TypeError: expected string or bytes-like object pandas variable

Question

I have dataset like this

import pandas as pd
df = pd.DataFrame({'word': ['abs e learning ', 'abs e-learning', 'abs e&learning', 'abs elearning']})

I want to get

      word
0   abs elearning
1   abs elearning
2   abs elearning
3   abs elearning

I do as bellow

re_map = {r'\be learning\b': 'elearning', r'\be-learning\b': 'elearning', r'\be&learning\b': 'elearning'}
import re
for r, map in re_map.items():
            df['word'] = re.sub(r, map, df['word'])

and error

TypeError                                 Traceback (most recent call last)
<ipython-input-42-fbf00d9a0cba> in <module>()
      3 s = df['word']
      4 for r, map in re_map.items():
----> 5             df['word'] = re.sub(r, map, df['word'])

C:\Users\Edward\Anaconda3\lib\re.py in sub(pattern, repl, string, count, flags)
    180     a callable, it's passed the match object and must return
    181     a replacement string to be used."""
--> 182     return _compile(pattern, flags).sub(repl, string, count)
    183 
    184 def subn(pattern, repl, string, count=0, flags=0):

TypeError: expected string or bytes-like object

I can apply str like this

for r, map in re_map.items():
            df['word'] = re.sub(r, map, str(df['word']))

There is no mistake but i cann't get pd.dataFrame as i wish

    word
0   0 0 0 abs elearning \n1 abs elearning\...\n1 0 0 abs elearning \n1 abs elearning\...\n2 0 0 abs elearning \n1 abs ele...
1   0 0 0 abs elearning \n1 abs elearning\...\n1 0 0 abs elearning \n1 abs elearning\...\n2 0 0 abs elearning \n1 abs ele...
2   0 0 0 abs elearning \n1 abs elearning\...\n1 0 0 abs elearning \n1 abs elearning\...\n2 0 0 abs elearning \n1 abs ele...
3   0 0 0 abs elearning \n1 abs elearning\...\n1 0 0 abs elearning \n1 abs elearning\...\n2 0 0 abs elearning \n1 abs ele...

how to improve it?

Answer 1

df['word'] is a list. Converting to string just destroys your list.

You need to apply regex on each member:

for r, map in re_map.items():
    df['word'] = [re.sub(r, map, e) for e in df['word']]:

classical alternate method without list comprehension:

 for r, map in re_map.items():
     d = df['word']
     for i,e in enumerate(d):
         d[i] = re.sub(r, map, e)

BTW you could simplify your regex list drastically:

re_map = {r'\be[\-& ]learning\b': 'elearning'}

By doing that you only have one regex and this becomes a one-liner:

 df['word'] = [re.sub(r'\be[\-& ]learning\b', 'elearning', e) for e in df['word']]:

could even be faster by pre-compiling the regex once for all substitutions:

 theregex = re.compile(r'\be[\-& ]learning\b')
 df['word'] = [theregex.sub('elearning', e) for e in df['word']]: