Comments (9)
That's the reason it's necessary to use multiple traits. We need lifetime templated traits.
Here a simple example in the playground
from soa-derive.
Here is what the code looks like (simplified):
1. Use traits to add functionality to the primitive type
trait PtrExt {
type PtrMut;
fn as_mut_ptr(&self) -> Self::PtrMut;
}
impl<T> PtrExt for *const T {
type PtrMut = *mut T;
fn as_mut_ptr(&self) -> Self::PtrMut {
self as *const T
}
}
impl ParticlePtr {
fn as_mut_ptr(&self) -> ParticleMutPtr {
use PtrExt;
ParticleMutPtr {
name: self.name.as_mut_ptr(),
mass: self.mass.as_mut_ptr(),
nested: self.nested.as_mut_ptr(),
}
}
// [...]
}
2. Use traits static methods
trait SoaPtr {
type PtrMut;
fn as_mut_ptr(&self) -> Self::PtrMut;
}
impl<T> SoaPtr for *const T {
type PtrMut = *mut T;
fn as_mut_ptr(&self) -> Self::PtrMut {
self as *const T
}
}
impl SoaPtr for ParticlePtr {
type PtrMut = ParticleMutPtr;
fn as_mut_ptr(&self) -> Self::PtrMut {
self.as_mut_ptr()
}
}
impl ParticlePtr {
fn as_mut_ptr(&self) -> ParticleMutPtr {
ParticleMutPtr {
name: SoaPtr::as_mut_ptr(self.name),
mass: SoaPtr::as_mut_ptr(self.mass),
nested: SoaPtr::as_mut_ptr(self.nested),
}
}
// [...]
}
3. Generating the proper expressions
impl ParticlePtr {
fn as_mut_ptr(&self) -> ParticleMutPtr {
ParticleMutPtr {
name: self.name as *mut String,
mass: self.mass as *mut f64,
nested: self.nested.as_mut_ptr(),
}
}
// [...]
}
from soa-derive.
I realized this while working on the single length. I think this is more useful so I'll make it before the single length :)
from soa-derive.
This sounds like a good idea, however it might not be possible to do it for slices without GAT, cf #25 (comment).
from soa-derive.
Well, that sounds good then =)
from soa-derive.
Hey! Sorry for the inactivity these days. Here is an update on what I'm working right now.
I'm using the discussed traits that define the nested fields' types and it's working wonderfully :)
The current problem is how to handle the difference in what kind of expression we need depending of the type, for example particle_ptr.mass as *mut f64
vs particle_ptr.as_mut_ptr()
. I'm in my third iteration and this is what I can say about them:
- Using a helper trait to "inject" a function only to the non-nested field. It's how I added support for indexing. The main problem is that it doesn't work properly if we try to do so to a type that already has a method named the same. It can still be used with standard types like pointers because we can check if we do so (even though it may broke if a method with the same name is added in the future, which can cause issues with index and we may have to change it).
- Using a trait to add a static method to the generated SoAs and the standard types. This avoids any name collision to the expense of having to generate a lot of traits implementations for every generated SoA vector (probably increases build times). Also there are times that it's not possible to use it. As an example
as_ref
on the base type (Particle
). - (Just started trying it out) Generate every expression depending on the underlying. This makes the generated code straightforward, cleaner and should work everywhere. But doing that we get worse generating code because we can no longer simply use a single
quote!
(making some kind of helper function or macros may help with this problem).
If you have any thoughts or ideas, let me know. I'll probably add a comment after the third version is done or I find a roadblock.
from soa-derive.
I'm not quite sure I understand exactly how every one of your points work, could you add some basic example code?
The main problem is that it doesn't work properly if we try to do so to a type that already has a method named the same
If we are using a trait, don't inherent methods (the one defined on the type) take priority? And we can (and should) generate the code to always use the full path of the trait (soa_derive::IndexVec::index(&vec, i)
instead of vec.index(i)
). Or I am misunderstanding something here?
from soa-derive.
In the first and second case I'm abusing traits to force overriding (in another language would simply be an overrided function for each type). Like you said, the first case may break with the addition of the wrong method into a type (would take preference). If we want to add a full path to the Trait we need every generated type to implement it. That's case 2.
Defining the correct lifetimes explicitly can be a nightmare when using traits, so generating the precise expression (case 3) is turning out to be the simplest and most reliable solution.
from soa-derive.
If we are using a trait, don't inherent methods (the one defined on the type) take priority? And we can (and should) generate the code to always use the full path of the trait (soa_derive::IndexVec::index(&vec, i) instead of vec.index(i)). Or I am misunderstanding something here?
Index is a bit different. We do not (and cannot) control the traits (Index
and IndexMut
). The only problem we may have is that they add index
or index_mut
to slice
or Vec
and make them do something different than the traits.
We could probably add the primitive types into our custom trait and then forward to the standard ones for the primitive type and to the custom implementation for generated ones. This would make the code even longer (take a look at index.rs
, it's already massive). For now I'd keep it as it is, maybe add a comment explaining this in the code.
from soa-derive.
Related Issues (20)
- performance of the slice iterator HOT 14
- Allow explicit name generation HOT 1
- Use one length and capacity variable for whole struct HOT 23
- Update versions of `syn` and `quote` HOT 3
- Provide FromIterator implementation HOT 3
- Use trait to link the origin struct to generated soa struct HOT 3
- Add access methods HOT 3
- Nitpick: doc for generated vec-like methods have extra spaces HOT 1
- Does not support generics. HOT 6
- #[soa_attr] on field
- #[serde(skip)] cause struct of array containing different length of field HOT 4
- Const array SoA variant HOT 2
- Cannot find trait `SoAIter` in crate `soa_derive` HOT 3
- Sorting methods for Slice variant HOT 2
- Derive std::iter::Extend HOT 1
- Is it possible to avoid duplicate capacity and size fields? HOT 1
- How do you go from `MyTypeRef` to `MyType` ? HOT 6
- How to set #[pyclass] attribute for soa-derived struct? HOT 2
- cannot move out of type `CheeseVec`, which implements the `Drop` trait HOT 2
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from soa-derive.