Inconsistent behavior in Timestamp for large negative values

[ssanderson]

A small, complete example of the issue

# Your code here
In [1]: import numpy as np; import pandas as pd
In [3]: pd.Timestamp(np.iinfo(np.int64).min + 80000000000000)
Out[3]: Timestamp('2262-04-11 22:26:03.145224192')

In [4]: pd.Timestamp(np.iinfo(np.int64).min + 90000000000000)
Out[4]: Timestamp('1677-09-22 01:12:43.145224192')

In [5]: pd.Timestamp(np.iinfo(np.int64).min + 85000000000000)
---------------------------------------------------------------------------
OutOfBoundsDatetime                       Traceback (most recent call last)
<ipython-input-5-47bf9ab2ad0f> in <module>()
----> 1 pd.Timestamp(np.iinfo(np.int64).min + 85000000000000)

pandas/tslib.pyx in pandas.tslib.Timestamp.__new__ (pandas/tslib.c:9932)()

pandas/tslib.pyx in pandas.tslib.convert_to_tsobject (pandas/tslib.c:26453)()

pandas/tslib.pyx in pandas.tslib._check_dts_bounds (pandas/tslib.c:30034)()

OutOfBoundsDatetime: Out of bounds nanosecond timestamp: 2262-04-11 23:49:23

Expected Output

I would expect all the values listed above to evaluate to dates in the 1600s. The behavior here looks like there might be a signed integer overflow happening in C somewhere, which is a little nerve-wracking since that's undefined behavior unless pandas is compiled with -fwrapv.

Output of pd.show_versions()

# Paste the output here

In [8]: pd.show_versions()

INSTALLED VERSIONS

commit: None
python: 2.7.10.final.0
python-bits: 64
OS: Linux
OS-release: 4.2.0-16-generic
machine: x86_64
processor: x86_64
byteorder: little
LC_ALL: None
LANG: en_US.UTF-8
LOCALE: None.None

pandas: 0.19.0
nose: None
pip: 8.1.2
setuptools: 28.3.0
Cython: None
numpy: 1.11.2
scipy: None
statsmodels: None
xarray: None
IPython: 5.1.0
sphinx: None
patsy: None
dateutil: 2.5.3
pytz: 2016.7
blosc: None
bottleneck: None
tables: None
numexpr: None
matplotlib: None
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

[chris-b1]

Also goes wrong with numpy - so probably something in here that's causing an overflow, although there really shouldn't be.

pandas/pandas/src/datetime/np_datetime.c

Line 797 in a40e185

convert_datetime_to_datetimestruct(pandas_datetime_metadata *meta,

In [131]: np.datetime64(np.iinfo(np.int64).min + 80000000000000, 'ns')
Out[131]: numpy.datetime64('2262-04-11T22:26:03.145224192')

[chris-b1]

Right here -

pandas/pandas/src/datetime/np_datetime.c

Line 928 in a40e185

set_datetimestruct_days((dt - (perday-1)) / perday, out);