Pandas to_numeric returns object type for string integers-CodePudding

I'm getting this dict as input, I want to make sure that columns that have all numbers are given float / integer types. to do that I'm using pd.to_numeric on this dataFrame:

df = pd.DataFrame({"a": ["1.1", "2"], "b": ["2", "blue"], "c": ["1", "2"]})

then I'm using this func -

def convert_string_columns_to_numeric(df: pd.DataFrame):
    for col in df.columns:
        series_column = pd.Series(df[col])
        df[col] = pd.to_numeric(series_column, errors='ignore')
        return df

I get these dtypes:

["float","object", "object"]

Why isn't "c" of int type?

CodePudding user response：

Its of int type:

df = pd.DataFrame(data = {"a":["1","1.2"],"b":["2","bla"],"c":["2","4"]})
df = df.apply(pd.to_numeric, errors='ignore')
print(df.dtypes)

Output:

a    float64
b     object
c      int64
dtype: object

CodePudding user response：

If you want column c to be of type int, you should not use quotation marks. Why not create the dataframe like this (note quotation marks removed from column a as well):

df = pd.DataFrame({"a":[1, 1.2],"b":["2","bla"], "c":[2, 4]})

Now df.dtypes will give you:

a    float64
b     object
c      int64