PISN & type Ia models - On hold

[jpkneller]

New pisn model classes and Jupyter workbook

[JostMigenda]

I’ll try to review this in details tomorrow; here are some initial, high-level thoughts:

• For both models, tests are missing.
• The loader classes for both models look nearly identical; which makes sense given they’re the exact same format. Are there relevant differences or would it make sense to de-duplicate that?
• The TypeIa class doesn’t match our Lastname_YYYY convention for model classes. Can we rename that to Wright_2016? (Which matches the year both preprints appeared on the arXiv; even if the PRD paper on the GCD model is from 2017.)
• I notice the files for PISN and type Ia are in a combined repo, but the classes are in separate files. I don’t have a strong preference either way; happy to leave it as is, but wanted to flag it in case people see advantages to one or the other. (If joint, maybe snewpy-models-exotics/snewpy.models.exotics would be a better name, just in case there’s ever any other exotic models.)

[JostMigenda]

If both t and E are multi-element arrays (e.g. if we do model.get_flux(times, energies, d) as demonstrated in FlavorTransformation.ipynb), this will fail.

The fix could be something like:

        t = u.Quantity(t, ndmin=1).to(u.s).value
        E = u.Quantity(E, ndmin=1).to(u.MeV).value
        tE_grid = np.stack(np.meshgrid(t, E, indexing='ij'), axis=-1)

        initial_spectra = {}
        for flavor in ThreeFlavor:
            initial_spectra[flavor] = self.interpolation[flavor](tE_grid) / (u.MeV * u.s)

(Note that we also need to remove the units in between, otherwise np.stack will complain that we’re mixing units.)

I’ve played around with a toy example to figure out the correct order of indices (self.interpolation[flavor](tE_grid)[i, j] == self.interpolation[flavor]((t[i], E[j]))), but since this is so easy to mix up, I would appreciate if someone could double-check that.

[JostMigenda]

The same applies to the TypeIa class.

[jpkneller]

* For both models, tests are missing.

I'll fix that.

* The loader classes for both models look nearly identical; which makes sense given they’re the exact same format. Are there relevant differences or would it make sense to de-duplicate that?

I recycled the SNOwGLoBES model class as a parent for both. This model class lives on.

* The `TypeIa` class doesn’t match our `Lastname_YYYY` convention for model classes. Can we rename that to `Wright_2016`? (Which matches the year both preprints appeared on the arXiv; even if the PRD paper on the GCD model is from 2017.)

I am okay with this suggestion. Do you want to keep TypeIa for backwards compatibility?

* I notice the files for PISN and type Ia are in a [combined repo](https://github.com/SNEWS2/snewpy-models-pisn_typeIa/), but the classes are in separate files. I don’t have a strong preference either way; happy to leave it as is, but wanted to flag it in case people see advantages to one or the other. (If joint, maybe `snewpy-models-exotics`/`snewpy.models.exotics` would be a better name, just in case there’s ever any other exotic models.)

I made a single repo so there would be fewer to manage: I'd like to keep the models in separate files.

[JostMigenda]

• The TypeIa class doesn’t match our Lastname_YYYY convention for model classes. Can we rename that to Wright_2016? (Which matches the year both preprints appeared on the arXiv; even if the PRD paper on the GCD model is from 2017.)

I am okay with this suggestion. Do you want to keep TypeIa for backwards compatibility?

The TypeIa class only ever appeared on this PR branch and was never included in any snewpy release; so I don’t think there is any compatibility issue if we rename it here