Convert all columns to float if certain column are digits python-CodePudding

I want to convert all my columns to float if the certain column are digits.

Here's what I've tried so far:

for columns in range(len(df.columns)):
  for rows in range(len(df)):
    if str(df.iloc[rows,columns]).replace('.','',1).isdigit() == True:
      df.iloc[rows,columns] = float(df.iloc[rows,columns])

It works well, but it takes a long time to run because of the data frame size. Does anyone have any idea for a simpler, much efficient code?

CodePudding user response：

Does this answer your question?

df =  df.apply(lambda i: i.apply(lambda x: float(x) if str(x).replace('.','',1).isdigit() else x))

CodePudding user response：

You can apply pandas.to_numeric to all the columns. By specifying errors='coerce' the non-numeric values will be converted to NaNs. https://pandas.pydata.org/docs/reference/api/pandas.to_numeric.html

Then pass the DataFrame to fillna to fill the NaNs with the original non-numeric values.

This should be faster than the other answer for large DataFrames.

>>> df = pd.DataFrame([["x", 1, "2.1"], [3.2, "y", "5."]], columns=list("ABC"))
>>> df

     A  B    C
0    x  1  2.1
1  3.2  y   5.

>>> df = (
       df.apply(pd.to_numeric, errors='coerce', downcast='float')
         .fillna(df)
)

>>> df

     A  B    C
0    x  1  2.1
1  3.2  y    5

# confirm that the values are floats
>>> type(df.at[0,'C'])
float
>>> type(df.at[1,'C'])
float