User can choose which variables to store in pltfiles

[psharda]

Describe the proposal
We currently don't have the capability to let the user choose which variables they want stored in pltfiles. We should add that in.

Describe alternatives you've considered
Alternatively, we have to load in large pltfiles while analyzing data which will become problematic for users wanting to do analysis on local machines.

Additional context
This will be crucial for simulations including chemistry with many species, because we don't want to store all the species otherwise the pltfiles will become unnecessarily large. Even for primordial chemistry (with 14 species), the size of a single pltfile with 20 AMR levels is 16 GB. This could be reduced by more than half if we only stored the species we really care about.

[BenWibking]

I think the solution is to not do the analysis on your local machine. 16 GB is small compared to runs on Frontier (>10 TB+ snapshots). You just can't analyze sims on laptops anymore.

[BenWibking]

Or output slices while it's running. You can choose the variables you want for that: https://quokka-astro.github.io/quokka/insitu_analysis/

[psharda]

IMO you shouldn't focus on the size of the pltfiles. The key thing here (that can be rectified easily) is to avoid outputting variables deemed unnecessary by the user. Several (if not all) major astro codes have this functionality. It is not satisfactory to ask a user not to analyze their simulations on a machine of their choice if this is avoidable. Personally, I like creating and testing scripts using one snapshot stored on my laptop, before porting them to clusters. My simulations are not in the same category as the Frontier ones (science or code wise), and quokka is/will be used by people from very different astro backgrounds. So I would advocate for the user to have the freedom to decide this themselves.

[psharda]

Not saying someone has to do it now, but it would be a useful functionality to have.

[BenWibking]

Ok, that's a good point. We could copy the idea of having a "small plotfile" in addition to the regular full-size plotfiles from Castro; see https://amrex-astro.github.io/Castro/docs/io.html#plotfile-outputting.