Background Cosmology API

[nstarman]

• BackgroundCosmologyAPI
  • Attributes:
    • scale_factor0 -> Array
    • Otot0 -> Array
    • critical_density0 -> Array
  • Methods:
    • scale_factor(z: Array, /) -> Array
    • Otot(z: Array, /) -> Array
    • critical_density(z: Array, /) -> Array
    • age(z: Array, /) -> Array
    • lookback_time(z: Array, /) -> Array
    • comoving_distance(z: Array, /) -> Array
    • comoving_transverse_distance(z: Array, /) -> Array
    • comoving_volume(z: Array, /) -> Array
    • differential_comoving_volume(z: Array, /) -> Array
    • angular_diameter_distance(z: Array, /) -> Array
    • luminosity_distance(z: Array, /) -> Array

Are there any attributes/methods we want to change? Or add?

[nstarman]

@ntessore I think we should add a hierarchy to the background cosmology, where the current BackgroundCosmologyAPI becomes HomogeneousIsotropicBackground, because a anisotropic background would require a full 3d coordinate, not just the redshift.
The hierarchy becomes:

class Background(Protocol):
    ...


class HomogeneousBackground(Background):
    ...


class IsotropicBackground(Background):
    ...


class HomogeneousIsotropicBackground(HomogeneousBackground, IsotropicBackground):
    ...

Now an anisotropic background cosmology can be defined, e.g.

class AnisotropicBackground(Background):
    Otot(z: Array, theta, phi/) -> Array: ...
    ...

[nstarman]

On a related note, we have a BackgroundCosmologyWrapperAPI class. I can't remember why this class exists. The parent and child class CosmologyWrapperAPI and StandardCosmologyWrapperAPI both make sense to me. Did we want to have compat shadow each of the main classes? OTOH with CosmologyWrapperAPI it's so easy to build a protocol to check. I'm not sure it's worth having any that aren't directly used, like StandardCosmologyWrapperAPI.

class StandardCosmologyWrapperAPI(
    HomogeneousIsotropicBackgroundWrapperAPI[ArrayT],
    StandardCosmologyAPI[ArrayT],
    Protocol,
):
   ... # nothing else is required.

[nstarman]

Alright. I remembered why we have BackgroundCosmologyWrapperAPI: it's for the astropy and camb wrappers, like class CAMBBackgroundCosmology(CAMBCosmology, BackgroundCosmologyWrapperAPI[NDFloating]):. But I still don't think it's necessary as BackgroundCosmologyWrapperAPI can be trivially replaced with HomogeneousIsotropicBackground.

[ntessore]

It don't think it makes sense to put these in a hierarchy, since you break the LSP.

[nstarman]

Do you mean how AnisotropicBackground has a different signatures than IsotropicBackground — that LSP? Or about the Wrapper hierarchy?

[ntessore]

Sorry, I replied here before I saw the PR. I thought you somehow wanted to have a full homogeneous interface, with the hom + iso subclasses effectively implementing a different one!