pd.date_range Attribute_Error

[vipatel10]

I'm using pandas 0.12 and numpy 1.7.1.

I'm trying to calculate the number of business days between two columns of dates. Not all rows have valid dates though, and pd.date_range does not have an ignore errors option, so I'm getting this error:

In [439]: td = pd.DataFrame({'a':[pd.Timestamp('2010-1-1'), pd.Timestamp('2010-2-1')], 'b':[pd.Timestamp('2010-1-10'), p
d.Timestamp('2010-2-9')]})

In [440]: td.apply(lambda row: len(pd.date_range(row['a'], row['b'], freq='B')), axis=1)
Out[440]:
0    6
1    7
dtype: int64

In [441]: td = pd.DataFrame({'a':[pd.Timestamp('2010-1-1'), pd.Timestamp('2010-2-1')], 'b':[pd.Timestamp('2010-1-10'), p
d.NaT]})

In [442]: td.apply(lambda row: len(pd.date_range(row['a'], row['b'], freq='B')), axis=1)
---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
<ipython-input-442-30b1500c1594> in <module>()
----> 1 td.apply(lambda row: len(pd.date_range(row['a'], row['b'], freq='B')), axis=1)

C:\Python27\lib\site-packages\pandas\core\frame.pyc in apply(self, func, axis, broadcast, raw, args, **kwds)
   4414                     return self._apply_raw(f, axis)
   4415                 else:
-> 4416                     return self._apply_standard(f, axis)
   4417             else:
   4418                 return self._apply_broadcast(f, axis)

C:\Python27\lib\site-packages\pandas\core\frame.pyc in _apply_standard(self, func, axis, ignore_failures)
   4489                     # no k defined yet
   4490                     pass
-> 4491                 raise e
   4492
   4493

AttributeError: ("'NaTType' object has no attribute 'tz'", u'occurred at index 1')

In [443]:

[jreback]

That is quite ineffficient as you'd have to construct the entire series. Try this instead.
Period handingling of NaT is not supported currently, but easy enough to roll your own.

In [27]: def f(row):
   ....:     if pd.isnull(row['a']) or pd.isnull(row['b']):
   ....:         return np.nan
   ....:     return pd.Period(row['a'],freq='B')-pd.Period(row['b'],freq='B')
   ....: 

In [29]: td.apply(f,axis=1)
Out[29]: 
0    -5
1   NaN
dtype: float64

[vipatel10]

Thanks so much!

On Sat, Oct 12, 2013 at 7:04 PM, jreback notifications@github.com wrote:

That is quite ineffficient as you'd have to construct the entire series.
Try this instead.
Period handingling of NaT is not supported currently, but easy enough to
roll your own.

In [27]: def f(row):
....: if pd.isnull(row['a']) or pd.isnull(row['b']):
....: return np.nan
....: return pd.Period(row['a'],freq='B')-pd.Period(row['b'],freq='B')
....:

In [29]: td.apply(f,axis=1)
Out[29]:
0 -5
1 NaN
dtype: float64

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/5200#issuecomment-26207925
.

[vipatel10]

Actually, I just noticed this below. Does that make sense to you? I would think all three should be 3, but notice one result is 4.

In [475]: a = pd.Period(pd.Timestamp('2013-10-1'), freq='B')

In [476]: b = pd.Period(pd.Timestamp('2013-10-4'), freq='B')

In [477]: b - a
Out[477]: 3L

In [478]: b = pd.Period(pd.Timestamp('2013-10-5'), freq='B')

In [479]: b - a
Out[479]: 4L

In [480]: b = pd.Period(pd.Timestamp('2013-10-6'), freq='B')

In [481]: b - a
Out[481]: 3L

In [482]:

[vipatel10]

This I think is the problem:

In [485]: b = pd.Period(pd.Timestamp('2013-10-4'), freq='B')

In [486]: b
Out[486]: Period('2013-10-04', 'B')

In [487]: b = pd.Period(pd.Timestamp('2013-10-5'), freq='B')

In [488]: b
Out[488]: Period('2013-10-07', 'B')

In [489]: b = pd.Period(pd.Timestamp('2013-10-6'), freq='B')

In [490]: b
Out[490]: Period('2013-10-04', 'B')

In [491]:

[jreback]

that is correct, business days are not on weekends (10/6) is a sunday

[vipatel10]

Right, but why does b = pd.Period(pd.Timestamp('2013-10-5'), freq='B') round up to 10-7-13, but

b = pd.Period(pd.Timestamp('2013-10-6'), freq='B') rounds down to 10-4-13?

[jreback]

might be a bug.....will create an issue about it....

[vipatel10]

Thanks!

On Sat, Oct 12, 2013 at 7:47 PM, jreback notifications@github.com wrote:

might be a bug.....will create an issue about it....

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/5200#issuecomment-26208464
.

[jreback]

see #5203