Allow overloaded literals

[brson]

let i: uint = 0 is not valid and this frustrates almost everybody. Maybe there's some compromise that solves the biggest problem.

[marijnh]

This would actually be pretty easy to do, but A) it'll only work in contexts where the type of the literal is known before the typechecker reaches the literal (so the above would work, as would i += 1, but not let i = 1), which might confuse people, and B) I think Graydon is allergic to this kind of thing.

[graydon]

I don't know the typechecker well enough to say how invasive this is, and/or how determinstic this can be made. I would want it to be very predictable.

[graydon]

Email summary of IRC discussion:

https://mail.mozilla.org/pipermail/rust-dev/2012-March/001461.html

Any other preferences? I believe niko is working on refactoring typeck presently.

[jamorton]

I would really really really like this to be implemented. I'm creating an emulator and I cast my u16's to ints before alting them because I can't read the pattern matching when it looks like:

0x00u16 to 0x0fu16 { }
0x10u16 to 0x20u16 { }
0x21u16 { }
0x22u16 { }

compared to

0x00 to 0x0f { }
0x10 to 0x20 { }
0x21 { }
0x22 { }

I find myself using ints so much less than uints, u8s, etc. The syntax may be logical but its extraordinarily hard to read.

[nikomatsakis]

I have some good ideas for implementing this.

[lkuper]

The pull request I just put in (#2469) is a stab at making it possible to drop the "u" suffix under certain conditions. From talking to @nikomatsakis, it sounds like a more principled approach would have to involve creating a new kind of type variable and doing unification on those, separately from the existing type vars; this is more of a quick hack.

[lkuper]

I'm getting started on a less hacky solution to this one.

[lkuper]

Sorry for all the commit spam on this issue. I'm getting close to having something that works well. e3f73e4 has some examples of tests that I have passing locally. I'm hoping to be able to un-xfail these and enable suffix inference in the compiler soon...

[lkuper]

As of 3cf582b, suffix inference has landed and seems to be working well enough to close this bug. (Yay!) Please open new bugs for specific issues you encounter with it.