CODEWITHSUNDEEP
python-3.xpandasdataframe
David
David

Reputation: 500

Convert floats to ints of a column with numbers and nans

I'm working with Python 3.6 and Pandas 1.0.3.

I would like to convert the floats from column "A" to int... This column has some nan values.

So i followed this post with the solution of @jezrael.

But I get the following error: "TypeError: cannot safely cast non-equivalent float64 to int64"

This is my code

import pandas as pd
import numpy as np

data = {'timestamp': [1588757760.0000, 1588757760.0161, 1588757764.7339, 1588757764.9234], 'A':[9087.6000, 9135.8000, np.nan, 9102.1000], 'B':[0.1648, 0.1649, '', 5.3379], 'C':['b', 'a', '', 'a']}
df = pd.DataFrame(data)
df['A'] = pd.to_numeric(df['A'], errors='coerce').astype('Int64')
print(df)

Did I miss something?

Upvotes: 3

Views: 1641

Answers (2)

MarianD
MarianD

Reputation: 14131

Your problem is that you have true float numbers, not integers in the float form. So for safety reasons pandas will not convert them, because you would be obtained other values.

So you need first explicitely round them to integers, and only then use the.astype() method:

df['A'] = pd.to_numeric(df['A'].round(), errors='coerce').astype('Int64')

Test:

print(df)

      timestamp     A       B  C
0  1.588758e+09  9088  0.1648  b
1  1.588758e+09  9136  0.1649  a
2  1.588758e+09   NaN           
3  1.588758e+09  9102  5.3379  a

Upvotes: 4

NYC Coder
NYC Coder

Reputation: 7594

One way to do it is to convert NaN to a integer:

df['A'] = df['A'].fillna(99999999).astype(np.int64, errors='ignore')
df['A'] = df['A'].replace(99999999, np.nan)
df

    timestamp   A   B   C
0   1.588758e+09    9087    0.1648  b
1   1.588758e+09    9135    0.1649  a
2   1.588758e+09    NaN     
3   1.588758e+09    9102    5.3379  a

Upvotes: 0

Related Questions