Closed thomaspinder closed 1 year ago
Do we need this py.typed
file?
Suggestion: While we (or you) are refactoring, would it be sensible to structure the kernels and the computations across two submodules rather than in the main directory? I see the computations growing in size in the near future (inter-domain, Kronecker, Markovian dynamics, ...). It might be nice to have the computations split up into separate files for better readability (like I did in JaxLinOp
)? You could also consider have submodules within the kernels (for e.g. stationary). It might make things easier to maintain/ reduce the amount of scrolling? Interested to hear your thoughts on this.
Do we need this
py.typed
file?
Not right now. Removed.
Suggestion: While we (or you) are refactoring, would it be sensible to structure the kernels and the computations across two submodules rather than in the main directory? I see the computations growing in size in the near future (inter-domain, Kronecker, Markovian dynamics, ...). It might be nice to have the computations split up into separate files for better readability (like I did in
JaxLinOp
)? You could also consider have submodules within the kernels (for e.g. stationary). It might make things easier to maintain/ reduce the amount of scrolling? Interested to hear your thoughts on this.
Yeah nice idea.
Suggestion: While we (or you) are refactoring, would it be sensible to structure the kernels and the computations across two submodules rather than in the main directory? I see the computations growing in size in the near future (inter-domain, Kronecker, Markovian dynamics, ...). It might be nice to have the computations split up into separate files for better readability (like I did in
JaxLinOp
)? You could also consider have submodules within the kernels (for e.g. stationary). It might make things easier to maintain/ reduce the amount of scrolling? Interested to hear your thoughts on this.Yeah nice idea.
@Daniel-Dodd - this is now done.
Sorry @thomaspinder these comment are a little out of date with regards to their file references.
This PR restructures JaxKern so that kernels are broken out by their type. Further, computation objects are given their own modules.
Resolves #31