This is potentially not safe. Rust will fill the unused char ranges with other values when it can do. (if not yet, it can do those optimization in future)
To correctly fix this problem, #3574 need to be finished.
are you interested in it?

I'm still looking for a better solution for this. Sure. I'll take a look at #3574. Thanks.

@garychia did you get any idea about it?

Currently the parser uses Rust chars and strings to handle Python string literals. This does not work with Unicode surrogate characters. We may need to change the way we store Python string literals. My suggestion is to replace Rust strings with our custom class (such as PyStr, I guess). But I guess this may require a lot of work to be done.

Perhaps the bstr crate can be useful.