How to create PointPartitioner from data that is read in ?

[nyue]

Hi,

After reading in an OpenVDB 4 file with point data, how does one create a partitioner ?

I am looking to the partitioner to provide indices to actual points (per bucket?) and query additional information like position, velocity and other attributes.

So far, I don't know what method in PointDataGrid returns object that can be use for the construction of the Partitioner, here is my current attempt, ::create() fails to compile as I am not providing what it needs

I see the example creating a struct call PointList within an anonymous namespace

`
openvdb::io::File fileIn(vdb_file);
fileIn.open();

    openvdb::GridPtrVecPtr grids = fileIn.getGrids();

    fileIn.close();

	std::cout << boost::format("Grid size = %1%") % grids->size() << std::endl;

	if (!grids->empty())
	{
		openvdb::points::PointDataGrid::Ptr inputPointDataGrid = openvdb::GridBase::grid<openvdb::points::PointDataGrid>((*grids)[0]);
		if (inputPointDataGrid)
		{
			if (!inputPointDataGrid->empty())
			{
			    const float voxelSize = 0.1f;
			    const openvdb::math::Transform::Ptr transform =
			            openvdb::math::Transform::createLinearTransform(voxelSize);

			    typedef openvdb::tools::UInt32PointPartitioner PointPartitioner;

			    PointPartitioner::Ptr partitioner =
			                PointPartitioner::create(pointList, *transform);

`

Cheers

[danrbailey]

Hi @nyue,

I'm currently working on getting the code samples and documentation ready for submission into OpenVDB, so that should help explain this better. In short, the PointPartitioner is only for use with a PointIndexGrid where the attribute data is stored in a linear external array.

Here's a super-simple example for iterating over position data in a PointDataGrid:

for (auto leafIter = grid.tree().cbeginLeaf(); leafIter; ++leafIter) {
    auto handle = openvdb::points::AttributeHandle<Vec3f>::create(leafIter->constAttributeArray("P"));
    for (auto iter = leafIter->beginIndexOn(); iter; ++iter) {
        const Vec3d voxelCoord = iter.getCoord().asVec3d();
        const Vec3d positionVoxelSpace = handle->get(*iter);
        const Vec3d positionWorldSpace = grid.transform().indexToWorld(positionVoxelSpace + voxelCoord);
        std::cerr << "Position: " << positionWorldSpace << std::endl;
    }
}

Hope that helps.

Dan

[nyue]

Thank you.

In your code

leafIter->constAttributeArray("P")

How do you know there will be an attribute call "P" ?

How does one programmatically find all the other attribute names ?

Cheers

[danrbailey]

"P" is a mandatory attribute.

For simple inspection, you can use some of the convenience methods on the PointDataLeaf:

bool test1 = leafIter->hasAttribute("P");
bool test2 = leafIter->hasAttribute(/*index=*/4);

Otherwise, you'll need to go through the Descriptor:

const AttributeSet::Descriptor& descriptor = leafIter->attributeSet().descriptor();
for (const auto& namePos : descriptor.map()) {
    const Name& name = namePos.first;
    const size_t index = namePos.second;
    const Name& valueType = descriptor.valueType(index);
}

Although the API doesn't enforce this, by convention we assume that the Descriptor is the same across all LeafNodes.