> Do you realize that making + do what the ordinary mathematical + does, is quite expensive? It requires arbitrary precision arithmetic and hence, potentially, memory allocations.
Yes I do. It's an acceptable cost in most cases (you will note much of the world runs on Python, a language in which + does behave sensibly, these days), and dangerous optimisations should be opt-in rather than opt-out.
In the worst case you should at least make your language fail-stop definite by making it error when the result of + is too large rather than silently continuing in an invalid state.
