ENH: Support datetime.date in Period Index

[max-sixty]

PeriodIndexs should resolve when datetime.date is passed into in, but doesn't. Example below.

Tracing this through, it looks like parse_time_string is expected to return a tuple in get_loc (which is called by __contains__): https://github.com/pydata/pandas/blob/master/pandas/tseries/period.py#L610

But if passed something that isn't a string, just returns it back as a single item: https://github.com/pydata/pandas/blob/master/pandas/tseries/tools.py#L447

...so in this instance the lookup raises an exception and mistakenly propagates up that the value can't be found.

What's the best way forward here? Should the return type always be the same? Or do we add a check in get_loc to stop it calling parse_time_string? Or broaden the except here to be any exception: https://github.com/pydata/pandas/blob/master/pandas/tseries/period.py#L612

My inclination is the first, but whether that will have some blast radius / what the standards are? The third seems like a reasonable cut through too.

In [9]:

period_index
Out[9]:
PeriodIndex(['2015-01-01', '2015-01-02', '2015-01-05', '2015-01-06',
             '2015-01-07', '2015-01-08', '2015-01-09', '2015-01-12',
             '2015-01-13', '2015-01-14', '2015-01-15', '2015-01-16',
             '2015-01-19', '2015-01-20', '2015-01-21', '2015-01-22',
             '2015-01-23', '2015-01-26', '2015-01-27', '2015-01-28',
             '2015-01-29', '2015-01-30'],
            dtype='int64', freq='B')
In [10]:

'2015-01-06' in period_index
Out[10]:
True
In [15]:

dt = datetime.datetime(2015,1,6) 
dt
Out[15]:
datetime.datetime(2015, 1, 6, 0, 0)
In [16]:

dt in period_index
Out[16]:
False
In [17]:

d = datetime.date(2015,1,6) 
d
Out[17]:
datetime.date(2015, 1, 6)
In [18]:

d in period_index
Out[18]:
False

[sinhrks]

I agree we should accept datetime-likes as the same as other comparison ops.

I don't think parse_time_string in get_loc is a problem. It raises TypeError when the returned value is split to tuple, then catched and directly converted to Period. So, removing string_types condition in __contains__ should work?

• https://github.com/pydata/pandas/blob/master/pandas/tseries/period.py#L59

idx = pd.period_range('2011-01-01', freq='D', periods=10)
idx.get_loc(datetime.date(2011, 1, 3))
# 2

pd.Period(datetime.date(2011, 1, 3), freq='D') in idx
# True

PR is welcome:)

[max-sixty]

@sinhrks perfect, thanks!

[jreback]

as discussed in #10801 . This API needs some clarification as to how it should iteract (if at all) with Timestamps/datetime.datetime and such.