BUG: error in string array generated with arange function

[m-aguena]

Describe the issue:

Arrays created with np.arange(n, dtype=str) for n>2 raise an error.

Reproduce the code example:

import numpy as np
np.arange(3, dtype=str)

Error message:

File "<stdin>", line 1, in <module>
ValueError: no fill-function for data-type.

NumPy/Python version information:

1.23.1, 3.10.4 (main, Mar 31 2022, 08:41:55) [GCC 7.5.0]

[seberg]

I am not quite convinced that arange on strings make much sense. Maybe you would want np.arange("a", "z") to work? You can cast to string later on with .astype(str).
It may be a bit weird that n==1 apparently works. But generally I would lean towards closing it as correct (with the caveat that maybe we should disallow the n==1 case as well.

[rkern]

I concur.

[m-aguena]

I agree with this solution, as long as there is consistency (n==2 also works btw).

[ddasilva]

I'm trying to work on this by adding a check to the multiarraymodule.c function for arange() which throws a TypeError if the dtype is string. I'm stuck trying to figure out how to check if PyArray_Descr *typecode is a string. Any advice?

[ddasilva]

Nevermind, figured it out. PR coming soon..

[rgommers]

PR ready, so that should just be reviewed (seems like not much work) - so removed the Close? label.

[seberg]

PR wasn't correctly linked, closing.