append_to_backend accesses private methods from MOI.Utilities

[odow]

See

Plasmo.jl/src/moi_backend_graph.jl

Lines 126 to 152 in fb7bda5

	function append_to_backend!(dest::MOI.ModelLike, src::MOI.ModelLike)
	vis_src = MOI.get(src, MOI.ListOfVariableIndices()) #returns vector of MOI.VariableIndex
	index_map = MOIU.IndexMap()
	# per the comment in MOI:
	# "The `NLPBlock` assumes that the order of variables does not change (#849)
	# Therefore, all VariableIndex and VectorOfVariable constraints are added
	# seprately, and no variables constrained-on-creation are added.""
	# Consequently, Plasmo avoids using the constrained-on-creation approach because
	# of the way it constructs the NLPBlock for the optimizer.
	# has_nlp = MOI.NLPBlock() in MOI.get(src, MOI.ListOfModelAttributesSet())
	# constraints_not_added = if has_nlp
	constraints_not_added = Any[
	MOI.get(src, MOI.ListOfConstraintIndices{F,S}()) for
	(F, S) in MOI.get(src, MOI.ListOfConstraintTypesPresent()) if
	MOIU._is_variable_function(F)
	]
	# else
	# Any[
	# MOIU._try_constrain_variables_on_creation(dest, src, index_map, S)
	# for S in MOIU.sorted_variable_sets_by_cost(dest, src)
	# ]
	# end
	# Copy free variables into graph optimizer
	MOI.Utilities._copy_free_variables(dest, index_map, vis_src)

x-ref jump-dev/MathOptInterface.jl#2520

[jalving]

Thanks for the heads up. I will be rolling out a new version soon that uses no private MOI methods.

[odow]

I'm reverting the change to MOI that removed them, so you're good for a while.

[jalving]

Plasmo.jl should no longer be using private MOI methods with v0.6, but please re-open if you spot an instance i missed.

Addressed with #105