Indexing DataFrame with DateOffset is nearly impossible

[JoergVanAken]

Code Sample, a copy-pastable example if possible

offsets = pd.Series(data=[-15, -10, -5, 0, 5, 10, 15], dtype=float).map(pd.DateOffset)
df = pd.DataFrame(index=[0, 1], columns=offsets)

df.loc[0, offsets[0]]          # raises an error
df.at[0, offsets[0]]           # this works
df.at[0, offsets[0]] = 1       # raises an error
df.set_value(0, offsets[0], 1) # this works but raises a FutureWarning

Problem description

It is very difficult to access cells of a DataFrame if the columns are of type DateOffset.

Expected Output

Accessing a cell via df.loc[0, offsets[0]] should be possible for read and write.

Output of pd.show_versions()

INSTALLED VERSIONS

commit: None
python: 3.6.2.final.0
python-bits: 64
OS: Windows
OS-release: 7
machine: AMD64
processor: Intel64 Family 6 Model 42 Stepping 7, GenuineIntel
byteorder: little
LC_ALL: None
LANG: None
LOCALE: None.None

pandas: 0.22.0
pytest: 3.2.1
pip: 9.0.1
setuptools: 36.4.0
Cython: 0.26
numpy: 1.13.1
scipy: 0.19.1
pyarrow: None
xarray: None
IPython: 6.1.0
sphinx: 1.6.3
patsy: 0.4.1
dateutil: 2.6.1
pytz: 2017.2
blosc: None
bottleneck: 1.2.1
tables: 3.4.2
numexpr: 2.6.2
feather: None
matplotlib: 2.0.2
openpyxl: 2.4.8
xlrd: 1.1.0
xlwt: 1.3.0
xlsxwriter: 0.9.8
lxml: 3.8.0
bs4: 4.6.0
html5lib: 0.9999999
sqlalchemy: 1.1.13
pymysql: None
psycopg2: None
jinja2: 2.9.6
s3fs: None
fastparquet: None
pandas_gbq: None
pandas_datareader: None

[jreback]

what are you trying to do? this is a very odd thing to actually try to index by

[JoergVanAken]

The above code is only a demonstration. In fact I have a couple of timeseries which I want to compare for example by correlation. Sometimes these series match better after introducing a time lag. The dataframe is used to store the results.

Of course it is more a kind of "sugar" and not really necessary.

First of all, I was just surprised that the code did not work as expected and it took me while to see that it was not my fault.

[chris-b1]

xref to #19211, probably some others.

Right now offsets aren't a first-class data type in pandas, they 'work' by being treated as generic python objects. Seems to be some kind of over-aggressive type inference here, you're welcome to take a look.