Pandas unpivot dataframe using datetime elements from column names - python-3.x

Question

Say I have a pandas dataframe as follows:
Here Store serves as id, Jan 18 - Mar 18 columns represent sales of said stores in respective years and months and Trading Area is an example of time-invariant feature of a store.
For simplicity assume sales column names are already converted to proper datetime format.
Expected result:
I was thinking about using pandas.melt, however I'm not sure how to properly use datetime information contained within column names to construct columns for year and month (obviously this can be done manually in a loop but I need to apply this to arbitrarily large dataframes and this is where it gets tedious, surely a more elegant solution exists).
Any help is appreciated.
Edit: data = pd.DataFrame({'Store':['A', 'B', 'C'], 'Jan 18':[100, 50, 60], 'Feb 18':[120, 70, 80], 'Mar 18':[140, 90, 100], 'Trading Area':[500, 800, 700]})

Answer 1

You could use melt in the following way:
# melt
melted = data.melt(id_vars=['Store', 'Trading Area'], var_name='Month', value_name='Sales')
# extract month and year
melted[['Month', 'Year']] = melted.Month.str.split(expand=True)
# format year
melted['Year'] = pd.to_datetime(melted.Year, yearfirst=True, format='%y').dt.year
print(melted.sort_values('Store'))
Output
Store Trading Area Month Sales Year
0 A 500 Jan 100 2018
3 A 500 Feb 120 2018
6 A 500 Mar 140 2018
1 B 800 Jan 50 2018
4 B 800 Feb 70 2018
7 B 800 Mar 90 2018
2 C 700 Jan 60 2018
5 C 700 Feb 80 2018
8 C 700 Mar 100 2018

Answer 2

You can do a wide_to_long followed by a stack:
(pd.wide_to_long(df=data,
stubnames=['Jan','Feb', 'Mar'],
i=['Store','Trading Area'],
j='Year',
sep=' '
)
.stack()
.reset_index(name='Sales')
.rename(columns={'level_3':'Month'})
)
Output:
Store Trading Area Year Month Sales
0 A 500 18 Jan 100
1 A 500 18 Feb 120
2 A 500 18 Mar 140
3 B 800 18 Jan 50
4 B 800 18 Feb 70
5 B 800 18 Mar 90
6 C 700 18 Jan 60
7 C 700 18 Feb 80
8 C 700 18 Mar 100