Handle patterns within closures #45

roxelo · 2021-01-30T21:14:18Z

Using the following test:

#![feature(capture_disjoint_fields)]

fn main() {
    let _z = 9;
    let t = (String::new(), String::new());

    let _c = ||  {
          let (_t1, _) = t;
    };
}

The relevant log output is as follows:

DEBUG rustc_typeck::check::upvar RFC2229 = upvar.rs Initialized delegate InferBorrowKind
DEBUG rustc_typeck::expr_use_visitor RFC2229 = expr_use_visitor.rs walk_pat introduce fake_read
DEBUG rustc_typeck::check::upvar RFC2229 = upvar.rs Adding fake read PlaceWithHirId { hir_id: HirId { owner: DefId(0:3 ~ rox[317d]::main), local_id: 20 }, place: Place { base_ty: (std::string::String, std
DEBUG rustc_typeck::check::upvar RFC2229 = upvar.rs write results out to typeck_results
DEBUG rustc_typeck::check::writeback RFC2229 = writeback.rs call visit_fake_read_map()
DEBUG rustc_typeck::check::writeback RFC2229 = writeback.rs visit_fake_reads_map
DEBUG rustc_mir_build::thir::cx::expr RFC2229 = Closure expr.rs ExprKind::Closure
DEBUG rustc_mir_build::thir::cx::expr RFC2229 = closure_fake_read {DefId(0:4 ~ rox[317d]::main::{closure#0}): [Place { base_ty: (std::string::String, std::string::String), base: Upvar(UpvarId(HirId { owner: DefId(0:3 ~ rox[317d]::main), local_id: 17 };`t`;DefId(0:4 ~ rox[317d]::main::{closure#0}))), projections: [] }]}
DEBUG rustc_mir_build::build::expr::as_rvalue RFC2229 = as_rvalue.rs ExprKind::Closure
DEBUG rustc_mir_build::build::expr::as_rvalue RFC2229 = as_rvalue.rs Place { base_ty: (std::string::String, std::string::String), base: Upvar(UpvarId(HirId { owner: DefId(0:3 ~ rox[317d]::main), local_id: 17 };`t`;DefId(0:4 ~ rox[317d]::main::{closure#0}))), projections: [] }
DEBUG rustc_mir_build::build::expr::as_rvalue RFC2229 = as_rvalue pushing following fake_read HirId { owner: DefId(0:3 ~ rox[317d]::main), local_id: 17 }
DEBUG rustc_mir_build::build::expr::as_rvalue RFC2229 = as_rvalue pushing following mir_place _2
DEBUG rustc_mir_build::build::expr::as_rvalue RFC2229 = as_rvalue.rs ExprKind::Closure fake_read processing over

arora-aman

Just things that I feel need to chage. Will do nits later

compiler/rustc_typeck/src/expr_use_visitor.rs

compiler/rustc_mir_build/src/thir/cx/expr.rs

compiler/rustc_mir_build/src/build/expr/as_rvalue.rs

compiler/rustc_mir_build/src/thir/cx/expr.rs

compiler/rustc_typeck/src/expr_use_visitor.rs

compiler/rustc_mir_build/src/build/expr/as_place.rs

compiler/rustc_typeck/src/expr_use_visitor.rs

compiler/rustc_mir_build/src/build/expr/as_place.rs

arora-aman · 2021-02-03T04:15:08Z

compiler/rustc_mir_build/src/build/expr/as_rvalue.rs

+                        let fake_read_upvar = this.hir.mirror(thir_place);
+                        let mir_place = unpack!(block = this.as_place(block, fake_read_upvar));
+                        debug!("RFC2229 = as_rvalue pushing following mir_place {:?}", mir_place);
+                        this.cfg.push_fake_read(block, source_info, FakeReadCause::ForLet, mir_place);


There are two fake read causes. One ForLet and one ForMatch. We probably should capture the correct one. I don't think it will be too much work to do that, but I think if this works fine for now, we should delay that until we fix bugs/ we get close to getting this to work.

compiler/rustc_typeck/src/check/upvar.rs

compiler/rustc_mir_build/src/thir/cx/expr.rs

compiler/rustc_typeck/src/check/upvar.rs

compiler/rustc_mir_build/src/build/expr/as_place.rs

src/test/ui/closures/2229_closure_analysis/run_pass/destructure_patterns-1.rs

compiler/rustc_mir_build/src/build/matches/mod.rs

compiler/rustc_mir_build/src/build/expr/as_place.rs

compiler/rustc_mir_build/src/thir/cx/expr.rs

compiler/rustc_typeck/src/expr_use_visitor.rs

arora-aman

initial set of comments

compiler/rustc_mir_build/src/thir/cx/expr.rs

compiler/rustc_mir_build/src/thir/mod.rs

arora-aman · 2021-02-25T20:50:18Z

compiler/rustc_typeck/src/check/upvar.rs

@@ -1102,6 +1108,7 @@ struct InferBorrowKind<'a, 'tcx> {
    /// Place { V1, [ProjectionKind::Field(Index=1, Variant=0)] } : CaptureKind { E2, MutableBorrow }
    /// ```
    capture_information: InferredCaptureInformation<'tcx>,
+    fake_reads: Vec<(Place<'tcx>, FakeReadCause)>,


Did you get a chance to try HashSet, because in case of

let c = || { let (t1, _) = t; let (_, t2) = t; }

we will introduce 2 fake reads, when one should suffice

compiler/rustc_typeck/src/expr_use_visitor.rs

arora-aman · 2021-02-25T20:52:40Z

compiler/rustc_typeck/src/expr_use_visitor.rs

@@ -608,6 +645,30 @@ impl<'a, 'tcx> ExprUseVisitor<'a, 'tcx> {
            ty::Closure(..) | ty::Generator(..)
        );

+        if let Some(fake_reads) = self.mc.typeck_results.closure_fake_reads.get(&closure_def_id) {


Note to self: Read this logic properly

compiler/rustc_mir_build/src/build/expr/as_place.rs

compiler/rustc_mir_build/src/build/expr/as_rvalue.rs

arora-aman · 2021-02-25T20:55:21Z

compiler/rustc_mir_build/src/build/expr/mod.rs

@@ -62,7 +62,7 @@

 mod as_constant;
 mod as_operand;
-mod as_place;
+pub mod as_place;


is this needed?

Yes,

error[E0603]: module `as_place` is private --> compiler/rustc_mir_build/src/build/mod.rs:2:25 | 2 | use crate::build::expr::as_place::PlaceBuilder; | ^^^^^^^^ private module | note: the module `as_place` is defined here --> compiler/rustc_mir_build/src/build/expr/mod.rs:65:1 | 65 | mod as_place; | ^^^^^^^^^^^^^ error[E0603]: module `as_place` is private --> compiler/rustc_mir_build/src/build/matches/mod.rs:8:25 | 8 | use crate::build::expr::as_place::PlaceBuilder; | ^^^^^^^^ private module | note: the module `as_place` is defined here --> compiler/rustc_mir_build/src/build/expr/mod.rs:65:1 | 65 | mod as_place; | ^^^^^^^^^^^^^ error[E0603]: module `as_place` is private --> compiler/rustc_mir_build/src/build/matches/simplify.rs:15:25 | 15 | use crate::build::expr::as_place::PlaceBuilder; | ^^^^^^^^ private module | note: the module `as_place` is defined here --> compiler/rustc_mir_build/src/build/expr/mod.rs:65:1 | 65 | mod as_place; | ^^^^^^^^^^^^^ error[E0603]: module `as_place` is private --> compiler/rustc_mir_build/src/build/matches/test.rs:8:25 | 8 | use crate::build::expr::as_place::PlaceBuilder; | ^^^^^^^^ private module | note: the module `as_place` is defined here --> compiler/rustc_mir_build/src/build/expr/mod.rs:65:1 | 65 | mod as_place; | ^^^^^^^^^^^^^ error[E0603]: module `as_place` is private --> compiler/rustc_mir_build/src/build/matches/util.rs:1:25 | 1 | use crate::build::expr::as_place::PlaceBuilder; | ^^^^^^^^ private module | note: the module `as_place` is defined here --> compiler/rustc_mir_build/src/build/expr/mod.rs:65:1 | 65 | mod as_place; | ^^^^^^^^^^^^^ error: aborting due to 5 previous errors

roxelo · 2021-02-25T23:34:22Z

Opened PR on rust-lang rust-lang#82536

arora-aman reviewed Jan 30, 2021

View reviewed changes

roxelo force-pushed the handle-patterns-take-2 branch from 194b37d to cadf23b Compare January 30, 2021 23:20

arora-aman requested changes Jan 30, 2021

View reviewed changes

compiler/rustc_mir_build/src/thir/cx/expr.rs Outdated Show resolved Hide resolved

compiler/rustc_mir_build/src/thir/cx/expr.rs Outdated Show resolved Hide resolved

compiler/rustc_typeck/src/expr_use_visitor.rs Outdated Show resolved Hide resolved

roxelo force-pushed the handle-patterns-take-2 branch from cadf23b to 92ae1ce Compare January 31, 2021 07:06

arora-aman reviewed Jan 31, 2021

View reviewed changes

compiler/rustc_mir_build/src/build/expr/as_place.rs Outdated Show resolved Hide resolved

roxelo force-pushed the handle-patterns-take-2 branch from 92ae1ce to 5f854d3 Compare February 3, 2021 02:08

arora-aman requested changes Feb 3, 2021

View reviewed changes

roxelo force-pushed the handle-patterns-take-2 branch 6 times, most recently from 751a599 to 84157b4 Compare February 19, 2021 05:27

arora-aman reviewed Feb 21, 2021

View reviewed changes

roxelo force-pushed the handle-patterns-take-2 branch 2 times, most recently from 523a104 to 31e3451 Compare February 25, 2021 20:40

arora-aman reviewed Feb 25, 2021

View reviewed changes

roxelo changed the title ~~Handle patterns within closures [WIP]~~ Handle patterns within closures Feb 25, 2021

roxelo force-pushed the handle-patterns-take-2 branch 3 times, most recently from d566cb4 to 9aca998 Compare March 1, 2021 02:11

roxelo force-pushed the handle-patterns-take-2 branch 3 times, most recently from 06ff810 to 459ebf8 Compare March 12, 2021 04:42

roxelo added 5 commits March 14, 2021 16:31

Introduce new fake reads

ec10b71

Delay use of Place in favor of PlaceBuilder

d4f8729

Attempt to deal with nested closures properly

b6cf070

Use the correct FakeReadCause

685a4c6

Only borrow place for matching under specific conditions

74fc643

Add fake_read() to clippy

fb3b77a

roxelo force-pushed the handle-patterns-take-2 branch from 459ebf8 to 88de9c2 Compare March 15, 2021 04:36

roxelo added 2 commits March 15, 2021 13:16

Add comments with examples and tests

22eaffe

Fix error after rebase

189d206

roxelo force-pushed the handle-patterns-take-2 branch from 88de9c2 to 189d206 Compare March 15, 2021 17:23

sexxi-bot merged commit f5d8117 into master Mar 22, 2021

Handle patterns within closures #45

Handle patterns within closures #45

Uh oh!

Conversation

roxelo commented Jan 30, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

arora-aman left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

arora-aman Feb 3, 2021

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

arora-aman left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

arora-aman Feb 25, 2021

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

arora-aman Feb 25, 2021

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

arora-aman Feb 25, 2021

Choose a reason for hiding this comment

Uh oh!

roxelo Feb 25, 2021

Choose a reason for hiding this comment

Uh oh!

roxelo commented Feb 25, 2021

Uh oh!

Uh oh!

roxelo commented Jan 30, 2021 •

edited

Loading