Add support for the reversed() built-in on appropriate iterators

[rpdelaney]

https://docs.python.org/3/library/functions.html#reversed

reversed(seq)

    Return a reverse iterator. seq must be an object which has 
a __reversed__() method or supports the sequence protocol 
(the __len__() method and the __getitem__() method with integer 
arguments starting at 0).

As discussed offline, I'm planning on implementing support for this built-in with a PR when I can get the time.

Discussion in #95 might be relevant.

[niklasf]

Thanks for your PR. I implemented it a little bit differently:

• Not using node[n] for now.
• node.mainline() is an iterator over the nodes starting at (and excluding) node.
• reversed(node.mainline()) is the corresponding reversed iterator.
• node.mainline().accept(visitor) works.

__getitem__, __len__ and maybe others could be added on node.mainline(). I didn't do this to start with, because it appears to be no less efficient than working with list(node.mainline()).

Note that master breaks backwards compability with 0.23.x.

[rpdelaney]

Cool, I'll test this out.

As for compatibility, it looks like that's broken due to doing s/main_line/mainline in various places. Is that all?

[niklasf]

s/main_line/mainline_moves/, some more renamed methods, node.accept(...) includes the move leading to node.

[rpdelaney]

Just curious, but is that rename really worth breaking compatibility? You could maintain compatibility by leaving the old form behind pointing to the new form, but ... idk, seems like a lot of extra work just to kill an innocent underscore. :)

[niklasf]

It does a different thing now: It iterates over nodes, where it previously was iterating over moves. So might as well change the name, which I always wanted to do anyway.