检测 Pandas Dataframe 中的浮点值

Question

I have a dataframe column containing integers, floating numbers and strings. I want to process this column depending on what type of data is present in a particular record.

Now the problem is that, I am able to separate out integer records by Series.str.isnumeric() call, but floating numbers return False here. How can I separate ints & floats together. Here is a basic code:

import numpy as np
import pandas as pd

d = {'A' : ['1234', '12.16', '1234m']}
df= pd.DataFrame(d)
df.A.str.isnumeric()

I get [True False False] as of now, I expect to get [True, True, False].

Answer 1

Use pd.to_numeric with argument errors="coerce" and check which values come out not NaN :

pd.to_numeric(df['A'],errors='coerce').notna()

0     True
1     True
2    False
Name: A, dtype: bool

---

If you want to use str.isnumeric , pandas does not automatically recognizes the . as a decimal, so we have to replace it:

df['A'].str.replace('\.', '').str.isnumeric()

0     True
1     True
2    False
Name: A, dtype: bool

---

If I think ahead and what you want to do, you can write a try except to convert each element to it's type without losing any rows to NaN :

def convert_numeric(x):
    try:
        return pd.to_numeric(x)
    except:
        return x

df['A'].apply(convert_numeric)

0     1234
1    12.16
2    1234m
Name: A, dtype: object

If we then check the types per value, we see it's mixed type now:

df['A'].apply(convert_numeric).apply(type)

0      <class 'numpy.int64'>
1    <class 'numpy.float64'>
2              <class 'str'>
Name: A, dtype: object

Answer 2

def my_func(x):
    try:
        float(x)
    except ValueError:
        return False
    return True

df['A'].apply(my_func)

0     True
1     True
2    False