current community

your communities

more stack exchange communities

Stack Exchange
tour help
stack overflow careers
Take the 2-minute tour ×
Stack Overflow is a question and answer site for professional and enthusiast programmers. It's 100% free, no registration required.
up vote 4 down vote favorite
1

I am reading in a file using numpy.genfromtxt which brings in columns of both strings and numeric values. One thing I need to do is detect the length of the input. This is all fine provided there are more than one value read into each array.

But...if there is only one element in the resulting array, the logic fails. I can recreate an example here:

import numpy as np
a = np.array(2.3)

len(a) returns an error saying:

TypeError: len() of unsized object

however, If a has 2 or more elements, len() behaves as one would expect.

import numpy as np
a = np.array([2.3,3.6])

len(a) returns 2

My concern here is, if I use some strange exception handling, I can't distinguish between a being empty and a having length = 1.

EDIT: @noskio suggested setting a = np.array([2.3]). The problem is, the actual genesis of a is by using numpy.genfromtxt. The code looks like this:

import numpy as np
indata = np.genfromtxt(some_filename, names=True,dtype=None)
a = indata['one_col_headername']

As a result, if indata is only one row in the file, a is a 0-d array.

share|improve this question
    
array([2]) is an array with one element and 1 dimension. array(2) is an array with zero rank or zero dimensions. –  endolith Dec 16 '12 at 5:12

4 Answers 4

up vote 18 down vote accepted

If you need a one-liner (assuming the answer you are expecting is 1):

In [1]: import numpy as np

In [2]: a = np.array(2.3)

In [3]: len(np.atleast_1d(a))
Out[3]: 1

This page explains why it was decided to implement 0-dimensional arrays in numpy.

share|improve this answer
    
Thanks @pberkes! That helps tons and solves the issue –  mishaF Dec 30 '10 at 20:15
    
Link is broken :( –  wim May 5 at 11:25
    
@wim it looks like the link has been removed, and I cannot find a replacement for it. –  pberkes May 5 at 12:03
up vote 6 down vote 
import numpy as np

tests=[np.array(2.3),np.array([]),np.array([2.3]),np.array([2.3,3.6])]

print('{a:30}{s:<10}{l:<10}{sl:<10}'.format(a='repr',s='shape',sl='len(shape)',l='length'))
for a in tests:
    s=a.shape
    l=len(a) if a.shape else 0
    sl=len(s)
    print('{a!r:30}{s:<10}{l:<10}{sl:<10}'.format(a=a,l=l,s=s,sl=sl))

yields

repr                          shape     length    len(shape)
array(2.2999999999999998)     ()        0         0         
array([], dtype=float64)      (0,)      0         1         
array([ 2.3])                 (1,)      1         1         
array([ 2.3,  3.6])           (2,)      2         1

You can distinguish between an "empty" array (e.g. np.array([])) and a numpy scalar (e.g. np.array(2.3)) by looking at the length of the shape.

share|improve this answer
    
Thanks @unubtu. That's cool, but how can I distinguish between a with length of 1 and an empty a? –  mishaF Dec 30 '10 at 20:10
    
Got it! That helps tons. –  mishaF Dec 30 '10 at 20:43
up vote 1 down vote

It looks like the size property of ndarrays will work in this case if you know that the array is one-dimensional. In my opinion, a.size is a lot more readable than len(np.atleast_1d(a)). However, note that the size property will return the total number of elements in the array if it has more than one dimension:

In [1]: import numpy as np

In [2]: np.array(2.3).size
Out[2]: 1

In [3]: np.array([1, 2]).size
Out[3]: 2

In [4]: np.array([[1,2], [3,4]]).size
Out[4]: 4
share|improve this answer
up vote 0 down vote 
a = np.array([2.3])
print len(a)
share|improve this answer
    
this works for a constant as you show, but not for an array input –  ecoe Aug 31 '13 at 22:39

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.