Use re.split with non-capturing groups instead of handwritten code

    return ''.join(up_to_brace), string

@fix_unicode_literals_in_doctest

def _split(string, sep):

    """

    >>> sep = re.compile(',')

    >>> list(_split('', sep))

    [u'']

    >>> list(_split('a,b,c', sep))

    [u'a', u'b', u'c']

    >>> list(_split('a,b,c,', sep))

    [u'a', u'b', u'c', u'']

    """

    while True:

        match = sep.search(string)

        if not match:

            yield string

            break

        else:

            yield string[:match.start()]

            string = string[match.end():]

# "\ " is a "control space" in TeX, i. e. "a space that is not to be ignored"

#     -- The TeXbook, Chapter 3: Controlling TeX, p 8

# ~ is a space character, according to BibTeX

# \~ is not a space character

BIBTEX_SPACE_RE = re.compile(r'(\\ |\s|(?<!\\)~)+')

BIBTEX_SPACE_RE = re.compile(r'(?:\\ |\s|(?<!\\)~)+')

BRACE_RE = re.compile(r'{|}')

        head, brace, string = string.partition('{')

        if head:

            head_parts = list(_split(head, sep))

            head_parts = sep.split(head)

            for word in head_parts[:-1]:

                result.append(''.join(word_parts + [word]))

                word_parts = []