Artem Pavlenko

@jgreenbaum - Thanks for reporting! Having both `register_ptr_to_python` and also registering conversion in `_class` constructor is causing runtime warnings. At least on OS X/python-3.5/boost_1_61 ``` python /Library/Frameworks/Python.framework/Versions/3.5/lib/python3.5/importlib/_bootstrap.py:222: RuntimeWarning: to-Python converter...

@jgreenbaum - ~~could you try latest master and see if this issue is solved~~ I'm seeing other issues creeping in - will take another look tomorrow

re: `get` - Lets not over engineer here, at least not now :). As stated in https://github.com/mapbox/variant/blob/master/README.md the goal is : "..is to maintain external API compatibility with boost::variant such...

@lightmare - I agree, "same-type alternatives" are pretty useless. My only concern is that compile check can be relatively expensive (thinking about mapnik::expression_node -https://github.com/mapnik/mapnik/blob/master/include/mapnik/expression_node_types.hpp#L169-L196). How would you propose checking type...

Found this one : http://stackoverflow.com/questions/18986560/check-variadic-templates-parameters-for-uniqueness

Maybe some `optimization` ideas are in http://www.boost.org/doc/libs/1_60_0/libs/mpl/doc/refmanual/set.html

@joto - great, will need to bench it! How about making this check an optional (default: yes) ? This way it'll protect new users and also will allow power users...

planning to bench finally /cc @joto @springmeyer @lightmare

@lightmare - ok, in this particular case both version will be likely inlined so not much difference, but I'm personally considering current version a better coding style. Why are you...

@lightmare - thanks. I can see it will work but I'm a bit hesitant to just go and change internal index to `N,...,1` Lets get a consensus on this :)...