TypeVisitor: use `std::ops::ControlFlow` instead of `bool` #78182

LeSeulArtichaut · 2020-10-21T12:33:57Z

Implements MCP rust-lang/compiler-team#374.

Blocked on FCP in rust-lang/compiler-team#374.
r? @lcnr cc @jonas-schievink

lcnr

Thanks.

Some nitpicks and suggestion but this is already looking a lot clearer to me ❤️

I do have to admit that my brain ended up skipping over some bool -> ControlFlow conversions so hopefully I didn't accidentially miss anything 😅

compiler/rustc_middle/src/macros.rs

lcnr · 2020-10-21T13:12:03Z

compiler/rustc_middle/src/mir/type_foldable.rs

            }
            Assert { ref cond, ref msg, .. } => {
-                if cond.visit_with(visitor) {
+                if cond.visit_with(visitor) == ControlFlow::BREAK {


that's weird, why are we only visiting here if cond.visit_with(...) returns true?

Is this a preexisting bug? cc @oli-obk @jonas-schievink maybe

yea this looks wrong. I don't know why it would be this way around. I mean this is essentially cond.visit_with(visitor) && {...} which is not something that is ever used anywhere

lcnr · 2020-10-21T13:14:00Z

compiler/rustc_middle/src/ty/fold.rs

@@ -74,7 +76,7 @@ pub trait TypeFoldable<'tcx>: fmt::Debug + Clone {
    }

    fn has_type_flags(&self, flags: TypeFlags) -> bool {
-        self.visit_with(&mut HasTypeFlagsVisitor { flags })
+        self.visit_with(&mut HasTypeFlagsVisitor { flags }) == ControlFlow::Break(())


Yeah, it might make sense to introduce a newtype here in the future, so this can be visit_with(...).break_value() == Some(FoundFlags).

Not part of this PR though

compiler/rustc_middle/src/ty/fold.rs

compiler/rustc_privacy/src/lib.rs

lcnr · 2020-10-21T13:35:13Z

compiler/rustc_trait_selection/src/traits/structural_match.rs

@@ -258,16 +248,16 @@ impl<'a, 'tcx> TypeVisitor<'tcx> for Search<'a, 'tcx> {
            let ty = self.tcx().normalize_erasing_regions(ty::ParamEnv::empty(), field_ty);
            debug!("structural-match ADT: field_ty={:?}, ty={:?}", field_ty, ty);

-            if ty.visit_with(self) {
+            if ty.visit_with(self) == ControlFlow::BREAK {


yeah, we probably want to store self.found in ControlFlow::Break here.

(also not part of this PR)

lcnr · 2020-10-21T13:40:19Z

compiler/rustc_typeck/src/check/wfcheck.rs

@@ -798,26 +800,26 @@ fn check_where_clauses<'tcx, 'fcx>(
                params: FxHashSet<u32>,
            }
            impl<'tcx> ty::fold::TypeVisitor<'tcx> for CountParams {


That folder seems weird to me, why are we storing the params in a hashmap if we only care about if there is more than one of them?

Using an option is probably more appropriate here but whatever 🤷 Isn't really relevant for this PR

lcnr · 2020-10-21T13:41:55Z

src/tools/clippy/clippy_lints/src/redundant_clone.rs

-    fn visit_region(&mut self, _: ty::Region<'_>) -> bool {
-        true
+    fn visit_region(&mut self, _: ty::Region<'_>) -> ControlFlow<(), ()> {
+        ControlFlow::BREAK


Interesting that we aren't ignoring ReStatic here.

compiler/rustc_middle/src/mir/type_foldable.rs

compiler/rustc_middle/src/ty/subst.rs

compiler/rustc_mir/src/monomorphize/polymorphize.rs

compiler/rustc_privacy/src/lib.rs

oli-obk · 2020-10-21T16:17:51Z

compiler/rustc_privacy/src/lib.rs

-    fn visit_def_id(&mut self, def_id: DefId, kind: &str, descr: &dyn fmt::Display) -> bool {
-        self.check_def_id(def_id, kind, descr)
+    fn visit_def_id(&mut self, def_id: DefId, kind: &str, descr: &dyn fmt::Display) -> ControlFlow<(), ()> {
+        if self.check_def_id(def_id, kind, descr) {


these are very common. Maybe we should have a ControlFlow::break_if(some_bool) and ControlFlow::continue_if(some_bool) for convenience? Or will that become infeasible with types other than ()?

yeah, I think adding these methods make sense. I am not yet sure how to generalize them to deal with types other than ControlFlow<(), ()>

There has been almost no thought put into what methods the ControlFlow type should have, so if you come up with reasonable ones definitely add them. It used to be an internal implementation detail, so doesn't even have the usual combinators.

@scottmcm I'd say definitely is_break and is_continue (name to be bikeshed of course). I'll file a PR

oli-obk · 2020-10-21T16:18:51Z

compiler/rustc_trait_selection/src/opaque_types.rs

@@ -691,26 +693,26 @@ impl<'tcx, OP> TypeVisitor<'tcx> for ConstrainOpaqueTypeRegionVisitor<OP>
 where
    OP: FnMut(ty::Region<'tcx>),
 {
-    fn visit_binder<T: TypeFoldable<'tcx>>(&mut self, t: &ty::Binder<T>) -> bool {
+    fn visit_binder<T: TypeFoldable<'tcx>>(&mut self, t: &ty::Binder<T>) -> ControlFlow<(), ()> {
        t.as_ref().skip_binder().visit_with(self);


ControlFlow should probably be #[must_use] so we'd be forced to use let _ = here, but that's for another PR

Yeah, making it #[must_use] seems like a good idea to me

I missed this in my review, but it seems weird to not propagate ControlFlow::Break here, I guess that this visitor never actually wants to break early so it is fine but it does seem strange.

Another reason for custom break types as we could use ! here to make this part of the API

Opened #78202

Another thing you could consider to avoid the conversation on the type would be to just mark the methods on the trait as #[must_use]. It looks like that works even for calls on concrete things that don't copy the attribute to their implementation. Probably still not this PR, though.

oli-obk · 2020-10-21T16:20:52Z

compiler/rustc_typeck/src/check/check.rs

            debug!("check_opaque_for_inheriting_lifetimes: (visit_ty) t={:?}", t);
-            if t != self.opaque_identity_ty && t.super_visit_with(self) {
+            if t != self.opaque_identity_ty && t.super_visit_with(self) == ControlFlow::BREAK {


This alos looks like a candidate for returning a value via Break

oli-obk · 2020-10-21T16:22:36Z

This PR is awesome (and exposes so many issues of different sublety

oli-obk · 2020-10-23T16:53:49Z

@bors r+

bors · 2020-10-23T16:53:51Z

📌 Commit 901c1f446e22f1d067d2c0948b95fdd5ebe3d39d has been approved by oli-obk

oli-obk · 2020-10-23T16:55:42Z

@bors r- I guess we should wait on the FCP 😅

LeSeulArtichaut · 2020-10-30T18:09:51Z

@oli-obk @lcnr The MCP finished its 10 days FCP

oli-obk · 2020-10-30T18:54:08Z

@bory r=lcnr,oli-obk

lcnr · 2020-10-30T19:58:19Z

@bors r=lcnr,oli-obk rollup=never

bors · 2020-10-30T19:58:20Z

📌 Commit 9433eb8 has been approved by lcnr,oli-obk

bors · 2020-10-30T22:54:05Z

⌛ Testing commit 9433eb8 with merge 0d033de...

bors · 2020-10-31T01:34:43Z

☀️ Test successful - checks-actions
Approved by: lcnr,oli-obk
Pushing 0d033de to master...

fix `super_visit_with` for `Terminator` fixes rust-lang#78182 (comment) r? `@oli-obk` cc `@LeSeulArtichaut`

…m-basics, r=m-ou-se Stabilize `ops::ControlFlow` (just the type) Tracking issue: rust-lang#75744 (which also tracks items *not* closed by this PR). With the new `?` desugar implemented, [it's no longer possible to mix `Result` and `ControlFlow`](https://play.rust-lang.org/?version=nightly&mode=debug&edition=2018&gist=13feec97f5c96a9d791d97f7de2d49a6). (At the time of making this PR, godbolt was still on the 2021-05-01 nightly, where you can see that [the mixing example compiled](https://rust.godbolt.org/z/13Ke54j16).) That resolves the only blocker I know of, so I'd like to propose that `ControlFlow` be considered for stabilization. Its basic existence was part of rust-lang/rfcs#3058, where it got a bunch of positive comments (examples [1](rust-lang/rfcs#3058 (comment)) [2](rust-lang/rfcs#3058 (review)) [3](rust-lang/rfcs#3058 (comment)) [4](rust-lang/rfcs#3058 (comment))). Its use in the compiler has been well received (rust-lang#78182 (comment)), and there are ecosystem updates interested in using it (rust-itertools/itertools#469 (comment), jonhoo/rust-imap#194). As this will need an FCP, picking a libs member manually: r? `@m-ou-se` ## Stabilized APIs ```rust #[derive(Debug, Clone, Copy, PartialEq)] pub enum ControlFlow<B, C = ()> { /// Exit the operation without running subsequent phases. Break(B), /// Move on to the next phase of the operation as normal. Continue(C), } ``` As well as using `?` on a `ControlFlow<B, _>` in a function returning `ControlFlow<B, _>`. (Note, in particular, that there's no `From::from`-conversion on the `Break` value, the way there is for `Err`s.) ## Existing APIs *not* stabilized here All the associated methods and constants: `break_value`, `is_continue`, `map_break`, [`CONTINUE`](https://doc.rust-lang.org/nightly/std/ops/enum.ControlFlow.html#associatedconstant.CONTINUE), etc. Some of the existing methods in nightly seem reasonable, some seem like they should be removed, and some need more discussion to decide. But none of them are *essential*, so [as in the RFC](https://rust-lang.github.io/rfcs/3058-try-trait-v2.html#methods-on-controlflow), they're all omitted from this PR. They can be considered separately later, as further usage demonstrates which are important.

LeSeulArtichaut added C-cleanup Category: PRs that clean code up or issues documenting cleanup. T-compiler Relevant to the compiler team, which will review and decide on the PR/issue. S-waiting-on-fcp Status: PR is in FCP and is awaiting for FCP to complete. labels Oct 21, 2020

rust-highfive assigned lcnr Oct 21, 2020

rust-highfive added the S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. label Oct 21, 2020

lcnr approved these changes Oct 21, 2020

View reviewed changes

LeSeulArtichaut force-pushed the ty-visitor-contolflow branch 3 times, most recently from 6d4777b to 8e58212 Compare October 21, 2020 15:31

oli-obk reviewed Oct 21, 2020

View reviewed changes

compiler/rustc_middle/src/mir/type_foldable.rs Outdated Show resolved Hide resolved

oli-obk reviewed Oct 21, 2020

View reviewed changes

compiler/rustc_middle/src/mir/type_foldable.rs Outdated Show resolved Hide resolved

LeSeulArtichaut force-pushed the ty-visitor-contolflow branch from 8e58212 to e0f17c2 Compare October 21, 2020 16:08

oli-obk reviewed Oct 21, 2020

View reviewed changes

LeSeulArtichaut force-pushed the ty-visitor-contolflow branch from e0f17c2 to 3ef1e8f Compare October 21, 2020 18:46

LeSeulArtichaut mentioned this pull request Oct 21, 2020

Make ControlFlow #[must_use] #78202

Closed

LeSeulArtichaut force-pushed the ty-visitor-contolflow branch from 3ef1e8f to e14a5d8 Compare October 21, 2020 20:20

LeSeulArtichaut mentioned this pull request Oct 21, 2020

Tracking issue for ControlFlow enum, for use with try_fold and in Try #75744

Closed

9 tasks

This comment has been minimized.

Sign in to view

LeSeulArtichaut force-pushed the ty-visitor-contolflow branch from e14a5d8 to 901c1f4 Compare October 22, 2020 08:20

bors added S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels Oct 23, 2020

bors added S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. and removed S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. labels Oct 23, 2020

LeSeulArtichaut removed the S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. label Oct 23, 2020

This comment has been minimized.

Sign in to view

bors added the S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. label Oct 24, 2020

LeSeulArtichaut added 4 commits October 30, 2020 12:27

TypeVisitor: use ControlFlow in rustc_{mir,privacy,traits,typeck}

61f8182

TypeVisitor: use ControlFlow in clippy

8e4cf0b

Use ControlFlow::is{break,continue}

24e1a7e

Remove implicit Continue type

9433eb8

LeSeulArtichaut force-pushed the ty-visitor-contolflow branch from 3244777 to 9433eb8 Compare October 30, 2020 11:28

This comment has been minimized.

Sign in to view

bors added S-waiting-on-bors Status: Waiting on bors to run and complete tests. Bors will change the label on completion. and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels Oct 30, 2020

bors added the merged-by-bors This PR was explicitly merged by bors. label Oct 31, 2020

bors merged commit 0d033de into rust-lang:master Oct 31, 2020

rustbot added this to the 1.49.0 milestone Oct 31, 2020

This was referenced Oct 31, 2020

polymorphize: remove predicate logic #75737

Closed

Add built-in implementations of Default for function definition and… #77688

Closed

LeSeulArtichaut deleted the ty-visitor-contolflow branch October 31, 2020 08:14

This was referenced Oct 31, 2020

Use ControlFlow in Visitor rust-lang/chalk#645

Merged

TypeVisitor: do not hard-code a ControlFlow<()> result rust-lang/compiler-team#383

Closed

Introduce TypeVisitor::BreakTy #78779

Merged

lcnr mentioned this pull request Nov 7, 2020

fix super_visit_with for Terminator #78830

Merged

Dylan-DPC-zz pushed a commit to Dylan-DPC-zz/rust that referenced this pull request Nov 9, 2020

Rollup merge of rust-lang#78830 - lcnr:mir-folder, r=oli-obk

7924ecc

fix `super_visit_with` for `Terminator` fixes rust-lang#78182 (comment) r? `@oli-obk` cc `@LeSeulArtichaut`

scottmcm mentioned this pull request May 23, 2021

Stabilize ops::ControlFlow (just the type) #85608

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

TypeVisitor: use `std::ops::ControlFlow` instead of `bool` #78182

TypeVisitor: use `std::ops::ControlFlow` instead of `bool` #78182

LeSeulArtichaut commented Oct 21, 2020

lcnr left a comment

lcnr Oct 21, 2020 •

edited

Loading

oli-obk Oct 21, 2020

lcnr Oct 21, 2020

lcnr Oct 21, 2020

lcnr Oct 21, 2020

lcnr Oct 21, 2020

oli-obk Oct 21, 2020

lcnr Oct 21, 2020 •

edited

Loading

scottmcm Oct 21, 2020

LeSeulArtichaut Oct 21, 2020

oli-obk Oct 21, 2020

lcnr Oct 21, 2020

lcnr Oct 21, 2020 •

edited

Loading

LeSeulArtichaut Oct 21, 2020

scottmcm Oct 23, 2020

oli-obk Oct 21, 2020

oli-obk commented Oct 21, 2020

This comment has been minimized.

oli-obk commented Oct 23, 2020

bors commented Oct 23, 2020

oli-obk commented Oct 23, 2020

This comment has been minimized.

LeSeulArtichaut commented Oct 30, 2020

oli-obk commented Oct 30, 2020

This comment has been minimized.

lcnr commented Oct 30, 2020

bors commented Oct 30, 2020

bors commented Oct 30, 2020

bors commented Oct 31, 2020

TypeVisitor: use std::ops::ControlFlow instead of bool #78182

TypeVisitor: use std::ops::ControlFlow instead of bool #78182

Conversation

LeSeulArtichaut commented Oct 21, 2020

lcnr left a comment

Choose a reason for hiding this comment

lcnr Oct 21, 2020 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

lcnr Oct 21, 2020 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

lcnr Oct 21, 2020 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

oli-obk commented Oct 21, 2020

This comment has been minimized.

oli-obk commented Oct 23, 2020

bors commented Oct 23, 2020

oli-obk commented Oct 23, 2020

This comment has been minimized.

LeSeulArtichaut commented Oct 30, 2020

oli-obk commented Oct 30, 2020

This comment has been minimized.

lcnr commented Oct 30, 2020

bors commented Oct 30, 2020

bors commented Oct 30, 2020

bors commented Oct 31, 2020

TypeVisitor: use `std::ops::ControlFlow` instead of `bool` #78182

TypeVisitor: use `std::ops::ControlFlow` instead of `bool` #78182

lcnr Oct 21, 2020 •

edited

Loading

lcnr Oct 21, 2020 •

edited

Loading

lcnr Oct 21, 2020 •

edited

Loading