I recently answered a question from Harsha Nalore on StackOverflow that involved using Ansible to extract the output of a command sent to a BigIP device of some sort. My solution – which I claim to be functional, but probably not optimal – involved writing an Ansible filter module to parse the output. That filter made use of a complex-looking regular expression. Harsha asked for some details on that regular expression works, and the existing StackOverflow answer didn’t really seem the write place for that: so, here we are.

The output in question looks like this:

gtm wideip a wideip {
    description wideip
    pool-lb-mode topology
    pools {
        test1-pool {
            order 1
        }
        test2-pool {
            order 0
        }
    }
}

The goal is to return a list of pool names. You can see the complete solution in my answer; for the purposes of this post we’re interesting in the following two regular expressions:

re_pools = re.compile('''
gtm \s+ wideip \s+ a \s+ (\S+) \s+ { \s+
(?P<parameters>(\S+ \s+ \S+ \s+)*)
pools \s+ { \s+ (?P<pools>
(?:
\S+ \s+ {  \s+
[^}]* \s+
} \s+
)+ \s+
)
}
''', flags=re.VERBOSE)

re_pool = re.compile('''
(\S+) \s+ { \s+ [^}]* \s+ } \s+
''', flags=re.VERBOSE)

VERBOSE mode

The first thing to note is that I’m using VERBOSE syntax for both of these expressions. That means that whitespace must be included explicitly in the expression. That’s what all of those \s+ markers are – that means “any white space character, one or more times”. For example, consider the following simple expression:

>>> re.match('this is a test', 'this is a test')

The pattern matches the string just fine. But if we were to enable the VERBOSE flag, the pattern would no longer match:

>>> re.match('this is a test', 'this is a test', flags=re.VERBOSE)

We would instead need to write it like this:

>>> re.match('this \s is \s a \s test', 'this is a test', flags=re.VERBOSE)

The advantage to VERBOSE mode is that you can split your regular expression across multiple lines for legibility:

>>> re.match('''
... this \s
... is \s
... a \s
... test''', 'this is a test', flags=re.VERBOSE)

Capture groups

In order to make it easier to extract information from the results of a match, I’m using named capture groups. A “capture group” is a part of the expression inside parentheses that can be extracted from the resulting match object. Unnamed groups can be extracted using their index. If we wanted to match the phrase this is a <noun>, rather than this is a test, and we wanted to extract the noun, we might write something like this:

>>> re_example = re.compile('this is a (\S+)')
>>> match = re_example.match('this is a frog')
>>> match.groups()
('frog',)
>>> match.group(1)
'frog'

The expression (\S+) is a capture group that will match any string of non-whitespace characters. This works fine for a simple expression, but keeping the index straight in a complex expression can be difficult. This is where named capture groups become useful. We could rewrite the above like this:

>>> re_example = re.compile('this is a (?P<noun>\S+)')
>>> match = re_example.match('this is a frog')
>>> match.groupdict()
{'noun': 'frog'}
>>> match.group('noun')
'frog'

Non-capture groups

Sometimes, you want to group part of a regular expression in a way that does not result in another capture group. This is what the (?: ...) expression is for. For example, we we were to write:

>>> re_example = re.compile('this (?:is|was) a (?P<noun>\S+)')

Then we could match the phrase this is a test or this was a test, but we would still only have a single capture group:

>>> match = re_example.match('this is a test')
>>> match.groupdict()
{'noun': 'test'}

Putting it all together

With all that in mind, let’s take a look at the regular expression in my answer:

re_pools = re.compile('''
gtm \s+ wideip \s+ a \s+ (\S+) \s+ { \s+
(?P<parameters>(\S+ \s+ \S+ \s+)*)
pools \s+ { \s+ (?P<pools>
(?:
\S+ \s+ {  \s+
[^}]* \s+
} \s+
)+ \s+
)
}
''', flags=re.VERBOSE)

The first line matches gtm wideip a <something> {:

gtm \s+ wideip \s+ a \s+ (\S+) \s+ { \s+

Next, we match the <key> <value> part of the output, which looks like this:

description wideip
pool-lb-mode topology

With this expression:

 (?P<parameters>(\S+ \s+ \S+ \s+)*)

That is a named capture group (“parameters”) that matches the expression (\S+ \s+ \S+ \s+) zero or more times (*). Since \S+ means “a string of non-whitespace characters” and \s+ means “a string of whitespace characters”, this correctly matches that part of the output.

Next, we match the entire pools {...} part of the output with this expression:

pools \s+ { \s+ (?P<pools>
(?:
\S+ \s+ {  \s+
[^}]* \s+
} \s+
)+ \s+
)

That creates a named capture group (“pools”) that looks for one or more occurrences of the pattern:

\S+ \s+ {  \s+
[^}]* \s+
} \s+

The first line will match a string like test1-pool1 {. The next line matches any sequence of characters that are not }, so that gathers up everthing between test1-pool { and the closing }. Because we have the entire thing wrapped in (?: ...)+, we are looking for one or more matches of that sub-expression, which gathers up all of the pool definitions.

Finally we match the closing brace:

}

When that expression matches, we end up with a match object that has a pools match group that will look like this:

>>> print(match.group('pools'))
test1-pool {
            order 1
        }
        test2-pool {
            order 0
        }

We now use a much simpler regular expression to extract the pool names from that content:

re_pool = re.compile('''
(\S+) \s+ { \s+ [^}]* \s+ } \s+
''', flags=re.VERBOSE)

That has a single capture group ((\S+)) that will match the pool name; the remainder of the expression takes care of matching the { <anythingthing not '}'> } part. We use re.findall to get all of the matches in one go:

>>> re_pool.findall(match.group('pools'))
['test1-pool', 'test2-pool']

And that’s it!

For more information

For more information on Python regular expressions: