Fix iterators for generic arrays

src/operators.jl

@@ -91,11 +91,11 @@ function Base.show(io::IO, i::IteratorValues)
 end
 
 function _tuple(axe)
-    if axe[1] isa Tuple
+    # axe can also be `keys(dict)`
+    if !isempty(axe) && first(axe) isa Tuple
         return axe
-    else
-        return tuple.(axe)
     end
+    return tuple.(axe)
 end
 
 function iterators(axes)

test/generic_index_set.jl

@@ -0,0 +1,62 @@
+# Copyright (c) 2024: Benoît Legat and contributors
+#
+# Use of this source code is governed by an MIT-style license that can be found
+# in the LICENSE.md file or at https://opensource.org/licenses/MIT.
+
+# Integration test for using a non-`Vector` index set (e.g. `keys(dict)`) with a GenOpt
+# `container` constraint.
+#
+# Why it is useful: models are frequently indexed by the keys of a `Dict` (component ids,
+# names, ...) rather than by `1:n`, exactly like PowerModels. `keys(dict)` is iterable but not
+# integer-indexable, so the iterator machinery must not assume `axe[1]`. This builds a
+# vectorized constraint indexed over `keys(dict)` and checks it solves.
+
+module TestGenericIndexSet
+
+using Test
+import JuMP
+import GenOpt
+import HiGHS
+import MathOptInterface as MOI
+
+function runtests()
+    for name in names(@__MODULE__; all = true)
+        if startswith("$(name)", "test_")
+            @testset "$(name)" begin
+                getfield(@__MODULE__, name)()
+            end
+        end
+    end
+    return
+end
+
+function _model()
+    inner = HiGHS.Optimizer()
+    MOI.set(inner, MOI.RawOptimizerAttribute("output_flag"), false)
+    optimizer = MOI.Bridges.full_bridge_optimizer(inner, Float64)
+    MOI.Bridges.add_bridge(optimizer, GenOpt.FunctionGeneratorBridge{Float64})
+    return JuMP.direct_model(optimizer)
+end
+
+function test_constraint_indexed_over_dict_keys()
+    demand = Dict(1 => 3.0, 2 => 5.0, 3 => 4.0)
+    model = _model()
+    JuMP.@variable(model, x[1:3, 1:1])
+    JuMP.@objective(model, Min, sum(x))
+    # Index the family of constraints by `keys(demand)`, not `1:n`.
+    JuMP.@constraint(
+        model,
+        [i in keys(demand)],
+        x[i, 1] >= demand[i],
+        container = GenOpt.ParametrizedArray,
+    )
+    JuMP.optimize!(model)
+    @test JuMP.termination_status(model) == MOI.OPTIMAL
+    @test JuMP.value(x[1, 1]) ≈ 3.0
+    @test JuMP.value(x[2, 1]) ≈ 5.0
+    @test JuMP.value(x[3, 1]) ≈ 4.0
+end
+
+end
+
+TestGenericIndexSet.runtests()