How to get a column with partial sum (grouped by two columns) in pandas data frame

Question

I have a table with the following columns:

    date       Product  category     units     xx_col       ..
0   2017-02-11  Books   heavy       2.          11.     
1   2017-02-11  Books   medium      1.          22.
2   2017-02-11  Books   light       1.          11.
3   2017-02-11  DVD     heavy       3.          11.
4   2017-02-11  DVD     medium      2.          4170.775    
5   2017-02-11  DVD     light       2.          4170.775    

Now, I want to create another two columns with a sum of units for a given product (all three categories) for every date, and then the fraction. as shown below. There are also other columns xx_col, which I want to preserve in the table ( I do not mention it in the second table, but it should be there).

    date       Product  category     units     unit_tot   unit_frac ..  
0   2017-02-11  Books   heavy       2.          4.        .5
1   2017-02-11  Books   medium      1.          4         .25
2   2017-02-11  Books   light       1.          4         .25
3   2017-02-11  DVD     heavy       3.          10        .3
4   2017-02-11  DVD     medium      2.          10        .2
5   2017-02-11  DVD     light       5.          10        .5

unit_tot: total_units- sum for a given product for a given date over the three (h, m, l) categories (2+1+1=4).
unit_frac: unit/ unit_tot I want both of these not just the unit_tot or unit_frac, and I want to keep this table as is. I could get two different pivot_tables, but I want to see if there is an easy way to add these two columns.

Answer 1

IIUC, just need to transform

df['unit_tot'] = df.groupby(['date', 'Product']).units.transform('sum')

Then divide both columns to get the fraction

df['unit_frac'] = df['units']/df['unit_tot']