Nicer printing of type and region variables

[lkuper]

Today on IRC several of us were talking about how the way that type variables and region variables are printed (<VN> and <RN>, respectively, where N is an id number) is rather ugly. Now that we have integral type variables, printing them as <VIN> (as is happening now) just adds insult to injury. "integer" and "unconstrained int" were suggested as nicer solutions for integral type variables. We should come up with something nicer for the others as well.

There's already a -Z ppregions compiler flag, useful for debugging, that prints the IDs of regions. Maybe there could be a -Z ppvarids flag that prints the ID numbers for times when we really need them.

[nikomatsakis]

this sounds good. I think we ought to just roll these flags together into one though.

[lkuper]

So, just to be clear, -Z ppregions prints IDs of regions, not region variables, right? The proposed -Z ppvarids would print IDs for type variables, integral type variables, and region variables. Does it make sense to conflate those two flags?

[nikomatsakis]

conceptually I think we just want one flag for "print types using internal notation". This would affect the printing of regions as well as type variables and so forth.

[graydon]

Agreed. We discussed this at a meeting recently and wanted to get tidied up. Possible relative of #2951 maybe?

IIRC the strings ? and ?N were suggested. Not sure if those are any better.

[catamorphism]

Nominating for milestone 5, production-ready

[graydon]

just a bug, removing milestone/nomination.

[pnkfelix]

visiting for triage, email from 2013 sep 09.

[flaper87]

Error messages containing generics were changed in this commit e25d706 regions haven't been changed.

Triaged

[huonw]

This bug really needs a test case or two. For type parameters, we current give an error message like

expected `<generic float #0>` but found `<generic integer #0>` (expected floating-point variable but found integral variable)

code:

fn main() { let 1 = 1.0; }

And for regions... I don't know how to get one to be printed.

[flaper87]

@nikomatsakis are region variables just printed in debug messages? I didn't find a case for that, TBH.

@huonw Agreed, I think, as you also mentioned on IRC earlier, that the current format shouldn't be the final format. We could start discussing a better format for that here. My first thought is that we could maybe get rid of generic in the current message.

[treeman]

Visiting for triage. Error message @huonw mentioned is still relevant.

Still would like some more test cases.