Remove Duplicate OpacityState object in numba_interface.py

[Rodot-]

Describe the bug
There are two copies of the OpacityState jitclass and its initializer. One is in opacity_state.py and the other in numba_interface.py. The version in numba_interface.py should be removed to complete the refactoring.

[karthik11135]

are we supposed to delete numba_interface file? because the implementations are same as opacitystatenumba in opacity_state file.

thanks for your time.

[Sonu0305]

Hi @andrewfullard ,
I notice one difference in the implementation of the opacity_state_initialize method between the two files,

In opacity_state.py (L385-L392):

    tau_sobolev_df = calculate_sobolev_line_opacity(
        plasma.atomic_data.lines,
        plasma.level_number_density,
        plasma.time_explosion,
        plasma.stimulated_emission_factor,
    )

    tau_sobolev = np.ascontiguousarray(tau_sobolev_df, dtype=np.float64)

Here, tau_sobolev is being calculated dynamically using the calculate_sobolev_line_opacity method.

In numba_interface.py (L167-L169):

    tau_sobolev = np.ascontiguousarray(
        plasma.tau_sobolevs.values.copy(), dtype=np.float64
    )

Here, tau_sobolev is copied from precomputed values stored in the plasma object.

Concern:

When we remove the opacity_state_initialize method from numba_interface.py, references that currently use this method will start using the implementation in opacity_state.py. This means that tau_sobolev will be calculated dynamically instead of using the precomputed values from plasma.tau_sobolevs.

Is this the intended behavior? Or should we preserve the ability to use precomputed tau_sobolevs in certain contexts?
Looking forward for your review, Thank You.

[andrewfullard]

Thanks for the extensive research. In this case, the opacity_state.py implementation is what we want to use. The critical check is to make sure that any uses of the numba_interface version of the class are replaced and tested against the relevant regression data.

[Sonu0305]

Thanks for the extensive research. In this case, the opacity_state.py implementation is what we want to use. The critical check is to make sure that any uses of the numba_interface version of the class are replaced and tested against the relevant regression data.

Thank You for the clarification, should I open a PR for the same now?

[andrewfullard]

No thank you, you have already produced a large number of PRs that we need to review.

[Sonu0305]

No thank you, you have already produced a large number of PRs that we need to review.

Thank you for your update. I've added this task to my To-Do list, and once we've completed the reviews for the current open PRs, I'll prioritize and open a PR for this one as well.