Is delete_cells_by_point_index parallelisable?

[JeffreyWardman]

Current code:

  def delete_cells_by_point_index(self, indices):
      """
      Delete a list of vertices identified by any of their vertex index.

      See also `delete_cells()`.

      Examples:
          - [delete_mesh_pts.py](https://github.com/marcomusy/vedo/tree/master/examples/basic/delete_mesh_pts.py)

              ![](https://vedo.embl.es/images/basic/deleteMeshPoints.png)
      """
      cell_ids = vtki.vtkIdList()
      self.dataset.BuildLinks()
      n = 0
      for i in np.unique(indices):
          self.dataset.GetPointCells(i, cell_ids)
          for j in range(cell_ids.GetNumberOfIds()):
              self.dataset.DeleteCell(cell_ids.GetId(j))  # flag cell
              n += 1

      self.dataset.RemoveDeletedCells()
      self.dataset.Modified()
      self.pipeline = OperationNode("delete_cells_by_point_index", parents=[self])
      return self

Are there any issues with parallelising these two for loops? Even if it's via setting the number of jobs with joblib? It doesn't scale well (e.g. deleting half of a 150,000 point mesh). I'm not sure how VTK datasets work under the hood.

[marcomusy]

I think that should be possible: the DeleteCell() call only flags a cell for deletion and the actual work is done by RemoveDeletedCells(). But I dont know which one is the bottleneck.