Stuck on an issue?

Lightrun Answers was designed to reduce the constant googling that comes with debugging 3rd party libraries. It collects links to all the places you might be looking at while hunting down a tough bug.

And, if you’re still stuck at the end, we’re happy to hop on a call to see how we can help out.

BUG: ascontiguousarray and asfortranarray make 0d scalars into 1d arrays

See original GitHub issue

Not sure if this is intended behavior or a bug, but I had been under the assumption that the result of np.ascontiguousarray would be the same shape as the input. This is true except in the case of scalars.

>>> scalar = np.float64(3.0)
>>> np.asarray(scalar).shape  # asarray doesn't change shape
()
>>> np.ascontiguousarray(scalar).shape  # ascontiguousarray does
(1,)

If this is what should be expected, it would be helpful to document this in its docstring.

Thanks!

Issue Analytics

State:
Created 9 years ago
Comments:19 (17 by maintainers)

Top GitHub Comments

3reactions

eric-wiesercommented, Nov 15, 2017

How about we just deprecate np.asfortranarray and np.ascontiguousarray in favor of np.asarray(..., order='C') and np.asarray(..., order='F'), since the former are:

More verbose
Both broken with respect to their documentation on 0d arrays

1reaction

charriscommented, Jun 21, 2015

Thinking more, I suspect this is intended for producing arrays for C and, in the case of asfortrancontiguous, for Fortran functions. Allowing zero dimensional arrays might well cause problems.