Crate slice_of_array
source ·Expand description
Extension traits for viewing a slice as a slice of arrays or vice versa.
Provides the following methods on [T]:
nest:&[T] -> &[[T; n]]flat:&[[T; n]] -> &[T]as_array:&[T] -> &[T; n](the reverse is already provided by a coercion)nest_mut,flat_mut,as_mut_arrayfor&mut [_].
Altogether, these let you swap between arbitrary representations
of contiguous, T-aligned streams of T data. For instance,
to view a [[i32; 6]; 5] as a &[[[i32; 3]; 2]; 5],
one could write
x.flat().nest().nest().as_array()Type inference generally works quite well, and as long as the final shape is unambiguous there is no need to annotate types in the middle of the method chain.
In cases where type inference is unable to determine the target
array size, one can use a turbofish: e.g .x.nest::<[_; 3]>().
use ::slice_of_array::prelude::*;
let vec = vec![[2i32, 2, 2], [7, 7, 7], [4, 4, 4], [1, 1, 1]];
assert_eq!(vec.flat(), &[2, 2, 2, 7, 7, 7, 4, 4, 4, 1, 1, 1]);
// note: this requires an annotation only due to polymorphism in PartialEq
let slc = vec.nest::<[_; 2]>();
assert_eq!(slc, &[[[2i32, 2, 2], [7, 7, 7]], [[ 4, 4, 4], [1, 1, 1]]]);nest and as_array panic on failure rather than returning options.
The rationale is that it is believed that these these conversions are
seldom needed on arbitrary user data which may be the wrong size; rather,
they are most likely used when bridging the gap between APIs that work
with flattened slices and APIs that work with slices of arrays.
Zero-cost conversions in owned data (e.g. between Vec<T>
and Vec<[T; n]>) are not provided, and are probably impossible
in consideration of e.g. custom allocators. If you need to
convert between such types, you can use these traits in tandem
with <[T]>::to_vec to perform a copy:
let vec = vec![[2i32, 2, 2], [7, 7, 7]];
// copying into a Vec<i32>
let flattened = vec.flat().to_vec();
assert_eq!(flattened, vec![2i32, 2, 2, 7, 7, 7]);Modules
Traits
Slice{Flat,Nest,Array}Ext.