On Fri, Aug 2, 2024 at 2:02 PM Danilo Krummrich <dakr@xxxxxxxxxx> wrote:
>
> On Fri, Aug 02, 2024 at 09:08:48AM +0200, Alice Ryhl wrote:
> > On Thu, Aug 1, 2024 at 5:37 PM Danilo Krummrich <dakr@xxxxxxxxxx> wrote:
> > >
> > > On Thu, Aug 01, 2024 at 05:10:22PM +0200, Alice Ryhl wrote:
> > > > On Thu, Aug 1, 2024 at 2:08 AM Danilo Krummrich <dakr@xxxxxxxxxx> wrote:
> > > > >
> > > > > Currently, we can't implement `FromIterator`. There are a couple of
> > > > > issues with this trait in the kernel, namely:
> > > > >
> > > > > - Rust's specialization feature is unstable. This prevents us to
> > > > > optimze for the special case where `I::IntoIter` equals `Vec`'s
> > > > > `IntoIter` type.
> > > > > - We also can't use `I::IntoIter`'s type ID either to work around this,
> > > > > since `FromIterator` doesn't require this type to be `'static`.
> > > > > - `FromIterator::from_iter` does return `Self` instead of
> > > > > `Result<Self, AllocError>`, hence we can't properly handle allocation
> > > > > failures.
> > > > > - Neither `Iterator::collect` nor `FromIterator::from_iter` can handle
> > > > > additional allocation flags.
> > > > >
> > > > > Instead, provide `IntoIter::collect`, such that we can at least convert
> > > > > `IntoIter` into a `Vec` again.
> > > > >
> > > > > Signed-off-by: Danilo Krummrich <dakr@xxxxxxxxxx>
> > > >
> > > > I'm not convinced a collect implementation specific to IntoIter is necessary?
> > >
> > > For the reasons above, we can't implement `FromIterator`. At some point we may
> > > want to implement our own kernel `FromIterator` trait, but that's out of scope
> > > for this series.
> > >
> > > For now, I just want to provide a way to get a `Vec` from `IntoIter` again,
> > > which without `Vec::collect` would be impossible.
> >
> > If you have a need for a collect on this particular kind of iterator, then okay.
>
> Even once we have our own `FromIterator` trait, we probably want to keep this
> specific one besides the generic `collect` for optimization. With the generic
> one we'd otherwise copy into a new `Vec`.
>
> >
> > > > > +
> > > > > + // SAFETY: `buf` points to the start of the backing buffer and `len` is guaranteed to be
> > > > > + // smaller than `cap`. Depending on `alloc` this operation may shrink the buffer or leaves
> > > > > + // it as it is.
> > > > > + ptr = match unsafe { A::realloc(Some(buf.cast()), layout, flags) } {
> > > >
> > > > Why would you shrink it? You can just keep the capacity.
> > >
> > > What if the vector was pretty huge and meanwhile as advanced by a lot? I think
> > > we don't want to waste those resources.
> > >
> > > Ideally the corresponding `Allocator` implements a proper heuristic for when to
> > > actually shrink. For instance, krealloc() never shrinks, and it's probably not
> > > worth it. For vrealloc() though we clearly want to shrink properly (i.e. unmap
> > > and free spare pages) at some point.
> >
> > The Rust Vec never shrinks unless explicitly requested. But I guess
> > it's okay either way.
>
> It actually does, see [1]. But Rust's `Vec` does it less efficiently. It either
> keeps the `Vec` as it is, or creates a whole new one.
>
> [1] https://github.com/rust-lang/rust/blob/master/library/alloc/src/vec/spec_from_iter.rs#L39
Huh, surprising.
Reviewed-by: Alice Ryhl <aliceryhl@xxxxxxxxxx>
Alice
