fix construction with partial graphelement information

src/component_functions.jl

@@ -760,8 +760,7 @@ set_metadata!(c::ComponentFunction, key::Symbol, val) = setindex!(metadata(c), v
 
 Checks if the edge or vetex function function has the `graphelement` metadata.
 """
-has_graphelement(c::EdgeFunction) = has_metadata(c, :graphelement)
-has_graphelement(c::VertexFunction) = has_metadata(c, :graphelement)
+has_graphelement(c::ComponentFunction) = has_metadata(c, :graphelement)
 """
     get_graphelement(c::EdgeFunction)::@NamedTuple{src::T, dst::T}
     get_graphelement(c::VertexFunction)::Int

src/construction.jl

@@ -160,7 +160,7 @@ end
 function _unique_name_dict(cfs::AbstractVector{<:ComponentFunction})
     # find all names to resolve
     names = getproperty.(cfs, :name)
-    dict = Dict(names .=> get_graphelement.(cfs))
+    dict = Dict(cf.name => get_graphelement(cf) for cf in cfs if has_graphelement(cf))
     # delete all names which occure multiple times
     for i in eachindex(names)
         if names[i] ∈ @views names[i+1:end]
@@ -344,6 +344,7 @@ end
     Network(nw::Network; kwargs...)
 
 Rebuild the Network with same graph and vertex/edge functions but possibly different kwargs.
+# FIXME : needs to take all Network kw arguments into acount!
 """
 function Network(nw::Network; kwargs...)
     Network(nw.im.g, nw.im.vertexf, nw.im.edgef; kwargs...)

src/initialization.jl

@@ -148,19 +148,25 @@ in the metadata field `init`.
 The `kwargs` are passed to the nonlinear solver.
 """
 function initialize_component!(cf; verbose=true, kwargs...)
-    prob = initialization_problem(cf)
-    sol = SciMLBase.solve(prob; kwargs...)
-
-    if sol.prob isa NonlinearLeastSquaresProblem && sol.retcode == SciMLBase.ReturnCode.Stalled
-        # https://github.com/SciML/NonlinearSolve.jl/issues/459
-        res = LinearAlgebra.norm(sol.resid)
-        @warn "Initialization for componend stalled with residual $(res)"
-    elseif !SciMLBase.successful_retcode(sol.retcode)
-        throw(ArgumentError("Initialization failed. Solver returned $(sol.retcode)"))
+    prob = initialization_problem(cf; verbose)
+
+    if !isempty(prob.u0)
+        sol = SciMLBase.solve(prob; kwargs...)
+
+        if sol.prob isa NonlinearLeastSquaresProblem && sol.retcode == SciMLBase.ReturnCode.Stalled
+            # https://github.com/SciML/NonlinearSolve.jl/issues/459
+            res = LinearAlgebra.norm(sol.resid)
+            @warn "Initialization for componend stalled with residual $(res)"
+        elseif !SciMLBase.successful_retcode(sol.retcode)
+            throw(ArgumentError("Initialization failed. Solver returned $(sol.retcode)"))
+        end
+        set_init!.(Ref(cf), SII.variable_symbols(sol), sol.u)
+        resid = sol.resid
+    else
+        resid = init_residual(cf; recalc=true)
     end
-    set_init!.(Ref(cf), SII.variable_symbols(sol), sol.u)
 
-    set_metadata!(cf, :init_residual, sol.resid)
+    set_metadata!(cf, :init_residual, resid)
 
     verbose && @info "Initialization successful with residual $(LinearAlgebra.norm(sol.resid))"
     cf

test/construction_test.jl

@@ -72,8 +72,17 @@ end
 
     set_graphelement!(e3, (;src=2,dst=1))
     Network([v1,v2,v3], [e1,e2,e3]) # throws waring about 1->2 and 2->1 beeing present
-end
 
+    v1 = ODEVertex(x->x^1, 2, 0; metadata=Dict(:graphelement=>1), name=:v1)
+    v2 = ODEVertex(x->x^2, 2, 0; name=:v2)
+    v3 = ODEVertex(x->x^3, 2, 0; name=:v3)
+    @test NetworkDynamics._unique_name_dict([v1,v2,v3]) == Dict(:v1=>1)
+
+    v1 = ODEVertex(x->x^1, 2, 0; metadata=Dict(:graphelement=>1), name=:v1)
+    v2 = ODEVertex(x->x^2, 2, 0; name=:v1)
+    v3 = ODEVertex(x->x^3, 2, 0; name=:v3)
+    @test NetworkDynamics._unique_name_dict([v1,v2,v3]) == Dict()
+end
 @testset "Vertex batch" begin
     using NetworkDynamics: BatchStride, VertexBatch, parameter_range
     vb = VertexBatch{ODEVertex, typeof(sum), Vector{Int}}([1, 2, 3, 4], # vertices