pd. to_ numeric

effect

effect ： Convert parameter to numeric type .

Default return dtype by float64 or int64, Depending on the data provided . Use downcast Parameter get other dtype.

Parameters to describe

Parameters describe
args Accept scalar, list, tuple, 1-d array, or Series type
errors Yes 3 Types {‘ignore’, ‘raise’, ‘coerce’}, The default is ‘raise’
downcast {‘integer’, ‘signed’, ‘unsigned’, ‘float’} , default None, Default return float64 or int64
Be careful downcast It means down conversion

errors Explanation of parameters in

’raise’ Parameters ： Invalid parsing will throw an exception

’corece’ Parameters ： Set invalid resolution to NaN

‘ignore’ Parameters ：** Invalid parsing will return input

downcast The meaning of parameters in

default None Just don't deal with it

’integer’ and ’signed’： The smallest signed integer dtype（ minimum value np.int8）

’unsigned’： The smallest unsigned int dtype（np.uint8）

’float’： The smallest float dtype（np.float32）

Return value ： If the parsing is successful , It's numbers . The return type depends on the input . If Series, Then for Series, Otherwise ndarray.

example

import pandas as pd
import numpy as np
s = pd.Series(['apple', '1.0', '2','2019-01-02',1, False,None,pd.Timestamp('2018-01-05')])

# to_numeric Is in object, Do conversion in the middle of time format , And then use astype do numeric Internal conversion of type 
pd.to_numeric(s, errors='raise') #  An error is reported when a non numeric string type is encountered ,bool Type error , The time type is converted to int
pd.to_numeric(s, errors='ignore') #  Convert only numeric strings , Other types are not converted , Include time type 
pd.to_numeric(s, errors='coerce')  #  Combine the time string with bool Type to number , Others are converted to NaN

# downcast  It can be further transformed into int perhaps float
pd.to_numeric(s) #  Default float64 type 
pd.to_numeric(s, downcast='signed') #  Convert to integer 

# astype Medium error No, `coerce` Options , So it's only suitable for `numeric` Conversion of internal types , For example, will int32 Convert to int64,int32 Convert to float32
#  Not suitable for object, Convert between time formats ,
s.astype('int32',errors='raise')
s.astype('int32',errors='ignore')  #  Yes object Invalid ,astype Only right numeric Type validation