Loading AMREX format data in Python

[henry2004y]

Hi, I am wondering if pyamrex is able to load the AMREX format field and particle data that are generated from a C++ executable? Is YT the recommended way to load and process AMReX data in Python?

[atmyers]

For data analysis and visualization, I recommend yt. If you want to restart from a checkpoint in Python, the VisMF::Read and ParticleContainer::restart methods are also accessible in pyamrex.

[henry2004y]

Thanks for the tips and suggestions! We are using yt currently (e.g. flekspy demo), but we also find it very hard to customize the visualizations as we need since yt comes with its own data structures and plotting routines, which may not be directly compatible with Matplotlib settings (e.g. tweaking the plotting range, ticks, titles, fonts, etc.). What we hope to have is a lightweight, easy-to-use data loader for just for AMReX in Python.

[atmyers]

There is a script for loading the particle data at least directly here.

Another option is to use yt as just a data loader, then pass the raw data to matplotlib for plotting, as done here.

[henry2004y]

This is very useful! I managed to create a simple AMReX particle loader based on the code in the first link. This would load all particle data into memory all at once.

One extra thing I am wondering is how yt handles the lazy particle data loading: _read_particles. It seems that there are no explicit codes for controlling which particle data files to read for a given region?

[atmyers]

In yt you load data through a "data object", which could represent the entire domain or some subset of it (slice, sphere, disk, etc.). When you query a data object for a field, internally it loops over all the grids in the hierarchy, intersects them with the data object, and loads only the data associated with the current grid if it's needed. This also happens when you use e.g. SlicePlot.

You can also loop over the grids manually like this:

for g in ds.index.grids:
    print(g['density'].max())