TensorFlow transform would throw on non-vector input

[TomFinley]

In the process of working on #1533, I found the following code in the tensorflow transform.

machinelearning/src/Microsoft.ML.TensorFlow/TensorflowTransform.cs

Lines 834 to 844 in 73762a8

	_isInputVector[i] = type.IsVector;
	var expectedType = TensorFlowUtils.Tf2MlNetType(_parent.TFInputTypes[i]);
	if (type.ItemType != expectedType)
	throw _host.ExceptSchemaMismatch(nameof(inputSchema), "input", _parent.Inputs[i], expectedType.ToString(), type.ToString());
	var originalShape = _parent.TFInputShapes[i];
	var shape = originalShape.ToIntArray();
	var colTypeDims = Enumerable.Range(0, type.AsVector.DimCount + 1).Select(d => d == 0 ? 1 : (long)type.AsVector.GetDim(d - 1)).ToArray();
	if (shape == null)
	_fullySpecifiedShapes[i] = new TFShape(colTypeDims);
	else if (type.AsVector.DimCount == 1)

Consider the above code. We have at line 834 this assignment to a boolean value depending on whether the input is of type vector, or not. This strongly suggests that the transform can accomodate non-vector types. However at lines 841 and 844, we have this type.AsVector.DimCount. Now, AsVector will be null in the case where the type is not a vector of course, so this would throw a null reference exception if we were to ever feed this transform a non-vector value.

So, there's something wrong here. Unfortunately the intent of what the author meant to write is somewhat hidden from me, so perhaps whoever wrote this code could check this out. Maybe even write a test to test this condition.

[najeeb-kazmi]

@abgoswam @yaeldekel Can you provide details on the original intent of this code?