New lint: prefer-ascii for Deno's internal JS

[magurotuna]

Optimization around deno_core is being made now in denoland/deno#9713. This optimization requires that all the JS code that is eventually bundled in Deno is fully written only in ASCII i.e. one-byte characters. So it would be great to have a new lint rule that makes sure of that.

Similar to prefer-primordials, the new lint will be dedicated to the internal of Deno. Not intended to be used by Deno users.

CC @AaronO @bartlomieju

[magurotuna]

I'd love to work on it

[grian32]

I believe this rule may also be of use to regular code, as you don't usually find yourself using chars outside of ASCII and if you do it might be a mistake(accidentally copy pasting a minus unicode character instead of - which while looking pretty similar, the minus unicode character is not valid javascript AFAIK).

[bartlomieju]

I believe this rule may also be of use to regular code, as you don't usually find yourself using chars outside of ASCII and if you do it might be a mistake(accidentally copy pasting a minus unicode character instead of - which while looking pretty similar, the minus unicode character is not valid javascript AFAIK).

Once denoland/deno#11776 lands it will be possible to enable this rule, but I don't think it should be part of "recommended" set

[grian32]

I don't think it should be part of "recommended" set

I agree @bartlomieju

[magurotuna]

I thought what @grian32 said made a lot of sense and tried to deal with characters outside of ASCII in JavaScript syntax, like the case where ‐ (this is not an ASCII, unicode code point is U+2010) is placed at the position of binary operation. However, it seems like we can't handle this case, because swc only accepts valid JavaScript. Therefore, even without this rule, using ‐ (U+2010) mistakenly would be detected by running the linter as a result of the parse error.