This is a short blogpost. I wanted to document this recipe for my own benefit, and hopefully it will help others. I was working with a very messy dataset with some columns containing non-alphanumeric characters such as #,!,\$^*) and even emojis.

numpy has two methods `isalnum` and `isalpha`.

`isalnum` returns True if all characters are alphanumeric, i.e. letters and numbers. documentation

`isalpha` returns True if all characters are alphabets (only alphabets, no numbers).documentation

```import numpy as np
import pandas as pd
```
```df = pd.DataFrame({'col':['abc', 'a b c', 'a_b_c', '#\$#\$abc', 'abc111', 'abc111#@\$@', '  abc   !!! 123', 'ABC']})
```
```df
```
col
0 abc
1 a b c
2 a_b_c
3 #\$#\$abc
4 abc111
5 abc111#@\$@
6 abc !!! 123
7 ABC

### Remove symbols and return alphanumerics

```def alphanum(element):

return "".join(filter(str.isalnum, element))
```
```df.loc[:,'alphanum'] = [alphanum(x) for x in df.col]
```
```df
```
col alphanum
0 abc abc
1 a b c abc
2 a_b_c abc
3 #\$#\$abc abc
4 abc111 abc111
5 abc111#@\$@ abc111
6 abc !!! 123 abc123
7 ABC ABC

### Remove symbols & numbers and return alphabets only

```def alphabets(element):

return "".join(filter(str.isalpha, element))
```
```df.loc[:,'alphabets'] = [alphabets(x) for x in df.col]
df
```
col alphanum alphabets
0 abc abc abc
1 a b c abc abc
2 a_b_c abc abc
3 #\$#\$abc abc abc
4 abc111 abc111 abc
5 abc111#@\$@ abc111 abc
6 abc !!! 123 abc123 abc
7 ABC ABC ABC

### Bonus: Remove symbols & characters and return numbers only

```def numbers(element):

return "".join(filter(str.isnumeric, element))
```
```df.loc[:,'num'] = [numbers(x) for x in df.col]
df
```
col alphanum alphabets num
0 abc abc abc
1 a b c abc abc
2 a_b_c abc abc
3 #\$#\$abc abc abc
4 abc111 abc111 abc 111
5 abc111#@\$@ abc111 abc 111
6 abc !!! 123 abc123 abc 123
7 ABC ABC ABC
```df.dtypes
```
```col          object
alphanum     object
alphabets    object
num          object
dtype: object```

Note that the `num` column is returned as an `object` (i.e. string) and not a number so be sure to convert it to `int`