Series with dtype='M8[us]' does not create NaT values

[adamsd5]

Using Pandas 0.10.1, python 2.7, and Numpy 1.6.2:

>>> npa = numpy.ndarray(1, dtype=numpy.int64)
>>> npa[0] = -9223372036854775808
>>> type(pandas.Series(npa, dtype='M8[ns]')[0])
<class 'pandas.tslib.NaTType'>
>>> type(pandas.Series(npa, dtype='M8[us]')[0])
<class 'pandas.tslib.Timestamp'>

The latter should also have class 'pandas.tslib.NaTType', right?

[wesm]

Units other than nanoseconds are not supported right now in indexing etc. This should be strictly enforced

[adamsd5]

Agreed. An exception or an automatic up-conversion to nanoseconds seems reasonable.

[jreback]

this was closed by #3516 in 0.12

In [6]: pandas.Series(npa, dtype='M8[us]')
TypeError: cannot convert datetimelike to dtype [datetime64[us]]