Turn `type_id` into a constant intrinsic #47892

Badel2 · 2018-01-30T22:47:55Z

The method get_type_id in Any is intended to support reflection. It's currently unstable in favor of using an associated constant instead. This PR makes the type_id intrinsic a constant intrinsic, the same as size_of and align_of, allowing TypeId::of to be a const fn, which will allow using an associated constant in Any.

rust-highfive · 2018-01-30T22:48:08Z

r? @dtolnay

(rust_highfive has picked a reviewer for you, use r? to override)

Badel2 · 2018-01-30T23:03:14Z

I basically copied @eddyb 's #42859, as it's my first time dealing with compiler instrinsics.

In miri/const_eval.rs there's an error called "NeedsRfc", which would suggest that adding intrinsics to const_eval needs some discussion, so I'm not sure if my PR would be allowed.

I added a few lines of code to an existing test because I don't know what exactly to test, so any suggestions welcome.

As for the Any trait, I tried adding a simple const TYPE_ID: TypeId = TypeId::of::<Self>(); but it didn't compile, I get an error (E0038) that says "the trait cannot contain associated consts" and I'm not sure why. Since that is the goal of this PR, I would like to make sure that adding an associated constant to Any is possible, and maybe do it in this PR.

eddyb · 2018-01-30T23:05:40Z

cc @oli-obk @durka This probably needs to indicate that it's not stably const fn.

oli-obk · 2018-01-31T06:05:52Z

Do the type IDs change between compilations?

According to the docs they'll change between releases and platforms. I guess this could be said of type_size, too, but that one is at least predictable in some manner.

Should definitely be unstable const fn until it can actually be added to Any

durka · 2018-01-31T06:15:11Z

As I understand they're basically keys to a hashmap, right? So completely unstable between compiler versions. Like enum discriminants but worse. However, that's already the case with TypeId::of.

…

On Wed, Jan 31, 2018 at 7:06 AM, Oliver Schneider ***@***.***> wrote: Do the type IDs change between compilations? According to the docs they'll change between releases and platforms. I guess this could be said of type_size, too, but that one is at least predictable in some manner. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#47892 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAC3n9yn9Ua4qxNMpFhnbnGmm1cbuBPRks5tQALWgaJpZM4RzI_i> .

durka · 2018-01-31T06:16:39Z

`#[rustc_const_unstable(feature="const_type_id")]` is what @eddyb is asking for.

…

On Wed, Jan 31, 2018 at 7:14 AM, Alex Burka ***@***.***> wrote: As I understand they're basically keys to a hashmap, right? So completely unstable between compiler versions. Like enum discriminants but worse. However, that's already the case with TypeId::of. On Wed, Jan 31, 2018 at 7:06 AM, Oliver Schneider < ***@***.***> wrote: > Do the type IDs change between compilations? > > According to the docs they'll change between releases and platforms. I > guess this could be said of type_size, too, but that one is at least > predictable in some manner. > > — > You are receiving this because you were mentioned. > Reply to this email directly, view it on GitHub > <#47892 (comment)>, > or mute the thread > <https://github.com/notifications/unsubscribe-auth/AAC3n9yn9Ua4qxNMpFhnbnGmm1cbuBPRks5tQALWgaJpZM4RzI_i> > . >

oli-obk · 2018-01-31T07:07:04Z

So completely unstable between compiler versions. Like enum discriminants but worse.

With miri we could make sure that no decisions are made on that value.

durka · 2018-01-31T08:01:38Z

Uh... define "decisions"? That sounds like a potentially really annoying restriction.

…

On Jan 31, 2018 08:07, "Oliver Schneider" ***@***.***> wrote: So completely unstable between compiler versions. Like enum discriminants but worse. With miri we could make sure that no decisions are made on that value. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#47892 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAC3n_sP8od1eq9k7Xn1ldjMLrKI7Xe9ks5tQBEtgaJpZM4RzI_i> .

oli-obk · 2018-01-31T08:16:11Z

You can't use it as an index to arrays, you can't compare it to anything but other values derived from type_id, ...

eddyb · 2018-01-31T08:18:36Z

I'd rather not go crazy about it - it's your fault if you misuse this random (but deterministic within one compilation) number that the compiler gave you.

Badel2 · 2018-01-31T12:19:20Z

#[rustc_const_unstable(feature="const_type_id")]

Thanks @durka, but it was really hard to find documentation about it. I assume that when you add it to a stable function, and that function is called in a non-const context, it works the same as before? Because TypeId::of is already stable so it should keep working. Or maybe do I need to add rustc_const_unstable to the intrinsic itself? I have run rg rustc_const_unstable src/ and there are zero uses of this feature outside of tests, so I don't know what to do.

oli-obk · 2018-01-31T15:06:24Z

src/test/run-pass/typeid-intrinsic.rs

@@ -91,4 +92,11 @@ pub fn main() {
    // Check fn pointer against collisions
    assert!(TypeId::of::<fn(fn(A) -> A) -> A>() !=
            TypeId::of::<fn(fn() -> A, A) -> A>());
+
+    // Check const
+    const T1: TypeId = TypeId::of::<A>();


This should be in a separate file to ensure that TypeId::of without the feature enabled still works

I agree, this test is only temporal as I didn't know what to test. (It still works btw).

dtolnay · 2018-01-31T16:56:41Z

Seems reasonable to me, @oli-obk please merge when ready. Thanks!

@bors delegate=oli-obk

bors · 2018-01-31T16:56:42Z

✌️ @oli-obk can now approve this pull request

Badel2 · 2018-02-01T19:06:07Z

ping @oli-obk

oli-obk

Just a style thing, Lgtm otherwise

oli-obk · 2018-02-01T21:31:18Z

src/test/compile-fail/const-typeid-of.rs

+
+fn main() {
+    const A_ID: TypeId = TypeId::of::<A>();
+    //~^ ERROR `std::any::TypeId::of` is not yet stable as a const fn


Ah, this is a good test that I forgot about. What I meant was an extra file that uses TypeId::of without the feature gate in a non const context.

....

And a quick search shows there are already files testing that. Don't mind me

oli-obk · 2018-02-01T21:37:29Z

src/librustc_trans/mir/constant.rs

+                                    let llval = C_u64(self.cx,
+                                        self.cx.tcx.type_id_hash(substs.type_at(0)));
+                                    Ok(Const::new(llval, tcx.types.u64))
+


Nit: remove this empty line

Add rustc_const_unstable attribute for `any::TypeId::of` Add test for `const fn TypeId::of`

dtolnay · 2018-02-04T17:20:01Z

@bors r=oli-obk

bors · 2018-02-04T17:20:02Z

📌 Commit 196fad0 has been approved by oli-obk

Turn `type_id` into a constant intrinsic rust-lang#27745 The method `get_type_id` in `Any` is intended to support reflection. It's currently unstable in favor of using an associated constant instead. This PR makes the `type_id` intrinsic a constant intrinsic, the same as `size_of` and `align_of`, allowing `TypeId::of` to be a `const fn`, which will allow using an associated constant in `Any`.

Rollup of 10 pull requests - Successful merges: #47862, #47877, #47896, #47912, #47947, #47958, #47978, #47996, #47999, #47892 - Failed merges:

rust-highfive assigned dtolnay Jan 30, 2018

kennytm added the S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. label Jan 31, 2018

oli-obk reviewed Jan 31, 2018

View reviewed changes

Badel2 force-pushed the const_type_id_of branch 2 times, most recently from 2da171f to 80cb8b3 Compare January 31, 2018 16:40

kennytm added S-waiting-on-review Status: Awaiting review from the assignee but also interested parties. and removed S-waiting-on-author Status: This is awaiting some action (such as code changes or more information) from the author. labels Feb 1, 2018

oli-obk requested changes Feb 1, 2018

View reviewed changes

Turn type_id into a constant intrinsic

196fad0

Add rustc_const_unstable attribute for `any::TypeId::of` Add test for `const fn TypeId::of`

Badel2 force-pushed the const_type_id_of branch from 80cb8b3 to 196fad0 Compare February 1, 2018 22:03

kennytm 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 Feb 4, 2018

kennytm mentioned this pull request Feb 4, 2018

Rollup of 10 pull requests #47998

Merged

bors added a commit that referenced this pull request Feb 4, 2018

Auto merge of #47998 - kennytm:rollup, r=kennytm

7dd3864

Rollup of 10 pull requests - Successful merges: #47862, #47877, #47896, #47912, #47947, #47958, #47978, #47996, #47999, #47892 - Failed merges:

bors added a commit that referenced this pull request Feb 4, 2018

Auto merge of #47998 - kennytm:rollup, r=kennytm

e7e982a

Rollup of 10 pull requests - Successful merges: #47862, #47877, #47896, #47912, #47947, #47958, #47978, #47996, #47999, #47892 - Failed merges:

bors merged commit 196fad0 into rust-lang:master Feb 5, 2018

KodrAus mentioned this pull request Sep 23, 2020

Tracking Issue for const fn type_id #77125

Open

4 tasks

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Turn `type_id` into a constant intrinsic #47892

Turn `type_id` into a constant intrinsic #47892

Badel2 commented Jan 30, 2018

rust-highfive commented Jan 30, 2018

Badel2 commented Jan 30, 2018

eddyb commented Jan 30, 2018

oli-obk commented Jan 31, 2018 •

edited

Loading

durka commented Jan 31, 2018 via email

durka commented Jan 31, 2018 via email

oli-obk commented Jan 31, 2018

durka commented Jan 31, 2018 via email

oli-obk commented Jan 31, 2018

eddyb commented Jan 31, 2018

Badel2 commented Jan 31, 2018

oli-obk Jan 31, 2018

Badel2 Jan 31, 2018

dtolnay commented Jan 31, 2018

bors commented Jan 31, 2018

Badel2 commented Feb 1, 2018

oli-obk left a comment

oli-obk Feb 1, 2018

oli-obk Feb 1, 2018

dtolnay commented Feb 4, 2018

bors commented Feb 4, 2018

Turn type_id into a constant intrinsic #47892

Turn type_id into a constant intrinsic #47892

Conversation

Badel2 commented Jan 30, 2018

rust-highfive commented Jan 30, 2018

Badel2 commented Jan 30, 2018

eddyb commented Jan 30, 2018

oli-obk commented Jan 31, 2018 • edited Loading

durka commented Jan 31, 2018 via email

durka commented Jan 31, 2018 via email

oli-obk commented Jan 31, 2018

durka commented Jan 31, 2018 via email

oli-obk commented Jan 31, 2018

eddyb commented Jan 31, 2018

Badel2 commented Jan 31, 2018

oli-obk Jan 31, 2018

Choose a reason for hiding this comment

Badel2 Jan 31, 2018

Choose a reason for hiding this comment

dtolnay commented Jan 31, 2018

bors commented Jan 31, 2018

Badel2 commented Feb 1, 2018

oli-obk left a comment

Choose a reason for hiding this comment

oli-obk Feb 1, 2018

Choose a reason for hiding this comment

oli-obk Feb 1, 2018

Choose a reason for hiding this comment

dtolnay commented Feb 4, 2018

bors commented Feb 4, 2018

Turn `type_id` into a constant intrinsic #47892

Turn `type_id` into a constant intrinsic #47892

oli-obk commented Jan 31, 2018 •

edited

Loading