Incorrect calculation of weekofyear?

[dhirschfeld]

The calculation of the week/weekofyear attribute of a Timestamp/DatetimeIndex differs from both python and mx.DateTime which both return week = 1 for 31-Dec-2013.

In [25]: import mx.DateTime
    ...: from datetime import datetime
    ...: d = datetime(2013,12,31)
    ...: 

In [26]: d.isocalendar()
Out[26]: (2014, 1, 2)

In [27]: mx.DateTime.DateTimeFrom(d).iso_week
Out[27]: (2014, 1, 2)

In [28]: pd.Timestamp(d).week
Out[28]: 53

[dhirschfeld]

Also inconsistent with scikits.timeseries:

In [36]: ts.Date('D', d).week
Out[36]: 1

[changhiskhan]

@dhirschfeld: just pushed a fix for this, give it shot on your end?

[dhirschfeld]

Taking mx.DateTime as the gold-standard it would still seem to disagree in a few instances:

In [4]: import mx.DateTime
   ...: dates = pd.date_range('01-Jan-1980','01-Jan-2020', freq='D')
   ...: year, week, day = asarray([mx.DateTime.DateTimeFrom(d).iso_week for d in dates]).T
   ...: errors = list(dates[(dates.week != week).nonzero()[0]].to_pydatetime())
   ...: 

In [5]: errors
Out[5]: 
[datetime.datetime(2000, 1, 1, 0, 0),
 datetime.datetime(2000, 1, 2, 0, 0),
 datetime.datetime(2005, 1, 1, 0, 0),
 datetime.datetime(2005, 1, 2, 0, 0)]

In [6]: [(mx.DateTime.DateTimeFrom(d).iso_week[1], pd.Timestamp(d).week) for d in errors]
Out[6]: [(52, 53), (52, 53), (53, 52), (53, 52)]

[changhiskhan]

ok should be good now. let me know if there's anything else. thanks