Shaping list of tuples into MxN dimensional ndarray in numpy

Question

I have the following list of tuples,

a=[('1A34', 'RBP', 0.0, 1.0, 0.0, 0.0, 0.0, 0.0),
   ('1A9N', 'RBP', 0.0456267, 0.0539268, 0.331932, 0.0464031, 4.41336e-06, 0.522107),
   ('1AQ3', 'RBP', 0.0444479, 0.201112, 0.268581, 0.0049757, 1.28505e-12, 0.480883),
   ('1AQ4', 'RBP', 0.0177232, 0.363746, 0.308995, 0.00169861, 0.0, 0.307837)]

What I want to do is to create a numpy nd.array with shape (4,8) That looks like this:

array([['1A34', 'RBP', 0.0, 1.0, 0.0, 0.0, 0.0, 0.0],
       ['1A9N', 'RBP', 0.0456267, 0.0539268, 0.331932, 0.0464031, 4.41336e-06, 0.522107],
       ['1AQ3', 'RBP', 0.0444479, 0.201112, 0.268581, 0.0049757, 1.28505e-12, 0.480883],
       ['1AQ4', 'RBP', 0.0177232, 0.363746, 0.308995, 0.00169861, 0.0, 0.307837]])

I tried the following code:

import numpy as np
x = np.array(a, dtype=('a10,a10,f4,f4,f4,f4,f4,f4'))

But it gives this shape instead:

In [37]: x.shape
Out[37]: (4,)

What's the right way to do it?

ebarr · Accepted Answer · 2014-06-30 09:37:04Z

1

What you have already done is by far the most logical way to do it. To achieve what you are asking for you need to create an object array:

z = np.array(a,dtype=np.object)
print z.shape
# (4, 8)

What is looks like you are asking for is an array with variable data type by column. This is exactly what you achieve with np.array(a, dtype=('a10,a10,f4,f4,f4,f4,f4,f4')). Internally you can think of this array like an array of structs in C, i.e. a 1-d array of dtype=('a10,a10,f4,f4,f4,f4,f4,f4') instances.

By using an object array you can request that numpy handles everything as a simple python object.

edited Jun 30, 2014 at 9:37

answered Jun 30, 2014 at 9:32

ebarr

7,8521 gold badge32 silver badges43 bronze badges

Sign up to request clarification or add additional context in comments.

4 Comments

neversaint Over a year ago

Thanks a lot. Is there any harm using object instead of ('a10,a10,f4,f4,f4,f4,f4,f4') ?

ebarr Over a year ago

I would assume that using object would be inherently slower due to the its variable nature. The "harm" here would be somewhat unpredictable behaviour, as it will not always be obvious what bit width numbers have. e.g. half your ints could be 64-bit and the other half 32-bit.

neversaint Over a year ago

Is there safer alternative than 'object' which produces the same results?

ebarr Over a year ago

The safer alternative is your original example, which creates a recarray. What is the problem with using that? Is there really a requirement that you have a (4,8) shape? The recarray can be indexed just like a normal 2-d array.

Collectives™ on Stack Overflow

Shaping list of tuples into MxN dimensional ndarray in numpy

1 Answer 1

4 Comments

Your Answer

Linked

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

4 Comments

Your Answer

Sign up or log in

Post as a guest

Linked

Related