numpy frombuffer - AttributeError: 'str' object has no attribute '__buffer__'

Question

Python Version: 3.5.2 Numpy Version : 1.12.1

Error:

import numpy as np
s = 'Hello World'
np.frombuffer(s, dtype='S1')
AttributeError: 'str' object has no attribute '__buffer__'

Things Tried:

1. Tried Online Ideone compiler, got same error in Python3.xx.
2. Referred scipy faqs for numpy and python compatible version, which states "NumPy support the Python 2.x series, (versions 2.6 and 2.7), as well as Python 3.2 and newer. The first release of NumPy to support Python 3 was NumPy 1.5.0."

Can't figure out the issue, tried stackoverflow for same issue but nothing found, may be i had missed it. Any suggestions or leads on why the error and how to resolve it in python3.xx.

Answer 1

In a PY3 session:

In [62]: np.frombuffer('hello world')
...
AttributeError: 'str' object has no attribute '__buffer__'
In [63]: np.frombuffer(b'hello world')
...
ValueError: buffer size must be a multiple of element size
In [64]: np.frombuffer(b'hello world',dtype='S1')
Out[64]: 
array([b'h', b'e', b'l', b'l', b'o', b' ', b'w', b'o', b'r', b'l', b'd'],  dtype='|S1')

In PY3, the default string type is unicode. The b is used to create and display bytestrings.

The np.frombuffer docs should be updated to reflect the difference. The 'hello world' example only works with PY2 or with PY3 bytestrings.

As I noted in the comments, there are few SO questions regarding frombuffer, indicating that it is rarely used. np.array is by far the most common way of making an array, even from strings:

In [80]: np.array('hello')
Out[80]: 
array('hello', 
      dtype='<U5')

or use list to split the string into characters:

In [81]: np.array(list('hello'))
Out[81]: 
array(['h', 'e', 'l', 'l', 'o'], 
      dtype='<U1')

In [82]: np.array(b'hello')
Out[82]: 
array(b'hello', 
      dtype='|S5')
In [83]: np.array(list(b'hello'))
Out[83]: array([104, 101, 108, 108, 111])

In [85]: np.fromiter('hello','S1')
Out[85]: 
array([b'h', b'e', b'l', b'l', b'o'], 
      dtype='|S1')
In [86]: np.fromiter('hello','U1')
Out[86]: 
array(['h', 'e', 'l', 'l', 'o'], 
      dtype='<U1')*

I created a bug issue: https://github.com/numpy/numpy/issues/8933