Remove ends of string entries in pandas DataFrame column

Question

I have a pandas Dataframe with one column a list of files

import pandas as pd
df = pd.read_csv('fname.csv')

df.head()

filename    A    B    C
fn1.txt   2    4    5
fn2.txt   1    2    1
fn3.txt   ....
....

I would like to delete the file extension .txt from each entry in filename. How do I accomplish this?

I tried:

df['filename'] = df['filename'].map(lambda x: str(x)[:-4])

but when I look at the column entries afterwards with df.head(), nothing has changed.

How does one do this?

Answer 1

I think you can use str.replace with regex .txt$' ( $ - matches the end of the string):

import pandas as pd

df = pd.DataFrame({'A': {0: 2, 1: 1}, 
                   'C': {0: 5, 1: 1}, 
                   'B': {0: 4, 1: 2}, 
                   'filename': {0: "txt.txt", 1: "x.txt"}}, 
                columns=['filename','A','B', 'C'])

print df
  filename  A  B  C
0  txt.txt  2  4  5
1    x.txt  1  2  1

df['filename'] = df['filename'].str.replace(r'.txt$', '')
print df
  filename  A  B  C
0      txt  2  4  5
1        x  1  2  1

df['filename'] = df['filename'].map(lambda x: str(x)[:-4])
print df
  filename  A  B  C
0      txt  2  4  5
1        x  1  2  1

df['filename'] = df['filename'].str[:-4]
print df
  filename  A  B  C
0      txt  2  4  5
1        x  1  2  1

EDIT:

rstrip can remove more characters, if the end of strings contains some characters of striped string (in this case ., t, x):

Example:

print df
  filename  A  B  C
0  txt.txt  2  4  5
1    x.txt  1  2  1

df['filename'] = df['filename'].str.rstrip('.txt')

print df
  filename  A  B  C
0           2  4  5
1           1  2  1

Answer 2

You can use str.rstrip to remove the endings:

df['filename'] = df['filename'].str.rstrip('.txt')

should work

Answer 3

You may want:

df['filename'] = df.apply(lambda x: x['filename'][:-4], axis = 1)

Answer 4

use list comprehension

df['filename'] = [x[:-4] for x in df['filename']]

Answer 5

update 2021 + speedtest

Starting from pandas 1.4, the equivalent of str.removesuffix, the pandas.Series.str.removesuffix is implemented, so one can use

df['filename'].str.removesuffix('.txt')

speed test

tl;dr: the fastest is

dat["fname"].map(lambda x: x[:-4] if x[-4:] == ".txt" else x)

In the speed test, I wanted to consider the different methods collected in this SO page. I excluded rstrip, because it would strip other than .txt endings too, and as regexp contains conditional, therefore it would be fair to modify the other functions too so that they remove the last 4 chars only if they are .txt.

The testing code is

import pandas as pd
import time

ITER = 10


def rm_re(dat: pd.DataFrame) -> pd.Series:
    """Use regular expression."""
    return dat["fname"].str.replace(r'.txt$', '', regex=True)


def rm_map(dat: pd.DataFrame) -> pd.Series:
    """Use pandas map, find occurrences and remove with []"""
    where = dat["fname"].str.endswith(".txt")
    dat.loc[where, "fname"] = dat["fname"].map(lambda x: x[:-4])
    return dat["fname"]


def rm_map2(dat: pd.DataFrame) -> pd.Series:
    """Use pandas map with lambda conditional."""
    return dat["fname"].map(lambda x: x[:-4] if x[-4:] == ".txt" else x)


def rm_apply_str_suffix(dat: pd.DataFrame) -> pd.Series:
    """Use str method suffix with pandas apply"""
    return dat["fname"].apply(str.removesuffix, args=(".txt",))


def rm_suffix(dat: pd.DataFrame) -> pd.Series:
    """Use pandas removesuffix from version 1.6"""
    return dat["fname"].str.removesuffix(".txt")


functions = [rm_map2, rm_apply_str_suffix, rm_map, rm_suffix, rm_re]
for base in range(12, 23):
    size = 2**base
    data = pd.DataFrame({"fname": ["fn"+str(i) for i in range(size)]})
    data.update(data.sample(frac=.5)["fname"]+".txt")
    for func in functions:
        diff = 0
        for _ in range(ITER):
            data_copy = data.copy()
            start = time.process_time()
            func(data_copy)
            diff += time.process_time() - start

        print(diff, end="\t")

The output is plotted below:

It can be seen from the plot that the slowest solution is the regexp, and the fastest is the pandas.Series.map with a conditional. In later versions of pandas, this may change and I'd expect an improvement in pandas.Series.str.removesuffix, as it has a greater potential in vectorization.

Pandas had to be installed from the source as of 2021-11-30, because version 1.4 is in the developement stage only. I installed it by following the instructions from pandas dev repo, by cloning the project and installing with python setup.py install.

My machine:

• AMD Ryzen 5 2400G with Radeon Vega Graphics, 3.60 GHz
• Windows 10 20H2
• Python 3.10.0, pandas.version '1.4.0.dev0+1267.gaee662a7e3', numpy.version '1.21.4'