0 
len = 5
results = np.empty(len)
for i in range(len):
    results[i] = ([999, 'sss'])

I would like to predefine the length of a list, which is 5 and then fill in some values which contains both numerical and string variables... However, whatever I use

for i in range(len):
    results[i] = ([[999], ['sss']])

or

for i in range(len):
    results[i] = [[999], ['sss']]

I got error message as 

ValueError                                Traceback (most recent call last)
<ipython-input-114-4cfd5264742d> in <module>()
      2 results = np.empty(len)
      3 for i in range(len):
----> 4     results[i] = ([999, 'sss'])

ValueError: setting an array element with a sequence.

How should i do?

Improve this question
1
  • What are you trying to create? A real Python list? A 1d numpy array? A 2d array? It you want an array you need to specify the dtype. Equally important, what are you going to do with this array/list? Commented Mar 29, 2018 at 19:08

2 Answers 2

Reset to default
2

you could use the dtype 'object' for your results array:

length = 5
results = np.empty(length, dtype='object')
for i in range(length):
    results[i] = ([999, 'sss'])

>>> results
array([list([999, 'sss']), list([999, 'sss']), list([999, 'sss']),
       list([999, 'sss']), list([999, 'sss'])], dtype=object)

If what you're putting into your results array is always of length 2, better to specify the appropriate shape when you make your np.empty:

length = 5
results = np.empty((length,2), dtype='object')
for i in range(length):
    results[i] = ([999, 'sss'])

>>> results
array([[999, 'sss'],
       [999, 'sss'],
       [999, 'sss'],
       [999, 'sss'],
       [999, 'sss']], dtype=object)

Note, try not to use len as a variable, as it is already a built-in function in python (I replaced it with length in my examples above)

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

Comments

1

Unless you really have a need for a numpy array with mixed dtype, I'd recommend making a list of lists:

In [96]: alist = []
In [97]: for i in range(5):
    ...:     alist.append([999, 'sss'])
    ...:     
In [98]: alist
Out[98]: [[999, 'sss'], [999, 'sss'], [999, 'sss'], [999, 'sss'], [999, 'sss']]

Or if you make a list of tuples, you can go on an make a structured array - 1d with 2 fields of the desired dtype:

In [99]: alist = []
In [100]: for i in range(5):
     ...:     alist.append((999, 'sss'))
     ...:     
     ...:     
In [101]: alist
Out[101]: [(999, 'sss'), (999, 'sss'), (999, 'sss'), (999, 'sss'), (999, 'sss')]
In [102]: np.array(alist, dtype='int,U3')
Out[102]: 
array([(999, 'sss'), (999, 'sss'), (999, 'sss'), (999, 'sss'),
       (999, 'sss')], dtype=[('f0', '<i8'), ('f1', '<U3')])

When trying to mix integers and strings you need to pay attention to the intended use.

Note what happens when you create an 'empty' array:

In [103]: arr = np.empty(4)
In [104]: arr
Out[104]: array([0.e+000, 0.e+000, 0.e+000, 5.e-324])
In [105]: arr.dtype
Out[105]: dtype('float64')

It creates an array of the desired size, but with a default dtype, float. The initial values are garbage - though displayed as floats.

That means that the only thing you can put in arr[0] is a float.

In [106]: arr[0]=999       # converted to float
In [107]: arr[1]='sss'
...
ValueError: could not convert string to float: 'sss'
In [108]: arr
Out[108]: array([9.99e+002, 0.00e+000, 0.00e+000, 4.94e-324])

And it won't accept a list or tuple either, a sequence. It has to be something that can be converted to a float.

Specifying an object dtype does let you put anything in elements, because the elements actually hold pointers - just like a list.

In [110]: arr = np.empty(4,dtype=object)
In [111]: arr
Out[111]: array([None, None, None, None], dtype=object)
In [112]: arr[0]=1
In [113]: arr[1]=.999
In [114]: arr[2]='sss'
In [115]: arr[3]=[999,'sss']
In [116]: arr
Out[116]: array([1, 0.999, 'sss', list([999, 'sss'])], dtype=object)

Such an array is either a glorified list, or a debased one, depending on your needs.

Improve this answer

1 Comment

Hei, you first solution is what I really want... but also where I started... because though I want the same result, I do not want to use append function... In reality, my len is not 5 but 1000000, so I believe, as a matlab user, append will make it really slow... so I want to initialize the size and then fill in all cells with the right stuff..

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.