Use numpy where on a multidimensional array

Question

I have a multidimensional array similar to the following and i'm trying to delete the strings that end with * (stars) such that I can convert it into an array of floats.

    array1 = np.column_stack((a, b, c, d)) 
    array1 = np.array([
       ['*0.70*', '21.59', '4.37', '21.70'],
       ['2.15', '21.42', '5.63', '22.33'],
       ['*8.00*', '21.17', '5.11', '22.40'],
       ['2.36', '22.88', '*2.54*', '*20.95*'],
       ['2.07', '22.64', '6.68', '22.26']
       ])

Is there a way to us np.where to give the coordinate within the array of the valuse highlighted with stars, and not just index so I can delete the entire row?

So Ideal output would be somthing along the lines ofK

fil1 = np.where(np.char.endswith(array1, "*") == True)

print(fil1) 
(0,0), (0,2), (2, 3), (3, 3)

Serge Ballesta · Accepted Answer · 2020-05-27 13:21:08Z

1

np.where returns 1 array per dimension. If you want to know the indices of the rows containing stars, just do:

starred_rows = np.unique(np.where(np.char.endswith(array1, "*") == True)[0])

To get the index pairs, you can use zip:

np.array(list(zip(*np.where(np.char.endswith(array1, "*") == True))))

answered May 27, 2020 at 13:21

Serge Ballesta

150k13 gold badges137 silver badges267 bronze badges

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

Comments

Karthik Radhakrishnan · Accepted Answer · 2020-05-27 13:41:41Z

0

To get the Ideal output mentioned above you need to use zip

fil1 = list(zip(*np.where(np.char.endswith(array1, "*") == True)))
print(fil1)  
[(0, 0), (2, 0), (3, 2), (3, 3)] //result

If you want to get only the index of rows, you can take the unique of the first element of the your result.

fil1 = list(set(np.where(np.char.endswith(array1, "*") == True)[0]))
print(fil1)
[0, 2, 3]  // result

answered May 27, 2020 at 13:41

Karthik Radhakrishnan

1,0548 silver badges11 bronze badges

1 Comment

hpaulj Over a year ago

np.transpose(np.where(...)) or np.argwhere(...)

hpaulj · Accepted Answer · 2020-05-27 15:02:10Z

In [81]: array1 = np.array([ 
    ...:        ['*0.70*', '21.59', '4.37', '21.70'], 
    ...:        ['2.15', '21.42', '5.63', '22.33'], 
    ...:        ['*8.00*', '21.17', '5.11', '22.40'], 
    ...:        ['2.36', '22.88', '*2.54*', '*20.95*'], 
    ...:        ['2.07', '22.64', '6.68', '22.26'] 
    ...:        ])

The char test returns a boolean array:

In [84]: mask = np.char.endswith(array1,"*")                                             
In [85]: mask                                                                            
Out[85]: 
array([[ True, False, False, False],
       [False, False, False, False],
       [ True, False, False, False],
       [False, False,  True,  True],
       [False, False, False, False]])

np.nonzero (aka np.where) finds the coordinates of the True values, one array per dimension:

In [86]: np.nonzero(mask)                                                                
Out[86]: (array([0, 2, 3, 3]), array([0, 0, 2, 3]))

If you want to delete the rows, the first array can be used (duplicate 3 apparently doesn't bother delete):

In [88]: np.delete(array1, np.nonzero(mask)[0], 0)                                       
Out[88]: 
array([['2.15', '21.42', '5.63', '22.33'],
       ['2.07', '22.64', '6.68', '22.26']], dtype='<U7')

But we can also find rows with any True with:

In [89]: mask.any(axis=1)                                                                
Out[89]: array([ True, False,  True,  True, False])

and use that to select those rows (boolean array indexing)

In [91]: array1[mask.any(axis=1)]                                                        
Out[91]: 
array([['*0.70*', '21.59', '4.37', '21.70'],
       ['*8.00*', '21.17', '5.11', '22.40'],
       ['2.36', '22.88', '*2.54*', '*20.95*']], dtype='<U7')

or select their not:

In [92]: array1[~mask.any(axis=1)]                                                       
Out[92]: 
array([['2.15', '21.42', '5.63', '22.33'],
       ['2.07', '22.64', '6.68', '22.26']], dtype='<U7')

np.nonzero(Out[89]) is (array([0, 2, 3]),), the desired delete rows.

Other answers have used the Python list version of transpose; numpy's own transpose works as well:

In [93]: np.argwhere(mask)                                                               
Out[93]: 
array([[0, 0],
       [2, 0],
       [3, 2],
       [3, 3]])
In [94]: np.transpose(np.nonzero(mask))                                                  
Out[94]: 
array([[0, 0],
       [2, 0],
       [3, 2],
       [3, 3]])

For the purpose of deleting rows this transpose isn't any more useful than the where.

Collectives™ on Stack Overflow

Use numpy where on a multidimensional 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