IncrementalInference.jl
IncrementalInference.jl copied to clipboard
TreeMsgLikelihoods should swap NamedTuple for Tuple
Could have a similar compile time implication as was found in:
- #1565
- #1513
- JuliaRobotics/Caesar.jl#835
Are you referring to sender
in this?
https://github.com/JuliaRobotics/IncrementalInference.jl/blob/60134f968860d5b3a0de6eb429adbb08ef54f24d/src/entities/CliqueTypes.jl#L43-L56
It is good as a NamedTuple.
Looking at .sender
I think that should be okay, since the type is hard coded.
Not recalling exactly, but upon looking through this again I think it was actually the NamedTuple
in the MsgRelativeType const that might be causing the Bayes tree compile time delays.
Just quickly, I made a very small runtime improvement:
- #1579
But does notyet get rid of the potential NamedTuple
compile time issue on tree message passing.
I'm also worried about the non-concrete union type on .cliqueLikelihood