Python正则表达式-转义

Question

I have a Python regex that takes a string (database connection URI) and splits it using named groups into username, password etc.

uri = 'username:password@host/database'
m = re.compile('^(?P<user>[^:@]+)(\:(?P<password>[^@]*))?@(?P<host>[^\:@/]+)(\:(?P<port>[0-9]+))?/(?P<db>[^\?]+)?$').match(uri)
print m.groupdict()
{'host': 'host', 'password': 'password', 'db': 'database', 'user': 'username', 'port': None}

This works fine. The problem is if the uri has a @ symbol in it, since that's used to split password and host. For example,

uri = 'username:p@ssword@host/database'

will not match, which is expected. However, I'd like to be able to escape the special character, eg:

uri = 'username:p\@ssword@host/database'

and have it match. My regex experience is pretty limited - I guess what I'd like to do is modify the

(?P<password>[^@]*)

group so that it will match any character that's not a @, unless it's preceded by a \\ character. Of course, some (most) connection strings will not contain a \\@ at all.

Any help much appreciated.

Answer 1

You could do:

(?P<password>([^\\@]|\\.)*)

This scans through your string and matches either: a non- \\ or non- @ , OR a backslash in which case it matches whatever follows too. The only way an '@' can be matched by that regex is if it sneaks in through the \\\\. regex, ie it's escaped.

As an aside, to write regex in python, use r"insert_regex_here".

Otherwise for a regex \\\\. , you have to write it in python like "\\\\\\\\." . TO avoid that you can do r"\\\\." .

Answer 2

I would recommend you to use re.split :

>>> print re.split(r"(?<!\\)@|/|:", r"username:password@host/database")
['username', 'password', 'host', 'database']
>>> print re.split(r"(?<!\\)@|/|:", r"username:p\@ssword@host/database")
['username', 'p\\@ssword', 'host', 'database']

Answer 3

My take is you want greedy matching, that is password is up the last @ and hostname is between last @ and first /

A simple way could be like this:

In [68]: re.match('((?P<user>.*):)((?P<pass>.*)@)((?P<host>.*)/)((?P<db>.*))', "username:p@ssword@host/data").groupdict()
Out[68]: {'db': 'data', 'host': 'host', 'pass': 'p@ssword', 'user': 'username'}

You might want to add optionals, that is (stuff)+ if eg username and password can be omitted.