search

CRPropa / CRPropa3

CRPropa is a public astrophysical simulation framework for propagating extraterrestrial ultra-high energy particles. https://crpropa.github.io/CRPropa3/
https://crpropa.desy.de
GNU General Public License v3.0
69 stars 68 forks source link

Variant improvements #427

Closed rafaelab closed 11 months ago

rafaelab commented 1 year ago

This PR improves the handling of variants in CRPropa. It updates the previous implementation making the code more efficient. Its main goal is to facilitate the lives of users aiming to write their own plugins. This PR also adds some more tests in testCore.cpp to check if variants are being treated correctly.

Currently some types like Vector3d cannot be treated as a variant. Instead, the user would require three double and consequently three operations to handle it. New types added are: complex<double>, complex<float>, Vector3d, Vector3f, Vector3c (alias for Vector3<complex<double>>, long double. A new type for a vector of variants (vector<Variant>) was also added.

For some yet-unknown reason these new types have to be called with variant.toVector3d(), for instance, whereas operations assigning a variant directly to some types are permitted for the more conventional types like double.

lukasmerten commented 11 months ago

Hi @rafaelab I have tested the implementation and could not find any problems. Although I still do not fully get the advantages of the new implementation, I don't have any reason to hold the PR back.

If nobody has a different opinion, I will merge it in the next days.

rafaelab commented 11 months ago

Thanks, @lukasmerten

Hi @rafaelab I have tested the implementation and could not find any problems. Although I still do not fully get the advantages of the new implementation, I don't have any reason to hold the PR back.

There isn't much new. The redesign just allows for a vector<Variant> and a few other types that were not available. This is particularly useful for plugins, which require Candidate properties that don't exist. The motivation for this was that I needed to use Vector3 and vector as a Variant-type to add as a candidate property. This is much more efficient than having, for example, 3 individual double variants to represent a Vector3d. It doesn't break in any way the previous code structure (and now there are also tests for it).