numpy IndexError: too many indices for array when indexing matrix with another

Question

I have a matrix a which I create like this:

>>> a = np.matrix("1 2 3; 4 5 6; 7 8 9; 10 11 12")

I have a matrix labels which I create like this:

>>> labels = np.matrix("1;0;1;1")

This is what the two matricies look like:

>>> a
matrix([[ 1,  2,  3],
        [ 4,  5,  6],
        [ 7,  8,  9],
        [10, 11, 12]])
>>> labels
matrix([[1],
        [0],
        [1],
        [1]])

As you can see, when I select all columns, there is no problem

>>> a[labels == 1, :]
matrix([[ 1,  7, 10]])

But when I try to specify a column I get an error

>>> a[labels == 1, 1]
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/usr/local/lib/python2.7/site-packages/numpy/matrixlib/defmatrix.py", line 305, in     __getitem__
    out = N.ndarray.__getitem__(self, index)
IndexError: too many indices for array
>>>   

Does anybody know why this is? I am aware there are similar questions to this already but none of them explain my problem well enough, neither are the answers helpful to me.

Answer 1

Since labels is a matrix when you do labels==1 you obtain a boolean matrix of the same shape. Then doing a[labels==1, :] will return you only the first column with the lines corresponding to the match. Note that your intention to get:

matrix([[ 1,  2,  3],
        [ 7,  8,  9],
        [10, 11, 12]])

was not achieved (you got only the first column), even though it worked for NumPy < 1.8 (as pointed out by @seberg).

In order to get what you want you can use a flattened view of labels:

a[labels.view(np.ndarray).ravel()==1, :]