Question

I have a variable that contains the string ‘long’. How can I create a numpy dtype object with some type equivalent to long from this string? I have a file with many numbers and the corresponding types. int, float etc. are no problems, only long doesn’t work. I don’t want to hardcode some replacement long -> int32 or so in my code.

>>> np.dtype('long')
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
TypeError: data type not understood

Also: Is there a way to create a variable type from a string in pure python? I mean, I want the inverse of int.__name__, which converts the type name into a string.

Answer 1

In this particular case, I think you can use getattr to get it from the numpy module itself:

>>> import numpy as np
>>> np.dtype('long')
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
TypeError: data type not understood

but:

>>> getattr(np, 'long')
<type 'long'>
>>> np.dtype(getattr(np, 'long'))
dtype('int64')
>>> np.dtype(getattr(np, 'int'))
dtype('int32')
>>> np.dtype(getattr(np, 'float64'))
dtype('float64')
>>> np.dtype(getattr(np, 'float'))
dtype('float64')