Issue 1633953: re.compile("(.*$){1,4}", re.MULTILINE) fails (original) (raw)

[forwarded from http://bugs.debian.org/289603]

Trying to match 1-4 lines of arbitrary content (as part of a larger regex) using the expression (.*$){1,4} and re.MULTILINE. This caused the re module to raise the error "nothing to repeat".

$ python2.5 Python 2.5 (release25-maint, Dec 13 2006, 16:21:45) [GCC 4.1.2 20061212 (prerelease) (Ubuntu 4.1.1-21ubuntu2)] on linux2 Type "help", "copyright", "credits" or "license" for more information.

import re re.compile("(.*$){1,4}", re.MULTILINE) Traceback (most recent call last): File "", line 1, in File "/usr/lib/python2.5/re.py", line 180, in compile return _compile(pattern, flags) File "/usr/lib/python2.5/re.py", line 233, in _compile raise error, v # invalid expression sre_constants.error: nothing to repeat

On first blush, this issue sounds quite similar to issue 2537, but I have been looking at different scenarios and found that there is a subtle difference because, grammatically:

(?m)(?:.$)(.$)

is the same as:

(?m)(.*$){2}

Yet the former compiles while the later raises the exception you list below. Thus, I think the issue YOU raise is indeed related to the redundant repeat operator issue numbered 2537, BUT, when I match an expression with the alternate form, I get an empty string in my capture group, since in a range repeat over a capture group, only the last group is captured, while the entire expression matches only the first line, without the end-line character. Thus, the other thing to remember is that ^ and $ are zero-width matches, so when you write .$, you are saying match up to, but not including, the end of the line. If you immediately follow that with another .$, you will start from the point "up to, but not including, the end of the line", which means the next character is an end of line. Thus, when you reach the second .$, you capture nothing because the . is allowed to be zero-length and you still haven't advanced PAST the end of the line.

As a working alternative, you could write r'(?m)(?:(.$)[\r\n]){1,4}' , since this would give you your 1-4 lines, but also consume the carriage return and line feed characters to get you to the next line.

Since we don't want to change the meaning of $ and ^ to make them capturing (custom POSIX character classes may make 'capturing' a new line character easier), and the 'redundant repeat operator' is already listed as a bug (your expression is essentially saying (.){1,4}$ because it does not capture the new-line character(s) and thus has a redundant repeat operation in the range repeat expression), I'm willing to call this a repeat (technically repeated by as this issue is older) of [issue 2537](issue2537 "[closed] re.compile(r'((x|y+))*') should not fail").