熊猫datetime索引错误或可能的时间旅行发现

Question

Is this a bug? It looks like pushing a datetime object into a dataframe, then indexing on the datetime column, does some scrambling of the date but produces no error (see line 97). Interestingly the view looks right so I'm guessing it's some sort of memory indexing thing.

This was on a fairly recent build of pandas: pandas-0.9.1.dev_85d982d-py2.7-linux-x86_64.egg.

In [93]: import datetime, pandas
In [94]: df = pandas.DataFrame([[datetime.datetime.today(), 12.1]], columns=['Date', 'Value'])

In [95]: df = df.set_index('Date')

In [96]: df
Out[96]: 
                            Value
Date                             
2012-11-22 12:12:40.905739   12.1    

In [97]: df.index
Out[97]: 
<class 'pandas.tseries.index.DatetimeIndex'>
[2190-12-31 02:18:44.941732032]
Length: 1, Freq: None, Timezone: None

In [98]: df = df.reset_index()

In [99]: df
Out[99]: 
                           Date  Value
0 2190-12-31 02:18:44.941732032   12.1

Answer 1

Updating to a newer version of numpy (and rebuilding pandas) seems to have fixed the problem.

numpy. version '1.8.0.dev-fd78546' pandas. version '0.9.1.dev-85d982d'

I posted and closed the issue here: https://github.com/pydata/pandas/issues/2329

Feel free to close this question on stackoverflow if you think it is superfluous, though who knows maybe someone else will run into it too.