Rename and expose LoopState as ControlFlow #76204

NoraCodes · 2020-09-01T16:46:21Z

Basic PR for #75744. Addresses everything there except for documentation; lots of examples are probably a good idea.

rust-highfive · 2020-09-01T16:46:24Z

Thanks for the pull request, and welcome! The Rust team is excited to review your changes, and you should hear from @dtolnay (or someone else) soon.

If any changes to this PR are deemed necessary, please add them as extra commits. This ensures that the reviewer can see what has changed since they last reviewed the code. Due to the way GitHub handles out-of-date commits, this should also make it reasonably obvious what issues have or haven't been addressed. Large or tricky changes may require several passes of review and changes.

Please see the contribution instructions for more information.

NoraCodes · 2020-09-01T16:46:38Z

r? @scottmcm

scottmcm · 2020-09-01T18:36:21Z

Ping @rust-lang/libs to check that this is a plausible addition (as unstable, so no full FCP needed):

This is helpful for writing try_fold closures that are more readable, as seen in the implementation of the iterator methods. And in general, ? is convenient for more than just errors -- for example this bit of itertools uses Err for success because it has the desired short-circuiting behaviour, but that would be far clearer with Break-for-success.
And in particular, this is good for graph traversals, both in NoraCodes's original request and as ecstatic-morse mentioned is used in the compiler
I would also like to have this enum for use in an a tweaked Try definition.

scottmcm · 2020-09-01T18:38:46Z

@NoraCodes Looks like tidy is complaining:

 tidy error: /checkout/library/core/src/iter/mod.rs: too many trailing newlines (2)
some tidy checks failed

Can you fix that, rebase to upstream/master (to avoid the merge conflict that kept CI from running initially), and force-push the new change, please?

ecstatic-morse

Thanks @NoraCodes! I left some minor notes/suggestions. You're free to ignore them until someone from the libs team weighs in.

ecstatic-morse · 2020-09-01T18:48:26Z

library/core/src/ops/control_flow.rs

+
+/// Used to make try_fold closures more like normal loops
+#[unstable(feature="control_flow_enum", reason="new API", issue="75744")]
+#[derive(Debug, Clone, Copy, PartialEq)]


Suggested change

#[derive(Debug, Clone, Copy, PartialEq)]

#[derive(Debug, Clone, Copy, PartialEq, Eq, Hash)]

Hash and especially Eq seem handy, although they are not needed to replace LoopState and are less essential for ControlFlow than for Option.

Yes, that seems sensible! I don't know when you'd want to hash this but I suppose there's no reason to prevent it.

ecstatic-morse · 2020-09-01T18:48:34Z

library/core/src/ops/control_flow.rs

+/// Used to make try_fold closures more like normal loops
+#[unstable(feature="control_flow_enum", reason="new API", issue="75744")]
+#[derive(Debug, Clone, Copy, PartialEq)]
+pub enum ControlFlow<C, B> {


I have found that C = () is very common when using this enum, common enough that it might deserve a defaulted type parameter. We could add one backwards compatibly but only if C follows B in the parameter list. Otherwise we would have to default B at the same time and anyone that wishes to override it would have to override C as well.

Good point! It's even usually that way in the uses this PR touches, like

I suppose we could even do ControlFlow<B = (), C = ()> the same way break; is break ();.

(Though I suppose the reorder might make the diff messier, so could also be a separate PR.)

I think it would be confusing to do this without changing the order of the variants as well, and we may as well do that too.

On consideration, I think a separate PR would be best.

ecstatic-morse · 2020-09-01T18:52:21Z

library/core/src/ops/control_flow.rs

+    }
+}
+
+impl<R: Try> ControlFlow<R::Ok, R> {


Shouldn't the second parameter be R::Error?

Yes, definitely.

Oh, I see - we can't do this because it's an unconstrained type parameter. Is there a good way to solve that?

No, the break part here is intentionally the full impl Try type (Result, etc). They're definitely strange APIs, though -- break_value is something that could plausibly stabilize, but these ones probably aren't.

Maybe leave this particular impl block over in iter as non-pub so they can still be used there, but don't show up in the rustdoc? (And won't need #[unstable] since they'll be unusable outside iter.)

I've been contemplating making a pass through the implementations to use feature(try_blocks) instead of explicit Try::foo calls where possible; this PR going in would be a good impetus to go do that -- and hopefully delete these methods while I'm at it.

Oh, I see. I misunderstood how these were being used. You're right, it's a very strange API.

library/core/src/iter/mod.rs

scottmcm · 2020-09-02T21:46:04Z

r=me for the code; waiting on a "seems plausible" comment from a libs member.

dtolnay · 2020-09-02T23:18:02Z

Looks plausible, so @bors r=scottmcm

I agree with #75744 (comment) that "a Break/Continue enum is the correct abstraction for the Try trait instead of Result". As I understand it, while this type is structurally identical to Result, its API is less expansive and better geared toward readable code in uses of try_fold, in a way that use Result::{self as ControlFlow, Ok as Continue, Err as Break}; wouldn't quite accomplish.

bors · 2020-09-02T23:18:04Z

📌 Commit 96eb5e1 has been approved by scottmcm

@ghost

Rollup of 12 pull requests Successful merges: - rust-lang#75150 (Add a note for Ipv4Addr::to_ipv6_compatible) - rust-lang#76120 (Add `[T; N]::as_[mut_]slice`) - rust-lang#76142 (Make all methods of `std::net::Ipv4Addr` const) - rust-lang#76164 (Link to slice pattern in array docs) - rust-lang#76167 (Replace MinGW library hack with heuristic controlling link mode) - rust-lang#76204 (Rename and expose LoopState as ControlFlow) - rust-lang#76238 (Move to intra-doc links for library/core/src/iter/traits/iterator.rs) - rust-lang#76242 (Read: adjust a FIXME reference) - rust-lang#76243 (Fix typos in vec try_reserve(_exact) docs) - rust-lang#76245 (inliner: Avoid query cycles when optimizing generators) - rust-lang#76255 (Update books) - rust-lang#76261 (Use intra-doc links in `core::marker`) Failed merges: r? @ghost

Demote `ControlFlow::{from|into}_try` to `pub(crate)` They have mediocre names and non-obvious semantics, so personally I don't think they're worth trying to stabilize, and thus might as well just be internal (they're used for convenience in iterator adapters), not something shown in the rustdocs. I don't think anyone actually wanted to use them outside `core` -- they just got made public-but-unstable along with the whole type in rust-lang#76204 that promoted `LoopState` from an internal type to the exposed `ControlFlow` type. cc rust-lang#75744, the tracking issue they mention.

Demote `ControlFlow::{from|into}_try` to `pub(crate)` They have mediocre names and non-obvious semantics, so personally I don't think they're worth trying to stabilize, and thus might as well just be internal (they're used for convenience in iterator adapters), not something shown in the rustdocs. I don't think anyone actually wanted to use them outside `core` -- they just got made public-but-unstable along with the whole type in rust-lang#76204 that promoted `LoopState` from an internal type to the exposed `ControlFlow` type. cc rust-lang#75744, the tracking issue they mention. cc rust-lang#85608, the PR where I'm proposing stabilizing the type.

rust-highfive assigned dtolnay Sep 1, 2020

rust-highfive added the S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. label Sep 1, 2020

rust-highfive assigned scottmcm and unassigned dtolnay Sep 1, 2020

jyn514 added the T-libs-api Relevant to the library API team, which will review and decide on the PR/issue. label Sep 1, 2020

ecstatic-morse reviewed Sep 1, 2020

View reviewed changes

Rename and expose LoopState as ControlFlow

d0af125

NoraCodes force-pushed the nora/control_flow_enum branch from 57c720f to d0af125 Compare September 1, 2020 21:02

scottmcm reviewed Sep 1, 2020

View reviewed changes

library/core/src/iter/mod.rs Outdated Show resolved Hide resolved

Format ControlFlow changes with rustfmt

96eb5e1

scottmcm added S-waiting-on-team Status: Awaiting decision from the relevant subteam (see the T-<team> label). and removed S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. labels Sep 2, 2020

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-team Status: Awaiting decision from the relevant subteam (see the T-<team> label). labels Sep 2, 2020

NoraCodes mentioned this pull request Sep 3, 2020

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

Open

9 tasks

Dylan-DPC-zz mentioned this pull request Sep 3, 2020

Rollup of 12 pull requests #76265

Merged

bors merged commit d059f26 into rust-lang:master Sep 3, 2020

scottmcm mentioned this pull request May 24, 2021

Demote ControlFlow::{from|into}_try to pub(crate) #85645

Merged

cuviper added this to the 1.48.0 milestone Nov 16, 2023

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Rename and expose LoopState as ControlFlow #76204

Rename and expose LoopState as ControlFlow #76204

NoraCodes commented Sep 1, 2020

rust-highfive commented Sep 1, 2020

NoraCodes commented Sep 1, 2020

scottmcm commented Sep 1, 2020

scottmcm commented Sep 1, 2020

ecstatic-morse left a comment

ecstatic-morse Sep 1, 2020

NoraCodes Sep 1, 2020

ecstatic-morse Sep 1, 2020

scottmcm Sep 1, 2020

NoraCodes Sep 1, 2020

NoraCodes Sep 2, 2020

ecstatic-morse Sep 1, 2020 •

edited

Loading

NoraCodes Sep 1, 2020

NoraCodes Sep 1, 2020

scottmcm Sep 2, 2020 •

edited

Loading

NoraCodes Sep 2, 2020

scottmcm commented Sep 2, 2020 •

edited

Loading

dtolnay commented Sep 2, 2020

bors commented Sep 2, 2020

	#[derive(Debug, Clone, Copy, PartialEq)]
	#[derive(Debug, Clone, Copy, PartialEq, Eq, Hash)]

Rename and expose LoopState as ControlFlow #76204

Rename and expose LoopState as ControlFlow #76204

Conversation

NoraCodes commented Sep 1, 2020

rust-highfive commented Sep 1, 2020

NoraCodes commented Sep 1, 2020

scottmcm commented Sep 1, 2020

scottmcm commented Sep 1, 2020

ecstatic-morse left a 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

Choose a reason for hiding this comment

ecstatic-morse Sep 1, 2020 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

scottmcm Sep 2, 2020 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

scottmcm commented Sep 2, 2020 • edited Loading

dtolnay commented Sep 2, 2020

bors commented Sep 2, 2020

ecstatic-morse Sep 1, 2020 •

edited

Loading

scottmcm Sep 2, 2020 •

edited

Loading

scottmcm commented Sep 2, 2020 •

edited

Loading