Crates.io | image-texel |
lib.rs | image-texel |
version | 0.3.1 |
source | src |
created_at | 2022-05-17 20:52:01.669061 |
updated_at | 2022-11-02 18:00:40.600283 |
description | A texel type and allocated buffers suitable for image data. |
homepage | |
repository | https://github.com/image-rs/canvas |
max_upload_size | |
id | 588607 |
size | 177,048 |
An allocated buffer suitable for image data. Compared to a Vec
of all
samples, it restricts the possible data types but on the other hand focusses on
a more efficient interface to casting the data or rearranging the contents.
The buffer types provided provide several utility operations out of the box, or they can be used as internal structures on which a more concrete interface is built.
After some discussion it
was concluded that there is likely no safe way to reinterpret the allocation of
a Vec<T>
for a different Vec<U>
, except in a very restricted set of cases¹.
This includes grouping samples in logic pixel structs, even if those are
annotated #[repr(C)]
or alike. This is hardly the fault of Vec<_>
, as
images and the necessary transmutations of the representative binary data are
far from general but Vec
must be.
¹This was slightly relaxed in Rust 1.61 where few casts such as Vec<[u8; 3]>
to Vec<u8>
are now permitted, and the reverse under a restricted set of
circumstances. This may provide us a good 'out' to perform zero-allocation
conversion into and from a standard vector. It's still far from flexible enough
for most high-performance needs.
It is not optimized to provide nd-algebra or other matrix style operations. This firstly keeps the implementation complexity low, allows other implementation details, and separates concerns. If you nevertheless find it useful for such purposes, rest assured that we find this incredibly cool even though we may not accept PRs to introduce additional complexity motivated solely by such use. Rather, use the code and free license to shape it to those other needs.