numpy.split on non-UTC, tz-aware data undergoes UTC roundtrip

[isacarnekvist]

Code Sample, a copy-pastable example if possible

import pandas as pd
import numpy as np
indices = pd.date_range('2016-01-01 00:00:00+0200', freq='S', periods=10)
np.split(indices, indices_or_sections=[])
Out:
[DatetimeIndex(['2015-12-31 20:00:00+02:00', '2015-12-31 20:00:01+02:00',
               '2015-12-31 20:00:02+02:00', '2015-12-31 20:00:03+02:00',
               '2015-12-31 20:00:04+02:00', '2015-12-31 20:00:05+02:00',
               '2015-12-31 20:00:06+02:00', '2015-12-31 20:00:07+02:00',
               '2015-12-31 20:00:08+02:00', '2015-12-31 20:00:09+02:00'],
              dtype='datetime64[ns, pytz.FixedOffset(120)]', freq='S')]

Expected Output

[DatetimeIndex(['2016-01-01 00:00:00+02:00', '2016-01-01 00:00:01+02:00',
               '2016-01-01 00:00:02+02:00', '2016-01-01 00:00:03+02:00',
               '2016-01-01 00:00:04+02:00', '2016-01-01 00:00:05+02:00',
               '2016-01-01 00:00:06+02:00', '2016-01-01 00:00:07+02:00',
               '2016-01-01 00:00:08+02:00', '2016-01-01 00:00:09+02:00'],
              dtype='datetime64[ns, pytz.FixedOffset(120)]', freq='S')]

output of pd.show_versions()

In [16]: pd.show_versions()

INSTALLED VERSIONS
------------------
commit: None
python: 2.7.6.final.0
python-bits: 64
OS: Linux
OS-release: 4.2.0-38-generic
machine: x86_64
processor: x86_64
byteorder: little
LC_ALL: None
LANG: en_US.UTF-8

pandas: 0.18.1
nose: 1.3.7
pip: 8.1.2
setuptools: 25.2.0
Cython: 0.24.1
numpy: 1.11.1
scipy: 0.18.0
statsmodels: None
xarray: None
IPython: 1.2.1
sphinx: None
patsy: None
dateutil: 2.5.3
pytz: 2016.6.1
blosc: None
bottleneck: None
tables: 3.1.1
numexpr: 2.6.1
matplotlib: 1.3.1
openpyxl: None
xlrd: None
xlwt: None
xlsxwriter: None
lxml: None
bs4: None
html5lib: 0.999
httplib2: 0.8
apiclient: None
sqlalchemy: None
pymysql: None
psycopg2: None
jinja2: None
boto: 2.42.0
pandas_datareader: None

[jreback]

This is fixed in master, I dont know exactly which issue, @sinhrks fixed __array__ though IIRC to propogate meta-data.

In [46]: indices = pd.date_range('2016-01-01 00:00:00+0200', freq='S', periods=10)
    ...: np.split(indices, indices_or_sections=[])
    ...: 
Out[46]: 
[DatetimeIndex(['2015-12-31 20:00:00+02:00', '2015-12-31 20:00:01+02:00', '2015-12-31 20:00:02+02:00', '2015-12-31 20:00:03+02:00', '2015-12-31 20:00:04+02:00', '2015-12-31 20:00:05+02:00',
                '2015-12-31 20:00:06+02:00', '2015-12-31 20:00:07+02:00', '2015-12-31 20:00:08+02:00', '2015-12-31 20:00:09+02:00'],
               dtype='datetime64[ns, pytz.FixedOffset(120)]', freq='S')]

If you'd like to do a PR with tests of this would be great.

[jorisvandenbossche]

@jreback the output you show is the incorrect output, so I don't think this is fixed. Seems there goes something wrong in np.swapaxes(indices, 0, 0)

[jreback]

oh, looks like it underwent a UTC round-trip then. I c. ok.

Ok, we can mark it, though @isacarnekvist we have no need for np.split as .groupby effectively does almost all partitioning (yes we have discussions of actually having a .partition, but that's another issue).