列切片熊猫

Question

Here is the dummy DataFrame:

d = {'col_1': [1, 2], 'col_n_1': [3, 4], 'col_2': [2, 1], 'col_n_2': [6, 3]}
df = pd.DataFrame(data=d)


   col_1    col_2   col_n_1   col_n_2
0      1        2         3         6
1      2        1         4         3
2      1        1         4         5

I am looking for a nice way to extract the values from col_n_1 where col_1 == 1 and col_n_2 where col_2 == 1 in a new column that would look like:

new_col
      3
      3
    4,5

Answer 1

Use where for get values by mask and then join columns together:

L = ['col_1','col_2']
L1 = ['col_n_1','col_n_2']
df['new'] = (df[L1].astype(str).where(df[L].eq(1).values, axis=1)
                  .apply(lambda x: ','.join(x.dropna()), 1))

Solution if only 2 columns:

L = ['col_1','col_2']
L1 = ['col_n_1','col_n_2']
df1 = df[L1].astype(str).where(df[L].eq(1).values, axis=1)
df['new'] = (df1['col_n_1'] .fillna('') + ',' + df1['col_n_2'] .fillna('')).str.strip(',')

Or solution with add , and then sum , last remove trailing , :

df['new'] = (df[L1].astype(str).where(df[L].eq(1).values)
                  .add(', ')
                  .fillna('')
                  .sum(axis=1)
                  .str.strip(', '))

---

print (df)
   col_1  col_2  col_n_1  col_n_2  new
0      1      2        3        6    3
1      2      1        4        3    3
2      1      1        4        5  4,5

Answer 2

Borrow the name list from Jez

df[L].eq(1).rename(columns=dict(zip(L,L1))).mul((df[L1].astype(str)+',')).sum(1).str[:-1]
Out[126]: 
0      3
1      3
2    4,5
dtype: object

Answer 3

This can be accomplished with the apply() method and a lambda function. apply() with the index parameter set to 1 will call a given function on each row of the dataframe. So the only trouble is writing that function -- I think the best solution is to create a list containing either the row's col_n_1 or col_n_2 , both, or neither, then joining the list with commas. Like this:

df['new'] = df.apply(lambda row: ','.join([str(row.col_n_1)] if row.col_1 == 1 else [] + [str(row.col_n_2)] if row.col_2 == 1 else []), axis = 1)