CODEWITHSUNDEEP
pythonpandasnumpy
Jesh Kundem
Jesh Kundem

Reputation: 976

Subtract each element in a data frame column, by a given number - python

I am looking to subtract each element in a data frame column by a specific number from that column. I am presently doing that by converting each column to a numpy array and it is not ideal.

As an example,

data = [[1, 10], [2, 20], [3, 30],[4, 40],[5, 50]]
# Existing dataframe
df = pd.DataFrame(data, columns=['column1', 'column2']) 

a = np.array([2,4]) # this is an array for the index of elements. 2 is for column 1, 4 is for column 2. 

# In column 1 with index=2, find the element, and subtract that from all the elements in column 1. 
#Similarly with column 2, with index = 4, find the element, and subtract that from all the elements in column 2



# Required Output dataframe
data2 = [[-2, -40], [-1, -30], [0, -20],[1, -10],[2, 0]]
df2 = pd.DataFrame(data2, columns=['column1', 'column2'])

Output

Existing data frame: 
   column1  column2
0        1       10
1        2       20
2        3       30
3        4       40
4        5       50

Required Output data fram
 column1  column2
0       -2      -40
1       -1      -30
2        0      -20
3        1      -10
4        2        0

Upvotes: 0

Views: 488

Answers (1)

Henry Ecker
Henry Ecker

Reputation: 35686

We can use numpy indexing to select the values from the DataFrame by converting DataFrame.to_numpy, then subtracting:

output = df - df.to_numpy()[a, np.arange(df.columns.size)]

Or with DataFrame.sub:

output = df.sub(df.to_numpy()[a, np.arange(df.columns.size)], axis='columns')

output:

   column1  column2
0       -2      -40
1       -1      -30
2        0      -20
3        1      -10
4        2        0

Values are selected using the row indexes in a:

a = np.array([2, 4])
# [2, 4]

A RangeIndex is created for the length of columns using np.arange and the Index.size:

col_index = np.arange(df.columns.size)
# [0 1]

These indices can be used together to select values from the DataFrame:

df.to_numpy()[a, col_index]
# [ 3 50]

Upvotes: 1

Related Questions