`bytes`/`string`s can be passed more efficiently

[heinrich5991]

E.g. passing bytes from Python to Rust involves two copies, when it could be zero copies if the Rust code accepted &[u8]. The other way around currently involves two copies as well, which could be reduced to one (the creation of the Python bytes).

┆Issue is synchronized with this Jira Task ┆Issue Number: UNIFFI-281

[arg0d]

Its not just about copies. Currently Rust bytes implementation uses for loop to copy individual bytes from RustBuffer into Vec<u8>. memcpy would be far more efficient.

[arg0d]

Do you think using memcpy instead of for loop is a separate issue or not?

[mhammond]

This issue title is broad enough I think it's fine to cover both cases. If it turns out a patch is submitted which only solves some inefficiencies but leaves others on the table, then it makes sense for a new more-specific issue to exist at that point.

[heinrich5991]

Its not just about copies. Currently Rust bytes implementation uses for loop to copy individual bytes from RustBuffer into Vec<u8>. memcpy would be far more efficient.

Have you verified that it's not actually optimized into a memcpy already?

[arg0d]

Good point, I didn't check that.