ERR: disallow direct tz setting

[nipunbatra]

related to #3714
tz should not be settable

df=pd.DataFrame({"events":np.ones(len(dates))},index=motor["timestamps"].astype("datetime64[s]"))
df.describe()

	events
count	2393
mean	1
std	0
min	1
25%	1
50%	1
75%	1
max	1

df

<class 'pandas.core.frame.DataFrame'>
DatetimeIndex: 2393 entries, 2013-05-30 11:58:36 to 2013-06-03 13:30:43
Data columns (total 1 columns):
events    2393  non-null values
dtypes: float64(1)


df.index.tz='Asia/Kolkata'
df.index

<class 'pandas.tseries.index.DatetimeIndex'>
[2013-05-30 17:28:36, ..., 2013-06-03 19:00:43]
Length: 2393, Freq: None, Timezone: Asia/Kolkata

df2=df[df.index.day==3]

---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
<ipython-input-39-23de8e1402be> in <module>()
----> 1 df2=df[df.index.day==3]

/usr/local/lib/python2.7/dist-packages/pandas/tseries/index.pyc in f(self)
 39             utc = _utc()
 40             if self.tz is not utc:
---> 41                 values = self._local_timestamps()
 42         return tslib.get_date_field(values, field)
 43     f.__name__ = name

/usr/local/lib/python2.7/dist-packages/pandas/tseries/index.pyc in     _local_timestamps(self)
399             values = self.asi8
400             indexer = values.argsort()
--> 401             result = tslib.tz_convert(values.take(indexer), utc, self.tz)
402 
403             n = len(indexer)

/usr/local/lib/python2.7/dist-packages/pandas/tslib.so in pandas.tslib.tz_convert     (pandas/tslib.c:20949)()

/usr/local/lib/python2.7/dist-packages/pandas/tslib.so in pandas.tslib._get_deltas (pandas/tslib.c:22754)()

/usr/local/lib/python2.7/dist-packages/pandas/tslib.so in pandas.tslib._get_utcoffset (pandas/tslib.c:11327)()

AttributeError: 'str' object has no attribute 'utcoffset'

A temporary fix which i did was to add seconds to data manually while creating the index.

[jreback]

Create a new index with the tz at the beginning;

This kind of assignment is not allowed (and should raise)

In [43]: df = DataFrame(randn(10,2),index=date_range('20130101',periods=10,tz='Asia/Kolkata'),columns=list('AB'))

In [44]: df
Out[44]: 
                                  A         B
2013-01-01 00:00:00+05:30  0.202436  0.409555
2013-01-02 00:00:00+05:30  0.537623  0.304266
2013-01-03 00:00:00+05:30  1.936195 -0.337695
2013-01-04 00:00:00+05:30  1.839265 -0.741898
2013-01-05 00:00:00+05:30 -0.126482 -0.599313
2013-01-06 00:00:00+05:30  0.228828 -0.252229
2013-01-07 00:00:00+05:30  0.909610 -0.728261
2013-01-08 00:00:00+05:30  0.487048 -0.056313
2013-01-09 00:00:00+05:30  0.582852  0.838708
2013-01-10 00:00:00+05:30 -1.220794  0.233096

In [45]: df.index
Out[45]: 
<class 'pandas.tseries.index.DatetimeIndex'>
[2013-01-01 00:00:00, ..., 2013-01-10 00:00:00]
Length: 10, Freq: D, Timezone: Asia/Kolkata

In [46]: df[df.index.day==3]
Out[46]: 
                                  A         B
2013-01-03 00:00:00+05:30  1.936195 -0.337695

[nipunbatra]

What would be the best way to create an index in the way i am doing from epoch timestamps and adding tz information.

[jreback]

pd.to_datetime(stamps_in_ns,tz=your_tz)

make sure that you stamps are in nanoseconds (prob just multipy by 1e6)

Timestamp will take the nanoseconds and give you the correct timestamp

In [6]: Series([Timestamp('20130101')])
Out[6]: 
0   2013-01-01 00:00:00
dtype: datetime64[ns]

In [7]: Series([Timestamp('20130101')]).astype('int')
Out[7]: 
0    1356998400000000000
dtype: int64

[nipunbatra]

Specifying timezone failed for me. Version 0.11.0

In [10]: t
Out[10]: 
array([1370445222, 1370445223, 1370445224, 1370445225, 1370445226,
    1370445227, 1370445228, 1370445229, 1370445230, 1370445231])

In [9]: pd.to_datetime((t*1e9).astype(int),tz='Asia/Kolkata')
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-9-ee97c24c5e36> in <module>()
----> 1 pd.to_datetime((t*1e9).astype(int),tz='Asia/Kolkata')

TypeError: to_datetime() got an unexpected keyword argument 'tz'

[jreback]

@nipunreddevil

from the cross-post

In [9]: pd.to_datetime((t*1e9).astype(int)).tz_localize('Asia/Kolkata')
Out[9]: 
<class 'pandas.tseries.index.DatetimeIndex'>
[2013-06-05 15:13:42, ..., 2013-06-05 15:13:51]
Length: 10, Freq: None, Timezone: Asia/Kolkata

[nipunbatra]

Please see timezone is changing, but not the time.

In [11]: pd.to_datetime((t*1e9).astype(int)).tz_localize('Asia/Kolkata')
Out[11]: 
<class 'pandas.tseries.index.DatetimeIndex'>
[2013-06-05 15:13:42, ..., 2013-06-05 15:13:51]
Length: 10, Freq: None, Timezone: Asia/Kolkata

In [12]: pd.to_datetime((t*1e9).astype(int)).tz_localize('UTC')
Out[12]: 
<class 'pandas.tseries.index.DatetimeIndex'>
[2013-06-05 15:13:42, ..., 2013-06-05 15:13:51]
Length: 10, Freq: None, Timezone: UTC

In [13]: pd.to_datetime((t*1e9).astype(int)).tz_localize('EST')
Out[13]: 
<class 'pandas.tseries.index.DatetimeIndex'>
[2013-06-05 15:13:42, ..., 2013-06-05 15:13:51]
Length: 10, Freq: None, Timezone: EST

Contrast this with

In [15]: datetime.datetime.fromtimestamp(t[0])
Out[15]: datetime.datetime(2013, 6, 5, 20, 43, 42)

[jreback]

This is why we need a method to do this......
the epoch stamps are 'utc', but after the to_datetime they are not in any timezone
so first have to put them in utc (tz_localize), then convert

In [5]: pd.to_datetime((t*1e9).astype(int)).tz_localize('UTC').tz_convert('Asia/Kolkata')
Out[5]: 
<class 'pandas.tseries.index.DatetimeIndex'>
[2013-06-05 20:43:42, ..., 2013-06-05 20:43:51]
Length: 10, Freq: None, Timezone: Asia/Kolkata

[nipunbatra]

Great!

[jreback]

Hmm, we should prevent this from being set. Changing this issue to an ERR one.

In [21]: df = DataFrame(randn(10,2),index=date_range('20130101',periods=10,tz='Asia/Kolkata'),columns=list('AB'))

In [22]: df
Out[22]: 
                                  A         B
2013-01-01 00:00:00+05:30  1.106337  0.257936
2013-01-02 00:00:00+05:30 -0.789945 -0.267006
2013-01-03 00:00:00+05:30 -1.002179 -1.844458
2013-01-04 00:00:00+05:30  0.638277  0.125499
2013-01-05 00:00:00+05:30 -1.733083  0.447643
2013-01-06 00:00:00+05:30  0.782098  0.906288
2013-01-07 00:00:00+05:30 -0.797574  0.418644
2013-01-08 00:00:00+05:30  1.550464 -0.957223
2013-01-09 00:00:00+05:30 -0.303808 -0.239765
2013-01-10 00:00:00+05:30  0.086044  0.485118

In [23]: df.index.tz
Out[23]: <DstTzInfo 'Asia/Kolkata' LMT+5:53:00 STD>

In [24]: df.index.tz = pytz.timezone('US/Eastern')

In [25]: df
Out[25]: 
                                  A         B
2012-12-31 13:30:00-05:00  1.106337  0.257936
2013-01-01 13:30:00-05:00 -0.789945 -0.267006
2013-01-02 13:30:00-05:00 -1.002179 -1.844458
2013-01-03 13:30:00-05:00  0.638277  0.125499
2013-01-04 13:30:00-05:00 -1.733083  0.447643
2013-01-05 13:30:00-05:00  0.782098  0.906288
2013-01-06 13:30:00-05:00 -0.797574  0.418644
2013-01-07 13:30:00-05:00  1.550464 -0.957223
2013-01-08 13:30:00-05:00 -0.303808 -0.239765
2013-01-09 13:30:00-05:00  0.086044  0.485118