mutter/cogl/cogl-path/tesselator/tess.c
2016-04-22 16:44:31 +02:00

633 lines
17 KiB
C

/*
* SGI FREE SOFTWARE LICENSE B (Version 2.0, Sept. 18, 2008)
* Copyright (C) 1991-2000 Silicon Graphics, Inc. All Rights Reserved.
*
* Permission is hereby granted, free of charge, to any person obtaining a
* copy of this software and associated documentation files (the "Software"),
* to deal in the Software without restriction, including without limitation
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
* and/or sell copies of the Software, and to permit persons to whom the
* Software is furnished to do so, subject to the following conditions:
*
* The above copyright notice including the dates of first publication and
* either this permission notice or a reference to
* http://oss.sgi.com/projects/FreeB/
* shall be included in all copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
* OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
* SILICON GRAPHICS, INC. BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
* WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF
* OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
* SOFTWARE.
*
* Except as contained in this notice, the name of Silicon Graphics, Inc.
* shall not be used in advertising or otherwise to promote the sale, use or
* other dealings in this Software without prior written authorization from
* Silicon Graphics, Inc.
*/
/*
** Author: Eric Veach, July 1994.
**
*/
#include "gluos.h"
#include <stddef.h>
#include <assert.h>
#include <setjmp.h>
#include "memalloc.h"
#include "tess.h"
#include "mesh.h"
#include "normal.h"
#include "sweep.h"
#include "tessmono.h"
#include "render.h"
#define GLU_TESS_DEFAULT_TOLERANCE 0.0
#define GLU_TESS_MESH 100112 /* void (*)(GLUmesh *mesh) */
#ifndef TRUE
#define TRUE 1
#endif
#ifndef FALSE
#define FALSE 0
#endif
/*ARGSUSED*/ static void GLAPIENTRY noBegin( GLenum type ) {}
/*ARGSUSED*/ static void GLAPIENTRY noEdgeFlag( GLboolean boundaryEdge ) {}
/*ARGSUSED*/ static void GLAPIENTRY noVertex( void *data ) {}
/*ARGSUSED*/ static void GLAPIENTRY noEnd( void ) {}
/*ARGSUSED*/ static void GLAPIENTRY noError( GLenum errnum ) {}
/*ARGSUSED*/ static void GLAPIENTRY noCombine( GLdouble coords[3], void *data[4],
GLfloat weight[4], void **dataOut ) {}
/*ARGSUSED*/ static void GLAPIENTRY noMesh( GLUmesh *mesh ) {}
/*ARGSUSED*/ void GLAPIENTRY __gl_noBeginData( GLenum type,
void *polygonData ) {}
/*ARGSUSED*/ void GLAPIENTRY __gl_noEdgeFlagData( GLboolean boundaryEdge,
void *polygonData ) {}
/*ARGSUSED*/ void GLAPIENTRY __gl_noVertexData( void *data,
void *polygonData ) {}
/*ARGSUSED*/ void GLAPIENTRY __gl_noEndData( void *polygonData ) {}
/*ARGSUSED*/ void GLAPIENTRY __gl_noErrorData( GLenum errnum,
void *polygonData ) {}
/*ARGSUSED*/ void GLAPIENTRY __gl_noCombineData( GLdouble coords[3],
void *data[4],
GLfloat weight[4],
void **outData,
void *polygonData ) {}
/* Half-edges are allocated in pairs (see mesh.c) */
typedef struct { GLUhalfEdge e, eSym; } EdgePair;
#undef MAX
#define MAX(a,b) ((a) > (b) ? (a) : (b))
#define MAX_FAST_ALLOC (MAX(sizeof(EdgePair), \
MAX(sizeof(GLUvertex),sizeof(GLUface))))
GLUtesselator * GLAPIENTRY
gluNewTess( void )
{
GLUtesselator *tess;
/* Only initialize fields which can be changed by the api. Other fields
* are initialized where they are used.
*/
if (memInit( MAX_FAST_ALLOC ) == 0) {
return 0; /* out of memory */
}
tess = (GLUtesselator *)memAlloc( sizeof( GLUtesselator ));
if (tess == NULL) {
return 0; /* out of memory */
}
tess->state = T_DORMANT;
tess->normal[0] = 0;
tess->normal[1] = 0;
tess->normal[2] = 0;
tess->relTolerance = GLU_TESS_DEFAULT_TOLERANCE;
tess->windingRule = GLU_TESS_WINDING_ODD;
tess->flagBoundary = FALSE;
tess->boundaryOnly = FALSE;
tess->callBegin = &noBegin;
tess->callEdgeFlag = &noEdgeFlag;
tess->callVertex = &noVertex;
tess->callEnd = &noEnd;
tess->callError = &noError;
tess->callCombine = &noCombine;
tess->callMesh = &noMesh;
tess->callBeginData= &__gl_noBeginData;
tess->callEdgeFlagData= &__gl_noEdgeFlagData;
tess->callVertexData= &__gl_noVertexData;
tess->callEndData= &__gl_noEndData;
tess->callErrorData= &__gl_noErrorData;
tess->callCombineData= &__gl_noCombineData;
tess->polygonData= NULL;
return tess;
}
static void MakeDormant( GLUtesselator *tess )
{
/* Return the tessellator to its original dormant state. */
if( tess->mesh != NULL ) {
__gl_meshDeleteMesh( tess->mesh );
}
tess->state = T_DORMANT;
tess->lastEdge = NULL;
tess->mesh = NULL;
}
#define RequireState( tess, s ) if( tess->state != s ) GotoState(tess,s)
static void GotoState( GLUtesselator *tess, enum TessState newState )
{
while( tess->state != newState ) {
/* We change the current state one level at a time, to get to
* the desired state.
*/
if( tess->state < newState ) {
switch( tess->state ) {
case T_DORMANT:
CALL_ERROR_OR_ERROR_DATA( GLU_TESS_MISSING_BEGIN_POLYGON );
gluTessBeginPolygon( tess, NULL );
break;
case T_IN_POLYGON:
CALL_ERROR_OR_ERROR_DATA( GLU_TESS_MISSING_BEGIN_CONTOUR );
gluTessBeginContour( tess );
break;
default:
;
}
} else {
switch( tess->state ) {
case T_IN_CONTOUR:
CALL_ERROR_OR_ERROR_DATA( GLU_TESS_MISSING_END_CONTOUR );
gluTessEndContour( tess );
break;
case T_IN_POLYGON:
CALL_ERROR_OR_ERROR_DATA( GLU_TESS_MISSING_END_POLYGON );
/* gluTessEndPolygon( tess ) is too much work! */
MakeDormant( tess );
break;
default:
;
}
}
}
}
void GLAPIENTRY
gluDeleteTess( GLUtesselator *tess )
{
RequireState( tess, T_DORMANT );
memFree( tess );
}
void GLAPIENTRY
gluTessProperty( GLUtesselator *tess, GLenum which, GLdouble value )
{
GLenum windingRule;
switch( which ) {
case GLU_TESS_TOLERANCE:
if( value < 0.0 || value > 1.0 ) break;
tess->relTolerance = value;
return;
case GLU_TESS_WINDING_RULE:
windingRule = (GLenum) value;
if( windingRule != value ) break; /* not an integer */
switch( windingRule ) {
case GLU_TESS_WINDING_ODD:
case GLU_TESS_WINDING_NONZERO:
case GLU_TESS_WINDING_POSITIVE:
case GLU_TESS_WINDING_NEGATIVE:
case GLU_TESS_WINDING_ABS_GEQ_TWO:
tess->windingRule = windingRule;
return;
default:
break;
}
case GLU_TESS_BOUNDARY_ONLY:
tess->boundaryOnly = (value != 0);
return;
default:
CALL_ERROR_OR_ERROR_DATA( GLU_INVALID_ENUM );
return;
}
CALL_ERROR_OR_ERROR_DATA( GLU_INVALID_VALUE );
}
/* Returns tessellator property */
void GLAPIENTRY
gluGetTessProperty( GLUtesselator *tess, GLenum which, GLdouble *value )
{
switch (which) {
case GLU_TESS_TOLERANCE:
/* tolerance should be in range [0..1] */
assert(0.0 <= tess->relTolerance && tess->relTolerance <= 1.0);
*value= tess->relTolerance;
break;
case GLU_TESS_WINDING_RULE:
assert(tess->windingRule == GLU_TESS_WINDING_ODD ||
tess->windingRule == GLU_TESS_WINDING_NONZERO ||
tess->windingRule == GLU_TESS_WINDING_POSITIVE ||
tess->windingRule == GLU_TESS_WINDING_NEGATIVE ||
tess->windingRule == GLU_TESS_WINDING_ABS_GEQ_TWO);
*value= tess->windingRule;
break;
case GLU_TESS_BOUNDARY_ONLY:
assert(tess->boundaryOnly == TRUE || tess->boundaryOnly == FALSE);
*value= tess->boundaryOnly;
break;
default:
*value= 0.0;
CALL_ERROR_OR_ERROR_DATA( GLU_INVALID_ENUM );
break;
}
} /* gluGetTessProperty() */
void GLAPIENTRY
gluTessNormal( GLUtesselator *tess, GLdouble x, GLdouble y, GLdouble z )
{
tess->normal[0] = x;
tess->normal[1] = y;
tess->normal[2] = z;
}
void GLAPIENTRY
gluTessCallback( GLUtesselator *tess, GLenum which, _GLUfuncptr fn)
{
switch( which ) {
case GLU_TESS_BEGIN:
tess->callBegin = (fn == NULL) ? &noBegin : (void (GLAPIENTRY *)(GLenum)) fn;
return;
case GLU_TESS_BEGIN_DATA:
tess->callBeginData = (fn == NULL) ?
&__gl_noBeginData : (void (GLAPIENTRY *)(GLenum, void *)) fn;
return;
case GLU_TESS_EDGE_FLAG:
tess->callEdgeFlag = (fn == NULL) ? &noEdgeFlag :
(void (GLAPIENTRY *)(GLboolean)) fn;
/* If the client wants boundary edges to be flagged,
* we render everything as separate triangles (no strips or fans).
*/
tess->flagBoundary = (fn != NULL);
return;
case GLU_TESS_EDGE_FLAG_DATA:
tess->callEdgeFlagData= (fn == NULL) ?
&__gl_noEdgeFlagData : (void (GLAPIENTRY *)(GLboolean, void *)) fn;
/* If the client wants boundary edges to be flagged,
* we render everything as separate triangles (no strips or fans).
*/
tess->flagBoundary = (fn != NULL);
return;
case GLU_TESS_VERTEX:
tess->callVertex = (fn == NULL) ? &noVertex :
(void (GLAPIENTRY *)(void *)) fn;
return;
case GLU_TESS_VERTEX_DATA:
tess->callVertexData = (fn == NULL) ?
&__gl_noVertexData : (void (GLAPIENTRY *)(void *, void *)) fn;
return;
case GLU_TESS_END:
tess->callEnd = (fn == NULL) ? &noEnd : (void (GLAPIENTRY *)(void)) fn;
return;
case GLU_TESS_END_DATA:
tess->callEndData = (fn == NULL) ? &__gl_noEndData :
(void (GLAPIENTRY *)(void *)) fn;
return;
case GLU_TESS_ERROR:
tess->callError = (fn == NULL) ? &noError : (void (GLAPIENTRY *)(GLenum)) fn;
return;
case GLU_TESS_ERROR_DATA:
tess->callErrorData = (fn == NULL) ?
&__gl_noErrorData : (void (GLAPIENTRY *)(GLenum, void *)) fn;
return;
case GLU_TESS_COMBINE:
tess->callCombine = (fn == NULL) ? &noCombine :
(void (GLAPIENTRY *)(GLdouble [3],void *[4], GLfloat [4], void ** )) fn;
return;
case GLU_TESS_COMBINE_DATA:
tess->callCombineData = (fn == NULL) ? &__gl_noCombineData :
(void (GLAPIENTRY *)(GLdouble [3],
void *[4],
GLfloat [4],
void **,
void *)) fn;
return;
case GLU_TESS_MESH:
tess->callMesh = (fn == NULL) ? &noMesh : (void (GLAPIENTRY *)(GLUmesh *)) fn;
return;
default:
CALL_ERROR_OR_ERROR_DATA( GLU_INVALID_ENUM );
return;
}
}
static int AddVertex( GLUtesselator *tess, GLdouble coords[3], void *data )
{
GLUhalfEdge *e;
e = tess->lastEdge;
if( e == NULL ) {
/* Make a self-loop (one vertex, one edge). */
e = __gl_meshMakeEdge( tess->mesh );
if (e == NULL) return 0;
if ( !__gl_meshSplice( e, e->Sym ) ) return 0;
} else {
/* Create a new vertex and edge which immediately follow e
* in the ordering around the left face.
*/
if (__gl_meshSplitEdge( e ) == NULL) return 0;
e = e->Lnext;
}
/* The new vertex is now e->Org. */
e->Org->data = data;
e->Org->coords[0] = coords[0];
e->Org->coords[1] = coords[1];
e->Org->coords[2] = coords[2];
/* The winding of an edge says how the winding number changes as we
* cross from the edge''s right face to its left face. We add the
* vertices in such an order that a CCW contour will add +1 to
* the winding number of the region inside the contour.
*/
e->winding = 1;
e->Sym->winding = -1;
tess->lastEdge = e;
return 1;
}
static void CacheVertex( GLUtesselator *tess, GLdouble coords[3], void *data )
{
CachedVertex *v = &tess->cache[tess->cacheCount];
v->data = data;
v->coords[0] = coords[0];
v->coords[1] = coords[1];
v->coords[2] = coords[2];
++tess->cacheCount;
}
static int EmptyCache( GLUtesselator *tess )
{
CachedVertex *v = tess->cache;
CachedVertex *vLast;
tess->mesh = __gl_meshNewMesh();
if (tess->mesh == NULL) return 0;
for( vLast = v + tess->cacheCount; v < vLast; ++v ) {
if ( !AddVertex( tess, v->coords, v->data ) ) return 0;
}
tess->cacheCount = 0;
tess->emptyCache = FALSE;
return 1;
}
void GLAPIENTRY
gluTessVertex( GLUtesselator *tess, GLdouble coords[3], void *data )
{
int i, tooLarge = FALSE;
GLdouble x, clamped[3];
RequireState( tess, T_IN_CONTOUR );
if( tess->emptyCache ) {
if ( !EmptyCache( tess ) ) {
CALL_ERROR_OR_ERROR_DATA( GLU_OUT_OF_MEMORY );
return;
}
tess->lastEdge = NULL;
}
for( i = 0; i < 3; ++i ) {
x = coords[i];
if( x < - GLU_TESS_MAX_COORD ) {
x = - GLU_TESS_MAX_COORD;
tooLarge = TRUE;
}
if( x > GLU_TESS_MAX_COORD ) {
x = GLU_TESS_MAX_COORD;
tooLarge = TRUE;
}
clamped[i] = x;
}
if( tooLarge ) {
CALL_ERROR_OR_ERROR_DATA( GLU_TESS_COORD_TOO_LARGE );
}
if( tess->mesh == NULL ) {
if( tess->cacheCount < TESS_MAX_CACHE ) {
CacheVertex( tess, clamped, data );
return;
}
if ( !EmptyCache( tess ) ) {
CALL_ERROR_OR_ERROR_DATA( GLU_OUT_OF_MEMORY );
return;
}
}
if ( !AddVertex( tess, clamped, data ) ) {
CALL_ERROR_OR_ERROR_DATA( GLU_OUT_OF_MEMORY );
}
}
void GLAPIENTRY
gluTessBeginPolygon( GLUtesselator *tess, void *data )
{
RequireState( tess, T_DORMANT );
tess->state = T_IN_POLYGON;
tess->cacheCount = 0;
tess->emptyCache = FALSE;
tess->mesh = NULL;
tess->polygonData= data;
}
void GLAPIENTRY
gluTessBeginContour( GLUtesselator *tess )
{
RequireState( tess, T_IN_POLYGON );
tess->state = T_IN_CONTOUR;
tess->lastEdge = NULL;
if( tess->cacheCount > 0 ) {
/* Just set a flag so we don't get confused by empty contours
* -- these can be generated accidentally with the obsolete
* NextContour() interface.
*/
tess->emptyCache = TRUE;
}
}
void GLAPIENTRY
gluTessEndContour( GLUtesselator *tess )
{
RequireState( tess, T_IN_CONTOUR );
tess->state = T_IN_POLYGON;
}
void GLAPIENTRY
gluTessEndPolygon( GLUtesselator *tess )
{
GLUmesh *mesh;
if (setjmp(tess->env) != 0) {
/* come back here if out of memory */
CALL_ERROR_OR_ERROR_DATA( GLU_OUT_OF_MEMORY );
return;
}
RequireState( tess, T_IN_POLYGON );
tess->state = T_DORMANT;
if( tess->mesh == NULL ) {
if( ! tess->flagBoundary && tess->callMesh == &noMesh ) {
/* Try some special code to make the easy cases go quickly
* (eg. convex polygons). This code does NOT handle multiple contours,
* intersections, edge flags, and of course it does not generate
* an explicit mesh either.
*/
if( __gl_renderCache( tess )) {
tess->polygonData= NULL;
return;
}
}
if ( !EmptyCache( tess ) ) longjmp(tess->env,1); /* could've used a label*/
}
/* Determine the polygon normal and project vertices onto the plane
* of the polygon.
*/
__gl_projectPolygon( tess );
/* __gl_computeInterior( tess ) computes the planar arrangement specified
* by the given contours, and further subdivides this arrangement
* into regions. Each region is marked "inside" if it belongs
* to the polygon, according to the rule given by tess->windingRule.
* Each interior region is guaranteed be monotone.
*/
if ( !__gl_computeInterior( tess ) ) {
longjmp(tess->env,1); /* could've used a label */
}
mesh = tess->mesh;
if( ! tess->fatalError ) {
int rc = 1;
/* If the user wants only the boundary contours, we throw away all edges
* except those which separate the interior from the exterior.
* Otherwise we tessellate all the regions marked "inside".
*/
if( tess->boundaryOnly ) {
rc = __gl_meshSetWindingNumber( mesh, 1, TRUE );
} else {
rc = __gl_meshTessellateInterior( mesh );
}
if (rc == 0) longjmp(tess->env,1); /* could've used a label */
__gl_meshCheckMesh( mesh );
if( tess->callBegin != &noBegin || tess->callEnd != &noEnd
|| tess->callVertex != &noVertex || tess->callEdgeFlag != &noEdgeFlag
|| tess->callBeginData != &__gl_noBeginData
|| tess->callEndData != &__gl_noEndData
|| tess->callVertexData != &__gl_noVertexData
|| tess->callEdgeFlagData != &__gl_noEdgeFlagData )
{
if( tess->boundaryOnly ) {
__gl_renderBoundary( tess, mesh ); /* output boundary contours */
} else {
__gl_renderMesh( tess, mesh ); /* output strips and fans */
}
}
if( tess->callMesh != &noMesh ) {
/* Throw away the exterior faces, so that all faces are interior.
* This way the user doesn't have to check the "inside" flag,
* and we don't need to even reveal its existence. It also leaves
* the freedom for an implementation to not generate the exterior
* faces in the first place.
*/
__gl_meshDiscardExterior( mesh );
(*tess->callMesh)( mesh ); /* user wants the mesh itself */
tess->mesh = NULL;
tess->polygonData= NULL;
return;
}
}
__gl_meshDeleteMesh( mesh );
tess->polygonData= NULL;
tess->mesh = NULL;
}
/*XXXblythe unused function*/
#if 0
void GLAPIENTRY
gluDeleteMesh( GLUmesh *mesh )
{
__gl_meshDeleteMesh( mesh );
}
#endif
/*******************************************************/
/* Obsolete calls -- for backward compatibility */
void GLAPIENTRY
gluBeginPolygon( GLUtesselator *tess )
{
gluTessBeginPolygon( tess, NULL );
gluTessBeginContour( tess );
}
/*ARGSUSED*/
void GLAPIENTRY
gluNextContour( GLUtesselator *tess, GLenum type )
{
gluTessEndContour( tess );
gluTessBeginContour( tess );
}
void GLAPIENTRY
gluEndPolygon( GLUtesselator *tess )
{
gluTessEndContour( tess );
gluTessEndPolygon( tess );
}