Input:
df = pd.DataFrame({
'a':['1',np.nan,np.nan, '2',np.nan,np.nan],
'b':['a',np.nan,'ddd',np.nan,'d','gg'],
'c':[np.nan,'aa','bb',np.nan,'d',np.nan],
})
print (df)
a b c
0 1 a NaN
1 NaN NaN aa
2 NaN ddd bb
3 2 NaN NaN
4 NaN d d
5 NaN gg NaN
Output:
a b c
0 1 a ddd aa bb
1 2 d gg d
If there is non missing value for start of each group use ffill for forward filling missing values and aggregate all values with join and removed missing values:
df = df.groupby(df['a'].ffill()).agg(lambda x: ' '.join(x.dropna())).reset_index(drop=True)
print (df)
a b c
0 1 a ddd aa bb
1 2 d gg d
Detail:
print (df['a'].ffill())
0 1
1 1
2 1
3 2
4 2
5 2
Name: a, dtype: object