dayfirst non working when constructing a pd.DatetimeIndex with an array

[gbakalian]

assert pd.DatetimeIndex(['10/02/2014'], dayfirst=True) == pd.DatetimeIndex(np.asarray(['10/02/2014'], dtype=object), dayfirst=True)

[jreback]

this is a known bug, see here: #3341

[jorisvandenbossche]

@jreback I think this is another issue:

In [5]: pd.to_datetime(['10/02/2014'], dayfirst=True)[0]
Out[5]: Timestamp('2014-02-10 00:00:00', tz=None)

In [6]: pd.to_datetime(np.asarray(['10/02/2014'], dtype='object'), dayfirst=True)[0]
Out[6]: Timestamp('2014-02-10 00:00:00', tz=None)

In [8]: pd.DatetimeIndex(np.asarray(['10/02/2014'], dtype='object'), dayfirst=True)[0]
Out[8]: Timestamp('2014-10-02 00:00:00', tz=None)

There seems to be a difference between to_datetime and DatetimeIndex.
Also, the other issue is about dayfirst not being strict, so no error for eg "10/14/2014". While here, it is just wrong even in a case that is perfectly possible for dayfirst.

[jreback]

@jorisvandenbossche I agree dayfirst is non-strict / wrong because dateutil is broken IIRC

I had closed this because it is a dupe of #3341 but not prob