DataFrame.from_records incorrectly up-converts dtypes to object.

[dalejung]

rows = []
rows.append([pd.datetime(2010, 1, 1), 1])
rows.append([pd.datetime(2010, 1, 2), 'hi']) # test col upconverts to obj

df2_obj = pd.DataFrame.from_records(rows, columns=['date', 'test'])

print df2_obj.date.dtype # object

If the 'hi' is changed to an int then df2_obj.date.dtype is equal to M8[ns]

[wesm]

Oh boy. This is caused by a real snake nest in pandas/core/internals.py. Punting til post 0.10.1 =/

[jreback]

@dalejung

This is fixed in current master (though I think I will add some tests in any event),
can you verify

In [6]: rows = []
In [7]: rows.append([pd.datetime(2010, 1, 1), 1])
In [8]: rows.append([pd.datetime(2010, 1, 2), 'hi']) # test col upconverts to obj
In [9]: df2_obj = pd.DataFrame.from_records(rows, columns=['date', 'test'])
In [10]: df2_obj.dtypes
Out[10]: 
date    datetime64[ns]
test            object
dtype: object

Ints

In [12]: In [7]: rows.append([pd.datetime(2010, 1, 1), 1])
In [13]: In [8]: rows.append([pd.datetime(2010, 1, 2),1])
In [14]: df2_int = pd.DataFrame.from_records(rows, columns=['date', 'test'])
In [15]: df2_int.dtypes
Out[15]: 
date    datetime64[ns]
test             int64
dtype: object

[dalejung]

Yar! Just tested against master and it works.

[jreback]

great!