rustc: Remove HirId from queries

[alexcrichton]

This'll allow us to reconstruct query parameters purely from the DepNode
they're associated with.

Closes#44414

[rust-highfive]

r? @arielb1

(rust_highfive has picked a reviewer for you, use r? to override)

[alexcrichton]

r? @michaelwoerister

[eddyb]

I believe there's DefIdMap and DefIdSet aliases which could be used for these changes.

[eddyb]

So with_freevars used to be a thing because of a RefCell. Can it be removed in favor of tcx.freevars(def_id)? All callers should, ironically, already have a closure DefId around.

[kennytm]

Could not compile core

[00:17:57] error: internal compiler error: unexpected panic
[00:17:57] [00:17:57] note: the compiler unexpectedly panicked. this is a bug.
[00:17:57] [00:17:57] note: we would appreciate a bug report: https://github.com/rust-lang/rust/blob/master/CONTRIBUTING.md#bug-reports
[00:17:57] [00:17:57] note: rustc 1.22.0-dev running on x86_64-unknown-linux-gnu
[00:17:57] [00:17:57] thread 'rustc' panicked at '/checkout/src/librustc/hir/map/mod.rs:330: local_def_id: no entry for `291005`, which has a map of `Some(EntryLifetime(NodeId(192109), DepNodeIndex { index: 4294967295 }, lifetime(291005: )))`', /checkout/src/librustc/session/mod.rs:889:25
[00:17:57] note: Run with `RUST_BACKTRACE=1` for a backtrace.
[00:17:57] [00:17:57] error: Could not compile `core`.

[bors]

☔ The latest upstream changes (presumably #44335) made this pull request unmergeable. Please resolve the merge conflicts.

[michaelwoerister]

Thanks, @alexcrichton!

Unfortunately it won't be as easy as just replacing HirId with DefId here because not every HirId has a corresponding DefId. For example, HIR expressions don't have a DefId, so moving in_scope_traits to DefId also involves storing things in a two-level map, so that it can look something like this:

implTyCtxt{fnin_scope_traits_of_expr(self,hir_id:HirId) -> Option<Rc<Vec<TraitCandidate>>>{// Get the map for the item containing the HirId. This is a query:let map_for_containing_item = self.in_scope_traits_of_item(hir_id.owner);// From that item-local map get actual return value
map_for_containing_item.get(hir_id.local_id).cloned()}}

[alexcrichton]

Heh yeah the panics figured that out pretty quickly here! I'll work on making these sub-maps.

[alexcrichton]

Alright I believe the necessary maps have been transitioned, re-r? @michaelwoerister

[eddyb]

Why not use DefId here? Closures have DefIds.

[alexcrichton]

Indeed! Got a bit overzealous with DefIndex

[michaelwoerister]

Looks good to me except for that one performance concern.

[michaelwoerister]

Whoops, I should better have listened to @arielb1 and marked find_node_for_hir_id as super-slow somehow. It does a linear search over all NodeIds 😰

But it seems that what you're doing here is equivalent to tcx.hir.as_local_node_id(id).unwrap()?

[eddyb]

Why does find_node_for_hir_id exist? I've introduced hir_to_node_id. (EDIT: I generated a reverse HashMap for that purpose)

[michaelwoerister]

Oh, I see. It does exist because I wanted to avoid building that reverse mapping because it was only needed in error reporting. It's probably cheap to build though, especially now that locals don't have DefIds anymore.

[alexcrichton]

@bors: r=michaelwoerister

[bors]

📌 Commit 772ca85 has been approved by michaelwoerister

[alexcrichton]

@bors: r=michaelwoerister

[bors]

📌 Commit caaf365 has been approved by michaelwoerister

[bors]

⌛ Testing commit caaf365 with merge efa3ec6...

[bors]

☀️ Test successful - status-appveyor, status-travis
Approved by: michaelwoerister
Pushing efa3ec6 to master...