Hopefully I'm not posting this again. Hello, I'm wrapping boost::regex with cython to be able to call it from python. Assuming a "Hello World" text and a regex (\w+)*. This results in two matches. groups ('Hello', 'Hello') lastindex 1 group:0 Hello start 0 end 5 group:1 Hello start 0 end 5 groups ('', '') lastindex 1 group:0 start 5 end 5 group:1 start -1 end -1 I understand, that there is always the main match in group0 and matches from capturing groups in group1 ... As we see, the second match, the zero-length-width match, reports two groups but the second group returns -1. Is it possible to prevent such matches to be reported beforehand? Or is it needed to iterate over the groups to eliminate those? Just in case it is needed to see my code logic, this is what I'm doing currently def unicode_research_iter(const wchar_t* text, wchar_t* pattern, int flags): cdef: wcmatch what size_t _length size_t _position wcregex_iterator start, end try: start = make_regex_iterator(text, <wregex>pattern, match_flags.match_perl) end = wcregex_iterator() while (start != end): what = <wcmatch>deref(start) if not what.empty(): match_object = UnicodeMatch.from_instance(what) yield match_object else: print('Empty match result: ??') inc(start) # increment except Exception as e: raise RuntimeError(f'{e}') Thank you Eren