Replace Ident with Name in for loops

[flaper87]

Cleaning up Ident usage in rustc. This patch replaces Ident usage in ExprFor and ExprForLoop

cc #6993

[flaper87]

FWIW, I'll keep working on this cleanup. If we want to hold this PR off and merge them all, that's fine.

[huonw]

Doesn't this remove the possibility for 'foo: for x in bar() { ... } to be hygienic? i.e.

macro_rules! breakfoo(() => {break'foo;})'foo:for x inbar(){breakfoo!()}

would compile and immediately break. I'm not sure (a) if I'm understanding this correctly, and (b) if we want this poor hygiene

(@paulstansifer@jbclements is my assessment above accurate: using Name rather than Ident makes it non-hygienic?)

[jbclements]

@huonw your understanding sounds correct to me.

[alexcrichton]

Is the issue correct in that we want to have a Name on label instead of an Ident? Or does this implementation just need some touch-up around the edges?

[flaper87]

@alexcrichton I'd like to hold this patch until I verify @huonw's concerns w.r.t this change making for-loops non-hygienic. FWIW, I don't think this specific change makes for loops non-hygienic because what it is doing is replacing a bunch of ident.name with name.

At least for for loops, it looks like the only thing we're actually using from Ident is the name. I'm not sure about removing Ident all around, though.

[huonw]

FWIW, I don't think this specific change makes for loops non-hygienic because what it is doing is replacing a bunch of ident.name with name.

That's exactly what the difference between being hygienic and not is; the non-Name part of the Ident is what tells the compiler about where an ident comes from (i.e. which "macro expansion context" the ident was created in). Our loops are probably not hygienic at the moment, but hopefully they will be in future and so removing this only to add it back in later seems counter-productive.

[flaper87]

@huonw I know. At this point I'd recommend discussing #6993 in the next meeting since it's targeting more than just for-loops and AFAIK some clean up was done already. /cc @alexcrichton

[huonw]

To be clear, lack of hygiene will mean (at least) one of the following will be incorrect:

macro_rules! foo {() => {// this should just break from this inner loop...'x:loop{break'x;}}}'x:loop{// ...but it may refer to the outer 'x when used inside herefoo!();println!("should be printed");}

or

macro_rules! foo {($e: expr) => {// $e shouldn't be able to interact with this 'x'x:loop{ $e }}}'x:loop{// i.e. this 'x should refer to the outer loop, but may be "captured"// by the 'x declaration inside foofoo!(break'x);println!("should not be printed")}

(I don't have a compiler with me to see if either of these are currently broken.)

[huonw]

@flaper87 also, it's worth noting that #6993 is specifically "replace Ident by Name after libsyntax", i.e. libsyntax (and presumably the resolve pass) still needs Idents.

[huonw]

That said, it seems break and continue explicitly use Names: #9103.

[flaper87]

also, it's worth noting that #6993 is specifically "replace Ident by Name after libsyntax", i.e. libsyntax (and presumably the resolve pass) still needs Idents.

Yeah, bad wording from my side!

I tested something similar to the first example yday and forgot to comment back here. Anyway, I gave both your examples a try with and without this change, as expected the behavior was the same in both cases. The first example works as expected whereas the second one doesn't.

Presumably, reverting #9103 should make our for-loops hygienic again.

I'll close this PR and work on bringing the hygienic behavior back.

Thanks @huonw and @alexcrichton