zcy
/
solvespace
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
solvespace/src/srf/boolean.cpp

958 lines
34 KiB
C++
Raw Normal View History

Changes in preparation for the release of SolveSpace under the GPL, to add that license, and change all copyright notices to me, not Useful Subset, LLC. [git-p4: depot-paths = "//depot/solvespace/": change = 2211]
2013-07-28 22:08:34 +00:00
//-----------------------------------------------------------------------------
// Top-level functions to compute the Boolean union or difference between
// two shells of rational polynomial surfaces.
Trim trailing whitespace.
2015-03-29 00:30:52 +00:00
//
Changes in preparation for the release of SolveSpace under the GPL, to add that license, and change all copyright notices to me, not Useful Subset, LLC. [git-p4: depot-paths = "//depot/solvespace/": change = 2211]
2013-07-28 22:08:34 +00:00
// Copyright 2008-2013 Jonathan Westhues.
//-----------------------------------------------------------------------------
Tear everything apart, moving away from meshes and toward shells. Add stubs for functions to perform Booleans, and get rid of mesh stuff, including the kd tree accelerated snap to vertex (which should not be required if the shell triangulation performs as it should). Also check that a sketch is not self-intersecting before extruding it or whatever. This is dead slow, needs n*log(n) implementation. [git-p4: depot-paths = "//depot/solvespace/": change = 1902]
2009-01-23 03:30:30 +00:00
#include "solvespace.h"
Enable and mollify -Wunused-variable. In my (whitequark's) experience this warning tends to expose copy-paste errors with a high SNR, so making a few fragments slightly less symmetric is worth it. Also mollify -Wlogical-op-parentheses while we're at it.
2016-02-14 00:36:54 +00:00
static int I;
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
Tear everything apart, moving away from meshes and toward shells. Add stubs for functions to perform Booleans, and get rid of mesh stuff, including the kd tree accelerated snap to vertex (which should not be required if the shell triangulation performs as it should). Also check that a sketch is not self-intersecting before extruding it or whatever. This is dead slow, needs n*log(n) implementation. [git-p4: depot-paths = "//depot/solvespace/": change = 1902]
2009-01-23 03:30:30 +00:00
void SShell::MakeFromUnionOf(SShell *a, SShell *b) {
Convert all enumerations to use `enum class`. Specifically, take the old code that looks like this: class Foo { enum { X = 1, Y = 2 }; int kind; } ... foo.kind = Foo::X; ... and convert it to this: class Foo { enum class Kind : uint32_t { X = 1, Y = 2 }; Kind kind; } ... foo.kind = Foo::Kind::X; (In some cases the enumeration would not be in the class namespace, such as when it is generally useful.) The benefits are as follows: * The type of the field gives a clear indication of intent, both to humans and tools (such as binding generators). * The compiler is able to automatically warn when a switch is not exhaustive; but this is currently suppressed by the default: ssassert(false, ...) idiom. * Integers and plain enums are weakly type checked: they implicitly convert into each other. This can hide bugs where type conversion is performed but not intended. Enum classes are strongly type checked. * Plain enums pollute parent namespaces; enum classes do not. Almost every defined enum we have already has a kind of ad-hoc namespacing via `NAMESPACE_`, which is now explicit. * Plain enums do not have a well-defined ABI size, which is important for bindings. Enum classes can have it, if specified. We specify the base type for all enums as uint32_t, which is a safe choice and allows us to not change the numeric values of any variants. This commit introduces absolutely no functional change to the code, just renaming and change of types. It handles almost all cases, except GraphicsWindow::pending.operation, which needs minor functional change.
2016-05-20 08:31:20 +00:00
MakeFromBoolean(a, b, SSurface::CombineAs::UNION);
Tear everything apart, moving away from meshes and toward shells. Add stubs for functions to perform Booleans, and get rid of mesh stuff, including the kd tree accelerated snap to vertex (which should not be required if the shell triangulation performs as it should). Also check that a sketch is not self-intersecting before extruding it or whatever. This is dead slow, needs n*log(n) implementation. [git-p4: depot-paths = "//depot/solvespace/": change = 1902]
2009-01-23 03:30:30 +00:00
}
void SShell::MakeFromDifferenceOf(SShell *a, SShell *b) {
Convert all enumerations to use `enum class`. Specifically, take the old code that looks like this: class Foo { enum { X = 1, Y = 2 }; int kind; } ... foo.kind = Foo::X; ... and convert it to this: class Foo { enum class Kind : uint32_t { X = 1, Y = 2 }; Kind kind; } ... foo.kind = Foo::Kind::X; (In some cases the enumeration would not be in the class namespace, such as when it is generally useful.) The benefits are as follows: * The type of the field gives a clear indication of intent, both to humans and tools (such as binding generators). * The compiler is able to automatically warn when a switch is not exhaustive; but this is currently suppressed by the default: ssassert(false, ...) idiom. * Integers and plain enums are weakly type checked: they implicitly convert into each other. This can hide bugs where type conversion is performed but not intended. Enum classes are strongly type checked. * Plain enums pollute parent namespaces; enum classes do not. Almost every defined enum we have already has a kind of ad-hoc namespacing via `NAMESPACE_`, which is now explicit. * Plain enums do not have a well-defined ABI size, which is important for bindings. Enum classes can have it, if specified. We specify the base type for all enums as uint32_t, which is a safe choice and allows us to not change the numeric values of any variants. This commit introduces absolutely no functional change to the code, just renaming and change of types. It handles almost all cases, except GraphicsWindow::pending.operation, which needs minor functional change.
2016-05-20 08:31:20 +00:00
MakeFromBoolean(a, b, SSurface::CombineAs::DIFFERENCE);
Preparatory work for Boolean. Make the u and v coordinates of the trim curves for all surfaces lie between 0 and 1. And add routines to merge the curves and surfaces from two shells into one, and to split the trim curves into their piecewise linear segments and then reassemble them into trim curves. [git-p4: depot-paths = "//depot/solvespace/": change = 1905]
2009-01-25 11:52:29 +00:00
}
Add Newton iterations to intersect a line with a surface at a point, and to intersect three surfaces at a point. So now when we split an edge, we can refine the split point to lie exactly on the trim curve, so I can do certain Booleans on curved surfaces. But surface-line intersection is globally broken, since I don't correctly detect the number of intersections or provide a good first guess. I maybe should test by bounding boxes and subdivision. [git-p4: depot-paths = "//depot/solvespace/": change = 1920]
2009-02-27 13:04:36 +00:00
//-----------------------------------------------------------------------------
// Take our original pwl curve. Wherever an edge intersects a surface within
// either agnstA or agnstB, split the piecewise linear element. Then refine
// the intersection so that it lies on all three relevant surfaces: the
// intersecting surface, srfA, and srfB. (So the pwl curve should lie at
// the intersection of srfA and srfB.) Return a new pwl curve with everything
// split.
//-----------------------------------------------------------------------------
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
static Vector LineStart, LineDirection;
static int ByTAlongLine(const void *av, const void *bv)
{
Fix stupidity in Point2d::DistanceToLine, and classify line segments in Boolean against the shell, not the intersection polygon. (We just cast a ray, and use the surface-line intersection function that already existed.) That's slow, but can be accelerated later. [git-p4: depot-paths = "//depot/solvespace/": change = 1911]
2009-02-01 13:01:28 +00:00
SInter *a = (SInter *)av,
*b = (SInter *)bv;
Trim trailing whitespace.
2015-03-29 00:30:52 +00:00
Fix stupidity in Point2d::DistanceToLine, and classify line segments in Boolean against the shell, not the intersection polygon. (We just cast a ray, and use the surface-line intersection function that already existed.) That's slow, but can be accelerated later. [git-p4: depot-paths = "//depot/solvespace/": change = 1911]
2009-02-01 13:01:28 +00:00
double ta = (a->p.Minus(LineStart)).DivPivoting(LineDirection),
tb = (b->p.Minus(LineStart)).DivPivoting(LineDirection);
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
return (ta > tb) ? 1 : -1;
}
Add Newton iterations to intersect a line with a surface at a point, and to intersect three surfaces at a point. So now when we split an edge, we can refine the split point to lie exactly on the trim curve, so I can do certain Booleans on curved surfaces. But surface-line intersection is globally broken, since I don't correctly detect the number of intersections or provide a good first guess. I maybe should test by bounding boxes and subdivision. [git-p4: depot-paths = "//depot/solvespace/": change = 1920]
2009-02-27 13:04:36 +00:00
SCurve SCurve::MakeCopySplitAgainst(SShell *agnstA, SShell *agnstB,
Add const qualifiers to functions where trivially possible. This will allow us in future to accept `const T &` anywhere it's necessary to reduce the amount of copying. This commit is quite conservative: it does not attempt very hard to refactor code that performs incidental mutation. In particular dogd and caches are not marked with the `mutable` keyword. dogd will be eliminated later, opening up more opportunities to add const qualifiers. This commit also doesn't introduce any uses of the newly added const qualifers. This will be done later.
2016-05-21 05:18:00 +00:00
SSurface *srfA, SSurface *srfB) const
Add Newton iterations to intersect a line with a surface at a point, and to intersect three surfaces at a point. So now when we split an edge, we can refine the split point to lie exactly on the trim curve, so I can do certain Booleans on curved surfaces. But surface-line intersection is globally broken, since I don't correctly detect the number of intersections or provide a good first guess. I maybe should test by bounding boxes and subdivision. [git-p4: depot-paths = "//depot/solvespace/": change = 1920]
2009-02-27 13:04:36 +00:00
{
Preparatory work for Boolean. Make the u and v coordinates of the trim curves for all surfaces lie between 0 and 1. And add routines to merge the curves and surfaces from two shells into one, and to split the trim curves into their piecewise linear segments and then reassemble them into trim curves. [git-p4: depot-paths = "//depot/solvespace/": change = 1905]
2009-01-25 11:52:29 +00:00
SCurve ret;
More poking at Booleans; generate the unsplit intersection curves for planes against planes. [git-p4: depot-paths = "//depot/solvespace/": change = 1907]
2009-01-27 07:59:58 +00:00
ret = *this;
Replace all ZERO and memset with C++11 brace-initialization. This will allow us to use non-POD classes inside these objects in future and is otherwise functionally equivalent, as well as more concise. Note that there are some subtleties with handling of brace-initialization. Specifically: On aggregates (e.g. simple C-style structures) using an empty brace-initializer zero-initializes the aggregate, i.e. it makes all members zero. On non-aggregates an empty brace-initializer calls the default constructor. And if the constructor doesn't explicitly initialize the members (which the auto-generated constructor doesn't) then the members will be constructed but otherwise uninitialized. So, what is an aggregate class? To quote the C++ standard (C++03 8.5.1 §1): An aggregate is an array or a class (clause 9) with no user-declared constructors (12.1), no private or protected non-static data members (clause 11), no base classes (clause 10), and no virtual functions (10.3). In SolveSpace, we only have to handle the case of base classes; Constraint and Entity have those. Thus, they had to gain a default constructor that does nothing but initializes the members to zero.
2015-03-27 15:31:23 +00:00
ret.pts = {};
Preparatory work for Boolean. Make the u and v coordinates of the trim curves for all surfaces lie between 0 and 1. And add routines to merge the curves and surfaces from two shells into one, and to split the trim curves into their piecewise linear segments and then reassemble them into trim curves. [git-p4: depot-paths = "//depot/solvespace/": change = 1905]
2009-01-25 11:52:29 +00:00
Add const qualifiers to functions where trivially possible. This will allow us in future to accept `const T &` anywhere it's necessary to reduce the amount of copying. This commit is quite conservative: it does not attempt very hard to refactor code that performs incidental mutation. In particular dogd and caches are not marked with the `mutable` keyword. dogd will be eliminated later, opening up more opportunities to add const qualifiers. This commit also doesn't introduce any uses of the newly added const qualifers. This will be done later.
2016-05-21 05:18:00 +00:00
const SCurvePt *p = pts.First();
Replace all oops() checks with ssassert()s. This includes explanation and context for non-obvious cases and shortens debug cycles when just-in-time debugging is not available (like on Linux) by immediately printing description of the assert as well as symbolized backtrace.
2016-05-18 22:51:36 +00:00
ssassert(p != NULL, "Cannot split an empty curve");
Generate intersection curves for surfaces of extrusion along a parallel axis (which are always lines parallel to that axis). Remove short pwl segments when possible, to avoid short edges that get misclassified. [git-p4: depot-paths = "//depot/solvespace/": change = 1952]
2009-05-18 07:26:51 +00:00
SCurvePt prev = *p;
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
ret.pts.Add(p);
p = pts.NextAfter(p);
for(; p; p = pts.NextAfter(p)) {
Replace all ZERO and memset with C++11 brace-initialization. This will allow us to use non-POD classes inside these objects in future and is otherwise functionally equivalent, as well as more concise. Note that there are some subtleties with handling of brace-initialization. Specifically: On aggregates (e.g. simple C-style structures) using an empty brace-initializer zero-initializes the aggregate, i.e. it makes all members zero. On non-aggregates an empty brace-initializer calls the default constructor. And if the constructor doesn't explicitly initialize the members (which the auto-generated constructor doesn't) then the members will be constructed but otherwise uninitialized. So, what is an aggregate class? To quote the C++ standard (C++03 8.5.1 §1): An aggregate is an array or a class (clause 9) with no user-declared constructors (12.1), no private or protected non-static data members (clause 11), no base classes (clause 10), and no virtual functions (10.3). In SolveSpace, we only have to handle the case of base classes; Constraint and Entity have those. Thus, they had to gain a default constructor that does nothing but initializes the members to zero.
2015-03-27 15:31:23 +00:00
List<SInter> il = {};
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
// Find all the intersections with the two passed shells
Generate intersection curves for surfaces of extrusion along a parallel axis (which are always lines parallel to that axis). Remove short pwl segments when possible, to avoid short edges that get misclassified. [git-p4: depot-paths = "//depot/solvespace/": change = 1952]
2009-05-18 07:26:51 +00:00
if(agnstA)
Discard intersection curves that lie entirely outside of one surface's domain of u, v in [0, 1]. Cache the starting guess when projecting a point into a ratpoly surface, to avoid brute force searching for a good one every time. Split edges even if they aren't quite inside the trim curve, since the trim boundaries are pwl, not exact; unnecessary splits won't hurt, but failure to split when necessary will. Make the triangulation code use a better (but not perfect) epsilon, to avoid "can't find ear" failures on very fine meshes. And turn on compiler optimization! I had somehow forgotten about that, and it's a ~2x improvement. [git-p4: depot-paths = "//depot/solvespace/": change = 2026]
2009-08-21 04:58:28 +00:00
agnstA->AllPointsIntersecting(prev.p, p->p, &il, true, false, true);
Generate intersection curves for surfaces of extrusion along a parallel axis (which are always lines parallel to that axis). Remove short pwl segments when possible, to avoid short edges that get misclassified. [git-p4: depot-paths = "//depot/solvespace/": change = 1952]
2009-05-18 07:26:51 +00:00
if(agnstB)
Discard intersection curves that lie entirely outside of one surface's domain of u, v in [0, 1]. Cache the starting guess when projecting a point into a ratpoly surface, to avoid brute force searching for a good one every time. Split edges even if they aren't quite inside the trim curve, since the trim boundaries are pwl, not exact; unnecessary splits won't hurt, but failure to split when necessary will. Make the triangulation code use a better (but not perfect) epsilon, to avoid "can't find ear" failures on very fine meshes. And turn on compiler optimization! I had somehow forgotten about that, and it's a ~2x improvement. [git-p4: depot-paths = "//depot/solvespace/": change = 2026]
2009-08-21 04:58:28 +00:00
agnstB->AllPointsIntersecting(prev.p, p->p, &il, true, false, true);
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
if(il.n > 0) {
Add code to subdivide (with de Castljau's algorithm) a surface, and use that for surface-line intersections. That has major problems with the heuristic on when to stop and do Newton polishing. There's also an issue with all the Newton stuff when surfaces join tangent. And update the wishlist to reflect current needs. [git-p4: depot-paths = "//depot/solvespace/": change = 1925]
2009-03-08 10:59:57 +00:00
// The intersections were generated by intersecting the pwl
// edge against a surface; so they must be refined to lie
// exactly on the original curve.
When splitting a curve against surfaces, don't split the curve against the surfaces that it supposedly borders; that will cause numerical trouble. [git-p4: depot-paths = "//depot/solvespace/": change = 1993]
2009-06-21 09:54:21 +00:00
il.ClearTags();
Fix stupidity in Point2d::DistanceToLine, and classify line segments in Boolean against the shell, not the intersection polygon. (We just cast a ray, and use the surface-line intersection function that already existed.) That's slow, but can be accelerated later. [git-p4: depot-paths = "//depot/solvespace/": change = 1911]
2009-02-01 13:01:28 +00:00
SInter *pi;
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
for(pi = il.First(); pi; pi = il.NextAfter(pi)) {
When splitting a curve against surfaces, don't split the curve against the surfaces that it supposedly borders; that will cause numerical trouble. [git-p4: depot-paths = "//depot/solvespace/": change = 1993]
2009-06-21 09:54:21 +00:00
if(pi->srf == srfA || pi->srf == srfB) {
// The edge certainly intersects the surfaces that it
// trims (at its endpoints), but those ones don't count.
// They are culled later, but no sense calculating them
// and they will cause numerical problems (since two
// of the three surfaces they're refined to lie on will
// be identical, so the matrix will be singular).
pi->tag = 1;
continue;
}
Discard intersection curves that lie entirely outside of one surface's domain of u, v in [0, 1]. Cache the starting guess when projecting a point into a ratpoly surface, to avoid brute force searching for a good one every time. Split edges even if they aren't quite inside the trim curve, since the trim boundaries are pwl, not exact; unnecessary splits won't hurt, but failure to split when necessary will. Make the triangulation code use a better (but not perfect) epsilon, to avoid "can't find ear" failures on very fine meshes. And turn on compiler optimization! I had somehow forgotten about that, and it's a ~2x improvement. [git-p4: depot-paths = "//depot/solvespace/": change = 2026]
2009-08-21 04:58:28 +00:00
Point2d puv;
(pi->srf)->ClosestPointTo(pi->p, &puv, false);
// Split the edge if the intersection lies within the surface's
// trim curves, or within the chord tol of the trim curve; want
// some slop if points are close to edge and pwl is too coarse,
// and it doesn't hurt to split unnecessarily.
Point2d dummy = { 0, 0 };
Convert all enumerations to use `enum class`. Specifically, take the old code that looks like this: class Foo { enum { X = 1, Y = 2 }; int kind; } ... foo.kind = Foo::X; ... and convert it to this: class Foo { enum class Kind : uint32_t { X = 1, Y = 2 }; Kind kind; } ... foo.kind = Foo::Kind::X; (In some cases the enumeration would not be in the class namespace, such as when it is generally useful.) The benefits are as follows: * The type of the field gives a clear indication of intent, both to humans and tools (such as binding generators). * The compiler is able to automatically warn when a switch is not exhaustive; but this is currently suppressed by the default: ssassert(false, ...) idiom. * Integers and plain enums are weakly type checked: they implicitly convert into each other. This can hide bugs where type conversion is performed but not intended. Enum classes are strongly type checked. * Plain enums pollute parent namespaces; enum classes do not. Almost every defined enum we have already has a kind of ad-hoc namespacing via `NAMESPACE_`, which is now explicit. * Plain enums do not have a well-defined ABI size, which is important for bindings. Enum classes can have it, if specified. We specify the base type for all enums as uint32_t, which is a safe choice and allows us to not change the numeric values of any variants. This commit introduces absolutely no functional change to the code, just renaming and change of types. It handles almost all cases, except GraphicsWindow::pending.operation, which needs minor functional change.
2016-05-20 08:31:20 +00:00
SBspUv::Class c = (pi->srf->bsp) ? pi->srf->bsp->ClassifyPoint(puv, dummy, pi->srf) : SBspUv::Class::OUTSIDE;
if(c == SBspUv::Class::OUTSIDE) {
Get rid of if(!this). According to the C++ standard, "this" is never NULL, so checks of the form "if(!this)" can be legally optimized out. This breaks SolveSpace on GCC 6, and probably on other compilers and configurations.
2016-03-25 14:16:58 +00:00
double d = VERY_POSITIVE;
if(pi->srf->bsp) d = pi->srf->bsp->MinimumDistanceToEdge(puv, pi->srf);
Discard intersection curves that lie entirely outside of one surface's domain of u, v in [0, 1]. Cache the starting guess when projecting a point into a ratpoly surface, to avoid brute force searching for a good one every time. Split edges even if they aren't quite inside the trim curve, since the trim boundaries are pwl, not exact; unnecessary splits won't hurt, but failure to split when necessary will. Make the triangulation code use a better (but not perfect) epsilon, to avoid "can't find ear" failures on very fine meshes. And turn on compiler optimization! I had somehow forgotten about that, and it's a ~2x improvement. [git-p4: depot-paths = "//depot/solvespace/": change = 2026]
2009-08-21 04:58:28 +00:00
if(d > SS.ChordTolMm()) {
pi->tag = 1;
continue;
}
}
// We're keeping the intersection, so actually refine it.
(pi->srf)->PointOnSurfaces(srfA, srfB, &(puv.x), &(puv.y));
pi->p = (pi->srf)->PointAt(puv);
Add code to subdivide (with de Castljau's algorithm) a surface, and use that for surface-line intersections. That has major problems with the heuristic on when to stop and do Newton polishing. There's also an issue with all the Newton stuff when surfaces join tangent. And update the wishlist to reflect current needs. [git-p4: depot-paths = "//depot/solvespace/": change = 1925]
2009-03-08 10:59:57 +00:00
}
When splitting a curve against surfaces, don't split the curve against the surfaces that it supposedly borders; that will cause numerical trouble. [git-p4: depot-paths = "//depot/solvespace/": change = 1993]
2009-06-21 09:54:21 +00:00
il.RemoveTagged();
Add Newton iterations to intersect a line with a surface at a point, and to intersect three surfaces at a point. So now when we split an edge, we can refine the split point to lie exactly on the trim curve, so I can do certain Booleans on curved surfaces. But surface-line intersection is globally broken, since I don't correctly detect the number of intersections or provide a good first guess. I maybe should test by bounding boxes and subdivision. [git-p4: depot-paths = "//depot/solvespace/": change = 1920]
2009-02-27 13:04:36 +00:00
Add code to subdivide (with de Castljau's algorithm) a surface, and use that for surface-line intersections. That has major problems with the heuristic on when to stop and do Newton polishing. There's also an issue with all the Newton stuff when surfaces join tangent. And update the wishlist to reflect current needs. [git-p4: depot-paths = "//depot/solvespace/": change = 1925]
2009-03-08 10:59:57 +00:00
// And now sort them in order along the line. Note that we must
// do that after refining, in case the refining would make two
// points switch places.
Generate intersection curves for surfaces of extrusion along a parallel axis (which are always lines parallel to that axis). Remove short pwl segments when possible, to avoid short edges that get misclassified. [git-p4: depot-paths = "//depot/solvespace/": change = 1952]
2009-05-18 07:26:51 +00:00
LineStart = prev.p;
LineDirection = (p->p).Minus(prev.p);
Add code to subdivide (with de Castljau's algorithm) a surface, and use that for surface-line intersections. That has major problems with the heuristic on when to stop and do Newton polishing. There's also an issue with all the Newton stuff when surfaces join tangent. And update the wishlist to reflect current needs. [git-p4: depot-paths = "//depot/solvespace/": change = 1925]
2009-03-08 10:59:57 +00:00
qsort(il.elem, il.n, sizeof(il.elem[0]), ByTAlongLine);
Trim trailing whitespace.
2015-03-29 00:30:52 +00:00
Add code to subdivide (with de Castljau's algorithm) a surface, and use that for surface-line intersections. That has major problems with the heuristic on when to stop and do Newton polishing. There's also an issue with all the Newton stuff when surfaces join tangent. And update the wishlist to reflect current needs. [git-p4: depot-paths = "//depot/solvespace/": change = 1925]
2009-03-08 10:59:57 +00:00
// And now uses the intersections to generate our split pwl edge(s)
Vector prev = Vector::From(VERY_POSITIVE, 0, 0);
for(pi = il.First(); pi; pi = il.NextAfter(pi)) {
Make Boolean union work when the shells have coincident plane faces. Still on planes only, no curved surface intersections. [git-p4: depot-paths = "//depot/solvespace/": change = 1912]
2009-02-09 12:40:48 +00:00
// On-edge intersection will generate same split point for
// both surfaces, so don't create zero-length edge.
if(!prev.Equals(pi->p)) {
Generate intersection curves for surfaces of extrusion along a parallel axis (which are always lines parallel to that axis). Remove short pwl segments when possible, to avoid short edges that get misclassified. [git-p4: depot-paths = "//depot/solvespace/": change = 1952]
2009-05-18 07:26:51 +00:00
SCurvePt scpt;
scpt.tag = 0;
scpt.p = pi->p;
scpt.vertex = true;
ret.pts.Add(&scpt);
Make Boolean union work when the shells have coincident plane faces. Still on planes only, no curved surface intersections. [git-p4: depot-paths = "//depot/solvespace/": change = 1912]
2009-02-09 12:40:48 +00:00
}
prev = pi->p;
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
}
}
Fix stupidity in Point2d::DistanceToLine, and classify line segments in Boolean against the shell, not the intersection polygon. (We just cast a ray, and use the surface-line intersection function that already existed.) That's slow, but can be accelerated later. [git-p4: depot-paths = "//depot/solvespace/": change = 1911]
2009-02-01 13:01:28 +00:00
il.Clear();
Preparatory work for Boolean. Make the u and v coordinates of the trim curves for all surfaces lie between 0 and 1. And add routines to merge the curves and surfaces from two shells into one, and to split the trim curves into their piecewise linear segments and then reassemble them into trim curves. [git-p4: depot-paths = "//depot/solvespace/": change = 1905]
2009-01-25 11:52:29 +00:00
ret.pts.Add(p);
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
prev = *p;
Preparatory work for Boolean. Make the u and v coordinates of the trim curves for all surfaces lie between 0 and 1. And add routines to merge the curves and surfaces from two shells into one, and to split the trim curves into their piecewise linear segments and then reassemble them into trim curves. [git-p4: depot-paths = "//depot/solvespace/": change = 1905]
2009-01-25 11:52:29 +00:00
}
return ret;
}
Trim trailing whitespace.
2015-03-29 00:30:52 +00:00
void SShell::CopyCurvesSplitAgainst(bool opA, SShell *agnst, SShell *into) {
Preparatory work for Boolean. Make the u and v coordinates of the trim curves for all surfaces lie between 0 and 1. And add routines to merge the curves and surfaces from two shells into one, and to split the trim curves into their piecewise linear segments and then reassemble them into trim curves. [git-p4: depot-paths = "//depot/solvespace/": change = 1905]
2009-01-25 11:52:29 +00:00
SCurve *sc;
for(sc = curve.First(); sc; sc = curve.NextAfter(sc)) {
Add Newton iterations to intersect a line with a surface at a point, and to intersect three surfaces at a point. So now when we split an edge, we can refine the split point to lie exactly on the trim curve, so I can do certain Booleans on curved surfaces. But surface-line intersection is globally broken, since I don't correctly detect the number of intersections or provide a good first guess. I maybe should test by bounding boxes and subdivision. [git-p4: depot-paths = "//depot/solvespace/": change = 1920]
2009-02-27 13:04:36 +00:00
SCurve scn = sc->MakeCopySplitAgainst(agnst, NULL,
surface.FindById(sc->surfA),
surface.FindById(sc->surfB));
Convert all enumerations to use `enum class`. Specifically, take the old code that looks like this: class Foo { enum { X = 1, Y = 2 }; int kind; } ... foo.kind = Foo::X; ... and convert it to this: class Foo { enum class Kind : uint32_t { X = 1, Y = 2 }; Kind kind; } ... foo.kind = Foo::Kind::X; (In some cases the enumeration would not be in the class namespace, such as when it is generally useful.) The benefits are as follows: * The type of the field gives a clear indication of intent, both to humans and tools (such as binding generators). * The compiler is able to automatically warn when a switch is not exhaustive; but this is currently suppressed by the default: ssassert(false, ...) idiom. * Integers and plain enums are weakly type checked: they implicitly convert into each other. This can hide bugs where type conversion is performed but not intended. Enum classes are strongly type checked. * Plain enums pollute parent namespaces; enum classes do not. Almost every defined enum we have already has a kind of ad-hoc namespacing via `NAMESPACE_`, which is now explicit. * Plain enums do not have a well-defined ABI size, which is important for bindings. Enum classes can have it, if specified. We specify the base type for all enums as uint32_t, which is a safe choice and allows us to not change the numeric values of any variants. This commit introduces absolutely no functional change to the code, just renaming and change of types. It handles almost all cases, except GraphicsWindow::pending.operation, which needs minor functional change.
2016-05-20 08:31:20 +00:00
scn.source = opA ? SCurve::Source::A : SCurve::Source::B;
Add Newton iterations to intersect a line with a surface at a point, and to intersect three surfaces at a point. So now when we split an edge, we can refine the split point to lie exactly on the trim curve, so I can do certain Booleans on curved surfaces. But surface-line intersection is globally broken, since I don't correctly detect the number of intersections or provide a good first guess. I maybe should test by bounding boxes and subdivision. [git-p4: depot-paths = "//depot/solvespace/": change = 1920]
2009-02-27 13:04:36 +00:00
Preparatory work for Boolean. Make the u and v coordinates of the trim curves for all surfaces lie between 0 and 1. And add routines to merge the curves and surfaces from two shells into one, and to split the trim curves into their piecewise linear segments and then reassemble them into trim curves. [git-p4: depot-paths = "//depot/solvespace/": change = 1905]
2009-01-25 11:52:29 +00:00
hSCurve hsc = into->curve.AddAndAssignId(&scn);
// And note the new ID so that we can rewrite the trims appropriately
sc->newH = hsc;
}
}
Add code to identify planes and cylindrical surfaces from a solid of revolution, and put them in the same form as if they had been draw by an extrusion (so that we can use all the same special case intersection curves). And add code to merge coincident faces into one. That turns out to be more than a cosmetic/efficiency thing, since edge splitting fails at the join between two coincident faces. [git-p4: depot-paths = "//depot/solvespace/": change = 1965]
2009-06-05 05:38:41 +00:00
void SSurface::TrimFromEdgeList(SEdgeList *el, bool asUv) {
Preparatory work for Boolean. Make the u and v coordinates of the trim curves for all surfaces lie between 0 and 1. And add routines to merge the curves and surfaces from two shells into one, and to split the trim curves into their piecewise linear segments and then reassemble them into trim curves. [git-p4: depot-paths = "//depot/solvespace/": change = 1905]
2009-01-25 11:52:29 +00:00
el->l.ClearTags();
Trim trailing whitespace.
2015-03-29 00:30:52 +00:00
Replace all ZERO and memset with C++11 brace-initialization. This will allow us to use non-POD classes inside these objects in future and is otherwise functionally equivalent, as well as more concise. Note that there are some subtleties with handling of brace-initialization. Specifically: On aggregates (e.g. simple C-style structures) using an empty brace-initializer zero-initializes the aggregate, i.e. it makes all members zero. On non-aggregates an empty brace-initializer calls the default constructor. And if the constructor doesn't explicitly initialize the members (which the auto-generated constructor doesn't) then the members will be constructed but otherwise uninitialized. So, what is an aggregate class? To quote the C++ standard (C++03 8.5.1 §1): An aggregate is an array or a class (clause 9) with no user-declared constructors (12.1), no private or protected non-static data members (clause 11), no base classes (clause 10), and no virtual functions (10.3). In SolveSpace, we only have to handle the case of base classes; Constraint and Entity have those. Thus, they had to gain a default constructor that does nothing but initializes the members to zero.
2015-03-27 15:31:23 +00:00
STrimBy stb = {};
Preparatory work for Boolean. Make the u and v coordinates of the trim curves for all surfaces lie between 0 and 1. And add routines to merge the curves and surfaces from two shells into one, and to split the trim curves into their piecewise linear segments and then reassemble them into trim curves. [git-p4: depot-paths = "//depot/solvespace/": change = 1905]
2009-01-25 11:52:29 +00:00
for(;;) {
// Find an edge, any edge; we'll start from there.
SEdge *se;
for(se = el->l.First(); se; se = el->l.NextAfter(se)) {
if(se->tag) continue;
Trim trailing whitespace.
2015-03-29 00:30:52 +00:00
break;
Preparatory work for Boolean. Make the u and v coordinates of the trim curves for all surfaces lie between 0 and 1. And add routines to merge the curves and surfaces from two shells into one, and to split the trim curves into their piecewise linear segments and then reassemble them into trim curves. [git-p4: depot-paths = "//depot/solvespace/": change = 1905]
2009-01-25 11:52:29 +00:00
}
if(!se) break;
se->tag = 1;
stb.start = se->a;
stb.finish = se->b;
stb.curve.v = se->auxA;
stb.backwards = se->auxB ? true : false;
// Find adjoining edges from the same curve; those should be
// merged into a single trim.
bool merged;
do {
merged = false;
for(se = el->l.First(); se; se = el->l.NextAfter(se)) {
if(se->tag) continue;
Use casts to bridge mismatches in integer-type sizes and signedness The compiler gets nervous when we (for example) pass in a size_t as an int parameter, or assign an int to a char, or assign -1 to an unsigned type. By adding appropriate casts, we inform the compiler that, yes, we know what we're doing. This change also upgrades a va_arg() type from char to int, as char is always promoted to int when passed through '...'.
2013-08-26 20:09:15 +00:00
if(se->auxA != (int)stb.curve.v) continue;
Preparatory work for Boolean. Make the u and v coordinates of the trim curves for all surfaces lie between 0 and 1. And add routines to merge the curves and surfaces from two shells into one, and to split the trim curves into their piecewise linear segments and then reassemble them into trim curves. [git-p4: depot-paths = "//depot/solvespace/": change = 1905]
2009-01-25 11:52:29 +00:00
if(( se->auxB && !stb.backwards) ||
(!se->auxB && stb.backwards)) continue;
if((se->a).Equals(stb.finish)) {
stb.finish = se->b;
se->tag = 1;
merged = true;
} else if((se->b).Equals(stb.start)) {
stb.start = se->a;
se->tag = 1;
merged = true;
}
}
} while(merged);
Add code to identify planes and cylindrical surfaces from a solid of revolution, and put them in the same form as if they had been draw by an extrusion (so that we can use all the same special case intersection curves). And add code to merge coincident faces into one. That turns out to be more than a cosmetic/efficiency thing, since edge splitting fails at the join between two coincident faces. [git-p4: depot-paths = "//depot/solvespace/": change = 1965]
2009-06-05 05:38:41 +00:00
if(asUv) {
stb.start = PointAt(stb.start.x, stb.start.y);
stb.finish = PointAt(stb.finish.x, stb.finish.y);
}
Preparatory work for Boolean. Make the u and v coordinates of the trim curves for all surfaces lie between 0 and 1. And add routines to merge the curves and surfaces from two shells into one, and to split the trim curves into their piecewise linear segments and then reassemble them into trim curves. [git-p4: depot-paths = "//depot/solvespace/": change = 1905]
2009-01-25 11:52:29 +00:00
// And add the merged trim, with xyz (not uv like the polygon) pts
trim.Add(&stb);
}
}
Convert all enumerations to use `enum class`. Specifically, take the old code that looks like this: class Foo { enum { X = 1, Y = 2 }; int kind; } ... foo.kind = Foo::X; ... and convert it to this: class Foo { enum class Kind : uint32_t { X = 1, Y = 2 }; Kind kind; } ... foo.kind = Foo::Kind::X; (In some cases the enumeration would not be in the class namespace, such as when it is generally useful.) The benefits are as follows: * The type of the field gives a clear indication of intent, both to humans and tools (such as binding generators). * The compiler is able to automatically warn when a switch is not exhaustive; but this is currently suppressed by the default: ssassert(false, ...) idiom. * Integers and plain enums are weakly type checked: they implicitly convert into each other. This can hide bugs where type conversion is performed but not intended. Enum classes are strongly type checked. * Plain enums pollute parent namespaces; enum classes do not. Almost every defined enum we have already has a kind of ad-hoc namespacing via `NAMESPACE_`, which is now explicit. * Plain enums do not have a well-defined ABI size, which is important for bindings. Enum classes can have it, if specified. We specify the base type for all enums as uint32_t, which is a safe choice and allows us to not change the numeric values of any variants. This commit introduces absolutely no functional change to the code, just renaming and change of types. It handles almost all cases, except GraphicsWindow::pending.operation, which needs minor functional change.
2016-05-20 08:31:20 +00:00
static bool KeepRegion(SSurface::CombineAs type, bool opA, SShell::Class shell, SShell::Class orig)
Additional poking at Booleans. At least this is a halfway rational way to think about the cases; I'm classifying the regions to the left and right of each edge, and keeping the edges if those regions (2d, surfaces) classify different. Still screws up with edge-on-edge intersections; but if I make the surface intersection stuff handle that, then might be more straightforward to use that info. [git-p4: depot-paths = "//depot/solvespace/": change = 1914]
2009-02-16 12:05:08 +00:00
{
Convert all enumerations to use `enum class`. Specifically, take the old code that looks like this: class Foo { enum { X = 1, Y = 2 }; int kind; } ... foo.kind = Foo::X; ... and convert it to this: class Foo { enum class Kind : uint32_t { X = 1, Y = 2 }; Kind kind; } ... foo.kind = Foo::Kind::X; (In some cases the enumeration would not be in the class namespace, such as when it is generally useful.) The benefits are as follows: * The type of the field gives a clear indication of intent, both to humans and tools (such as binding generators). * The compiler is able to automatically warn when a switch is not exhaustive; but this is currently suppressed by the default: ssassert(false, ...) idiom. * Integers and plain enums are weakly type checked: they implicitly convert into each other. This can hide bugs where type conversion is performed but not intended. Enum classes are strongly type checked. * Plain enums pollute parent namespaces; enum classes do not. Almost every defined enum we have already has a kind of ad-hoc namespacing via `NAMESPACE_`, which is now explicit. * Plain enums do not have a well-defined ABI size, which is important for bindings. Enum classes can have it, if specified. We specify the base type for all enums as uint32_t, which is a safe choice and allows us to not change the numeric values of any variants. This commit introduces absolutely no functional change to the code, just renaming and change of types. It handles almost all cases, except GraphicsWindow::pending.operation, which needs minor functional change.
2016-05-20 08:31:20 +00:00
bool inShell = (shell == SShell::Class::INSIDE),
inSame = (shell == SShell::Class::COINC_SAME),
inOpp = (shell == SShell::Class::COINC_OPP),
inOrig = (orig == SShell::Class::INSIDE);
Additional poking at Booleans. At least this is a halfway rational way to think about the cases; I'm classifying the regions to the left and right of each edge, and keeping the edges if those regions (2d, surfaces) classify different. Still screws up with edge-on-edge intersections; but if I make the surface intersection stuff handle that, then might be more straightforward to use that info. [git-p4: depot-paths = "//depot/solvespace/": change = 1914]
2009-02-16 12:05:08 +00:00
bool inFace = inSame || inOpp;
// If these are correct, then they should be independent of inShell
// if inFace is true.
if(!inOrig) return false;
switch(type) {
Convert all enumerations to use `enum class`. Specifically, take the old code that looks like this: class Foo { enum { X = 1, Y = 2 }; int kind; } ... foo.kind = Foo::X; ... and convert it to this: class Foo { enum class Kind : uint32_t { X = 1, Y = 2 }; Kind kind; } ... foo.kind = Foo::Kind::X; (In some cases the enumeration would not be in the class namespace, such as when it is generally useful.) The benefits are as follows: * The type of the field gives a clear indication of intent, both to humans and tools (such as binding generators). * The compiler is able to automatically warn when a switch is not exhaustive; but this is currently suppressed by the default: ssassert(false, ...) idiom. * Integers and plain enums are weakly type checked: they implicitly convert into each other. This can hide bugs where type conversion is performed but not intended. Enum classes are strongly type checked. * Plain enums pollute parent namespaces; enum classes do not. Almost every defined enum we have already has a kind of ad-hoc namespacing via `NAMESPACE_`, which is now explicit. * Plain enums do not have a well-defined ABI size, which is important for bindings. Enum classes can have it, if specified. We specify the base type for all enums as uint32_t, which is a safe choice and allows us to not change the numeric values of any variants. This commit introduces absolutely no functional change to the code, just renaming and change of types. It handles almost all cases, except GraphicsWindow::pending.operation, which needs minor functional change.
2016-05-20 08:31:20 +00:00
case SSurface::CombineAs::UNION:
Additional poking at Booleans. At least this is a halfway rational way to think about the cases; I'm classifying the regions to the left and right of each edge, and keeping the edges if those regions (2d, surfaces) classify different. Still screws up with edge-on-edge intersections; but if I make the surface intersection stuff handle that, then might be more straightforward to use that info. [git-p4: depot-paths = "//depot/solvespace/": change = 1914]
2009-02-16 12:05:08 +00:00
if(opA) {
return (!inShell && !inFace);
} else {
return (!inShell && !inFace) || inSame;
}
Convert all enumerations to use `enum class`. Specifically, take the old code that looks like this: class Foo { enum { X = 1, Y = 2 }; int kind; } ... foo.kind = Foo::X; ... and convert it to this: class Foo { enum class Kind : uint32_t { X = 1, Y = 2 }; Kind kind; } ... foo.kind = Foo::Kind::X; (In some cases the enumeration would not be in the class namespace, such as when it is generally useful.) The benefits are as follows: * The type of the field gives a clear indication of intent, both to humans and tools (such as binding generators). * The compiler is able to automatically warn when a switch is not exhaustive; but this is currently suppressed by the default: ssassert(false, ...) idiom. * Integers and plain enums are weakly type checked: they implicitly convert into each other. This can hide bugs where type conversion is performed but not intended. Enum classes are strongly type checked. * Plain enums pollute parent namespaces; enum classes do not. Almost every defined enum we have already has a kind of ad-hoc namespacing via `NAMESPACE_`, which is now explicit. * Plain enums do not have a well-defined ABI size, which is important for bindings. Enum classes can have it, if specified. We specify the base type for all enums as uint32_t, which is a safe choice and allows us to not change the numeric values of any variants. This commit introduces absolutely no functional change to the code, just renaming and change of types. It handles almost all cases, except GraphicsWindow::pending.operation, which needs minor functional change.
2016-05-20 08:31:20 +00:00
case SSurface::CombineAs::DIFFERENCE:
Additional poking at Booleans. At least this is a halfway rational way to think about the cases; I'm classifying the regions to the left and right of each edge, and keeping the edges if those regions (2d, surfaces) classify different. Still screws up with edge-on-edge intersections; but if I make the surface intersection stuff handle that, then might be more straightforward to use that info. [git-p4: depot-paths = "//depot/solvespace/": change = 1914]
2009-02-16 12:05:08 +00:00
if(opA) {
return (!inShell && !inFace);
} else {
return (inShell && !inFace) || inSame;
}
Enable exhaustive switch coverage warnings as an error, and use them. Specifically, this enables -Wswitch=error on GCC/Clang and its MSVC equivalent; the exact way it is handled varies slightly, but what they all have in common is that in a switch statement over an enumeration, any enumerand that is not explicitly (via case:) or implicitly (via default:) handled in the switch triggers an error. Moreover, we also change the switch statements in three ways: * Switch statements that ought to be extended every time a new enumerand is added (e.g. Entity::DrawOrGetDistance(), are changed to explicitly list every single enumerand, and not have a default: branch. Note that the assertions are kept because it is legal for a enumeration to have a value unlike any of its defined enumerands, and we can e.g. read garbage from a file, or an uninitialized variable. This requires some rearranging if a default: branch is undesired. * Switch statements that ought to only ever see a few select enumerands, are changed to always assert in the default: branch. * Switch statements that do something meaningful for a few enumerands, and ignore everything else, are changed to do nothing in a default: branch, under the assumption that changing them every time an enumerand is added or removed would just result in noise and catch no bugs. This commit also removes the {Request,Entity,Constraint}::UNKNOWN and Entity::DATUM_POINT enumerands, as those were just fancy names for zeroes. They mess up switch exhaustiveness checks and most of the time were not the best way to implement what they did anyway.
2016-05-25 06:55:50 +00:00
default: ssassert(false, "Unexpected combine type");
Additional poking at Booleans. At least this is a halfway rational way to think about the cases; I'm classifying the regions to the left and right of each edge, and keeping the edges if those regions (2d, surfaces) classify different. Still screws up with edge-on-edge intersections; but if I make the surface intersection stuff handle that, then might be more straightforward to use that info. [git-p4: depot-paths = "//depot/solvespace/": change = 1914]
2009-02-16 12:05:08 +00:00
}
}
Convert all enumerations to use `enum class`. Specifically, take the old code that looks like this: class Foo { enum { X = 1, Y = 2 }; int kind; } ... foo.kind = Foo::X; ... and convert it to this: class Foo { enum class Kind : uint32_t { X = 1, Y = 2 }; Kind kind; } ... foo.kind = Foo::Kind::X; (In some cases the enumeration would not be in the class namespace, such as when it is generally useful.) The benefits are as follows: * The type of the field gives a clear indication of intent, both to humans and tools (such as binding generators). * The compiler is able to automatically warn when a switch is not exhaustive; but this is currently suppressed by the default: ssassert(false, ...) idiom. * Integers and plain enums are weakly type checked: they implicitly convert into each other. This can hide bugs where type conversion is performed but not intended. Enum classes are strongly type checked. * Plain enums pollute parent namespaces; enum classes do not. Almost every defined enum we have already has a kind of ad-hoc namespacing via `NAMESPACE_`, which is now explicit. * Plain enums do not have a well-defined ABI size, which is important for bindings. Enum classes can have it, if specified. We specify the base type for all enums as uint32_t, which is a safe choice and allows us to not change the numeric values of any variants. This commit introduces absolutely no functional change to the code, just renaming and change of types. It handles almost all cases, except GraphicsWindow::pending.operation, which needs minor functional change.
2016-05-20 08:31:20 +00:00
static bool KeepEdge(SSurface::CombineAs type, bool opA,
SShell::Class indir_shell, SShell::Class outdir_shell,
SShell::Class indir_orig, SShell::Class outdir_orig)
Additional poking at Booleans. At least this is a halfway rational way to think about the cases; I'm classifying the regions to the left and right of each edge, and keeping the edges if those regions (2d, surfaces) classify different. Still screws up with edge-on-edge intersections; but if I make the surface intersection stuff handle that, then might be more straightforward to use that info. [git-p4: depot-paths = "//depot/solvespace/": change = 1914]
2009-02-16 12:05:08 +00:00
{
Revamp the edge classification for Booleans. I no longer make a separate polygon of coincident (with same or opposite normal) faces; I instead test all the edges against the other shell, and have extended the classify-against-shell stuff to handle those cases. And the normals are now perturbed a bit numerically, to either side of the edge, to distinguish tangency from a coincident surface. This seems to work fairly well, although things still tend to fail when the piecewise linear tolerance is too coarse. [git-p4: depot-paths = "//depot/solvespace/": change = 1964]
2009-06-04 03:59:40 +00:00
bool keepIn = KeepRegion(type, opA, indir_shell, indir_orig),
keepOut = KeepRegion(type, opA, outdir_shell, outdir_orig);
Additional poking at Booleans. At least this is a halfway rational way to think about the cases; I'm classifying the regions to the left and right of each edge, and keeping the edges if those regions (2d, surfaces) classify different. Still screws up with edge-on-edge intersections; but if I make the surface intersection stuff handle that, then might be more straightforward to use that info. [git-p4: depot-paths = "//depot/solvespace/": change = 1914]
2009-02-16 12:05:08 +00:00
// If the regions to the left and right of this edge are both in or both
// out, then this edge is not useful and should be discarded.
if(keepIn && !keepOut) return true;
return false;
}
Convert all enumerations to use `enum class`. Specifically, take the old code that looks like this: class Foo { enum { X = 1, Y = 2 }; int kind; } ... foo.kind = Foo::X; ... and convert it to this: class Foo { enum class Kind : uint32_t { X = 1, Y = 2 }; Kind kind; } ... foo.kind = Foo::Kind::X; (In some cases the enumeration would not be in the class namespace, such as when it is generally useful.) The benefits are as follows: * The type of the field gives a clear indication of intent, both to humans and tools (such as binding generators). * The compiler is able to automatically warn when a switch is not exhaustive; but this is currently suppressed by the default: ssassert(false, ...) idiom. * Integers and plain enums are weakly type checked: they implicitly convert into each other. This can hide bugs where type conversion is performed but not intended. Enum classes are strongly type checked. * Plain enums pollute parent namespaces; enum classes do not. Almost every defined enum we have already has a kind of ad-hoc namespacing via `NAMESPACE_`, which is now explicit. * Plain enums do not have a well-defined ABI size, which is important for bindings. Enum classes can have it, if specified. We specify the base type for all enums as uint32_t, which is a safe choice and allows us to not change the numeric values of any variants. This commit introduces absolutely no functional change to the code, just renaming and change of types. It handles almost all cases, except GraphicsWindow::pending.operation, which needs minor functional change.
2016-05-20 08:31:20 +00:00
static void TagByClassifiedEdge(SBspUv::Class bspclass, SShell::Class *indir, SShell::Class *outdir)
Additional poking at Booleans. At least this is a halfway rational way to think about the cases; I'm classifying the regions to the left and right of each edge, and keeping the edges if those regions (2d, surfaces) classify different. Still screws up with edge-on-edge intersections; but if I make the surface intersection stuff handle that, then might be more straightforward to use that info. [git-p4: depot-paths = "//depot/solvespace/": change = 1914]
2009-02-16 12:05:08 +00:00
{
switch(bspclass) {
Convert all enumerations to use `enum class`. Specifically, take the old code that looks like this: class Foo { enum { X = 1, Y = 2 }; int kind; } ... foo.kind = Foo::X; ... and convert it to this: class Foo { enum class Kind : uint32_t { X = 1, Y = 2 }; Kind kind; } ... foo.kind = Foo::Kind::X; (In some cases the enumeration would not be in the class namespace, such as when it is generally useful.) The benefits are as follows: * The type of the field gives a clear indication of intent, both to humans and tools (such as binding generators). * The compiler is able to automatically warn when a switch is not exhaustive; but this is currently suppressed by the default: ssassert(false, ...) idiom. * Integers and plain enums are weakly type checked: they implicitly convert into each other. This can hide bugs where type conversion is performed but not intended. Enum classes are strongly type checked. * Plain enums pollute parent namespaces; enum classes do not. Almost every defined enum we have already has a kind of ad-hoc namespacing via `NAMESPACE_`, which is now explicit. * Plain enums do not have a well-defined ABI size, which is important for bindings. Enum classes can have it, if specified. We specify the base type for all enums as uint32_t, which is a safe choice and allows us to not change the numeric values of any variants. This commit introduces absolutely no functional change to the code, just renaming and change of types. It handles almost all cases, except GraphicsWindow::pending.operation, which needs minor functional change.
2016-05-20 08:31:20 +00:00
case SBspUv::Class::INSIDE:
*indir = SShell::Class::INSIDE;
*outdir = SShell::Class::INSIDE;
Revamp the edge classification for Booleans. I no longer make a separate polygon of coincident (with same or opposite normal) faces; I instead test all the edges against the other shell, and have extended the classify-against-shell stuff to handle those cases. And the normals are now perturbed a bit numerically, to either side of the edge, to distinguish tangency from a coincident surface. This seems to work fairly well, although things still tend to fail when the piecewise linear tolerance is too coarse. [git-p4: depot-paths = "//depot/solvespace/": change = 1964]
2009-06-04 03:59:40 +00:00
break;
Additional poking at Booleans. At least this is a halfway rational way to think about the cases; I'm classifying the regions to the left and right of each edge, and keeping the edges if those regions (2d, surfaces) classify different. Still screws up with edge-on-edge intersections; but if I make the surface intersection stuff handle that, then might be more straightforward to use that info. [git-p4: depot-paths = "//depot/solvespace/": change = 1914]
2009-02-16 12:05:08 +00:00
Convert all enumerations to use `enum class`. Specifically, take the old code that looks like this: class Foo { enum { X = 1, Y = 2 }; int kind; } ... foo.kind = Foo::X; ... and convert it to this: class Foo { enum class Kind : uint32_t { X = 1, Y = 2 }; Kind kind; } ... foo.kind = Foo::Kind::X; (In some cases the enumeration would not be in the class namespace, such as when it is generally useful.) The benefits are as follows: * The type of the field gives a clear indication of intent, both to humans and tools (such as binding generators). * The compiler is able to automatically warn when a switch is not exhaustive; but this is currently suppressed by the default: ssassert(false, ...) idiom. * Integers and plain enums are weakly type checked: they implicitly convert into each other. This can hide bugs where type conversion is performed but not intended. Enum classes are strongly type checked. * Plain enums pollute parent namespaces; enum classes do not. Almost every defined enum we have already has a kind of ad-hoc namespacing via `NAMESPACE_`, which is now explicit. * Plain enums do not have a well-defined ABI size, which is important for bindings. Enum classes can have it, if specified. We specify the base type for all enums as uint32_t, which is a safe choice and allows us to not change the numeric values of any variants. This commit introduces absolutely no functional change to the code, just renaming and change of types. It handles almost all cases, except GraphicsWindow::pending.operation, which needs minor functional change.
2016-05-20 08:31:20 +00:00
case SBspUv::Class::OUTSIDE:
*indir = SShell::Class::OUTSIDE;
*outdir = SShell::Class::OUTSIDE;
Revamp the edge classification for Booleans. I no longer make a separate polygon of coincident (with same or opposite normal) faces; I instead test all the edges against the other shell, and have extended the classify-against-shell stuff to handle those cases. And the normals are now perturbed a bit numerically, to either side of the edge, to distinguish tangency from a coincident surface. This seems to work fairly well, although things still tend to fail when the piecewise linear tolerance is too coarse. [git-p4: depot-paths = "//depot/solvespace/": change = 1964]
2009-06-04 03:59:40 +00:00
break;
Additional poking at Booleans. At least this is a halfway rational way to think about the cases; I'm classifying the regions to the left and right of each edge, and keeping the edges if those regions (2d, surfaces) classify different. Still screws up with edge-on-edge intersections; but if I make the surface intersection stuff handle that, then might be more straightforward to use that info. [git-p4: depot-paths = "//depot/solvespace/": change = 1914]
2009-02-16 12:05:08 +00:00
Convert all enumerations to use `enum class`. Specifically, take the old code that looks like this: class Foo { enum { X = 1, Y = 2 }; int kind; } ... foo.kind = Foo::X; ... and convert it to this: class Foo { enum class Kind : uint32_t { X = 1, Y = 2 }; Kind kind; } ... foo.kind = Foo::Kind::X; (In some cases the enumeration would not be in the class namespace, such as when it is generally useful.) The benefits are as follows: * The type of the field gives a clear indication of intent, both to humans and tools (such as binding generators). * The compiler is able to automatically warn when a switch is not exhaustive; but this is currently suppressed by the default: ssassert(false, ...) idiom. * Integers and plain enums are weakly type checked: they implicitly convert into each other. This can hide bugs where type conversion is performed but not intended. Enum classes are strongly type checked. * Plain enums pollute parent namespaces; enum classes do not. Almost every defined enum we have already has a kind of ad-hoc namespacing via `NAMESPACE_`, which is now explicit. * Plain enums do not have a well-defined ABI size, which is important for bindings. Enum classes can have it, if specified. We specify the base type for all enums as uint32_t, which is a safe choice and allows us to not change the numeric values of any variants. This commit introduces absolutely no functional change to the code, just renaming and change of types. It handles almost all cases, except GraphicsWindow::pending.operation, which needs minor functional change.
2016-05-20 08:31:20 +00:00
case SBspUv::Class::EDGE_PARALLEL:
*indir = SShell::Class::INSIDE;
*outdir = SShell::Class::OUTSIDE;
Revamp the edge classification for Booleans. I no longer make a separate polygon of coincident (with same or opposite normal) faces; I instead test all the edges against the other shell, and have extended the classify-against-shell stuff to handle those cases. And the normals are now perturbed a bit numerically, to either side of the edge, to distinguish tangency from a coincident surface. This seems to work fairly well, although things still tend to fail when the piecewise linear tolerance is too coarse. [git-p4: depot-paths = "//depot/solvespace/": change = 1964]
2009-06-04 03:59:40 +00:00
break;
Additional poking at Booleans. At least this is a halfway rational way to think about the cases; I'm classifying the regions to the left and right of each edge, and keeping the edges if those regions (2d, surfaces) classify different. Still screws up with edge-on-edge intersections; but if I make the surface intersection stuff handle that, then might be more straightforward to use that info. [git-p4: depot-paths = "//depot/solvespace/": change = 1914]
2009-02-16 12:05:08 +00:00
Convert all enumerations to use `enum class`. Specifically, take the old code that looks like this: class Foo { enum { X = 1, Y = 2 }; int kind; } ... foo.kind = Foo::X; ... and convert it to this: class Foo { enum class Kind : uint32_t { X = 1, Y = 2 }; Kind kind; } ... foo.kind = Foo::Kind::X; (In some cases the enumeration would not be in the class namespace, such as when it is generally useful.) The benefits are as follows: * The type of the field gives a clear indication of intent, both to humans and tools (such as binding generators). * The compiler is able to automatically warn when a switch is not exhaustive; but this is currently suppressed by the default: ssassert(false, ...) idiom. * Integers and plain enums are weakly type checked: they implicitly convert into each other. This can hide bugs where type conversion is performed but not intended. Enum classes are strongly type checked. * Plain enums pollute parent namespaces; enum classes do not. Almost every defined enum we have already has a kind of ad-hoc namespacing via `NAMESPACE_`, which is now explicit. * Plain enums do not have a well-defined ABI size, which is important for bindings. Enum classes can have it, if specified. We specify the base type for all enums as uint32_t, which is a safe choice and allows us to not change the numeric values of any variants. This commit introduces absolutely no functional change to the code, just renaming and change of types. It handles almost all cases, except GraphicsWindow::pending.operation, which needs minor functional change.
2016-05-20 08:31:20 +00:00
case SBspUv::Class::EDGE_ANTIPARALLEL:
*indir = SShell::Class::OUTSIDE;
*outdir = SShell::Class::INSIDE;
Revamp the edge classification for Booleans. I no longer make a separate polygon of coincident (with same or opposite normal) faces; I instead test all the edges against the other shell, and have extended the classify-against-shell stuff to handle those cases. And the normals are now perturbed a bit numerically, to either side of the edge, to distinguish tangency from a coincident surface. This seems to work fairly well, although things still tend to fail when the piecewise linear tolerance is too coarse. [git-p4: depot-paths = "//depot/solvespace/": change = 1964]
2009-06-04 03:59:40 +00:00
break;
Additional poking at Booleans. At least this is a halfway rational way to think about the cases; I'm classifying the regions to the left and right of each edge, and keeping the edges if those regions (2d, surfaces) classify different. Still screws up with edge-on-edge intersections; but if I make the surface intersection stuff handle that, then might be more straightforward to use that info. [git-p4: depot-paths = "//depot/solvespace/": change = 1914]
2009-02-16 12:05:08 +00:00
Add a menu item to rotate an imported part by ninety degrees about the coordinate axis closest to the screen normal. [git-p4: depot-paths = "//depot/solvespace/": change = 1994]
2009-06-21 20:39:42 +00:00
default:
dbp("TagByClassifiedEdge: fail!");
Convert all enumerations to use `enum class`. Specifically, take the old code that looks like this: class Foo { enum { X = 1, Y = 2 }; int kind; } ... foo.kind = Foo::X; ... and convert it to this: class Foo { enum class Kind : uint32_t { X = 1, Y = 2 }; Kind kind; } ... foo.kind = Foo::Kind::X; (In some cases the enumeration would not be in the class namespace, such as when it is generally useful.) The benefits are as follows: * The type of the field gives a clear indication of intent, both to humans and tools (such as binding generators). * The compiler is able to automatically warn when a switch is not exhaustive; but this is currently suppressed by the default: ssassert(false, ...) idiom. * Integers and plain enums are weakly type checked: they implicitly convert into each other. This can hide bugs where type conversion is performed but not intended. Enum classes are strongly type checked. * Plain enums pollute parent namespaces; enum classes do not. Almost every defined enum we have already has a kind of ad-hoc namespacing via `NAMESPACE_`, which is now explicit. * Plain enums do not have a well-defined ABI size, which is important for bindings. Enum classes can have it, if specified. We specify the base type for all enums as uint32_t, which is a safe choice and allows us to not change the numeric values of any variants. This commit introduces absolutely no functional change to the code, just renaming and change of types. It handles almost all cases, except GraphicsWindow::pending.operation, which needs minor functional change.
2016-05-20 08:31:20 +00:00
*indir = SShell::Class::OUTSIDE;
*outdir = SShell::Class::OUTSIDE;
Add a menu item to rotate an imported part by ninety degrees about the coordinate axis closest to the screen normal. [git-p4: depot-paths = "//depot/solvespace/": change = 1994]
2009-06-21 20:39:42 +00:00
break;
Additional poking at Booleans. At least this is a halfway rational way to think about the cases; I'm classifying the regions to the left and right of each edge, and keeping the edges if those regions (2d, surfaces) classify different. Still screws up with edge-on-edge intersections; but if I make the surface intersection stuff handle that, then might be more straightforward to use that info. [git-p4: depot-paths = "//depot/solvespace/": change = 1914]
2009-02-16 12:05:08 +00:00
}
}
Various fixes for warnings and minutia This commit consists of numerous small changes, none significant enough to merit a commit on their own: * Added extra braces to quash for-loop variable scoping issues for older compilers (or "g++ -fno-for-scope") * Appeased "unreachable code" warnings, spurious or otherwise * Added casts to fix integer-variable signedness warnings * Added a dummy virtual method to the VectorFileWriter class to silence the -Wweak-vtables warning from Clang++ * Renamed some parameters in the Expr and GraphicsWindow classes to eliminate "parameter shadows a field" warnings * Removed an inert "0 ||" from a conditional, and changed a "&& 0" into an "#if 0" * Added missing elements to array/struct/class initializers to zap further warnings * Indented some cpp conditionals where appropriate * Qualified some variables and functions as static to quiet "no previous declaration" warnings * toolbar.cpp needed to #include<icons-proto.h> to fix those same "no previous declaration" warnings from icons.h * Added some casts and const qualifiers to the Win32 code to address warnings produced by g++ when compiling under MinGW * Rewrote Cnf{Freeze,Thaw}Float() to use a union rather than pointer aliasing; this makes g++ a lot happier * Removed redundant #includes from win32/w32util.cpp * With Jonathan's blessing, shortened the last line of the About dialog text to better match the preceding lines
2013-10-19 05:36:45 +00:00
static void DEBUGEDGELIST(SEdgeList *sel, SSurface *surf) {
Add code to subdivide (with de Castljau's algorithm) a surface, and use that for surface-line intersections. That has major problems with the heuristic on when to stop and do Newton polishing. There's also an issue with all the Newton stuff when surfaces join tangent. And update the wishlist to reflect current needs. [git-p4: depot-paths = "//depot/solvespace/": change = 1925]
2009-03-08 10:59:57 +00:00
dbp("print %d edges", sel->l.n);
More coincident fixing; test for edge-on-edge, fix some gross stupidity. [git-p4: depot-paths = "//depot/solvespace/": change = 1915]
2009-02-17 11:17:12 +00:00
SEdge *se;
for(se = sel->l.First(); se; se = sel->l.NextAfter(se)) {
Vector mid = (se->a).Plus(se->b).ScaledBy(0.5);
Vector arrow = (se->b).Minus(se->a);
Use C++ std::{min,max,swap} instead of custom ones. The main benefit is that std::swap will ensure that the type of arguments is copy-constructible and move-constructible. It is more concise as well. When min and max are defined as macros, they will conflict with STL header files included by other C++ libraries; in this case STL will #undef any other definition.
2015-03-27 15:43:28 +00:00
swap(arrow.x, arrow.y);
More coincident fixing; test for edge-on-edge, fix some gross stupidity. [git-p4: depot-paths = "//depot/solvespace/": change = 1915]
2009-02-17 11:17:12 +00:00
arrow.x *= -1;
I had been using LENGTH_EPS as the tolerance on both xyz points and uv points. This is inconsistent, unless the surface happens to be a plane square with side length one. So modify the SBspUv tests to take a surface, and measure distance linearized in that surface. That fixes at least one mis-classification bug, and doesn't seem to break anything. [git-p4: depot-paths = "//depot/solvespace/": change = 2005]
2009-07-02 03:32:17 +00:00
arrow = arrow.WithMagnitude(0.01);
More coincident fixing; test for edge-on-edge, fix some gross stupidity. [git-p4: depot-paths = "//depot/solvespace/": change = 1915]
2009-02-17 11:17:12 +00:00
arrow = arrow.Plus(mid);
SS.nakedEdges.AddEdge(surf->PointAt(se->a.x, se->a.y),
surf->PointAt(se->b.x, se->b.y));
SS.nakedEdges.AddEdge(surf->PointAt(mid.x, mid.y),
surf->PointAt(arrow.x, arrow.y));
}
}
Group edges into chains (that don't intersect edges from the other contour, except at the ends of the chain), and classify the entire chain. That's much faster than going edge by edge. [git-p4: depot-paths = "//depot/solvespace/": change = 2002]
2009-06-27 05:53:56 +00:00
//-----------------------------------------------------------------------------
// We are given src, with at least one edge, and avoid, a list of points to
// avoid. We return a chain of edges (that share endpoints), such that no
// point within the avoid list ever occurs in the middle of a chain. And we
// delete the edges in that chain from our source list.
//-----------------------------------------------------------------------------
Trim trailing whitespace.
2015-03-29 00:30:52 +00:00
void SSurface::FindChainAvoiding(SEdgeList *src, SEdgeList *dest,
Group edges into chains (that don't intersect edges from the other contour, except at the ends of the chain), and classify the entire chain. That's much faster than going edge by edge. [git-p4: depot-paths = "//depot/solvespace/": change = 2002]
2009-06-27 05:53:56 +00:00
SPointList *avoid)
{
Replace all oops() checks with ssassert()s. This includes explanation and context for non-obvious cases and shortens debug cycles when just-in-time debugging is not available (like on Linux) by immediately printing description of the assert as well as symbolized backtrace.
2016-05-18 22:51:36 +00:00
ssassert(src->l.n > 0, "Need at least one edge");
Group edges into chains (that don't intersect edges from the other contour, except at the ends of the chain), and classify the entire chain. That's much faster than going edge by edge. [git-p4: depot-paths = "//depot/solvespace/": change = 2002]
2009-06-27 05:53:56 +00:00
// Start with an arbitrary edge.
dest->l.Add(&(src->l.elem[0]));
src->l.ClearTags();
src->l.elem[0].tag = 1;
bool added;
do {
added = false;
// The start and finish of the current edge chain
Vector s = dest->l.elem[0].a,
f = dest->l.elem[dest->l.n - 1].b;
// We can attach a new edge at the start or finish, as long as that
// start or finish point isn't in the list of points to avoid.
bool startOkay = !avoid->ContainsPoint(s),
finishOkay = !avoid->ContainsPoint(f);
// Now look for an unused edge that joins at the start or finish of
// our chain (if permitted by the avoid list).
SEdge *se;
for(se = src->l.First(); se; se = src->l.NextAfter(se)) {
if(se->tag) continue;
if(startOkay && s.Equals(se->b)) {
dest->l.AddToBeginning(se);
s = se->a;
se->tag = 1;
startOkay = !avoid->ContainsPoint(s);
} else if(finishOkay && f.Equals(se->a)) {
dest->l.Add(se);
f = se->b;
se->tag = 1;
finishOkay = !avoid->ContainsPoint(f);
} else {
continue;
}
added = true;
}
} while(added);
src->l.RemoveTagged();
}
Revamp the edge classification for Booleans. I no longer make a separate polygon of coincident (with same or opposite normal) faces; I instead test all the edges against the other shell, and have extended the classify-against-shell stuff to handle those cases. And the normals are now perturbed a bit numerically, to either side of the edge, to distinguish tangency from a coincident surface. This seems to work fairly well, although things still tend to fail when the piecewise linear tolerance is too coarse. [git-p4: depot-paths = "//depot/solvespace/": change = 1964]
2009-06-04 03:59:40 +00:00
void SSurface::EdgeNormalsWithinSurface(Point2d auv, Point2d buv,
Vector *pt,
Vector *enin, Vector *enout,
Vector *surfn,
Use C99 integer types and C++ boolean types/values This change comprehensively replaces the use of Microsoft-standard integer and boolean types with their C99/C++ standard equivalents, as the latter is more appropriate for a cross-platform application. With matter-of-course exceptions in the Win32-specific code, the types/values have been converted as follows: QWORD --> uint64_t SQWORD --> int64_t DWORD --> uint32_t SDWORD --> int32_t WORD --> uint16_t SWORD --> int16_t BYTE --> uint8_t BOOL --> bool TRUE --> true FALSE --> false The following related changes are also included: * Added C99 integer type definitions for Windows, as stdint.h is not available prior to Visual Studio 2010 * Changed types of some variables in the SolveSpace class from 'int' to 'bool', as they actually represent boolean settings * Implemented new Cnf{Freeze,Thaw}Bool() functions to support boolean variables in the Registry * Cnf{Freeze,Thaw}DWORD() are now Cnf{Freeze,Thaw}Int() * TtfFont::Get{WORD,DWORD}() are now TtfFont::Get{USHORT,ULONG}() (names inspired by the OpenType spec) * RGB colors are packed into an integer of type uint32_t (nee DWORD), but in a few places, these were represented by an int; these have been corrected to uint32_t
2013-10-02 05:45:13 +00:00
uint32_t auxA,
Add beginnings of marching surface intersection; I can find all the boundary points, at least. That required some changes to what gets passed around (for example because to project a point onto this inexact curve, we need to know which two surfaces it trims so that we can do a Newton's method on them). And fix stupidity in the way that I calculated edge normals; I just did normal in uv space, and there's no particular reason why that would be normal in xyz. So edges in long skinny surfaces failed, for example. [git-p4: depot-paths = "//depot/solvespace/": change = 1990]
2009-06-19 07:56:33 +00:00
SShell *shell, SShell *sha, SShell *shb)
Revamp the edge classification for Booleans. I no longer make a separate polygon of coincident (with same or opposite normal) faces; I instead test all the edges against the other shell, and have extended the classify-against-shell stuff to handle those cases. And the normals are now perturbed a bit numerically, to either side of the edge, to distinguish tangency from a coincident surface. This seems to work fairly well, although things still tend to fail when the piecewise linear tolerance is too coarse. [git-p4: depot-paths = "//depot/solvespace/": change = 1964]
2009-06-04 03:59:40 +00:00
{
// the midpoint of the edge
Point2d muv = (auv.Plus(buv)).ScaledBy(0.5);
*pt = PointAt(muv);
// If this edge just approximates a curve, then refine our midpoint so
// so that it actually lies on that curve too. Otherwise stuff like
// point-on-face tests will fail, since the point won't actually lie
// on the other face.
hSCurve hc = { auxA };
SCurve *sc = shell->curve.FindById(hc);
if(sc->isExact && sc->exact.deg != 1) {
double t;
sc->exact.ClosestPointTo(*pt, &t, false);
*pt = sc->exact.PointAt(t);
Oops, was computing numerical by perturbing a point in uv; but the xyz point that I subtracted off had been refined to lie exactly on our edge's curve, and the uv point that I started with had not. So normals got randomly screwed up. [git-p4: depot-paths = "//depot/solvespace/": change = 1978]
2009-06-10 08:04:35 +00:00
ClosestPointTo(*pt, &muv);
Add beginnings of marching surface intersection; I can find all the boundary points, at least. That required some changes to what gets passed around (for example because to project a point onto this inexact curve, we need to know which two surfaces it trims so that we can do a Newton's method on them). And fix stupidity in the way that I calculated edge normals; I just did normal in uv space, and there's no particular reason why that would be normal in xyz. So edges in long skinny surfaces failed, for example. [git-p4: depot-paths = "//depot/solvespace/": change = 1990]
2009-06-19 07:56:33 +00:00
} else if(!sc->isExact) {
SSurface *trimmedA = sc->GetSurfaceA(sha, shb),
Add const qualifiers to functions where trivially possible. This will allow us in future to accept `const T &` anywhere it's necessary to reduce the amount of copying. This commit is quite conservative: it does not attempt very hard to refactor code that performs incidental mutation. In particular dogd and caches are not marked with the `mutable` keyword. dogd will be eliminated later, opening up more opportunities to add const qualifiers. This commit also doesn't introduce any uses of the newly added const qualifers. This will be done later.
2016-05-21 05:18:00 +00:00
*trimmedB = sc->GetSurfaceB(sha, shb);
Add beginnings of marching surface intersection; I can find all the boundary points, at least. That required some changes to what gets passed around (for example because to project a point onto this inexact curve, we need to know which two surfaces it trims so that we can do a Newton's method on them). And fix stupidity in the way that I calculated edge normals; I just did normal in uv space, and there's no particular reason why that would be normal in xyz. So edges in long skinny surfaces failed, for example. [git-p4: depot-paths = "//depot/solvespace/": change = 1990]
2009-06-19 07:56:33 +00:00
*pt = trimmedA->ClosestPointOnThisAndSurface(trimmedB, *pt);
ClosestPointTo(*pt, &muv);
Revamp the edge classification for Booleans. I no longer make a separate polygon of coincident (with same or opposite normal) faces; I instead test all the edges against the other shell, and have extended the classify-against-shell stuff to handle those cases. And the normals are now perturbed a bit numerically, to either side of the edge, to distinguish tangency from a coincident surface. This seems to work fairly well, although things still tend to fail when the piecewise linear tolerance is too coarse. [git-p4: depot-paths = "//depot/solvespace/": change = 1964]
2009-06-04 03:59:40 +00:00
}
Oops, was computing numerical by perturbing a point in uv; but the xyz point that I subtracted off had been refined to lie exactly on our edge's curve, and the uv point that I started with had not. So normals got randomly screwed up. [git-p4: depot-paths = "//depot/solvespace/": change = 1978]
2009-06-10 08:04:35 +00:00
*surfn = NormalAt(muv.x, muv.y);
Add beginnings of marching surface intersection; I can find all the boundary points, at least. That required some changes to what gets passed around (for example because to project a point onto this inexact curve, we need to know which two surfaces it trims so that we can do a Newton's method on them). And fix stupidity in the way that I calculated edge normals; I just did normal in uv space, and there's no particular reason why that would be normal in xyz. So edges in long skinny surfaces failed, for example. [git-p4: depot-paths = "//depot/solvespace/": change = 1990]
2009-06-19 07:56:33 +00:00
// Compute the edge's inner normal in xyz space.
Vector ab = (PointAt(auv)).Minus(PointAt(buv)),
enxyz = (ab.Cross(*surfn)).WithMagnitude(SS.ChordTolMm());
// And based on that, compute the edge's inner normal in uv space. This
// vector is perpendicular to the edge in xyz, but not necessarily in uv.
Vector tu, tv;
TangentsAt(muv.x, muv.y, &tu, &tv);
Point2d enuv;
enuv.x = enxyz.Dot(tu) / tu.MagSquared();
enuv.y = enxyz.Dot(tv) / tv.MagSquared();
Revamp the edge classification for Booleans. I no longer make a separate polygon of coincident (with same or opposite normal) faces; I instead test all the edges against the other shell, and have extended the classify-against-shell stuff to handle those cases. And the normals are now perturbed a bit numerically, to either side of the edge, to distinguish tangency from a coincident surface. This seems to work fairly well, although things still tend to fail when the piecewise linear tolerance is too coarse. [git-p4: depot-paths = "//depot/solvespace/": change = 1964]
2009-06-04 03:59:40 +00:00
// Compute the inner and outer normals of this edge (within the srf),
// in xyz space. These are not necessarily antiparallel, if the
// surface is curved.
Add beginnings of marching surface intersection; I can find all the boundary points, at least. That required some changes to what gets passed around (for example because to project a point onto this inexact curve, we need to know which two surfaces it trims so that we can do a Newton's method on them). And fix stupidity in the way that I calculated edge normals; I just did normal in uv space, and there's no particular reason why that would be normal in xyz. So edges in long skinny surfaces failed, for example. [git-p4: depot-paths = "//depot/solvespace/": change = 1990]
2009-06-19 07:56:33 +00:00
Vector pin = PointAt(muv.Minus(enuv)),
pout = PointAt(muv.Plus(enuv));
Revamp the edge classification for Booleans. I no longer make a separate polygon of coincident (with same or opposite normal) faces; I instead test all the edges against the other shell, and have extended the classify-against-shell stuff to handle those cases. And the normals are now perturbed a bit numerically, to either side of the edge, to distinguish tangency from a coincident surface. This seems to work fairly well, although things still tend to fail when the piecewise linear tolerance is too coarse. [git-p4: depot-paths = "//depot/solvespace/": change = 1964]
2009-06-04 03:59:40 +00:00
*enin = pin.Minus(*pt),
*enout = pout.Minus(*pt);
}
Preparatory work for Boolean. Make the u and v coordinates of the trim curves for all surfaces lie between 0 and 1. And add routines to merge the curves and surfaces from two shells into one, and to split the trim curves into their piecewise linear segments and then reassemble them into trim curves. [git-p4: depot-paths = "//depot/solvespace/": change = 1905]
2009-01-25 11:52:29 +00:00
//-----------------------------------------------------------------------------
// Trim this surface against the specified shell, in the way that's appropriate
// for the specified Boolean operation type (and which operand we are). We
// also need a pointer to the shell that contains our own surface, since that
// contains our original trim curves.
//-----------------------------------------------------------------------------
Add beginnings of marching surface intersection; I can find all the boundary points, at least. That required some changes to what gets passed around (for example because to project a point onto this inexact curve, we need to know which two surfaces it trims so that we can do a Newton's method on them). And fix stupidity in the way that I calculated edge normals; I just did normal in uv space, and there's no particular reason why that would be normal in xyz. So edges in long skinny surfaces failed, for example. [git-p4: depot-paths = "//depot/solvespace/": change = 1990]
2009-06-19 07:56:33 +00:00
SSurface SSurface::MakeCopyTrimAgainst(SShell *parent,
SShell *sha, SShell *shb,
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
SShell *into,
Convert all enumerations to use `enum class`. Specifically, take the old code that looks like this: class Foo { enum { X = 1, Y = 2 }; int kind; } ... foo.kind = Foo::X; ... and convert it to this: class Foo { enum class Kind : uint32_t { X = 1, Y = 2 }; Kind kind; } ... foo.kind = Foo::Kind::X; (In some cases the enumeration would not be in the class namespace, such as when it is generally useful.) The benefits are as follows: * The type of the field gives a clear indication of intent, both to humans and tools (such as binding generators). * The compiler is able to automatically warn when a switch is not exhaustive; but this is currently suppressed by the default: ssassert(false, ...) idiom. * Integers and plain enums are weakly type checked: they implicitly convert into each other. This can hide bugs where type conversion is performed but not intended. Enum classes are strongly type checked. * Plain enums pollute parent namespaces; enum classes do not. Almost every defined enum we have already has a kind of ad-hoc namespacing via `NAMESPACE_`, which is now explicit. * Plain enums do not have a well-defined ABI size, which is important for bindings. Enum classes can have it, if specified. We specify the base type for all enums as uint32_t, which is a safe choice and allows us to not change the numeric values of any variants. This commit introduces absolutely no functional change to the code, just renaming and change of types. It handles almost all cases, except GraphicsWindow::pending.operation, which needs minor functional change.
2016-05-20 08:31:20 +00:00
SSurface::CombineAs type)
Preparatory work for Boolean. Make the u and v coordinates of the trim curves for all surfaces lie between 0 and 1. And add routines to merge the curves and surfaces from two shells into one, and to split the trim curves into their piecewise linear segments and then reassemble them into trim curves. [git-p4: depot-paths = "//depot/solvespace/": change = 1905]
2009-01-25 11:52:29 +00:00
{
Add beginnings of marching surface intersection; I can find all the boundary points, at least. That required some changes to what gets passed around (for example because to project a point onto this inexact curve, we need to know which two surfaces it trims so that we can do a Newton's method on them). And fix stupidity in the way that I calculated edge normals; I just did normal in uv space, and there's no particular reason why that would be normal in xyz. So edges in long skinny surfaces failed, for example. [git-p4: depot-paths = "//depot/solvespace/": change = 1990]
2009-06-19 07:56:33 +00:00
bool opA = (parent == sha);
SShell *agnst = opA ? shb : sha;
Preparatory work for Boolean. Make the u and v coordinates of the trim curves for all surfaces lie between 0 and 1. And add routines to merge the curves and surfaces from two shells into one, and to split the trim curves into their piecewise linear segments and then reassemble them into trim curves. [git-p4: depot-paths = "//depot/solvespace/": change = 1905]
2009-01-25 11:52:29 +00:00
SSurface ret;
// The returned surface is identical, just the trim curves change
ret = *this;
Replace all ZERO and memset with C++11 brace-initialization. This will allow us to use non-POD classes inside these objects in future and is otherwise functionally equivalent, as well as more concise. Note that there are some subtleties with handling of brace-initialization. Specifically: On aggregates (e.g. simple C-style structures) using an empty brace-initializer zero-initializes the aggregate, i.e. it makes all members zero. On non-aggregates an empty brace-initializer calls the default constructor. And if the constructor doesn't explicitly initialize the members (which the auto-generated constructor doesn't) then the members will be constructed but otherwise uninitialized. So, what is an aggregate class? To quote the C++ standard (C++03 8.5.1 §1): An aggregate is an array or a class (clause 9) with no user-declared constructors (12.1), no private or protected non-static data members (clause 11), no base classes (clause 10), and no virtual functions (10.3). In SolveSpace, we only have to handle the case of base classes; Constraint and Entity have those. Thus, they had to gain a default constructor that does nothing but initializes the members to zero.
2015-03-27 15:31:23 +00:00
ret.trim = {};
Preparatory work for Boolean. Make the u and v coordinates of the trim curves for all surfaces lie between 0 and 1. And add routines to merge the curves and surfaces from two shells into one, and to split the trim curves into their piecewise linear segments and then reassemble them into trim curves. [git-p4: depot-paths = "//depot/solvespace/": change = 1905]
2009-01-25 11:52:29 +00:00
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
// First, build a list of the existing trim curves; update them to use
// the split curves.
STrimBy *stb;
for(stb = trim.First(); stb; stb = trim.NextAfter(stb)) {
STrimBy stn = *stb;
stn.curve = (parent->curve.FindById(stn.curve))->newH;
ret.trim.Add(&stn);
}
Preparatory work for Boolean. Make the u and v coordinates of the trim curves for all surfaces lie between 0 and 1. And add routines to merge the curves and surfaces from two shells into one, and to split the trim curves into their piecewise linear segments and then reassemble them into trim curves. [git-p4: depot-paths = "//depot/solvespace/": change = 1905]
2009-01-25 11:52:29 +00:00
Convert all enumerations to use `enum class`. Specifically, take the old code that looks like this: class Foo { enum { X = 1, Y = 2 }; int kind; } ... foo.kind = Foo::X; ... and convert it to this: class Foo { enum class Kind : uint32_t { X = 1, Y = 2 }; Kind kind; } ... foo.kind = Foo::Kind::X; (In some cases the enumeration would not be in the class namespace, such as when it is generally useful.) The benefits are as follows: * The type of the field gives a clear indication of intent, both to humans and tools (such as binding generators). * The compiler is able to automatically warn when a switch is not exhaustive; but this is currently suppressed by the default: ssassert(false, ...) idiom. * Integers and plain enums are weakly type checked: they implicitly convert into each other. This can hide bugs where type conversion is performed but not intended. Enum classes are strongly type checked. * Plain enums pollute parent namespaces; enum classes do not. Almost every defined enum we have already has a kind of ad-hoc namespacing via `NAMESPACE_`, which is now explicit. * Plain enums do not have a well-defined ABI size, which is important for bindings. Enum classes can have it, if specified. We specify the base type for all enums as uint32_t, which is a safe choice and allows us to not change the numeric values of any variants. This commit introduces absolutely no functional change to the code, just renaming and change of types. It handles almost all cases, except GraphicsWindow::pending.operation, which needs minor functional change.
2016-05-20 08:31:20 +00:00
if(type == SSurface::CombineAs::DIFFERENCE && !opA) {
Additional poking at Booleans. At least this is a halfway rational way to think about the cases; I'm classifying the regions to the left and right of each edge, and keeping the edges if those regions (2d, surfaces) classify different. Still screws up with edge-on-edge intersections; but if I make the surface intersection stuff handle that, then might be more straightforward to use that info. [git-p4: depot-paths = "//depot/solvespace/": change = 1914]
2009-02-16 12:05:08 +00:00
// The second operand of a Boolean difference gets turned inside out
ret.Reverse();
}
More coincident fixing; test for edge-on-edge, fix some gross stupidity. [git-p4: depot-paths = "//depot/solvespace/": change = 1915]
2009-02-17 11:17:12 +00:00
// Build up our original trim polygon; remember the coordinates could
Revamp the edge classification for Booleans. I no longer make a separate polygon of coincident (with same or opposite normal) faces; I instead test all the edges against the other shell, and have extended the classify-against-shell stuff to handle those cases. And the normals are now perturbed a bit numerically, to either side of the edge, to distinguish tangency from a coincident surface. This seems to work fairly well, although things still tend to fail when the piecewise linear tolerance is too coarse. [git-p4: depot-paths = "//depot/solvespace/": change = 1964]
2009-06-04 03:59:40 +00:00
// be changed if we just flipped the surface normal, and we are using
// the split curves (not the original curves).
Replace all ZERO and memset with C++11 brace-initialization. This will allow us to use non-POD classes inside these objects in future and is otherwise functionally equivalent, as well as more concise. Note that there are some subtleties with handling of brace-initialization. Specifically: On aggregates (e.g. simple C-style structures) using an empty brace-initializer zero-initializes the aggregate, i.e. it makes all members zero. On non-aggregates an empty brace-initializer calls the default constructor. And if the constructor doesn't explicitly initialize the members (which the auto-generated constructor doesn't) then the members will be constructed but otherwise uninitialized. So, what is an aggregate class? To quote the C++ standard (C++03 8.5.1 §1): An aggregate is an array or a class (clause 9) with no user-declared constructors (12.1), no private or protected non-static data members (clause 11), no base classes (clause 10), and no virtual functions (10.3). In SolveSpace, we only have to handle the case of base classes; Constraint and Entity have those. Thus, they had to gain a default constructor that does nothing but initializes the members to zero.
2015-03-27 15:31:23 +00:00
SEdgeList orig = {};
Convert all enumerations to use `enum class`. Specifically, take the old code that looks like this: class Foo { enum { X = 1, Y = 2 }; int kind; } ... foo.kind = Foo::X; ... and convert it to this: class Foo { enum class Kind : uint32_t { X = 1, Y = 2 }; Kind kind; } ... foo.kind = Foo::Kind::X; (In some cases the enumeration would not be in the class namespace, such as when it is generally useful.) The benefits are as follows: * The type of the field gives a clear indication of intent, both to humans and tools (such as binding generators). * The compiler is able to automatically warn when a switch is not exhaustive; but this is currently suppressed by the default: ssassert(false, ...) idiom. * Integers and plain enums are weakly type checked: they implicitly convert into each other. This can hide bugs where type conversion is performed but not intended. Enum classes are strongly type checked. * Plain enums pollute parent namespaces; enum classes do not. Almost every defined enum we have already has a kind of ad-hoc namespacing via `NAMESPACE_`, which is now explicit. * Plain enums do not have a well-defined ABI size, which is important for bindings. Enum classes can have it, if specified. We specify the base type for all enums as uint32_t, which is a safe choice and allows us to not change the numeric values of any variants. This commit introduces absolutely no functional change to the code, just renaming and change of types. It handles almost all cases, except GraphicsWindow::pending.operation, which needs minor functional change.
2016-05-20 08:31:20 +00:00
ret.MakeEdgesInto(into, &orig, MakeAs::UV);
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
ret.trim.Clear();
More coincident fixing; test for edge-on-edge, fix some gross stupidity. [git-p4: depot-paths = "//depot/solvespace/": change = 1915]
2009-02-17 11:17:12 +00:00
// which means that we can't necessarily use the old BSP...
I had been using LENGTH_EPS as the tolerance on both xyz points and uv points. This is inconsistent, unless the surface happens to be a plane square with side length one. So modify the SBspUv tests to take a surface, and measure distance linearized in that surface. That fixes at least one mis-classification bug, and doesn't seem to break anything. [git-p4: depot-paths = "//depot/solvespace/": change = 2005]
2009-07-02 03:32:17 +00:00
SBspUv *origBsp = SBspUv::From(&orig, &ret);
Preparatory work for Boolean. Make the u and v coordinates of the trim curves for all surfaces lie between 0 and 1. And add routines to merge the curves and surfaces from two shells into one, and to split the trim curves into their piecewise linear segments and then reassemble them into trim curves. [git-p4: depot-paths = "//depot/solvespace/": change = 1905]
2009-01-25 11:52:29 +00:00
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
// And now intersect the other shell against us
Replace all ZERO and memset with C++11 brace-initialization. This will allow us to use non-POD classes inside these objects in future and is otherwise functionally equivalent, as well as more concise. Note that there are some subtleties with handling of brace-initialization. Specifically: On aggregates (e.g. simple C-style structures) using an empty brace-initializer zero-initializes the aggregate, i.e. it makes all members zero. On non-aggregates an empty brace-initializer calls the default constructor. And if the constructor doesn't explicitly initialize the members (which the auto-generated constructor doesn't) then the members will be constructed but otherwise uninitialized. So, what is an aggregate class? To quote the C++ standard (C++03 8.5.1 §1): An aggregate is an array or a class (clause 9) with no user-declared constructors (12.1), no private or protected non-static data members (clause 11), no base classes (clause 10), and no virtual functions (10.3). In SolveSpace, we only have to handle the case of base classes; Constraint and Entity have those. Thus, they had to gain a default constructor that does nothing but initializes the members to zero.
2015-03-27 15:31:23 +00:00
SEdgeList inter = {};
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
SSurface *ss;
for(ss = agnst->surface.First(); ss; ss = agnst->surface.NextAfter(ss)) {
SCurve *sc;
for(sc = into->curve.First(); sc; sc = into->curve.NextAfter(sc)) {
Convert all enumerations to use `enum class`. Specifically, take the old code that looks like this: class Foo { enum { X = 1, Y = 2 }; int kind; } ... foo.kind = Foo::X; ... and convert it to this: class Foo { enum class Kind : uint32_t { X = 1, Y = 2 }; Kind kind; } ... foo.kind = Foo::Kind::X; (In some cases the enumeration would not be in the class namespace, such as when it is generally useful.) The benefits are as follows: * The type of the field gives a clear indication of intent, both to humans and tools (such as binding generators). * The compiler is able to automatically warn when a switch is not exhaustive; but this is currently suppressed by the default: ssassert(false, ...) idiom. * Integers and plain enums are weakly type checked: they implicitly convert into each other. This can hide bugs where type conversion is performed but not intended. Enum classes are strongly type checked. * Plain enums pollute parent namespaces; enum classes do not. Almost every defined enum we have already has a kind of ad-hoc namespacing via `NAMESPACE_`, which is now explicit. * Plain enums do not have a well-defined ABI size, which is important for bindings. Enum classes can have it, if specified. We specify the base type for all enums as uint32_t, which is a safe choice and allows us to not change the numeric values of any variants. This commit introduces absolutely no functional change to the code, just renaming and change of types. It handles almost all cases, except GraphicsWindow::pending.operation, which needs minor functional change.
2016-05-20 08:31:20 +00:00
if(sc->source != SCurve::Source::INTERSECTION) continue;
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
if(opA) {
if(sc->surfA.v != h.v || sc->surfB.v != ss->h.v) continue;
Many changes: * Rewrite surface handles in curves, so that Booleans beyond the first don't screw up. * If an intersection curve is identical to an existing curve (as happens when faces are coincident), take the piecewise linearization of the existing curve; this stops us from screwing up when different shells are pwl'd at different chord tols. * Hook up the plane faces again. * Remove coincident (parallel or anti-parallel) edges from the coincident-face edge lists when doing Booleans; those may happen if two faces are coincident with ours. * Miscellaneous bugfixes. It doesn't seem to screw up very much now, although tangent edges (and insufficient pwl resolution) may still cause problems. [git-p4: depot-paths = "//depot/solvespace/": change = 1929]
2009-03-15 23:04:45 +00:00
} else {
if(sc->surfB.v != h.v || sc->surfA.v != ss->h.v) continue;
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
}
Trim trailing whitespace.
2015-03-29 00:30:52 +00:00
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
int i;
for(i = 1; i < sc->pts.n; i++) {
Generate intersection curves for surfaces of extrusion along a parallel axis (which are always lines parallel to that axis). Remove short pwl segments when possible, to avoid short edges that get misclassified. [git-p4: depot-paths = "//depot/solvespace/": change = 1952]
2009-05-18 07:26:51 +00:00
Vector a = sc->pts.elem[i-1].p,
b = sc->pts.elem[i].p;
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
Point2d auv, buv;
ss->ClosestPointTo(a, &(auv.x), &(auv.y));
ss->ClosestPointTo(b, &(buv.x), &(buv.y));
Convert all enumerations to use `enum class`. Specifically, take the old code that looks like this: class Foo { enum { X = 1, Y = 2 }; int kind; } ... foo.kind = Foo::X; ... and convert it to this: class Foo { enum class Kind : uint32_t { X = 1, Y = 2 }; Kind kind; } ... foo.kind = Foo::Kind::X; (In some cases the enumeration would not be in the class namespace, such as when it is generally useful.) The benefits are as follows: * The type of the field gives a clear indication of intent, both to humans and tools (such as binding generators). * The compiler is able to automatically warn when a switch is not exhaustive; but this is currently suppressed by the default: ssassert(false, ...) idiom. * Integers and plain enums are weakly type checked: they implicitly convert into each other. This can hide bugs where type conversion is performed but not intended. Enum classes are strongly type checked. * Plain enums pollute parent namespaces; enum classes do not. Almost every defined enum we have already has a kind of ad-hoc namespacing via `NAMESPACE_`, which is now explicit. * Plain enums do not have a well-defined ABI size, which is important for bindings. Enum classes can have it, if specified. We specify the base type for all enums as uint32_t, which is a safe choice and allows us to not change the numeric values of any variants. This commit introduces absolutely no functional change to the code, just renaming and change of types. It handles almost all cases, except GraphicsWindow::pending.operation, which needs minor functional change.
2016-05-20 08:31:20 +00:00
SBspUv::Class c = (ss->bsp) ? ss->bsp->ClassifyEdge(auv, buv, ss) : SBspUv::Class::OUTSIDE;
if(c != SBspUv::Class::OUTSIDE) {
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
Vector ta = Vector::From(0, 0, 0);
Vector tb = Vector::From(0, 0, 0);
Additional poking at Booleans. At least this is a halfway rational way to think about the cases; I'm classifying the regions to the left and right of each edge, and keeping the edges if those regions (2d, surfaces) classify different. Still screws up with edge-on-edge intersections; but if I make the surface intersection stuff handle that, then might be more straightforward to use that info. [git-p4: depot-paths = "//depot/solvespace/": change = 1914]
2009-02-16 12:05:08 +00:00
ret.ClosestPointTo(a, &(ta.x), &(ta.y));
ret.ClosestPointTo(b, &(tb.x), &(tb.y));
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
Additional poking at Booleans. At least this is a halfway rational way to think about the cases; I'm classifying the regions to the left and right of each edge, and keeping the edges if those regions (2d, surfaces) classify different. Still screws up with edge-on-edge intersections; but if I make the surface intersection stuff handle that, then might be more straightforward to use that info. [git-p4: depot-paths = "//depot/solvespace/": change = 1914]
2009-02-16 12:05:08 +00:00
Vector tn = ret.NormalAt(ta.x, ta.y);
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
Vector sn = ss->NormalAt(auv.x, auv.y);
More coincident fixing; test for edge-on-edge, fix some gross stupidity. [git-p4: depot-paths = "//depot/solvespace/": change = 1915]
2009-02-17 11:17:12 +00:00
// We are subtracting the portion of our surface that
// lies in the shell, so the in-plane edge normal should
// point opposite to the surface normal.
bool bkwds = true;
Additional poking at Booleans. At least this is a halfway rational way to think about the cases; I'm classifying the regions to the left and right of each edge, and keeping the edges if those regions (2d, surfaces) classify different. Still screws up with edge-on-edge intersections; but if I make the surface intersection stuff handle that, then might be more straightforward to use that info. [git-p4: depot-paths = "//depot/solvespace/": change = 1914]
2009-02-16 12:05:08 +00:00
if((tn.Cross(b.Minus(a))).Dot(sn) < 0) bkwds = !bkwds;
Convert all enumerations to use `enum class`. Specifically, take the old code that looks like this: class Foo { enum { X = 1, Y = 2 }; int kind; } ... foo.kind = Foo::X; ... and convert it to this: class Foo { enum class Kind : uint32_t { X = 1, Y = 2 }; Kind kind; } ... foo.kind = Foo::Kind::X; (In some cases the enumeration would not be in the class namespace, such as when it is generally useful.) The benefits are as follows: * The type of the field gives a clear indication of intent, both to humans and tools (such as binding generators). * The compiler is able to automatically warn when a switch is not exhaustive; but this is currently suppressed by the default: ssassert(false, ...) idiom. * Integers and plain enums are weakly type checked: they implicitly convert into each other. This can hide bugs where type conversion is performed but not intended. Enum classes are strongly type checked. * Plain enums pollute parent namespaces; enum classes do not. Almost every defined enum we have already has a kind of ad-hoc namespacing via `NAMESPACE_`, which is now explicit. * Plain enums do not have a well-defined ABI size, which is important for bindings. Enum classes can have it, if specified. We specify the base type for all enums as uint32_t, which is a safe choice and allows us to not change the numeric values of any variants. This commit introduces absolutely no functional change to the code, just renaming and change of types. It handles almost all cases, except GraphicsWindow::pending.operation, which needs minor functional change.
2016-05-20 08:31:20 +00:00
if(type == SSurface::CombineAs::DIFFERENCE && !opA) bkwds = !bkwds;
Additional poking at Booleans. At least this is a halfway rational way to think about the cases; I'm classifying the regions to the left and right of each edge, and keeping the edges if those regions (2d, surfaces) classify different. Still screws up with edge-on-edge intersections; but if I make the surface intersection stuff handle that, then might be more straightforward to use that info. [git-p4: depot-paths = "//depot/solvespace/": change = 1914]
2009-02-16 12:05:08 +00:00
if(bkwds) {
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
inter.AddEdge(tb, ta, sc->h.v, 1);
Trim trailing whitespace.
2015-03-29 00:30:52 +00:00
} else {
Additional poking at Booleans. At least this is a halfway rational way to think about the cases; I'm classifying the regions to the left and right of each edge, and keeping the edges if those regions (2d, surfaces) classify different. Still screws up with edge-on-edge intersections; but if I make the surface intersection stuff handle that, then might be more straightforward to use that info. [git-p4: depot-paths = "//depot/solvespace/": change = 1914]
2009-02-16 12:05:08 +00:00
inter.AddEdge(ta, tb, sc->h.v, 0);
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
}
}
}
}
}
Group edges into chains (that don't intersect edges from the other contour, except at the ends of the chain), and classify the entire chain. That's much faster than going edge by edge. [git-p4: depot-paths = "//depot/solvespace/": change = 2002]
2009-06-27 05:53:56 +00:00
// Record all the points where more than two edges join, which I will call
// the choosing points. If two edges join at a non-choosing point, then
// they must either both be kept or both be discarded (since that would
// otherwise create an open contour).
Replace all ZERO and memset with C++11 brace-initialization. This will allow us to use non-POD classes inside these objects in future and is otherwise functionally equivalent, as well as more concise. Note that there are some subtleties with handling of brace-initialization. Specifically: On aggregates (e.g. simple C-style structures) using an empty brace-initializer zero-initializes the aggregate, i.e. it makes all members zero. On non-aggregates an empty brace-initializer calls the default constructor. And if the constructor doesn't explicitly initialize the members (which the auto-generated constructor doesn't) then the members will be constructed but otherwise uninitialized. So, what is an aggregate class? To quote the C++ standard (C++03 8.5.1 §1): An aggregate is an array or a class (clause 9) with no user-declared constructors (12.1), no private or protected non-static data members (clause 11), no base classes (clause 10), and no virtual functions (10.3). In SolveSpace, we only have to handle the case of base classes; Constraint and Entity have those. Thus, they had to gain a default constructor that does nothing but initializes the members to zero.
2015-03-27 15:31:23 +00:00
SPointList choosing = {};
Group edges into chains (that don't intersect edges from the other contour, except at the ends of the chain), and classify the entire chain. That's much faster than going edge by edge. [git-p4: depot-paths = "//depot/solvespace/": change = 2002]
2009-06-27 05:53:56 +00:00
SEdge *se;
for(se = orig.l.First(); se; se = orig.l.NextAfter(se)) {
choosing.IncrementTagFor(se->a);
choosing.IncrementTagFor(se->b);
}
for(se = inter.l.First(); se; se = inter.l.NextAfter(se)) {
choosing.IncrementTagFor(se->a);
choosing.IncrementTagFor(se->b);
}
SPoint *sp;
for(sp = choosing.l.First(); sp; sp = choosing.l.NextAfter(sp)) {
if(sp->tag == 2) {
sp->tag = 1;
} else {
sp->tag = 0;
}
}
choosing.l.RemoveTagged();
// The list of edges to trim our new surface, a combination of edges from
// our original and intersecting edge lists.
Replace all ZERO and memset with C++11 brace-initialization. This will allow us to use non-POD classes inside these objects in future and is otherwise functionally equivalent, as well as more concise. Note that there are some subtleties with handling of brace-initialization. Specifically: On aggregates (e.g. simple C-style structures) using an empty brace-initializer zero-initializes the aggregate, i.e. it makes all members zero. On non-aggregates an empty brace-initializer calls the default constructor. And if the constructor doesn't explicitly initialize the members (which the auto-generated constructor doesn't) then the members will be constructed but otherwise uninitialized. So, what is an aggregate class? To quote the C++ standard (C++03 8.5.1 §1): An aggregate is an array or a class (clause 9) with no user-declared constructors (12.1), no private or protected non-static data members (clause 11), no base classes (clause 10), and no virtual functions (10.3). In SolveSpace, we only have to handle the case of base classes; Constraint and Entity have those. Thus, they had to gain a default constructor that does nothing but initializes the members to zero.
2015-03-27 15:31:23 +00:00
SEdgeList final = {};
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
Group edges into chains (that don't intersect edges from the other contour, except at the ends of the chain), and classify the entire chain. That's much faster than going edge by edge. [git-p4: depot-paths = "//depot/solvespace/": change = 2002]
2009-06-27 05:53:56 +00:00
while(orig.l.n > 0) {
Replace all ZERO and memset with C++11 brace-initialization. This will allow us to use non-POD classes inside these objects in future and is otherwise functionally equivalent, as well as more concise. Note that there are some subtleties with handling of brace-initialization. Specifically: On aggregates (e.g. simple C-style structures) using an empty brace-initializer zero-initializes the aggregate, i.e. it makes all members zero. On non-aggregates an empty brace-initializer calls the default constructor. And if the constructor doesn't explicitly initialize the members (which the auto-generated constructor doesn't) then the members will be constructed but otherwise uninitialized. So, what is an aggregate class? To quote the C++ standard (C++03 8.5.1 §1): An aggregate is an array or a class (clause 9) with no user-declared constructors (12.1), no private or protected non-static data members (clause 11), no base classes (clause 10), and no virtual functions (10.3). In SolveSpace, we only have to handle the case of base classes; Constraint and Entity have those. Thus, they had to gain a default constructor that does nothing but initializes the members to zero.
2015-03-27 15:31:23 +00:00
SEdgeList chain = {};
Group edges into chains (that don't intersect edges from the other contour, except at the ends of the chain), and classify the entire chain. That's much faster than going edge by edge. [git-p4: depot-paths = "//depot/solvespace/": change = 2002]
2009-06-27 05:53:56 +00:00
FindChainAvoiding(&orig, &chain, &choosing);
// Arbitrarily choose an edge within the chain to classify; they
// should all be the same, though.
se = &(chain.l.elem[chain.l.n/2]);
Revamp the edge classification for Booleans. I no longer make a separate polygon of coincident (with same or opposite normal) faces; I instead test all the edges against the other shell, and have extended the classify-against-shell stuff to handle those cases. And the normals are now perturbed a bit numerically, to either side of the edge, to distinguish tangency from a coincident surface. This seems to work fairly well, although things still tend to fail when the piecewise linear tolerance is too coarse. [git-p4: depot-paths = "//depot/solvespace/": change = 1964]
2009-06-04 03:59:40 +00:00
Point2d auv = (se->a).ProjectXy(),
buv = (se->b).ProjectXy();
Make Boolean union work when the shells have coincident plane faces. Still on planes only, no curved surface intersections. [git-p4: depot-paths = "//depot/solvespace/": change = 1912]
2009-02-09 12:40:48 +00:00
Revamp the edge classification for Booleans. I no longer make a separate polygon of coincident (with same or opposite normal) faces; I instead test all the edges against the other shell, and have extended the classify-against-shell stuff to handle those cases. And the normals are now perturbed a bit numerically, to either side of the edge, to distinguish tangency from a coincident surface. This seems to work fairly well, although things still tend to fail when the piecewise linear tolerance is too coarse. [git-p4: depot-paths = "//depot/solvespace/": change = 1964]
2009-06-04 03:59:40 +00:00
Vector pt, enin, enout, surfn;
ret.EdgeNormalsWithinSurface(auv, buv, &pt, &enin, &enout, &surfn,
Add beginnings of marching surface intersection; I can find all the boundary points, at least. That required some changes to what gets passed around (for example because to project a point onto this inexact curve, we need to know which two surfaces it trims so that we can do a Newton's method on them). And fix stupidity in the way that I calculated edge normals; I just did normal in uv space, and there's no particular reason why that would be normal in xyz. So edges in long skinny surfaces failed, for example. [git-p4: depot-paths = "//depot/solvespace/": change = 1990]
2009-06-19 07:56:33 +00:00
se->auxA, into, sha, shb);
Revamp the edge classification for Booleans. I no longer make a separate polygon of coincident (with same or opposite normal) faces; I instead test all the edges against the other shell, and have extended the classify-against-shell stuff to handle those cases. And the normals are now perturbed a bit numerically, to either side of the edge, to distinguish tangency from a coincident surface. This seems to work fairly well, although things still tend to fail when the piecewise linear tolerance is too coarse. [git-p4: depot-paths = "//depot/solvespace/": change = 1964]
2009-06-04 03:59:40 +00:00
Convert all enumerations to use `enum class`. Specifically, take the old code that looks like this: class Foo { enum { X = 1, Y = 2 }; int kind; } ... foo.kind = Foo::X; ... and convert it to this: class Foo { enum class Kind : uint32_t { X = 1, Y = 2 }; Kind kind; } ... foo.kind = Foo::Kind::X; (In some cases the enumeration would not be in the class namespace, such as when it is generally useful.) The benefits are as follows: * The type of the field gives a clear indication of intent, both to humans and tools (such as binding generators). * The compiler is able to automatically warn when a switch is not exhaustive; but this is currently suppressed by the default: ssassert(false, ...) idiom. * Integers and plain enums are weakly type checked: they implicitly convert into each other. This can hide bugs where type conversion is performed but not intended. Enum classes are strongly type checked. * Plain enums pollute parent namespaces; enum classes do not. Almost every defined enum we have already has a kind of ad-hoc namespacing via `NAMESPACE_`, which is now explicit. * Plain enums do not have a well-defined ABI size, which is important for bindings. Enum classes can have it, if specified. We specify the base type for all enums as uint32_t, which is a safe choice and allows us to not change the numeric values of any variants. This commit introduces absolutely no functional change to the code, just renaming and change of types. It handles almost all cases, except GraphicsWindow::pending.operation, which needs minor functional change.
2016-05-20 08:31:20 +00:00
SShell::Class indir_shell, outdir_shell, indir_orig, outdir_orig;
Revamp the edge classification for Booleans. I no longer make a separate polygon of coincident (with same or opposite normal) faces; I instead test all the edges against the other shell, and have extended the classify-against-shell stuff to handle those cases. And the normals are now perturbed a bit numerically, to either side of the edge, to distinguish tangency from a coincident surface. This seems to work fairly well, although things still tend to fail when the piecewise linear tolerance is too coarse. [git-p4: depot-paths = "//depot/solvespace/": change = 1964]
2009-06-04 03:59:40 +00:00
Convert all enumerations to use `enum class`. Specifically, take the old code that looks like this: class Foo { enum { X = 1, Y = 2 }; int kind; } ... foo.kind = Foo::X; ... and convert it to this: class Foo { enum class Kind : uint32_t { X = 1, Y = 2 }; Kind kind; } ... foo.kind = Foo::Kind::X; (In some cases the enumeration would not be in the class namespace, such as when it is generally useful.) The benefits are as follows: * The type of the field gives a clear indication of intent, both to humans and tools (such as binding generators). * The compiler is able to automatically warn when a switch is not exhaustive; but this is currently suppressed by the default: ssassert(false, ...) idiom. * Integers and plain enums are weakly type checked: they implicitly convert into each other. This can hide bugs where type conversion is performed but not intended. Enum classes are strongly type checked. * Plain enums pollute parent namespaces; enum classes do not. Almost every defined enum we have already has a kind of ad-hoc namespacing via `NAMESPACE_`, which is now explicit. * Plain enums do not have a well-defined ABI size, which is important for bindings. Enum classes can have it, if specified. We specify the base type for all enums as uint32_t, which is a safe choice and allows us to not change the numeric values of any variants. This commit introduces absolutely no functional change to the code, just renaming and change of types. It handles almost all cases, except GraphicsWindow::pending.operation, which needs minor functional change.
2016-05-20 08:31:20 +00:00
indir_orig = SShell::Class::INSIDE;
outdir_orig = SShell::Class::OUTSIDE;
Revamp the edge classification for Booleans. I no longer make a separate polygon of coincident (with same or opposite normal) faces; I instead test all the edges against the other shell, and have extended the classify-against-shell stuff to handle those cases. And the normals are now perturbed a bit numerically, to either side of the edge, to distinguish tangency from a coincident surface. This seems to work fairly well, although things still tend to fail when the piecewise linear tolerance is too coarse. [git-p4: depot-paths = "//depot/solvespace/": change = 1964]
2009-06-04 03:59:40 +00:00
agnst->ClassifyEdge(&indir_shell, &outdir_shell,
Trim trailing whitespace.
2015-03-29 00:30:52 +00:00
ret.PointAt(auv), ret.PointAt(buv), pt,
Revamp the edge classification for Booleans. I no longer make a separate polygon of coincident (with same or opposite normal) faces; I instead test all the edges against the other shell, and have extended the classify-against-shell stuff to handle those cases. And the normals are now perturbed a bit numerically, to either side of the edge, to distinguish tangency from a coincident surface. This seems to work fairly well, although things still tend to fail when the piecewise linear tolerance is too coarse. [git-p4: depot-paths = "//depot/solvespace/": change = 1964]
2009-06-04 03:59:40 +00:00
enin, enout, surfn);
Many changes: * Rewrite surface handles in curves, so that Booleans beyond the first don't screw up. * If an intersection curve is identical to an existing curve (as happens when faces are coincident), take the piecewise linearization of the existing curve; this stops us from screwing up when different shells are pwl'd at different chord tols. * Hook up the plane faces again. * Remove coincident (parallel or anti-parallel) edges from the coincident-face edge lists when doing Booleans; those may happen if two faces are coincident with ours. * Miscellaneous bugfixes. It doesn't seem to screw up very much now, although tangent edges (and insufficient pwl resolution) may still cause problems. [git-p4: depot-paths = "//depot/solvespace/": change = 1929]
2009-03-15 23:04:45 +00:00
Trim trailing whitespace.
2015-03-29 00:30:52 +00:00
if(KeepEdge(type, opA, indir_shell, outdir_shell,
Revamp the edge classification for Booleans. I no longer make a separate polygon of coincident (with same or opposite normal) faces; I instead test all the edges against the other shell, and have extended the classify-against-shell stuff to handle those cases. And the normals are now perturbed a bit numerically, to either side of the edge, to distinguish tangency from a coincident surface. This seems to work fairly well, although things still tend to fail when the piecewise linear tolerance is too coarse. [git-p4: depot-paths = "//depot/solvespace/": change = 1964]
2009-06-04 03:59:40 +00:00
indir_orig, outdir_orig))
{
Group edges into chains (that don't intersect edges from the other contour, except at the ends of the chain), and classify the entire chain. That's much faster than going edge by edge. [git-p4: depot-paths = "//depot/solvespace/": change = 2002]
2009-06-27 05:53:56 +00:00
for(se = chain.l.First(); se; se = chain.l.NextAfter(se)) {
final.AddEdge(se->a, se->b, se->auxA, se->auxB);
}
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
}
Group edges into chains (that don't intersect edges from the other contour, except at the ends of the chain), and classify the entire chain. That's much faster than going edge by edge. [git-p4: depot-paths = "//depot/solvespace/": change = 2002]
2009-06-27 05:53:56 +00:00
chain.Clear();
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
}
Group edges into chains (that don't intersect edges from the other contour, except at the ends of the chain), and classify the entire chain. That's much faster than going edge by edge. [git-p4: depot-paths = "//depot/solvespace/": change = 2002]
2009-06-27 05:53:56 +00:00
while(inter.l.n > 0) {
Replace all ZERO and memset with C++11 brace-initialization. This will allow us to use non-POD classes inside these objects in future and is otherwise functionally equivalent, as well as more concise. Note that there are some subtleties with handling of brace-initialization. Specifically: On aggregates (e.g. simple C-style structures) using an empty brace-initializer zero-initializes the aggregate, i.e. it makes all members zero. On non-aggregates an empty brace-initializer calls the default constructor. And if the constructor doesn't explicitly initialize the members (which the auto-generated constructor doesn't) then the members will be constructed but otherwise uninitialized. So, what is an aggregate class? To quote the C++ standard (C++03 8.5.1 §1): An aggregate is an array or a class (clause 9) with no user-declared constructors (12.1), no private or protected non-static data members (clause 11), no base classes (clause 10), and no virtual functions (10.3). In SolveSpace, we only have to handle the case of base classes; Constraint and Entity have those. Thus, they had to gain a default constructor that does nothing but initializes the members to zero.
2015-03-27 15:31:23 +00:00
SEdgeList chain = {};
Group edges into chains (that don't intersect edges from the other contour, except at the ends of the chain), and classify the entire chain. That's much faster than going edge by edge. [git-p4: depot-paths = "//depot/solvespace/": change = 2002]
2009-06-27 05:53:56 +00:00
FindChainAvoiding(&inter, &chain, &choosing);
// Any edge in the chain, same as above.
se = &(chain.l.elem[chain.l.n/2]);
Trim trailing whitespace.
2015-03-29 00:30:52 +00:00
Make Boolean union work when the shells have coincident plane faces. Still on planes only, no curved surface intersections. [git-p4: depot-paths = "//depot/solvespace/": change = 1912]
2009-02-09 12:40:48 +00:00
Point2d auv = (se->a).ProjectXy(),
buv = (se->b).ProjectXy();
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
Revamp the edge classification for Booleans. I no longer make a separate polygon of coincident (with same or opposite normal) faces; I instead test all the edges against the other shell, and have extended the classify-against-shell stuff to handle those cases. And the normals are now perturbed a bit numerically, to either side of the edge, to distinguish tangency from a coincident surface. This seems to work fairly well, although things still tend to fail when the piecewise linear tolerance is too coarse. [git-p4: depot-paths = "//depot/solvespace/": change = 1964]
2009-06-04 03:59:40 +00:00
Vector pt, enin, enout, surfn;
ret.EdgeNormalsWithinSurface(auv, buv, &pt, &enin, &enout, &surfn,
Add beginnings of marching surface intersection; I can find all the boundary points, at least. That required some changes to what gets passed around (for example because to project a point onto this inexact curve, we need to know which two surfaces it trims so that we can do a Newton's method on them). And fix stupidity in the way that I calculated edge normals; I just did normal in uv space, and there's no particular reason why that would be normal in xyz. So edges in long skinny surfaces failed, for example. [git-p4: depot-paths = "//depot/solvespace/": change = 1990]
2009-06-19 07:56:33 +00:00
se->auxA, into, sha, shb);
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
Convert all enumerations to use `enum class`. Specifically, take the old code that looks like this: class Foo { enum { X = 1, Y = 2 }; int kind; } ... foo.kind = Foo::X; ... and convert it to this: class Foo { enum class Kind : uint32_t { X = 1, Y = 2 }; Kind kind; } ... foo.kind = Foo::Kind::X; (In some cases the enumeration would not be in the class namespace, such as when it is generally useful.) The benefits are as follows: * The type of the field gives a clear indication of intent, both to humans and tools (such as binding generators). * The compiler is able to automatically warn when a switch is not exhaustive; but this is currently suppressed by the default: ssassert(false, ...) idiom. * Integers and plain enums are weakly type checked: they implicitly convert into each other. This can hide bugs where type conversion is performed but not intended. Enum classes are strongly type checked. * Plain enums pollute parent namespaces; enum classes do not. Almost every defined enum we have already has a kind of ad-hoc namespacing via `NAMESPACE_`, which is now explicit. * Plain enums do not have a well-defined ABI size, which is important for bindings. Enum classes can have it, if specified. We specify the base type for all enums as uint32_t, which is a safe choice and allows us to not change the numeric values of any variants. This commit introduces absolutely no functional change to the code, just renaming and change of types. It handles almost all cases, except GraphicsWindow::pending.operation, which needs minor functional change.
2016-05-20 08:31:20 +00:00
SShell::Class indir_shell, outdir_shell, indir_orig, outdir_orig;
Additional poking at Booleans. At least this is a halfway rational way to think about the cases; I'm classifying the regions to the left and right of each edge, and keeping the edges if those regions (2d, surfaces) classify different. Still screws up with edge-on-edge intersections; but if I make the surface intersection stuff handle that, then might be more straightforward to use that info. [git-p4: depot-paths = "//depot/solvespace/": change = 1914]
2009-02-16 12:05:08 +00:00
Convert all enumerations to use `enum class`. Specifically, take the old code that looks like this: class Foo { enum { X = 1, Y = 2 }; int kind; } ... foo.kind = Foo::X; ... and convert it to this: class Foo { enum class Kind : uint32_t { X = 1, Y = 2 }; Kind kind; } ... foo.kind = Foo::Kind::X; (In some cases the enumeration would not be in the class namespace, such as when it is generally useful.) The benefits are as follows: * The type of the field gives a clear indication of intent, both to humans and tools (such as binding generators). * The compiler is able to automatically warn when a switch is not exhaustive; but this is currently suppressed by the default: ssassert(false, ...) idiom. * Integers and plain enums are weakly type checked: they implicitly convert into each other. This can hide bugs where type conversion is performed but not intended. Enum classes are strongly type checked. * Plain enums pollute parent namespaces; enum classes do not. Almost every defined enum we have already has a kind of ad-hoc namespacing via `NAMESPACE_`, which is now explicit. * Plain enums do not have a well-defined ABI size, which is important for bindings. Enum classes can have it, if specified. We specify the base type for all enums as uint32_t, which is a safe choice and allows us to not change the numeric values of any variants. This commit introduces absolutely no functional change to the code, just renaming and change of types. It handles almost all cases, except GraphicsWindow::pending.operation, which needs minor functional change.
2016-05-20 08:31:20 +00:00
SBspUv::Class c_this = (origBsp) ? origBsp->ClassifyEdge(auv, buv, &ret) : SBspUv::Class::OUTSIDE;
Revamp the edge classification for Booleans. I no longer make a separate polygon of coincident (with same or opposite normal) faces; I instead test all the edges against the other shell, and have extended the classify-against-shell stuff to handle those cases. And the normals are now perturbed a bit numerically, to either side of the edge, to distinguish tangency from a coincident surface. This seems to work fairly well, although things still tend to fail when the piecewise linear tolerance is too coarse. [git-p4: depot-paths = "//depot/solvespace/": change = 1964]
2009-06-04 03:59:40 +00:00
TagByClassifiedEdge(c_this, &indir_orig, &outdir_orig);
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
Revamp the edge classification for Booleans. I no longer make a separate polygon of coincident (with same or opposite normal) faces; I instead test all the edges against the other shell, and have extended the classify-against-shell stuff to handle those cases. And the normals are now perturbed a bit numerically, to either side of the edge, to distinguish tangency from a coincident surface. This seems to work fairly well, although things still tend to fail when the piecewise linear tolerance is too coarse. [git-p4: depot-paths = "//depot/solvespace/": change = 1964]
2009-06-04 03:59:40 +00:00
agnst->ClassifyEdge(&indir_shell, &outdir_shell,
Trim trailing whitespace.
2015-03-29 00:30:52 +00:00
ret.PointAt(auv), ret.PointAt(buv), pt,
Revamp the edge classification for Booleans. I no longer make a separate polygon of coincident (with same or opposite normal) faces; I instead test all the edges against the other shell, and have extended the classify-against-shell stuff to handle those cases. And the normals are now perturbed a bit numerically, to either side of the edge, to distinguish tangency from a coincident surface. This seems to work fairly well, although things still tend to fail when the piecewise linear tolerance is too coarse. [git-p4: depot-paths = "//depot/solvespace/": change = 1964]
2009-06-04 03:59:40 +00:00
enin, enout, surfn);
Trim trailing whitespace.
2015-03-29 00:30:52 +00:00
if(KeepEdge(type, opA, indir_shell, outdir_shell,
Revamp the edge classification for Booleans. I no longer make a separate polygon of coincident (with same or opposite normal) faces; I instead test all the edges against the other shell, and have extended the classify-against-shell stuff to handle those cases. And the normals are now perturbed a bit numerically, to either side of the edge, to distinguish tangency from a coincident surface. This seems to work fairly well, although things still tend to fail when the piecewise linear tolerance is too coarse. [git-p4: depot-paths = "//depot/solvespace/": change = 1964]
2009-06-04 03:59:40 +00:00
indir_orig, outdir_orig))
{
Group edges into chains (that don't intersect edges from the other contour, except at the ends of the chain), and classify the entire chain. That's much faster than going edge by edge. [git-p4: depot-paths = "//depot/solvespace/": change = 2002]
2009-06-27 05:53:56 +00:00
for(se = chain.l.First(); se; se = chain.l.NextAfter(se)) {
final.AddEdge(se->a, se->b, se->auxA, se->auxB);
}
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
}
Group edges into chains (that don't intersect edges from the other contour, except at the ends of the chain), and classify the entire chain. That's much faster than going edge by edge. [git-p4: depot-paths = "//depot/solvespace/": change = 2002]
2009-06-27 05:53:56 +00:00
chain.Clear();
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
}
Many changes: * Rewrite surface handles in curves, so that Booleans beyond the first don't screw up. * If an intersection curve is identical to an existing curve (as happens when faces are coincident), take the piecewise linearization of the existing curve; this stops us from screwing up when different shells are pwl'd at different chord tols. * Hook up the plane faces again. * Remove coincident (parallel or anti-parallel) edges from the coincident-face edge lists when doing Booleans; those may happen if two faces are coincident with ours. * Miscellaneous bugfixes. It doesn't seem to screw up very much now, although tangent edges (and insufficient pwl resolution) may still cause problems. [git-p4: depot-paths = "//depot/solvespace/": change = 1929]
2009-03-15 23:04:45 +00:00
// Cull extraneous edges; duplicates or anti-parallel pairs. In particular,
// we can get duplicate edges if our surface intersects the other shell
// at an edge, so that both surfaces intersect coincident (and both
// generate an intersection edge).
final.CullExtraneousEdges();
More coincident fixing; test for edge-on-edge, fix some gross stupidity. [git-p4: depot-paths = "//depot/solvespace/": change = 1915]
2009-02-17 11:17:12 +00:00
Additional poking at Booleans. At least this is a halfway rational way to think about the cases; I'm classifying the regions to the left and right of each edge, and keeping the edges if those regions (2d, surfaces) classify different. Still screws up with edge-on-edge intersections; but if I make the surface intersection stuff handle that, then might be more straightforward to use that info. [git-p4: depot-paths = "//depot/solvespace/": change = 1914]
2009-02-16 12:05:08 +00:00
// Use our reassembled edges to trim the new surface.
Add code to identify planes and cylindrical surfaces from a solid of revolution, and put them in the same form as if they had been draw by an extrusion (so that we can use all the same special case intersection curves). And add code to merge coincident faces into one. That turns out to be more than a cosmetic/efficiency thing, since edge splitting fails at the join between two coincident faces. [git-p4: depot-paths = "//depot/solvespace/": change = 1965]
2009-06-05 05:38:41 +00:00
ret.TrimFromEdgeList(&final, true);
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
Replace all ZERO and memset with C++11 brace-initialization. This will allow us to use non-POD classes inside these objects in future and is otherwise functionally equivalent, as well as more concise. Note that there are some subtleties with handling of brace-initialization. Specifically: On aggregates (e.g. simple C-style structures) using an empty brace-initializer zero-initializes the aggregate, i.e. it makes all members zero. On non-aggregates an empty brace-initializer calls the default constructor. And if the constructor doesn't explicitly initialize the members (which the auto-generated constructor doesn't) then the members will be constructed but otherwise uninitialized. So, what is an aggregate class? To quote the C++ standard (C++03 8.5.1 §1): An aggregate is an array or a class (clause 9) with no user-declared constructors (12.1), no private or protected non-static data members (clause 11), no base classes (clause 10), and no virtual functions (10.3). In SolveSpace, we only have to handle the case of base classes; Constraint and Entity have those. Thus, they had to gain a default constructor that does nothing but initializes the members to zero.
2015-03-27 15:31:23 +00:00
SPolygon poly = {};
Many changes: * Rewrite surface handles in curves, so that Booleans beyond the first don't screw up. * If an intersection curve is identical to an existing curve (as happens when faces are coincident), take the piecewise linearization of the existing curve; this stops us from screwing up when different shells are pwl'd at different chord tols. * Hook up the plane faces again. * Remove coincident (parallel or anti-parallel) edges from the coincident-face edge lists when doing Booleans; those may happen if two faces are coincident with ours. * Miscellaneous bugfixes. It doesn't seem to screw up very much now, although tangent edges (and insufficient pwl resolution) may still cause problems. [git-p4: depot-paths = "//depot/solvespace/": change = 1929]
2009-03-15 23:04:45 +00:00
final.l.ClearTags();
if(!final.AssemblePolygon(&poly, NULL, true)) {
If a Boolean fails, then make a note of it in the group's text window screen, and remind the user that they could 'fix' the problem by working with meshes instead. [git-p4: depot-paths = "//depot/solvespace/": change = 1962]
2009-05-30 08:49:09 +00:00
into->booleanFailed = true;
Group edges into chains (that don't intersect edges from the other contour, except at the ends of the chain), and classify the entire chain. That's much faster than going edge by edge. [git-p4: depot-paths = "//depot/solvespace/": change = 2002]
2009-06-27 05:53:56 +00:00
dbp("failed: I=%d, avoid=%d", I, choosing.l.n);
Add a special case for line-cylinder intersection, solving in closed form. This is a fairly good speedup, and handles tangency well. But that shows that tangency has other problems; need to classify edges correctly (whether they point to a coincident surface) in curved surfaces too. I need to tweak SShell::ClassifyPoint(). [git-p4: depot-paths = "//depot/solvespace/": change = 1933]
2009-03-19 17:40:11 +00:00
DEBUGEDGELIST(&final, &ret);
Many changes: * Rewrite surface handles in curves, so that Booleans beyond the first don't screw up. * If an intersection curve is identical to an existing curve (as happens when faces are coincident), take the piecewise linearization of the existing curve; this stops us from screwing up when different shells are pwl'd at different chord tols. * Hook up the plane faces again. * Remove coincident (parallel or anti-parallel) edges from the coincident-face edge lists when doing Booleans; those may happen if two faces are coincident with ours. * Miscellaneous bugfixes. It doesn't seem to screw up very much now, although tangent edges (and insufficient pwl resolution) may still cause problems. [git-p4: depot-paths = "//depot/solvespace/": change = 1929]
2009-03-15 23:04:45 +00:00
}
poly.Clear();
Group edges into chains (that don't intersect edges from the other contour, except at the ends of the chain), and classify the entire chain. That's much faster than going edge by edge. [git-p4: depot-paths = "//depot/solvespace/": change = 2002]
2009-06-27 05:53:56 +00:00
choosing.Clear();
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
final.Clear();
inter.Clear();
orig.Clear();
Preparatory work for Boolean. Make the u and v coordinates of the trim curves for all surfaces lie between 0 and 1. And add routines to merge the curves and surfaces from two shells into one, and to split the trim curves into their piecewise linear segments and then reassemble them into trim curves. [git-p4: depot-paths = "//depot/solvespace/": change = 1905]
2009-01-25 11:52:29 +00:00
return ret;
}
Convert all enumerations to use `enum class`. Specifically, take the old code that looks like this: class Foo { enum { X = 1, Y = 2 }; int kind; } ... foo.kind = Foo::X; ... and convert it to this: class Foo { enum class Kind : uint32_t { X = 1, Y = 2 }; Kind kind; } ... foo.kind = Foo::Kind::X; (In some cases the enumeration would not be in the class namespace, such as when it is generally useful.) The benefits are as follows: * The type of the field gives a clear indication of intent, both to humans and tools (such as binding generators). * The compiler is able to automatically warn when a switch is not exhaustive; but this is currently suppressed by the default: ssassert(false, ...) idiom. * Integers and plain enums are weakly type checked: they implicitly convert into each other. This can hide bugs where type conversion is performed but not intended. Enum classes are strongly type checked. * Plain enums pollute parent namespaces; enum classes do not. Almost every defined enum we have already has a kind of ad-hoc namespacing via `NAMESPACE_`, which is now explicit. * Plain enums do not have a well-defined ABI size, which is important for bindings. Enum classes can have it, if specified. We specify the base type for all enums as uint32_t, which is a safe choice and allows us to not change the numeric values of any variants. This commit introduces absolutely no functional change to the code, just renaming and change of types. It handles almost all cases, except GraphicsWindow::pending.operation, which needs minor functional change.
2016-05-20 08:31:20 +00:00
void SShell::CopySurfacesTrimAgainst(SShell *sha, SShell *shb, SShell *into, SSurface::CombineAs type) {
Preparatory work for Boolean. Make the u and v coordinates of the trim curves for all surfaces lie between 0 and 1. And add routines to merge the curves and surfaces from two shells into one, and to split the trim curves into their piecewise linear segments and then reassemble them into trim curves. [git-p4: depot-paths = "//depot/solvespace/": change = 1905]
2009-01-25 11:52:29 +00:00
SSurface *ss;
for(ss = surface.First(); ss; ss = surface.NextAfter(ss)) {
SSurface ssn;
Add beginnings of marching surface intersection; I can find all the boundary points, at least. That required some changes to what gets passed around (for example because to project a point onto this inexact curve, we need to know which two surfaces it trims so that we can do a Newton's method on them). And fix stupidity in the way that I calculated edge normals; I just did normal in uv space, and there's no particular reason why that would be normal in xyz. So edges in long skinny surfaces failed, for example. [git-p4: depot-paths = "//depot/solvespace/": change = 1990]
2009-06-19 07:56:33 +00:00
ssn = ss->MakeCopyTrimAgainst(this, sha, shb, into, type);
Many changes: * Rewrite surface handles in curves, so that Booleans beyond the first don't screw up. * If an intersection curve is identical to an existing curve (as happens when faces are coincident), take the piecewise linearization of the existing curve; this stops us from screwing up when different shells are pwl'd at different chord tols. * Hook up the plane faces again. * Remove coincident (parallel or anti-parallel) edges from the coincident-face edge lists when doing Booleans; those may happen if two faces are coincident with ours. * Miscellaneous bugfixes. It doesn't seem to screw up very much now, although tangent edges (and insufficient pwl resolution) may still cause problems. [git-p4: depot-paths = "//depot/solvespace/": change = 1929]
2009-03-15 23:04:45 +00:00
ss->newH = into->surface.AddAndAssignId(&ssn);
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
I++;
Preparatory work for Boolean. Make the u and v coordinates of the trim curves for all surfaces lie between 0 and 1. And add routines to merge the curves and surfaces from two shells into one, and to split the trim curves into their piecewise linear segments and then reassemble them into trim curves. [git-p4: depot-paths = "//depot/solvespace/": change = 1905]
2009-01-25 11:52:29 +00:00
}
}
More poking at Booleans; generate the unsplit intersection curves for planes against planes. [git-p4: depot-paths = "//depot/solvespace/": change = 1907]
2009-01-27 07:59:58 +00:00
void SShell::MakeIntersectionCurvesAgainst(SShell *agnst, SShell *into) {
SSurface *sa;
Many changes: * Rewrite surface handles in curves, so that Booleans beyond the first don't screw up. * If an intersection curve is identical to an existing curve (as happens when faces are coincident), take the piecewise linearization of the existing curve; this stops us from screwing up when different shells are pwl'd at different chord tols. * Hook up the plane faces again. * Remove coincident (parallel or anti-parallel) edges from the coincident-face edge lists when doing Booleans; those may happen if two faces are coincident with ours. * Miscellaneous bugfixes. It doesn't seem to screw up very much now, although tangent edges (and insufficient pwl resolution) may still cause problems. [git-p4: depot-paths = "//depot/solvespace/": change = 1929]
2009-03-15 23:04:45 +00:00
for(sa = surface.First(); sa; sa = surface.NextAfter(sa)) {
More poking at Booleans; generate the unsplit intersection curves for planes against planes. [git-p4: depot-paths = "//depot/solvespace/": change = 1907]
2009-01-27 07:59:58 +00:00
SSurface *sb;
Many changes: * Rewrite surface handles in curves, so that Booleans beyond the first don't screw up. * If an intersection curve is identical to an existing curve (as happens when faces are coincident), take the piecewise linearization of the existing curve; this stops us from screwing up when different shells are pwl'd at different chord tols. * Hook up the plane faces again. * Remove coincident (parallel or anti-parallel) edges from the coincident-face edge lists when doing Booleans; those may happen if two faces are coincident with ours. * Miscellaneous bugfixes. It doesn't seem to screw up very much now, although tangent edges (and insufficient pwl resolution) may still cause problems. [git-p4: depot-paths = "//depot/solvespace/": change = 1929]
2009-03-15 23:04:45 +00:00
for(sb = agnst->surface.First(); sb; sb = agnst->surface.NextAfter(sb)){
More poking at Booleans; generate the unsplit intersection curves for planes against planes. [git-p4: depot-paths = "//depot/solvespace/": change = 1907]
2009-01-27 07:59:58 +00:00
// Intersect every surface from our shell against every surface
// from agnst; this will add zero or more curves to the curve
// list for into.
Many changes: * Rewrite surface handles in curves, so that Booleans beyond the first don't screw up. * If an intersection curve is identical to an existing curve (as happens when faces are coincident), take the piecewise linearization of the existing curve; this stops us from screwing up when different shells are pwl'd at different chord tols. * Hook up the plane faces again. * Remove coincident (parallel or anti-parallel) edges from the coincident-face edge lists when doing Booleans; those may happen if two faces are coincident with ours. * Miscellaneous bugfixes. It doesn't seem to screw up very much now, although tangent edges (and insufficient pwl resolution) may still cause problems. [git-p4: depot-paths = "//depot/solvespace/": change = 1929]
2009-03-15 23:04:45 +00:00
sa->IntersectAgainst(sb, this, agnst, into);
More poking at Booleans; generate the unsplit intersection curves for planes against planes. [git-p4: depot-paths = "//depot/solvespace/": change = 1907]
2009-01-27 07:59:58 +00:00
}
}
}
Rewrite declarations of form f(void) as f(). In C++ there is no difference and newly added functions are all declared as f(), so this brings back consistency.
2016-05-05 05:54:05 +00:00
void SShell::CleanupAfterBoolean() {
More poking at Booleans; generate the unsplit intersection curves for planes against planes. [git-p4: depot-paths = "//depot/solvespace/": change = 1907]
2009-01-27 07:59:58 +00:00
SSurface *ss;
for(ss = surface.First(); ss; ss = surface.NextAfter(ss)) {
Revamp the edge classification for Booleans. I no longer make a separate polygon of coincident (with same or opposite normal) faces; I instead test all the edges against the other shell, and have extended the classify-against-shell stuff to handle those cases. And the normals are now perturbed a bit numerically, to either side of the edge, to distinguish tangency from a coincident surface. This seems to work fairly well, although things still tend to fail when the piecewise linear tolerance is too coarse. [git-p4: depot-paths = "//depot/solvespace/": change = 1964]
2009-06-04 03:59:40 +00:00
ss->edges.Clear();
More poking at Booleans; generate the unsplit intersection curves for planes against planes. [git-p4: depot-paths = "//depot/solvespace/": change = 1907]
2009-01-27 07:59:58 +00:00
}
}
Add code to assemble two shells into one, without checking for any intersections or otherwise trying to make the result not self-intersecting. [git-p4: depot-paths = "//depot/solvespace/": change = 1955]
2009-05-20 03:04:36 +00:00
//-----------------------------------------------------------------------------
// All curves contain handles to the two surfaces that they trim. After a
// Boolean or assembly, we must rewrite those handles to refer to the curves
Trim trailing whitespace.
2015-03-29 00:30:52 +00:00
// by their new IDs.
Add code to assemble two shells into one, without checking for any intersections or otherwise trying to make the result not self-intersecting. [git-p4: depot-paths = "//depot/solvespace/": change = 1955]
2009-05-20 03:04:36 +00:00
//-----------------------------------------------------------------------------
void SShell::RewriteSurfaceHandlesForCurves(SShell *a, SShell *b) {
SCurve *sc;
for(sc = curve.First(); sc; sc = curve.NextAfter(sc)) {
Add beginnings of marching surface intersection; I can find all the boundary points, at least. That required some changes to what gets passed around (for example because to project a point onto this inexact curve, we need to know which two surfaces it trims so that we can do a Newton's method on them). And fix stupidity in the way that I calculated edge normals; I just did normal in uv space, and there's no particular reason why that would be normal in xyz. So edges in long skinny surfaces failed, for example. [git-p4: depot-paths = "//depot/solvespace/": change = 1990]
2009-06-19 07:56:33 +00:00
sc->surfA = sc->GetSurfaceA(a, b)->newH,
sc->surfB = sc->GetSurfaceB(a, b)->newH;
Add code to assemble two shells into one, without checking for any intersections or otherwise trying to make the result not self-intersecting. [git-p4: depot-paths = "//depot/solvespace/": change = 1955]
2009-05-20 03:04:36 +00:00
}
}
//-----------------------------------------------------------------------------
// Copy all the surfaces and curves from two different shells into a single
// shell. The only difficulty is to rewrite all of their handles; we don't
// look for any surface intersections, so if two objects interfere then the
// result is just self-intersecting. This is used for assembly, since it's
// much faster than merging as union.
//-----------------------------------------------------------------------------
void SShell::MakeFromAssemblyOf(SShell *a, SShell *b) {
If a Boolean fails, then make a note of it in the group's text window screen, and remind the user that they could 'fix' the problem by working with meshes instead. [git-p4: depot-paths = "//depot/solvespace/": change = 1962]
2009-05-30 08:49:09 +00:00
booleanFailed = false;
Add code to assemble two shells into one, without checking for any intersections or otherwise trying to make the result not self-intersecting. [git-p4: depot-paths = "//depot/solvespace/": change = 1955]
2009-05-20 03:04:36 +00:00
Vector t = Vector::From(0, 0, 0);
Quaternion q = Quaternion::IDENTITY;
int i = 0;
SShell *ab;
Trim trailing whitespace.
2015-03-29 00:30:52 +00:00
Add code to assemble two shells into one, without checking for any intersections or otherwise trying to make the result not self-intersecting. [git-p4: depot-paths = "//depot/solvespace/": change = 1955]
2009-05-20 03:04:36 +00:00
// First, copy over all the curves. Note which shell (a or b) each curve
// came from, but assign it a new ID.
SCurve *c, cn;
for(i = 0; i < 2; i++) {
ab = (i == 0) ? a : b;
for(c = ab->curve.First(); c; c = ab->curve.NextAfter(c)) {
A big nasty change, originally just to add paste transformed. So it does that, and adds a scale factor to that transformation (instead of just mirroring, as before), but also: * Replace the "import mirrored" mechanism with a scale factor, which if negative corresponds to a reflection as well. * Fix self-intersection checker to report a meaningful point when edges are collinear. * Don't blow an assertion on some types of invalid file; instead provide a nice error message to the user. * Clear the naked edges before each regen. * Don't create zero-length line segments by snapping a line segment's end to its beginning. [git-p4: depot-paths = "//depot/solvespace/": change = 2086]
2009-12-15 12:26:22 +00:00
cn = SCurve::FromTransformationOf(c, t, q, 1.0);
Convert all enumerations to use `enum class`. Specifically, take the old code that looks like this: class Foo { enum { X = 1, Y = 2 }; int kind; } ... foo.kind = Foo::X; ... and convert it to this: class Foo { enum class Kind : uint32_t { X = 1, Y = 2 }; Kind kind; } ... foo.kind = Foo::Kind::X; (In some cases the enumeration would not be in the class namespace, such as when it is generally useful.) The benefits are as follows: * The type of the field gives a clear indication of intent, both to humans and tools (such as binding generators). * The compiler is able to automatically warn when a switch is not exhaustive; but this is currently suppressed by the default: ssassert(false, ...) idiom. * Integers and plain enums are weakly type checked: they implicitly convert into each other. This can hide bugs where type conversion is performed but not intended. Enum classes are strongly type checked. * Plain enums pollute parent namespaces; enum classes do not. Almost every defined enum we have already has a kind of ad-hoc namespacing via `NAMESPACE_`, which is now explicit. * Plain enums do not have a well-defined ABI size, which is important for bindings. Enum classes can have it, if specified. We specify the base type for all enums as uint32_t, which is a safe choice and allows us to not change the numeric values of any variants. This commit introduces absolutely no functional change to the code, just renaming and change of types. It handles almost all cases, except GraphicsWindow::pending.operation, which needs minor functional change.
2016-05-20 08:31:20 +00:00
cn.source = (i == 0) ? SCurve::Source::A : SCurve::Source::B;
Add code to assemble two shells into one, without checking for any intersections or otherwise trying to make the result not self-intersecting. [git-p4: depot-paths = "//depot/solvespace/": change = 1955]
2009-05-20 03:04:36 +00:00
// surfA and surfB are wrong now, and we can't fix them until
// we've assigned IDs to the surfaces. So we'll get that later.
c->newH = curve.AddAndAssignId(&cn);
}
}
// Likewise copy over all the surfaces.
SSurface *s, sn;
for(i = 0; i < 2; i++) {
ab = (i == 0) ? a : b;
for(s = ab->surface.First(); s; s = ab->surface.NextAfter(s)) {
A big nasty change, originally just to add paste transformed. So it does that, and adds a scale factor to that transformation (instead of just mirroring, as before), but also: * Replace the "import mirrored" mechanism with a scale factor, which if negative corresponds to a reflection as well. * Fix self-intersection checker to report a meaningful point when edges are collinear. * Don't blow an assertion on some types of invalid file; instead provide a nice error message to the user. * Clear the naked edges before each regen. * Don't create zero-length line segments by snapping a line segment's end to its beginning. [git-p4: depot-paths = "//depot/solvespace/": change = 2086]
2009-12-15 12:26:22 +00:00
sn = SSurface::FromTransformationOf(s, t, q, 1.0, true);
Add code to assemble two shells into one, without checking for any intersections or otherwise trying to make the result not self-intersecting. [git-p4: depot-paths = "//depot/solvespace/": change = 1955]
2009-05-20 03:04:36 +00:00
// All the trim curve IDs get rewritten; we know the new handles
// to the curves since we recorded them in the previous step.
STrimBy *stb;
for(stb = sn.trim.First(); stb; stb = sn.trim.NextAfter(stb)) {
stb->curve = ab->curve.FindById(stb->curve)->newH;
}
s->newH = surface.AddAndAssignId(&sn);
}
}
// Finally, rewrite the surfaces associated with each curve to use the
// new handles.
RewriteSurfaceHandlesForCurves(a, b);
}
Convert all enumerations to use `enum class`. Specifically, take the old code that looks like this: class Foo { enum { X = 1, Y = 2 }; int kind; } ... foo.kind = Foo::X; ... and convert it to this: class Foo { enum class Kind : uint32_t { X = 1, Y = 2 }; Kind kind; } ... foo.kind = Foo::Kind::X; (In some cases the enumeration would not be in the class namespace, such as when it is generally useful.) The benefits are as follows: * The type of the field gives a clear indication of intent, both to humans and tools (such as binding generators). * The compiler is able to automatically warn when a switch is not exhaustive; but this is currently suppressed by the default: ssassert(false, ...) idiom. * Integers and plain enums are weakly type checked: they implicitly convert into each other. This can hide bugs where type conversion is performed but not intended. Enum classes are strongly type checked. * Plain enums pollute parent namespaces; enum classes do not. Almost every defined enum we have already has a kind of ad-hoc namespacing via `NAMESPACE_`, which is now explicit. * Plain enums do not have a well-defined ABI size, which is important for bindings. Enum classes can have it, if specified. We specify the base type for all enums as uint32_t, which is a safe choice and allows us to not change the numeric values of any variants. This commit introduces absolutely no functional change to the code, just renaming and change of types. It handles almost all cases, except GraphicsWindow::pending.operation, which needs minor functional change.
2016-05-20 08:31:20 +00:00
void SShell::MakeFromBoolean(SShell *a, SShell *b, SSurface::CombineAs type) {
If a Boolean fails, then make a note of it in the group's text window screen, and remind the user that they could 'fix' the problem by working with meshes instead. [git-p4: depot-paths = "//depot/solvespace/": change = 1962]
2009-05-30 08:49:09 +00:00
booleanFailed = false;
Revamp the edge classification for Booleans. I no longer make a separate polygon of coincident (with same or opposite normal) faces; I instead test all the edges against the other shell, and have extended the classify-against-shell stuff to handle those cases. And the normals are now perturbed a bit numerically, to either side of the edge, to distinguish tangency from a coincident surface. This seems to work fairly well, although things still tend to fail when the piecewise linear tolerance is too coarse. [git-p4: depot-paths = "//depot/solvespace/": change = 1964]
2009-06-04 03:59:40 +00:00
a->MakeClassifyingBsps(NULL);
b->MakeClassifyingBsps(NULL);
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
Preparatory work for Boolean. Make the u and v coordinates of the trim curves for all surfaces lie between 0 and 1. And add routines to merge the curves and surfaces from two shells into one, and to split the trim curves into their piecewise linear segments and then reassemble them into trim curves. [git-p4: depot-paths = "//depot/solvespace/": change = 1905]
2009-01-25 11:52:29 +00:00
// Copy over all the original curves, splitting them so that a
// piecwise linear segment never crosses a surface from the other
// shell.
Add Newton iterations to intersect a line with a surface at a point, and to intersect three surfaces at a point. So now when we split an edge, we can refine the split point to lie exactly on the trim curve, so I can do certain Booleans on curved surfaces. But surface-line intersection is globally broken, since I don't correctly detect the number of intersections or provide a good first guess. I maybe should test by bounding boxes and subdivision. [git-p4: depot-paths = "//depot/solvespace/": change = 1920]
2009-02-27 13:04:36 +00:00
a->CopyCurvesSplitAgainst(true, b, this);
b->CopyCurvesSplitAgainst(false, a, this);
Preparatory work for Boolean. Make the u and v coordinates of the trim curves for all surfaces lie between 0 and 1. And add routines to merge the curves and surfaces from two shells into one, and to split the trim curves into their piecewise linear segments and then reassemble them into trim curves. [git-p4: depot-paths = "//depot/solvespace/": change = 1905]
2009-01-25 11:52:29 +00:00
// Generate the intersection curves for each surface in A against all
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
// the surfaces in B (which is all of the intersection curves).
More poking at Booleans; generate the unsplit intersection curves for planes against planes. [git-p4: depot-paths = "//depot/solvespace/": change = 1907]
2009-01-27 07:59:58 +00:00
a->MakeIntersectionCurvesAgainst(b, this);
Generate intersection curves for surfaces of extrusion along a parallel axis (which are always lines parallel to that axis). Remove short pwl segments when possible, to avoid short edges that get misclassified. [git-p4: depot-paths = "//depot/solvespace/": change = 1952]
2009-05-18 07:26:51 +00:00
SCurve *sc;
for(sc = curve.First(); sc; sc = curve.NextAfter(sc)) {
Add beginnings of marching surface intersection; I can find all the boundary points, at least. That required some changes to what gets passed around (for example because to project a point onto this inexact curve, we need to know which two surfaces it trims so that we can do a Newton's method on them). And fix stupidity in the way that I calculated edge normals; I just did normal in uv space, and there's no particular reason why that would be normal in xyz. So edges in long skinny surfaces failed, for example. [git-p4: depot-paths = "//depot/solvespace/": change = 1990]
2009-06-19 07:56:33 +00:00
SSurface *srfA = sc->GetSurfaceA(a, b),
Add const qualifiers to functions where trivially possible. This will allow us in future to accept `const T &` anywhere it's necessary to reduce the amount of copying. This commit is quite conservative: it does not attempt very hard to refactor code that performs incidental mutation. In particular dogd and caches are not marked with the `mutable` keyword. dogd will be eliminated later, opening up more opportunities to add const qualifiers. This commit also doesn't introduce any uses of the newly added const qualifers. This will be done later.
2016-05-21 05:18:00 +00:00
*srfB = sc->GetSurfaceB(a, b);
Add beginnings of marching surface intersection; I can find all the boundary points, at least. That required some changes to what gets passed around (for example because to project a point onto this inexact curve, we need to know which two surfaces it trims so that we can do a Newton's method on them). And fix stupidity in the way that I calculated edge normals; I just did normal in uv space, and there's no particular reason why that would be normal in xyz. So edges in long skinny surfaces failed, for example. [git-p4: depot-paths = "//depot/solvespace/": change = 1990]
2009-06-19 07:56:33 +00:00
Generate intersection curves for surfaces of extrusion along a parallel axis (which are always lines parallel to that axis). Remove short pwl segments when possible, to avoid short edges that get misclassified. [git-p4: depot-paths = "//depot/solvespace/": change = 1952]
2009-05-18 07:26:51 +00:00
sc->RemoveShortSegments(srfA, srfB);
}
Trim trailing whitespace.
2015-03-29 00:30:52 +00:00
Revamp the edge classification for Booleans. I no longer make a separate polygon of coincident (with same or opposite normal) faces; I instead test all the edges against the other shell, and have extended the classify-against-shell stuff to handle those cases. And the normals are now perturbed a bit numerically, to either side of the edge, to distinguish tangency from a coincident surface. This seems to work fairly well, although things still tend to fail when the piecewise linear tolerance is too coarse. [git-p4: depot-paths = "//depot/solvespace/": change = 1964]
2009-06-04 03:59:40 +00:00
// And clean up the piecewise linear things we made as a calculation aid
a->CleanupAfterBoolean();
b->CleanupAfterBoolean();
// Remake the classifying BSPs with the split (and short-segment-removed)
// curves
a->MakeClassifyingBsps(this);
b->MakeClassifyingBsps(this);
Compute surface intersections in a way that is closer to what I will do for real; now handling the special cases of plane against a surface of extrusion. Still need to fix up line-surface intersection to work for curved things, but then some simple curved cases should work (as well as plane-plane). [git-p4: depot-paths = "//depot/solvespace/": change = 1919]
2009-02-23 10:06:02 +00:00
if(b->surface.n == 0 || a->surface.n == 0) {
Add beginnings of marching surface intersection; I can find all the boundary points, at least. That required some changes to what gets passed around (for example because to project a point onto this inexact curve, we need to know which two surfaces it trims so that we can do a Newton's method on them). And fix stupidity in the way that I calculated edge normals; I just did normal in uv space, and there's no particular reason why that would be normal in xyz. So edges in long skinny surfaces failed, for example. [git-p4: depot-paths = "//depot/solvespace/": change = 1990]
2009-06-19 07:56:33 +00:00
I = 1000000;
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
} else {
Revamp the edge classification for Booleans. I no longer make a separate polygon of coincident (with same or opposite normal) faces; I instead test all the edges against the other shell, and have extended the classify-against-shell stuff to handle those cases. And the normals are now perturbed a bit numerically, to either side of the edge, to distinguish tangency from a coincident surface. This seems to work fairly well, although things still tend to fail when the piecewise linear tolerance is too coarse. [git-p4: depot-paths = "//depot/solvespace/": change = 1964]
2009-06-04 03:59:40 +00:00
I = 0;
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
}
Add beginnings of marching surface intersection; I can find all the boundary points, at least. That required some changes to what gets passed around (for example because to project a point onto this inexact curve, we need to know which two surfaces it trims so that we can do a Newton's method on them). And fix stupidity in the way that I calculated edge normals; I just did normal in uv space, and there's no particular reason why that would be normal in xyz. So edges in long skinny surfaces failed, for example. [git-p4: depot-paths = "//depot/solvespace/": change = 1990]
2009-06-19 07:56:33 +00:00
// Then trim and copy the surfaces
a->CopySurfacesTrimAgainst(a, b, this, type);
b->CopySurfacesTrimAgainst(a, b, this, type);
More poking at Booleans; generate the unsplit intersection curves for planes against planes. [git-p4: depot-paths = "//depot/solvespace/": change = 1907]
2009-01-27 07:59:58 +00:00
Many changes: * Rewrite surface handles in curves, so that Booleans beyond the first don't screw up. * If an intersection curve is identical to an existing curve (as happens when faces are coincident), take the piecewise linearization of the existing curve; this stops us from screwing up when different shells are pwl'd at different chord tols. * Hook up the plane faces again. * Remove coincident (parallel or anti-parallel) edges from the coincident-face edge lists when doing Booleans; those may happen if two faces are coincident with ours. * Miscellaneous bugfixes. It doesn't seem to screw up very much now, although tangent edges (and insufficient pwl resolution) may still cause problems. [git-p4: depot-paths = "//depot/solvespace/": change = 1929]
2009-03-15 23:04:45 +00:00
// Now that we've copied the surfaces, we know their new hSurfaces, so
// rewrite the curves to refer to the surfaces by their handles in the
// result.
Add code to assemble two shells into one, without checking for any intersections or otherwise trying to make the result not self-intersecting. [git-p4: depot-paths = "//depot/solvespace/": change = 1955]
2009-05-20 03:04:36 +00:00
RewriteSurfaceHandlesForCurves(a, b);
Many changes: * Rewrite surface handles in curves, so that Booleans beyond the first don't screw up. * If an intersection curve is identical to an existing curve (as happens when faces are coincident), take the piecewise linearization of the existing curve; this stops us from screwing up when different shells are pwl'd at different chord tols. * Hook up the plane faces again. * Remove coincident (parallel or anti-parallel) edges from the coincident-face edge lists when doing Booleans; those may happen if two faces are coincident with ours. * Miscellaneous bugfixes. It doesn't seem to screw up very much now, although tangent edges (and insufficient pwl resolution) may still cause problems. [git-p4: depot-paths = "//depot/solvespace/": change = 1929]
2009-03-15 23:04:45 +00:00
More poking at Booleans; generate the unsplit intersection curves for planes against planes. [git-p4: depot-paths = "//depot/solvespace/": change = 1907]
2009-01-27 07:59:58 +00:00
// And clean up the piecewise linear things we made as a calculation aid
a->CleanupAfterBoolean();
b->CleanupAfterBoolean();
Tear everything apart, moving away from meshes and toward shells. Add stubs for functions to perform Booleans, and get rid of mesh stuff, including the kd tree accelerated snap to vertex (which should not be required if the shell triangulation performs as it should). Also check that a sketch is not self-intersecting before extruding it or whatever. This is dead slow, needs n*log(n) implementation. [git-p4: depot-paths = "//depot/solvespace/": change = 1902]
2009-01-23 03:30:30 +00:00
}
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
//-----------------------------------------------------------------------------
// All of the BSP routines that we use to perform and accelerate polygon ops.
//-----------------------------------------------------------------------------
Revamp the edge classification for Booleans. I no longer make a separate polygon of coincident (with same or opposite normal) faces; I instead test all the edges against the other shell, and have extended the classify-against-shell stuff to handle those cases. And the normals are now perturbed a bit numerically, to either side of the edge, to distinguish tangency from a coincident surface. This seems to work fairly well, although things still tend to fail when the piecewise linear tolerance is too coarse. [git-p4: depot-paths = "//depot/solvespace/": change = 1964]
2009-06-04 03:59:40 +00:00
void SShell::MakeClassifyingBsps(SShell *useCurvesFrom) {
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
SSurface *ss;
for(ss = surface.First(); ss; ss = surface.NextAfter(ss)) {
Revamp the edge classification for Booleans. I no longer make a separate polygon of coincident (with same or opposite normal) faces; I instead test all the edges against the other shell, and have extended the classify-against-shell stuff to handle those cases. And the normals are now perturbed a bit numerically, to either side of the edge, to distinguish tangency from a coincident surface. This seems to work fairly well, although things still tend to fail when the piecewise linear tolerance is too coarse. [git-p4: depot-paths = "//depot/solvespace/": change = 1964]
2009-06-04 03:59:40 +00:00
ss->MakeClassifyingBsp(this, useCurvesFrom);
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
}
}
Revamp the edge classification for Booleans. I no longer make a separate polygon of coincident (with same or opposite normal) faces; I instead test all the edges against the other shell, and have extended the classify-against-shell stuff to handle those cases. And the normals are now perturbed a bit numerically, to either side of the edge, to distinguish tangency from a coincident surface. This seems to work fairly well, although things still tend to fail when the piecewise linear tolerance is too coarse. [git-p4: depot-paths = "//depot/solvespace/": change = 1964]
2009-06-04 03:59:40 +00:00
void SSurface::MakeClassifyingBsp(SShell *shell, SShell *useCurvesFrom) {
Replace all ZERO and memset with C++11 brace-initialization. This will allow us to use non-POD classes inside these objects in future and is otherwise functionally equivalent, as well as more concise. Note that there are some subtleties with handling of brace-initialization. Specifically: On aggregates (e.g. simple C-style structures) using an empty brace-initializer zero-initializes the aggregate, i.e. it makes all members zero. On non-aggregates an empty brace-initializer calls the default constructor. And if the constructor doesn't explicitly initialize the members (which the auto-generated constructor doesn't) then the members will be constructed but otherwise uninitialized. So, what is an aggregate class? To quote the C++ standard (C++03 8.5.1 §1): An aggregate is an array or a class (clause 9) with no user-declared constructors (12.1), no private or protected non-static data members (clause 11), no base classes (clause 10), and no virtual functions (10.3). In SolveSpace, we only have to handle the case of base classes; Constraint and Entity have those. Thus, they had to gain a default constructor that does nothing but initializes the members to zero.
2015-03-27 15:31:23 +00:00
SEdgeList el = {};
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
Convert all enumerations to use `enum class`. Specifically, take the old code that looks like this: class Foo { enum { X = 1, Y = 2 }; int kind; } ... foo.kind = Foo::X; ... and convert it to this: class Foo { enum class Kind : uint32_t { X = 1, Y = 2 }; Kind kind; } ... foo.kind = Foo::Kind::X; (In some cases the enumeration would not be in the class namespace, such as when it is generally useful.) The benefits are as follows: * The type of the field gives a clear indication of intent, both to humans and tools (such as binding generators). * The compiler is able to automatically warn when a switch is not exhaustive; but this is currently suppressed by the default: ssassert(false, ...) idiom. * Integers and plain enums are weakly type checked: they implicitly convert into each other. This can hide bugs where type conversion is performed but not intended. Enum classes are strongly type checked. * Plain enums pollute parent namespaces; enum classes do not. Almost every defined enum we have already has a kind of ad-hoc namespacing via `NAMESPACE_`, which is now explicit. * Plain enums do not have a well-defined ABI size, which is important for bindings. Enum classes can have it, if specified. We specify the base type for all enums as uint32_t, which is a safe choice and allows us to not change the numeric values of any variants. This commit introduces absolutely no functional change to the code, just renaming and change of types. It handles almost all cases, except GraphicsWindow::pending.operation, which needs minor functional change.
2016-05-20 08:31:20 +00:00
MakeEdgesInto(shell, &el, MakeAs::UV, useCurvesFrom);
I had been using LENGTH_EPS as the tolerance on both xyz points and uv points. This is inconsistent, unless the surface happens to be a plane square with side length one. So modify the SBspUv tests to take a surface, and measure distance linearized in that surface. That fixes at least one mis-classification bug, and doesn't seem to break anything. [git-p4: depot-paths = "//depot/solvespace/": change = 2005]
2009-07-02 03:32:17 +00:00
bsp = SBspUv::From(&el, this);
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
el.Clear();
Revamp the edge classification for Booleans. I no longer make a separate polygon of coincident (with same or opposite normal) faces; I instead test all the edges against the other shell, and have extended the classify-against-shell stuff to handle those cases. And the normals are now perturbed a bit numerically, to either side of the edge, to distinguish tangency from a coincident surface. This seems to work fairly well, although things still tend to fail when the piecewise linear tolerance is too coarse. [git-p4: depot-paths = "//depot/solvespace/": change = 1964]
2009-06-04 03:59:40 +00:00
Replace all ZERO and memset with C++11 brace-initialization. This will allow us to use non-POD classes inside these objects in future and is otherwise functionally equivalent, as well as more concise. Note that there are some subtleties with handling of brace-initialization. Specifically: On aggregates (e.g. simple C-style structures) using an empty brace-initializer zero-initializes the aggregate, i.e. it makes all members zero. On non-aggregates an empty brace-initializer calls the default constructor. And if the constructor doesn't explicitly initialize the members (which the auto-generated constructor doesn't) then the members will be constructed but otherwise uninitialized. So, what is an aggregate class? To quote the C++ standard (C++03 8.5.1 §1): An aggregate is an array or a class (clause 9) with no user-declared constructors (12.1), no private or protected non-static data members (clause 11), no base classes (clause 10), and no virtual functions (10.3). In SolveSpace, we only have to handle the case of base classes; Constraint and Entity have those. Thus, they had to gain a default constructor that does nothing but initializes the members to zero.
2015-03-27 15:31:23 +00:00
edges = {};
Convert all enumerations to use `enum class`. Specifically, take the old code that looks like this: class Foo { enum { X = 1, Y = 2 }; int kind; } ... foo.kind = Foo::X; ... and convert it to this: class Foo { enum class Kind : uint32_t { X = 1, Y = 2 }; Kind kind; } ... foo.kind = Foo::Kind::X; (In some cases the enumeration would not be in the class namespace, such as when it is generally useful.) The benefits are as follows: * The type of the field gives a clear indication of intent, both to humans and tools (such as binding generators). * The compiler is able to automatically warn when a switch is not exhaustive; but this is currently suppressed by the default: ssassert(false, ...) idiom. * Integers and plain enums are weakly type checked: they implicitly convert into each other. This can hide bugs where type conversion is performed but not intended. Enum classes are strongly type checked. * Plain enums pollute parent namespaces; enum classes do not. Almost every defined enum we have already has a kind of ad-hoc namespacing via `NAMESPACE_`, which is now explicit. * Plain enums do not have a well-defined ABI size, which is important for bindings. Enum classes can have it, if specified. We specify the base type for all enums as uint32_t, which is a safe choice and allows us to not change the numeric values of any variants. This commit introduces absolutely no functional change to the code, just renaming and change of types. It handles almost all cases, except GraphicsWindow::pending.operation, which needs minor functional change.
2016-05-20 08:31:20 +00:00
MakeEdgesInto(shell, &edges, MakeAs::XYZ, useCurvesFrom);
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
}
Rewrite declarations of form f(void) as f(). In C++ there is no difference and newly added functions are all declared as f(), so this brings back consistency.
2016-05-05 05:54:05 +00:00
SBspUv *SBspUv::Alloc() {
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
return (SBspUv *)AllocTemporary(sizeof(SBspUv));
}
static int ByLength(const void *av, const void *bv)
{
SEdge *a = (SEdge *)av,
*b = (SEdge *)bv;
Trim trailing whitespace.
2015-03-29 00:30:52 +00:00
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
double la = (a->a).Minus(a->b).Magnitude(),
lb = (b->a).Minus(b->b).Magnitude();
// Sort in descending order, longest first. This improves numerical
// stability for the normals.
return (la < lb) ? 1 : -1;
}
Add const qualifiers to functions where trivially possible. This will allow us in future to accept `const T &` anywhere it's necessary to reduce the amount of copying. This commit is quite conservative: it does not attempt very hard to refactor code that performs incidental mutation. In particular dogd and caches are not marked with the `mutable` keyword. dogd will be eliminated later, opening up more opportunities to add const qualifiers. This commit also doesn't introduce any uses of the newly added const qualifers. This will be done later.
2016-05-21 05:18:00 +00:00
I had been using LENGTH_EPS as the tolerance on both xyz points and uv points. This is inconsistent, unless the surface happens to be a plane square with side length one. So modify the SBspUv tests to take a surface, and measure distance linearized in that surface. That fixes at least one mis-classification bug, and doesn't seem to break anything. [git-p4: depot-paths = "//depot/solvespace/": change = 2005]
2009-07-02 03:32:17 +00:00
SBspUv *SBspUv::From(SEdgeList *el, SSurface *srf) {
Replace all ZERO and memset with C++11 brace-initialization. This will allow us to use non-POD classes inside these objects in future and is otherwise functionally equivalent, as well as more concise. Note that there are some subtleties with handling of brace-initialization. Specifically: On aggregates (e.g. simple C-style structures) using an empty brace-initializer zero-initializes the aggregate, i.e. it makes all members zero. On non-aggregates an empty brace-initializer calls the default constructor. And if the constructor doesn't explicitly initialize the members (which the auto-generated constructor doesn't) then the members will be constructed but otherwise uninitialized. So, what is an aggregate class? To quote the C++ standard (C++03 8.5.1 §1): An aggregate is an array or a class (clause 9) with no user-declared constructors (12.1), no private or protected non-static data members (clause 11), no base classes (clause 10), and no virtual functions (10.3). In SolveSpace, we only have to handle the case of base classes; Constraint and Entity have those. Thus, they had to gain a default constructor that does nothing but initializes the members to zero.
2015-03-27 15:31:23 +00:00
SEdgeList work = {};
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
SEdge *se;
for(se = el->l.First(); se; se = el->l.NextAfter(se)) {
work.AddEdge(se->a, se->b, se->auxA, se->auxB);
}
qsort(work.l.elem, work.l.n, sizeof(work.l.elem[0]), ByLength);
SBspUv *bsp = NULL;
for(se = work.l.First(); se; se = work.l.NextAfter(se)) {
Get rid of if(!this). According to the C++ standard, "this" is never NULL, so checks of the form "if(!this)" can be legally optimized out. This breaks SolveSpace on GCC 6, and probably on other compilers and configurations.
2016-03-25 14:16:58 +00:00
bsp = InsertOrCreateEdge(bsp, (se->a).ProjectXy(), (se->b).ProjectXy(), srf);
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
}
work.Clear();
return bsp;
}
I had been using LENGTH_EPS as the tolerance on both xyz points and uv points. This is inconsistent, unless the surface happens to be a plane square with side length one. So modify the SBspUv tests to take a surface, and measure distance linearized in that surface. That fixes at least one mis-classification bug, and doesn't seem to break anything. [git-p4: depot-paths = "//depot/solvespace/": change = 2005]
2009-07-02 03:32:17 +00:00
//-----------------------------------------------------------------------------
// The points in this BSP are in uv space, but we want to apply our tolerances
// consistently in xyz (i.e., we want to say a point is on-edge if its xyz
// distance to that edge is less than LENGTH_EPS, irrespective of its distance
// in uv). So we linearize the surface about the point we're considering and
// then do the test. That preserves point-on-line relationships, and the only
// time we care about exact correctness is when we're very close to the line,
// which is when the linearization is accurate.
//-----------------------------------------------------------------------------
Add const qualifiers to functions where trivially possible. This will allow us in future to accept `const T &` anywhere it's necessary to reduce the amount of copying. This commit is quite conservative: it does not attempt very hard to refactor code that performs incidental mutation. In particular dogd and caches are not marked with the `mutable` keyword. dogd will be eliminated later, opening up more opportunities to add const qualifiers. This commit also doesn't introduce any uses of the newly added const qualifers. This will be done later.
2016-05-21 05:18:00 +00:00
void SBspUv::ScalePoints(Point2d *pt, Point2d *a, Point2d *b, SSurface *srf) const {
I had been using LENGTH_EPS as the tolerance on both xyz points and uv points. This is inconsistent, unless the surface happens to be a plane square with side length one. So modify the SBspUv tests to take a surface, and measure distance linearized in that surface. That fixes at least one mis-classification bug, and doesn't seem to break anything. [git-p4: depot-paths = "//depot/solvespace/": change = 2005]
2009-07-02 03:32:17 +00:00
Vector tu, tv;
srf->TangentsAt(pt->x, pt->y, &tu, &tv);
double mu = tu.Magnitude(), mv = tv.Magnitude();
pt->x *= mu; pt->y *= mv;
a ->x *= mu; a ->y *= mv;
b ->x *= mu; b ->y *= mv;
}
Add const qualifiers to functions where trivially possible. This will allow us in future to accept `const T &` anywhere it's necessary to reduce the amount of copying. This commit is quite conservative: it does not attempt very hard to refactor code that performs incidental mutation. In particular dogd and caches are not marked with the `mutable` keyword. dogd will be eliminated later, opening up more opportunities to add const qualifiers. This commit also doesn't introduce any uses of the newly added const qualifers. This will be done later.
2016-05-21 05:18:00 +00:00
Trim trailing whitespace.
2015-03-29 00:30:52 +00:00
double SBspUv::ScaledSignedDistanceToLine(Point2d pt, Point2d a, Point2d b,
Add const qualifiers to functions where trivially possible. This will allow us in future to accept `const T &` anywhere it's necessary to reduce the amount of copying. This commit is quite conservative: it does not attempt very hard to refactor code that performs incidental mutation. In particular dogd and caches are not marked with the `mutable` keyword. dogd will be eliminated later, opening up more opportunities to add const qualifiers. This commit also doesn't introduce any uses of the newly added const qualifers. This will be done later.
2016-05-21 05:18:00 +00:00
SSurface *srf) const
I had been using LENGTH_EPS as the tolerance on both xyz points and uv points. This is inconsistent, unless the surface happens to be a plane square with side length one. So modify the SBspUv tests to take a surface, and measure distance linearized in that surface. That fixes at least one mis-classification bug, and doesn't seem to break anything. [git-p4: depot-paths = "//depot/solvespace/": change = 2005]
2009-07-02 03:32:17 +00:00
{
ScalePoints(&pt, &a, &b, srf);
Point2d n = ((b.Minus(a)).Normal()).WithMagnitude(1);
double d = a.Dot(n);
return pt.Dot(n) - d;
}
Add const qualifiers to functions where trivially possible. This will allow us in future to accept `const T &` anywhere it's necessary to reduce the amount of copying. This commit is quite conservative: it does not attempt very hard to refactor code that performs incidental mutation. In particular dogd and caches are not marked with the `mutable` keyword. dogd will be eliminated later, opening up more opportunities to add const qualifiers. This commit also doesn't introduce any uses of the newly added const qualifers. This will be done later.
2016-05-21 05:18:00 +00:00
I had been using LENGTH_EPS as the tolerance on both xyz points and uv points. This is inconsistent, unless the surface happens to be a plane square with side length one. So modify the SBspUv tests to take a surface, and measure distance linearized in that surface. That fixes at least one mis-classification bug, and doesn't seem to break anything. [git-p4: depot-paths = "//depot/solvespace/": change = 2005]
2009-07-02 03:32:17 +00:00
double SBspUv::ScaledDistanceToLine(Point2d pt, Point2d a, Point2d b, bool seg,
Add const qualifiers to functions where trivially possible. This will allow us in future to accept `const T &` anywhere it's necessary to reduce the amount of copying. This commit is quite conservative: it does not attempt very hard to refactor code that performs incidental mutation. In particular dogd and caches are not marked with the `mutable` keyword. dogd will be eliminated later, opening up more opportunities to add const qualifiers. This commit also doesn't introduce any uses of the newly added const qualifers. This will be done later.
2016-05-21 05:18:00 +00:00
SSurface *srf) const
I had been using LENGTH_EPS as the tolerance on both xyz points and uv points. This is inconsistent, unless the surface happens to be a plane square with side length one. So modify the SBspUv tests to take a surface, and measure distance linearized in that surface. That fixes at least one mis-classification bug, and doesn't seem to break anything. [git-p4: depot-paths = "//depot/solvespace/": change = 2005]
2009-07-02 03:32:17 +00:00
{
ScalePoints(&pt, &a, &b, srf);
return pt.DistanceToLine(a, b, seg);
}
Add const qualifiers to functions where trivially possible. This will allow us in future to accept `const T &` anywhere it's necessary to reduce the amount of copying. This commit is quite conservative: it does not attempt very hard to refactor code that performs incidental mutation. In particular dogd and caches are not marked with the `mutable` keyword. dogd will be eliminated later, opening up more opportunities to add const qualifiers. This commit also doesn't introduce any uses of the newly added const qualifers. This will be done later.
2016-05-21 05:18:00 +00:00
SBspUv *SBspUv::InsertOrCreateEdge(SBspUv *where, Point2d ea, Point2d eb, SSurface *srf) {
Get rid of if(!this). According to the C++ standard, "this" is never NULL, so checks of the form "if(!this)" can be legally optimized out. This breaks SolveSpace on GCC 6, and probably on other compilers and configurations.
2016-03-25 14:16:58 +00:00
if(where == NULL) {
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
SBspUv *ret = Alloc();
ret->a = ea;
ret->b = eb;
return ret;
}
Get rid of if(!this). According to the C++ standard, "this" is never NULL, so checks of the form "if(!this)" can be legally optimized out. This breaks SolveSpace on GCC 6, and probably on other compilers and configurations.
2016-03-25 14:16:58 +00:00
where->InsertEdge(ea, eb, srf);
return where;
}
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
Get rid of if(!this). According to the C++ standard, "this" is never NULL, so checks of the form "if(!this)" can be legally optimized out. This breaks SolveSpace on GCC 6, and probably on other compilers and configurations.
2016-03-25 14:16:58 +00:00
void SBspUv::InsertEdge(Point2d ea, Point2d eb, SSurface *srf) {
I had been using LENGTH_EPS as the tolerance on both xyz points and uv points. This is inconsistent, unless the surface happens to be a plane square with side length one. So modify the SBspUv tests to take a surface, and measure distance linearized in that surface. That fixes at least one mis-classification bug, and doesn't seem to break anything. [git-p4: depot-paths = "//depot/solvespace/": change = 2005]
2009-07-02 03:32:17 +00:00
double dea = ScaledSignedDistanceToLine(ea, a, b, srf),
deb = ScaledSignedDistanceToLine(eb, a, b, srf);
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
if(fabs(dea) < LENGTH_EPS && fabs(deb) < LENGTH_EPS) {
// Line segment is coincident with this one, store in same node
SBspUv *m = Alloc();
m->a = ea;
m->b = eb;
m->more = more;
more = m;
} else if(fabs(dea) < LENGTH_EPS) {
// Point A lies on this lie, but point B does not
if(deb > 0) {
Get rid of if(!this). According to the C++ standard, "this" is never NULL, so checks of the form "if(!this)" can be legally optimized out. This breaks SolveSpace on GCC 6, and probably on other compilers and configurations.
2016-03-25 14:16:58 +00:00
pos = InsertOrCreateEdge(pos, ea, eb, srf);
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
} else {
Get rid of if(!this). According to the C++ standard, "this" is never NULL, so checks of the form "if(!this)" can be legally optimized out. This breaks SolveSpace on GCC 6, and probably on other compilers and configurations.
2016-03-25 14:16:58 +00:00
neg = InsertOrCreateEdge(neg, ea, eb, srf);
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
}
} else if(fabs(deb) < LENGTH_EPS) {
// Point B lies on this lie, but point A does not
if(dea > 0) {
Get rid of if(!this). According to the C++ standard, "this" is never NULL, so checks of the form "if(!this)" can be legally optimized out. This breaks SolveSpace on GCC 6, and probably on other compilers and configurations.
2016-03-25 14:16:58 +00:00
pos = InsertOrCreateEdge(pos, ea, eb, srf);
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
} else {
Get rid of if(!this). According to the C++ standard, "this" is never NULL, so checks of the form "if(!this)" can be legally optimized out. This breaks SolveSpace on GCC 6, and probably on other compilers and configurations.
2016-03-25 14:16:58 +00:00
neg = InsertOrCreateEdge(neg, ea, eb, srf);
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
}
} else if(dea > 0 && deb > 0) {
Get rid of if(!this). According to the C++ standard, "this" is never NULL, so checks of the form "if(!this)" can be legally optimized out. This breaks SolveSpace on GCC 6, and probably on other compilers and configurations.
2016-03-25 14:16:58 +00:00
pos = InsertOrCreateEdge(pos, ea, eb, srf);
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
} else if(dea < 0 && deb < 0) {
Get rid of if(!this). According to the C++ standard, "this" is never NULL, so checks of the form "if(!this)" can be legally optimized out. This breaks SolveSpace on GCC 6, and probably on other compilers and configurations.
2016-03-25 14:16:58 +00:00
neg = InsertOrCreateEdge(neg, ea, eb, srf);
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
} else {
// New edge crosses this one; we need to split.
I had been using LENGTH_EPS as the tolerance on both xyz points and uv points. This is inconsistent, unless the surface happens to be a plane square with side length one. So modify the SBspUv tests to take a surface, and measure distance linearized in that surface. That fixes at least one mis-classification bug, and doesn't seem to break anything. [git-p4: depot-paths = "//depot/solvespace/": change = 2005]
2009-07-02 03:32:17 +00:00
Point2d n = ((b.Minus(a)).Normal()).WithMagnitude(1);
double d = a.Dot(n);
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
double t = (d - n.Dot(ea)) / (n.Dot(eb.Minus(ea)));
Point2d pi = ea.Plus((eb.Minus(ea)).ScaledBy(t));
if(dea > 0) {
Get rid of if(!this). According to the C++ standard, "this" is never NULL, so checks of the form "if(!this)" can be legally optimized out. This breaks SolveSpace on GCC 6, and probably on other compilers and configurations.
2016-03-25 14:16:58 +00:00
pos = InsertOrCreateEdge(pos, ea, pi, srf);
neg = InsertOrCreateEdge(neg, pi, eb, srf);
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
} else {
Get rid of if(!this). According to the C++ standard, "this" is never NULL, so checks of the form "if(!this)" can be legally optimized out. This breaks SolveSpace on GCC 6, and probably on other compilers and configurations.
2016-03-25 14:16:58 +00:00
neg = InsertOrCreateEdge(neg, ea, pi, srf);
pos = InsertOrCreateEdge(pos, pi, eb, srf);
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
}
}
Get rid of if(!this). According to the C++ standard, "this" is never NULL, so checks of the form "if(!this)" can be legally optimized out. This breaks SolveSpace on GCC 6, and probably on other compilers and configurations.
2016-03-25 14:16:58 +00:00
return;
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
}
Convert all enumerations to use `enum class`. Specifically, take the old code that looks like this: class Foo { enum { X = 1, Y = 2 }; int kind; } ... foo.kind = Foo::X; ... and convert it to this: class Foo { enum class Kind : uint32_t { X = 1, Y = 2 }; Kind kind; } ... foo.kind = Foo::Kind::X; (In some cases the enumeration would not be in the class namespace, such as when it is generally useful.) The benefits are as follows: * The type of the field gives a clear indication of intent, both to humans and tools (such as binding generators). * The compiler is able to automatically warn when a switch is not exhaustive; but this is currently suppressed by the default: ssassert(false, ...) idiom. * Integers and plain enums are weakly type checked: they implicitly convert into each other. This can hide bugs where type conversion is performed but not intended. Enum classes are strongly type checked. * Plain enums pollute parent namespaces; enum classes do not. Almost every defined enum we have already has a kind of ad-hoc namespacing via `NAMESPACE_`, which is now explicit. * Plain enums do not have a well-defined ABI size, which is important for bindings. Enum classes can have it, if specified. We specify the base type for all enums as uint32_t, which is a safe choice and allows us to not change the numeric values of any variants. This commit introduces absolutely no functional change to the code, just renaming and change of types. It handles almost all cases, except GraphicsWindow::pending.operation, which needs minor functional change.
2016-05-20 08:31:20 +00:00
SBspUv::Class SBspUv::ClassifyPoint(Point2d p, Point2d eb, SSurface *srf) const {
I had been using LENGTH_EPS as the tolerance on both xyz points and uv points. This is inconsistent, unless the surface happens to be a plane square with side length one. So modify the SBspUv tests to take a surface, and measure distance linearized in that surface. That fixes at least one mis-classification bug, and doesn't seem to break anything. [git-p4: depot-paths = "//depot/solvespace/": change = 2005]
2009-07-02 03:32:17 +00:00
double dp = ScaledSignedDistanceToLine(p, a, b, srf);
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
if(fabs(dp) < LENGTH_EPS) {
Add const qualifiers to functions where trivially possible. This will allow us in future to accept `const T &` anywhere it's necessary to reduce the amount of copying. This commit is quite conservative: it does not attempt very hard to refactor code that performs incidental mutation. In particular dogd and caches are not marked with the `mutable` keyword. dogd will be eliminated later, opening up more opportunities to add const qualifiers. This commit also doesn't introduce any uses of the newly added const qualifers. This will be done later.
2016-05-21 05:18:00 +00:00
const SBspUv *f = this;
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
while(f) {
Point2d ba = (f->b).Minus(f->a);
I had been using LENGTH_EPS as the tolerance on both xyz points and uv points. This is inconsistent, unless the surface happens to be a plane square with side length one. So modify the SBspUv tests to take a surface, and measure distance linearized in that surface. That fixes at least one mis-classification bug, and doesn't seem to break anything. [git-p4: depot-paths = "//depot/solvespace/": change = 2005]
2009-07-02 03:32:17 +00:00
if(ScaledDistanceToLine(p, f->a, ba, true, srf) < LENGTH_EPS) {
if(ScaledDistanceToLine(eb, f->a, ba, false, srf) < LENGTH_EPS){
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
if(ba.Dot(eb.Minus(p)) > 0) {
Convert all enumerations to use `enum class`. Specifically, take the old code that looks like this: class Foo { enum { X = 1, Y = 2 }; int kind; } ... foo.kind = Foo::X; ... and convert it to this: class Foo { enum class Kind : uint32_t { X = 1, Y = 2 }; Kind kind; } ... foo.kind = Foo::Kind::X; (In some cases the enumeration would not be in the class namespace, such as when it is generally useful.) The benefits are as follows: * The type of the field gives a clear indication of intent, both to humans and tools (such as binding generators). * The compiler is able to automatically warn when a switch is not exhaustive; but this is currently suppressed by the default: ssassert(false, ...) idiom. * Integers and plain enums are weakly type checked: they implicitly convert into each other. This can hide bugs where type conversion is performed but not intended. Enum classes are strongly type checked. * Plain enums pollute parent namespaces; enum classes do not. Almost every defined enum we have already has a kind of ad-hoc namespacing via `NAMESPACE_`, which is now explicit. * Plain enums do not have a well-defined ABI size, which is important for bindings. Enum classes can have it, if specified. We specify the base type for all enums as uint32_t, which is a safe choice and allows us to not change the numeric values of any variants. This commit introduces absolutely no functional change to the code, just renaming and change of types. It handles almost all cases, except GraphicsWindow::pending.operation, which needs minor functional change.
2016-05-20 08:31:20 +00:00
return Class::EDGE_PARALLEL;
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
} else {
Convert all enumerations to use `enum class`. Specifically, take the old code that looks like this: class Foo { enum { X = 1, Y = 2 }; int kind; } ... foo.kind = Foo::X; ... and convert it to this: class Foo { enum class Kind : uint32_t { X = 1, Y = 2 }; Kind kind; } ... foo.kind = Foo::Kind::X; (In some cases the enumeration would not be in the class namespace, such as when it is generally useful.) The benefits are as follows: * The type of the field gives a clear indication of intent, both to humans and tools (such as binding generators). * The compiler is able to automatically warn when a switch is not exhaustive; but this is currently suppressed by the default: ssassert(false, ...) idiom. * Integers and plain enums are weakly type checked: they implicitly convert into each other. This can hide bugs where type conversion is performed but not intended. Enum classes are strongly type checked. * Plain enums pollute parent namespaces; enum classes do not. Almost every defined enum we have already has a kind of ad-hoc namespacing via `NAMESPACE_`, which is now explicit. * Plain enums do not have a well-defined ABI size, which is important for bindings. Enum classes can have it, if specified. We specify the base type for all enums as uint32_t, which is a safe choice and allows us to not change the numeric values of any variants. This commit introduces absolutely no functional change to the code, just renaming and change of types. It handles almost all cases, except GraphicsWindow::pending.operation, which needs minor functional change.
2016-05-20 08:31:20 +00:00
return Class::EDGE_ANTIPARALLEL;
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
}
} else {
Convert all enumerations to use `enum class`. Specifically, take the old code that looks like this: class Foo { enum { X = 1, Y = 2 }; int kind; } ... foo.kind = Foo::X; ... and convert it to this: class Foo { enum class Kind : uint32_t { X = 1, Y = 2 }; Kind kind; } ... foo.kind = Foo::Kind::X; (In some cases the enumeration would not be in the class namespace, such as when it is generally useful.) The benefits are as follows: * The type of the field gives a clear indication of intent, both to humans and tools (such as binding generators). * The compiler is able to automatically warn when a switch is not exhaustive; but this is currently suppressed by the default: ssassert(false, ...) idiom. * Integers and plain enums are weakly type checked: they implicitly convert into each other. This can hide bugs where type conversion is performed but not intended. Enum classes are strongly type checked. * Plain enums pollute parent namespaces; enum classes do not. Almost every defined enum we have already has a kind of ad-hoc namespacing via `NAMESPACE_`, which is now explicit. * Plain enums do not have a well-defined ABI size, which is important for bindings. Enum classes can have it, if specified. We specify the base type for all enums as uint32_t, which is a safe choice and allows us to not change the numeric values of any variants. This commit introduces absolutely no functional change to the code, just renaming and change of types. It handles almost all cases, except GraphicsWindow::pending.operation, which needs minor functional change.
2016-05-20 08:31:20 +00:00
return Class::EDGE_OTHER;
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
}
}
f = f->more;
}
// Pick arbitrarily which side to send it down, doesn't matter
Convert all enumerations to use `enum class`. Specifically, take the old code that looks like this: class Foo { enum { X = 1, Y = 2 }; int kind; } ... foo.kind = Foo::X; ... and convert it to this: class Foo { enum class Kind : uint32_t { X = 1, Y = 2 }; Kind kind; } ... foo.kind = Foo::Kind::X; (In some cases the enumeration would not be in the class namespace, such as when it is generally useful.) The benefits are as follows: * The type of the field gives a clear indication of intent, both to humans and tools (such as binding generators). * The compiler is able to automatically warn when a switch is not exhaustive; but this is currently suppressed by the default: ssassert(false, ...) idiom. * Integers and plain enums are weakly type checked: they implicitly convert into each other. This can hide bugs where type conversion is performed but not intended. Enum classes are strongly type checked. * Plain enums pollute parent namespaces; enum classes do not. Almost every defined enum we have already has a kind of ad-hoc namespacing via `NAMESPACE_`, which is now explicit. * Plain enums do not have a well-defined ABI size, which is important for bindings. Enum classes can have it, if specified. We specify the base type for all enums as uint32_t, which is a safe choice and allows us to not change the numeric values of any variants. This commit introduces absolutely no functional change to the code, just renaming and change of types. It handles almost all cases, except GraphicsWindow::pending.operation, which needs minor functional change.
2016-05-20 08:31:20 +00:00
Class c1 = neg ? neg->ClassifyPoint(p, eb, srf) : Class::OUTSIDE;
Class c2 = pos ? pos->ClassifyPoint(p, eb, srf) : Class::INSIDE;
Fix stupidity in Point2d::DistanceToLine, and classify line segments in Boolean against the shell, not the intersection polygon. (We just cast a ray, and use the surface-line intersection function that already existed.) That's slow, but can be accelerated later. [git-p4: depot-paths = "//depot/solvespace/": change = 1911]
2009-02-01 13:01:28 +00:00
if(c1 != c2) {
dbp("MISMATCH: %d %d %08x %08x", c1, c2, neg, pos);
}
return c1;
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
} else if(dp > 0) {
Convert all enumerations to use `enum class`. Specifically, take the old code that looks like this: class Foo { enum { X = 1, Y = 2 }; int kind; } ... foo.kind = Foo::X; ... and convert it to this: class Foo { enum class Kind : uint32_t { X = 1, Y = 2 }; Kind kind; } ... foo.kind = Foo::Kind::X; (In some cases the enumeration would not be in the class namespace, such as when it is generally useful.) The benefits are as follows: * The type of the field gives a clear indication of intent, both to humans and tools (such as binding generators). * The compiler is able to automatically warn when a switch is not exhaustive; but this is currently suppressed by the default: ssassert(false, ...) idiom. * Integers and plain enums are weakly type checked: they implicitly convert into each other. This can hide bugs where type conversion is performed but not intended. Enum classes are strongly type checked. * Plain enums pollute parent namespaces; enum classes do not. Almost every defined enum we have already has a kind of ad-hoc namespacing via `NAMESPACE_`, which is now explicit. * Plain enums do not have a well-defined ABI size, which is important for bindings. Enum classes can have it, if specified. We specify the base type for all enums as uint32_t, which is a safe choice and allows us to not change the numeric values of any variants. This commit introduces absolutely no functional change to the code, just renaming and change of types. It handles almost all cases, except GraphicsWindow::pending.operation, which needs minor functional change.
2016-05-20 08:31:20 +00:00
return pos ? pos->ClassifyPoint(p, eb, srf) : Class::INSIDE;
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
} else {
Convert all enumerations to use `enum class`. Specifically, take the old code that looks like this: class Foo { enum { X = 1, Y = 2 }; int kind; } ... foo.kind = Foo::X; ... and convert it to this: class Foo { enum class Kind : uint32_t { X = 1, Y = 2 }; Kind kind; } ... foo.kind = Foo::Kind::X; (In some cases the enumeration would not be in the class namespace, such as when it is generally useful.) The benefits are as follows: * The type of the field gives a clear indication of intent, both to humans and tools (such as binding generators). * The compiler is able to automatically warn when a switch is not exhaustive; but this is currently suppressed by the default: ssassert(false, ...) idiom. * Integers and plain enums are weakly type checked: they implicitly convert into each other. This can hide bugs where type conversion is performed but not intended. Enum classes are strongly type checked. * Plain enums pollute parent namespaces; enum classes do not. Almost every defined enum we have already has a kind of ad-hoc namespacing via `NAMESPACE_`, which is now explicit. * Plain enums do not have a well-defined ABI size, which is important for bindings. Enum classes can have it, if specified. We specify the base type for all enums as uint32_t, which is a safe choice and allows us to not change the numeric values of any variants. This commit introduces absolutely no functional change to the code, just renaming and change of types. It handles almost all cases, except GraphicsWindow::pending.operation, which needs minor functional change.
2016-05-20 08:31:20 +00:00
return neg ? neg->ClassifyPoint(p, eb, srf) : Class::OUTSIDE;
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
}
}
Convert all enumerations to use `enum class`. Specifically, take the old code that looks like this: class Foo { enum { X = 1, Y = 2 }; int kind; } ... foo.kind = Foo::X; ... and convert it to this: class Foo { enum class Kind : uint32_t { X = 1, Y = 2 }; Kind kind; } ... foo.kind = Foo::Kind::X; (In some cases the enumeration would not be in the class namespace, such as when it is generally useful.) The benefits are as follows: * The type of the field gives a clear indication of intent, both to humans and tools (such as binding generators). * The compiler is able to automatically warn when a switch is not exhaustive; but this is currently suppressed by the default: ssassert(false, ...) idiom. * Integers and plain enums are weakly type checked: they implicitly convert into each other. This can hide bugs where type conversion is performed but not intended. Enum classes are strongly type checked. * Plain enums pollute parent namespaces; enum classes do not. Almost every defined enum we have already has a kind of ad-hoc namespacing via `NAMESPACE_`, which is now explicit. * Plain enums do not have a well-defined ABI size, which is important for bindings. Enum classes can have it, if specified. We specify the base type for all enums as uint32_t, which is a safe choice and allows us to not change the numeric values of any variants. This commit introduces absolutely no functional change to the code, just renaming and change of types. It handles almost all cases, except GraphicsWindow::pending.operation, which needs minor functional change.
2016-05-20 08:31:20 +00:00
SBspUv::Class SBspUv::ClassifyEdge(Point2d ea, Point2d eb, SSurface *srf) const {
SBspUv::Class ret = ClassifyPoint((ea.Plus(eb)).ScaledBy(0.5), eb, srf);
if(ret == Class::EDGE_OTHER) {
Many changes: * Rewrite surface handles in curves, so that Booleans beyond the first don't screw up. * If an intersection curve is identical to an existing curve (as happens when faces are coincident), take the piecewise linearization of the existing curve; this stops us from screwing up when different shells are pwl'd at different chord tols. * Hook up the plane faces again. * Remove coincident (parallel or anti-parallel) edges from the coincident-face edge lists when doing Booleans; those may happen if two faces are coincident with ours. * Miscellaneous bugfixes. It doesn't seem to screw up very much now, although tangent edges (and insufficient pwl resolution) may still cause problems. [git-p4: depot-paths = "//depot/solvespace/": change = 1929]
2009-03-15 23:04:45 +00:00
// Perhaps the edge is tangent at its midpoint (and we screwed up
// somewhere earlier and failed to split it); try a different
// point on the edge.
I had been using LENGTH_EPS as the tolerance on both xyz points and uv points. This is inconsistent, unless the surface happens to be a plane square with side length one. So modify the SBspUv tests to take a surface, and measure distance linearized in that surface. That fixes at least one mis-classification bug, and doesn't seem to break anything. [git-p4: depot-paths = "//depot/solvespace/": change = 2005]
2009-07-02 03:32:17 +00:00
ret = ClassifyPoint(ea.Plus((eb.Minus(ea)).ScaledBy(0.294)), eb, srf);
Many changes: * Rewrite surface handles in curves, so that Booleans beyond the first don't screw up. * If an intersection curve is identical to an existing curve (as happens when faces are coincident), take the piecewise linearization of the existing curve; this stops us from screwing up when different shells are pwl'd at different chord tols. * Hook up the plane faces again. * Remove coincident (parallel or anti-parallel) edges from the coincident-face edge lists when doing Booleans; those may happen if two faces are coincident with ours. * Miscellaneous bugfixes. It doesn't seem to screw up very much now, although tangent edges (and insufficient pwl resolution) may still cause problems. [git-p4: depot-paths = "//depot/solvespace/": change = 1929]
2009-03-15 23:04:45 +00:00
}
return ret;
More work on Booleans. This works only for planes, and only for non-coincident faces. There's also a problem when I don't generate the full intersection polygon of shell B against a given surface in shell A; I need to modify the code to not require that. [git-p4: depot-paths = "//depot/solvespace/": change = 1910]
2009-02-01 05:13:43 +00:00
}
Add const qualifiers to functions where trivially possible. This will allow us in future to accept `const T &` anywhere it's necessary to reduce the amount of copying. This commit is quite conservative: it does not attempt very hard to refactor code that performs incidental mutation. In particular dogd and caches are not marked with the `mutable` keyword. dogd will be eliminated later, opening up more opportunities to add const qualifiers. This commit also doesn't introduce any uses of the newly added const qualifers. This will be done later.
2016-05-21 05:18:00 +00:00
double SBspUv::MinimumDistanceToEdge(Point2d p, SSurface *srf) const {
Discard intersection curves that lie entirely outside of one surface's domain of u, v in [0, 1]. Cache the starting guess when projecting a point into a ratpoly surface, to avoid brute force searching for a good one every time. Split edges even if they aren't quite inside the trim curve, since the trim boundaries are pwl, not exact; unnecessary splits won't hurt, but failure to split when necessary will. Make the triangulation code use a better (but not perfect) epsilon, to avoid "can't find ear" failures on very fine meshes. And turn on compiler optimization! I had somehow forgotten about that, and it's a ~2x improvement. [git-p4: depot-paths = "//depot/solvespace/": change = 2026]
2009-08-21 04:58:28 +00:00
Get rid of if(!this). According to the C++ standard, "this" is never NULL, so checks of the form "if(!this)" can be legally optimized out. This breaks SolveSpace on GCC 6, and probably on other compilers and configurations.
2016-03-25 14:16:58 +00:00
double dn = (neg) ? neg->MinimumDistanceToEdge(p, srf) : VERY_POSITIVE;
double dp = (pos) ? pos->MinimumDistanceToEdge(p, srf) : VERY_POSITIVE;
Discard intersection curves that lie entirely outside of one surface's domain of u, v in [0, 1]. Cache the starting guess when projecting a point into a ratpoly surface, to avoid brute force searching for a good one every time. Split edges even if they aren't quite inside the trim curve, since the trim boundaries are pwl, not exact; unnecessary splits won't hurt, but failure to split when necessary will. Make the triangulation code use a better (but not perfect) epsilon, to avoid "can't find ear" failures on very fine meshes. And turn on compiler optimization! I had somehow forgotten about that, and it's a ~2x improvement. [git-p4: depot-paths = "//depot/solvespace/": change = 2026]
2009-08-21 04:58:28 +00:00
Point2d as = a, bs = b;
ScalePoints(&p, &as, &bs, srf);
double d = p.DistanceToLine(as, bs.Minus(as), true);
return min(d, min(dn, dp));
}