How to control dimensions of empty arrays in Numpy

Question

I was trying to concatenate 1-D two arrays in Python, using numpy. One of the arrays might potentially be empty (a2 in this case). a1 and a2 are the results from some computation over which I have no control. When a1 and a2 are non-empty they both have shapes of the form (n,2), so concatenation is not a problem. However it could turn out that one of them is empty, in which case its size becomes (0,). Hence the concatenation throws up an error.

s1=array(a1).shape
s2=array(a2).shape
print(s1) #(5,2) 
print(s2) #(0,)
s3=hstack((a1, a2))
s3=concatenate((a1, a2), 0)

Error: ValueError: all the input arrays must have same number of dimensions

I see other stackoverflow questions where it is said that it is possible to concatenate an empty array. How do I ensure that the empty array's size is (0,2)? Can someone help me out?

Answer 1

The error message tells you what you need to know. It's not enough that the array is empty - they have to have the same number of dimensions. You are looking only at the first element of shape - but shape can have more than one element:

numpy.array([[]]).shape   # (1L, 0L)
numpy.array([[]]).transpose.shape  # (0L, 1L)
numpy.array([]).shape     # (0L, )

So you see, empty arrays can have different numbers of dimensions. This may be your problem.

EDIT solution to create an empty array of the right size is to reshape it:

a2.shape()      # (0L,)
a2 = a2.reshape((0,2))
a2.shape()      # (0L, 2L)

This should solve your problem.