more general indexing

src/karray.jl

@@ -494,9 +494,6 @@ end
 # function _getindex(l::LinearIndexing, A::AbstractArray, I::Union{Real, AbstractArray, Colon}...)
 # in abstractarray.jl:487,multidimensional.jl:184.
 
-if VERSION < v"0.5.0"
-    @typealias6 AbstractUnitRange UnitRange
-end
 
 function getindex{T}(A::KnetArray{T}, I::AbstractUnitRange)
     if !(1 <= first(I) <= last(I) <= length(A)); throw(BoundsError(A,I)); end
@@ -552,6 +549,22 @@ function setindex!{T}(A::KnetArray{T}, v, I::Colon)
     unsafe_copy!(A,1,v,1,length(A))
 end
 
+## General Indexing Fallback to linear indexing
+
+function getindex(a::KnetArray, I...) 
+    crange = CartesianRange(to_indices(a, I))
+    linind = [sub2ind(size(a), t.I...) for t in crange]
+    b = getindex(a, vec(linind))
+    shape = size(crange) # TODO drop scalar dimension
+    reshape(b, shape)
+end
+
+function setindex!(a::KnetArray, v, I...) 
+    crange = CartesianRange(to_indices(a, I))
+    linind = [sub2ind(size(a), t.I...) for t in crange]
+    setindex!(a, v, vec(linind))
+end
+
 for F in (32,64); T=Symbol("Float$F"); @eval begin
 
 ## Indexing with KnetArray{Int32}: low level, only Int32 supported, no bounds checking

test/karray.jl

@@ -123,6 +123,26 @@ if gpu() >= 0
             @test size(reshape(a, :, 4)) == size(reshape(a, (:, 4))) == (2, 4)
             @test size(reshape(a, :, 1, 4)) == (2, 1,  4)
         end
+
+        @testset "general indexing" begin
+        # TODO
+        #
+        #julia> a=KnetArray(rand(3,3,2));
+        #
+        #julia> grad(a->sum(a[1:2,:,1]))(a)
+        #3×3×2 Knet.KnetArray{Float64,3}:
+        #[:, :, 1] =
+        # 1.0  1.0  1.0
+        # 1.0  1.0  1.0
+        # 0.0  0.0  0.0
+        #
+        #[:, :, 2] =
+        # 0.0  0.0  0.0
+        # 0.0  0.0  0.0
+        # 0.0  0.0  0.0
+        end
+
+
     end
 end