XII-IP : Transform, applymap and apply in Pandas

Transform

Takes one function that is expected to be applied to a column and return a column of equal size. Pandas DataFrame.transform() function call func on self producing a DataFrame with transformed values and that has the same axis length as self.

import pandas as pd

d1 = {'rollno':[101,101,103,102,104], 'maths': [80,70,80,60,70],\

      'physics':[90,40,50,90,65],'chem':[75,80,60,85,60] }

df = pd.DataFrame(d1)

display(df)

r = df.transform(func=lambda x:x+10)

print('--------------Transform----------------')

display(r)

  rollno maths physics chem

0 101 80 90 75

1 101 70 40 80

2 103 80 50 60

3 102 60 90 85

4 104 70 65 60

--------------Transform----------------

  rollno maths physics chem

0 111 90 100 85

1 111 80 50 90

2 113 90 60 70

3 112 70 100 95

4 114 80 75 70

Applymap

This takes a function and returns a new dataframe with the results of that function being applied to the value in each cell and replacing the value of the cell with the result.

import numpy as np

s = df.applymap(np.sqrt)

print('--------------applymap----------------')

display(s)

--------------applymap----------------

  rollno     maths    physics    chem

0 10.049876 8.944272 9.486833 8.660254

1 10.049876 8.366600 6.324555 8.944272

2 10.148892 8.944272 7.071068 7.745967

3 10.099505 7.745967 9.486833 9.219544

4 10.198039 8.366600 8.062258 7.745967

Apply

Takes a function and applies it either row-wise or column-wise depending on the value of axis parameter. Apply may give the same result as transfom() or applymap() with parameter set to non-aggregate function.

print('----------apply behaving as applymap-----------')

x = df.apply(np.sqrt)

display(x)

print('----------apply with axis =0-----------')

t = df.apply(np.mean)

display(t)

print('----------apply with axis =1-----------')

t = df.apply(np.mean, axis=1)

display(t)

output:

----------apply behaving as applymap-----------

     rollno maths     physics chem

0 10.049876 8.944272 9.486833 8.660254

1 10.049876 8.366600 6.324555 8.944272

2 10.148892 8.944272 7.071068 7.745967

3 10.099505 7.745967 9.486833 9.219544

4 10.198039 8.366600 8.062258 7.745967

----------apply with axis =0-----------

rollno     102.2

maths       72.0

physics     67.0

chem        72.0

dtype: float64

----------apply with axis =1-----------

0    86.50

1    72.75

2    73.25

3    84.25

4    74.75

dtype: float64

Another example of applymap and transform

d=df.transform(func=lambda x:x+10)

display(d)

e=df.applymap(lambda x:x+10)

display(e)

output:

  rollno maths physics chem

0 111 90 100 85

1 111 80 50 90

2 113 90 60 70

3 112 70 100 95

4 114 80 75 70

  rollno maths physics chem

0 111 90 100 85

1 111 80 50 90

2 113 90 60 70

3 112 70 100 95

4 114 80 75 70