Repeated names

README.md

@@ -106,7 +106,3 @@ end
 
 You can do this to your own functions in your own packages, to add `NamedDimsArray` support.
 If you implement it for any functions in a standard library, a PR would be very appreciated.
-
-### Caveats
-
-If multiple dimensions have the same names, indexing by name is considered undefined behaviour and should not be relied upon.

src/name_core.jl

@@ -180,8 +180,11 @@ order_named_inds(val::Val{L}; kw...) where {L} = order_named_inds(val, kw.data)
 
 @generated function order_named_inds(val::Val{L}, ni::NamedTuple{K}) where {L,K}
     tuple_issubset(K, L) || throw(DimensionMismatch("Expected subset of $L, got $K"))
+    seen = []
     exs = map(L) do n
         if Base.sym_in(n, K)
+            n in seen && throw(ArgumentError("Can't index with :$n as this is repeated in $L"))
+            push!(seen, n)
             qn = QuoteNode(n)
             :(getfield(ni, $qn))
         else

test/functions.jl

@@ -28,6 +28,13 @@ using Statistics
         end
     end
 
+    @testset "repeated names" begin
+        ndt = NamedDimsArray(reshape(1:12, 2,3,2), (:x, :y, :x))
+        @test sum(ndt, dims=:y) == sum(ndt, dims=2)
+        @test sum(ndt, dims=:x) == sum(ndt, dims=1) # chooses the first
+        @test dropdims(sum(ndt, dims=:x), dims=:x) == dropdims(sum(ndt, dims=1), dims=:x)
+    end
+
     @testset "sort!" begin
         a = [1 9; 7 3]
         nda = NamedDimsArray(a, (:x, :y))

test/wrapper_array.jl

@@ -137,6 +137,12 @@ end
         @test nda[findall(iseven, nda)] isa Vector
         @test dimnames(ndv[findall(iseven, ndv)]) == (:x,)
     end
+
+    @testset "repeated names" begin
+        ndt = NamedDimsArray(rand(2,3,2), (:x, :y, :x))
+        @test ndt[y=1] == ndt[:, 1, :]
+        @test_throws ArgumentError ndt[x=1]
+    end
 end