""", re.VERBOSE | re.MULTILINE)

_header_next_check = re.compile(br'''(?!@)(?!literal )(?!delta )''')

def _get_header(vcs, diff_chunk):

    """

    Parses a Git diff for a single file (header and chunks) and returns a tuple with:

    1. A dict with meta info:

        a_path, b_path, similarity_index, rename_from, rename_to,

        old_mode, new_mode, new_file_mode, deleted_file_mode,

        a_blob_id, b_blob_id, b_mode, a_file, b_file

    2. An iterator yielding lines with simple HTML markup.

    """

    match = None

    if vcs == 'git':

        match = _git_header_re.match(diff_chunk)

    elif vcs == 'hg':

        match = _hg_header_re.match(diff_chunk)

    if match is None:

    meta_info = {k: None if v is None else safe_str(v) for k, v in match.groupdict().items()}

    rest = diff_chunk[match.end():]

    if rest:

        if _header_next_check.match(rest):

            raise Exception('cannot parse %s diff header: %r followed by %r' % (vcs, safe_str(bytes(diff_chunk[:match.end()])), safe_str(bytes(rest[:1000]))))

        if rest[-1:] != b'\n':

            # The diff will generally already have trailing \n (and be a memoryview). It might also be huge so we don't want to allocate it twice. But in this very rare case, we don't care.

            rest = bytes(rest) + b'\n'

    diff_lines = (_escaper(safe_str(m.group(1))) for m in re.finditer(br'(.*)\n', rest))

    return meta_info, diff_lines

_chunk_re = re.compile(r'^@@ -(\d+)(?:,(\d+))? \+(\d+)(?:,(\d+))? @@(.*)')

_newline_marker = re.compile(r'^\\ No newline at end of file')

def _parse_lines(diff_lines):

    """

    Given an iterator of diff body lines, parse them and return a dict per

    line and added/removed totals.

    """

    added = deleted = 0

    old_line = old_end = new_line = new_end = None