[maartenscholl]

The author claims to be an expert in C++ but begins the article with a detour that incorrectly states that static and dynamic dispatch are necessarily orthogonal in C++. In reality you can do both within the same hierarchy using CRTP combined with a virtual base method. Compile-time inlining eliminates the overhead associated with the virtual call, so you can match Rusts flexibility with the same performance.

[wakawaka28]

The cases where you can inline a virtual method seem narrow. So narrow in fact, that if it is possible then you should probably only be doing CRTP or simple static dispatch. See: https://stackoverflow.com/questions/733737/are-inline-virtua...

If you try to use CRTP + virtual on polymorphic types then one has to wonder if it will work as intended for both use cases (when used as a static object or a polymorphic one).

I'm not the absolute most expert C++ programmer, but I'm no noob. The idea of deliberately introducing dynamic polymorphism only to try to optimize it out seems like a bad idea. It's unnecessarily complicated and confusing. If you want to go fast just use CRTP straight up and forget all about dynamic dispatch and potential cute optimizations.

Edit: I think this explains my objection: https://www.codeproject.com/Tips/537606/Cplusplus-Prefer-Cur... (I found that in the StackOverflow post I linked to above.)

[klysm]

I feel like expert in C++ is a wide range. There are corners of C++ that probably 5 people know about total