En Fri, 14 Dec 2007 06:06:21 -0300, Sean DiZazzo <[EMAIL PROTECTED]> escribió:
> On Dec 14, 12:04 am, Marc 'BlackJack' Rintsch <[EMAIL PROTECTED]> wrote: >> On Thu, 13 Dec 2007 17:49:20 -0800, Sean DiZazzo wrote: >> > I'm wrapping up a command line util that returns xml in Python. The >> > util is flaky, and gives me back poorly formed xml with different >> > problems in different cases. Anyway I'm making progress. I'm not >> > very good at regular expressions though and was wondering if someone >> > could help with initially splitting the tags from the stdout returned >> > from the util. >> >> Flaky XML is often produced by programs that treat XML as ordinary text >> files. If you are starting to parse XML with regular expressions you are >> making the very same mistake. XML may look somewhat simple but >> producing correct XML and parsing it isn't. Sooner or later you stumble >> across something that breaks producing or parsing the "naive" way. >> > It's not really complicated xml so far, just tags with attributes. > Still, using different queries against the program sometimes offers > differing results...a few examples: > > <id 123456 /> > <tag name="foo" /> > <tag2 name="foo" moreattrs="..." /tag2> > <tag3 name="foo" moreattrs="..." tag3/> Ouch... only the second is valid xml. Most tools require at least a well formed document. You may try using BeautifulStoneSoup, included with BeautifulSoup http://crummy.com/software/BeautifulSoup/ > I found something that works, although I couldn't tell you why it > works. :) > retag = re.compile(r'<.+?>', re.DOTALL) > tags = retag.findall(retag) > Why does that work? That means: "look for a less-than sign (<), followed by the shortest sequence of (?) one or more (+) arbitrary characters (.), followed by a greater-than sign (>)" If you never get nested tags, and never have a ">" inside an attribute, that expression *might* work. But please try BeautifulStoneSoup, it uses a lot of heuristics trying to guess the right structure. Doesn't work always, but given your input, there isn't much one can do... -- Gabriel Genellina -- http://mail.python.org/mailman/listinfo/python-list
