Sergey, I believe the IndexMeshVisitor attempts to detect and remove duplicate vertices. Could that be the reason?
Glenn Waldron / @glennwaldron On Thu, Apr 25, 2013 at 11:16 AM, Sergey Polischuk <[email protected]> wrote: > Hi all > > once i've wrote small utility to optimize models with rig geometry, which, > among other things, apply INDEX_MESH, VERTEX_PRETRANSFORM, > VERTEX_POSTTRANSFORM optimizations and remaps vertex influences of rig > geometry to new vertex indices. Now it suddenly dont work any more. > > Problem is that after indexing geometry, there are not enough matches > between old vertex attributes and new vertex attributes (some old vertex > attribute values cant be found in new vertex attribute arrays after > indexing, which means some data is lost). > > here is some simplified code to show problem: > > ////////////////////////////////////////////////////////////// > > template <typename T> bool compare(T* r, unsigned ri, T* l, unsigned li) > //compares two array elements > { > if ((*r)[ri] < (*l)[li]) > return false; > if ((*l)[li] < (*r)[ri]) > return false; > > return true; > } > > bool cmp (osg::Geometry::ArrayList& rhs, unsigned rhsi, > osg::Geometry::ArrayList& lhs, unsigned lhsi) //compares all [rhsi] > elements from rhs array list with all [lhsi] elements in lhs array list > { > if (rhs.size() != lhs.size()) > throw std::logic_error("different vertex attrib arrays > count before and after optimizations, cant do anything"); > bool result = true; > for (unsigned i = 0; i< rhs.size(); i++) > { > bool compared = false; > if (!result) > return result; > { > osg::Vec3Array* r = > dynamic_cast<osg::Vec3Array*>(rhs[i]); > osg::Vec3Array* l = > dynamic_cast<osg::Vec3Array*>(lhs[i]); > if (r && l) > { > result = result && compare<>(r, rhsi, l, > lhsi); > compared = true; > } > } > { > osg::Vec2Array* r = > dynamic_cast<osg::Vec2Array*>(rhs[i]); > osg::Vec2Array* l = > dynamic_cast<osg::Vec2Array*>(lhs[i]); > if (r && l) > { > result = result && compare<>(r, rhsi, l, > lhsi); > compared = true; > } > } > if (!compared) > return false; > } > return result; > } > > void acceptMesh(osgAnimation::RigGeometry* rg) > { > osg::Geometry* g = rg->getSourceGeometry(); > > osg::ref_ptr<osg::Geometry> origG = new osg::Geometry(*g, > osg::CopyOp::DEEP_COPY_ALL); > > osg::Geometry::ArrayList alold; > origG->getArrayList(alold); > osg::ref_ptr<osg::Vec3Array> origVerts = > dynamic_cast<osg::Vec3Array*>(origG->getVertexArray()); > > osgUtil::IndexMeshVisitor imv; > imv.makeMesh(*g); > > osg::Geometry::ArrayList alnew; > g->getArrayList(alnew); > > std::map<int, int> vertsMap;//store matches between old > and new vertexattrib arrays in pairs [old_index, new_index] > osg::ref_ptr<osg::Vec3Array> newVerts = > dynamic_cast<osg::Vec3Array*>(g->getVertexArray()); > > unsigned id = 0; > for (osg::Vec3Array::iterator itr = origVerts->begin(); > itr != origVerts->end(); itr++) > { > unsigned nid = 0; > for (osg::Vec3Array::iterator itr2 = > newVerts->begin(); itr2 != newVerts->end(); itr2++) > { > if (cmp(alold, id, alnew, nid))//found > match > { > vertsMap[id] = nid; > break; > } > nid++; > } > id++; > } > std::cerr<<"origVerts size:"<<origVerts->size()<<std::endl; > std::cerr<<"verts map size:"<<vertsMap.size()<<std::endl; > } > > ///////////////////////////////////////////////////////////// > > here size of vertsMap map should be same as original vertex count, and it > isnt, which means some data is lost when indexing geometry (and comparing > floating point values with thresholds does not change anything, i've tested > it). > > Any ideas about what could go wrong? May be there was some changes to the > way IndexMeshVisitor works? I'm testing on osg svn version updated last > week or so. > > Cheers, > Sergey. > _______________________________________________ > osg-users mailing list > [email protected] > http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org >
_______________________________________________ osg-users mailing list [email protected] http://lists.openscenegraph.org/listinfo.cgi/osg-users-openscenegraph.org
