bpo-20180: Simplify char_converter in Argument Clinic. - #9828

@serhiy-storchaka, I wrote the existing escaping mechanism based on my reading of the C standard, since self.c_default needs to be a valid C character literal. It seemed to be a safer way to ensure creating a valid C char literal than relying on e.g. repr(bytes(...)), which we might decide to change in the future.

I recall also being surprised by '?' having an escape sequence, but I don't remember why I decided to keep it at the end; perhaps just for completeness, having included all of the other escape sequences.

I definitely defer to your judgement on this. I'm not a C expert and so I was extra-careful when I wrote this code, perhaps overly so.

I think that the repr of str and bytes was intentionally made mostly compatible with C (except that they can use either singular or double quotes). It is very unlikely that they will be made incompatible in future.

? can be need to be escaped because of trigraph sequences like ??=. But in our case ? is a single character.

>>> repr(b'\0')
"b'\\x00'"

Assuming '\x00' is a valid C char literal, it's not as nice as '\0'. IMO not a deal-breaker, for your consideration.

Same goes for '\1' etc.