pandas使用concat基于单个列添加和重命名多个列

Question

I have this df:

  group owner  failed granted_pe  slots
0    g1    u1       0     single      1
1   g50   u92       0     shared      8
2   g50   u92       0     shared      1

df can be created using this code:

df = pd.DataFrame([['g1', 'u1', 0, 'single', 1],
                   ['g50', 'u92', '0', 'shared', '8'],
                   ['g50', 'u92', '0', 'shared', '1']], 
                  columns=['group', 'owner', 'failed','granted_pe', 'slots'])
df = (df.astype(dtype={'group':'str', 'owner':'str','failed':'int', 'granted_pe':'str', 'slots':'int'}))
print(df)

Using groupby I create three columns calculated on the "slots" column:

df_calculated = pd.concat([
    df.loc[:,['group', 'slots']].groupby(['group']).sum(),
    df.loc[:,['group', 'slots']].groupby(['group']).mean(),
    df.loc[:,['group', 'slots']].groupby(['group']).max()
    ], axis=1)
print(df_calculated)

       slots  slots  slots
group                     
g1         1    1.0      1
g50        9    4.5      8

Issue 1 : Naming the new columns appropriately
Can I add an argument to concat to name these columns "slots_sum", "slots_avg", and "slots_max"?

Issue 2 : Add columns to df
I would prefer to add the new columns to the df just to the right of the "source" column ("slots" in this case). Desired output would look something like this:

  group owner  failed granted_pe  slots  slots_sum  slots_avg  slots_max
0    g1    u1       0     single      1          1        1.0          1
1   g50   u92       0     shared      8          9        4.5          8
2   g50   u92       0     shared      1  

My actual df is 4.5 mil rows, 23 cols. I will want to do something similar for other columns.

Answer 1

Using agg with add_prefix then merge it back

yourdf=df.merge(df.groupby('group')['slots'].agg(['sum','mean','max']).add_prefix('slots_').reset_index(),how='left')
Out[86]: 
  group owner  failed    ...     slots_sum  slots_mean  slots_max
0    g1    u1       0    ...             1         1.0          1
1   g50   u92       0    ...             9         4.5          8
2   g50   u92       0    ...             9         4.5          8

Answer 2

Another way is to use keys parameter in pd.concat then merge multiindex column headers

df = pd.DataFrame([['g1', 'u1', 0, 'single', 1],
                   ['g50', 'u92', '0', 'shared', '8'],
                   ['g50', 'u92', '0', 'shared', '1']], 
                  columns=['group', 'owner', 'failed','granted_pe', 'slots'])
df = (df.astype(dtype={'group':'str', 'owner':'str','failed':'int', 'granted_pe':'str', 'slots':'int'}))

df_calculated = pd.concat([
    df.loc[:,['group', 'slots']].groupby(['group']).sum(),
    df.loc[:,['group', 'slots']].groupby(['group']).mean(),
    df.loc[:,['group', 'slots']].groupby(['group']).max()
    ], axis=1, keys=['sum','mean','max'])
df_calculated.columns = [f'{j}_{i}' for i,j in df_calculated.columns]
print(df_calculated)

Output:

       slots_sum  slots_mean  slots_max
group                                  
g1             1         1.0          1
g50            9         4.5          8