Add a string prefix to each value in a pandas string column

问题描述：

I would like to prepend a string to the start of each value in a said column of a pandas dataframe. I am currently using:

df.ix[(df['col'] != False), 'col'] = 'str' + df[(df['col'] != False), 'col']

This seems an inelegant method. Do you know any other way (which maybe also adds the character to rows where that column is 0 or NaN)?

As an example, I would like to turn:

    col
1     a
2     0

into:

       col
1     stra
2     str0

解决方案 1：

df['col'] = 'str' + df['col'].astype(str)

Example:

>>> df = pd.DataFrame({'col':['a',0]})
>>> df
  col
0   a
1   0
>>> df['col'] = 'str' + df['col'].astype(str)
>>> df
    col
0  stra
1  str0

解决方案 2：

As an alternative, you can also use an apply combined with format (or better with f-strings) which I find slightly more readable if one e.g. also wants to add a suffix or manipulate the element itself:

df = pd.DataFrame({'col':['a', 0]})

df['col'] = df['col'].apply(lambda x: "{}{}".format('str', x))

which also yields the desired output:

    col
0  stra
1  str0

If you are using Python 3.6+, you can also use f-strings:

df['col'] = df['col'].apply(lambda x: f"str{x}")

yielding the same output.

The f-string version is almost as fast as @RomanPekar's solution (python 3.6.4):

df = pd.DataFrame({'col':['a', 0]*200000})

%timeit df['col'].apply(lambda x: f"str{x}")
117 ms ± 451 µs per loop (mean ± std. dev. of 7 runs, 10 loops each)

%timeit 'str' + df['col'].astype(str)
112 ms ± 1.04 ms per loop (mean ± std. dev. of 7 runs, 10 loops each)

Using format, however, is indeed far slower:

%timeit df['col'].apply(lambda x: "{}{}".format('str', x))
185 ms ± 1.07 ms per loop (mean ± std. dev. of 7 runs, 10 loops each)

解决方案 3：

You can use pandas.Series.map :

df['col'].map('str{}'.format)

In this example, it will apply the word str before all your values.

解决方案 4：

If you load you table file with dtype=str

or convert column type to string df['a'] = df['a'].astype(str)

then you can use such approach:

df['a']= 'col' + df['a'].str[:]

This approach allows prepend, append, and subset string of df.

Works on Pandas v0.23.4, v0.24.1. Don't know about earlier versions.

解决方案 5：

Another solution with .loc:

df = pd.DataFrame({'col': ['a', 0]})
df.loc[df.index, 'col'] = 'string' + df['col'].astype(str)

This is not as quick as solutions above (>1ms per loop slower) but may be useful in case you need conditional change, like:

mask = (df['col'] == 0)
df.loc[mask, 'col'] = 'string' + df['col'].astype(str)

解决方案 6：

Contributing to prefixing columns while controlling NaNs for things like human readable values on csv export.

"_" + df['col1'].replace(np.nan,'').astype(str)

Example:

import sys
import platform
import pandas as pd
import numpy as np

print("python {}".format(platform.python_version(), sys.executable))
print("pandas {}".format(pd.__version__))
print("numpy {}".format(np.__version__))

df = pd.DataFrame({
    'col1':["1a","1b","1c",np.nan],
    'col2':["2a","2b",np.nan,"2d"], 
    'col3':[31,32,33,34],
    'col4':[np.nan,42,43,np.nan]})

df['col1_prefixed'] = "_" + df['col1'].replace(np.nan,'no value').astype(str)
df['col4_prefixed'] = "_" + df['col4'].replace(np.nan,'no value').astype(str)

print(df)

python 3.7.3
pandas 1.2.3
numpy 1.18.5
  col1 col2  col3  col4 col1_prefixed col4_prefixed
0   1a   2a    31   NaN           _1a     _no value
1   1b   2b    32  42.0           _1b         _42.0
2   1c  NaN    33  43.0           _1c         _43.0
3  NaN   2d    34   NaN     _no value     _no value

(Sorry for the verbosity, I found this Q while working on an unrelated column type issue and this is my reproduction code)

解决方案 7：

You can use radd() to element-wise add a string to each value in a column (N.B. make sure to convert the column into a string column using astype() if the column contains mixed types). An example:

df = pd.DataFrame({'col': ['a', 0]})
df['col'] = df['col'].astype('string').radd('str')

which outputs

    col
0  stra
1  str0

It has two advantages over concatenation via +:

1. Null handling: If the column contains NaN values, + simply returns NaN. For example:

df = pd.DataFrame({'col': ['a', float('nan')]})
df['col'] = 'str' + df['col']

which outputs

    col
0  stra
1   NaN

which forces you to handle the NaN later using fillna() etc.

However, with radd(), you can directly pass fill_value= kwarg to handle the NaN values in one function call. For the above example, we can pass fill_value='' to treat NaN values as an empty string, so that when we add the prefix string, we get a column of strings:

df['col'] = df['col'].radd('str', fill_value='')

which outputs

    col
0  stra
1   str

As a side note, there's a difference between using astype(str) and astype('string'); one important difference is related to null handling; you can read more about that here.

1. Method chaining: If you were adding prefixes to strings in a column as part of a method in a pipeline, it might be important to be able to do it using method chaining. + forces you to move out of the pipeline whereas radd clearly shows that the prepending prefixes come after a chain of methods. For example, we can do the following:

df = pd.DataFrame({'col': ['a', 0]})
df.reset_index().astype({'col': 'string'}).radd({'index': 0, 'col': 'str'})

which outputs

  index   col
0     0  stra
1     1  str0