yang686526

floating point and tranformation matrix
floating point and tranformation matrix
hi,
we experienced a problem to use transformby method with oddb3dsolidptr. when we create a transformation matrix with this code. (the xaxis, yaxis, and zaxis comes from our orthonormal structure)
code:
odgematrix3d dwgdatabase::gettransformmatrix(const odgepoint3d& origin, const odgepoint3d& xaxis, const odgepoint3d& yaxis, const odgepoint3d& zaxis)
{
odgevector3d xvector = odgevector3d(xaxis.x, xaxis.y, xaxis.z);
odgevector3d yvector = odgevector3d(yaxis.x, yaxis.y, yaxis.z);
odgevector3d zvector = odgevector3d(zaxis.x, zaxis.y, zaxis.z);
odgematrix3d xfm;
xfm.setcoordsystem(odgepoint3d(origin.x, origin.y, origin.z), xvector, yvector, zvector);
return xfm;
}
then use this matrix to transform a oddb3dsolid object, we got an odresult of ecannotscalenonuniformly. i think we need to do something with the tolerance for the floating point, but i cannot find a way to set the tolerance for the matrix or the oddb3dsolid object. any idea of how to fix this problem?
thank you very much.
hello,
transformby uses next tests for input matrix. please, check your input parameters for isuniscaledortho.
code:
if ( xform.issingular() ) return edegenerategeometry;
if ( !xform.isuniscaledortho() ) return ecannotscalenonuniformly;
also
code:
/////////////////////////////////////////////////////////////////////////////////
// test scaling effects of matrix
// 12.05.2004 gu in acad this function doesn't take perspective into account
// 30.12.2004 d.novikov: acad treats zero vectors as perpendicular && codirectional
// 07.03.2007 sv: acad uses tolerance for vectors while comparing length of coordinate system vectors
bool odgematrix3d::isuniscaledortho(const odgetol& tol) const
{
odgevector3d x, y, z;
odgepoint3d origin;
getcoordsystem(origin, x, y, z);
double xlen = x.length();
double ylen = y.length();
double zlen = z.length();
if ( odequal(xlen, ylen, tol.equalvector()) &&
odequal(xlen, zlen, tol.equalvector()) &&
odequal(ylen, zlen, tol.equalvector())
)
{
return x.isperpendicularto(y, tol) &&
y.isperpendicularto(z, tol) &&
z.isperpendicularto(x, tol);
}
return false;
}
hi alexander,
the problem is transformby uses the test method xform.isuniscaledortho() not xform.isuniscaledortho(tol). take a look at this code
code:
odgetol tol = odgetol(0.0001f);
// test for orthonormal with tolerance, it is ok.
if(xform.isuniscaledortho(tol))
{
odresult result = p3dsolid->transformby(xform);
// transformby failed with result of ecannotscalenonuniformly
}
is there a way to force transformby use xform.isuniscaledortho(tol) to test the input matrix?
transformby() has no tolerance parameter. it uses default tolerance (1.e-10) to check transformation matrix for being uniscaled and ortho.
sergey slezkin
our orthonormal structure uses tolerance of 1.e-5 that is why we got the problem. beside using setcoordsystem, we try to use settorotation(angle, axis, center) to create a transformation matrix, but we cannot get a correct oddb3dsolid after transformation. can you please explain how use settorotation and its parameters?
what does incorrect oddb3dsolid mean? does matrix::settoidentity(or settotranslation) generate correct 3dsolid after transformby?
i just got it correct. i have odgepoint3d origin, odgevector3d xaxis, yaxis, and zaxis. instead of using setcoordsyste to create a transformation matrix, i can use this code to create a transformation matrix with the rotation axis and angle coming from orthonormal odgevector3d xaxis, yaxis, and zaxis.
code:
odgematrix3d* xfm = new odgematrix3d();
odgevector3d vec = odgevector3d(axis.x, axis.y, axis.z);
xfm->settorotation(angle, vec);
xfm->settranslation(odgevector3d(origin.x, origin.y, origin.z));
this help to avoid the floating problem with the transformation matrix.
thank you.