Thank you everyone for your advice.
Some kinds of names are restricted to unicode alphabetic/numeric in Rust.
And the current definition of the table defined in libcpp/ucind.h lacks
some rows representing which characters are alphabetic/numeric.
But it is not a problem because it seems to be easy to add missing rows to
the table and use it in the Rust frontend.

2023年3月16日(木) 21:59 Mark Wielaard <mark@klomp.org>:

> You might want to research whether NFC normalization of identifiers is
> required to be done by the lexer or parser in Rust and how it interacts
> with proc macros.


Yes, NFC normalization must be done by the lexer, which may be complex and
hard to implement.
libunistring can also be used for normalization, so is it good to use
libunistring only in the normalization process?

Raiki Tamura