I believe I've found a bug in the `re` module -- specifically, in the 3.7+ support for splitting on zero-width patterns. Compare Java's behavior...

    jshell> "1211".split("(?<=(\\d))(?!\\1)(?=\\d)");
    $1 ==> String[3] { "1", "2", "11" }

...with Python's:

    >>> re.split(r'(?<=(\d))(?!\1)(?=\d)', '1211')
    ['1', '1', '2', '2', '11']

(The pattern itself is pretty straightforward in design, but regex syntax can cloud things, so to be totally clear: it finds any point that follows a digit and precedes a *different* digit.)

* Tested on 3.7.1 win10 and 3.7.0 linux.

From the docs:

"""If capturing parentheses are used in pattern, then the text of all groups in the pattern are also returned as part of the resulting list."""

The pattern does contain a capture, so that's why the result has additional '1' and '2'.

Presumably, Java's split doesn't do that.

Not a bug.

Thank you. Was too zeroed-in on the idea that it was from the zero-width pattern, and I forgot to consider the group. Looks like `re.sub(pattern, 'some-delim', s).split('some-delim')` is a way to do this if it's not possible to use a non-capturing group

The list alternates between substrings (s, between the splits) and captures (c):

['1', '1', '2', '2', '11']
 -s-  -c-  -s-  -c-  -s--

You can use slicing to extract the substrings:

>>> re.split(r'(?<=(\d))(?!\1)(?=\d)', '12111')[ : : 2]
['1', '2', '111']

stage: resolved