ability to read unstructured data from legacy vtk format

[dbochkov-flexcompute]

• adds the ability to read from legacy VTK format .vtk in addition to existing .vtu
• adds an option to not error when loading unstructured grid with non-triangle/-tetrahedral cells (for example, to skip triangle elements that are not relevant)
• fixes a bug when loading from vtk file without any point data

todo:

• unit tests
• changelog entry
• missing docstrings

Greptile Summary

Enhances VTK file support by adding legacy .vtk format compatibility and flexible cell type handling in unstructured grids.

• Added support for reading legacy VTK format (.vtk) files in addition to existing .vtu format
• Introduced ignore_invalid_cells parameter to gracefully handle non-triangular/-tetrahedral cells instead of erroring
• Fixed bug when loading VTK files without point data by properly handling empty coordinate lists
• Modified cell type validation in _from_vtk_obj to support mixed cell types when ignore_invalid_cells=True
• Missing unit tests, changelog entry, and docstrings need to be addressed

[github-actions]

Diff Coverage

Diff: origin/develop...HEAD, staged and unstaged changes

• tidy3d/components/data/unstructured/base.py (42.9%): Missing lines 490-493,495,596-597,640
• tidy3d/components/data/unstructured/tetrahedral.py (28.6%): Missing lines 115-118,122
• tidy3d/components/data/unstructured/triangular.py (33.3%): Missing lines 151-153,157

Summary

• Total: 27 lines
• Missing: 17 lines
• Coverage: 37%

tidy3d/components/data/unstructured/base.py

Lines 486-499

  486     @staticmethod
  487     @requires_vtk
  488     def _read_vtkLegacyFile(fname: str):
  489         """Load a grid from a legacy `.vtk` file."""
! 490         reader = vtk["mod"].vtkGenericDataObjectReader()
! 491         reader.SetFileName(fname)
! 492         reader.Update()
! 493         grid = reader.GetOutput()
  494 
! 495         return grid
  496 
  497     @classmethod
  498     @abstractmethod
  499     @requires_vtk

---

Lines 592-601

  592         -------
  593         UnstructuredGridDataset
  594             Unstructured data.
  595         """
! 596         grid = cls._read_vtkLegacyFile(file)
! 597         return cls._from_vtk_obj(
  598             grid,
  599             field=field,
  600             remove_degenerate_cells=remove_degenerate_cells,
  601             remove_unused_points=remove_unused_points,

---

Lines 636-644

  636             log.warning(
  637                 "No point data is found in a VTK object. '.values' will be initialized to zeros."
  638             )
  639             values_numpy = np.zeros(num_points)
! 640             values_coords = {"index": np.arange(num_points)}
  641             values_name = None
  642 
  643         else:
  644             field_ind = field if isinstance(field, str) else 0

---

tidy3d/components/data/unstructured/tetrahedral.py

Lines 111-126

  111         # verify cell_types
  112         cells_types = vtk["vtk_to_numpy"](vtk_obj.GetCellTypesArray())
  113         invalid_cells = cells_types != cls._vtk_cell_type()
  114         if any(invalid_cells):
! 115             if ignore_invalid_cells:
! 116                 cell_offsets = vtk["vtk_to_numpy"](vtk_obj.GetCells().GetOffsetsArray())
! 117                 valid_cell_offsets = cell_offsets[:-1][invalid_cells == 0]
! 118                 cells_numpy = cells_numpy[
  119                     np.ravel(valid_cell_offsets[:, None] + np.arange(4, dtype=int)[None, :])
  120                 ]
  121             else:
! 122                 raise DataError("Only tetrahedral 'vtkUnstructuredGrid' is currently supported")
  123 
  124         # pack point and cell information into Tidy3D arrays
  125         num_cells = len(cells_numpy) // cls._cell_num_vertices()
  126         cells_numpy = np.reshape(cells_numpy, (num_cells, cls._cell_num_vertices()))

---

tidy3d/components/data/unstructured/triangular.py

Lines 147-161

  147         # verify cell_types
  148         cell_offsets = vtk["vtk_to_numpy"](cells_vtk.GetOffsetsArray())
  149         invalid_cells = np.diff(cell_offsets) != cls._cell_num_vertices()
  150         if any(invalid_cells):
! 151             if ignore_invalid_cells:
! 152                 valid_cell_offsets = cell_offsets[:-1][invalid_cells == 0]
! 153                 cells_numpy = cells_numpy[
  154                     np.ravel(valid_cell_offsets[:, None] + np.arange(3, dtype=int)[None, :])
  155                 ]
  156             else:
! 157                 raise DataError(
  158                     "Only triangular 'vtkUnstructuredGrid' or 'vtkPolyData' can be converted into "
  159                     "'TriangularGridDataset'."
  160                 )

---

[greptile-apps]

_{3 files reviewed, 1 comment
Edit PR Review Bot Settings | Greptile}

[greptile-apps]

style: Error message should specify which invalid cell types were found to help debugging

[momchil-flex]

Looks good to me and seems to work in my separate meshing pipeline PR for both 3D and 2D meshes. Just needs some tests yeah.

[marc-flex]

Thanks Daniil. This looks good.

In terms of testing, how do you plan on doing this? I'm should probably test a function in a different PR and I didn't really know how to do it.

[momchil-flex]

Closing in favor of #2493