XArray
is an abstract data type functioning like an expansive array of items where each item must be an 8-byte object, such as Arc<T>
or Box<T>
.
User-stored pointers must have a minimum alignment of 4 bytes. XArray
facilitates efficient sequential access to adjacent entries,
supporting multiple concurrent reads and exclusively allowing one write operation at a time.
- Cursors: Provide cursors for precise and efficient iteration over the array. Cursors have both immutable and mutable versions. One can hold multiple immutable cursors or hold a mutable cursor exclusively at a time.
- Marking: Provide ability to mark entries and the XArray itself for easy state tracking.
- Generics: Generic implementation that can work with any entry type that fits the use case.
- Copy-on-Write (COW): Efficient cloning of XArrays with shared structure until mutation.
Add this to your Cargo.toml
:
[dependencies]
xarray = "0.1.0"
This crate is developed in no_std
environment, but std users can still use this crate with --feature="std"
:
The following section covers how to interact with XArray
including creating an XArray
, using cursors, marking, cloning, and more.
// In std environment
extern crate alloc;
use alloc::sync::Arc;
use xarray::XArray;
// Create a new XArray instance
let mut xarray: XArray<Arc<i32>> = XArray::new();
- Users should declare the type of items (Arc) stored in the XArray, and the item type should implement
ItemEntry
trait. - We implement
ItemEntry
foralloc::sync::Arc
andalloc::sync::Box
by default, hence std users can use them directly.
extern crate alloc;
use alloc::sync::Arc;
use xarray::XArray;
let mut xarray_arc: XArray<Arc<i32>> = XArray::new();
let mut cursor = xarray_arc.cursor_mut(0);
// Store the Arc at the index range 0~10000.
for i in 0..10000 {
let value = Arc::new(i * 2);
cursor.store(value);
cursor.next();
}
cursor.reset_to(0);
for i in 0..10000 {
let value = cursor.load().unwrap();
assert!(*value.as_ref() == i * 2);
cursor.next();
}
Here is an example of using marks for the stored pages in the XArray, where PageMark represents the states of each individual Page:
extern crate alloc;
use alloc::sync::Arc;
use xarray::{XArray, XMark, StdMutex};
#[derive(Clone, Copy)]
enum PageMark {
DirtyPage
...
}
impl From<PageState> for XMark {
fn from(mark: PageState) -> Self {
match mark {
PageState::DirtyPage => Self::Mark0,
...
}
}
}
let mut pages: XArray<Page, StdMutex, PageState> = XArray::new();
let mut cursor = pages.cursor_mut(1000);
cursor.store(Page::alloc_zero());
// Mark the Page as DirtyPage.
cursor.set_mark(PageState::DirtyPage).unwrap();
assert!(cursor.is_marked(PageState::DirtyPage));
- Items and the
XArray
can have up to three distinct marks by default, with each mark independently maintained. - Users need to use a struct to represent the marks that need to be used. For the situation where multiple marks are required, these marks are typically encapsulated within an enumeration class.
- If users want to use a struct
M
for marks, they should implementFrom<M>
trait forXMark
and declareM
in the generics list of XArray.
use std::sync::Arc;
use xarray::{XArray};
let mut xarray: XArray<Arc<i32>> = XArray::new();
// Store values
let value = Arc::new(10);
xarray.store(1, value.clone());
assert_eq!(*xarray.load(1).unwrap().as_ref(), 10);
// Clone the XArray
let mut xarray_clone = xarray.clone();
assert_eq!(*xarray_clone.load(1).unwrap().as_ref(), 10);
// Store a new value in the clone
let new_value = Arc::new(100);
xarray_clone.store(1, new_value);
// The original XArray is unaffected by changes in the clone
assert_eq!(*xarray.load(1).unwrap().as_ref(), 10);
assert_eq!(*xarray_clone.load(1).unwrap().as_ref(), 100);
use std::sync::Arc;
use xarray::XArray;
let mut xarray: XArray<Arc<i32>> = XArray::new();
// Store item to even index in the range 100~200.
for i in 100..200 {
if i % 2 == 0 {
let value = Arc::new(i * 2);
cursor.store(value);
}
cursor.next();
}
// Iterate at the range 100~200.
let mut count = 0;
for item in xarray.range(100..200) {
count += 1;
}
assert_eq!(count == 50);