Lifetime issues with `slice` methods
還沒有人認領這個 Issue。
評估
研究方向
Start by reproducing the minimal Rust example from the playground link, comparing the slice_axis behavior in foo with slice_axis_inplace in bar and the explicit lifetimes in baz. Trace the slice and slice_axis APIs to determine which lifetime is tied to the returned view; done means the intended foo pattern is either supported with appropriate coverage or documented as requiring explicit lifetimes.
由索引模型根據 Issue 內容生成。
描述
Hello folks,
I think I found an issue in the way lifetimes are inferred when using slice or slice_axis on array views.
Here is a minimal example with three functions, only bar and baz compile, while foo doesn't. I would argue that foo should be able to compile.
use ndarray; // 0.15.6
use ndarray::{ArrayBase, ViewRepr, Axis, Dimension, Slice, ArrayView};
fn main() {
let a = ndarray::array![[1, 2, 3], [4, 5, 6]];
foo(a.view(), ndarray::Axis(1));
}
fn foo<T, D>(arrayview: ArrayView<T, D>, axis: ndarray::Axis) -> ArrayView<T, D>
where
D: Dimension,
{
arrayview.slice_axis(axis, Slice::from(1..2))
}
fn bar<T, D>(arrayview: &mut ArrayView<T, D>, axis: Axis)
where
D: Dimension,
{
arrayview.slice_axis_inplace(axis, Slice::from(1..2));
}
fn baz<'a, 'data: 'a, T, D>(
arrayview: &'a ArrayBase<ViewRepr<&'data T>, D>,
axis: Axis,
) -> ArrayBase<ViewRepr<&'a T>, D>
where
D: Dimension,
{
arrayview.slice_axis(axis, Slice::from(1..2))
}
(playground link: https://play.rust-lang.org/?version=stable&mode=debug&edition=2021&gist=7d62b20eec45d257e629dae07840544b)
The error is the following:
error[[E0311]](https://doc.rust-lang.org/stable/error-index.html#E0311): the parameter type `D` may not live long enough
--> src/main.rs:14:5
|
14 | arrayview.slice_axis(axis, Slice::from(1..2))
| ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
|
note: the parameter type `D` must be valid for the anonymous lifetime defined here...
--> src/main.rs:10:25
|
10 | fn foo<T, D>(arrayview: ArrayView<T, D>, axis: ndarray::Axis) -> ArrayView<T, D>
| ^^^^^^^^^^^^^^^
note: ...so that the type `D` will meet its required lifetime bounds
--> src/main.rs:14:5
|
14 | arrayview.slice_axis(axis, Slice::from(1..2))
| ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
help: consider adding an explicit lifetime bound...
|
12 | D: Dimension + 'a,
| ++++
error[[E0515]](https://doc.rust-lang.org/stable/error-index.html#E0515): cannot return reference to function parameter `arrayview`
--> src/main.rs:14:5
|
14 | arrayview.slice_axis(axis, Slice::from(1..2))
| ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^ returns a reference to data owned by the current function
For more information about this error, try `rustc --explain E0515`.
error: could not compile `playground` due to 2 previous errors
My understanding of the problem is the following: the lifetime of the data represented by the array view passed in as argument to foo is not propagated correctly to the one obtained by slicing it, and the output of the slice function is actually depending on the lifetime of the view passed as argument and not the lifetime of the data referenced by the view.
In bar I show how to circumnavigate the issue by slicing in place. Given that views are cheap to create and clone, this is the alternative I'm going with for now.
In baz, I show a version where I explicitly annotate lifetimes to make it work, note how I have to constrain 'a to 'data and how I need to make the argument a reference to avoid having the function drop it while the returned value still references it.
Also, I'm not sure why the error message is angry about the lifetime of D in the first place, I don't see where it's being defined as a reference or a pointer.
I'm not super good at understanding lifetimes but it feels like it should be possible to slice a view and obtain something that doesn't depend on the view itself but on the underlying data only. Or am I missing something here ?
- 主要語言
- Rust
- 星號
- 4.3k
- 分支
- 391
- PR 合併指標
- 30 天內沒有已合併 PR
環境準備
這個專案沒有提供開發容器、Dockerfile 或貢獻指南,環境需要你自己搭建:先看它的 README,通用步驟見我們的新手貢獻指南。
從這裡開始
- 先讀完整個 Issue,再讀專案的貢獻指南。
- 在 Issue 下留言說明你要接手 —— 這能避免兩個人做同樣的事。
- Fork 儲存庫,在一個分支上完成修改。
- 送出 Pull Request,並在描述裡引用這個 Issue 編號。
rust-ndarray/ndarray 的其他 Issue
-
.is_all_nan() returns true for an all-finite array可能已有人在做 @youdie006 於 52 天前認領。 未關閉
難度 2/5 1-3 小時 新手友好度 72/100
rust-ndarray/ndarray#1612 · 1 則留言 · 1 個 reaction ·
-
難度 4/5 3-5 天 新手友好度 48/100
rust-ndarray/ndarray#1617 · 1 則留言 ·
-
Stack overflow in `triu`可能已有人在做 關聯的 PR 仍在進行中或已合併。 未關閉bug good first issue
難度 3/5 1-2 天 新手友好度 68/100
rust-ndarray/ndarray#1615 · 1 則留言 ·
-
難度 4/5 3-5 天 新手友好度 48/100
rust-ndarray/ndarray#1610 ·
-
Empty Array lead to "The strides must not allow any element to be referenced by two different indices"可能已有人在做 關聯的 PR 仍在進行中或已合併。 未關閉
難度 3/5 1-2 天 新手友好度 72/100
rust-ndarray/ndarray#1609 ·
查看 rust-ndarray/ndarray 的全部 Issue
相似的 Issue
-
難度 2/5 1-3 小時 新手友好度 68/100
維護者通常 1 天內回覆
-
security-scan
難度 1/5 1 小時以內 新手友好度 85/100
維護者通常 1 天內回覆
-
content good first issue
難度 2/5 1-3 小時 新手友好度 75/100
StudentSuite/awesome-skills-plugins-for-students#293 ·
維護者通常 1 天內回覆
-
難度 1/5 1 小時以內 新手友好度 72/100
gfx-rs/wgpu-native#636 ·
-
難度 2/5 1-3 小時 新手友好度 72/100
維護者通常 1 天內回覆