Consider this code snippet:


from re import match, fullmatch

pattern = '".+?"'
string = '"hello" "again"'

print(match(pattern, string))
print(fullmatch(pattern, string))


Which prints:
<_sre.SRE_Match object; span=(0, 7), match='"hello"'>
<_sre.SRE_Match object; span=(0, 15), match='"hello" "again"'>

The fullmatch function seems to ignore the non-greedy modifier.

From the fullmatch docstring I expected that fullmatch is equivalent to:

def fullmatch(pattern, string):
    match = re.match(pattern, string)
    if match:
        if match.start() == 0 and match.end() == len(string):
            return match
        else:
            return None
    else:
        return None

No, this is not how fullmatch() works. Otherwise there wouldn't be need to add it into the stdlib. You could just check end() of the result.

fullmatch() acts like match() with added \Z. But if the pattern contains multiple alternates you need to add \Z at the end of every branch (r"foo\Z|bar\Z") or wrap the entire patch in a non-capturing group (r"(?:foo|bar)\Z").