reorganize elixirs

[ranocha]

This gives

examples/
├── p4est_2d_dgsem
│   ├── elixir_advection_amr_solution_independent.jl
│   │...
│   └── elixir_euler_nonperiodic_unstructured.jl
├── p4est_3d_dgsem
│   ├── elixir_advection_amr.jl
│   │...
│   └── elixir_euler_nonperiodic.jl
├── paper_self_gravitating_gas_dynamics
│   ├── elixir_euler_eoc.jl
│   │...
│   ├── elixir_hypdiff_eoc.jl
│   └── README.md
├── README.md
├── special_elixirs
│   └── elixir_euler_ad.jl
├── structured_1d_dgsem
│   ├── elixir_advection_b
│   │...
│   └── elixir_euler_source_terms.jl
├── structured_2d_dgsem
│   ├── elixir_advection_basic.jl
│   │...
│   └── elixir_mhd_ec.jl
├── structured_3d_dgsem
│   ├── elixir_advection_basic.jl
│   │...
│   └── elixir_mhd_ec.jl
├── tree_1d_dgsem
│   ├── elixir_advection_amr.jl
│   │...
│   └── elixir_mhd_torrilhon_shock_tube.jl
├── tree_2d_dgsem
│   ├── elixir_advection_amr_coarsen_twice.jl
│   │...
│   └── elixir_mhd_rotor.jl
├── tree_2d_fdsbp
│   └── elixir_advection_extended.jl
├── tree_3d_dgsem
│   ├── elixir_advection_amr.jl
│   │...
│   └── elixir_mhd_ec.jl
└── unstructured_2d_dgsem
    ├── elixir_advection_basic.jl
    │...
    └── elixir_mhd_ec.jl

12 directories, 203 files

TODO:

• Adapt benchmarks
• Change expected check to paper_self_gravitating_gas_dynamics (with underscores)
• After merging this: Adapt paths to elixirs in Trixi2Vtk and Trixi2Img (if necessary)

Closes #569

[sloede]

Overall this LGTM! Should we wait with the merge until after the meeting on Tuesday?

Some inconsistencies I noticed:

• Several eleixirs for convergence tests are named _convergence_test, while others are named _eoc. I think we should unify those while we're at it.
• There are many files named _ec or _es. I think here we need to make use of the second clause of the "no abbreviations unless absolutely necessary" rule and leave it as it is; I just wanted to bring it to your attention.

[sloede]

And thanks a lot for doing this @ranocha! I know this was originally my TODO... 🙈

[ranocha]

And thanks a lot for doing this @ranocha! I know this was originally my TODO...

You're welcome. I just wanted to get this out of our way before preparing our demo for JuliaCon etc.

[ranocha]

Several eleixirs for convergence tests are named _convergence_test, while others are named _eoc. I think we should unify those while we're at it.

I renamed them using the suffix _eoc in 53b5e48.

[efaulhaber]

I'd like to point out that elixir_euler_nonperiodic.jl and elixir_euler_source_terms.jl are essentially the same elixir (affects both tree and structured). I think it's a bit inconsistent that elixir_euler_source_terms.jl with non-periodic boundaries is called elixir_euler_nonperiodic.jl, while the same elixir on a rotated domain is called elixir_euler_source_terms_rotated.jl.

Also, while you're at it, would it make sense to merge all (most) restart elixirs to one? Maybe in the parent directory?
With elixir_file and restart_file as variables, we could just include different elixirs in the test files.

[ranocha]

Also, while you're at it, would it make sense to merge all (most) restart elixirs to one? Maybe in the parent directory?
With elixir_file and restart_file as variables, we could just include different elixirs in the test files.

@sloede?

[ranocha]

I'd like to point out that elixir_euler_nonperiodic.jl and elixir_euler_source_terms.jl are essentially the same elixir (affects both tree and structured). I think it's a bit inconsistent that elixir_euler_source_terms.jl with non-periodic boundaries is called elixir_euler_nonperiodic.jl, while the same elixir on a rotated domain is called elixir_euler_source_terms_rotated.jl.

@sloede? I didn't choose those names, so I don't have any strong opinion here... I renamed them to elixir_euler_source_terms_nonperiodic.

[sloede]

Also, while you're at it, would it make sense to merge all (most) restart elixirs to one? Maybe in the parent directory?
With elixir_file and restart_file as variables, we could just include different elixirs in the test files.

@sloede?

Well, restarting is such a "basic" operation that I'd like to keep a basic version of it around, i.e., one without magic. But we could use one at the top level that actually works (e.g., by calling default_example() for the elixir name), and then override it everywhere like @efaulhaber proposed. However, I suggest to do this in a different PR to make this one not even bigger (but it's fine by me to do it here).

[sloede]

I'd like to point out that elixir_euler_nonperiodic.jl and elixir_euler_source_terms.jl are essentially the same elixir (affects both tree and structured). I think it's a bit inconsistent that elixir_euler_source_terms.jl with non-periodic boundaries is called elixir_euler_nonperiodic.jl, while the same elixir on a rotated domain is called elixir_euler_source_terms_rotated.jl.

@sloede? I didn't choose those names, so I don't have any strong opinion here... I renamed them to elixir_euler_source_terms_nonperiodic.

Good choice. One can argue that source_terms is the setup here and nonperiodic is a configuration detail, so this seems consistent to me.

[sloede]

LGTM now. I'd prefer to wait until someone comments on this eoc vs convergence_test question, but let's say that nobody does so by tomorrow at noon, I think we should just merge it.

Also, some tests are still failing...

[ranocha]

LGTM now. I'd prefer to wait until someone comments on this eoc vs convergence_test question, but let's say that nobody does so by tomorrow at noon, I think we should just merge it.

I renamed _eoc to _convergence

[sloede]

Thanks a lot! LGTM!

[jlchan]

Thanks for this reorganization! This'll make navigation a lot easier for me personally as well.