CODEWITHSUNDEEP
pythonpandas
mre
mre

Reputation: 137

TypeError: string indices must be integers when calling pd.to_numeric

I created the following class to convert objects to a numeric type:

class ConvertObjectToNum:
   def convert(dataframe, column):
       dataframe[column] = dataframe[column].apply(lambda x: pd.to_numeric(x, errors = 'coerce'))

When calling the class like this

data_set['Qual'] = ConvertObjectToNum.convert('data_set', 'Qual')

I get the following error:

<ipython-input-20-4d0c321cdb16> in convert(dataframe, column)
     1 class ConvertObjectToNum:
     2     def convert(dataframe, column):
----> 3         dataframe[column] = dataframe[column].apply(lambda x: pd.to_numeric(x, errors = 'coerce'))
     4 
     5 

TypeError: string indices must be integers

However, calling the to_numeric function outside of my class works just fine. So I guess it has to be a problem with how I constructed the class:

data_set['Qual'] = data_set['Qual'].apply(lambda x: pd.to_numeric(x, errors = 'coerce'))

Thanks for your help

E:

The values of the column are numbers in a String format. With some '-' mixed in, hence the Errors = 'coerce' parameter.

Upvotes: 0

Views: 188

Answers (2)

rajah9
rajah9

Reputation: 12339

Here's a solution.

import pandas as pd
class ConvertObjectToNum:
    @staticmethod
    def convert(dataframe, column):
        dataframe[column] = dataframe[column].apply(lambda x: pd.to_numeric(x, errors = 'coerce'))

df = pd.DataFrame({'Weight': [45, 88, 56, 15, 71]})
ConvertObjectToNum.convert(df, 'Weight')
print (df.dtypes) #  Weight    int64

I made the method a staticmethod.

Upvotes: 1

Nand0san
Nand0san

Reputation: 511

Maybe this approach will do the work:

df['col'] = pd.to_numeric(df['col'])

The class:

class ConvertObjectToNum:
   def convert(dataframe, column):
       dataframe[column] = pd.to_numeric(df[column])

Upvotes: 0

Related Questions