Inconsistent sorting with the .sort_values() function

[deanhystad]

I would do the conversion when reading the csv file.

import pandas as pd
import io

data = io.StringIO(
    """
Name, Percent
A ,5%
B, 100%
C, 20%
D ,3%
"""
)

def percent_to_float(valuestr):
    """Convert percent to floats.  Strip trailing %, convert to float and divide by 100"""
    return float(valuestr.replace("%", "")) / 100.0

df = pd.read_csv(data, converters={"Name": str.strip, "Percent": percent_to_float})
print(df.sort_values(by=["Percent"]))

Output:
  Name  Percent
3    D     0.03
0    A     0.05
2    C     0.20
1    B     1.00