deepcopy failure on empty dataframes with non-empty column set (numpy 1.12 compatibility)

[bmcfee]

Code Sample, a copy-pastable example if possible

# Your code here
import pandas as pd
import copy

# This one works
df = pd.DataFrame(data=[], index=[])
df_copy = copy.deepcopy(df)

# This one fails
df2 = pd.DataFrame(data=[], index=[], columns=['a'])
df2_copy = copy.deepcopy(df2)

Problem description

This only occurs with numpy 1.12 (and, presumably above): when deepcopying an empty dataframe with a non-empty column set, it fails with the following:

ValueError: Iteration of zero-sized operands is not enabled

If the column set is also empty, everything works as expected.

On older numpy versions (1.11), it also works as expected.

Expected Output

Not failing.

Output of pd.show_versions()

# Paste the output here pd.show_versions() here INSTALLED VERSIONS ------------------ commit: None python: 3.4.0.final.0 python-bits: 64 OS: Linux OS-release: 4.4.0-62-generic machine: x86_64 processor: x86_64 byteorder: little LC_ALL: None LANG: en_US.UTF-8 LOCALE: en_US.UTF-8

pandas: 0.19.2
nose: 1.3.7
pip: 9.0.1
setuptools: 32.3.1.post20170108
Cython: 0.25.2
numpy: 1.12.0
scipy: 0.18.1
statsmodels: None
xarray: None
IPython: 5.1.0
sphinx: None
patsy: None
dateutil: 2.6.0
pytz: 2016.10
blosc: None
bottleneck: None
tables: None
numexpr: None
matplotlib: 2.0.0
openpyxl: None
xlrd: None
xlwt: None
xlsxwriter: None
lxml: None
bs4: None
html5lib: None
httplib2: None
apiclient: None
sqlalchemy: None
pymysql: None
psycopg2: None
jinja2: None
boto: None
pandas_datareader: None

[jreback]

not shocked this doesn't work, see #8571

we are not using an override (which we should be), to simply call our custom methods.

e.g. basically

def __copy__(self):
    return self.copy()

if you'd like to take a crack at this issue (and add this as a test) would be appreciated!

[bmcfee]

if you'd like to take a crack at this issue (and add this as a test) would be appreciated!

Not sure I know how best to fix it, as I haven't mucked around the internals of dataframes much. It's not obvious to me that this should be fixed in dataframe, ndframe, or somewhere else?

[jreback]

@bmcfee simply need to define __copy__ and __deepcopy__, can be done in pandas/core/generic/NDFrame. These are already defined for Index (in pandas/indexes/base.py (it may simply be possible to move those to pandas/core/base though as we have a mixin that goes thru the entire hierarchy PandasObject.