CODEWITHSUNDEEP
pythonpandasdataframevectorization
Dan Boucher
Dan Boucher

Reputation: 192

Round off numbers in one column from the number of decimal places shown in another column using Pandas

I want to round numbers in one column from the number of decimal places (not all the same) shown in another column using Pandas.

My data

numbers  decimal
1.2345  2
2.3456  3
3.4567  2

Expected output:

   numbers decimal  newcolA
0   1.2345       2  1.23
1   2.3456       3  2.346
2   3.4567       2  3.46

My code #1

import pandas as pd
df = pd.DataFrame(
   data = {
       'numbers' : [1.2345, 2.3456, 3.4567],
       'decimal'  : [2,3,2]
   }
)
df['newcolA'] = round(df['numbers'] , df['decimal'])

I get the following error: TypeError: cannot convert the series to <class 'int'>

However, the following similar code works

My code #2

import pandas as pd
df = pd.DataFrame(
   data = {
       'numbers' : [1.2345, 2.3456, 3.4567],
       'decimal'  : [2,3,2]
   }
)
df['newcolB'] = df['numbers']*df['decimal'] #The only difference
df


  numbers  decimal  newcolB
0   1.2345       2  2.4690
1   2.3456       3  7.0368
2   3.4567       2  6.9134

What am I not understanding? Why code 2 works, but not the first

Upvotes: 1

Views: 231

Answers (3)

Dan Boucher
Dan Boucher

Reputation: 192

I have adapted the solution of @Corralien so as not to introduce any superfluous 0 (the numbers are all non-integer reals numbers greater than 1, so I believe my noZero function is correct)

import pandas as pd
from math import log10, floor

def noZero(x, rd):               # rd digits at right
   ld = floor(log10(x))+1        # ld digits at left
   x = round(x,rd)
   fmt = "{:<0"  + str(ld + 1 + rd) + "." +  str(rd) + "f}"
   x_str = fmt.format(x)
   return x_str

df = pd.DataFrame(
   data = {
      'numbers' : [1.2345, 2.3456, 3.4567, 3.399],
      'decimal' : [2, 3, 2, 2]
   }
)

df['newcolA'] = df.apply(lambda x: round(x['numbers'], int(x['decimal'])), axis=1)
df['happy :)'] = df.apply(lambda x: noZero(x['newcolA'], int(x['decimal'])), axis=1)
df

   numbers   decimal   newcolA   happy :)
0   1.2345      2       1.230    1.23
1   2.3456      3       2.346    2.346
2   3.4567      2       3.460    3.46
3   3.3990      2       3.400    3.40

Upvotes: 0

Vaishali
Vaishali

Reputation: 38415

Since your question is more about understanding two different behaviors, will focus on that.

Code #1

df['newcolA'] = round(df['numbers'] , df['decimal'])

df['numbers'] & df['decimal'] are of type Series. Effectively, you are passing a Series to round but it expects a number. Hence the error: TypeError: cannot convert the series to <class 'int'>

Code #2

df['numbers']*df['decimal']

Pandas allows various operations between two series of same length using vectorized operations.

Solution There are multiple possible solutions, the most idiomatic would be to use apply (already posted by @Corralien)

Upvotes: 1

Corralien
Corralien

Reputation: 120479

Try:

>>> df.apply(lambda x: round(x['numbers'], int(x['decimal'])), axis=1)
0    1.230
1    2.346
2    3.460
dtype: float64

Upvotes: 2

Related Questions