recarray from record has no shape, cannot be sliced.

Creating new recarrays from a single record results in a recarray with shape = (). It cannot be indexed into and it has no length. For now I’m working around this by reshaping the new recarray to (1,).

Reproducing code example:

import numpy as np
ra = np.rec.array(np.ones(4, dtype='f4,f4'))
r = ra[0]
ra2 = np.core.records.array(r)
ra2.shape  # returns ()
ra2[0]
# np.array_equal(ra2.reshape(1,)[0], ra[0])  # True

Error message:

Traceback (most recent call last):
  File "/home/chad/miniconda3/envs/np/lib/python3.6/site-packages/IPython/core/interactiveshell.py", line 2963, in run_code
    exec(code_obj, self.user_global_ns, self.user_ns)
  File "<ipython-input-19-18a4fa8bb93d>", line 6, in <module>
    ra2[0]
  File "/home/chad/miniconda3/envs/np/lib/python3.6/site-packages/numpy/core/records.py", line 505, in __getitem__
    obj = super(recarray, self).__getitem__(indx)
IndexError: too many indices for array

Numpy/Python version information:

1.14.3 3.6.3 |Intel Corporation| (default, May 4 2018, 04:22:28) [GCC 4.8.2 20140120 (Red Hat 4.8.2-15)]

[eric-wieser]

Good find on that archive page! I’m thinking we should revive that as an informational NEP, perhaps? @mattip?

[charris]

There was also a pretty long discussion on the mailing list back around the time of the first NumPy release, I forget who pushed the idea, it was one of the early contributors.