It’s very easy to run pypy on your own repls. Open the System dependencies tool and install pypy3, then show hidden files and change or add the run command in .replit file. (This will probably be PyPy 3.9.)
A more difficult alternative to PyPy for performance is Cython.
When I open the Systems Dependencies Tool there’s no option to actually install anything. I could just create the Nix file and add the pkgs.pypy310 dependency - is that what you mean?
Edit: I’ve just tried adding the Nix file with the PyPy-dependency and it worked.
I’ve used Cython quite a bit, but for my use cases PyPy turned out to be the better (faster) option , and I found it much easier to handle.
The System Dependencies tool is not guaranteed to the search and install functionality.
I believe the specific nix channel that the repl uses (which can be viewed and edited in .replit file) determines this, 23.05 has it while 23.11 doesn’t. (But 23.05 does not have pypy 3.10)
It certainly depends on the use case. For example, I was almost required to use Cython for my brute force algorithm rather than python or pypy, and it has indeed given me the performance of C. However, the code has to be adapted more to gain the full performance and it has all the drawbacks of coding in C.