adjTomo future features

[bch0w]

This is a Discussion tab for future features that we would like to include in the adjTomo software suite. These would either constitute individual packages, additions to current packages, or amendments to current features. Users/developers can also place wish lists here to discuss potential features.

To start out the discussion I would put a few main categories and how they might be handled:

[bch0w]

Cartesian Meshing

For simple Cartesian domains, we tend towards the open-source internal mesher (Meshfem). However there are a lot of internal rules and requirements needed to create a valid finite-element mesh with Meshfem. There are also (atleast) 3 separate files that need to be edited in parallel when meshing choices are made. This results in a very difficult time for new users who just want to generate a simple, layered mesh.

https://github.com/bch0w/simutils/tree/master/meshing/meshfem

During my PhD I developed this Python script that simplified this entire procedure. Users use a parameter file to set a bounding box, desired element size, and rough locations for doubling layers, interfaces etc. The script would then output all required Meshfem files (except topography), with meshes that more or less followed the constraints of Meshfem and created low-skew meshes.

https://github.com/bch0w/simutils/blob/master/meshing/meshfem/prepare_meshfem.py

The script is fast and has detailed logging, meaning a User can quickly prototype mesh configurations without running xmeshfem??. This is invaluable as things like cluster queue time might make this mesh development procedure take an inordinate amount of time.

This would be great to roll into the adjTomo capabilities. Not sure if it deserves it's own package, perhaps it can be rolled into Pyatoa, or another utility package? Script is mostly there, just needs a little polishing to be placed in the adjTomo suite.

[bch0w]

Mesh/Model/Gradient Visualization

Visualizing 3D mesh/model/gradients is necessary for understanding inversions, and writing publications. However there is no current standard on how to do this in an efficient manner. From my understanding most people take the native .VTK files from SPECFEM and import them into ParaView, and then manually manipulate them to get the desired figures.

This is a difficult procedure because ParaView is finnicky and has a steep learning curve, .vtk files can be large and difficult to move on filesystems or manipulate in ParaView, and ParaView can be difficult to automate so repeat figures are not simple to create.

Options include:

1. Streamlining ParaView using bash scripts and state files (.pvsm): This the most manual task, where you set up your figure and then manually change file names in the state files (e.g., https://github.com/bch0w/simutils/blob/master/paraview/modify_state_file.py).
2. Automating ParaView using ParaView Python: ParaView has it's own Python interface which can be used to automate file manipulation. It is not intuitive, and difficult to write. I did this in my PhD (https://github.com/bch0w/simutils/blob/master/tools/pvpyplot.py) so there is a starting point, but I think it would be difficult for new Users to learn how to do this, it would likely constitute a black box tool.
3. PyVista (https://docs.pyvista.org/): Heard this name thrown around a few times. Seems like a nice Python interface for interacting with .vtk files
4. Mayavi (https://docs.enthought.com/mayavi/mayavi/): Similar to PyVista, Python interface to interacting with .vtk files. I've tried this out in Pyatoa (https://github.com/adjtomo/pyatoa/blob/master/pyatoa/visuals/vtk_modeler.py) but hit some road blocks and never finished it.