Maybe there is a bug on Handle(Geom_Curve)->value()? Hope I am wrong T.T

[jianbaoxia]

Hi Bros, I'm JianBaoxia.
Today, while using Handle(Geom_Curve)->value() to obtain some points on a B-spline curve, I encountered some strange phenomena. Therefore, I couldn't wait to discuss and share with you. To illustrate the problem, I wrote a simple demo, which is as follows:
I read in a B-spline curve with parameter values ranging from 0 to 1. Using Handle(Geom_Curve)->value(), I generated five points on the curve with parameter values of {0, 0.005, 0.5, 0.995, 1}. Then, I used "GeomAPI_ProjectPointOnCurve" to calculate the distance from each of the five points to the curve to determine whether they were located on the curve. The answer was affirmative, as the maximum distance between the points and the curve was at the order of 1e-15, which is far smaller than the tolerance of 1e-7. However, when I outputted the points and the curve together, I found that only the first and last points were truly on the curve, while the middle three had some distance away from the curve. I measured the distances of the middle three points on FreeCAD and obtained the values of 1.2e-1, 4.07e-7, and 8.21e-5, respectively. Clearly, the errors of these three points are all greater than the tolerance of 1e-7. As show below:

So, I'm very puzzled as to where the problem lies. According to the formula for NURBS curves, there should be no errors in calculating the points on the curve. Could it be a display issue? Are these five points actually completely on the curve? However, if that were the case, there should be some visible deviation for the first and last points on FreeCAD, but in fact, these two points look completely on the curve.
There is my demo、B-spline curve file: "aEdge.step" and result file: "aSewShape.step":

int main()
{
    // read the file
   
    STEPControl_Reader reader;
    reader.ReadFile("../aEdge.step");
    reader.TransferRoots();
    TopoDS_Shape aShape = reader.OneShape();

   
    TopoDS_Edge aEdge;
    for (TopExp_Explorer exp(aShape, TopAbs_EDGE); exp.More(); exp.Next())
        aEdge = TopoDS::Edge(exp.Current());
    Standard_Real _1, _2;    // no use
    Handle(Geom_Curve) aCur = BRep_Tool::Curve(aEdge, _1, _2);

    // sample five points on curve
    TColStd_Array1OfReal sampArr(1, 5);
    TColgp_Array1OfPnt sampPntsArr(1, 5);
    sampArr[1] = 0.;
    sampArr[2] = 0.005;
    sampArr[3] = 0.5;
    sampArr[4] = 0.995;
    sampArr[5] = 1.;
    for (int i = 1; i <= 5; i++)
    {
        sampPntsArr = aCur->Value(sampArr);
    }

    // test whether pnt on curve
    for (int i = 1; i <= 5; i++)
    {
        gp_Pnt pnt = sampPntsArr;
        GeomAPI_ProjectPointOnCurve gppc(pnt, aCur);        // get the shortest distance bettwen point and curve
        if (gppc.LowerDistance() <= Precision::Confusion())
            std::cout << "point[" << i << "]: on the curve, the shortest distance is: " << gppc.LowerDistance() << std::endl;
        else
            std::cout << "! ERROR: point[" << i << "]: not on the curve, the shortest distance is" << gppc.LowerDistance() << std::endl;

    }
    std::cout << "The Precision::Confusion(): " << Precision::Confusion() << std::endl;
    
    // out put the point and edge for view
    BRepBuilderAPI_Sewing aSewMaker;
    aSewMaker.Add(aEdge);
    for (int i = 1; i <= 5; i++)
        aSewMaker.Add(BRepBuilderAPI_MakeVertex(sampPntsArr));
    aSewMaker.Perform();
    TopoDS_Shape aSewShape = aSewMaker.SewedShape();

    STEPControl_Writer writer;
    Interface_Static::SetCVal("write.step.schema", "AP203");
    IFSelect_ReturnStatus status = writer.Transfer(aSewShape, STEPControl_AsIs);
    status = writer.Write("../aSewShape.step");

    return 0;
}

 

[jianbaoxia]

I used the NURBS curve formula to verify the values of these five points, and it turned out that the values were correct. It is possible that the way I integrated the points and lines together was not appropriate, but if I don't use "BRepBuilderAPI_Sewing", what should I use instead? Or perhaps there are other reasons that have resulted in this less than ideal outcome?

 

[blobfish]

No you don't want to run that through sew. Can't you just write each shape individual to the step file? If not, you can wrap them in a compound.

 

[jianbaoxia]

No you don't want to run that through sew. Can't you just write each shape individual to the step file? If not, you can wrap them in a compound.

Sorry, I made a low-level mistake again. I am self-learning OCCT online, and as a self-taught person, I always make stupid mistakes. I used to think that "STEPControl_Writer" can only output one TopoDS_Shape at a time, but it turns out that it's not true. Therefore, I have been using "BRepBuilderAPI_Sewing" to achieve it, and I feel ridiculous about myself.