Value	Description
`"integer"`	Convert type to `np.int8`.
`"signed"`	Convert type to `np.int8`. Same as above.
`"unsigned"`	Convert type to `np.uint8`.
`"float"`	Convert type to `np.float32`.
`None`	Do not perform any downcasting.

Note that downcasting is performed after the main numeric conversion, and so if there are parsing issues during downcasting, then an error will be raised regardless of what you specified for errors.

By default, downcast=None.

Return Value

If arg is a Series, then a new Series is returned. Otherwise, a new Numpy array is returned.

Examples

Basic usage

To convert the type of all values in a Series to numeric type:


        
        
            
                
                
                    s = pd.Series(["1.","2.0",3])
pd.to_numeric(s)
                
            
            0    1.0
1    2.0
2    3.0
dtype: float64

Note that the source Series s is left intact and a new Series is returned here.

Error handling

By default, errors="raise", which means that when a problem occurs while converting to numeric type, an error will be thrown:


        
        
            
                
                
                    s = pd.Series(["2", "2.3.4"])
pd.to_numeric(s)
                
            
            ValueError: Unable to parse string "2.3.4" at position 1

Instead of throwing an error, we can convert invalid values into NaN, by specifying errors="coerce" like so:


        
        
            
                
                
                    s = pd.Series(["2", "2.3.4"])
pd.to_numeric(s, errors="coerce")
                
            
            0    2.0
1    NaN
dtype: float64

We could also leave invalid values intact by using errors="ignore":


        
        
            
                
                
                    s = pd.Series(["2", "2.3.4"])
pd.to_numeric(s, errors="ignore")
                
            
            0        2
1    2.3.4
dtype: object

Notice how the dtype of the Series is object. This is because a Series that contains even one non-numeric type ("2.3.4" in this case) must be upcast to a more general type, that is, object.

Downcasting

By default, numerics are converted to either int64 or float64:


        
        
            
                
                
                    s = pd.Series(["1.", "2.0", 3])
pd.to_numeric(s)
                
            
            0    1.0
1    2.0
2    3.0
dtype: float64

Here, float64 is used since "2.0" is converted to a float instead of int under the hood.

We can convert this to float32 instead by passing in downcast="float" like so:


        
        
            
                
                
                    s = pd.Series(["1.", "2.0", 3])
pd.to_numeric(s, downcast="float")
                
            
            0    1.0
1    2.0
2    3.0
dtype: float32

In this case, since 2.0 can be represented as an int as well, we can also pass downcast="integer" to convert the values into type int8:


        
        
            
                
                
                    s = pd.Series(["1.", "2.0", 3])
pd.to_numeric(s, downcast="integer")
                
            
            0    1
1    2
2    3
dtype: int8

Published by Isshin Inada

Edited by 0 others

Did you find this page useful?

thumb_up

thumb_down

Comment

Citation

Ask a question or leave a feedback...

Official Pandas Documentation

https://pandas.pydata.org/pandas-docs/stable/reference/api/pandas.to_numeric.html

thumb_up

thumb_down

chat_bubble_outline

settings

Enjoy our search

Hit / to insta-search docs and recipes!