The root of the problem is incorrect processing of cases when intersection line goes through the apex(es) of sphere. The fix improves this situation. The algorithm is taken from DecomposeResult(...) function (see IntPatch_ImpPrmIntersection.cxx file). Before the fix, faltering steps were done to solve this problem. As result, it worked in some particular cases. Now, its possibilities have been extended significantly.
Following changes have been made in the fix:
1. Class IntPatch_ALineToWLine has been rewritten cardinally. It touches as interfaces of existing methods as adding/removing some methods/fields. Correction touches both cases: going through seam of Cone/Sphere and through pole(s) of sphere. Old interface did not allow making some actions with analytical line (ALine), e.g. splitting it on several Walking-lines (WLine).
2. Restriction-line support has been removed from Implicit-Implicit intersection result (see IntPatch_Intersection::GeomGeomPerfom(...) method). It connects with the fact that the intersection algorithm itself returns precise intersection line in analytical cases (in compare with parametric intersector). Therefore, we do not need in additional (restriction) line.
3. New class IntPatch_SpecialPoints has been added. This class contains methods to add some special points (such as apex of cone, pole of sphere, point on surface boundary etc.) in intersection line (IntPatch_PointLine). It is based on the static functions, which already exist in IntPatch_ImpPrmIntersection.cxx file (these functions have been moved to the new class).
4. Method IntPatch_WLineTool::ExtendTwoWlinesToEachOther(...) has been renamed to IntPatch_WLineTool::ExtendTwoWLines(...). It is connected with changing main idea of the method. Now it allows extending WLine to the surface boundary or to the singular point (if it is possible): cone apex, sphere pole etc. Interface of this method has been corrected, too. At that, old functionality (extending to each other) has been kept. For implementation of this algorithm, new enumeration "IntPatchWT_WLsConnectionType" has been created.
5. Method IntPatch_PointLine::CurvatureRadiusOfIntersLine(...) has been added. See IntPatch_PointLine.hxx for detail information. It allows correct step computing depended on the local curvature of the intersection line. This method uses geometrical properties of intersected surfaces to compute local curvature. Therefore, it can be applied in wide range of cases even if the intersection curve is not represented in explicit form (e.g. in case of param-param-intersection).
6. Method IntSurf::SetPeriod(...) has been created.
7. Additional check has been added in Draft_Modification::Perform() method for better choice of correct fragment of intersection line for processing DRAFT operation.
8. New overload method IntPatch_Point::SetValue() has been added.
9. Some refactoring of the code has been made.
Creation of test case for issue #27431.
---------------------------------------------------------------------------------------------
Some test cases have been adjusted according to their new behavior.
tests\bugs\modalg_4\bug62
It is really IMPROVEMENT (but fortuitous).
tests\bugs\modalg_5\bug25838
The behavior of this test has been reverted to the state before fixing the issue #27341. Main problem has not been fixed in #27341. It was fortuitous improvement.
tests\bugs\moddata_2\bug565
Quality of intersection curve was not checked. And the curve is bad on both MASTER and FIX. Input data are really wrong: plane-like-cone. However, on the MASTER, four intersection curves (the quality is insignificant) are expected. On the fix, not empty intersection result is expected simply.
tests\boolean\volumemaker\A8
Differences in images and CPU is expected. Difference in images is expected to be fixed in the issue #26020. Now, we should apply this behavior.
Much CPU time is spent by IntTools_FaceFace::ComputeTolReached3d(...) and GeomInt_IntSS::BuildPCurves(...) methods calling. These methods are not touched by the algorithm. It is the result of change of intersection curve(s) form. However, the new Curve(s) seems to be valid and can be applied. As result, new behavior can be applied, too.
tests\boolean\volumemaker\F8
tests\boolean\volumemaker\F9
tests\boolean\volumemaker\G5
tests\boolean\volumemaker\G6
CPU difference is expected. Much CPU time is spent by IntPatch_PointLine::CurvatureRadiusOfIntersLine(...) method calling. This method is really new (it does not exist on the MASTER) and is really useful. Therefore, we should apply new behavior.
tests\boolean\volumemaker\G1
CPU difference is expected. Much CPU time is spent by IntTools_WLineTool::DecompositionOfWLine(...) and IntTools_FaceFace::ComputeTolReached3d(...) methods calling. These methods are not touched by the algorithm. It is the result of change of intersection curve(s) form. However, the new Curve(s) seems to be valid and can be applied. As result, new behavior can be applied, too.
tests\bugs\modalg_6\bug26619
Differences in images is expected. The test keeps its BAD status on the FIX. But the result on the fix is nearer to expected than on the MASTER. Issue #27014 is still actual. As before, it is not clear, why the number of entities is different. The number of section curves has not been changed. Interfered entities are the same as on the MASTER.
tests\bugs\modalg_5\bug25319_1(2)
The reason is described in the issue #27896.
Small correction in the test case
1. VRange of intersection curve has been limited. As result, too oblong intersection curve(s) will be never returned.
2. Now, purger algorithm is not called for lines obtained by Geom-Geom intersection method.
3. New statuses are entered in IntPatch_ImpImpIntersection class. It makes intersection algorithm more informative and flexible for using.
4. Method IntPatch_ImpImpIntersection::GetStatus() has been created.
Tuning of test case bugs modalg_6/bug26894
1. Unification of trimmed and not-trimmed cylinders processing (IntPatch_Intersection::GeomGeomPerfomTrimSurf() method has been removed).
2. Interface of IntPatch_ImpImpIntersection::Perform(...) method has been changed.
3. Now, WLine purging is forbidden for Geom-Geom-Intersection.
4. Bnd_Range class has been created. See Bnd_Range.hxx for detail information.
5. Algorithm of AddBoundaryPoint function has been improved in order to obtain intersection points in both boundaries (VFirst and VLast of every surface).
6. Earlier, method Geom2dConvert::ConcatG1(...) increased resulted B-spline degree (in case of not succession of previous iteration). Now increased value has been limited by Geom2d_BSplineCurve::MaxDegree() value (max degree = 25).
7. Algorithm of B-spline closure definition has been changed in the methods Geom2dConvert::C0BSplineToC1BSplineCurve(...) and Geom2dConvert::C0BSplineToArrayOfC1BSplineCurve(...).
Creation of test case for this issue.
Adjusting test cases according to their new behavior.
Small correction in the code according to KGV's remark.
The matter was that with starting point paased into intersector the walking line goes one point outside of the surface domain. Then during purging this extra point is removed from the line but its geometry is used for the last vertex. This makes a set of points invalid for approximation, and as a result we obtain the curve with reversed tangent direction at the end.
The API of the method IntPatch_WLineTool::ComputePurgedWLine has been changed to insert a new Boolean parameter RestrictLine. If this parameter is false than the step of removing of outside points is skipped, and the result line is not distorted. This flag is determined inside IntTools_FaceFace to tell the intersector if it is needed to limit intersection line by surface domain.
Test case has been added.
1. In frame of the fix for #27282 issue, we have obtained several prolonged curves, which have common point(s). Fix for this issue joins these curves if it is possible.
2. ElCLib::InPeriod(...) method has been improved. Now it has become more faster (in general cases) and more reliable (in frame of FLT_OVERFLOW and DIVISION_BY_ZERO cases processing).
Creation of test case for issue #27302
Test case tests\bugs\modalg_6\bug27282_2 has been adjusted in accordance with its new behavior.
1. The reason of the regression is not-closed intersection result. This problem has been solved (in this fix) by adding joint point to the both neighbors intersection lines (lines were extended to the this intersection point). It is made in IntPatch_WLineTool::ExtendTwoWlinesToEachOther(...) method.
2. Interface of IntPatch_PointLine and inherited classes has been changed. Methods ChangeVertex(...) and RemoveVertex(...) have been added.
Test cases for this issue have been created.
Small correction in the code.
Calling IntPatch_WLineTool::ComputePurgedWLine() algorithm is forbidden if Walking-line is obtained from analytic line.
Creation of test case for this issue.
Method DownCast() is made template, to be available only when argument is actually a pointer or handle to a base class.
For compatibility with existing code, method DownCast() that can be used for the same type, derived, or unrelated class (i.e. where there is no actual down casting) is still available, its use shall cause "deprecated" compiler warning.
OCCT code is updated to remove meaningless DownCast()s; a few places where DownCast() was used with argument of unrelated type are corrected.
DRAW command QAHandleCast is removed (it was useful only during redesign of handles).
1. Step of Walking-line has been increased as possible and now computes by iteratively adaptive algorithm (every iteration checks if current step is too big/small and decreases/increases one).
2. Interface of IntWalk_PWalking class has been changed
2.1. Method MaxStep() has been added in order to know about maximal distance between 2D-points.
2.2. Method ComputePasInit(...) has been added in order to initial step value computation.
2.3. Fields myTolTang (tolerance for intersection algorithm) and myStepMin (minimal step value) have been added.
Correction of some test cases.
Creation test case for this issue.
Changes to eliminate compiler warnings.
This branch contains fixes for 26675 and 26431 bugs.
1. Normalization has been eliminated.
2. Interfaces of AppDef_Compute::Parametrization(...) and BRepAlgo_BooleanOperations::SetApproxParameters() methods have been changed.
3. Overloaded methods for ApproxInt_Approx::SetParameters(...), TopOpeBRepTool_GeomTool::GetTolerances(...) and TopOpeBRepTool_GeomTool::SetTolerances(...) have been removed (because some fields of these classes are not used more).
4. Comments for some methods have been changed in BRepApprox_TheMultiLineOfApprox.hxx and GeomInt_TheMultiLineOfWLApprox.hxx files.
5. Some fields have been deleted from ApproxInt_MultiLine class. Kept members have become constant.
6. Interface of ksection DRAW-command has been changed.
7. Now, 2dintersect DRAW-command prints information about found segments.
8. Some code fragments have been rewritten to make them easier.
9. Algorithm of splitting WLine, which goes through pole of sphere has been improved.
10. Improve approximation algorithm in order to it will compute correct 2D- and 3D-tangent at the end of bezier constraints (including case when curve goes through or finishes on singular points).
11. Interface of IntPatch_WLine::Dump(...) method has been corrected.
12. Some methods for working with Walking-line are made more universal (available for both GeomInt and IntTools packages).
13. Problem in BRepLib::SameParameter(...) method has been fixed (see corresponding comment).
14. Small correction in Draft package.
15. Any outputs in IntPatch_Intersection::Dump(...) method have become disabled because they are useless. If anybody need in this outputs he/she will correct this method himself/herself.
Adjusting some test cases according to their new behavior.
Creation of new test cases.
----------------------------------------------------------------------------------------------------------------------------
Some explanation of new behavior of some test cases:
1. Regressions:
a) blend simple X4
The problem is described in the issue #0026740. According to this description, the result on the current MASTER seems to be wrong indeed.
b) boolean bcommon_complex C7 and boolean bcut_complex Q1
These test case use same shapes with different Boolean operation (COMMON and CUT). They are already BAD (on the MASTER). Now, some sub-shapes have become not-shared, simply. In my opinion, we shall apply new behavior of these tests.
c) boolean bsection M3
The problem described in the issue #0026777 exists even on the current MASTER.
d) boolean bsection M9
The problem is described in the message http://tracker.dev.opencascade.org/view.php?id=26815#c47546. Here, we have really regression in the picture.
e) boolean bsection N2
The problem is described in issue #0026814.
f) boolean volumemaker G1
The problem is described in issue #26020.
g) bugs modalg_1 bug1255 (and bug1255_1)
The problem is described in issue #26815.
h) bugs modalg_2 bug5805_18, bugs modalg_2 bug5805_42, bugs modalg_2 bug5805_46
The problem is described in issue #25925.
i) bugs modalg_3 bug602
The problem is describes in issue #602.
j) bugs modalg_5 bug24915
The problem is described in the message http://tracker.dev.opencascade.org/view.php?id=25929#c48565. It is not fixed by this issue.
k) bugs modalg_5 bug25838
The main reason is described in issue #0026816.
----------------------------------------------------------------------------
2. Improvements:
a) boolean volumemaker F9
b) bugs modalg_1 bug10160_3
c) bugs modalg_2 bug22557
d) bugs modalg_5 bug25319_1 (_2)
e) draft angle G2
f) offset shape A1
g) offset with_intersect_80 N7
1. JoinWLines algorithm has been improved.
2. Reference to the V-boundaries is deleted when computing step.
3. Decreasing the tolerance when computing parameters of WLine.
4. Adding boundary point is forbidden if it lies in prolongation of found ones.
5. Possible reason of exception has been eliminated.
6. Processing of critical point has been improved.
Test cases for this issue have been created.
Correction of some test case(s).
Test case bugs/modalg_6/bug26310_4 is added
Update of test-cases for issue #26310
Automatic upgrade of OCCT code by command "occt_upgrade . -nocdl":
- WOK-generated header files from inc and sources from drv are moved to src
- CDL files removed
- All packages are converted to nocdlpack
Automatic update by command "occt_upgrade . -downcast"
C-style cast of Handle to that of derived type (now illegal) is replaced by call to DownCast()
Const reference local variables of Handle type initialized by result of DownCast are replaced by normal variables.
1. Conditions for adjusting and for breaking Walking-lines have been amended.
2. Processing of case when WLine should be broken has been changed.
Test cases for issues 26193 and 26208 have been added
Cosmetic correction of test-cases
Modification of test-case according to the new behavior.
1. Restriction line is processed in IntTools_FaceFace with using methods of GeomInt_IntSS class.
2. Check, if Restriction- and Walking-lines (or Restriction-Restriction lines) are coincided, has been added in IntPatch_ImpPrmIntersection.cxx (at that RLine is considered to be isoline only).
3. Check, if RLine and GLine are coincided, has been added in IntPatch_ImpImpIntersection.cxx.
4. Create new class IntPatch_PointLine, which is inherited from IntPatch_Line.
5. The reason of exception (in DEBUG MODE) has been eliminated.
New test cases for issue #25416 were added.
tests/bugs/modalg_5/bug24650 was modified.
1. Curve boundaries were more precised.
2. Some test cases were changed in accordance of their new behavior.
3. Code fragment for WLines joining was moved to higher level (for more universal using in the future).
Code was changed in accordance with the last remark.
Added test cases bugs/modalg_5/bug25715_1 bug25715_2 bug25715_3
Existing method of Cylinder-Cylinder intersection computing is based on finding the analytic line (as a function of one argument) and converting one into the walking-line with set of equidistant (along the line parameter) points.
The main advantage of applied method is using adaptively computed step. Necessary step is computed into every point of the obtained walking-line. At that we receive final walking-line directly (without preliminary analytic line) and we determine moments more precisely, when it should be split (see IntPatch_ImpImpIntersection_4.gxx).
The main disadvantages is bad working this method for non-trimmed cylinders (with infinite bounds), because step value is depend on the boundaries values.
More over, new method always returns walking-line, while intersection result can be an analytic curve (lines, circle, ellipse). That is NO good. Therefore, analytic curve is computed by existing method.
In conclusion, in spite of covering almost all more often meeting cases, new method has limited application. Then we should use the existing old method.
Additionally, method MinMax() is added (see Standard_Real.hxx file). It uses into new algorithm.
Some test cases is changed according to their new behavior.
Test case for issue CR24915 is added.
Into GeometryTest_APICommands.cxx only tabulations were chaged.
"Extending" of isolines (see Geom2dHatch_Hatcher.cxx).
Small correction of test case for issue CR24915.
Includes of Handle*.hxx headers replaced by include of corresponding class (where possible without redesign of the classes).
Minor correction against compiler warning (vc10 64-bit) in BOPAlgo
Remove double header inclusions
Correction of minor mistake
Main changes are in "IntWalk_IWalking::TestDeflection(...)" function (IntWalk_IWalking_5.gxx).
Some test cases were corrected according to their new behavior.
License statement text corrected; compiler warnings caused by Bison 2.41 disabled for MSVC; a few other compiler warnings on 54-bit Windows eliminated by appropriate type cast
Wrong license statements corrected in several files.
Copyright and license statements added in XSD and GLSL files.
Copyright year updated in some files.
Obsolete documentation files removed from DrawResources.
The intersection between Cone and one of the following surfaces: Cylinder, Sphere, Cone and Torus
is treated as analytical when the axes of the surfaces are collinear.
Test case for issue CR24505
Modifications:
1. The validity of the intersection point between edge and face is checked with sum of the tolerance values of the edge and face.
2. The intersection between toroidal surface and one of the following surfaces: Plane, Cylinder, Sphere, Cone or Torus
is treated as analytical in the following cases:
1) Torus and Plane: a. Axes of the surfaces are parallel;
b. Axes of the surfaces are perpendicular and location of the torus is lying on the plane;
2) Torus and Sphere: The location of the sphere is lying on the line made from toruses axis;
3) Torus and Cone, Cylinder or Torus: The axis of the surfaces are collinear.
In all cases the intersection line(s) is(are) circle(s).
Added test cases bugs/modalg_5/bug24470
The intersection between torus with minor radius more than (or equal to) the major radius and any other surface is considered as parametric.
Eliminated warnings -Wswitch shown by gcc
Eliminated two new warnings about wrong initialization order
Eliminated warnings about unused functions.
A few more ICC warnings (type conversions) fixed
