Variable named ∇x gives "unknown start of token" compiler error

[Danvil]

I tried this code:

let ∇x = 1;

I expected the code to compiles but instead I get the compiler error message "unknown start of token \u{2207}".

This is surprising as variable names starting with Greek letters are fine:

let Δλ = 1;

I believe the cause is that Rust identifiers need to start with a XID_Start unicode characters, however the "Nabla" ∇ (0x2207) does not seem to be on that list.

It would be great to have the "Nabla" operator as a valid start token for identifier as it very commonly used in physics and mathematics to denote the derivative of a multi-variable function.

A possible workaround is to use the "Canadian syllabics e" ᐁ (0x1401).

[Jules-Bertholet]

See rust-lang/rfcs#3402 and https://www.unicode.org/reports/tr31/proposed.html#Mathematical_Compatibility_Notation_Profile

@rustbot label T-lang -C-bug C-enhancement

[Manishearth]

This would need an RFC to extend the current identifier profile (the default one from UAX 31) to use the mathematical notation profile.

This would add these characters to the identifier profile (with the superscripts and subscripts not being allowed at the beginning of an identifier)

All of these would get linted on by the uncommon_codepoints lint since they have Identifier_Type=Not_NFKC.

(A change I want to make is for uncommon_codepoints to have slightly different lint text based on the category that is triggered: #120228)

[bend-n]

Hey if were getting mathematical characters can i just say i would really love it if we had
≥ ≤ ∈ ∉ ∣ ∤ ∋ ∌ ∅ ∪ ∖ ∆ ∩ ¬ ∧ ∨ ⊻ ⊼ ⊽
is there a unicode profile for these?

Also what about the emoji profile 😀

[Manishearth]

There isn't for the math operators because those are considered operatorlike.

As for emoji it's unlikely. Rust would have to put together its own set.

Emoji identifiers are a complicated can of worms.

[CraftSpider]

Personally I'd be interested in math operators at least tokenizing, so they could reach macros as Punct or such, but I figure that's somewhat unlikely to actually happen.

[gilescope]

mathematical notation profile support would be great - It would be very helpful that subscript and superscript numbers are allowed to be used as identifiers ( E.g. let 𝐞₁ = ). One of the cool things about Julia is that it has good mathematical symbol support.

[Jules-Bertholet]

If you want this, write an RFC for it.

[bend-n]

simply RFC this to unicode

[Manishearth]

No, the mathematical notation profile is Unicode's solution to this problem. It allows consumers of the specification, like Rust, to make this type of choice on its own merits. Writing further proposals to Unicode would not be productive.