I have code in Pandas Python like below:

df (both columns are string - "object" data type):

code:

nominal_variables = ['col1', 'col2']
list_of_numbers = ['0', '1', '2', '3', '4', '5', '6', '7', '8', '9']
variables_with_numbers = list()

for col in nominal_variables:
    for el in list_of_numbers:
        if df[col].str.contains(f'{el}').any():
           variables_with_numbers.append(col)

I need to fill my list "variables_with_numbers" by columns wich contain numbers (string).

1. First of all I make a loop by all kolumns in my df
2. Secondly I make a loop by list of numbers in string which I will search in values in columns from the first loop
3. Thirtly I am looking for in columns from the tirst loop elements from the second loop and if it is True I add this col to my list

Nevertheless using my code I have Error: AttributeError: Can only use .str accessor with string values! but I do not know why, because all columns in my DF are "object".

So as a result I need: variables_with_numbers = ['col1']

How can I modify my code in Python Pandas ?

CodePudding user response：

IIUC, you want to check which columns of your data frame contain at least one element from a defined list e.g. list_of_numbers. If that is true, you can start joining your list by | and check columns row-wise if one value contains one element from your list. This results in a boolean mask that can be used to query your data frame. Then, just drop nan values column-wisely (if they only contain nan values) and call the columns argument.

Code:

import pandas as pd

df = pd.DataFrame({
    "col1": ["123445", "653144", "aaaBc"],
    "col2": ["one", "bbt", "tro"]
})

list_of_numbers = ['0', '1', '2', '3', '4', '5', '6', '7', '8', '9']
list(df[df.apply(lambda x: x.str.contains("|".join(list_of_numbers)), axis=1)].dropna(axis=1, how="all").columns)

# Or this (equivalent to the line above):
# list(df.columns[df.apply(lambda x: x.str.contains("|".join(list_of_numbers)), axis=1).any()])

# Result: ['col1']

You can also produce a series with booleans for columns that contain at least one value from your list:

df.apply(lambda x: x.str.contains("|".join(list_of_numbers)), axis=1).any()

--------------
col1    True
col2    False
dtype: bool
--------------

If you want to check which columns contain numeric values, you can simple use this approach:

list(df[df.apply(lambda x: x.str.isnumeric(), axis=1)].dropna(axis=1, how="all").columns)