Masking a numpy array after selection with a non-numpy array

Question

I have the following code:

# unicorns is a numpy array with several fields
idx = (1, 2, 3, 5, 7)
unicorns=uni[idx]
# now i have only the first, second, third, ... unicorn
print unicorns

However if I want to subselect this unicorn array

unicorns['color'=='white']['Name']

which should give me the names of the unicorns that are white, numpy interprets only the color==white part as False, which goes to 0 and it returns the first entry of my array.

How can I fix this code, so that it does what I want it to, selecting the white unicorns?

I would prefer everything stays as numpy, so I can also select other properties oft the unicorns.

Edit

Here is an example for the arrays:

    unicorns=[(1, black, 0.0, 'Pinky', 1) (2, black, 0.0, 'Winky', 1)
 (3, white, 0.0, 'Lala', 1) (4, white, 0.0, 'Merlin', 1)
 (5, black, 0.0, 'Meriva', 1) (6, white, 0.0, 'Panda', 1)]
    idx = [  0 ,  3  , 6 ]

Could you post a minimal version of unicorns that can be used to reproduce the problem? — Fred Foo
– Fred Foo, Commented Sep 26, 2012 at 9:20

unutbu · Accepted Answer · 2012-09-26 09:38:00Z

I had to modify your array a bit to make it valid Python code. If I converted it properly, then I think what you are looking for is:

unicorns[unicorns['color'] == 'white']['name']

import numpy as np

unicorns=[(1, 'black', 0.0, 'Pinky', 1), (2, 'black', 0.0, 'Winky', 1),
          (3, 'white', 0.0, 'Lala', 1), (4, 'white', 0.0, 'Merlin', 1),
          (5, 'black', 0.0, 'Meriva', 1), (6, 'white', 0.0, 'Panda', 1),
          ]
unicorns = np.array(unicorns,
                    dtype = [('id', '<i4'),
                             ('color', 'S10'),
                             ('val1', '<f4'),
                             ('name', 'S10'),
                             ('val2', '<i4')])

print(unicorns['color'] == 'white')
# [False False  True  True False  True]

print(unicorns[unicorns['color'] == 'white']['name'])
# ['Lala' 'Merlin' 'Panda']

Pierre GM · Accepted Answer · 2012-09-26 09:41:36Z

1

What you probabply want to use is the numpy.where function. Use it like this:

    >>>unicorns=np.array([[1, "black", 0.0, 'Pinky', 1] ,
                       [2, "black", 0.0, 'Winky', 1],
                       [3, "white", 0.0, 'Lala', 1],
                       [4, "white", 0.0, 'Merlin', 1],
                       [5, "black", 0.0, 'Meriva', 1],
                       [6, "white", 0.0, 'Panda', 1]])
    >>> np.where(unicorns[:,1] == "black")
    (array([0, 1, 4]),)
    >>> unicorns[np.where(unicorns[:,1] == "black")]
    array([['1', 'black', '0.0', 'Pinky', '1'],
    ['2', 'black', '0.0', 'Winky', '1'],
    ['5', 'black', '0.0', 'Meriva', '1']], 
    dtype='|S8')

edited Sep 26, 2012 at 9:41

Pierre GM

20.5k3 gold badges58 silver badges67 bronze badges

answered Sep 26, 2012 at 9:40

Schuh

1,0955 silver badges10 bronze badges

1 Comment

Pierre GM Over a year ago

You're transforming the input as an array of strings... Stick with structured arrays.

Daniel · Accepted Answer · 2012-09-26 20:45:56Z

You can also take a look at pandas which is really awesome for such kinds of slicing and querying operations. Your problem could be solved with it like that:

In [12]: df = pd.DataFrame(unicorns)

In [13]: df.columns = ['id','color','speed','name','tails']

In [14]: df
Out[14]: 
  id  color speed    name tails
0  1  black   0.0   Pinky     1
1  2  black   0.0   Winky     1
2  3  white   0.0    Lala     1
3  4  white   0.0  Merlin     1
4  5  black   0.0  Meriva     1
5  6  white   0.0   Panda     1

In [16]: df[df.color == 'black']
Out[16]: 
  id  color speed    name tails
0  1  black   0.0   Pinky     1
1  2  black   0.0   Winky     1
4  5  black   0.0  Meriva     1

In [17]: df[df.color == 'black'].name
Out[17]: 
0     Pinky
1     Winky
4    Meriva
Name: name

Collectives™ on Stack Overflow

Masking a numpy array after selection with a non-numpy array

3 Answers 3

Comments

1 Comment

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

3 Answers 3

Comments

1 Comment

Comments

Your Answer

Sign up or log in

Post as a guest

Related