CODEWITHSUNDEEP
pythonpandasnumpy
Asma Damani
Asma Damani

Reputation: 197

How to subtract a column value with every value in another column (pandas)

I have two columns A and B. I want to subtract column B value with every value in column A and create a new column without using for-loop.

Below is my Dataframe

    A   B
0   5   3
1   3   2
2   8   1

Desired output 

    A   B   C   D   E
0   5   3   2   3   4   
1   3   2   0   1   2
2   8   1   5   6   7

C = A - B[0]
D = A - B[1]
E = A - B[2]

Upvotes: 3

Views: 2768

Answers (3)

kuzand
kuzand

Reputation: 9806

Using numpy's array broadcasting:

df = pd.DataFrame({'A':[5, 3, 8],
                   'B':[3, 2, 1]})

df2 = pd.DataFrame(df['A'].values[:, None] - df['B'].values, columns=['C', 'D', 'E'])

df = df.join(df2)

Result:

   A  B  C  D  E
0  5  3  2  3  4
1  3  2  0  1  2
2  8  1  5  6  7

Explanation:

>>> df['A'].values[:, None]

array([[5],
       [3],
       [8]])

>>> df['B'].values

array([3, 2, 1])

When subtracting them, numpy "stretches" df['A'].values[:, None] to:

array([[5, 5, 5],
       [3, 3, 3],
       [8, 8, 8]])

and df['B'].values to:

array([[3, 2, 1],
       [3, 2, 1],
       [3, 2, 1]])

and the result of subtraction is:

array([[2, 3, 4],
       [0, 1, 2],
       [5, 6, 7]])

Upvotes: 4

Josh Friedlander
Josh Friedlander

Reputation: 11657

Here you go:

d = pd.DataFrame.from_dict({'A': {0: 5, 1: 3, 2: 8}, 'B': {0: 3, 1: 2, 2: 1}})
m = d.shape[0]
cols = [chr(67 + x) for x in range(m)]
d.join(pd.DataFrame(np.broadcast_to(d['A'], (m, m)).T - np.broadcast_to(d['B'], (m, m)), columns=cols))

Explanation: broadcast each series into a 3x3 matrix and subtracts them, make it into a dataframe and join it to original. The columns are auto-generated.

Upvotes: 0

sammywemmy
sammywemmy

Reputation: 28699

This might help:
1. replicate column A, according to the length of the dataframe
2. convert B to a numpy array
3. Subtract B from A, which should get u ur subtraction per row
4. concat back to main data

temp = pd.concat([df.A]*len(df), axis=1).sub(df.B.to_numpy())
final = pd.concat([df,temp], axis=1).set_axis(['A','B','C','D','E'],axis='columns')
final

    A   B   C   D   E
0   5   3   2   3   4
1   3   2   0   1   2
2   8   1   5   6   7

Upvotes: 0

Related Questions