I'm pretty sure changing the associated types here is a semver-breaking change and so shouldn't be done in a patch release. Personally I'm against that if possible as it would mean leptos would also need to release a new minor version, which breaks the whole ecosystem again etc.

I think that, at least at present, using AlignedVec::into_vec() to keep the associated types the same would be preferable. At least in the leptos context we only use this encoding and decoding process once (going from server to client) and it is part of a larger encoding/decoding process which includes encoding binary data to base64 to be included as part of the HTML payload anyway — given the entire context (encoding to rkyv, encoding the rkyv binary to base64 to include it in the HTML file, sending it over the network, running it through the JavaScript interpreter as a string, copying it over the JS/WASM barrier, decoding the base64 to binary, decoding the rkyv binary format) eating an additional bytewise copy is worth it to fix a bug without semver breakage.

@thestarmaker Thank you very much for investigating and this fix!

I agree with @gbj
The design target of this crate is not to get the highest performance in any situation with any underlying tech but to make it easy to switch while giving good enough performance for many use-cases.

@thestarmaker could I bother you to implement this and also fix the rustfmt issue?

PR comments have been addressed; please merge and update Leptos when ready

Awesome, thank you!

@gbj It's released

Thanks! Can confirm that a simple cargo update to codee 0.3.2 now fixes the issue.