Consider allowing limited type inference in static declarations

[sfackler]

Currently, statics must have a type annotation as doing global type inference to resolve the type is a bad idea. However, this leads to some annoying duplication, especially when defining a static struct instance. For example, from extra::base64:

pub static STANDARD: Config =
    Config {char_set: Standard, pad: true, line_length: None};

There's no ambiguity in the type of the RHS, but the LHS must repeat the type anyways.

Would it be possible/reasonable to infer the type of statics from the RHS of the assignment where possible? Things like this would be work:

pub static STANDARD =
    Config {char_set: Standard, pad: true, line_length: None};

pub static HELLO = "hello, world!";

pub static HELLO_BYTES = bytes!("hello, world!");

pub static BAR = 18u8;

but

pub static BAZ = 10;

would not since the type isn't specified (or maybe it would resolve to int?). Untyped integer literals and generic types containing them are actually the only things I can think of that would require a type annotation.

[emberian]

I think I'd be ok with this, given the current very limited form of constant expressions. It's not possible to change the type of the static without editing its initializer. This is backwards compat, though, so not a priority.

[nikomatsakis]

cc me

[rust-highfive]

This issue has been moved to the RFCs repo: rust-lang/rfcs#296