apply with .loc[:, col] method failed when the DataFrame only has one row and the column's type is datetime64

[ppiiyyaa4149]

Code Sample, a copy-pastable example if possible

import pandas as pd
import datetime as dt

# only one row
s = pd.DataFrame(columns=['a', 'b', 'c'], index=[1])
s.loc[:, 'a'] = dt.datetime.now()
s.loc[:, 'b'] = 1
s.loc[:, 'c'] = dt.datetime.now()

s.loc[:, 'b'] = s.loc[:, 'b'].apply(lambda x: x + 1)

# no problem
s.loc[:, 'c'] = s.loc[:, 'c'].apply(lambda x: x.date().strftime("%Y-%m-%d"))

# raise TypeError 
s.loc[:, 'a'] = s.loc[:, 'a'].apply(lambda x: x.date().strftime("%Y-%m-%d"))

# raise TypeError 
temp = s.loc[:, 'a'].apply(lambda x: x.date().strftime("%Y-%m-%d"))
s.loc[:, 'a'] = temp



# no problems when DataFrame has more than 2 rows
s = pd.DataFrame(columns=['a', 'b', 'c'], index=[1, 2])
s.loc[:, 'a'] = dt.datetime.now()
s.loc[:, 'b'] = 1
s.loc[:, 'c'] = dt.datetime.now()

s.loc[:, 'b'] = s.loc[:, 'b'].apply(lambda x: x + 1)
temp = s.loc[:, 'a'].apply(lambda x: x.date().strftime("%Y-%m-%d"))
s.loc[:, 'a'] = temp
# s.loc[:, 'a'] = s.loc[:, 'a'].apply(lambda x: x.date().strftime("%Y-%m-%d"))
s.loc[:, 'c'] = s.loc[:, 'c'].apply(lambda x: x.date().strftime("%Y-%m-%d"))

Problem description

In my situation, I need to get user's login history from the database, then trans the type of the login time(timestamp in MySQL db, datetime64 in pandas) to string("%Y-%m-%d"). But when a user only has one login record, I have this problem occurred.
Really got confused

Expected Output

In[3]: s
Out[3]: 
            a  b           c
1  2017-10-30  2  2017-10-30

Output of pd.show_versions()

INSTALLED VERSIONS

commit: None
python: 3.5.3.final.0
python-bits: 64
OS: Windows
OS-release: 10
machine: AMD64
processor: Intel64 Family 6 Model 60 Stepping 3, GenuineIntel
byteorder: little
LC_ALL: None
LANG: None
LOCALE: None.None
pandas: 0.19.2
nose: None
pip: 9.0.1
setuptools: 36.2.7
Cython: None
numpy: 1.12.1
scipy: 0.19.0
statsmodels: None
xarray: None
IPython: 5.3.0
sphinx: None
patsy: None
dateutil: 2.6.0
pytz: 2016.10
blosc: None
bottleneck: None
tables: None
numexpr: None
matplotlib: 2.0.2
openpyxl: 2.4.8
xlrd: 1.0.0
xlwt: None
xlsxwriter: None
lxml: 3.7.3
bs4: 4.6.0
html5lib: 0.9999999
httplib2: None
apiclient: None
sqlalchemy: 1.1.6
pymysql: 0.7.10.None
psycopg2: None
jinja2: 2.9.5
boto: None
pandas_datareader: 0.4.0

[gfyoung]

@zjhjz : Thanks for the report! Could you do us a quick favor and post the error message that you get?

[jreback]

The problem is not .apply, which FYI your usage is quite non-idiomatic.

Rather the inference on indexing is not coercing in the n=1 case.
this is a duplicate of #6942