Gruber’s URL Regular Expression in Python

Question

How do I rewrite this new way to recognise addresses to work in Python?

\b(([\w-]+://?|www[.])[^\s()<>]+(?:\([\w\d]+\)|([^[:punct:]\s]|/)))

Why don't you put your code in the question, with examples of what fails — Vinko Vrsalovic, Dec 31 '09 at 17:32
@vanity, the answer with Unicode character properties won't work. See http://stackoverflow.com/questions/1832893/python-regex-matching-unicode-properties — Peter Hansen, Dec 31 '09 at 17:57

Peter Hansen · Accepted Answer · 2010-01-01T12:01:00.677

12

The original source for that states "This pattern should work in most modern regex implementations" and specifically Perl. Python's regex implementation is modern and similar to Perl's but is missing the [:punct:] character class. You can easily build that using this:

>>> import string, re
>>> pat = r'\b(([\w-]+://?|www[.])[^\s()<>]+(?:\([\w\d]+\)|([^%s\s]|/)))'
>>> pat = pat % re.sub(r'([-\\\]])', r'\\\1', string.punctuation)

The re.sub() call escapes certain characters inside the character set as required.

Edit: Using re.escape() works just as well, since it just sticks a backslash in front of everything. That felt crude to me at first, but certainly works fine for this case.

>>> pat = pat % re.escape(string.punctuation)

edited Jan 01 '10 at 12:01

answered Dec 31 '09 at 16:55

Peter Hansen

21,046
5
50
72

This passes all of Gruber's tests, as does pat = pat % re.escape(string.punctuation) – Tobias Dec 31 '09 at 18:04
@vanity, updated to mention that. Note the obvious, that if your data source is Unicode a pure-ASCII solution like string.punctuation may give imperfect results. – Peter Hansen Dec 31 '09 at 18:11
It works with non-ASCII domains and paths. I don't have test data with non-English punctuation. – Tobias Jan 01 '10 at 17:20

YOU · Answer 2 · 2009-12-31T17:08:58.263

5

I don't think python have this expression

[:punct:]

Wikipedia says [:punct:] is same to

[-!\"#$%&\'()*+,./:;<=>?@\\[\\\\]^_`{|}~]

edited Dec 31 '09 at 17:08

answered Dec 31 '09 at 16:48

YOU

120,166
34
186
219

Wikipedia is wrong. It's missing the caret, according to http://www.regular-expressions.info/posixbrackets.html. – Peter Hansen Dec 31 '09 at 17:05
Yeah, Updated my post, Thanks. Somebody updated Wikipedia too. Great! – YOU Dec 31 '09 at 17:09

Vinko Vrsalovic · Answer 3 · 2010-01-01T04:25:35.203

2

Python doesn't have the POSIX bracket expressions.

The [:punct:] bracket expression is equivalent in ASCII to

[!"#$%&'()*+,\-./:;<=>?@[\\\]^_`{|}~]

edited Jan 01 '10 at 04:25

answered Dec 31 '09 at 16:52

Vinko Vrsalovic

330,807
53
334
373

Make sure you use a "raw" string (prefix with `r`) when using that, as the backslash escapes won't be correct otherwise. – Peter Hansen Dec 31 '09 at 17:10
Also note that Python does not support those Unicode character properties: http://stackoverflow.com/questions/1832893 – Peter Hansen Dec 31 '09 at 17:56
Indeed, they compile fine but don't do what you expect – Tobias Dec 31 '09 at 18:00
Python's regex engine is a very strange beast. Fixed the answer. – Vinko Vrsalovic Jan 01 '10 at 04:30

Gruber’s URL Regular Expression in Python

3 Answers3

Linked