Francesco Bonazzi

> @Upabjojr would you have time to do a video call with @naveensaigit and I? We have been working on this lately. We think it should be possible to generate,...

Please remember that unless there is support for the coroutines introduced in C++20, any attempt at using RUBI in SymEngine will either timeout or cause a memory error.

Maybe we need a way to automatically translate the code into C++.

> (only 42 rules as compared to 3000 for Rubi-4) Have you tried to use the submodule `matchPyCpp` on those 42 rules? It's a Python submodule in SymEngine that is...

Also, as a reminder, this work could be useful to SymPy as well...

> However, I'm not sure if that's sufficient; we need to make sure that something reasonable happens if the same wildcard is used in functions with different neutral elements, for...

> ["Neutral element"](https://en.wikipedia.org/wiki/Identity_element) or "identity element" (or just "identity") makes much more sense. That makes sense within an algebraic context. Consider that MatchPy's scope is a lot wider... you could...

Maybe we still need the definition of identity elements in MatchPy, but for a different reason: I'm realizing that `Wildcard.star` is not working in SymPy in case of empty sequence...

> Are you planning to do any experiments to compare the performance of the generated C++ code to the generated Python code? I would be curious about that. At the...

Are you willing to port the code to C++20?