CODEWITHSUNDEEP
pythonpandas
john
john

Reputation: 129

type conversion in python from float to int

I am trying to change data_df which is type float64 to int.

data_df['grade'] = data_df['grade'].astype(int)

I get the following error.

invalid literal for int() with base 10: '17.44'

Upvotes: 9

Views: 14625

Answers (4)

Ruperto
Ruperto

Reputation: 338

From:

data_df['grade'] = data_df['grade'].astype(int)

Need to change int into 'int' 

data_df['grade'] = data_df['grade'].astype('int')

Upvotes: 2

befpy
befpy

Reputation: 41

I found this to work for me where none of the other earlier answers did the job for me:

data_df['grade'] = data_df['grade'].apply(np.int)

Upvotes: 1

jezrael
jezrael

Reputation: 863741

I think you need to_numeric first because float cannot be cast to int:

data_df['grade'] = pd.to_numeric(data_df['grade']).astype(int)

Another solution is first cast to float and then to int:

data_df['grade'] = data_df['grade'].astype(float).astype(int)

Sample:

data_df = pd.DataFrame({'grade':['10','20','17.44']})
print (data_df)
   grade
0     10
1     20
2  17.44

data_df['grade'] = pd.to_numeric(data_df['grade']).astype(int)
print (data_df)
   grade
0     10
1     20
2     17

data_df['grade'] = data_df['grade'].astype(float).astype(int)
print (data_df)
   grade
0     10
1     20
2     17

---

If some values cannot be converted and after to_numeric get error:

ValueError: Unable to parse string

is possible add parameter errors='coerce' for convert non numeric to NaN.

If NaN values then cast to int is not possible see docs:

data_df = pd.DataFrame({'grade':['10','20','17.44', 'aa']})
print (data_df)
   grade
0     10
1     20
2  17.44
3     aa

data_df['grade'] = pd.to_numeric(data_df['grade'], errors='coerce')
print (data_df)
   grade
0  10.00
1  20.00
2  17.44
3    NaN

If want change NaN to some numeric e.g. 0 use fillna:

data_df['grade'] = pd.to_numeric(data_df['grade'], errors='coerce')
                     .fillna(0)
                     .astype(int)
print (data_df)
   grade
0     10
1     20
2     17
3      0

Small advice:

Before using errors='coerce' check all rows where is impossible casting to numeric by boolean indexing:

print (data_df[pd.to_numeric(data_df['grade'], errors='coerce').isnull()])
  grade
3    aa

Upvotes: 11

zabeltech
zabeltech

Reputation: 971

what works is data_df['grade'] = int(pd.to_numeric(data_df['grade'])) The method as_type(int) throws and error because it want's to tell you, that no exact conversion from float to integer is possible and you will lose information. My solution will truncate the integer (i.e. 1.9 will become 1), so you might want to specifiy in your question wether you want to convert float to integer by truncation or by rounding (i.e. 1.9 will become 2)

Upvotes: 3

Related Questions