BUG: dt.strftime causes TypeError when there is only one row in DataFrame

[yp514]

Code below outlines the issue, if I have a DataFrame with a date column and want to transfer the datetime to string using df.loc[:,'date'] = df['date'].dt.strftime('%Y-%m-%d'), it works fine as long as there is more than 1 row in the Data Frame. If there is only one row then you get a Type Error (see below for full error). If, on the other hand, I use df['date']=df['date'].dt.strftime('%Y-%m-%d') then it works fine for one or more rows. This was observed on both 0.17.1 and 0.19.2.

----- Error Generated -----
Traceback (most recent call last):
File "/usr/local/lib/python3.4/site-packages/pandas/core/internals.py", line 2265, in _try_coerce_args
other = other.astype('i8', copy=False).view('i8')
ValueError: invalid literal for int() with base 10: '2017-02-23'

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
File "./testpd.py", line 35, in
do_stuff(d1)
File "./testpd.py", line 17, in do_stuff
df.loc[:,'date'] = df['date'].dt.strftime('%Y-%m-%d')
File "/usr/local/lib/python3.4/site-packages/pandas/core/indexing.py", line 141, in setitem
self._setitem_with_indexer(indexer, value)
File "/usr/local/lib/python3.4/site-packages/pandas/core/indexing.py", line 533, in _setitem_with_indexer
setter(labels[0], value)
File "/usr/local/lib/python3.4/site-packages/pandas/core/indexing.py", line 473, in setter
s._data = s._data.setitem(indexer=pi, value=v)
File "/usr/local/lib/python3.4/site-packages/pandas/core/internals.py", line 3168, in setitem
return self.apply('setitem', **kwargs)
File "/usr/local/lib/python3.4/site-packages/pandas/core/internals.py", line 3056, in apply
applied = getattr(b, f)(**kwargs)
File "/usr/local/lib/python3.4/site-packages/pandas/core/internals.py", line 668, in setitem
values, _, value, _ = self._try_coerce_args(self.values, value)
File "/usr/local/lib/python3.4/site-packages/pandas/core/internals.py", line 2270, in _try_coerce_args
raise TypeError
TypeError

----- Code to Reproduce Issue -----

#!/usr/bin/env python3
""" Demonstrate Pandas issue """

import datetime as dt
import pandas as pd

def do_stuff(d):
    df = pd.DataFrame(d)
    df['date'] = df['date'].dt.strftime('%Y-%m-%d')
    df['id'] = df['id'].astype(str)
    df['fl'] = df['fl'].astype(str)
    print("df[df]: " + str(df))

    df = pd.DataFrame(d)
    df.loc[:,'id'] = df['id'].astype(str)
    df.loc[:,'fl'] = df['fl'].astype(str)
    df.loc[:,'date'] = df['date'].dt.strftime('%Y-%m-%d')
    print("df.loc[df]: " + str(df))


""" With 2 rows in Data Frame, works fine """
d2 = { 'date': pd.Series([dt.datetime.today(), dt.datetime.today()]),
      'bsym': pd.Series(['blah', 'blah2']),
      'id': pd.Series([1, 2]),
      'fl': pd.Series([1.5, 3]) }
print("D2...")
do_stuff(d2)

""" With 1 row in Data Frame, exception occurs on date column """
d1 = { 'date': pd.Series([dt.datetime.today()]),
      'bsym': pd.Series(['blah']),
      'id': pd.Series([1]),
      'fl': pd.Series([1.5]) }
print("D1...")
do_stuff(d1)

[jorisvandenbossche]

More comprehensive reproducible example:

In [14]: df1 = pd.DataFrame({'date': pd.Series([dt.datetime.today()])})

In [15]: df1
Out[15]: 
                        date
0 2017-02-24 10:41:07.209562

In [18]: df1['date'] = 'string'

In [19]: df1
Out[19]: 
     date
0  string

In [20]: df1 = pd.DataFrame({'date': pd.Series([dt.datetime.today()])})

In [21]: df1.loc[:,'date'] = 'string'
...
TypeError: 

And indeed works when having more than 1 row.

[jreback]

yeah .loc does inference on the resulting column to potentially change the dtype, looks like an uncaught path.

[yp514]

I am relatively new to pandas... Is the workaround suggested in the original script (df['date'] = df['date'].dt...) completely equivalent ? Thank You

…

Sent from my iPhone

On Feb 24, 2017, at 07:11, Jeff Reback ***@***.***> wrote: yeah .loc does inference on the resulting column to potentially change the dtype, looks like an uncaught path. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or mute the thread.

[jorisvandenbossche]

Yes, df['date'] and df.loc[:, 'date'] should be equivalent

[jschendel]

This appears to be fixed on 0.22.0 (so really probably on 0.21.x too):

In [1]: import pandas as pd; import datetime as dt

In [2]: pd.__version__
Out[2]: '0.22.0'

In [3]: df1 = pd.DataFrame({'date': pd.Series([dt.datetime.today()])})

In [4]: df1.loc[:,'date'] = 'string'

In [5]: df1
Out[5]:
     date
0  string

Could still use a test to ensure there's not a regression.