python - AttributeError: 'str' object has no attribute 'isnumeric'

Question

Slightly confused as I'm positive I've had this working before.

I've created the following method...

def p2f(x):
    if x.strip('%').isnumeric():
        return float(x.strip('%'))/100
    elif x in ['SUPP', 'NEW', 'LOWCOV', 'NA', '']:
        return 0.0
    else:
        return x

but when I run it on my imported CSV file, it produces this error:

AttributeError: 'str' object has no attribute 'isnumeric'

Although I can see that isnumeric is an attribute of str in the documention:

https://pandas.pydata.org/pandas-docs/stable/generated/pandas.Series.str.isnumeric.html?highlight=isnumeric#pandas.Series.str.isnumeric

Unless I'm not interpreting the information correctly?

Answer 1

str.isnumeric() is only available on Python 3. The error indicates you are using Python 2 instead, where only unicode.isnumeric() exists.

You should really use str.isdecimal(), or better yet, use exception handling:

def p2f(x):
    try:
        return float(x.strip('%'))/100
    except ValueError:
        return 0.0 if x in ('SUPP', 'NEW', 'LOWCOV', 'NA', '') else x

.isnumeric() matches 430 Unicode codepoints in the BMP that float() won't accept, and there are codepoints that .isdigit() returns true for that are also not convertible.

You can generate your own table to check with:

for i in range(2 ** 16):
    c = chr(i)
    if c.isnumeric() or c.isdigit() or c.isdecimal():
        try:
            f = float(c)
        except ValueError:
            f = '<not convertible>'
        di, de, nu = ('u2705' if test() else 'u274c' for test in (c.isdigit, c.isdecimal, c.isnumeric))
        print(f'{c!a:<6} {c}digit: {di}   decimal: {de}   numeric: {nu}  float: {f}')

which produces output like:

'0'    0    digit: ?   decimal: ?   numeric: ?  float: 0.0
'1'    1    digit: ?   decimal: ?   numeric: ?  float: 1.0
'2'    2    digit: ?   decimal: ?   numeric: ?  float: 2.0
'3'    3    digit: ?   decimal: ?   numeric: ?  float: 3.0
'4'    4    digit: ?   decimal: ?   numeric: ?  float: 4.0
'5'    5    digit: ?   decimal: ?   numeric: ?  float: 5.0
'6'    6    digit: ?   decimal: ?   numeric: ?  float: 6.0
'7'    7    digit: ?   decimal: ?   numeric: ?  float: 7.0
'8'    8    digit: ?   decimal: ?   numeric: ?  float: 8.0
'9'    9    digit: ?   decimal: ?   numeric: ?  float: 9.0
'xb2' 2    digit: ?   decimal: ?   numeric: ?  float: <not convertible>
'xb3' 3    digit: ?   decimal: ?   numeric: ?  float: <not convertible>
'xb9' 1    digit: ?   decimal: ?   numeric: ?  float: <not convertible>
'xbc' ?    digit: ?   decimal: ?   numeric: ?  float: <not convertible>
'xbd' ?    digit: ?   decimal: ?   numeric: ?  float: <not convertible>
'xbe' ?    digit: ?   decimal: ?   numeric: ?  float: <not convertible>

and you'll find that only the decimal column has crosses for all non-convertible codepoints.

If you want to use isdecimal() in Python 2, you'd have to decode your bytestring to Unicode first.