How do you figure out the types at run time with zero overhead? I guess we have different concept of "zero".

As for the price of my skills - my paycheck told me that. And sure, it could be low demand (there are vacancies in my field, that had been open for years, but I don't really know what kind of demand there is for Python, could be decades for all I care). I am just happy for my compensation and if others have more expensive skills - I am happy for them too :)

Nothing is truly zero overhead. C++ templates are neither, because they create some overhead in code size which is obviously non-free (both space-wise and time-wise). Zero-overhead for me = optimized to the point that noone can notice. There is no problem for a VM to generate 10 versions of a tight loop, every one optimised for particular type, with a switch for the type right before entering the loop, or the whole method, or even a whole call-tree.

We also mean different things by overhead. For me (and the C++ committee) it means something that is done in addition. So for me templates create zero overhead as the do not produce more code than the same, non-template code. In practice, templates produce less code as there is a higher chance to fold common code from the instantiations of the same template than from many handwritten functions.

Comparing to other generics implementations - they do add a lot of code, much more than is really needed to achieve good performance. Sure, some of this code can be reduced at linking stage, but for this to work you need to statically link everything which has its own set of problems (overhead of repeated code in every binary).

IMHO the approach is a great example of premature optimisation: generate 2^N versions then try to fold some similar ones to save space. Much better approach is: generate 1 version and specialize small parts when really needed. Technically, assuming sufficiently smart compiler, both would end up with exactly the same code, performing exactly the same, but the latter solution is preferable for other non-performance wise reasons (like better type checking, ability to distribute binary libraries, faster compile times, etc.).

BTW: I still often find Scala containers to be often much faster than C++ STL. But this is probably for much different reasons than template/generics implementation.

I have lost you. Are you still arguing that C++ is internally inconsistent because it's not doing stuff that Java does? Are you arguing that templates add a lot more code than a VM and JIT compiler and languages with generic functions are as efficient as C++?

In the first case - I fail to see how it follows. In the second case - our levels of expertise is too far apart to discuss anything. I am not saying you are not qualified, I might be just too out of the loop and missing some dramatic advances in Java, Scala etc.