change Singleton / Pairwise to AbstractPrior / AbstractRelativeFactor

[dehann]

Or something similar, we can discuss here

Also see

• Manifest VariableTypes (Singleton model) and new FMD IncrementalInference.jl#783

[dehann]

JuliaRobotics/IncrementalInference.jl#783 (comment)

[Affie]

The applicable code:

##==============================================================================
## Abstract Types
##==============================================================================

abstract type InferenceType end
abstract type PackedInferenceType end
abstract type FunctorInferenceType <: Function end
abstract type FactorOperationalMemory <: Function end
abstract type FunctorSingleton <: FunctorInferenceType end
abstract type FunctorPairwise <: FunctorInferenceType end
abstract type FunctorPairwiseMinimize <: FunctorInferenceType end

If a FunctorSingleton is always a prior, I would suggest 'Prior' then. It might be less of a learning curve to stick with familiar names? It took me a while to figure out what it was.

Also is the naming of Functor needed in FunctorInferenceType and FunctorSingleton
maybe it can be reduced to AbstractPriorFactor or even AbstractPrior and AbstractRelativeFactor ?

[dehann]

Yes, these two sound best to me -- let's do that:

AbstractPrior and AbstractRelativeFactor

Additional Info

Functor was previously introduced as part of an upgrade -- although I think the original Singleton and Pairwise stuff have long been deprecated. Dropping Functor should be fine and actually preferable.

[GearsAD]

Personally I like simpler/shorter, AbstractPrior and AbstractRelativeFactor sound like a plan

[dehann]

Lets do a deprecation cycle on the names because RoME and Caesar are both using Singleton and Pairwise abstracts at the moment. So basically just create the new ones next to it and place the old ones in the Deprecated.jl file.