okenize a source reading Python code as unicode strings.

    This has the same API as tokenize(), except that it expects the *readline*
    callable to return str objects instead of bytes.
    N)