Add const generics to infer (and transitive dependencies) #59008

varkor · 2019-03-08T01:24:21Z

Split out from #53645. This work is a collaborative effort with @yodaldevoid.

There are a number of stubs. These are mainly to ensure we don't overlook them when completing the implementation, but are not necessary for the initial implementation. We plan to address these in follow up PRs.

r? @eddyb / @nikomatsakis

eddyb

LGTM, modulo comments, some of which require someone who understands certain typesystem interactions - I've pinged @nikomatsakis but I suspect @rust-lang/wg-traits can also help here.

src/librustc/infer/canonical/mod.rs

src/librustc/infer/combine.rs

src/librustc/infer/const_variable.rs

src/librustc/infer/mod.rs

src/librustc/ty/error.rs

eddyb · 2019-03-12T18:51:20Z

src/librustc/ty/relate.rs

+                }
+            }
+        }
+        // FIXME(const_generics): this is probably wrong (regarding TyProjection)


This is wrong because it is a projection, like TyProjection. cc @nikomatsakis

Well, to be quite honest, I'm a bit confused by what the "normalization policy" is here. If this were doing eager normalization, as I would expect at the moment, then this doesn't seem wrong. But if it's lazy normalization, then indeed it is wrong. It seems like we can land it as is, but I think that overall it would be good to try and document and talk out the design of constants and const generics.

Design meeting! :)

src/librustc/infer/const_variable.rs

eddyb · 2019-03-13T09:15:29Z

src/librustc/infer/const_variable.rs

+#[derive(Copy, Clone, Debug)]
+pub enum ConstVariableValue<'tcx> {
+    Known { value: &'tcx ty::LazyConst<'tcx> },
+    Unknown { universe: ty::UniverseIndex },


Is this enough? Don't we also need the type?

Well, I think the type is carried in the ty::Const wrapper, for better or worse:

#[inline] pub fn mk_const_var(self, v: ConstVid<'tcx>, ty: Ty<'tcx>) -> &'tcx Const<'tcx> { self.mk_const(ty::Const { val: ConstValue::Infer(InferConst::Var(v)), ty, }) }

It's not carried around in Unknown, which I think is the part @eddyb was referring to. (There's some other discussion about this elsewhere, but there's a FIXME in the code about this now.)

src/librustc/infer/const_variable.rs

eddyb · 2019-03-13T09:29:21Z

src/librustc/infer/canonical/mod.rs

+                self.tcx.mk_lazy_const(ty::LazyConst::Evaluated(
+                    ty::Const {
+                        val: ConstValue::Placeholder(placeholder_mapped),
+                        ty: self.tcx.types.err, // FIXME(const_generics)


Sounds like PlaceholderConst should contain the type.

Yeah, it probably should. The reason I didn't do it earlier was it ended up propagating <'tcx> everywhere and I wanted to make sure it was definitely necessary before going down that route.

I attempted propagating <'tcx> around when originally adding CanonicalVarKind::Const and found it to be impossible because CanonicalVarInfos are allocated in the global arena.

Sound like we need to bother @nikomatsakis.

But also, aren't const types supposed to be always be "global"?

AFAIK, types are only global after they have been lifted and their dependencies on type parameters and such are taken care of.

Type parameters are "global" in this sense, only inference variables aren't.
I'm saying that I don't think we should ever have a constant value in the type system with an uninferred type.

Hmm. Something feels wrong here indeed. I'm trying too put my finger on what it is. I guess in short I don't like the idea of there being a type in the canonical-var-info -- I feel like types should in the canonicalized value.

I think what I would expect is that the placeholder doesn't replace the entire ty::Const, but rather just the ConstValue. In this way, it can be instantiated without needing the type.

(Similarly, in the existential case that appears above, you wouldn't make a "Fresh type variable" to represent its type -- that also seems wrong.)

bors · 2019-03-13T15:10:45Z

☔ The latest upstream changes (presumably #56864) made this pull request unmergeable. Please resolve the merge conflicts.

varkor · 2019-03-18T21:02:37Z

This is currently blocked on rust-lang/ena#19 and response from @nikomatsakis regarding the review comments above.

crlf0710 · 2019-03-19T17:08:12Z

ena 0.12.0 is released.

varkor · 2019-03-20T10:11:14Z

I missed a detail in the previous pull request to ena: this is now blocked on rust-lang/ena#21.

bors · 2019-03-24T18:00:05Z

☔ The latest upstream changes (presumably #58305) made this pull request unmergeable. Please resolve the merge conflicts.

eddyb · 2019-03-25T14:15:15Z

@varkor This should be unblocked now, just released ena v0.13.

varkor · 2019-03-25T16:00:32Z

This is now blocked on #59415.

varkor · 2019-03-27T17:54:33Z

I've addressed all the comments that aren't waiting on feedback from @nikomatsakis.

Centril · 2019-03-30T13:07:58Z

Reassigning the remainder of the review to r? @nikomatsakis then.

eddyb · 2019-03-30T13:28:27Z

Anyone else from @rust-lang/wg-traits want to take a look at the questions I left for @nikomatsakis, too?

WiSaGaN · 2019-04-12T05:07:46Z

Does this still block on @nikomatsakis and @rust-lang/wg-traits?

crlf0710 · 2019-04-12T09:20:54Z

@WiSaGaN Yes, althrough niko has responded that he'll add it to his list last week:
https://rust-lang.zulipchat.com/#narrow/stream/144729-wg-traits/topic/.2359008.20changes.20to.20infer

Co-Authored-By: Gabriel Smith <yodaldevoid@users.noreply.github.com>

eddyb · 2019-05-01T23:49:33Z

@bors r+

bors · 2019-05-01T23:49:34Z

📌 Commit a68ed06 has been approved by eddyb

bors · 2019-05-02T04:47:43Z

⌛ Testing commit a68ed06 with merge 92b5e20...

@yodaldevoid

Add const generics to infer (and transitive dependencies) Split out from #53645. This work is a collaborative effort with @yodaldevoid. There are a number of stubs. These are mainly to ensure we don't overlook them when completing the implementation, but are not necessary for the initial implementation. We plan to address these in follow up PRs. r? @eddyb / @nikomatsakis

bors · 2019-05-02T07:38:23Z

☀️ Test successful - checks-travis, status-appveyor
Approved by: eddyb
Pushing 92b5e20 to master...

@eddyb

… r=eddyb Refactor `TypeVariableOrigin` Removes some unused variants and extracts the common `Span` field. As suggested in rust-lang#59008 (comment). r? @eddyb

@eddyb

Refactor `TypeVariableOrigin` Removes some unused variants and extracts the common `Span` field. As suggested in #59008 (comment). r? @eddyb

rust-highfive assigned eddyb Mar 8, 2019

rust-highfive added the S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. label Mar 8, 2019

varkor mentioned this pull request Mar 8, 2019

The Genesis of Generic Germination #53645

Merged

HadrienG2 mentioned this pull request Mar 12, 2019

Tracking issue for const generics (RFC 2000) #44580

Closed

16 tasks

eddyb suggested changes Mar 12, 2019

View reviewed changes

eddyb reviewed Mar 13, 2019

View reviewed changes

src/librustc/infer/const_variable.rs Outdated Show resolved Hide resolved

eddyb reviewed Mar 13, 2019

View reviewed changes

src/librustc/infer/const_variable.rs Outdated Show resolved Hide resolved

eddyb reviewed Mar 13, 2019

View reviewed changes

src/librustc/infer/const_variable.rs Outdated Show resolved Hide resolved

eddyb reviewed Mar 13, 2019

View reviewed changes

varkor force-pushed the const-generics-infer branch from e1d5810 to 94b7984 Compare March 18, 2019 21:01

This comment has been minimized.

Sign in to view

varkor mentioned this pull request Mar 19, 2019

Factor out "value" variants of ConstValue #59210

Closed

varkor mentioned this pull request Mar 21, 2019

Refactor TypeVariableOrigin #59331

Merged

varkor mentioned this pull request Mar 26, 2019

Refactor InferenceFudger (née RegionFudger) #59415

Merged

varkor force-pushed the const-generics-infer branch from eb2e115 to a5d19dc Compare March 27, 2019 17:51

rust-highfive assigned nikomatsakis and unassigned eddyb Mar 30, 2019

varkor and others added 13 commits May 1, 2019 23:11

Add const-types test

67176f7

Remove spurious assertion

97c0c66

Fix known-known const unification case

b872e63

Fix unused variable warning

d8b1ddb

Correct name of constrained_generic_params alias

388f823

Resolve FIXME in probe.rs

beb2f84

Fix issue with const params in operand.rs

a29eca5

Add a test for const arguments

c7deb5b

Fold const in writeback

16d6ee3

Add a test for incorrect numbers of const args

a188850

Add a test for const args that cannot be inferred

1369132

Co-Authored-By: Gabriel Smith <yodaldevoid@users.noreply.github.com>

Create ShallowResolver

541de81

Co-Authored-By: Gabriel Smith <yodaldevoid@users.noreply.github.com>

Split ct_err out into CommonConsts

a68ed06

Co-Authored-By: Gabriel Smith <yodaldevoid@users.noreply.github.com>

varkor force-pushed the const-generics-infer branch from aa95b41 to a68ed06 Compare May 1, 2019 22:17

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 May 1, 2019

bors added the merged-by-bors This PR was explicitly merged by bors. label May 2, 2019

bors merged commit a68ed06 into rust-lang:master May 2, 2019

bors mentioned this pull request May 2, 2019

[Step 2] Implement "small substs optimization" for substs of length 1 #59312

Closed

varkor deleted the const-generics-infer branch May 2, 2019 08:28

bors added a commit that referenced this pull request Jun 6, 2019

Auto merge of #59331 - varkor:TypeVariableOrigin-refactor, r=eddyb

1bec46c

Refactor `TypeVariableOrigin` Removes some unused variants and extracts the common `Span` field. As suggested in #59008 (comment). r? @eddyb

JohnTitor mentioned this pull request Jun 21, 2021

Update "Inference variables" section rust-lang/rustc-dev-guide#1145

Merged

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Add const generics to infer (and transitive dependencies) #59008

Add const generics to infer (and transitive dependencies) #59008

varkor commented Mar 8, 2019

eddyb left a comment

eddyb Mar 12, 2019

nikomatsakis Apr 26, 2019

nikomatsakis Apr 26, 2019

eddyb Mar 13, 2019

nikomatsakis Apr 26, 2019

varkor Apr 26, 2019

eddyb Mar 13, 2019

varkor Mar 13, 2019

yodaldevoid Mar 13, 2019

eddyb Mar 13, 2019

yodaldevoid Mar 13, 2019

eddyb Mar 13, 2019

nikomatsakis Apr 26, 2019

bors commented Mar 13, 2019

varkor commented Mar 18, 2019

This comment has been minimized.

crlf0710 commented Mar 19, 2019

varkor commented Mar 20, 2019

bors commented Mar 24, 2019

eddyb commented Mar 25, 2019

varkor commented Mar 25, 2019

varkor commented Mar 27, 2019

Centril commented Mar 30, 2019

eddyb commented Mar 30, 2019

WiSaGaN commented Apr 12, 2019

crlf0710 commented Apr 12, 2019

eddyb commented May 1, 2019

bors commented May 1, 2019

bors commented May 2, 2019

bors commented May 2, 2019

Add const generics to infer (and transitive dependencies) #59008

Add const generics to infer (and transitive dependencies) #59008

Conversation

varkor commented Mar 8, 2019

eddyb 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

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

bors commented Mar 13, 2019

varkor commented Mar 18, 2019

This comment has been minimized.

crlf0710 commented Mar 19, 2019

varkor commented Mar 20, 2019

bors commented Mar 24, 2019

eddyb commented Mar 25, 2019

varkor commented Mar 25, 2019

varkor commented Mar 27, 2019

Centril commented Mar 30, 2019

eddyb commented Mar 30, 2019

WiSaGaN commented Apr 12, 2019

crlf0710 commented Apr 12, 2019

eddyb commented May 1, 2019

bors commented May 1, 2019

bors commented May 2, 2019

bors commented May 2, 2019