Standardize Abstract types

[Affie]

The Abstract prefix came up in #1146

Listing current abstract types for "do we use Abstract prefix?":

AbstractType	JustType
AbstractDFG
AbstractParams
AbstractBlobstore
	DFGNode
AbstractDFGVariable
AbstractDFGFactor
AbstractPackedFactorObservation
AbstractFactorObservation
AbstractPrior
AbstractRelative
	FactorSolverCache
	VariableStateType
	PackedBelief
AbstractPointParametricEst	will be obsolete

Note: DFGVariable and DFGFactor conflicts with names to be deprecated and is too similar to VariableDFG.

Julia uses the Abstract prefix regularly, and it's a recommendation in the SciML style guide: https://docs.sciml.ai/SciMLStyle/stable/#General-Naming-Principles

---

DFG v1.0 implementation :

Old	AbstractType	Alias
	AbstractDFG	--
AbstractParams	AbstractDFGParams	DFGParams
	AbstractBlobstore	Blobstore
DFGNode	AbstractGraphNode	GraphNode
AbstractDFGVariable	AbstractGraphVariable	GraphVariable
AbstractDFGFactor	AbstractGraphFactor	GraphFactor
AbstractPackedFactorObservation	AbstractPackedObservation	PackedObservation
AbstractFactorObservation	AbstractObservation	Observation
AbstractPrior	AbstractPriorObservation	PriorObservation
AbstractRelative	AbstractRelativeObservation	RelativeObservation
FactorSolverCache	AbstractFactorCache	FactorCache
VariableStateType	AbstractStateType	StateType
	AbstractPackedBelief	PackedBelief

[Affie]

We can maybe just use Abstract in definitions and then alias:

julia> abstract type AbstractFoo end
julia> const Foo = AbstractFoo
AbstractFoo
julia> Foo
AbstractFoo

[Affie]

My vote is to including the Abstract prefix (and then alias) for these reasons:

• The style guide recommendation
• Easy autocomplete to get to the abstract type if you can't remember it.
• Reading the code on dispatch seems easier for me as well.
• Abstract prefix might make bad design clearer, eg. bad performance of container of abstract types.
• Alias is convenient.

We should still follow the reasoning not to have the prefix to get the best possible noun, ie. the prefix should not be needed if possible. We can then still alias types if needed.

[Affie]

From triage call today, we agreed on #1153 (comment) with the aliases as general design. The exact names are still to be finalized.

[Affie]

@dehann and @mei-cheung, please give inputs on naming if you have any.