issue_comments

do you want to add this to the ecosystem page in the docs?

I'd love to! Thanks for the suggestion @keewis! I'll track this in https://github.com/pyvista/pyvista-xarray/issues/21 to contribute after I make my next round of improvements to pyvista-xarray

Following up to mention that I've developed pyvista-xarray which provides an interface between vtkRectilinearGrids and xarrays DataArrays via PyVista: https://github.com/pyvista/pyvista-xarray

This is very much in its early stages and I’m hoping to take it quite a bit further. But first, I’d love to gather feedback and find how it works or doesn’t work for the types of data folks have.

At present, this accessor is specifically for rectilinear-style grids but if you have some other types of data, do share so that I can work towards supporting it in pyvista-xarray by opening an issue or discussion over there.

(there is limited support for vtkStructuredGrids as well)

Just have to find some time to start.

Ok, so a discussion:-

https://github.com/pyvista/pyvista/discussions/2375

I am very interested in this sort of functionality as an xarray accessor. If I can help in any way, please let me know. Ideally this work would come in very handy to visualize Oxygen Minimum Zones in the global ocean as isosurfaces of a 3D oxygen array.

e.g. for a Uniform Grid

python data = result['density'].reshape(result.dimensions[2],result.dimensions[1],result.dimensions[0]) XC = [result.bounds[0]+200*i for i in range(result.dimensions[0])] YC = [result.bounds[2]+200*i for i in range(result.dimensions[1])] ZC = [result.bounds[4]+200*i for i in range(result.dimensions[2])] da = xr.DataArray(data=data, dims=["z", "y","x"], coords={"z": ZC, "y": YC, "x": XC})

So reversing that would require storing the metadata etc. and whatever the meshgrid reversal is?

I use this sort of thing a lot, anyway. https://banesullivan.com/pyvista/examples/raster.html#sphx-glr-pyvista-examples-raster-py

Yes, that sounds good. Definitely be useful to put this in some workflows.

I think a first step would be to make a data accessor between xarray and VTK/PyVista in accordance with https://xarray.pydata.org/en/stable/internals/extending-xarray.html

This keeps coming up again and again for me. @RichardScottOZ, if you're interested in kicking this off, I'd be happy to chat and help implement this and then outline some future direction for this sort of "integration"

And again mesh to data array and back, by variety?

Ryan what would you start with? I had a use case this week.

The same dataset as in that Visualizing_Meshes notebook is also shown in https://examples.pyviz.org/bay_trimesh, which is more self-contained and may be a better starting point.

Also note that VTK and pyvista are well supported by Panel, as illustrated in various examples at https://panel.holoviz.org/gallery. It would be great to update those examples to include xarray data sources to more fully capture the pipeline from data to display.

I'm not sure what the specific pain points are with UGRID or what specifically is being requested at https://github.com/pyviz-topics/EarthSim/issues/326, but we'd be happy to have examples either in EarthSim or at examples.pyviz.org that start with UGRID datasets and go via Xarray to either 2D or VTK-based 3D visualizations.

I'm really excited about this effort, thanks @rabernat! One thing I'd like to do before I forget is mention a project from the Software Underground (SWUNG) to create a new subsurface package: https://github.com/softwareunderground/subsurface

Over there, we are working to create a set of low-level data structures for sharing spatial data between various subsurface modeling/analysis software (e.g. GemPy, SimPEG, PyGIMLi, Fatiando, Welly, and more) with close ties to PyVista (VTK) for 3D visualization. At present, we are using xarray as the base data container for structured data types. This morning, @rabernat gave me some great insight into how xarray can be used to contain unstructured data as well and so I will be investigating how to do that so we leverage xarray as the base data container all around in subsurface. Then with subsurface's close ties to PyVista, there comes a bridge between xarray and VTK (via PyVista).

cc @Leguark and @prisae

This would be a proof of concept for the round-trip data interoperability going between xarray and VTK leaving us with some good lessons on what the value would be of a direct interface between VTK and xarray for better handling of large data and leveraging Dask.

Either way, it'd be awesome to have the subsurface development team in the loop here as there appears to be a lot of shared goals