Reading single index DataFrame with pandas > 0.20 causes MultiIndex issue

[sdementen]

During a training session this morning, a call on the laptop of a colleague of the function

@xlwings.func
@xlwings.arg('df', pandas.DataFrame, expand='vertical')
@xlwings.ret(expand="table", index=True)
def my_func(df):
    df.info()
    print(df.columns)

called with a range pointing to the first row of

month_start	month
1/01/2018	Jan-18
1/02/2018	Feb-18
1/03/2018	Mar-18
1/04/2018	Apr-18
1/05/2018	May-18
1/06/2018	Jun-18
1/07/2018	Jul-18
1/08/2018	Aug-18
1/09/2018	Sep-18
1/10/2018	Oct-18
1/11/2018	Nov-18
1/12/2018	Dec-18
1/01/2019	Jan-19

prints the following

<class 'pandas.core.frame.DataFrame'>
DatetimeIndex: 13 entries, 2018-01-01 to 2019-01-01
Data columns (total 1 columns):
(month,) 13 non-null object
dtypes: object(1)
memory usage: 208.0+ bytes
MultiIndex(levels=[['month', 'month_start']],
           labels=[[0]])

The MultiIndex part is weird as on my laptop I get

<class 'pandas.core.frame.DataFrame'>
DatetimeIndex: 13 entries, 2018-01-01 to 2019-01-01
Data columns (total 1 columns):
month    13 non-null object
dtypes: object(1)
memory usage: 208.0+ bytes
Index(['month'], dtype='object')

We were both using xlwings=0.11.4 on a windows 7.0 laptop with Excel 2010.
Any hint of what could give this different behaviours ?

[fzumstein]

different version of pandas?

[sdementen]

it fails on my colleague machine with pandas 0.21.0.
it works on mine with pandas 0.20.3

[sdementen]

indeed, downgrading pandas to 0.20.3 fixes the issues
do I close the issue ?

[fzumstein]

well it should also work with the latest version so leave it open Am 21.12.2017 6:47 vorm. schrieb "sdementen" <notifications@github.com>:

…

indeed, downgrading pandas to 0.20.3 fixes the issues do I close the issue ? — You are receiving this because you commented. Reply to this email directly, view it on GitHub <https://github.com/ZoomerAnalytics/xlwings/issues/786#issuecomment-353306649>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ACKRoOyweKs9sf5h-GcdHvcLkTVy6lTMks5tCiktgaJpZM4RCLKn> .

[pauliacomi]

I had the same problem. Looking at the pandas 0.21.0 changelog, the issue seems to be related to
this particular change in how a single level MultiIndex is treated.

I put an extra check in the pandas_conv.py module and it now works for me. I could submit a PR if you want.

[fzumstein]

sure, PR is always helpful!

[pauliacomi]

Opened #789. Ran tests on both pandas 0.20 and 0.21 and works on both.

[andymcarter]

Hi, ran across this problem a week or so ago and got very confused. Was unknowingly building single level multi-index columns (which look identical to standard indices), then df['col_name'] was returning a dataframe!

Glad to see a fix is on the way for this.