将 pivot 格式的 python df 转换为正确的行列格式

Question

i have the following dataframe

id a_1_1, a_1_2, a_1_3, a_1_4, b_1_1, b_1_2, b_1_3, c_1_1, c_1_2, c_1_3
1    10     20      30     40   90      80     70   Nan     Nan     Nan     
2    33     34      35     36   nan    nan    nan    11     12      13

and i want my result to be as follow

id  col_name 1    2    3
1    a       10   20   30
1    b       90   80   70
2    a       33   34   35
2    c       11   12   13

I am trying to use pd.melt function, but not yielding correct result?

Answer 1

IIUC, you can reshape using an intermediate MultiIndex after extracting the letter and last digit from the original column names:

(df.set_index('id')
   .pipe(lambda d: d.set_axis(pd.MultiIndex.from_frame(
                               d.columns.str.extract(r'^([^_]+).*(\d+)'),
                            names=['col_name', None]
                              ), axis=1))
   .stack('col_name')
   .dropna(axis=1)  # assuming you don't want columns with NaNs
   .reset_index()
)

Variant using janitor 's pivot_longer :

# pip install janitor
import janitor

(df
 .pivot_longer(index='id', names_to=('col name', '.value'),
               names_pattern=r'([^_]+).*(\d+)')
 .pipe(lambda d: d.dropna(thresh=d.shape[1]-2))
 .dropna(axis=1)
)

output:

   id col_name     1     2     3
0   1        a  10.0  20.0  30.0
1   1        b  90.0  80.0  70.0
2   2        a  33.0  34.0  35.0
3   2        c  11.0  12.0  13.0

Answer 2

Code:

df = df1.melt(id_vars=["id"], 
        var_name="Col_name", 
        value_name="Value").dropna()


df['Num'] = df['Col_name'].apply(lambda x: x[-1])
df['Col_name'] = df['Col_name'].apply(lambda x: x[0])

df = df.pivot(index=['id','Col_name'], columns='Num', values='Value').reset_index().dropna(axis=1)
df

Output:

Num id  Col_name    1   2       3
0   1   a          10.0 20.0    30.0
1   1   b          90.0 80.0    70.0
2   2   a          33.0 34.0    35.0
3   2   c          11.0 12.0    13.0