BUG: pandas.tools.util.cartesian_product doesn't work on DatetimeIndex objects

[shoyer]

This breaks the new pandas.MultiIndex.from_product function when one of the product arrays is DatetimeIndex. Observe:

>>> import pandas as pd
>>> from pandas.tools.util import cartesian_product
>>> x, y = cartesian_product([[1, 2], pd.date_range('2000-01-01', periods=2)])
>>> print x
[1 1 2 2]
>>> print y.values
['1999-12-31T16:00:00.000000000-0800' '1999-12-31T16:00:00.000000000-0800'
 '2000-01-01T16:00:00.000000000-0800' '2000-01-01T16:00:00.000000000-0800']

If cartesian_product was working properly, the dates should actually be [1999, 2000, 1999, 2000].

The simple fix would be to convert everything into a numpy array before calculating the cartesian product.

[jreback]

In [3]: MultiIndex.from_product([[1, 2], pd.date_range('2000-01-01', periods=2)]).values
Out[3]: 
array([(1, Timestamp('2000-01-01 00:00:00', tz=None)),
       (1, Timestamp('2000-01-01 00:00:00', tz=None)),
       (2, Timestamp('2000-01-02 00:00:00', tz=None)),
       (2, Timestamp('2000-01-02 00:00:00', tz=None))], dtype=object)

In [4]: from pandas.tools.util import cartesian_product

In [5]: >>> x, y = cartesian_product([[1, 2], pd.date_range('2000-01-01', periods=2)])

In [6]: x
Out[6]: array([1, 1, 2, 2])

In [7]: y
Out[7]: 
<class 'pandas.tseries.index.DatetimeIndex'>
[2000-01-01, ..., 2000-01-02]
Length: 4, Freq: None, Timezone: None

In [8]: y.values
Out[8]: 
array(['1999-12-31T19:00:00.000000000-0500',
       '1999-12-31T19:00:00.000000000-0500',
       '2000-01-01T19:00:00.000000000-0500',
       '2000-01-01T19:00:00.000000000-0500'], dtype='datetime64[ns]')

looks ok to me, what is the problem?

[jorisvandenbossche]

It's the order that is not OK:

In [22]: x, y = cartesian_product([[1, 2], pd.date_range('2000-01-01', periods=2)])
In [23]: x
Out[23]: array([1, 1, 2, 2])
In [25]: y.day
Out[25]: array([1, 1, 2, 2], dtype=int64)

In [26]: x, y = cartesian_product([[1, 2], [1,2]])
In [28]: x
Out[28]: array([1, 1, 2, 2])
In [29]: y
Out[29]: array([1, 2, 1, 2])

See difference between y.day in first case and y in second case.

[jreback]

ahh....I c...ok marking as a bug

[jreback]

@shoyer pull-request for this?

[shoyer]

Sure - I'll send something in tomorrow or Monday.

On Sat, Feb 22, 2014 at 6:18 AM, jreback [email protected] wrote:

@shoyer pull-request for this?

Reply to this email directly or view it on GitHub:
#6439 (comment)