LWN: Comments on "Rust 1.58.0 released"

Rust 1.58.0 released

mathstuf — Mon, 17 Jan 2022 12:09:20 +0000

There's a thread on IRLO[1] about this. The thing is that Python has the concept of a "simple" expression (basically what lambdas allow) to limit how crazy these expressions can get. Rust has no such restraint and if it is just "any expression", you end up allowing things like custom derive and macro expansion inside of these things. See my reply[2] with an example which does this. Is this allowed in a format string?

{
#[derive(Serialize)]
struct JsonWrapper { field1: bool, field2: u32 };
let wrap = JsonWrapper { field1: some_local_bool, field2: some_local_int };
serde_json::to_string(wrap)
}

because it is an expression. FWIW, even "simple" things end up needing a full expression parser because of raw string literals like `r##"{(string.add(r#"))}"#))}"##` so "just count quotes and braces" needs to have an active string literal parser to figure things out.

[1] https://internals.rust-lang.org/t/how-to-allow-arbitrary-...
[2] https://internals.rust-lang.org/t/how-to-allow-arbitrary-...

Rust 1.58.0 released

suckfish — Mon, 17 Jan 2022 04:39:22 +0000

Yes, as of today only a bare identifier is supported.

I don't see any fundamental reason why more complex expressions can't be supported, I guess it is a matter of taking time, thinking about the corner cases, and making sure it gets done properly.

Rust 1.58.0 released

darmengod — Sun, 16 Jan 2022 23:19:55 +0000

I remember being rather happy when Python3 introduced f-strings in 3.6.

>Format strings can only capture plain identifiers, not arbitrary paths or expressions.

I guess struct members are out of the question, then?