Remove `\xXX` char escapes from the language

[Valloric]

\xXX is very misleading in Rust since it actually works exactly like \u00XX instead of the way it works in C, C++ and other languages. Example:

// this FAILS because left is [195, 191]
assert_eq!( bytes!( "\xFF"), bytes!( 255 ) ); 
// this SUCCEEDS
assert_eq!( bytes!( "\xFF"), bytes!( "\u00FF" ) );  

I understand the reasoning behind this (Rust strings are always UTF-8), but then \xXX shouldn't exist in the language. It brings nothing but confusion and it's functionality as implemented is the same as \u00XX.

[alexcrichton]

Closing, this was previously decided in #2800 to be working as intended.

[Valloric]

#2800 was about changing \xXX to mean utf8 code unit instead of unicode codepoint. I agree with the conclusion in that issue that that change is not a good idea since it isn't useful.

But this issue is separate; it's about removing \xXX from the language entirely. \uXXXX has to exist because it covers a larger range of values and means "unicode codepoint" in every language. \xXX in Rust has no use because it's equivalent to \u00XX but causes confusion because \xXX in C and C++ means raw byte hex.

So in aggregate it's worse than useless, it's a net negative. It addresses no use case and provides no benefit but comes with a cost; the only thing it does successfully is confuse users coming from Rust's primary market, C & C++ developers.

I honestly can't see why it's being kept.

[lilyball]

I'm strongly in favor of a modified form of this, where we allow \xXX for ASCII characters but disallow it for non-ASCII. This was suggested recently in rust-lang/rfcs#69, and apparently was also suggested back in #2800.

Keeping \xXX for codepoints U+0000 through U+007F seems like a good idea, because it means the same thing regardless of whether \xXX is interpreted as a codepoint or a code unit, and it's a convenient syntax for referring to ASCII characters. But interpreting \x80-\xFF as unicode only serves to be confusing. And not just to C/C++ programmers; even though I've been using Rust for quite a while, the other day I caught myself using \x80 in a string and expecting to get the byte 0x80.

If we restrict it to ASCII characters now, that also makes the behavior of the proposed byte string literals (rust-lang/rfcs#69) make more sense, where \x80 will definitely want to refer to the byte 0x80.

[chris-morgan]

I am in favour of restricting it for non-bytestring-literals also.

[emberian]

This is surprising, I assumed this was only a byte literal. I agree with @kballard here.

[Valloric]

As I mentioned on rust-lang/rfcs#69, I agree with @kballard's proposal.

[rust-highfive]

This issue has been moved to the RFCs repo: rust-lang/rfcs#312