Trying to understand this statement in Python

Question

I came across the following statement in Python:

a = m[:,1].a - m[:,0].a

If I remove the .a parts, I understand that we are trying to subtract the first column in the array from the second column.

What does adding .a do?

Thanks.

Answer 1

When writing my comment I realized that this notation is consistent with a recarray .

Define a dtype with a couple of fields

In [28]: dt = np.dtype([('a',int),('b',float)])

Initialize a recarray with this dtype :

In [29]: arr = np.ones((3,2), dtype=dt).view(np.recarray)
In [30]: arr['a'] = np.arange(6).reshape(3,2)
In [31]: arr
Out[31]: 
rec.array([[(0,  1.), (1,  1.)],
           [(2,  1.), (3,  1.)],
           [(4,  1.), (5,  1.)]], 
          dtype=[('a', '<i4'), ('b', '<f8')])
In [32]: arr.a
Out[32]: 
array([[0, 1],
       [2, 3],
       [4, 5]])

The first field can be accessed with arr['a'] or arr.a (structured array or recarray attribute). Now we can write an expression like yours:

In [34]: arr[:,0].a - arr[:,1].a
Out[34]: array([-1, -1, -1])

Structured arrays are common on SO, especially when loading data from a CSV file. The recarray variant may actually be older, but has fallen out of use. For example, the recfunctions library requires a special import. Apparently the attribute notation isn't that special or useful.

If this isn't right, you need to tell us more about the source of the quote.