OTOH, the more pythonic one will probably perform worse, but I’m not familiar enough with Python internals to make that claim without benchmarks.

I’ll try it out and add the data in the article