[Numpy-discussion] Creating a 1-d structured array

[Derek Homeier]

Hi,

just a comment since I first thought the solution below might not be  
what Bruce
was looking for, but having realised it's probably what he's been  
asking for...

On 13 May 2011, at 17:20, josef.pktd at gmail.com wrote:

> On Fri, May 13, 2011 at 10:58 AM, Bruce Southey <bsouthey at gmail.com>  
> wrote:
>> Hi,
>> How do you create a 'single' structured array using np.array()?
>> Basically I am attempting to do something like this that does not  
>> work:
>> a=np.array([1,2, 3,4, 5,6], dtype=np.dtype([('foo', int)]))
>>
>> I realize that this is essentially redundant as if A is an 1-d array
>> then a structured array with a named field 'foo' is the same thing  
>> - A
>> would be A['foo'], just shorter.
...
>
> Using a view works, (and direct assignment of dtype)
>
>>>> a=np.array([1,2, 3,4, 5,6]).view(([('foo', int)]))
>>>> a
> array([(1,), (2,), (3,), (4,), (5,), (6,)],
>      dtype=[('foo', '<i4')])
>>>> b = a.copy()
>>>> b
> array([(1,), (2,), (3,), (4,), (5,), (6,)],
>      dtype=[('foo', '<i4')])
>
>
>>>> a1 = np.array([1,2, 3,4, 5,6]).astype([('foo', int)])
>
> Traceback (most recent call last):
>  File "<pyshell#36>", line 1, in <module>
>    a1 = np.array([1,2, 3,4, 5,6]).astype([('foo', int)])
> TypeError: expected a readable buffer object
>
>>>> a1 = np.array([1,2, 3,4, 5,6])
>>>> a1.dtype
> dtype('int32')
>>>> a1.dtype = np.dtype([('foo', int)])
>>>> a1
> array([(1,), (2,), (3,), (4,), (5,), (6,)],
>      dtype=[('foo', '<i4')])

This is a 1-d structured array, yet a is in fact not the same as  
a['foo']:

 >>> a['foo']
array([ 1, 2, 3, 4, 5, 6])
 >>> a['foo'].shape
(6,)
 >>> a.shape
(6,)
 >>> a['foo']+np.arange(6)
array([ 1,  3,  5,  7,  9, 11])
 >>> a+np.arange(6)
Traceback (most recent call last):
   File "<stdin>", line 1, in <module>
TypeError: unsupported operand type(s) for +: 'numpy.ndarray' and  
'numpy.ndarray'
 >>> a[0]+1
Traceback (most recent call last):
   File "<stdin>", line 1, in <module>
TypeError: unsupported operand type(s) for +: 'numpy.void' and 'int'

Thus I am wondering why broadcasting should not be possible in this  
case,
and if it really isn't, the first error message certainly is not very  
helpful... (maybe
inevitably though, since type() of a structured array is the same as  
for a "plain"
array (unlike for a record array).

Cheers,
						Derek