Python Pandas create a new dataframe from all row combinations of existing dataframes

Question

I have 2 input dataframes ( df1 and df2 ), with identical structure, and I want to create a 3rd one ( output_df ), with all row combinations of the input dataframes.

df1 = pd.DataFrame([["John","18","a"],["Jane","19","b"],["Jim","20","c"]],columns=['Name','Age','Function'])
df2 = pd.DataFrame([["Don","21","d"],["Diana","22","e"],["Dave","23","f"]],columns=['Name','Age','Function'])

output_df=pd.DataFrame([["John_Don","18_21","a_d"],
["John_Diana","18_22","a_e"],
["John_Dave","18_23","a_f"],
["Jane_Don","19_21","b_d"],
["Jane_Diana","19_22","b_e"],
["Jane_Dave","19_23","b_f"],
["Jim_Don","20_21","c_d"],
["Jim_Diana","20_22","c_e"],
["Jim_Dave","20_23","c_f"]],columns=['Name','Age','Function'])

The new dataframe would have the sum ("+") of the corresponding columns of the initial dataframe. (I am aware strings get concatenated - that is what I am after if inputs are strings)

The below code creates the output_df , but it is empty and the code is taking too long to run . The below sample code only runs for 2x10 records as input. Eventually, I will be dealing with thousands of records as input from each dataframe.

Q1: what am I missing when populating the output dataframe?

Q2: how can I make my code more efficient?

output_df=pandas.DataFrame(columns=['Name','Age','Function'])
i=0
for lendf1 in range (10):
    for lendf2 in range(10):
        output_df=output_df.append(pandas.Series(),ignore_index=True)
        i=i+1
        for column in output_df:
            output_df[column][i]=df1[column][lendf1:lendf1+1]+df2[column][lendf2:lendf2+1]

Answer 1

I believe you are looking for this:

first = pd.Series(['a', 'b', 'c', 'd', 'e'])
second = pd.Series(['f', 'g', 'h', 'i', 'j'])
pd.DataFrame(np.add.outer(first, second))

Output:

    0   1   2   3   4
0  af  ag  ah  ai  aj
1  bf  bg  bh  bi  bj
2  cf  cg  ch  ci  cj
3  df  dg  dh  di  dj
4  ef  eg  eh  ei  ej

Note that the input should be of type pd.Series and not dataframes.

Answer 2

I think you are trying to concatenate both the dataframe's columns. Please try the following code works for you.

import pandas as pd

df1 = pd.DataFrame([["John","18","a"],["Jane","19","b"],["Jim","20","c"]],columns=['Name','Age','Function'])
df2 = pd.DataFrame([["Don","21","d"],["Diana","22","e"],["Dave","23","f"]],columns=['Name','Age','Function'])

cols = list(df1)

out_list = []
for ind1, row1 in df1.iterrows():
    for ind2, row2 in df2.iterrows():
        in_list = []
        for i in range(0, len(cols)):
            in_list.append(row1[cols[i]] + '_' + row2[cols[i]])
        out_list.append(in_list)

outdf = pd.DataFrame(out_list, columns=cols)
print outdf