You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Right now, the methods used for numerical computations are somehow hardcoded and their parameters are all contained in the OrbitalParameters structure.
It would be great to free the following methods (probably using multi-dispatch ?):
anomaly (hardcoded to Hénon)
integration method (hardcoded to RK4 for actions and frequencies)
differentiation method (hardcoded to naive finite differences)
inversion method (hardcoded to Newton-Raphson)
One way to currently bypass this for a specific potential is to define all the mappings using multi-dispatch on the model type. This is already used for the analytic computations for Isochrone. However, this is not very convenient.
We should keep in mind the following requirements:
Signatures of the functions need to be the same (while, e.g., in some analytical cases, there is no real need for an 'inversion' method)
The text was updated successfully, but these errors were encountered:
MathieuRoule
changed the title
Disentangling mapping methods
Disentangling mapping from their methods
Mar 5, 2024
MathieuRoule
changed the title
Disentangling mapping from their methods
Disentangling mappings from their methods (anomaly, differentiation, integration, inversion)
Mar 5, 2024
Right now, the methods used for numerical computations are somehow hardcoded and their parameters are all contained in the OrbitalParameters structure.
It would be great to free the following methods (probably using multi-dispatch ?):
One way to currently bypass this for a specific potential is to define all the mappings using multi-dispatch on the model type. This is already used for the analytic computations for Isochrone. However, this is not very convenient.
We should keep in mind the following requirements:
The text was updated successfully, but these errors were encountered: