Refactored new ScopeData for improved abstraction. #19033

pnkfelix · 2014-11-17T15:41:49Z

(Previously, scopes were solely identified with NodeId's; this
refactoring prepares for a future where that does not hold.)

Ground work for a proper fix to #8861.

pnkfelix · 2014-11-17T15:58:57Z

r? @nikomatsakis (but please feel free to reassign; so much of this is mechanical that it probably could be "r? anyone")

nikomatsakis · 2014-11-17T16:31:27Z

r+ (deliberately did not put on the commit)

Only thing is that I think we should rename ScopeData, which to me suggests "data about scopes" and not "a single scope". On IRC @pnkfelix and I settled on CodeRegion; I'm happy with that or RegionScope or even Extent.

I also think we should probably find a shorter way to go from a node-id to a CodeRegion/RegionScope/Extent, because region::Foo::from_node_id(...) is quite long. Directly importing the name would help, so at least one could write Extent::from_node_id(...), or perhaps defining a free fn in region, so one could write region::extent(node-id). I suspect many of these lines are > 100 characters (I see that travis is failing, haven't investigated, but I bet that is why).

gereeter · 2014-11-17T17:57:55Z

src/librustc/middle/region.rs

@@ -76,11 +117,11 @@ The region maps encode information about region relationships.
  for dynamic checks and/or arbitrary amounts of stack space.
 */
 pub struct RegionMaps {
-    scope_map: RefCell<NodeMap<ast::NodeId>>,
-    var_map: RefCell<NodeMap<ast::NodeId>>,
+    scope_map: RefCell<HashMap<ScopeData, ScopeData>>,


I believe this should be a FnvHashMap.

(Previously, statically identifiable scopes/regions were solely identified with NodeId's; this refactoring prepares for a future where that 1:1 correspondence does not hold.)

…ctor, r=nikomatsakis (Previously, scopes were solely identified with NodeId's; this refactoring prepares for a future where that does not hold.) Ground work for a proper fix to #8861.

fix: Make proc_macro span's line & column 1-indexed, as documented

gereeter reviewed Nov 17, 2014
View reviewed changes

pnkfelix force-pushed the fsk-introduce-scopedata-via-refactor branch 3 times, most recently from 48f24d5 to dd95fa9 Compare November 19, 2014 16:02

Refactored new CodeExtent type for improved abstraction.

5ff9087

(Previously, statically identifiable scopes/regions were solely identified with NodeId's; this refactoring prepares for a future where that 1:1 correspondence does not hold.)

pnkfelix force-pushed the fsk-introduce-scopedata-via-refactor branch from dd95fa9 to 5ff9087 Compare November 20, 2014 12:10

bors closed this Nov 20, 2014

bors merged commit 5ff9087 into rust-lang:master Nov 20, 2014

lnicola pushed a commit to lnicola/rust that referenced this pull request Jan 27, 2025

Merge pull request rust-lang#19033 from ChayimFriedman2/span-1

bbf9ff6

fix: Make proc_macro span's line & column 1-indexed, as documented

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Refactored new ScopeData for improved abstraction. #19033

Refactored new ScopeData for improved abstraction. #19033

pnkfelix commented Nov 17, 2014

pnkfelix commented Nov 17, 2014

nikomatsakis commented Nov 17, 2014

gereeter Nov 17, 2014

Refactored new ScopeData for improved abstraction. #19033

Refactored new ScopeData for improved abstraction. #19033

Conversation

pnkfelix commented Nov 17, 2014

pnkfelix commented Nov 17, 2014

nikomatsakis commented Nov 17, 2014

gereeter Nov 17, 2014

Choose a reason for hiding this comment