regression in np.concatenate with object dtypes in 1.10

[jreback]

This was working as of this commit: 1.10.0.dev0+95b2c24 (and in 1.9.2), so just broken very recently

In [1]: np.__version__
Out[1]: '1.10.0.dev0+3d2b8ca'

In [3]: np.concatenate([np.array(['a','b'],dtype=object),np.array(['a','b'],dtype=object)],axis=1)
IndexError: axis 1 out of bounds [0, 1)

In [3]:  np.concatenate([np.array(['a','b'],dtype=object),np.array(['a','b'],dtype=object)],axis=1)
Out[3]: array(['a', 'b', 'a', 'b'], dtype=object)

In [4]: np.__version__
Out[4]: '1.9.2'

[charris]

The axis=1 has been deprecated for 1-D arrays for years, we just made is an error instead of a warning ;) Use axis=0 (or default) instead.

[charris]

For instance, in 1.9.2

n [1]: import warnings

In [2]: warnings.simplefilter('always')

In [3]: a = np.array(['a','b'],dtype=object)

In [4]: concatenate([a, a], axis=1)
/home/charris/.local/bin/ipython:1: DeprecationWarning: axis != 0 for ndim == 1; this will raise an error in future versions of numpy
  #!/usr/bin/python
Out[4]: array(['a', 'b', 'a', 'b'], dtype=object)

[jreback]

hmm you must have just changed this yes?
ok then I'll fix this up then

[charris]

Yep. It's been deprecated since 2012-09-13. I don't know if it made it into the 1.7 release or not, that was a confusing period. If not, the message is only 1 3/4 years old, so maybe borderline for deprecation removal.

[jreback]

no that's fine
I on purpose have depreciations fail but only for testing