TST: start tests for PeriodConverter

[jorisvandenbossche]

@jreback The tests I started related to PR #9050

[jreback]

b69b9d3 (obviously you have more)

[jorisvandenbossche]

yep, I saw that. But I thought to add it both as a test for PeriodConverter (which then seemed to not yet exist), and as a plotting test (but only came around to the first one yesterday, and you already did the second).

[jreback]

@jorisvandenbossche thought you had merged this, looks ok

[jorisvandenbossche]

there are two parts skipped for now in this PR:

• the datetime64 part is commented out -> there is already an issue about that: ENH: Period to accept datetime64 value? #9054
• the unicode test (test_convert_accepts_unicode) is failing, so I was skipping it for now (I copied the test from DatetimeConverter): @jreback this is something that should work?

[jreback]

Never used PeriodConverter but the undelrying supports, so maybe just a small issue

In [1]: pd.Timestamp(u'20121201')
Out[1]: Timestamp('2012-12-01 00:00:00')

In [2]: pd.Period(u'20121201',freq='D')
Out[2]: Period('2012-12-01', 'D')

[jorisvandenbossche]

It is actually a quite simple reason. From https://github.com/pydata/pandas/blob/master/pandas/tseries/converter.py#L112:

valid_types = (str, datetime, Period, pydt.date, pydt.time)
if isinstance(values, valid_types):
....

here of course testing for str does not work for unicode

I will add this

[jorisvandenbossche]

I just have to use pd.compat.string_types instead of str?

[jreback]

yep