ValueError raised on specific input

[elgehelge]

Not sure if you regard this as a bug or not.

When given input that:

1. evaluates as a digit (input_string.isdigit())
2. cannot be type casted to int (int(input_string))

...a ValueError is raised instead of a stdnum.exceptions.ValidationError.

Example:

from stdnum import isbn
isbn.validate('978-9024538²70')  # notice the superscript of 2 ('²')

[arthurdejong]

Thanks for reporting this. Yes I consider this a bug.

It turns out that str.isdigit() returns True for certain Unicode code points that represent numbers. It also turns out that it also returns True for non-ASCII digits that int() can handle such as int('᭓') == 3 which is not what I would expect for most numbers.

I'm looking into a nice way to solve this so that this raises a ValidationError:

isbn.validate('978-90245᭓ 8270')

[elgehelge]

Great job. Looking forward to next release! 👏

[elgehelge]

@arthurdejong Instead of the regex you might like this:

def isdigits(input):
    return str.isdigit(input) and str.isascii(input)

my guess is that it is faster, but I did not test

[arthurdejong]

Thanks. I hadn't really considered that. I did some performance tests and it turns out that number.isdigit() and number.isascii() is the fastest that seems to have the expected result. The biggest problem is that that function is not present on Python 2 :( so I'll stick with the regular expression solution for now.

code	relative performance
number.isdigit()	1
all(x in '0123456789' for x in number)	6.3
all(x in digits_set for x in number)	6.3
bool(re.match(r'^[0-9]+$', number))	7.7
bool(digits_re.match(number))	3.4
number.isdigit() and number.isascii()	1.2
str.isdigit(number) and str.isascii(number)	1.6