Create a new consolidated API to pass types across the ffi-boundary #421

lilizoey · 2023-09-23T18:05:56Z

We can now use GodotCompatible, ToGodot, and FromGodot to pass all values across the ffi-boundary. So for instance we can now pass custom user-defined types across the ffi-boundary like this:

struct MyType {
  a: i64,
  b: f64,
}

impl GodotCompatible for MyType {
  type Via = Dictionary;
}

impl ToGodot for MyType {
  fn to_godot(&self) -> Dictionary {
    dict! {
      "a": self.a,
      "b": self.b,
    }
  }
}

impl FromGodot for MyType {
  fn try_from_godot(d: Dictionary) -> Option<Self> {
    let a = d.get("a")?.try_to::<i64>().ok()?;
    let b = d.get("b")?.try_to::<f64>().ok()?;
    Ok(Self { a, b })
  }
}

Will keep on draft until i've cleaned up the RawGd and Gd code.

I've tried to avoid changing filenames as much as possible to avoid confusing git, some is unavoidable though. And we probably should rename some files before merging.

Added Traits

GodotCompatible
ToGodot
FromGodot
GodotFfiVariant, Defines the primitive conversions of godot types into/from Variant.

Removed Traits

FromVariant
ToVariant
GodotFuncMarshal
All of these have now been folded into From/ToGodot + GodotCompatible.

Renamed/Replaced traits

VariantMetadata -> GodotType, variant_type has been moved into GodotFfi.
GodotNullablePtr -> GodotNullableFfi, now has methods used to implement conversions for Option<T>, could also be implemented for Variant if we want.

`RawGd`

We need a primitive type to represent nullable objects, so i added RawGd to represent that. This is currently pretty messily implemented as i just copy pasted over a bunch of code from Gd and i will clean it up before making ready to merge.

Limitations

Currently ToGodot conversions are assumed infallible, and FromGodot only returns an Option. In the future, we should make these both return Results to improve error messages from gdext. This PR also likely degrades error messages in some places because of this.

Property is currently independent of this new API, this should be changed but will likely require a unification and rewrite of the To/FromGodot and Property derive macros.

The To/FromGodot macros are just a hacky rewrite of the original To/FromVariant derive macros, and as such converts everything to Variant instead of more appropriate types.

closes #414

GodotRust · 2023-09-23T18:12:06Z

API docs are being generated and will be shortly available at: https://godot-rust.github.io/docs/gdext/pr-421

ogapo · 2023-09-25T15:08:37Z

I realize null is a variant type in Godot, but in the interest of being more idiomatic on the Rust side, would it make sense to represent nullable types as just Option (or Option<Gd> in the case where it's specialized to an object)? This would be similar to how Rust handles nullable pointers on the FFI side. There might be a way to take advantage of the null pointer optimization even.

Bromeon · 2023-09-26T10:09:59Z

but in the interest of being more idiomatic on the Rust side, would it make sense to represent nullable types as [...] Option<Gd>

That's already done at the moment, see e.g. try_load().

godot-codegen/src/util.rs

lilizoey

Added some comments to RawGd where i remember there being significant changes from what was before. Since git doesn't quite get that a lot of it was copied over.

godot-core/src/obj/raw.rs

lilizoey · 2023-10-03T00:11:14Z

godot-core/src/obj/raw.rs

+    /// # Safety
+    /// Does not transfer ownership and is thus unsafe. Also operates on shared ref. Either the parameter or
+    /// the return value *must* be forgotten (since reference counts are not updated).
+    pub(super) unsafe fn ffi_cast<U>(&self) -> Option<RawGd<U>>
+    where
+        U: GodotClass,
+    {
+        if self.is_null() {
+            // Null can be cast to anything.
+            // Forgetting a null doesn't do anything, since dropping a null also does nothing.
+            return Some(RawGd::new_null());
+        }
+
+        let class_tag = interface_fn!(classdb_get_class_tag)(U::class_name().string_sys());
+        let cast_object_ptr = interface_fn!(object_cast_to)(self.obj_sys(), class_tag);
+
+        // Create weak object, as ownership will be moved and reference-counter stays the same
+        sys::ptr_then(cast_object_ptr, |ptr| RawGd::from_obj_sys_weak(ptr))
+    }


Basically unchanged except null-check

unsafe fn ffi_cast<U>(&self) -> Option<Gd<U>> where U: GodotClass, { let class_tag = interface_fn!(classdb_get_class_tag)(U::class_name().string_sys()); let cast_object_ptr = interface_fn!(object_cast_to)(self.obj_sys(), class_tag); // Create weak object, as ownership will be moved and reference-counter stays the same sys::ptr_then(cast_object_ptr, |ptr| Gd::from_obj_sys_weak(ptr)) }

I'm curious, when do we need to cast between null objects?

A Gd<T> can't hold nulls, and Option<Gd<T>> would be None.

Similar thoughts apply to other methods doing null checks. If we don't expect those, an assert!(!null) could potentially prevent logic errors.

We use casting when we turn Variant into RawGd, and so if you turn a Variant into an Option<Gd<T>> then you'd first turn the Variant into a RawGd<Object>, then cast that into RawGd<T> then turn that into Option<Gd<T>>. So if the Variant is null we'd be casting a null object.

Makes sense! Can you write down this conversion chain in a comment?
It's enough to do it here in ffi_cast.

godot-core/src/obj/raw.rs

lilizoey · 2023-10-03T00:14:50Z

godot-core/src/obj/raw.rs

+            "called free() on Gd<Object> which points to a RefCounted dynamic type; free() only supported for manually managed types."
+        );
+
+        assert!(!self.is_null(), "called free() on null object");


Added this here, otherwise unchanged

godot-core/src/obj/raw.rs

Bromeon

Fantastic work again, thanks so much for this huge but necessary refactoring!

I will likely review the PR over several sessions due to its sheer size, here is my first batch 🙂

godot-core/src/builtin/meta/godot_compat/mod.rs

godot-codegen/src/util.rs

godot-core/src/builtin/meta/godot_compat/impls.rs

Bromeon · 2023-10-04T22:16:06Z

godot-core/src/builtin/meta/mod.rs

+    impl<T: GodotClass> Sealed for RawGd<T> {}
+    impl<T: GodotClass> Sealed for Gd<T> {}


Do both need to be GodotTypes here?

We might be able to only make Gd a GodotType actually ig. I'll try that

No it isn't really possible, RawGd must implement GodotType. And unless we want people to use RawGd if they ever want to implement GodotCompatible with objects as the backing type, we'd need Gd to implement GodotType. And while i dont think people would want to use objects as the backing type often (it's likely easier to just make a GodotClass type), i dont see a particular reason to completely prevent people from doing so, and i'd rather people not use RawGd directly since it's meant to be a low-level type.

Sounds good, let's leave it then 👍

Bromeon · 2023-10-04T22:20:10Z

godot-ffi/src/godot_ffi.rs

+/// Used to blanket implement various conversions over `Option<T>`.
+pub trait GodotNullableFfi: Sized + GodotFfi {
+    fn flatten_option(opt: Option<Self>) -> Self;
+    fn is_null(&self) -> bool;
 }


I would also mention for which (kinds of) types this trait is implemented.

A user probably won't need to face this directly?

Yeah they shouldn't. We could consider implementing this for Variant, since they have a meaningful null-state. But for now it's only implemented for RawGd, since that's the least amount of changes from the current behavior.

Bromeon · 2023-10-04T22:21:47Z

godot-core/src/builtin/meta/godot_compat/mod.rs

+    // TODO: better error
+    /// Performs the conversion.
+    fn try_from_godot(via: Self::Via) -> Option<Self>;


Even if we don't have an error yet, should we not return Result<Self, ()> for now? That would just mean changing the Err type later, rather than changing from Option to Result.

It would also come with #[must_use] etc.

Makes sense yeah.

Actually thinking about it, it might be better to not do that. I can add a #[must_use], but turning this into a Result<Self, ()> means that i have to change every use-case to handling Results instead usually by ignoring the Err value. Which means that when we change this later we can't rely on the compiler to tell us where we're not properly handling the error value, since we're just ignoring them everywhere.

But do we have many places that ignore errors now but shouldn't ignore once they are Results? From a surface level, that seems wrong. On the other hand, I think the error type is much less important than the fact that there is an error. In gamedev practice, Err is mostly for debugging information; I don't think people frequently dispatched on the variant conversion errors in gdnative, for example.

But if you think there are too many changes or too high risk of missing result handling, we can also do it later (maybe make TODO more descriptive in that case).

The reason this PR doesn't contain the change to using Result initially is that it requires us to change a lot of code to make sure the Err is propagated properly and all error info is preserved throughout call chains where desired.

So instead of doing that, this PR is currently just failing with generic error messages in those cases. If we then change to Result in preparation for the future change, then we'd either need to just ignore the Err and fail with generic error messages still, or we'd need to actually make sure Err values are propagated through everything as they should. But propagating Err as we should was what i was trying to avoid doing so that the PR wouldn't get too big in the first place. And just ignoring Err and failing with generic error messages isn't really much better than just using Option. And it also means that we're more likely to miss result handling when we do make the change in the future.

I was planning on working on the error handling after this PR btw so it should hopefully not be too long a delay either way.

I see. Makes sense to split it then to a later change, thanks for elaboration!

Either way, it's not very urgent. Having this PR is a great base! 😊

Bromeon · 2023-10-04T22:24:40Z

godot-core/src/builtin/meta/mod.rs

+/// This trait cannot be implemented for custom user types, for that you should see [`GodotCompatible`]
+/// instead.


Possibly a short sentence how the two differ on a semantic level could help here. If it's difficult to explain without also mentioning 12 other things, maybe give an example?

Similar to the explanation of GodotType vs GodotFfi below.

Bromeon · 2023-10-04T22:28:25Z

godot-core/src/builtin/meta/mod.rs

@@ -7,27 +7,120 @@
 pub mod registration;

 mod class_name;
+mod godot_compat;


I would name it godot_compatible, otherwise it sounds like a GDExtension compat module.

Bromeon

Next round, mostly around RawGd<T> 😎

I see that a type that represents "objects" in the GDScript sense (i.e. nullable) has some benefits, but I'm still not sure if it makes sense to move all the high-level Gd functionality, such as cast/bind/instance_id/free etc. into RawGd.

Now it looks like we're actively losing type safety and have to do null checks, which are also not free. I'm generally not too concerned about these performance things, but still it's something we don't need with the current design. The changes here would also leave Gd as mostly a delegate/facade without much own business logic.

Where do you intend to use RawGd other than in Gd, that could also benefit from these high-level functions? Because I'm thinking maybe these should stay in Gd itself, and RawGd could focus on the FFI conversions and the nullability-related logic. It would split responsibilities and also reduce the magnitude and complexity of this changeset.

godot-core/src/obj/raw.rs

Bromeon · 2023-10-05T21:16:56Z

godot-core/src/obj/raw.rs

+    /// # Safety
+    /// Does not transfer ownership and is thus unsafe. Also operates on shared ref. Either the parameter or
+    /// the return value *must* be forgotten (since reference counts are not updated).
+    pub(super) unsafe fn ffi_cast<U>(&self) -> Option<RawGd<U>>
+    where
+        U: GodotClass,
+    {
+        if self.is_null() {
+            // Null can be cast to anything.
+            // Forgetting a null doesn't do anything, since dropping a null also does nothing.
+            return Some(RawGd::new_null());
+        }
+
+        let class_tag = interface_fn!(classdb_get_class_tag)(U::class_name().string_sys());
+        let cast_object_ptr = interface_fn!(object_cast_to)(self.obj_sys(), class_tag);
+
+        // Create weak object, as ownership will be moved and reference-counter stays the same
+        sys::ptr_then(cast_object_ptr, |ptr| RawGd::from_obj_sys_weak(ptr))
+    }


I'm curious, when do we need to cast between null objects?

A Gd<T> can't hold nulls, and Option<Gd<T>> would be None.

godot-core/src/obj/raw.rs

Bromeon · 2023-10-05T21:33:00Z

godot-core/src/obj/raw.rs

+    /// # Safety
+    /// Does not transfer ownership and is thus unsafe. Also operates on shared ref. Either the parameter or
+    /// the return value *must* be forgotten (since reference counts are not updated).
+    pub(super) unsafe fn ffi_cast<U>(&self) -> Option<RawGd<U>>
+    where
+        U: GodotClass,
+    {
+        if self.is_null() {
+            // Null can be cast to anything.
+            // Forgetting a null doesn't do anything, since dropping a null also does nothing.
+            return Some(RawGd::new_null());
+        }
+
+        let class_tag = interface_fn!(classdb_get_class_tag)(U::class_name().string_sys());
+        let cast_object_ptr = interface_fn!(object_cast_to)(self.obj_sys(), class_tag);
+
+        // Create weak object, as ownership will be moved and reference-counter stays the same
+        sys::ptr_then(cast_object_ptr, |ptr| RawGd::from_obj_sys_weak(ptr))
+    }


Similar thoughts apply to other methods doing null checks. If we don't expect those, an assert!(!null) could potentially prevent logic errors.

lilizoey · 2023-10-08T12:58:44Z

Next round, mostly around RawGd<T> 😎

I see that a type that represents "objects" in the GDScript sense (i.e. nullable) has some benefits, but I'm still not sure if it makes sense to move all the high-level Gd functionality, such as cast/bind/instance_id/free etc. into RawGd.

owned_cast is needed to turn Variant into RawGd.
ffi_cast is needed for as_refcounted and as_object
as_refcounted is needed to manage the refcount
as_object is needed for printing (which is strictly speaking not needed in RawGd but can help with error messages)
instance_id_or_none is needed in a couple of different places, but maybe we can remove it with the new deref liveness checking? i haven't tried that properly.
free could be moved into Gd
bind_mut is used in scoped_mut which is used in as_refcounted and as_object
bind could be moved into Gd but it feels weird to only move bind and not bind_mut

Now it looks like we're actively losing type safety and have to do null checks, which are also not free. I'm generally not too concerned about these performance things, but still it's something we don't need with the current design. The changes here would also leave Gd as mostly a delegate/facade without much own business logic.

I dont think we have to do a lot more null-checks than before. But we could get rid of the null-checks in some places by using Option<T> and unwrap_unchecked.

Where do you intend to use RawGd other than in Gd, that could also benefit from these high-level functions? Because I'm thinking maybe these should stay in Gd itself, and RawGd could focus on the FFI conversions and the nullability-related logic. It would split responsibilities and also reduce the magnitude and complexity of this changeset.

Either RawGd must manage the refcount, which is what requires at least some of the high-level stuff. Or we need to start adding drop-guards to methods that convert Gd into RawGd. I dont like the idea of adding drop-guards, it adds more complexity and types and such, especially since i think Gd is the only type that would need them.

Edit:

free could be moved into Gd without issue.
instance_id_or_none could also be moved, but i still needed a way to check if a RawGd is valid, for purposes of the Debug impl on RawGd.

PgBiel · 2023-10-08T23:17:48Z

Quick design Q: if one wanted to make it possible to cast more than one possible Godot type to your Rust type (e.g. allow both int and float to convert to your custom type), and say your Rust type's ToGodot always casts to a dictionary, they'd have to implement a casting from dictionary on FromGodot (since that's their Self::Via), but also override from_variant to accept other types than just dictionary, correct?

Just asking this because it might be interesting to ponder how flexible we want this API to be (preferably in some way to avoid Variant). I wonder if it'd be possible to replace Self::Via: GodotType by something like Self::Via: GodotCompatible, which could allow , for instance, using Rust enums to signify multiple possible types that can be received (and we'd just keep calling .to_godot() / .from_godot() until we reached a primitive type, which could use some separate method in GodotCompatible, such as is_primitive(), to signal we have reached the end of the chain). That could, however, be non-viable due to the inner workings of GDExtension and Godot FFI, which could just end up imposing one to use Variant in this situation instead, so this is mostly just a thought.

(Here I'm being kinda inspired from some previous experience as a contributor to the Typst project compiler, where casts of Rust types from and to Value - equivalent to Godot's Variant - mostly involve using match to accept different value types. E.g., it might make sense to accept both an int and a float to cast to a custom "real" type, and even be able to output a third type when casting to Value / Variant.)

lilizoey · 2023-10-09T08:17:18Z

Quick design Q: if one wanted to make it possible to cast more than one possible Godot type to your Rust type (e.g. allow both int and float to convert to your custom type), and say your Rust type's ToGodot always casts to a dictionary, they'd have to implement a casting from dictionary on FromGodot (since that's their Self::Via), but also override from_variant to accept other types than just dictionary, correct?

That's not how it's intended to be used. If you want to accept more than one type you should use Variant as the Via. Otherwise you could get confusing type errors where a call to a rust function from gdscript may succeed or fail depending on whether you statically type your variables or not.

Just asking this because it might be interesting to ponder how flexible we want this API to be (preferably in some way to avoid Variant). I wonder if it'd be possible to replace Self::Via: GodotType by something like Self::Via: GodotCompatible, which could allow , for instance, using Rust enums to signify multiple possible types that can be received (and we'd just keep calling .to_godot() / .from_godot() until we reached a primitive type, which could use some separate method in GodotCompatible, such as is_primitive(), to signal we have reached the end of the chain). That could, however, be non-viable due to the inner workings of GDExtension and Godot FFI, which could just end up imposing one to use Variant in this situation instead, so this is mostly just a thought.

(Here I'm being kinda inspired from some previous experience as a contributor to the Typst project compiler, where casts of Rust types from and to Value - equivalent to Godot's Variant - mostly involve using match to accept different value types. E.g., it might make sense to accept both an int and a float to cast to a custom "real" type, and even be able to output a third type when casting to Value / Variant.)

As far as i know there is no way to make rust generate the proper calls to to_godot() until we reach a GodotType (as this may be arbitrarily many calls, and may not even terminate possibly generating an infinitely long call graph). so we'd basically need to add a where Via: GodotType anywhere we'd use it to interface with godot anyway. which would make To/FromGodot basically just a poor version of Into/From.

Bromeon

owned_cast is needed to turn Variant into RawGd.

ffi_cast is needed for as_refcounted and as_object

as_refcounted is needed to manage the refcount

as_object is needed for printing (which is strictly speaking not needed in RawGd but can help with error messages)

instance_id_or_none is needed in a couple of different places, but maybe we can remove it with the new deref liveness checking? i haven't tried that properly.

free could be moved into Gd

bind_mut is used in scoped_mut which is used in as_refcounted and as_object

bind could be moved into Gd but it feels weird to only move bind and not bind_mut

Thanks for this great list! It looks like as_refcounted blocks everything else from being moved one level up. We're somewhat facing the OOP problem of "everything in the base class", just with composition. Gd then doesn't have much purpose beyond fulfilling some traits and delegating functionality. But it's fine, better than exposing the raw one directly 🙂

Maybe you could move up free and if possible, display/debug? Although I don't know if the latter would then make error reporting harder. They should also stay together. Either way, this can also be changed again later, so it's not crucial.

Can you make RawGd #[doc(hidden)]? Users should not need to know about it.
Also rename GodotCompatible -> GodotConvert.

godot-core/src/obj/raw.rs

godot-core/src/obj/traits.rs

godot-core/src/engine.rs

Bromeon · 2023-10-09T14:50:51Z

godot-core/src/obj/raw.rs

+    /// # Safety
+    /// Does not transfer ownership and is thus unsafe. Also operates on shared ref. Either the parameter or
+    /// the return value *must* be forgotten (since reference counts are not updated).
+    pub(super) unsafe fn ffi_cast<U>(&self) -> Option<RawGd<U>>
+    where
+        U: GodotClass,
+    {
+        if self.is_null() {
+            // Null can be cast to anything.
+            // Forgetting a null doesn't do anything, since dropping a null also does nothing.
+            return Some(RawGd::new_null());
+        }
+
+        let class_tag = interface_fn!(classdb_get_class_tag)(U::class_name().string_sys());
+        let cast_object_ptr = interface_fn!(object_cast_to)(self.obj_sys(), class_tag);
+
+        // Create weak object, as ownership will be moved and reference-counter stays the same
+        sys::ptr_then(cast_object_ptr, |ptr| RawGd::from_obj_sys_weak(ptr))
+    }


Makes sense! Can you write down this conversion chain in a comment?
It's enough to do it here in ffi_cast.

PgBiel · 2023-10-09T15:28:19Z

That's not how it's intended to be used. If you want to accept more than one type you should use Variant as the Via. Otherwise you could get confusing type errors where a call to a rust function from gdscript may succeed or fail depending on whether you statically type your variables or not.

Okay, that's fair. Thanks for clearing this up!

As far as i know there is no way to make rust generate the proper calls to to_godot() until we reach a GodotType (as this may be arbitrarily many calls, and may not even terminate possibly generating an infinitely long call graph). so we'd basically need to add a where Via: GodotType anywhere we'd use it to interface with godot anyway. which would make To/FromGodot basically just a poor version of Into/From.

It's possible, but I agree it'd probably be very hacky, so it's better to just not do this haha. Thanks for the answer!
I believe this design should work just fine then, with those considerations. 👍

- Rename VariantMetadata -> GodotMetadata - Add GodotCompatible, ToGodot, FromGodot - Add RawGd - Change everything so that To/FromGodot is the canonical way of passing values over the ffi-boundary - Change `GodotMetadata` to sealed `GodotType` - Implement `GodotType` for various types - Add `GodotFfiVariant` - Add tests for creating newtype wrappers around godot types - Fix to/from variant derives

…ssion

Bromeon mentioned this pull request Sep 25, 2023

Opt-in support for functions that return Result<T,GodotError> to indicate failable operations #425

Open

lilizoey force-pushed the feature/godot-compat branch 4 times, most recently from ca9dea5 to 7f47fcc Compare October 2, 2023 13:05

lilizoey marked this pull request as ready for review October 3, 2023 00:01

lilizoey commented Oct 3, 2023

View reviewed changes

godot-codegen/src/util.rs Outdated Show resolved Hide resolved

lilizoey requested a review from Bromeon October 3, 2023 00:03

lilizoey commented Oct 3, 2023

View reviewed changes

lilizoey force-pushed the feature/godot-compat branch from 7f47fcc to c8666af Compare October 3, 2023 12:20

Bromeon mentioned this pull request Oct 4, 2023

Separate out most of the low-level logic from Gd into RawGd #349

Closed

13 tasks

Bromeon reviewed Oct 4, 2023

View reviewed changes

Bromeon added feature Adds functionality to the library c: core Core components c: ffi Low-level components and interaction with GDExtension API labels Oct 4, 2023

lilizoey force-pushed the feature/godot-compat branch 3 times, most recently from 402569d to b7b2183 Compare October 5, 2023 13:18

Bromeon reviewed Oct 5, 2023

View reviewed changes

Bromeon mentioned this pull request Oct 9, 2023

Allow #[cfg] to be used with #[godot_api] - Part 1: inherent impls #443

Merged

Bromeon reviewed Oct 9, 2023

View reviewed changes

Bromeon mentioned this pull request Oct 10, 2023

Vector4i doesn't implement "into Variant" godot-rust/gdnative#1058

Closed

lilizoey force-pushed the feature/godot-compat branch from 2a103a0 to 2dc78d2 Compare October 13, 2023 11:56

lilizoey force-pushed the feature/godot-compat branch from 2dc78d2 to 3470977 Compare October 13, 2023 12:04

lilizoey added 2 commits October 13, 2023 22:58

Various cleanup and changes to GodotConvert and related after discu…

ffe241a

…ssion

lilizoey force-pushed the feature/godot-compat branch from 3470977 to ffe241a Compare October 13, 2023 21:02

Bromeon added this pull request to the merge queue Oct 13, 2023

github-merge-queue bot removed this pull request from the merge queue due to failed status checks Oct 13, 2023

Bromeon added this pull request to the merge queue Oct 13, 2023

Merged via the queue into godot-rust:master with commit 844f0f3 Oct 13, 2023
14 checks passed

This was referenced Oct 14, 2023

Make ToGodot and FromGodot conversions use Result #451

Closed

Rewrite the ToGodot and FromGodot derive macros #452

Closed

Integrate Property with the GodotConvert API #453

Closed

Bromeon mentioned this pull request Dec 13, 2023

Project status overview #24

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Create a new consolidated API to pass types across the ffi-boundary #421

Create a new consolidated API to pass types across the ffi-boundary #421

lilizoey commented Sep 23, 2023

GodotRust commented Sep 23, 2023

ogapo commented Sep 25, 2023 •

edited

Loading

Bromeon commented Sep 26, 2023

lilizoey left a comment

lilizoey Oct 3, 2023

Bromeon Oct 5, 2023

Bromeon Oct 5, 2023

lilizoey Oct 8, 2023

Bromeon Oct 9, 2023

lilizoey Oct 3, 2023

Bromeon left a comment

Bromeon Oct 4, 2023

lilizoey Oct 5, 2023

lilizoey Oct 5, 2023

Bromeon Oct 5, 2023

Bromeon Oct 4, 2023

lilizoey Oct 5, 2023

Bromeon Oct 4, 2023

lilizoey Oct 5, 2023

lilizoey Oct 5, 2023

Bromeon Oct 5, 2023

lilizoey Oct 5, 2023

Bromeon Oct 6, 2023

Bromeon Oct 4, 2023

Bromeon Oct 4, 2023

Bromeon left a comment •

edited

Loading

Bromeon Oct 5, 2023

Bromeon Oct 5, 2023

lilizoey commented Oct 8, 2023 •

edited

Loading

PgBiel commented Oct 8, 2023

lilizoey commented Oct 9, 2023

Bromeon left a comment

Bromeon Oct 9, 2023

PgBiel commented Oct 9, 2023

		impl<T: GodotClass> Sealed for RawGd<T> {}
		impl<T: GodotClass> Sealed for Gd<T> {}

		/// This trait cannot be implemented for custom user types, for that you should see [`GodotCompatible`]
		/// instead.

Create a new consolidated API to pass types across the ffi-boundary #421

Create a new consolidated API to pass types across the ffi-boundary #421

Conversation

lilizoey commented Sep 23, 2023

Added Traits

Removed Traits

Renamed/Replaced traits

RawGd

Limitations

GodotRust commented Sep 23, 2023

ogapo commented Sep 25, 2023 • edited Loading

Bromeon commented Sep 26, 2023

lilizoey 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

Bromeon 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

Choose a reason for hiding this comment

Bromeon left a comment • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

lilizoey commented Oct 8, 2023 • edited Loading

PgBiel commented Oct 8, 2023

lilizoey commented Oct 9, 2023

Bromeon left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

PgBiel commented Oct 9, 2023

`RawGd`

ogapo commented Sep 25, 2023 •

edited

Loading

Bromeon left a comment •

edited

Loading

lilizoey commented Oct 8, 2023 •

edited

Loading