
if (dx1 == 0 && dy1 == 0) { /* a more pathological
|
|
* case */
|
|
dx1 = ge->ix3 - ge->prev->ix3;
|
|
dy1 = ge->iy3 - ge->prev->iy3;
|
|
}
|
|
} else {
|
|
dx1 = ge->ix3 - ge->prev->ix3;
|
|
dy1 = ge->iy3 - ge->prev->iy3;
|
|
}
|
|
if (nge->type == GE_CURVE) {
|
|
dx2 = ge->ix3 - nge->ix1;
|
|
dy2 = ge->iy3 - nge->iy1;
|
|
if (dx1 == 0 && dy1 == 0) { /* a pathological case */
|
|
dx2 = ge->ix3 - nge->ix2;
|
|
dy2 = ge->iy3 - nge->iy2;
|
|
}
|
|
if (dx1 == 0 && dy1 == 0) { /* a more pathological
|
|
* case */
|
|
dx2 = ge->ix3 - nge->ix3;
|
|
dy2 = ge->iy3 - nge->iy3;
|
|
}
|
|
} else {
|
|
dx2 = ge->ix3 - nge->ix3;
|
|
dy2 = ge->iy3 - nge->iy3;
|
|
}
|
|
|
|
/* compare angles */
|
|
cont[i].direction = DIR_INNER;
|
|
if (dy1 == 0) {
|
|
if (dx1 < 0)
|
|
cont[i].direction = DIR_OUTER;
|
|
} else if (dy2 == 0) {
|
|
if (dx2 > 0)
|
|
cont[i].direction = DIR_OUTER;
|
|
} else if (dx2 * dy1 < dx1 * dy2)
|
|
cont[i].direction = DIR_OUTER;
|
|
|
|
cont[i].ymin = ymin[i];
|
|
cont[i].xofmin = xofmin[i];
|
|
}
|
|
|
|
/* save the information that may be needed further */
|
|
g->ncontours = ncont;
|
|
if (ncont > 0) {
|
|
g->contours = malloc(sizeof(CONTOUR) * ncont);
|
|
if (g->contours == 0) {
|
|
fprintf(stderr, "***** Memory allocation error *****\n");
|
|
exit(255);
|
|
}
|
|
memcpy(g->contours, cont, sizeof(CONTOUR) * ncont);
|
|
}
|
|
}
|
|
|
|
#endif
|
|
|
|
/*
|
|
*
|
|
*/
|
|
|