MantisBT - Open CASCADE
View Issue Details
0029699Open CASCADE[OCCT] OCCT:Application Frameworkpublic2018-04-17 12:442020-09-10 19:00
ssv 
ssv 
normalminor 
feedbackopen 
WindowsVC++ 201364 bit
 
[OCCT] 7.6.0* 
0029699: Application Framework - OCAF ignores orientation of shape in TNaming
The programmatic scenario is given in "Steps To Reproduce" section. Here is the overview:

Imagine you store a FORWARD face in OCAF using TNaming_Builder. Then imagine you reverse this face and store it again in the same attribute. Here is a bug: OCAF ignores the attempt to store the REVERSED face and keeps storing the FORWARD one.

I can imagine that TNaming_Builder::Generated() is not probably the best way to set shapes in OCAF (is there an alternative?). On the other hand, this problem is not reproduced for shapes with different TShape pointers (that is probably why we do not have this bug in parametric modelers).
// Take any label, where doc is TDocStd_Document.
TDF_Label aDataLab = doc->Main();

// Create just any face.
TopoDS_Shape
  face = BRepBuilderAPI_MakeFace( gp_Pln( gp::Origin(), gp::DZ() ) );

// Orientation is FORWARD initially.
std::cout << "Expected orientation: "
          << ( (face.Orientation() == TopAbs_FORWARD) ? "forward" : "reversed" ) << std::endl;

// Set shape.
doc->NewCommand();
{
  TNaming_Builder aNBuilder(aDataLab);
  aNBuilder.Generated(face);
}
doc->CommitCommand();

// Get shape to check its orientation.
Handle(TNaming_NamedShape) aShapeAttr;
aDataLab.FindAttribute(TNaming_NamedShape::GetID(), aShapeAttr);
face = aShapeAttr->Get();

// Orientation is forward (OK).
std::cout << "Orientation in OCAF: "
          << ( (face.Orientation() == TopAbs_FORWARD) ? "forward" : "reversed" ) << std::endl;

// Reverse shape before storing.
face.Reverse();

// Orientation is now REVERSED.
std::cout << "Expected orientation: "
          << ( (face.Orientation() == TopAbs_FORWARD) ? "forward" : "reversed" ) << std::endl;

// Set reversed shape.
doc->NewCommand();
{
  TNaming_Builder aNBuilder(aDataLab);
  aNBuilder.Generated(face);
}
doc->CommitCommand();

// Get shape to check orientation.
aDataLab.FindAttribute(TNaming_NamedShape::GetID(), aShapeAttr);
face = aShapeAttr->Get();

// BUG: Orientation is still forward, while it should be reversed (NOK).
std::cout << "Orientation in OCAF: "
          << ( (face.Orientation() == TopAbs_FORWARD) ? "forward" : "reversed" ) << std::endl;
No tags attached.
Issue History
2018-04-17 12:44ssvNew Issue
2018-04-17 12:44ssvAssigned To => mpv
2018-04-23 10:26mpvNote Added: 0075594
2018-04-23 10:26mpvStatusnew => acknowledged
2019-09-04 18:05kgvSummaryOCAF ignores orientation of shape in TNaming => Application Framework - OCAF ignores orientation of shape in TNaming
2019-09-05 04:58abvNote Added: 0086764
2019-09-05 04:58abvStatusacknowledged => assigned
2019-09-05 04:58abvTarget Version7.4.0 => 7.5.0
2019-09-05 09:21mpvNote Added: 0086772
2019-09-05 09:21mpvAssigned Tompv => ssv
2019-09-05 09:21mpvStatusassigned => feedback
2019-09-05 10:34ssvNote Added: 0086780
2020-09-10 19:00ssvTarget Version7.5.0 => 7.6.0*

Notes
(0075594)
mpv   
2018-04-23 10:26   
This is known and norma behavior of the NamedShapeAttribute, since it keep all the shapes in the UsedShapes attribute, which contains the not-oriented shapes of map.
Ther naming mechanism never works with orientation of shapes and trying to make this map oriented map of shapes will casue huge number of regressions in the naming and in application that uses it.

So, I may only propose to use additional flag (or integer if FORWARD/REVERSED is not enough) attribute to keep the orientation of the shape, or create a new attribute, which will have no relation to the topological naming.
(0086764)
abv   
2019-09-05 04:58   
Mikhail, the described behavior is clearly inconsistent and may not be considered as normal. Please consider possible ways to fix it.
(0086772)
mpv   
2019-09-05 09:21   
Sergey, is it help you if TDataXtd_Shape attribute (which stores the full shape only) will store the shape orientation, but not TNaming_NamedShape, which is related more to a local selection?
(0086780)
ssv   
2019-09-05 10:34   
Mikhail, the issue is quite old and is already workarounded by deep copying a face whenever its orientation is reversed. It is up to you to decide whether and how to fix the reported inconsistency. I am sure that I had a look at TDataXtd_Shape as we did not really need the naming service. However I do not remember why it was not suitable (probably, it did not work well on undo/redo or anything else related to backup). Anyway, the issue deserves deeper investigation which I cannot conduct right now.